7th Proceedings Indiana Conference

George K.
Stookey Editor
Proceedings of the 7th Indiana Conference Indianapolis, Indiana
Clinical Models Workshop: Remin-Demin, Precavitation, Caries
George K. Stookey Indiana University School of Dentistry Therametric Technologies, Inc. 351 W. 10th Street, Suite 222 Indianapolis, IN 46202-4119 U.S.A.
Library of Congress Cataloging-in-Publication Data Library of Congress Control Number: 2006908914 Clinical models workshop: Remin-demin, precavitation, caries: proceedings of the 7th Indiana Conference initiated by Indiana University School of Dentistry, Indianapolis, Indiana; supported by the National Institute of Dental and Craniofacial Research (Grant Number 1 R43 DE016767 and the Task Force for Design and Analysis in Dental and Oral Research; and held at the Indianapolis Marriott Downtown, Indianapolis, July 3-5, 2005. George K. Stookey, editor. Includes bibliographic references. ISBN 0-9655149-5-1 2005 Indiana University School of Dentistry All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Therametric Technologies, Indiana University Emerging Technologies Center, 351 West Tenth Street, Indianapolis, IN 46202-4119, USA), except for brief excerpts in connection with reviews of scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The publisher, editors, and sponsor cannot assume any legal responsibility for information on instrumentation or on drug dosage and administration contained in this publication. The respective user must check its accuracy by consulting other sources of reference in each individual case. The use of general descriptive names, registered names, trademarks, etc., in this publication, even if not identified as such, does not imply that they are exempt from the relevant protective laws and regulations or free for general use. Cover design by Mark. A. Dirlam. Interior design by Linda Levandoski, Shawna Ros, Dorothy Lyon, Scott Bennett, Robin Holtel and Evie Groll. Camera-ready copy prepared by Karen E. Wilczewski, Marketing Communications Consultants. Printed and bound by Printing Partners. Printed in the United States of America. ISBN 0-9655149-5-1
Acknowledgment
The Editor gratefully acknowledges the National Institute of Dental and Craniofacial Research (Grant Number 1 R43 DE016767 and the Task Force for Design and Analysis in Dental and Oral Research for sponsoring the 7th Indiana Conference, and for sharing our commitment to uphold the highest standards and ideals in the pursuit of knowledge. A special word of thanks go to Theresa J. Norris, administrative assistant, and Kristi A. Dobson, consultant, for their invaluable assistance in the planning and arrangements for this Indiana Conference, and Karen E. Wilczewski, Marketing Communications Consultants, for developing the camera-ready copy of this material.
Preface
This monograph contains the proceedings of the 7th Indiana Conference held July 3-5, 2005 at the Indianapolis Marriott Downtown Hotel, Indianapolis, Indiana, U.S.A. The program was designed as a workshop, with a series of presentations by recognized experts on assigned topics, followed by lengthy discussions by assigned discussion groups of 18 to 20 invited experts and conference attendees; final reports were then presented from each of the discussion groups. Six of the 12 primary oral presenters were from outside the United States, and the total of 150 invited scientists and conference attendees were from 18 different countries in North America, South America, Europe, and Asia. Copies of these oral presentations and the reports from the discussion groups are included in this monograph. This conference was made possible through grants from the National Institute of Dental and Craniofacial Research (Grant Number 1 R43 DE016767) and the Task Force for Design and Analysis in Dental and Oral Research. We are most appreciative for the support provided by these organizations. George K. Stookey, Ph.D. Distinguished Professor Emeritus Indiana University School of Dentistry
CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES
II
INDIANA CONFERENCE 2005
Table of Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .i Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .iii List of Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .vi Introductory Remarks & Workshop Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xi Paper Presentations Traditional Models for Clinical Caries Trials David Banting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Conclusions from the 2002 Scotland Concensus Conference (ICW-CCT) Nigel B. Pitts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17 ROC Analysis with Applications to Dental Research Nancy A. Obuchowski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35 Including Non-cavitated Lesions in Pivotal Clinical Trials: Issues and Concerns Phillippe P. Hujoel, Walter Bretz, and James Bader . . . . . . . . . . . . . . . . . . . . . .49 Distinguishing Between Individuals with High or Low Risk of Developing Dental Caries Hannu Hausen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65 Acceptance Criteria for Clinical Caries Models Albert Kingman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79 The Nyvad Criteria for Assessment of Caries Lesion Activity Bente Nyvad, V. Machiulskiene, and V. Baelum . . . . . . . . . . . . . . . . . . . . . . . . . . .99 The Chesters Model Using the Simplified DTSM Richard K. Chesters, I. Balciuniene, G. Matuliene, N.B. Pitts, and J.R. Matheson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .117
III
The Biesbrock-Bartizek Caries Clinical Model Aaron R. Biesbrock, and R.D. Bartizek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .143 Rationale and Evidence for the International Caries Detection and Assessment System (ICDAS II) Amid I. Ismail, and the International Caries Detection and Assessment System (ICDAS) Coordinating Committee . . . . . . . . . . . . . . . . . . . . . . . . . . . . .161 Appendix I. Criteria Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .190 Appendix II. Participants in the ICDAS Baltimore Workshop . . . . . . . . . . . . . . . . .209 Quantitative Light Induced Fluorescence in Caries Clinical Trials Susan M. Higham, I.A. Pretty, and P.W. Smith . . . . . . . . . . . . . . . . . . . . . . . . .223 Other Instrumental Detection Methods for Caries Clinical Trials James S. Wefel, Masatoshi Ando, and Jeffrey D. Harless . . . . . . . . . . . . . . . . .237 Workshop Procedures & Breakout to Workshop Groups Workshop Group on Statistical Modeling Strategies Albert Kingman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .251 The Nyvad Criteria Breakout Group Report Bente Nyvad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .257 2005 Indiana Conference Workshop Session: Chesters Model Richard Chesters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .259 2005 Indiana Conference Workshop Session: The Biesbrock-Bartizek Model Aaron R. Biesbrock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .263 Indiana Conference ICDAS Workshop Group Report Nigel B. Pitts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .265
IV
The Quantitative Light Fluorescence (QLF) Model Susan M. Higham . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .273 Other Instrumental Detection Methods for Clinical Caries Trials James S. Wefel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .277 Concluding Remarks George K. Stookey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .283
List of Participants
CONFERENCE HOST
Dr. George K. Stookey Distinguished Professor Emeritus Indiana University School of Dentistry 1121 W. Michigan Street Indianapolis, IN 46202-5186 USA telephone: 001 317 278 7876 facsimile: 001 317 278 7880 e-mail: gstookey@iupui.edu
PRIMARY AUTHORS
Dr. David Banting School of Dentistry Faculty of Medicine & Dentistry University of Western Ontario London, ON N6A 5Cl Canada telephone: 001 519 661 2111, ext. 86130 facsimile: 001 519 661 3885 e-mail: dbanting@uwo.ca Dr. Aaron Biesbrock Procter & Gamble Company Health Care Research Center Mason, OH 45040-9462 USA telephone: 001 513 622 0316 facsimile: 001 513 622 0550 e-mail: biesbrock.ar@pg.com
VI
Richard K. Chesters Colgate-Palmolive Europe 13-15, Cours de Rive 1204 Geneva Switzerland telephone: 41 22 722 0784 facsimile: 41 22 722 0703 e-mail: richard_chesters@colpal.com Dr. Hannu Hausen Institute of Dentistry P.O. Box 5281 FIN-90014 University of Lulu Finland telephone: 358 8 537 5582 facsimile: 358 8 537 5560 e-mail: Hannu.Hausen@oulu.fi Dr. Susan M. Higham Cariology and Toothwear Research Group School of Dental Studies The University of Liverpool UK telephone: 44 0151 706 5251/5288 facsimile: 44 0151 706 5937 e-mail: S.M.Higham@liv.ac.uk
Dr. Philippe Hujoel Department of Dental Public Health Sciences School of Dentistry University of Washington Seattle, WA 98195 USA telephone: 001 206 543 2034 facsimile: 001 206 685 4258 e-mail: hujoel@u.washington.edu
VII
Dr. Amid I. Ismail Department of Cariology Restorative Sciences and Endodontics School of Dentistry, D2361 1011 N. University University of Michigan Ann Arbor, MI 48109-1078 USA telephone: 001 734 647 9190 facsimile: 001 734 936 1597 e-mail: ismailai@umich.edu Dr. Albert Kingman Biostatistics Core Division of Clinical Research and Health Promotion National Institute of Dental and Craniofacial Research Bethesda, MD 20892-2190 USA telephone: 001 301 594 4820 facsimile: 001 301 480 8322 e-mail: kingmana@de45.nidr.nih.gov Dr. Bente Nyvad Faculty of Health Sciences University of Aarhus Vennelyst Boulevard 9 DK 8000 Aarhus C Denmark telephone: 45 8942 4074 e-mail: Nyvad@odont.au.dk Dr. Nancy A. Obuchowski Department of Quantitative Health Sciences.Wb4 The Cleveland Clinic Foundation 9500 Euclid Avenue Cleveland, OH 44195 USA telephone: 001 216 445 9549 facsimile: 001 216 444 3466 e-mail: nobuchow@bio.ri.ccf.org
VIII
Dr. Nigel B. Pitts Dental Health Services Research Unit & Centre for Clinical Innovations University of Dundee, Health Infomatics Centre Kirsty Semple Way Dundee DD2 4BF Scotland, UK telephone: 44 1382 635959 facsimile: 44 1382 226550 e-mail: n.b.pitts@dundee.ac.uk Dr. James S. Wefel Dows Institute for Dental Research College of Dentistry N413 DSB University of Iowa Iowa City, IA 52242 USA telephone: 001 319 335 7895 facsimile: 001 319 335 7376 e-mail: james-wefel@uiowa.edu
IX
Introductory Remarks & Workshop Goals

Welcome It is indeed my pleasure to welcome you to this Indiana Conference. This is the 7th Indiana Conference and has a somewhat different mission, format, and source of funding as compared to our previous conferences. The overall mission of the workshop is to review the status of various models being used for clinical caries trials. Funding for the entire workshop is being provided through grants from the National Institute for Dental and Craniofacial Research (Grant No. 1 R43 DE016767) and the Task Force for Design and Analysis in Dental and Oral Research. These grants allowed us to cover the costs for the primary speakers from academia and nearly all of the 58 invited expert discussants. We are most appreciative of these organizations for providing this support. We have limited attendance to 150, and we are pleased that we have a full house to address this important topic. Rather than a series of presentations on specific topics, this conference will follow a workshop format, with ample time for assigned group discussions of each topic. History of Indiana Conferences Indiana Conferences were initiated in 1996 as a means of having international scientists offer state-of-the-art presentations on selected topics and to provide a forum for scientists to discuss concerns and establish collaborations. The themes for the conferences have included early caries detection (1996, 1999, 2003), microbial pathogenesis (1997), dental and orthopedic implants (1998), and changing practices in restorative dentistry (2000). The proceedings of each conference has been published as a hard cover monograph and is provided to all attendees. Unrestricted funding to help cover travel costs for speakers and publication costs for these prior conferences were provided by the Procter & Gamble Company. The 2005 Indiana Conference is the fourth conference that has focused on the early detection of dental caries. Each of the previous conferences has involved a series of presentations by investigators of new and developing caries detection methods and technologies. These presentations have included virtually all of the technologies that have been, or are being George K. Stookey investigated for the detection of dental caries. CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES XI
Technologies that have been presented and discussed have included fiber optic transillumination (FOTI), digital imaging fiber optic trans-illumination (DIFOTI), quantitative laser/light fluorescence (QLF), infra-red laser fluorescence (DIAGNOdent), dyeenhanced laser fluorescence (DELF), electrical conductance method (ECM), direct digital radiography (DDR), ultrasonic methods, polarization sensitive optical coherence tomography (OCT), multi-photon imaging systems, electro-mechanical approaches, infra-red photothermal radiometry, and infra-red thermography. Conference Goals It is well known that relatively few, if any, significant advances in the development of more effective measures for the prevention of dental caries have been identified during the past 25 years. As noted during the recent Scotland workshop, the primary reasons for this have been the relative inefficiency of accepted clinical models available for assessing the merits of improved caries-preventive agents or treatment strategies. This inefficiency has resulted in major costs, totaling several millions of dollars for a single controlled clinical trial, and a common time period of at least four years from initiation to final report. Needless to say, such a commitment of funding and time is extremely difficult to justify by either NIDCR or dental products manufacturers. Especially during the past three-to-four years, considerable attention has been given to the need to identify more efficient models for controlled clinical trials. In January 2002, Drs. Nigel Pitts and John Stamm organized an International Consensus Workshop on Caries Clinical Trials in Glasgow, Scotland, and the proceedings, which were recently published in a special issue of the Journal of Dental Research, emphasized the need for more efficient models for clinical caries trials. In May 2003, our most recent Indiana Conference involved a series of state-of-the-art presentations on current research involving early caries detection methods. In May 2004, NIH-NIDCR convened a major conference entitled Methods for Enhancing the Efficiency of Dental/Oral Health Clinical TrialsCurrent Status, Future Possibilities, which focused on efforts to develop surrogate markers and other strategies to improve the efficiency of clinical trials. These prior conferences, as well as numerous literature reports, have recognized that the caries process is a continuum, beginning with enamel demineralization and progressing over a period of time to cavitation. While actual cavitation has been the measurable end-point in traditional clinical caries trials, it is recognized that this late stage of the caries process is no longer acceptable as the primary assessment measure, and there should be earlier stages in the process that may be used for this purpose. Dental scientists have proposed several approaches to improving the efficiency of clinical caries trials. These approaches have included:
XII
increased supervision and control of panelist compliance with the treatment regimens required of the investigation; clinical assessments of the caries process at the pre-cavitation stage of development; and, the use of instrumental methods capable of detecting and monitoring the caries process at the early enamel demineralization stage, well in advance of cavitation. During the past few years, several controlled clinical caries trials have included these methodologies, and the results of several of these investigations have been published or have become available at different conferences. The 2005 Indiana Conference has been designed as a workshop for established cariologists to review the available supporting data and determine whether there are now more efficient clinical models available for future assessments of caries-preventive measures. Thus, the goals of this workshop are to: review the scientific information available for the different strategies that have been explored to improve the efficiency of controlled clinical caries trials; and, determine whether there are adequate data to support the use of these methodologies in future clinical trials of caries-preventive measures and, if not, the type of additional information that is needed for these approaches. Workshop Format The format for this workshop will involve a series of presentations by persons considered to be among the leading experts on the designated topic, followed by workshop discussions of each topic by 18 to 20 invited scientists and general workshop participants. The designated topics include several presentations considered to be important background reviews for the discussions and presentations of the following clinical models: Statistical Modeling Strategies The Nyvad Model The Chesters Model The Biesbrock-Bartizek Model The ICDAS Model QLF: Quantitative Light Fluorescence Other Models for Clinical Trials The workshop will conclude with reports and recommendations from each of the discussion groups. We expect that these discussions and recommendations will be very important for identifying scientifically acceptable clinical models for future clinical caries trials.
XIII
XIV
Traditional Models for Clinical Caries Trials

David Banting School of Dentistry, Faculty of Medicine & Dentistry, University of Western Ontario, london, Ontario, Canada
Introduction One good thing about being around for a long time is that it affords one a unique perspective, an opportunity to judge events over time, which is not otherwise possible. I had the privilege a few years ago to speak at Don Lewis retirement and remarked then that, in the field of health services research, concepts that Don Lewis and other visionaries were researching, presenting, and promoting 30 years ago are now being widely accepted and implemented; concepts like critical appraisal of the literature, quality assessment, and the evidence-based approach.
The Traditional Models for Clinical Caries Trials Just last year, Ismail [2004] conducted a systematic review of criteria systems used for the detection and classification of dental caries from 1950 to 2000. This was the first time that the content validity of these existing systems had been looked at. What I found fascinating about this review was that it revealed 29 unique criteria systems and then went on to discuss their origin, date of introduction, and the criteria used. Ismail also made some interesting observations concerning whether or not these systems measured both the active and inactive stages of the caries process, whether an explorer was used for probing, and whether the teeth were cleaned and dried prior to applying the criteria. There are 3 messages resulting from this review that I feel are pertinent to this presentation: 1. There was considerable variability in the disease processes measured by the different systems.
BANTING
2. Generally, European systems include both the non-cavitated (early) and advanced (cavitated) stages of the caries process; whereas, North American systems focus on measuring cavitated lesions. 3. The sensibility of the European criteria is towards characterizing the disease process; whereas, the sensibility of the North American system is towards achieving reliability of caries diagnosis. Rightly or wrongly, I have selected 3 of these 29 systems as the traditional models. These models were developed in the 1960s, 1970s, and 1980s and have, as far as I can determine, been widely used throughout the world in clinical trials and surveys relating to dental caries. The 3 models that I have designated as traditional models are Radikes criteria [1968], the WHO Oral Health Surveys Basic Methods criteria [1971], and the NIDR criteria [1987].
Radike Criteria for Diagnosis of Dental Caries The Conference on Clinical Testing of Cariostatic Agents was convened at the American Dental Association in October 1968. Some in the audience were fortunate to be a part of that conference. When you look down the list of names of conference attendees, you cannot help but be impressed with who was there. I will not name them all, but names such as Finn Brudevold, Neal Chilton, Bob Grainger, Hershel Horowitz, Hans Muhlemann, Charles Gish, Stan Heifetz, Walter Zackeral, James Carlos, Thomas Mathaler, Harold Englander, Robert Glass, Geoffery Slack, Paul DePaola, Louis Ripa, George Stookey, and Tony Volpe are names we associate, not only with dental caries, but with clinical studies of dental caries. Arthur Radike worked for the Procter & Gamble, and he presented 2 papers at the 1968 ADA conference. The first paper was entitled, Criteria for the Diagnosis of Dental Caries and the second paper was entitled, Examiner Error and Reversals in Diagnosis. Although the first paper is the major focus of this presentation, I will refer to both papers. Radikes paper on Criteria for Diagnosis of Dental Caries was actually a report of a Task Group on Caries Management that had gathered together for several work sessions in the year preceding the conference; The Task Groups report was reviewed during the conference and ultimately appeared in a well known ADA document entitled, Proceedings of the Conference on Clinical Testing of Cariostatic Agents that was published 4 years later in 1972.
2 INDIANA CONFERENCE 2005
TRADITIONAL MODELS
FOR
CLINICAL CARIES TRIALS
The criteria, which became known as Radikes criteria, gave us such familiar descriptors for dental caries as a frank lesion, gross cavitation, the catch of an explorer, and a discontinuity or break in the continuity of the enamel surface. All of these, you will notice, deal with a hole in the tooth.
WHO Criteria In 1961, the World Health Organization (WHO) convened a WHO Expert Committee on Dental Health and formulated standard methods for the reporting of dental diseases (WHO, Technical Report Series, 1962, No. 242). This was followed a decade later by a manual on Basic Methods for Oral Health Surveys. As part of that manual, criteria for diagnosing dental caries were presented. These criteria are, no doubt, equally as familiar as that of Radike to most people in this audience. With the WHO criteria, in order for caries to be detected, a lesion must have a detectably softened floor, undermined enamel or softened wall. On an inter-proximal surface, the probe point must enter a lesion with certainty. The WHO criteria specifically stated: where any doubt exists, caries should not be diagnosed as present.
NIDR Criteria The 1987 version of the NIDR criteria were based on criteria developed in 1938 by H.T. Klein, C.E. Palmer, and J.W. Knutson for the DMFT index. However, the NIDR criteria also provided for the detection of what was termed incipient lesions and subdivided these into 3 categories according to location: pits and fissures, smooth areas on buccal and lingual surfaces, and proximal surfaces. The NIDR criteria described carious lesions on the root surfaces of teeth for the first time. Carious lesions that occur on the roots of teeth were described a being discolored, either with or without cavitation present. An explorer was used to detect a softness to the lesion. Rules were provided to determine whether adjacent coronal surfaces were to be deemed carious and to determine whether adjacent root surfaces were also carious.
BANTING
Characteristics of the Traditional Models for Caries Detection The details of the criteria for each of these traditional models can be found in the appendices and will not be reviewed in detail. However, it is informative to contrast some of the characteristics of these traditional models in order to illustrate not only similarities, but some evolution in the detection procedure. The Radike and WHO criteria relate only to coronal surfaces. The NIDR criteria included criteria for both coronal and root surfaces. All 3 of these models considered only cavitated lesions as caries. This is very much a North American sensibility, as Ismail has aptly pointed out. The NIDR criteria, however, include both gross and incipient lesions as caries, but the explorer must be used, and there must be softnessdetected by tactile sensation. Unfortunately, none of these models evaluate carious lesion severity or depth. No distinction is made between carious lesions that involve only the enamel or those that extend into the dentine and, except for white spots lesions that are soft to the explorer touch in the NIDR criteria, enamel lesions are precluded from all of these models. The Radike and NIDR procedures recommend that the teeth be dry when viewed, and the WHO procedure does not mention drying the teeth, probably because it is designed to be used under field conditions. All 3 detection models advocate that a (sharp) explorer be used with moderate-to-firm pressure for tactile purposes; all 3 ignore any judgment regarding lesion activity, and all 3 specify conditions, primarily stain, and white spots that are not defined as dental caries (Table 1).
Model Surfaces Evaluated Stages Measured Severity cavitated Lesion Depth or Wet or Dry no Explorer Used/Viewed Activity Other Conditions Excluded stain, pigmentation, fractures, erosion, abrasion, hypoplasia, mottled enamel, certain enamel opacities white or chalky spots, discolored or rough spots, hard pits or fissures that catch on the explorer point stain, hypoplasia, sealants
Radike
coronal
yes/dry
no
WHO
coronal
cavitated
no
yes/wet
no
NIDR
coronal and root
cavitated
no
yes/dry
no
TABLE 1. 4
Characteristic of the traditional caries detection models. INDIANA CONFERENCE 2005
TRADITIONAL MODELS
FOR
The Reality of Dental Caries Detection and Classification Today Since these traditional models were developed, our thinking about dental caries has evolved such that there are 10 principles that I feel reflect todays reality about dental caries 1. dental cariesa continuum: a process that begins with the loss of ions and ultimately ends with destruction and loss of 1 or more of the 3 calcified tissues of the tooth. 2. intermediatein this continuum there is a non-cavitated stage of dental caries (with substages) that can be identified and described using clinical signs [Ekstrand, 2004]. 3. superimposed on all of this is the notion that the dental caries process is reversible (demineralization/remineralization), at least in its early stages and that the surface contour, once lost, cannot be naturally re-established. 4. the time it takes for dental caries to progress/regress from one stage to another and through any stage is widely variable among individuals and groups. 5. surgical intervention (restoration) is generally considered to be a last resort in the management of dental caries, and the mere presence of cavitation does not necessarily invoke this type of management. 6. a carious lesion can either be in an active state or an inactive state at any one point in time. 7. there is a cut point, a point before which dental caries cannot be detected clinically using visual criteria. 8. the use of tactile sensation using a sharp instrument with moderate pressure is only marginally useful for detection and may adversely affect the carious lesion. 9. measurement of dental caries prevalence should take into account both non-cavitated and cavitated lesions. 10. similarly, measurement of dental caries incidence or increment should take into account transitions between all the clinical stages in the dental caries continuum, including biologically plausible reversals (remineralization).
BANTING
The Traditional Caries Detection Models and Todays Reality So, how well do the traditional models for diagnosing dental caries fulfill our present day needs for caries detection? Clearly, the Radike and WHO models do not provide any information on non-cavitated carious lesions or carious lesions on the roots of teeth. Investigators today want to know not only whether or not caries is present and on what tooth surface(s), but they need to know how severe the lesion is, whether it is active or arrested, and whether it is progressing or improving with treatment (remineralizing) in order to provide an accurate and valid picture of what is happening and initiate and evaluate different management strategies. It is readily apparent that all 3 of the traditional dental caries detection models fall short of what is needed to meet current needs regarding the detection of dental caries (Table 2). Therefore, we need to develop new models of visual caries detection. This is perhaps more easily said than done, because these new models would need to possess, at minimum, the following characteristics: 1. be backwardly compatible with traditional caries detection models for comparative purposes. 2. encompass the full range of conditions or stages deemed to be clinical dental caries. 3. be shown to be reproducible (reliable) at a specified level. 4. be valid.
Requirements Non-cavitated lesion coronal root Cavitated lesion coronal root
RADIKE
WHO
NIDR
no no yes no
no no yes no
yes yes yes yes
Severity (depth) Activity (progresses/arrests) Treatment options
RADIKE no no
WHO no no
NIDCR no no
surgical only
surgical only
medical and surgical
TABLE 2.
Traditional caries detection models and present-day measurement and reporting requirements. INDIANA CONFERENCE 2005
TRADITIONAL MODELS
FOR
differentiate active from arrested or remineralized lesions. incorporate severity levels that can be directly translatable into treatment management protocols. 7. be versatile in order to meet the specific needs of epidemiology/public health, research, and clinical practice. Even if new models were developed that met these guidelines, there are several issues related to dental caries detection that remain unresolved, and the resolution of these issues should precede the development and/or acceptance of any new diagnostic criteria for dental caries. These issues are: 1. a clear distinction between the use of the terms detection and diagnosis, guidance as to when each should be used, and what type of personnel can credibly perform these tasks. 2. clearly defined cut points in the dental caries continuum that can be directly translated into clinical decisions regarding observation, medical management, and surgical management of dental caries. 3. updating of undergraduate dental programs regarding current evidence relating to the detection and diagnosis of dental caries. 4. guidance and education of the dental profession as to when and how adjunctive clinical tests (laser fluorescence, FOTI, ultrasound) should be used to supplement the clinical detection/diagnosis of dental caries. 5. guidelines regarding the measurement and reporting of dental caries increment in clinical studies of anti-caries agents that are capable of arresting and/or reversing (remineralizing) a caries lesion.
5. 6.
A Final Word About the Measurement of Caries Increment My final comments relate to the last point. Radikes second paper at the Conference on the Clinical Testing of Cariostatic Agents dealt with examiner agreement or reproducibility. For anyone involved in clinical caries studies, this is a critical issue. In his paper, Radike made the point that, with the criteria that the Task Group proposed for the diagnosis of dental caries, errors would occur in an
BANTING
examiners calls, particularly reversals in diagnosis. And, Radike made the further point that these reversals will be more prevalent in the transition zone between sound and carious enamel. In fact, Radike estimated that 25% of the conditions in this transition zone will result in a reversal in diagnosis. Measurement of caries increment must take into consideration: 1. examiner inconsistency related particularly to non-cavitated lesions, 2. remineralization as a plausible treatment outcome, and 3. length of the observation period. One of the justifications for detecting earlier (non-cavitated) stages of dental caries is to shorten the observation period. This has the obvious advantage of clinical trial efficiency, with the accompanying funding implications. However, in order to determine what is actually happening in a clinical trial, a longer observation time may actually be required. For instance, consider the possible examination-to-examination transition strings for a given tooth surface (table at the top of page 9). A single (usually 6-month) examination can only provide information on a single transition, and the validity of that transition cannot be determined. Longer observation periods allow for more transitions, which, in turn, provide more information. However, at the same time, other measurement dilemmas are introduced such as: 1. When does a transition string become an increment? 2. What constitutes a plausible treatment reversal? 3. What is an illogical reversal? 4. How do you differentiate a treatment reversal from an examine error? 5. How many repeat calls are needed to ensure the validity of a call? 6. How is lesion activity determined? These are the fascinating and, in my mind, vital issues to be addressed in clinical visual detection of dental caries today. And, just to confuse things a bit further, caries increment can be calculated in several different ways, depending on whether reversals are included and whether or not more than 1 event can take place on a given surface over time. Some possible caries increment measures include:
TRADITIONAL MODELS
Transition String S, D1w S, D1w, D1d Observation Time 6 months 12 months A positive increment
FOR
What is Going On?
A positive increment followed by a plausible reversal (negative increment) OR A positive increment followed by an examiner error Should 2 positive and 1 negative increment be counted here? OR Is there just 1 positive increment with an examiner error that should be ignored?
S, D1w, D1d, D1w
18 months
S, D1w, D1d, D1w, D2
24 months
There is obvious progression of the lesion but: Which end point(s) do you count? To what stage do you allow the lesion to progress without treatment?
S, D1w, D1d, D1w, D2, D1w
30 months
Was the 24-month call an examiner error? OR Is this a plausible reversal (negative increment)?
D1d, S D1d, S, D1d
6 months 12 months
Starting with an early lesion, is this a negative increment (reversal)? Has this lesion become active again or was the 6-month call an examiner error? How do you score this? Is this a single, positive increment or a series of positive and negative increments? Does consistency of the call at 24- and 30 month examinations confirm that there was a positive caries increment at 24-months?
D1d, S, D1d, S D1d, S, D1d, S, D2 D1d, S, D1d, S, D2,D2
18 months 24 months
30 months
F, S
6 months
An obvious illogical transition; however, which call is correct if the trial does not continue?
Crude Incrementfirst event/surface, reversals not included Crude Cumulative Incrementmultiple events possible/surface, reversals not included Net Incrementfirst event/surface, reversals included Net Cumulative Incrementmultiple events possible/surface, reversals included I do not believe that any consensus exists as to which caries increment measure is considered to be the most appropriate. Actually, the most appropri-
BANTING
ate measure may depend on several factors that may vary among clinical trials. The traditional models of caries detection, as faithfully as they have served us for the past 30 years, can no longer help us with the measurement issues we face today. It is time to move on.
References
Ekstrand KR: Improving clinical visual detection-potential for caries clinical trials. J Dent Res 2004;83(Spec. Issue C): C67-C71. Ismail, A: Visual and visual-tactile detection of dental caries. J Dent Res 2004;83(Spec. Issue C): C56-C66. Proceedings of a Conference on the Clinical Testing of Cariostatic Agents 1968. World Health Organization, Oral Health Surveys, Basic Methods, Geneva: World Health Organization 1971. U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, Oral Health of United States Adults, NIH Publication No. 87-2868, August 1987.
10
TRADITIONAL MODELS
FOR
Radikes Criteria (1968)

I. Frank lesions gross cavitation (defined as a discontinuity of the enamel surface caused by loss of tooth substance) is present. cavitation must be distinguished from fractures and smooth lesions of erosion and abrasion. Lesions not showing frank cavitation Pits and fissure lesions of the occlusal, facial, and lingual surfaces. 1. area is carious when the explorer catches or resists removal after insertion into a pit or fissure using moderate to firm pressure and when accompanied by 1 or more of the following signs of caries: a softness at the base of the area opacity adjacent to the pit or fissure as evidence of undermining or demineralization softened enamel adjacent to the pit or fissure, which may be scraped away with an explorer 2. the area is carious if there is loss of the normal translucency of enamel adjacent to the pit, which is in contrast to the surrounding tooth structure Lesions on smooth areas of facial and lingual surfaces 1. area is carious if the surface is etched or if there is a white spot as evidence of subsurface demineralization and if the area is found to be soft by: penetration with an explorer scraping away enamel with an explorer 2. area is sound when there is apparent evidence of demineralization (etching or white spot) but no evidence of softness Lesions on proximal surfaces 1. for areas exposed to direct visual and tactile examination, these lesions are diagnosed with the same criteria as smooth lesions on facial and lingual surfaces 2. for hidden areas not exposed to direct visual-tactile examination:
11
II A.
B.
C.
BANTING
- if the marginal ridge shows an opacity as evidence of undermined enamel, the proximal surface is carious - any discontinuity of the enamel in which an explorer can enter is carious if it also shows other evidence of decay, such as: softness, shadow by transillumination, or loss of translucency - with radiographs, any definite radiolucency, indicating a break in the continuity of the enamel surface is carious - with transillumination, a loss of translucency producing a characteristic shadow in a calculus-free and stain-free proximal surface, is adequate evidence of caries It is interesting to also examine some of the caveats that the Task Group placed on these criteria. For instance, the report indicated that: Not every member agreed on all aspects of diagnosis. More variability was found in the examination procedures used for proximal surfaces than for other surfaces. Lesion depth is not a component of these criteria. Stain and pigmentation should not be regarded as evidence of caries since they occur in sound teeth. In borderline conditions, a decision to classify as sound or carious must be made. When in doubt, call it sound. Erosion, abrasion, hypoplasia, attrition, fractures, mottled enamel, and certain enamel opacities on exposed hard surfaces should not be classified as caries.
12
TRADITIONAL MODELS
FOR
WHO Basic Methods (1971)

A. caries will be considered present in a tooth when any lesion has a detectably softened floor, undermined enamel, or a softened wall. B. on an interproximal surface, the probe point must enter a lesion with certainty. The caveats that WHO placed on its criteria were: Where any doubt exists, caries should not be diagnosed as being present. The stages that precede cavitation and other conditions similar to the early stages of caries should be deliberately excluded, because they cannot be diagnosed positively and reliably. Defects not to be counted as caries are: (1) white and/or chalky spots, (2) discoloured or rough spots, and (3) hard-stained pits or fissures in the enamel that catch on the explorer point but do not have a detectably softened cavity, undermined enamel, or softening in the walls of the pit or fissure.
13
BANTING
NIDR Criteria (1987) Coronal Caries

Frank lesions are detected as gross cavitation. Incipient lesions may be subdivided into 3 categories according to location, each with special diagnostic considerations. These categories are: A. pits and fissures on occlusal, buccal, and lingual surfaces: these areas are diagnosed as carious when the explorer catches after insertion with moderate to firm pressure and when the catch is accompanied by 1 or more of the following signs of decay: i. softness at the base of the area. ii. opacity adjacent to the area, providing evidence of undermining or demineralization. iii. softened enamel adjacent to the area, which may be scraped away with the explorer. B. smooth areas on buccal (labial) or lingual surfaces: these areas are carious if they are decalcified or if there is a white spot as evidence of subsurface demineralization and if the area is found to be soft by: i. penetration with the explorer, or ii. scraping away the enamel with the explorer. These areas should be diagnosed as sound when there is only visual evidence of demineralization, but no evidence of softness. C. proximal surfaces: for areas exposed to direct visual and tactile examnation, as when there is no adjacent tooth, the criteria are the same as those for smooth areas on facial or lingual surfaces. For areas not available for direct visual-tactile examination, the following criterion applies: a discontinuity of the enamel in which the explorer will catch is carious if there is softness. In posterior teeth, visual evidence of undermining under a marginal ridge is not acceptable evidence of a proximal lesion unless a surface break can be entered with the explorer. In anterior teeth, however, transillumination can serve as a useful aid in discovering proximal lesions. If a characteristic shadow or loss of translucency is seen on the proximal surface, this is indicative of caries on that surface. Ideally, the actual diagnosis should be confirmed with the explorer; however, clear visualization of a lesion by transillumination can justify a positive diagnosis.
14
TRADITIONAL MODELS
FOR
Root Surface Caries Active carious lesions in root surfaces are yellow/orange, tan, or light brown in color. Lesions in remission tend to be darker, sometimes almost black. In some incipient lesion, the carious area of the root surface may merely be discolored without cavitation, but the area will be soft to exploration. Cavitation with jagged margins and a roughened, but soft floor or base usually occurring in advanced (lesions), will yield to pressure from the tip of an explorer. Areas of root caries, however, are softer than surrounding cementum; therefore, it is possible to differentiate sound from carious cementum based on tactile sense. In the presence of root caries, an explorer penetrates the tissue but usually can be removed easily. However, if the explorer penetrates but resists withdrawal or sticks, the surface is usually sound cementum. N.B. for areas without gross cavitation, visual criteria related to location, shape, and discoloration of the suspected area do not, in themselves, define root caries. The tactile criteria of softness to an explorer tip must be met for a definitive diagnosis of root caries to be made.
15
16
Conclusions from the 2002 Scotland Consensus Conference (ICW-CCT)*

*Based on a presentation made at the Indiana ConferenceClinical Models Workshop: Remin-Demin, Precavitation, Caries.
Nigel B. Pitts Dental Health Services Research Unit and Centre for Clinical Innovations, University of Dundee, Scotland, UK
Abstract Conclusions from the 2002 Scotland Consensus Conference Introduction: The International Consensus Workshop on Caries Clinical TrialsAgreeing Where the Evidence Leads (ICW-CCT) reached consensus on caries clinical trials that are scientifically acceptable as pivotal evidence of the anti-caries efficacy of oral care products. Methodology: Twentyfive review papers were debated over 3.5 days by 95 participants from 23 countries to meet the workshop objectives (to critically review modern caries definitions and concepts, the evidence on caries-diagnostic methods, potential designs for modern caries clinical trials, and the statistical approaches to analyzing modern trial data) and formulate key elements of protocol(s) for shorter, more efficient modern caries clinical trials and a framework for validating them endorsed by the group of leading international experts present at the meeting. Results: Key Consensus Statements: Caries assessment methods: in future CCT protocols, caries measurement methods are used that are capable of accurately capturing, at any given point in time, the manifestations of the caries process in dental hard tissues (enamel and dentin); when applied sequentially, these caries assessment methods can monitor definitive changes in manifestations of the caries process over time, over and above any background noise from normal levels of de- and remineralization, or from variations attributable to the caries detection system(s) used; and when applied sequentially, they can differentiate from among actual product effects in terms of group differences in lesion initiation and lesion behaviour (progression, arrest, and/or regression). If they are
17
PITTS
valid and contribute to the ability to improve trial efficiency, the new caries assessment technologies should be refined further and adopted in modern caries clinical trials. Validation of new assessment methods with clinical trials: Methods capable of recording the continuum of the caries process (including non-cavitated lesions) should be evaluated and their results compared with those of conventional caries assessment methods over a 2- to 3-year study; new caries assessment methods should have the ability to measure demineralization and remineralization of noncavitated lesions; while there are many other ways in which the design of CCTs might be improved further, through better diagnostic, design, and analytical techniques, it is paramount that the overriding principle behind CCT design validation must be that the results and conclusions from any new design are in line with those shown previously by conventional CCTs; any new design of caries clinical trial must not compromise the standard of proof of either efficacy or safety.
Introduction The International Consensus Workshop on Caries Clinical Trials Agreeing Where the Evidence Leadswas held at Cameron House on the shores of Loch Lomand, Scotland, between January 6-10, 2002. The Workshop was jointly organised and co-chaired by Nigel Pitts of the University of Dundee and John Stamm of the University of North Carolina and supported by the International Association for Dental Research (IADR), The National Institute for Dental and Craniofacial Research (NIDCR), the International Dental Federation (FDI), and major dentifrice companies. The Scientific Programme Committee comprised the co-chairs and representatives of all the supporting organisations. The purpose of this paper is to set the scene for the 2005 Indiana ConferenceClinical Models Workshop by providing an overview of what happened at the 2002 International Consensus Workshop on Caries Clinical Trials, ICW-CCT. The Mission of the 2005 Indiana Conference states: The Primary purpose of this Indiana Conference is to follow-up and build upon the results of the 2002 Loch Lomand Conference. The overall goal of the 2005 conference/workshop is to develop a general consensus among the scientific com
18
CONCLUSIONS FROM THE 2002 SCOTLAND CONSENSUS CONFERENCE (ICW-CCT)
munity regarding the acceptability of new, more efficient models for clinical trials designed to assess the impact of caries-preventive measures on remineralization, precavitation, and caries. Although the entire proceedings of the 2002 ICW-CCT Workshop are now published in a supplement to the Journal of Dental Research for reference [Pitts and Stamm, 2004], this paper highlights the main points and key Consensus Statements relating to the Indiana Conference. There is, however, much valuable detail in the 25 individual review papers that are published, in addition to the final Consensus Statements in this special issue.
Methodology: The ICW-CCT Process The Mission of the International Collaborative Workshop on Caries Clinical Trials (ICW-CCT) was to reach consensus about the design of protocols for caries clinical trials that are scientifically acceptable as pivotal evidence of the anti-caries efficacy of oral care products. A recurring theme of the ICW CCT was the desirability of identifying and evaluating methods that would lead to valid and reliable, yet shorter, more efficient caries clinical trials. The Scientific Programme Committee assembled an international panel of 25 speakers to review and present the contemporary evidence from across the disciplines who are concerned with improving the quality and efficiency of clinical trials for caries preventive agents and procedures. This evidence was then debated at length in a series of facilitated and repeated discussion sessions over 3.5 days by 95 participants from 23 countries. There was international representation from academia, industry, statisticians, methodologists, and regulators. Draft Context and Consensus Statements about the evidence and key elements of the design of future trials were presented to, debated, modified, and finally agreed upon by all participants on each evening and on the final morning. The papers outlining the reviews were then peer-reviewed prior to publication. To accomplish the ICW CCT mission and structure of the Workshop, the following 5 following objectives were set out for participants: 1. Critically review modern caries definitions and measurement concepts. 2. Critically review the evidence on caries diagnostic methods. 3. Debate potential designs for modern caries clinical trials to provide pivotal evidence of anti-caries activity.
CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 19
PITTS
Critically review the statistical approaches that could be used to analyze modern caries clinical trial data. 5. Ensure that, in meeting objectives 1-4 identified in the Workshop, participants formulate key elements of protocol(s) for shorter, more efficient modern caries clinical trials and a framework for validating them be endorsed by the leading group of experts present at the meeting. The full list of review papers that were presented to inform discussion is set out in the Journal of Dental Research Supplement. Presentations relating directly to the aims of the 2005 Indiana Conference included those on: NIH Consensus Development Conference on the Diagnosis and Management of Dental Caries throughout Life Washington D.C., March 2001 [Horowitz, 2004] The international evidence base for the continuum of the caries process [Featherstone, 2004] and the histopathology of caries related to biofilm [Kidd and Fejerskov, 2004] A systematic approach to visual caries detection [Ismail, 2004] Modern methods of caries measurement [Pitts, 2004] The potential of optical methods [Stookey, 2004] The potential of electrical methods [Longbottom and Huysmans, 2004] When designing clinical trials, it is important to understand the clinical context in which preventive caries control interventions are now delivered. Figure 1 shows a graphic representation of the 3 types of modern caries measurement: lesion measurement (assessing the extent of the lesion from early to later stages), lesion monitoring (assessing the behavior of lesions over time), and caries activity measures (which assess whether a lesion is active at a particular time point) [Pitts, 2004]. This focus on control of early stages in the disease process has been responsible for an increased focus on white (and brown) spot lesions of caries in enamel. Figure 2 shows clinical examples of non-cavitated (or pre-cavitated) caries lesions.
4.
20
FIG. 1.
Three types of modern caries measurement.
FIG. 2.
Non-cavitated or pre-cavitated caries is caries; it is amenable to preventive care, which minimises future fillings and future costs. 21
PITTS
Results of the ICW-CCT Process Consensus Statements for Objective 1: Critically Review Modern Caries Definitions and Measurement Concepts Context Dental caries is a process involving an imbalance of the normal molecular interactions between the tooth surface/subsurface and the adjacent microbial biofilm. This imbalance is manifested over time as cumulative demineralization of the tooth, which, if unchecked, has the potential to produce cavitation of the enamel and collateral damage to the dentin and pulp. Active caries is a process whereby an overly acidic environment caused by the presence of cariogenic organisms fuelled by the consumption of fermentable carbohydrates leads to the relatively rapid net destruction of hard tooth structure. Enamel caries, root caries, and dentinal caries are all variations on the same theme of bacterially induced demineralization and proteolysis. Consensus Statements (1.1) The Caries Process occurs as an interaction between biofilm and the tooth surface and subsurface. The Caries Lesion is the manifestation of the stage of the process at one point in time. [Note to readers: For clinical and epidemiological purposes, it is useful to know at one point in time whether a lesion is active, but within the clinical trial context, consecutive measures of lesion behavior provide sufficient information to establish efficacy between products.] (1.2) Caries progression occurs when the demineralization and remineralization equilibrium is out of balance, leading to net mineral loss. (1.3) Remineralization can arrest or reverse progression of disease and can lead to changes in mineral quality. (1.4) An understanding of the caries process has progressed far beyond the point of restricting evidence for dental caries to the D2 (caries in enamel only) or D3 (caries in enamel and dentin) level of cavitation. (1.5) One of the desired outcome measures in future caries clinical trials should be the arrestment or reversal of the progress of mineral loss. (1.6) Caries in enamel, dentin, and on root surfaces are all variations on the same theme of bacterially induced demineralization, alternating with remineralization, over many cycles in a lifetime.
m p t n
OBJECTIVE 2: Critically Review the Evidence on Caries-diagnostic Methods Context Existing caries-diagnostic methods used for many years in clinical practice and in clinical caries trials have centered on either clinical visuo-tactile or, in Europe in particular, clinical visual examinations. Since clinical examinations alone were shown to be relatively insensitive (although highly specific), they have often been supplemented in trials with bitewing radiography and, in some countries, Fiber-Optic Transillumination (FOTI). There have been attempts over many years to find improved methods of caries detection, and ordinally graded or continuous measures of mineral density have been developed or are under development for the measurement of net mineral loss in caries lesions over time. Proposed measures are based on disciplined standardized observation according to established techniques, technologically enhanced visual observations, new imaging technologies, and different combinations of these. There is some confusion with the terminology used in the literature around caries diagnosis (which should imply a human professional summation of all available data), lesion detection (which implies some objective method of determining whether or not disease is present), and lesion assessment (which aims to characterize or monitor a lesion, once it has been detected). Consensus Statements Existing Methods of Caries Detection and Assessment (2.1) For future clinical trials, recording only cavitated lesions as an outcome measure is becoming outmoded. (2.2) Visual inspection is the standard of caries diagnosis in Europe. Its use should continue to be evaluated in clinical trials to differentiate between stages of the carious process in different surfaces. A systematic review of the varied criteria for clinical visual assessment would help to build on the extensive literature already available. Detailed protocols for training and calibration with visual-only examination systems should be established and tested. The aim
23
PITTS
(2.3) (2.4)
(2.5) (2.6)
is to consistently achieve the high levels of inter- and intraexaminer agreement shown to be possible with such systems. The use of additional methods of caries lesion assessment to supplement visual techniques should be explored further. Bitewing radiography may add information about many of the clinical stages of the carious process at approximal surfaces and, in the case of more advanced stages, occlusal surfaces. Diagnostic yield with contemporary caries presentations must be evaluated and, for use in caries clinical trials, the prospects of radiographic examinations adding to product discrimination should be assessed. Low-dose radiographic systems and/or subtraction radiography should be considered when radiography is used for caries trials. Fiber-Optic Transillumination (FOTI) has been used successfully for caries diagnosis in some countries. Its wider use in caries clinical trials should be considered.
Validation of Methods of Caries Detection and Assessment (2.7) In vitro and/or in situ studies are still required, both should be instituted to develop and to evaluate new diagnostic techniques. (2.8) For extrapolation of the results of laboratory studies on caries diagnostic techniques to clinical use, the clinical environment, e.g., soft-tissue equivalents, moisture, etc., should be simulated as closely as possible within the laboratory setting. (2.9) There is concern about the so-called gold standards currently used in caries diagnosis/detection research, particularly for assessing occlusal and approximal surfaces. (2.10) Given the deficiency of current gold standards, and the challenges related to achieving appropriate validation, new reference standards and validation protocols should be developed. When the accuracy and precision of new caries-diagnostic methods are determined, it is necessary that the distribution of caries lesions in the in vitro or in situ assessment include all disease stages in which the diagnostic method will be used in vivo. (2.11) It should be appreciated that validation studies using immature teeth or only premolars and third molars may not be sufficiently representative to provide accurate and precise data for validation purposes.
Newer Methods of Caries Detection and Assessment (2.12) The assessment of non-cavitated lesions is essential for future CCTs, and new caries assessment methods have the prospect of helping in this task. (2.13) It would be desirable that the stage at which the carious process is measured with new diagnostic techniques relate to the extent to which the particular technique has been validated at the corresponding level of sensitivity. (2.14) New assessment techniques should, in the clinical context, be capable of measuring a wide range of lesion depths and changes in degree of mineralization, either directly or indirectly. Ideally, a continuous scale should be used. (2.15) The inherent nature of some of the techniques restricts surface types, which can be assessed, and the range of lesions, which can be measured. (2.16) Ideally, new caries assessment technologies should have the capacity to monitor changes on a highly targeted surface and on a sitespecific basis. (2.17) New measurement methods should continue to be developed. Electro-physical agents and other waves may have the potential to detect lesions and quantify changes in the caries process. Current examples are in various stages of development and include Electronic Caries Monitors (e.g., ECM), DIAGNOdent, Quantitative Light Fluorescence (QLF), Digitized Fiber Optic Transillumination (DIFOTI), and Optical Coherence Tomography (OCT). Other techniques are under development. (2.18) The new technologies should be further refined and adopted in modern caries clinical trials if they demonstrate the ability to improve trial outcomes, reduce subject numbers, and/or reduce costs. (2.19) Since some trials may, in the future, be conducted in general practice settings, and in order to optimize translation of research into practice while minimizing the risk of over-prescription, developers of new diagnostic techniques should specify the appropriate uses in caries clinical trials and/or in dental practice.
25
PITTS
(2.20) The importance of assessing newly erupted teeth with immature enamel and their related challenges should be recognized fully. (2.21) All new methodologies should have an in vitro assessment on natural caries lesions as part of the validation process. (2.22) Some new diagnostic techniques may exceed the accuracy, sensitivity, and resolution of existing validation reference standards. (2.23) Further investigation into how new technologies assess root, recurrent/secondary, and residual caries is required. (2.24) This field should be reviewed periodically to chart developments in technology and determine whether the new methods can improve on current methodology in the caries clinical trial context. OBJECTIVE 3: To Debate Potential Designs for Modern Caries Clinical Trials to Provide Pivotal Evidence of Anti-caries Efficacy Context Since the fundamental caries process varies only in detail across lesion sites and depths, and since surface demineralization is a prerequisite to dentinal caries and further damage to the tooth, the effectiveness of anticaries interventions can and should be evaluated, not only by its effect on cavitation, but also by its impact on non-cavitated lesion progression. Therefore, in future caries clinical trials, attempts should be made to incorporate appropriately standardized measurement systems to detect the effects of pre-cavitation demineralization and remineralization. Prevention of further demineralization and/or stimulation of remineralization in both non-cavitated and cavitated lesions, substantially beyond background noise, should therefore be sufficient to establish efficacy of an anticaries agent. The choice of caries measure(s) should be appropriate to the subject population, distribution of disease, and mechanism of action of the therapies studied. Site-specific determinations of mineral content and quality are not surrogate measures, but rather are primary indicators of the cumulative destructive impact of the caries process. It is an ongoing necessity to validate such measures for their power to discriminate between therapies relative to currently accepted observational methods. Clear and specific study aims are always essential to achieving a successful caries clinical trial. In addition, the use of ordinal or continuous measures of cumulative disease progression/regression, or rates of their occurrence when adequately standardized and precise, coupled with improved subject retention, offer
the hope of increased efficiency in the conduct of clinical trials of anticaries therapies. Consensus Statements For confirmatory caries clinical trials, the statements below were the object of broad consensus by the assembled workshop participants: Study Objectives and Efficacy Variables (3.1) In conventional caries trials, the primary objective is to reduce DMFS increment; and the primary efficacy variable is, accordingly, reduction of DMFS increment (over a 2- to 3-year period). (3.2) In new caries trials, the study objectives should be directed at measuring changes in the continuum of the caries process. The efficacy variables should be a measure of further demineralization and/or stimulation of remineralization in lesions (substantially beyond what would naturally occur). (3.3) Measures of mineral density change are not surrogate outcomes, but rather are primary indicators of the cumulative status of the dental caries lesion. Surrogate variables are unacceptable as primary endpoints in caries clinical trials. Target/Sample Population (3.4) As is the case in all disease-focused clinical trials, caries studies need to identify subjects who are at high caries risk for inclusion in the study. This may mean, for example, that there will be age restrictions in the studys inclusion criteria. (3.5) Within the context set by the above criteria, caries studies should attempt to have broad representation with respect to gender, race, and socio-economic status. These inclusions are to promote generalizability to the target population. Study Designs (3.6) The randomized, maximally blinded parallel group design is the preferred study design for caries trials. (3.7) As in other areas of health sciences, multi-center caries trials may be necessary to achieve sample size goals or they may contribute
PITTS
certain other advantages in some circumstances. (3.8) Factorial designs can permit the possibility of investigating multiple therapeutic objectives. (3.9) All studies should be adequately sized and powered (conventionally, 0.05 level of significance and power at least equal to 0.80). This should be the case for superiority, non-inferiority, and equivalence trials. (3.10) In non-inferiority trials, the demonstration of compliance and adherence to protocols, a definition of non-inferiority margin, and a rigorous attempt to demonstrate that the active control was effective, are essential. Randomization (3.11) The randomization process should be clearly stated and documented. Blinding/masking should be maximal to the extent possible, considering the constraints inherent in trial design and the therapies to be used. Trial Monitoring (3.12) The use of an Ethics Committee/Institutional Review Board (IRB) is essential. In addition, a Data Safety Monitoring Board (DSMB) may be required for monitoring safety and efficacy. OBJECTIVE 4: Critically Review the Statistical Approaches That Could be Used to Analyze Modern Caries Clinical Trial Data Context The data generated by caries clinical trials present numerous challenges during the statistical analysis phase. The detailed measuring of the caries status at each specific tooth surface provides extensive and potentially rich information on each study participant. That information traditionally has been compressed during statistical analysis into a single outcome data point. The evolution of statistical methods that can capitalize on correlated intra-subject data has significantly broadened opportunities for the application of modern, powerful statistical methods that offer the prospect of more efficient analyses for caries trials. The participants in caries trials generally are healthy volunteers rather than patients. The healthy volunteer has less at stake in the study intervention than has
the ill patient, often creating higher participant attrition in caries trials. On the other hand, caries-preventive treatments are generally benign and almost universally well accepted. Therefore, caries trial participants generally do not drop out due to effects linked to a poorly tolerated treatment regimen. Managing participant attrition and regimen compliance without incurring bias are important issues for data analysis of caries trials. Although caries trials have generally been analyzed by the treatment of caries increment as a continuous variable, validation studies of existing ordinally scaled clinical visual assessments should be conducted based on existing data from 2- to 3-year studies, comparing the results from ordinal scaling analyses with those from traditional caries increment analyses. Consensus Statements Following the presentation of all statistical methods papers, and the resultant plenary discussions, the Workshop participants elaborated on the following consensus statements relative to Objective 4. (4.1) The statistical analysis of a caries trial should utilize the maximum information available for the scale of measurement and the properties of the response variable (ordinal, counts, or continuous). (4.2) The statistical analysis should rely on robust techniques and be reported in a clinically interpretable manner. (4.3) Numerous statistical methodologies are available for analysis of the data from ordinal or continuous measures of change in mineralization. From among these, statistical methods should be chosen that appropriately reflect the experimental designs and efficiently capture the information produced. These include scoring of ordinal data, ordinal categorical data models, general and generalized linear models, and survival analysis methodology. (4.4) All methods should properly account for the possibility of intra-subject correlation. (4.5) For caries clinical trials, dichotomization of outcomes is often inefficient and should be avoided unless specifically justified by the declared goals of the trial. (4.6) The Intention-to-Treat (ITT) approach is problematic for extended caries trials, in view of the virtual unavoidability of extensive subject attrition, the typical loss of efficiency relative to analysis
PITTS
restricted to the subset of subjects who complete the full protocol (per protocol [PP] analysis), and the strong rationale and experiential basis for believing that dropouts occur randomly with respect to treatment. However, ITT analysis is generally recognized as the scientific standard for controlling bias and enhancing validity across the full range of medical clinical trials. (4.7) Modern statistical imputation methods may be useful in reducing the loss of efficiency from ITT relative to PP analyses and should be evaluated for their utility in caries trials. (4.8) To ensure that pivotal trials are sized adequately, the acceptability of primary PP analysis should be confirmed with regulatory authorities prior to data collection. (4.9) The appropriateness of a primary PP analysis should be explicitly justified in reports that take this approach, and a secondary ITT analysis should be provided. (4.10) In long-term caries trials reporting a primary ITT analysis, a secondary PP analysis may yield further understanding of the data. (4.11) It is essential that all trial participants are accounted for as part of the reporting process, as recommended by the CONSORT guidelines. (4.12) A considerable quantity of statistical literature exists on analytical methods when multiple outcomes are necessary to track a common underlying biological or target effect. Such methods include: (a) the use of a multivariate test statistic accounting for outcome variable correlation, or (b) appropriately standardized linear or non-linear combinations of outcomes to generate a composite variable or scale for analysis utilizing a global test statistic. (4.13) The magnitude of statistically significant treatment effects should be estimated and interpreted in clinical context. The interpretation of caries trials in terms of clinical significance requires further consideration OBJECTIVE 5: To ensure that, in meeting Objectives 1-4 (critically reviewing modern caries definitions and concepts, the evidence on cariesdiagnostic methods, potential designs for modern caries clinical trials, and the statistical approaches to analyze modern trial data), the Workshop for30 INDIANA CONFERENCE 2005
mulates key elements of protocol(s) for shorter and more efficient modern caries clinical trials and a framework for validating them endorsed by the group of leading international experts present at the meeting. Context The 4 sections above have, under the different objectives, already outlined several of the key elements that should be introduced into protocols for efficient, modern, caries clinical trials. In addition, there are several over-arching elements that were discussed and agreed on at the Workshop. These include the philosophy that clinical trials of anticaries agents should be considered within the same framework as clinical trials in other health care arenas. Hence, both the design and implementation of future caries trials should be considered in the full context of the modern science of clinical trial methodology, applied as appropriate to the special needs of oral health studies. Consensus Statements (5.1) In light of the evidence reviewed, both here and elsewhere pertaining to modern caries definitions and measurement concepts, the participants supported a statement recommending that, in future CCT protocols, caries measurement methods are to be used, which: are capable of accurately capturing, at any given point in time, the manifestations of the caries process in dental hard tissues (enamel and dentin); when applied sequentially, can monitor definitive changes in manifestations of the caries process over time, over and above any background noise from normal levels of de- and remineralization, or from variations attributable to the caries detection system(s) used; and when applied sequentially, can differentiate among actual product effects in terms of group differences in lesion initiation and lesion behaviour (progression, arrest, and/or regression). (5.2) If they are valid and contribute to the ability to improve trial efficiency, the new caries assessment technologies should be refined further and adopted in modern caries clinical trials.
31
PITTS
Validation of New Assessment Methods with Clinical Trials (5.3) Methods capable of recording the continuum of the caries process (including non-cavitated lesions) should be evaluated, and their results compared with those of conventional caries assessment methods over a 2- to 3-year study. (5.4) New caries assessment methods should have the ability to measure demineralization and remineralization of noncavitated lesions. (5.5) While there are many other ways in which the design of CCTs might be improved further, through better diagnostic, design, and analytical techniques, it is paramount that the overriding principle behind CCT design validation must be that the results and conclusions from any new design are in line with those shown previously by conventional CCTs. (5.6) Any new design of the caries clinical trial must not compromise the standard of proof of either efficacy or safety. Learnings From the ICW-CCT This meeting represented what participants claimed was an all too rare opportunity to hold an open, joint forum involving academic and commercial cariologists with input from industry, statisticians, and regulators. It employed an Evidence-Based approach, using systematic reviews of the literature, wherever possible, as well as incorporating much discussion and peer group synthesis. There will be a need for continuing activities to overcome inertia in this field and a push to implement new methods, which respond to new knowledge in the cariology field and which unlock the innovation needed to establish and test even more effective means of caries control. There is also a recognised need to ensure that any new standards or processes are updateable and to evolve as new data comes on stream from around the world. Following the ICW-CCT meeting, an unfunded ad hoc group of volunteers met to try to refine some of the tasks identified by the Workshop, which related to clinical visual examination, then to take this agenda forward. This initiative led to the formation of the initial ICDAS Committee in the spring of 2002. Overview Who participated? Ninety-five researchers from 23 countries, with representation from academia, industry, statistics, and regulatorsthis appears to
have been the first comprehensive position review of caries clinical trials since 1968. What was the outcome? A unique multidisciplinary forum produced and agreed to detailed Consensus Statements, including the one that states that: For future clinical trials, recording cavitated lesions only as an outcome measure is becoming outmodedthe assessment of non-cavitated lesions is essential for future CCTs It is hoped that the ICW CCT proceedings and consensus statements will provide an increased understanding and guidance for the future conduct of caries clinical trials.
Acknowledgements The author is indebted to many individuals and organisations for contributing material presented in this paper. In particular, thanks are due to: John Stamm, a tireless co-chair and source of advice The Programme Committee, who chose extraordinarily well The Speakers, who all fulfilled their charge The Session chairs, who helped marshal the extended discussions The Delegates, who made the Consensus Statements coherent The Sponsors, who made the Workshop possible: - IADR, NIDCR, FDI - GlaxoSmithKline Inc., Procter & Gamble, Unilever, ColgatePalmolive Company, Sunstar, Gaba International Lion, and Wrigley This paper expresses the personal views of the author.
References
Featherstone JDB: The continuum of dental cariesevidence for a dynamic disease process. J Dent Res 2004;83 (Spec. Issue C): C39-C42. Horowitz AM: A report on the NIH Consensus Development Conference on Diagnosis and Management of Dental Caries Throughout Life. J Dent Res 2004;83 (Spec. Issue C): C15-C17. Ismail AI: Visual and visuo-tactile detection of dental caries. J Dent Res 2004;83 (Spec. Issue C): C56-C66.
33
PITTS
Kidd EAM, Fejerskov O: What constitutes dental caries? Histopathology of carious enamel and dentin related to the action of cariogenic biofilms. J Dent Res 2004;83 (Spec. Issue C): C35-C38. Longbottom C, Huysmans M-C DNJM: Electrical measurements for use in caries clinical trials. J Dent Res 2004;83 (Spec. Issue C): C76-C79. Pitts NB: Modern concepts of caries measurement. J Dent R 2004a;83 Spec. Issue C: 43-47. Pitts NB, Stamm J: ICW-CCT consensus statements. J Dent R 2004;83 Spec. Issue C: 125-128. Stookey G: Optical methodsQuantitative light fluorescence. J Dent Res 2004;83 Spec. Issue C: C84-C88.
34
ROC Analysis with Applications to Dental Research

Nancy A. Obuchowski Department of Quantitative Health Sciences The Cleveland Clinic Foundation Cleveland, OH, USA
Introduction Receiver Operating Characteristic (ROC) curve analysis has been used in many medical and non-medical fields to describe and compare the performance of predictors. For example, in diagnostic radiology, ROC curves are used to assess and compare imaging tests (the predictors) for diagnosing and screening for disease. ROC analysis is a powerful tool, and many statistical methods and software have been developed for it [Zhou et al., 2002; Pepe, 2003; ROC analysis software]. In this article, I discuss aspects of ROC curve analysis that are particularly relevant to dental research. Consider the following example,1 which I use throughout the paper to illustrate various ideas and statistical methods. Two hundred and twenty-seven sites on 83 exfoliated teeth from 35 patients were evaluated by each of 3 diagnostic methods (the predictors) for detecting caries [Katz Personal communication]. The 3 diagnostic methods were: 1) the DIAGNOdent reading, 2) delta_q, which is one of the measures generated by QLF, and 3) area, another measure generated by QLF. In this example, the study goal is to compare the 3 diagnostic methods to determine which is better at detecting caries. Thin sectioning was used to establish definitively whether or not caries was present. Such a procedure is called the gold standard or reference standard [Zhou et al., 2002; Weinstein et al., 2005]. For each site, a Plm result, scored on a 0-4 scale, was obtained. A Plm result of 1 or greater is considered as definitive for caries. Thus, the gold standard was used to establish between 2 truth states: caries absent vs. caries pres1
The data used in this example are a subset from a much larger study and, thus, are merely illustrative of the statistical methods described in this paper.
35
OBUCHOWSKI
ent (so-called binary-scale gold standard). We will also consider situations where the gold standard is on an ordinal-scale. In this discussion, we briefly address the situation when no gold standard data is available. In this example, multiple sites on the same tooth were tested, and multiple teeth from the same patient were included in the study sample. Multiple observations from the same patient are called clustered data. Observations from the same patient are often correlated, at least to some small degree. Ignoring this correlation leads to estimates of variance and standard errors that are artificially small, which can result in misleading conclusions. I will discuss methods for clustered ROC curve data. Finally, this example illustrates the comparison of 3 diagnostic tests performed, not on different samples of patients (unpaired or independent design), but rather on the same sample of patients (paired design). I will discuss a statistical test that takes advantage of this paired design. I conclude with a brief discussion of the 3 phases of assessment for evaluating diagnostic tests.
MATERIALS AND METHODS Describing a Diagnostic Tests Performance Sensitivity and Specificity There are 2 basic measures of the inherent accuracy of a diagnostic test: sensitivity and specificity. Sensitivity is the probability of a positive test result (e.g., the test indicates the presence of caries) for a site with caries. Specificity is the probability of a negative test result (e.g., the test does not indicate the presence of caries) for a site without caries. Table 1 illustrates the definitions of sensitivity and specificity. The rows of the table give the results of the diagnostic test as either positive for the presence of caries or negative for caries. The columns indicate the true caries status. True positives (TP) are those sites where the Plm result is positive for caries and the diagnostic test is positive for caries. True negatives (TN) are sites without caries that test negative. False negatives (FN) are those with caries, but the test falsely indicates that caries is not present. False positives (FP) are those without caries, but the test falsely indicates the presence of caries. Sensitivity, then, is the probability of a TP among sites with caries (TPs + FNs); i.e., sensitivity =
ROC ANALYSIS WITH APPLICATIONS TO DENTAL RESEARCH

Caries Present1 Test positive2 Test negative2
1
2
Caries Absent1 3 (FP) 175 (TN)
6 (TP) 43 (FN)
based on a Plm score <1 vs. 1; based on a DIAGNOdent reading <15 vs. 15.
TABLE 1.
Calculating sensitivity and specificity.
6/(6 + 43) = 0.12. Specificity is the probability of a TN among sites without caries (TNs + FPs); i.e., specificity = 175/(175 + 3) = 0.98. It is often difficult to compare 2 diagnostic tests using the accuracy metrics of sensitivity and specificity. First, what if 1 test has better sensitivity than the other, but the other test is more specific? Second, what cutoff should be used to define a positive vs. negative test result? ROC Curves DIAGNOdent readings actually range from 0 to 89, and, thus, are not inherently positiveor negative as illustrated in table 1. Consider the subset of DIAGNOdent results given in table 2. Each unique DIAGNOdent value can be used as a cutoff, or decision threshold. We can calculate the sensitivity and specificity at each such cutoff. Sites with DIAGNOdent values less than or equal to the cutoff are called negative for caries; sites with DIAGNOdent values greater than the cutoff are called positive for caries. The data in table 2 are presented in table 3, sorted by DIAGNOdent reading value. The third and fourth columns in table 3 show the calculated sensitivity and specificity for the cutoffs listed in the second column of table 3. Note that, as the value of the cutoff increases, the specificity decreases, while the sensitivity increases. The pairs of sensitivity and specificity calculated in table 3 are plotted together in fig. 1. The y-axis is the sensitivity; the x-axis is 1 minus the specificity, or the false positive rate [FPR]. The Receiver Operating Characteristic [ROC] curve is formed by connecting the sensitivity/FPR points with line segments [Zhou et al., 2002; Pepe, 2003; Zweig and Campbell, 1993]. A diagnostic test (i.e., predictor) with an ROC curve that lies near the diagonal line (chance diagonal) has no true ability to distinguish between sites with and
37
OBUCHOWSKI
FIG. 1.
without caries. A diagnostic test whose ROC curve is near the upper left corner (i.e., near 100% sensitivity and 0% FPR [100% specificity]) has almost perfect discrimination of caries from no caries. DIAGNOdent reading has moderate accuracy, lying between the diagonal and the upper left corner. Area Under the ROC Curve The area under the ROC curve is often used to quantify the ability of a diagnostic test to discriminate disease from no disease. It incorporates both the sensitivity and specificity of the diagnostic test into a single metric of accuracy. An ROC area near 0.5 describes a diagnostic test with no discrimination ability, beyond mere guessing; whereas, an ROC area of 1 has perfect discrimination ability. The area under the ROC curve has the following formal interpretation [Hanley and McNeil, 1982]: Of 2 randomly chosen sites, 1 with caries and 1 without, the ROC area is the probability that the site with caries will have a higher (greater suspicion of caries) result on the diagnostic test than the site

Patient 1 2 Plm Result1 0 2 0 0 0 3 4 0 0 3 4 1 0 0 1 DIAGNOdent 1 10 0 3 0 5 1 1 0 25 5 1 21 8 89
3 4 5 6
TABLE 2.
Diagnostic test data for 6 patients.
Plm Result 0 0 0 0 0 >0 >0 0 >0 >0 0 >0 0 >0 >0
DIAGNOdent 0 0 0 1 1 1 1 3 5 5 8 10 21 25 89
Sensitivity 1.0 1.0 1.0 0.714 0.714 0.714 0.714 0.714 0.429 0.429 0.429 0.286 0.286 0.143 0.0
Specificity 0.375 0.375 0.375 0.625 0.625 0.625 0.625 0.750 0.750 0.750 0.875 0.875 1.0 1.0 1.0
FPR 0.625 0.625 0.625 0.375 0.375 0.375 0.375 0.250 0.250 0.250 0.125 0.125 0.0 0.0 0.0
FPR = false positive rate, or 1- specificity
TABLE 3.
Diagnostic test data sorted by DIAGNOdent result.
without caries. The ROC area can also be interpreted as the average sensitivity for all FPRs, or the average FPR over all sensitivities [Metz, 1989]. The area under the ROC curve is calculated as follows: Let Xi denote the result of the diagnostic test (e.g., DIAGNOdent reading) for the i-th site with caries (as determined by the gold standard), and let Yj denote the result of the diagnostic test for the j-th site without caries. A nonparametric estimator of the area under the ROC curve, AUC, is: AUC = [D + 0.5 x E]/(m x n)
39
OBUCHOWSKI
where n is the number of sites with caries (as determined by the gold standard) and m is the number of sites without caries in the study sample. If one were to compare the diagnostic tests results of each of the nsites with caries to each of the m sites without caries, then there would be m x n such comparisons. D is the number of these comparisons, where the site with caries has a higher (more suspicious) score on the diagnostic test than the site without caries; E is the number of comparisons, where the 2 sites have the same score on the diagnostic test. Table 4 illustrates the calculation of the ROC area for the 15 observations given in table 2. In the next 2 sections, I will discuss methods for formally testing the null hypothesis that 2 diagnostic tests have the same accuracy versus the alternative hypothesis that their accuracies differ. Note that I have discussed nonparametric AUC estimates, meaning that no assumptions are made about the underlying distribution of the diagnostic test results. Parametric methods also exist [Dorfman and Alf, 1968, 1969], along with software (ROCKIT). Clustered Data Clustered data are common in many medical fields, e.g., 2 eyes, 4 heart chambers, multiple segments of a vessel, etc. In the example described in this paper, there are 2 ways in which the observations are clustered. First, the sample contains multiple teeth from the same patients. Specifically, there are, on average, 2.4 teeth per patient. Second, there are multiple sites tested from the same tooth. There are, on average, 2.7 sites per tooth. Clustered data does not affect calculation of the AUC. Clustered data does, however, affect calculation of the variance of the AUC. Calculation of the variDIAGNOdent Reading for Caries Site 1 1 5 5 10 25 89 sum of column
1
Calculation of D1 3 3 6 6 7 8 8 D = 41
Calculation of E2 2 2 0 0 0 0 0 E=4
For each caries site, the DIAGNOdent reading is compared to the DIAGNOdent reading of all sites without caries. This column indicates the number of such comparisons where the DIAGNOdent reading at the caries site exceeds the DIAGNOdent reading at the non-caries site. 2 For each caries site, the DIAGNOdent reading is compared to the DIAGNOdent reading of all sites without caries. This column indicates the number of such comparisons where the DIAGNOdent readings at the two sites are equal. Then, AUC = [41 + 0.5 x 4]/(8 x 7) = 0.768.
TABLE 4.
Calculation of AUC from 15 observations in Table 3.
40
ance of AUC for unclustered data is described by DeLong et al. [1988]. Here, I describe calculation of the variance of AUC for clustered data [Obuchowski, 1997]. Let I denote the number of clusters in the study sample (e.g., number of patients, I = 35). Let I1 denote the number of clusters with at least 1 site with caries, and I2 denote the number of clusters with at least 1 site without caries. Note that I1 + I2 often do not equal I. Here, there are 22 patients with 1 or more caries (I1 = 22) and 34 patients with 1 or more sites without caries (I2 = 34). Let ni denote the number of sites with caries in the ith patient, and mi denote the number of sites without caries in the ith patient. There are a total of n = 49 and m = 178 sites with and without caries in the sample, respectively. The variance of the AUC, estimated from clustered data, is: Variance (AUC) = (1/n) x C1 + (1/m) x C2 + (2/mn) x C3 where C1 = [I1/(n x (I1-1))] x [i (B1i - ni x AUC)2], where the summation is over the I1 patients with at least 1 caries site, C2 = [I2/(m x (I2-1))] x [i (B2i - mi x AUC)2], where the summation is over the I2 patients with at least 1 site without caries, and C3 = [I/(I-1)] x [i (B1i - ni x AUC) x (B2i - mi x AUC)], where the summation is over all I patients. Now consider a particular caries site (labeled k) in the ith patient. Compare the diagnostic test result of the k-th site in the i-th patient to each of the m sites without caries. Let A1ik denote the number of these comparisons, where the site with caries has a higher (more suspicious) score on the diagnostic test than the site without caries, and let T1ik denote the number of these comparisons, where the sites with and without caries have the same test results. Similarly, consider a particular site without caries (labeled l) in the i-th patient. Compare the diagnostic test result of the l-th site in the i-th patient to each of the n sites with caries. Let A2il denote the number of these comparisons, where the site without caries has a lower (less suspicious) score on the diagnostic test than the site with caries, and let T2il denote the number of these comparisons, where the sites with and without caries have the same test results. B1i is the sum of the A1iks and 0.5 x T1iks for the ith patient, divided by m. B2i is the sum of the A2ils and 0.5 x T2ils for the i-th patient, divided by n. In particular, B1i = [1m] [k(A1ik + 0.5 x T1ik)], where the summation is over the ni sites with caries in the i-th patient; B2i = [1/n] [k(A2il + 0.5 x T2il)], where the summation is over the mi sites without caries in the ith patient. Table 5 illustrates the
41
OBUCHOWSKI
calculation of the variance of the AUC for DIAGNOdent reading based on the 15 observations in table 2. Comparing the Accuracy of Diagnostic Tests Suppose that we want to compare the accuracy of DIAGNOdent reading to delta_q. Since the diagnostic tests were performed on the same sites, the AUC of DIAGNOdent is correlated to the AUC of delta_q. We capture this correlation by the covariance between the 2 AUCs. In this example, the covariance needs also to take into account the clustered nature of the data.
Patient i = 1
Plm result 0
DIAGNOdent 1 >0
A1ik -
T1ik 10
A2il 5
T2il 2 7
B1i 0 0.875
B2i -0.857
Patient i = 2
0 0 0 >0
0 3 0 5
7 5 7 -
0 0 0 0.750 2.714
Patient i = 3
>0
0.500 0.857
Patient I = 4 Patient I = 5
0 0 >0 >0
1 0 25 5
8 6
0 0
5 7 -
2 0 -
1.750 Patient I = 6 >0 0 0 >0 1 21 8 89 3 8 2 0 2 3 0 0 1.500
1.000
0.714
A1ik is the number of comparisons, where the site with caries has a higher score than the site without caries. T1ik is the number of comparisons, where the sites with and without caries have the same test results. A2il is the number of comparisons, where the site without caries has a lower score than the site with caries. T2il denote the number of comparisons, where the sites with and without caries have the same test results. B1i is the sum of the A1iks and 0.5 x T1iks for the ith patient, divided by m. B2i is the sum of the A2iks and 0.5 x T2iks for the ith patient, divided by n. Then, C1 = [I1/(n x (I1-1))] x [i (Bli - ni x AUC)2] = [5/(7 x (5-1))] x [(0.875-0.768)2 + (0.750-0.768)2 + (0.5000.768)2 + (1.750-2 x 0.768)2 + (1.500-2 x 0.768)2] = 0.0233 C2 = [I2/(m x (I2-1))] x [i (B2i - mi x AUC)2] = [5/(8 x(5-1))] x [(0.857-0.768)2 + (2.714-3 x 0.768)2 + (0.8570.768)2 + (1.0-0.768)2 + (0.714-2 x 0.768)2] = 0.1427 C3 = [I/(I-1)] x [i (Bli - ni x AUC) x (B2i - mi x AUC)] = [6/(6-1)] x [(0.875-0.768) x (0.857-0.768) + (0.750-0.768) x (2.714-3 x 0.768) + (1.750-2 x 0.768) x (1.0-0.768) + (1.500-2 x 0.768) x (0.714-2 x 0.768)] = 0.09766 Finally, the estimated Variance (AUC) = (1/n) x C1 + (1/m) x C2 + (2/mn) x C3 = (1/7) x 0.0233 + (1/8) x 0.1427+(2/56) x 0.0976 = 0.0247.
TABLE 5.
Calculation of the Variance of AUC for 15 observations in Table 2.
42
The covariance between 2 ROC areas, AUC1 and AUC2, can be estimated by [Obuchowski, 1997]. Covariance (AUC1, AUC2) = (1/n) x S1 + (1/m) x S2 + (1/mn) x S3+ (1/mn) x S4 where S1 = [I1/((I1-1) x n)] x [i (B11i - ni x AUC1) x (B21i - ni x AUC2)], where the summation is over the I1 patients with at least 1 caries site, S2 = [I2/((I2-1) x m)] x [i (B12i - mi x AUC1) x (B22i - mi x AUC2)], where the summation is over the I2 patients with at least 1 site without caries, S3 = [I/(I-1)] x [i (B11i - ni x AUC1) x (B22i - mi x AUC2)], where the summation is over all I patients, and S4 = [I/(I-1)] x [i (B12i - mi x AUC1) x (B21i - ni x AUC2)], where the summation is over all I patients. To test the null hypothesis that the accuracy of the DIAGNOdent reading equals the accuracy of delta_q, versus the alternative hypothesis that the diagnostic tests accuracies differ, we use the following test statistic: Z = [AUC1 AUC2] [var1 + var2 2 x cov] where var1 is the estimated variance of AUC1 (DIAGNOdent reading), var2 is the estimated variance of AUC2 (delta_q) and cov is the covariance between AUC1 and AUC2. The test statistic Z follows a standard normal distribution. For a 2-tailed test with significance level of 0.05, the critical values are 1.96 and +1.96. If Z is less than -1.96, then we conclude that the accuracy of delta_q is superior to that of the DIAGNOdent reading; if Z exceeds +1.96, then we conclude that the accuracy of the DIAGNOdent reading is superior to that of delta_q. When the Gold Standard is Not Binary-Scale Sometimes, the gold standard does not inherently produce binary results (e.g., caries versus not caries), but rather is ordinal or continuous. For example, the Plm result is on an ordinal scale with 5 ranked values: 0 to 4. A simple measure of accuracy, with an interpretation similar to the area under the ROC curve, can be calculated for diagnostic tests whose gold standard is ordinal- or continuous-scale [Obuchowski, 2005]. Let T denote the total number of ordered values of the gold standard. For Plm, T = 5. We denote the ordered values as t = 1, 2, 3, T. From T = 5 ordered values of the gold standard, we can calculate 10 ROC curves and their associated areas: t = 1 vs. 2, t = 1 vs. 3, t = 1 vs. 4, t = 1 vs. 5, t = 2 vs. 3, t = 2 vs. 4,
43
OBUCHOWSKI
t = 2 vs. 5, t = 3 vs. 4, t = 3 vs. 5, and t = 4 vs. 5. A measure of accuracy for ordinal-scale gold standards is: Accuracy = stwst x AUCst (st) where the summations are over the T ordered values of the gold standard, AUCst is the area under the ROC curve for gold standard values of s vs. t, and wst is a weight for gold standard values of s vs. t. The weights are often based on the relative sample sizes in the study sample as follows: wst = (ns x nt)/[gh (ng x nh)] (where h>g) where the summations are over the T-ordered values of the gold standard, and nt is the number of patients with gold standard results of t. The interpretation is as follows: of two randomly chosen patients, sampled from different truth states, the measured accuracy is the probability that the patient with the higher truth state (e.g., higher Plm result) has a higher test score (e.g., higher DIAGNOdent result) than the patient with the lower truth state. The variance of this measure of accuracy is the sum of the estimated variances and covariances of the T outcomes. See Obuchowski, 2005 for details.
RESULTS As we saw in table 1, the sensitivity of the DIAGNOdent reading, based on a cutoff of 15, is: DIAGNOdent reading: sensitivity = 0.12, specificity = 0.98. Using cutoffs of <0 and >0 for defining positive results on the delta_q and area, respectively, the sensitivity and specificity of these tests are: delta_q: sensitivity = 0.45, specificity = 0.54 area: sensitivity = 0.43, specificity = 0.60. Comparing the 3 diagnostic tests, delta_q has the best sensitivity but lowest specificity; the DIAGNOdent reading has the best specificity but the lowest sensitivity. Thus, it is difficult to identify the most accurate diagnostic test based on sensitivity and specificity. The ROC methods are now applied to the 227 observations in the dataset. Figure 2 illustrates the ROC curves of the DIAGNOdent reading and delta_q. The ROC curve of area is similar to the ROC curve of delta_q, and, thus, is not plotted. Note that the ROC curve of the DIAGNOdent reading is superior to that of the delta_q at all false positive rates.
FIG. 2.
The estimated AUC for the DIAGNOdent reading is 0.688, for delta_q, the AUC equals 0.538, and for area, the AUC is 0.516. The variances of the AUCs for the DIAGNOdent reading, delta_q, and area are 0.00200, 0.00229, and 0.00230, respectively. The covariances are as follows: DIAGNOdent reading and delta_q is 0.00015, DIAGNOdent reading and area is 0.000004, and area and delta_q is 0.002038. The Z statistic for comparing the accuracies of the DIAGNOdent reading and delta_q equals +2.4. Thus, we conclude that the accuracy of the DIAGNOdent reading exceeds the accuracy of delta_q, with a p-value of 0.016. Using similar methods, the Z statistic for comparing DIAGNOdent and area is 2.6, and for the comparison of area to delta_q, the Z statistic is -1.0. Thus, we conclude that the DIAGNOdent reading is more accurate than both the delta_q and area, and the delta_q and area have similar accuracies. Now consider the Plm results as ordered values, rather than binary values. There are 178 sites with a Plm result of 0, 25 sites with a result of 1, and 7, 12, and 5 sites with Plm results of 2, 3, and 4, respectively. Sites with Plm results of 2-4 will be pooled because of the small sample size. Table 6 summarizes the ROC areas for comparing Plm results of 0 to 1, 0 to >1, and 1 to >1. Note that
OBUCHOWSKI
the accuracies of the delta_q and area are all near 0.5 (i.e., no diagnostic ability). The accuracy of the DIAGNOdent reading is also poor for distinguishing Plm results of 0 vs. 1 but is quite good for distinguishing Plm results of 0 vs. >1 and distinguishing 1 vs. >1. DISCUSSION
ROC Area Estimates Plm_result 0 vs. 1 0 vs. >1 1 vs. >1 Summary weight 0.477 0.458 0.064 DIAGNOdent 0.543 0.839 0.814 0.695 Delta_q 0.529 0.491 0.518 0.510 Area 0.517 0.486 0.503 0.501
TABLE 6.
ROC area estimates for contrasting different Plm results.
With any statistical analyses, it is important that the analytical methods are applied in the correct context and the generated results are interpreted appropriately. In diagnostic medicine, 3 phases to assessment of a diagnostic tests accuracy have been identified [Zhou et al., 2002]; these phases guide interpretation of the study results. I briefly discuss these 3 phases as they relate to dental research. The first phase is the exploratory phase, where the new diagnostic test is first performed on patients. These studies are small, perhaps 10 to 20 patients. The sites selected for the study samples are often sites with classical overt disease [e.g., symptomatic caries] and sites with no indication of disease. If the new diagnostic test is not able to distinguish overt caries from healthy teeth, then it is not worth pursuing the diagnostic test further. The ROC area is a useful measure of accuracy for these studies. If the ROC area does not exceed 0.5, then the diagnostic test has no value. Note that the measured accuracy from exploratory studies is often inflated, because the study samples, by design, are easy to diagnose. It is important to recognize this when interpreting the study results. The second phase is the challenge phase. In this phase, we select patients with subtle, or early disease, with comorbidities that could interfere with the diagnostic test [Ransohoff and Feinstein, 1978]. The goal of these studies is to challenge the diagnostic test and identify problematic subpopulations. These studies often include competing diagnostic tests to see how the new test com46 INDIANA CONFERENCE 2005
pares in difficult cases. ROC curves and their areas can be used to assess and compare the tests. If the diagnostic test shows good accuracy for these difficult cases, then it can be considered for the third phase of assessment. The third phase is the advanced phase. These are prospective studies, often involving large numbers of patients (maybe 100s). The patient sample for these studies must be representative of the target population. For example, if the diagnostic test is intended for screening, then the study sample must be representative of asymptomatic people. Advanced phase studies provide reliable estimates of a diagnostic tests accuracy for the target population. I conclude with some remarks about gold standards. In caries research, studies with a valid gold standard, such as thin sectioning of exfoliated teeth, are uncommon. There are several modeling approaches to ROC analysis without a gold standard [Zhou et al., 2002; Qu et al., 1996; and Espeland et al., 1989]. These methods usually require larger sample sizes and may require modification for clustered data. Huysmans and Longbottom [2004] discuss the challenges of selecting and validating gold standards in dental research.
References
DeLong ER, DeLong DM, Clarke-Pearson DL: Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 1988;44:837-844. Dorfman DD, Alf E: Maximum likelihood estimation of parameters of signal detection theorya direct solution. Psychometrika 1968;33:117-124. Dorfman DD, Alf E: Maximum-likelihood estimation of parameters of signal detection theory and determination of confidence intervalsrating method data. Journal of Mathematical Psychology 1969;6:487-496. Espeland MA, Platt OS, Gallagher D: Joint estimation of incidence and diagnosis error rates from irregular longitudinal data. JASA 1989; 84:972-979. Hanley JA, McNeil BJ: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29-36. Huysmans MC, Longbottom C: The challenges of validating diagnostic methods and selecting appropriate gold standards. J Dent Res 2004;83C. Metz CE: Some practical issues of experimental design and data analysis in radiologic ROC studies. Investigative Radiol 1989;24:234-245. Obuchowski NA: Nonparametric analysis of clustered ROC curve data. Biometrics 1997;53:170-180.
47
OBUCHOWSKI
Obuchowski NA: Estimating and comparing diagnostic tests accuracy when the gold standard is not binary. Acad Radiol. In press. Pepe MS: The statistical evaluation of medical tests for classification and prediction. Oxford University Press, United Kingdom 2003. Qu Y, Tan M, Kutner MH: Random effects models in latent class analysis for evaluating accuracy of diagnostic tests. Biometrics 1996;52:797-810. Ransohoff DJ, Feinstein AR: Problems of spectrum and verification bias in evaluating the efficacy of diagnostic tests. NEJM 1978;299:926-930. ROC Analysis. http://www.bio.ri.ccf.org/Research/ROC/index.html ROCKIT, available at http://xray.bsd.uchicago.edu/krl/KRL_ROC/software_index.htm. Weinstein S, Obuchowski NA, Lieber ML: Clinical evaluation of diagnostic tests. AJR, in press. Zhou XH, Obuchowski NA, McClish DK: Statistical methods in diagnostic medicine. Wiley-Interscience 2002. Zweig MH, Campbell G: Receiver-Operating Characteristic (ROC) Plots: a fundamental evaluation tool in clinical medicine. Clin Chem 1993;39:561-577.
48
Including Non-cavitated Lesions in Pivotal Clinical Trials: Issues and Concerns

Philippe P. Hujoel,1-2 Walter Bretz,3-4 and James Bader5
Department of Dental Public Health Sciences, School of Dentistry Department of Epidemiology University of Washington, Seattle, WA, USA
2
Division of Pediatrics & Developmental Dental Sciences School of Dental Medicine 4 Department of Epidemiology, Graduate School of Public Health University of Pittsburgh Pittsburgh, PA, USA
3
Cecil G. Sheps Center for Health Services Research University of North Carolina at Chapel Hill Chapel Hill , NC, USA
5
Abstract Pressure is increasing to approve caries preventive agents based on trials of shorter duration and smaller patient sample size. One approach to achieve this goal to incorporate non-cavitated lesions in the caries scoring. Experiences with subtle anatomical changes in other chronic disease areas suggest that such an approach should be taken cautiously. First, novel caries treatments may have billions of cradle-to-grave users and must be safe; yet, their safety cannot adequately be assessed in small-sample, short-term studies. Still smaller sample sizes and durations of caries trials, thanks to the inclusion of non-cavitated lesions in the caries evaluation, will make it even more challenging to assess long-term safety. Second, both the existence of effective caries treatments, combined with the non-life threatening nature of caries, minimizes ethical pressures
49
HUJOEL
ET AL.
for fast drug-approval, i.e., there are no ethical pressure to include subtle caries measures to shorten clinical trials. Third, there is sufficient evidence to question whether treatments that have a beneficial impact on non-cavitated lesions will have a beneficial impact on the real, tangible outcomes of the caries process: pain and loss of function. Further efforts should be spent on validating non-cavitated lesions so that promising caries preventive agents can be efficiently and reliably identified in exploratory trials aimed at screening for caries preventive agents or determining optimal dosing. Whether non-cavitated lesions can be used in pivotal trials conducted with the aim of obtaining regulatory drug approval will remain controversial and may be influenced by the type of claims desired, safety concerns, and experiences in other chronic disease areas.
Introduction Caries is one of the most common complex chronic diseases affecting mankind [Petersen, 2003]. Caries can cause pain, tooth loss and edentulism, facial disfigurement, difficulty in chewing, and limitation in food choices, all of which can impact quality of life. Because caries affects billions of otherwise healthy individuals, evaluation of the safety and efficacy of caries prevention and therapy is complex. For example, decisions to expose all inhabitants of large regions or countries to water fluoridation from cradle to grave are controversial; some countries such as Sweden or Japan do not fluoridate or have stopped fluoridation, while professional organizations or governments in other countries such as the U.S. or Canada endorse fluoridation [Carmona, 2004]. Similarly, different countries have different regulations regarding the use of mercury-containing dental amalgams during pregnancy [Hujoel et al., 2005]. The most reliable tool to assess the safety and efficacy of treatments is the randomized controlled clinical trial (RCT), and this approach has been used widely in caries research to assess a range of interventions including oral hygiene practices, topical and systemic fluorides, sugar substitutes, antibacterial agents, and probiotic interventions. Typically, pivotal clinical trials enroll more than a thousand subjects and last for at least 2 years [Kingman, 2005]. The search for populations with high caries rates and simple follow-up mechanisms often result in conducting trials in children or adolescents within a school set50 INDIANA CONFERENCE 2005
NON-CAVITATED LESIONS
IN
PIVOTAL CLINICAL TRIALS: ISSUES
AND
CONCERNS
ting, and investigators preferably having to conduct such studies in countries where the caries rates have remained high. There is a perception that such clinical trials are becoming too expensive and that various aspects of the study design need to be modified to reduce costs and speed up the regulatory process in drug approval. One important regulatory issue in this search for more efficient study designs is how treatment effectiveness should be assessed in pivotal trials those clinical trials designed and conducted with the aim of leading to drug approval. At one extreme, caries preventive drugs could be approved based on subtle surrogate endpoints, such as non-cavitated caries lesions detected by technologies approved by the Food and Drug Administration, such as Quantitative Light Fluorescence (QLF) or DIAGNOdent [Stookey and Gonzlez-Cabezas, 2001]. At the other extreme, clinical trials could be required to assess clinically relevant endpoints such as pain, discomfort, or frank cavitations. Deciding what measure of treatment efficacy to use has a substantial impact on the cost and duration of the study. The earlier the outcome of the caries process can be measured, the more subtle the selected caries measure, the smaller the required patient sample size, and/or the shorter the duration of the trial, with the result that the conduct of the trial becomes less expensive and greater uncertainty exists as to whether the subtle benefits will translate into real clinical benefits. The selection of a measure of treatment efficacy can be controversial for any complex chronic disease [Fleming and DeMets, 1996]. Should cardiovascular medications be evaluated based on their effect on blood pressure or based on the extent to which they increase lifespan? Should clinical trials on osteoporotic drugs measure bone density or observe fractures? The discovery that a medication that lowered high blood lipid levels did not necessarily increase lifespan [Report of the Committee of Principal Investigators, 1984] or that a treatment that increased bone density did not necessarily make bones more resistant to fracture [Haguenauer et al., 2000] led to a re-assessment on how treatment efficacy should be measured in clinical trials. Selection of the measure of treatment efficacy has surfaced as a controversial and consequential decision in clinical trial design and interpretation [Nowak, 1994]. The aim of this essay is to discuss the issues surrounding endpoint selection in caries trials aimed at obtaining drug approval.
HUJOEL
ET AL.
Defining True and Surrogate Endpoints In a clinical trial, an endpoint can be any measurement that is plausibly related to a disease process and that is used to assess treatment efficacy. For caries research, endpoints may range from microbiological diagnoses of infection through subtle sub-clinical markers of enamel demineralization to edentulism. Two types of endpoints have been defined: true endpoints and surrogate endpoints. The Food and Drug Administration has defined true endpoints for a regulatory setting as outcomes that directly measure how a patient feels, functions, or survives [Temple, 1995]. Statistical works define true endpoints as those outcome measures which are tangible to the patient, with the word tangible indicating that the measure of treatment efficacy is capable of being precisely identified or realized by the (patients) mind [Fleming and DeMets, 1996]. How does caries affect the way a patient feels, functions, or survives? Tooth loss, cold and sweet sensitivity, pain or discomfort and, one could suspect that even frank, clinically relevant cavitations are examples of events that a patient feels can precisely be identified or realized by the patients mind, i.e., that are tangible. True endpoints are sometimes referred to as clinically relevant endpoints, clinically meaningful endpoints, terminal endpoints, or ultimate endpoints. A surrogate endpoint is a laboratory measurement or physical sign used in therapeutic trials as a substitute for a clinically meaningful endpoint [Temple, 1995]. The International Conference on Harmonization defined a validated surrogate endpoint as an endpoint that allows prediction of a clinically important outcome but, in itself, does not measure a clinical benefit [Chakravarty, 2005]. Changes in tumor size, bone density or blood lipid levels are typical examples of surrogate endpoints, because the changes cannot be identified or realized by the mind; hence, they are intangible, they are surrogates. Surrogate endpoints in caries research are those aspects of the caries process that the patient does not feel or which do not affect his function and may include white spots, sticky fissures, and non-cavitated caries lesions diagnosed by diagnostic equipment such as radiographs. Surrogate endpoints are often objective, because they can be measured by the clinician (rather than relying on self-report by patients) or by laboratory methods. Surrogate endpoints are sometimes referred to as intermediate endpoints, biological markers, or biomarkers.
52
IN
AND
CONCERNS
Both at the caries conference in Loch Lomond [Pitts and Stamm, 2004] and at the Indiana Conference in 2005, it was suggested that enamel demineralizations should be considered true endpoints. Since these claims were unreferenced, it is unclear what definition of a surrogate disease marker, if any, was used to support this claim. Such a definition would be at odds with how the term surrogate and true endpoint has been defined in statistical or regulatory settings (e.g., for overview see [Burzykowski et al., 2005]). If a change in tumor size expressed in centimeters is a surrogate, if a change in bone mineral density expressed in grams is a surrogate, then a change in enamel density expressed in micrometers or nanograms is similarly a surrogate. As long as enamel demineralizations do not affect the way a patient feels or functions, as long as enamel demineralizations are intangible to the patient, such endpoints will be considered examples of surrogate endpoints. If standard statistical terminology will be uniquely redefined for caries, it will obscure the scientific issues that are involved in selecting surrogate endpoints, deprive caries research from the methodology available to determine reliability of surrogates, and lead to logical inconsistencies when clinical trials are conducted with the goal of validating non-cavitated lesions as an endpoint for regulatory drug approval [Chesters et al., 2004]. Downstream and Upstream Surrogate Endpoints Surrogate endpoints are often further characterized with respect to their temporal location in the natural disease history. Downstream endpoints are those endpoints that occur late in the natural disease history and are, therefore, close to the clinically relevant endpoints. Examples of downstream caries endpoints may include cavitations into > 1/2 of the dentin. Upstream endpoints are those endpoints that occur early in the natural disease history. Subsurface enamel demineralization only diagnosable by more novel diagnostic methods are examples of upstream or events. In general, the farther removed the endpoint is from clinically relevant endpoints (i.e., the more upstream), the more unreliable the surrogate becomes [Chakravarty, 2005]. Why is the Issue of Surrogate Endpoints Important? The use of surrogate endpoints allows for conducting clinical trials of shorter duration and/or smaller sample size. If the goal of a clinical trial is to reduce the study cost or to shorten the time to drug approval, surrogate end-
53
HUJOEL
ET AL.
points offer one way to achieve this goal. This increased efficiency in terms of reduced time to drug approval, in terms of reduced patient sample size and reduced study cost comes at a pricethe possibility that the intangible benefits do not translate into tangible benefits, or worse, that the intangible benefits are associated with real tangible harms. Because surrogates can be misleading, their use in clinical trials is controversial, with ethical issues to be considered both for subjects participating in the study and those populations that will benefit or be harmed by the treatments identified based on surrogate endpoints. Surrogate endpoints can be misleading, because the causal pathway connecting the treatment to the surrogate to the true endpoints is often misunderstood (Table 1). Knowing that a treatment has an impact on a surrogate endpoint does not mean that the treatment provides true endpoint benefits. Iressa drastically shrank lung tumors, but through some unknown mechanisms, it did not improve survival [Couzin, 2002]. Cardiac anti-arrhythmic drugs normalized heartbeats, but through some unknown mechanism, it increased mortality risk (The Cardiac Arrhytmia suppression trial [CAST] investigators, 1989). Chlorhexidine varnish wiped out mutans streptococci [Sandham et al., 1991], yet failed to have an impact on frank cavitations [Forgie et al., 2000]. Fluorides succeeded at preventing enamel demineralization, yet failed to prevent frank cavitations [Zimmer et al., 2001]. These are examples of surrogate endpoints leading to false-positive conclusions; surrogates suggested that treatment was effective (positive result), while the true endpoint indicated that treatment was ineffective or actually harmful. False-negative conclusions also occur. For instance, thalidomide healed AIDS associated aphthous ulcers (a positive outcome) but did not decrease TNF-a production, as the effective treatment was expected to do [Jacobson et al., 1997]. Finally, surrogate endpoints can lead to the correct conclusions. HAART treatment was approved based on surrogate endpoints and, subsequently, it was shown that HAART treatment also improves survival [Mocroft et al., 2003]. Surrogates can be misleading in caries research, because the assumed biological pathways between treatments, the surrogate (enamel demineralization), and the true endpoint (pain) are more complex than anticipated. Enamel demineralization may be a poor surrogate, because treatment effects on subtle enamel mineralization may not necessarily prevent cavitations [Zimmer et al., 2001], because a majority of the frank cavitations may occur without an observable
IN
AND
CONCERNS
intervening enamel demineralization [Katz et al., 2005], because an assumed good remineralization could increase the chance for a hidden caries lesions [Koulourides and Cameron, 1980] or because the demineralizations that can be successfully treated will not progress, and the demineralizations that need to be treated cannot be successfully arrested. Unless the validity of surrogate endpoints is evaluated in clinical trials, leaps of faith are required that the frequency by which such events occur are inconsequential to the results of the trial. More importantly, it has to be assumed that surrogate endpoints are not misleading in more insidious and unanticipated ways than presented here. Because surrogates always have the potential to be misleading, their use in clinical trials remains controversial. Some have suggested that the acceptance of drugs based on surrogate endpoints may potentially reflect a fundamental flaw in the drug approval process and that, possibly, a major drug tragedy will be required to lead to revisions in the drug-approval process in the United States [Psaty et al., 1999]. Others have opined that the evaluation of tomorrows drugs will be based primarily on biomarkers, rather than on the longer-term, harder clinical endpoints [Molenberghs et al., 2005] Should Caries-effectiveness be Evaluated Based on Non-cavitated (surrogate) Endpoints? In light of such controversies, regulatory agencies use a set of criteria to decide whether to accept drugs based on surrogate endpoints. Criteria commonly used in decision making include biological plausibility and epidemiological evidence in support of the surrogate endpoint, the extent to which the surrogate endpoints have been validated in clinical trials, and the Zrisk-benefit considerations related to drug approval from a public health perspective. Some of these criteria are now being reviewed [Chakravarty, 2005]. 1) Support for surrogate endpoints based on biological plausibility: Consistent Epidemiology: The lack of consistent progression from noncavitated lesions to pain argues against the use of non-cavitated lesions. Studies have indicated that subtle enamel lesions, white spots diagnosed clinically, frank cavitations, caries lesions into the dentin arrest or reverse course frequently [Backer Dirks, 1966; Makinen et al., 1995; Levine et al., 2003]. Caries may be more like prostate cancer (where the presence of a histologically identified cancer does not imply that the patient has a high probability of dying from
56
Disease/Conditions Experimental Treatment Control Treatment Effect on Surrogate Endpoint A significant 16% increased lumbar spine bone mineral density Dramatic tumor shrinkage in 10% of patients While thalidomide was expected to decrease TNF-a production, a significant 4.4 pg/ml increase TNF-a production occurred, suggesting harm Substantial elimination of tumor mass Significant reduction of salivary mutans streptococci by an average of 3 logs (99.9%) Reduction of D1, D2 lesions compared to controls Pain diminished and ability to eat improved No effect Non-vertebral fracture rates increased by 85% False-positive Effect on True Endpoint Misleading Conclusion Osteoporosis Fluoride Placebo Lung cancer Placebo ZD1839 (Iressa) False-positive Aphthous ulcers Thalidomide Placebo False-negative Prostate cancer Radical prostatectomy Watchful waiting No effect on overall mortality risk More D3 caries lesions in CHX group than in the placebo group False-positive Caries 40% chlorhexidine varnish Placebo False positive Caries Professional toothcleaning +0.1% F varnish +POH POH No difference in D3, D4 lesions compared to controls False-positive
HUJOEL
References
[Haguenauer et al., 2000]
ET AL.
[Couzin, 2002]
[Jacobson et al., 1997]
[Holmberg et al., 2002] [Forgie et al., 2000] [Sandham et al., 1991]
[Zimmer et al., 2001]
INDIANA CONFERENCE 2005 TABLE 1.
Examples of potentially misleading surrogates. For some examples, the experimental treatment led to improvements in surrogate endpoints, while the true endpoint was either unaffected or worsened (a false-positive conclusion). For other examples, the experimental treatment had no impact on the surrogate endpoint, while the true endpoint improved (a false-negative conclusion).
Endpoint True endpoints
IN

Examples
AND
CONCERNS
Sensitivity, pain, discomfort, tooth loss, edentulism, esthetics (visible discolorations and cavitations), decreased function. Lesions diagnosed with FOTI, DIANOdent, or QLF, white spots, sticky fissures, tooth migration, radiographically defined lesions or any measure of caries process that that patient does not feel or perceive.
Surrogate endpoints
TABLE 2.
Examples of true and surrogate endpoints in the caries process.
the cancer) than HIV (where the presence of an HIV infection in the absence of treatment has a high case-fatality rate). Quantitative Epidemiology: For caries, there is some evidence that both active and inactive non-cavitated lesions are a strong predictor for cavitation [Nyvad et al., 2003]. There is no direct evidence that lesions identified by DIAGNOdent or QLF are predictors for frank cavitation, but some indirect evidence suggests that this may be the case. The Food and Drug Administration stated that there must be a demonstrable linkage between the surrogate and the ultimate endpoint. They must be strong leadsmore than just a notion [Henney, 1999]. A systematic review of available evidence could determine whether current leads, indeed, do represent strong leads. Surrogate Relatively Late (downstream) in the Biological Path: The larger the cavity, the more likely it is associated with pain and discomfort, and the more likely the surrogate is valid. In general, the closer the surrogate endpoint is to the true endpoint, the more likely the surrogate is valid. Prostate cancer with metastases (downstream) is more informative on mortality than a cancer limited to the prostrate gland (upstream). The current trend to go further away from frank cavitation (to move upstream from the clinically relevant events) increases the likelihood that the selected surrogates are invalid and argues against the use of non-cavitated lesions. 2) Support for surrogates in clinical trials: The correspondence between fluoride treatment results on non-cavitated and cavitated lesions has been evaluated in 2 trials, 1 which failed to identify a correspondence of conclusions [Zimmer et al., 2001] and another trial that identified a correspondence [Chesters et al., 2002]. No caries trials have been conCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 57
HUJOEL
ET AL.
ducted where the correspondence of conclusions on non-cavitated and cavitated lesions was evaluated for products other than fluorides. None of the available methods to formally assess the validity of surrogate endpoints both at the individual trial level and within a meta-analytic framework have been utilized [Buyse and Molenberghs, 1998; Buyse et al., 2000a; Buyse et al., 2000b; Freedman et al., 1992; Prentice, 1989]. Most importantly, it has not been determined what percent of the effect of treatment on frank cavitations is mediated through observable, non-cavitated lesions. Estimating the proportion of treatment effect explained by non-cavitated lesions is key to understanding the utility of non-cavitated lesions in RCTs, and no work in this area has yet been reported. 3) Risk-benefit considerations in the use of surrogates for caries clinical trials: Caries as a Life-threatening Disease: The more a disease is life threatening, the more likely regulatory organizations may accept surrogate endpoints. Caries, unlike AIDS, cancer, or cardiovascular disease, cannot be considered a life threatening disease and, therefore, minimal pressures exist to use non-cavitated lesions. Caries and Effective Treatment: For those diseases where no effective treatment exists, there is pressure to provide access to drugs quickly (i.e., to approve drugs based on surrogates). Such an argument in favor of surrogates is difficult to make for caries given the evidence on topical fluoride [Marinho et al., 2004]. Large Databases on Safety: Dental diseases are common and treatments, such as fluoride, are among the most widely used medications in the history of medicine. Experiences, such as the increased risks for aggressive leukoplakia with sanguinaria tooth paste [Mascarenhas et al., 2001], or an increased mortality risk with beta-carotene [Omenn, 2004], suggests the need for having the highest quality evidence of safetynamely randomized trials. The global usage of caries preventive treatments puts intense pressure on the accurate characterization of safety, which can most reliably be achieved by true endpoint pivotal RCTs. Caries and Short-term Use of Drug: Because caries-preventive agents are used long-term, often, life-long surrogates are contra-indicated, because short58 INDIANA CONFERENCE 2005
IN
AND
CONCERNS
term surrogate-based trials do not allow for evaluating what type of health outcomes are associated with long-term usage. Caries and Difficulty in Studying Clinically Relevant Endpoints: While the incidence of caries has declined dramatically, caries remains one of the most common diseases in mankind, and clinically relevant endpoints such as pain or discomfort remain frequent occurrences. Caries is not like HIV or prostate cancer, where the true endpoints, such as mortality, represent a much higher challenge in the conduct of RCTs. Drugs used on healthy people and their anticipated benefits may take a long time to occur: Caries-preventive agents are used long-term, often life-long, and a significant percentage of healthy people in whom the drugs are used may not be at risk for caries. Such situations once again favor the use of true endpoint pivotal RCTs.
Conclusions To our knowledge, no caries trials have been conducted that have evaluated whether caries preventive agents provide tangible patient benefits. Whether such trials will ever be conducted is unclear and will most likely be determined by experiences and legislation driven by events in other areas of chronic disease. Just like the thalidomide crisis caused the introduction of randomized controlled trials for all diseases and drugs, including caries, so may a disaster with a widely used drug unrelated to dental disease lead to novel legislation on the use of surrogates in clinical trials that will affect dental trials. During the last decades, the Radike criteria were widely accepted as a good surrogate endpoint. Frank cavitation was an irreversible stage in the caries process; it was considered far enough downstream to be closely related to clinically relevant outcomes such as sensitivity and pain, and it was sufficiently easy to diagnose (i.e., to obtain acceptable inter- and intra-examiner agreement). Currently, there are financial pressures to included non-cavitated caries lesions in caries scoring. The hope exists that these subtle lesions will provide reliable data on the efficacy of the products, but insufficient data are available to indicate that this will indeed be the case. Both experiences in other chronic disease areas, available data on caries dynamics, and safety considerations suggest one
59
HUJOEL
ET AL.
should be careful with accepting that treatment effects on subtle enamel demineralizations will be safe and that they translate into prevention of frank cavitations. The current Radike criteria can be changed in 1 of 2 fundamentally opposing directions. The Radike criteria could be changed toward more clinically relevant endpoints, franker cavitation. Redefining caries to a more clinically relevant sized lesion would improve intra- and extra-examiner agreement, increase the likelihood that the surrogate is valid, and lead to an almost universal agreement that the endpoint studied is clinically relevant. Diagnosing a cavity in a tooth for a pivotal trial would have the same simplicity as diagnosing a bone fracture, a mortality event, or any other true endpoint typically used in pivotal trials. Simple and widely used time-to-event analyses techniques could then be employed. Alternately, the Radike criteria could be moved farther away from clinically relevant endpoints, and subtle non-cavitated caries lesions could be accepted into the diagnostic criteria for caries. Such a change could increase the difficulty in obtaining intra- and inter-examiner agreement and raise questions and uncertainty regarding clinical relevance, resulting in ordinal statistical analyses which may not be easily interpretable. Only a coterie of highly trained examiners with repeated training sessions during the trial would be able to diagnose caries on a multi-level scale with some reliability. The scientific issues associated with selecting subtle, non-cavitated lesions in pivotal trials are diverse. Is remineralization necessary for those subtle caries lesions that can be reversed, and is remineralization possible for those subtle caries lesions that will progress to clinically relevant outcomes? Surrogate-based trials do not allow for finding an answer to such questions. The natural disease process of caries is highly variable, and quantitative epidemiological information on the relationship between surrogate and true caries endpoints is minimal. No formal assessment of the validity of surrogate endpoints in caries clinical trials has been reported. The risk-benefit criteria that are commonly used by regulatory organizations suggests that caries may not be an appropriate disease for using surrogates. Currently, there is insufficient evidence to conclude that treatment effects observed on non-cavitated lesions provide clinically relevant information. Whether non-cavitated lesions can be used in pivotal trials may remain controversial and may be influenced by the type of claims desired, safety concerns,
IN
AND
CONCERNS
and experiences in other chronic disease areas. Large-scale definitive studies using frank cavitations offer the most reliable evidence on both the safety and efficacy of caries preventive treatments.
Note: This work was in part supported by NIDCR R-01 DE13912.

References
Backer Dirks O: Posteruptive changes in dental enamel. J Dent R 1966;45:503-511. Burzykowski T, Molenberghs G, Buyse M: The evaluation of surrogate endpoints. Springer, 2005. Buyse M, Molenberghs G: Criteria for the validation of surrogate endpoints in randomized experiments. Biometrics 1998;54:1014-1029. Buyse M, Molenberghs G, Burzykowski T, Renard D, Geys H: The validation of surrogate endpoints in metaanalyses of randomized experiments. Biostatistics 2000a;1:49-67. Buyse M, Thirion P, Carlson RW, Burzykowski T, Molenberghs G, Piedbois P: Relation between tumour response to first-line chemotherapy and survival in advanced colorectal cancer: a meta-analysis. MetaAnalysis Group in Cancer. Lancet 2000b;356:373-378. Carmona R: Surgeon general statement on community water fluoridation; Office of the Surgeon General, 2004, p. 1. Chakravarty A: Regulatory aspects in using surrogate markers in clinical trials; in M. Buyse, (ed), The evaluation of surrogate endpoints; Springer 2005, pp 13-51. Chesters RK, Ellwood RP, Biesbrock AR, Smith SR: Potential modern alternative designs for caries clinical trials (CCTs) and how these can be validated against the conventional model. J Dent Res 2004;83 Spec. No. C:C122-124. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson JR, Nicholson JA, Gendvilyte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002;81:637-640. Couzin J: Cancer drugs. Smart weapons prove tough to design. Science 2002;298:522-525. Fleming TR, DeMets DL: Surrogate end points in clinical trials: are we being misled? Ann Intern Med 1996;125:605-613. Forgie AH, Paterson M, Pine CM, Pitts NB, Nugent ZJ: A randomised controlled trial of the caries-preventive efficacy of a chlorhexidine-containing varnish in high-caries-risk adolescents. Caries Res 2000;34:432439. Freedman LS, Graubard BI, Schatzkin A: Statistical validation of intermediate endpoints for chronic diseases. Stat Med 1992;11:167-178.
61
HUJOEL
ET AL.
Haguenauer D, Welch V, Shea B, Tugwell P, Wells G: Fluoride for treating postmenopausal osteoporosis. Cochrane Database Syst Rev 2000:CD002825. Henney JE: Remarks by Commissioner of Food and Drugs International Conference on Surrogate Endpoints and Biomarkers, 1999. Holmberg L, Bill-Axelson A, Helgesen F, Salo JO, Folmerz P, Haggman M, Andersson SO, Spangberg A, Busch C, Nordling S, Palmgren J, Adami HO, Johansson JE, Norlen BJ: A randomized trial comparing radical prostatectomy with watchful waiting in early prostate cancer. N Engl J Med 2002;347:781-789. Hujoel PP, Lydon-Rochelle M, Bollen AM, Woods JS, Geurtsen W, del Aguila MA: Mercury exposure from dental filling placement during pregnancy and low birth weight risk. Am J Epidemiol 2005;161:734-740. Jacobson JM, Greenspan JS, Spritzler J, Ketter N, Fahey JL, Jackson JB, Fox L, Chernoff M, Wu AW, MacPhail LA, Vasquez GJ, Wohl DA: Thalidomide for the treatment of oral aphthous ulcers in patients with human immunodeficiency virus infection. National Institute of Allergy and Infectious Diseases AIDS Clinical Trials Group. N Engl J Med 1997;336:1487-1493. Katz BP, Ofner S, Eckert GJ, Ferreira-Zandon AG, Ando M, Eggertson H, Mau MS, Kelly SA, Doi T, Lukantsova L, Wefel JS, Stookey GK: De- and Re-Mineralization. Caries Res 2005;39:319. Kingman A: Acceptance criteria for Clinical Caries Models, 2005. Koulourides T, Cameron B: Enamel remineralization as a factor in the pathogenesis of dental caries. J Oral Pathol 1980;9:255-269. Levine RS, Nugent ZJ, Pitts NB: Pain prediction for preventive non-operative management of dentinal caries in primary teeth in general dental practice. Br Dent J 2003;195:202-206; discussion 197. Makinen KK, Makinen PL, Pape HR, Jr., Allen P, Bennett CA, Isokangas PJ, Isotupa KP: Stabilisation of rampant caries: polyol gums and arrest of dentine caries in two long-term cohort studies in young subjects. Int Dent J 1995;45:93-107. Marinho VC, Higgins JP, Sheiham A, Logan S: Combinations of topical fluoride (toothpastes, mouthrinses, gels, varnishes) versus single topical fluoride for preventing dental caries in children and adolescents. Cochrane Database Syst Rev 2004:CD002781. Mascarenhas AK, Allen CM, Loudon J: The association between Viadent use and oral leukoplakia. Epidemiology 2001;12:741-743. Mocroft A, Ledergerber B, Katlama C, Kirk O, Reiss P, dArminio Monforte A, Knysz B, Dietrich M, Phillips AN, Lundgren JD: Decline in the AIDS and death rates in the EuroSIDA study: an observational study. Lancet 2003;362:22-29. Molenberghs G, Buyse M, Burzykowski T: Introduction; in T. Burzykowski, G. Molenberghs, M. Buyse, (eds), The evaluation of surrogate endpoints; Springer, 2005, pp 13-51. Nowak R: Problems in clinical trials go far beyond misconduct. Science 1994;264:1538-1541. Nyvad B, Machiulskiene V, Baelum V: Construct and predictive validity of clinical caries diagnostic criteria assessing lesion activity. J Dent Res 2003;82:117-122. Omenn GS: Human lung cancer chemoprevention strategies: Parker B. Francis lecture. Chest 2004;125:123S127S.
62
IN
AND
CONCERNS
Petersen PE: The World Oral Health Report 2003: continuous improvement of oral health in the 21st centurythe approach of the WHO Global Oral Health Programme. Community Dent Oral Epidemiol 2003;31 Suppl. 1:3-23. Pitts NB, Stamm JW: International Consensus Workshop on Caries Clinical Trials (ICW-CCT)final consensus statements: agreeing where the evidence leads. J Dent Res 2004;83 Spec. No. C:C125-128. Prentice RL: Surrogate endpoints in clinical trials: definition and operational criteria. Stat Med 1989;8:431440. Psaty BM, Weiss NS, Furberg CD, Koepsell TD, Siscovick DS, Rosendaal FR, Smith NL, Heckbert SR, Kaplan RC, Lin D, Fleming TR, Wagner EH: Surrogate end points, health outcomes, and the drug-approval process for the treatment of risk factors for cardiovascular disease. JAMA 1999;282:786-790. Report of the Committee of Principal Investigators: WHO cooperative trial on primary prevention of ischaemic heart disease with clofibrate to lower serum cholesterol: final mortality follow-up. The Lancet 1984;2:600-604. Sandham HJ, Brown J, Chan KH, Phillips HI, Burgess RC, Stokl AJ: Clinical trial in adults of an antimicrobial varnish for reducing mutans streptococci. J Dent Res 1991;70:1401-1408. Stookey GK, Gonzlez-Cabezas C: Emerging methods of caries diagnosis. J Dent Educ 2001;65:1001-1006. Temple RJ: A regulatory authoritys opinion about surrogate endpoints; in W.S. Nimmo, G.T. Tucker, (eds), Clinical Measurement in Drug Evaluation; J Wiley, 1995. The Cardiac Arrhytmia suppression trial (CAST) investigators: Preliminary report: effect of encainide and flecainide on mortality in randomized trial of arrhytmia suppression after myocardial infarction. New England Journal of Medicine 1989;321:406-412. Zimmer S, Bizhang M, Seemann R, Witzke S, Roulet JF: The effect of a preventive program, including the application of low-concentration fluoride varnish, on caries control in high-risk children. Clin Oral Investig 2001;5:40-44.
63
64
Distinguishing Between Individuals with High or Low Risk of Developing Dental Caries
Hannau Hausen Institute of Dentistry, University of Oulu, Oulu, Finland
Abstract The burden of caries is unevenly distributed among the contemporary, low-caries populations. It would be highly desirable to distinguish in advance the individuals who have a high and low risk of developing cavities. For instance, subjects who are recruited for proof-of-principle caries trials should have the potential for developing enough lesions to allow the effect of an intervention to show up at a feasible cost. Currently, it is not possible to distinguish very accurately between high and low-risk individuals. Among different predictors, the ones that consider the clinical signs of the caries experience have proved most powerful, but even their performance is modest. The power of other factors varies from setting to setting, but often their contribution to the accuracy of predictions is insignificant. Consequently, individual risk assessments are likely to bring about many errors, false negatives, and false positives. However, it is possible to form groups with a high and low average caries risk by using simple measures. It has proven difficult, however, to achieve a significant effect in trials among high-risk groups that have been selected by screening populations with a low average risk, especially if the subjects have been schoolchildren. If the aim is to give a caries-control measure the best possible chance to show its effect, trials in high-risk areas seem to be preferable to trials among high-risk individuals from low-risk areas. However, caries trials are still needed, even in the low-risk areas, to find out whether interventions that have proven efficacious in high-risk settings are effective even among populations with a low average caries risk.
65
HAUSEN The need for distinguishing between individuals with a high and low risk of developing caries lesions stems from the fact that the burden of caries is unevenly distributed among the contemporary low-caries populations. In 19881991, one quarter of the U.S. 5 to 17-year-olds had at least 1 permanent tooth accounting for about 80% of the caries experienced in permanent teeth [Kaste et al., 1996]. Correspondingly, the worse quarter of 12-year-old children in 2 Finnish cities had 70-80% of all DMF surfaces in 1998 [Seppa et al., 2000]. Different parties among the dental community have different motives for trying to distinguish between individuals with a high and low risk of caries. The clinicians would like to offer the high-risk susceptible individuals protection against the development of cavities. They should also be able to identify the low-risk individuals in order to avoid providing them with prevention that has little or no purpose. The researchers interested in evaluating the efficacy of caries-control measures, in turn, want to make sure that candidates for their trials have the potential for developing enough lesions to allow the effect of an intervention to show up at a feasible cost. The baseline level of risk can also be treated as a potential effect-modifying or confounding factor when participants with multiple levels of risk are recruited for a trial. Distinguishing Between High- and Low-risk IndividualsState of the Art Evaluating the caries predictive power of different risk indicators has been a very popular study subject. Even the number of related reviews and compilations is extensive [Vanderas, 1986; Anonymous, 1988; Beck et al., 1988; Krasse, 1988; Winter, 1988; Bader, 1990; Demers et al., 1990; Eriksen and Bjertness, 1991; Johnson, 1991; Stamm et al., 1991; Banting, 1993; van Houte, 1993; Anonymous, 1995; Moss and Zero, 1995; Tinanoff, 1995; Kidd, 1998; Pitts, 1998; Powell, 1998a,b; Reich et al., 1999; Messer, 2000; Hausen, 2003; Giannoni et al., 2005]. Despite the vast research efforts, our current ability to distinguish between high- and low-risk individuals is modest. Individual risk assessments are likely to bring about many errors, false negatives, and false positives. In a conference on risk assessment in dentistry held at The University of North Carolina Chapel Hill in 1989, a suggestion was made that, for a risk model, the sum of sensitivity and specificity should be at least 160% before a
66
CARIES RISK
caries risk marker can be considered a legitimate candidate for targeting individualized prevention [Kingman, 1990]. This was equal to the proposal by Wilson and Ashley [1989], whereby a sensitivity and specificity of 80% would be acceptable for practical use in the community. These arbitrary suggestions indicate implicitly our modest ability to distinguish individually between a high and low caries risk. If accurate measures for risk assessment would be available, a false negative and false positive rate of 20% would undoubtedly not be considered acceptable. Among the vast literature on caries risk assessment, a combined sensitivity and specificity of 160% has only been achieved occasionally. An example is the study by Alaluusua and Malmivirta [1994], in which an attempt was made to identify in advance those preschool children who would have at least 1 dentinal caries lesion at the age of 3 years. At baseline, the children were divided into a high- and low-risk group so that those children who had visible plaque on the labial surfaces of all their upper incisors were considered as having a high risk. The remaining children were believed to have a low risk (table 1). Among the 12 children who had caries upon completion of the follow-up, 10 had had visible plaque at baseline. The sensitivity was 83%, specificity 92%, false positive rate 8%, and false negative rate 17%. This result exemplifies the fact that caries risk can be assessed more accurately among preschool-age children than among other age groups. In small children, even a mutans streptococci test can be passably predictive [Thibodeau et al., 1999; Pienihakkinen et al., 2004]. Among schoolchildren and adults, however, an ms-test is not very helpful for distinguishing between a high and low risk [Sderholm and Birkhed, 1988; Wilson and Ashley, 1989; Alaluusua et al., 1990; Scheinin et al., 1994]. This is true for other microbiological tests as well [for a review, see e.g., Hausen, 2003].
Visible Plaque at 19 mo
At Least 1 dmf Surface at 36 mo Yes No 6 73 79
Total
Yes No Total
10 2 12
16 75 91
TABLE 1.
Distribution of the subjects in the study by Alaluusua and Malmivirta [1994] according to the presence of dental plaque at baseline and caries at the end of the follow-up. 67
HAUSEN
FIG. 1.
ROC curves for predictors that were used for identifying the high-risk individuals among a group of initially 13-year-old Finnish children [Hausen, 1997]. Two-year approximal DMFS increment (0-1 vs. 2+) was used for validation.
Fig. 1 shows ROC curves from a study in which an attempt was made to identify the high-risk individuals among 13 year-old children [Hausen, 1997]. Baseline DMFS, salivary ms-level, salivary flow rate, sucrose intake frequency score, and social group were used as predictors and 2-year approximal DMFS increment for validation. Only past caries experience had any clinically significant predictive power. Even DMFS, however, failed to achieve the simultaneous sensitivity and specificity of 80%, which has been marked with a circle. The clearly stronger predictor of past caries experience as compared to the other predictors is in accordance with the literature [Disney et al., 1992]. Among different predictors of future caries, the ones that consider the clinical signs of caries experience have proved most powerful. The predictive power of other factors varies from setting to setting [for reviews, see e.g., Zero et al., 2001; Hausen, 2003], but generally, their contribution to the accuracy of predictions is modest compared to that of past caries experience. Therefore, their potential will not be discussed any further in the current context.
68
CARIES RISK
As for past caries experience, the status of the most recently erupted or exposed tooth surface is probably the most predictive [Powell, 1998; van Palenstein Helderman et al., 2001], considering initial caries lesions, in addition to other signs of caries experience, increase the accuracy of caries risk assessment [van Palenstein Helderman et al., 1989]. The improvement may be marginal, however, if all initial lesions are counted without distinguishing between the active and arrested ones [Seppa and Hausen, 1988]. Subjects with active initial caries lesions might be the most promising candidates for future caries trials, since they have the potential for developing more advanced lesions during a reasonable time-span. In conclusion, our current risk assessments are so inaccurate that, in clinical practice, the targeting of caries-control efforts cannot be mechanically based on the use of any single predictor or risk function carrying the information of multiple predictors. Many risk indicators have some predictive power, however. If they are used for screening a target population, the resulting group of test positives will subsequently have a higher average caries experience than the test negatives. Table 2 shows a cross-classification of a group of Finnish children according to baseline DMFS (1/0) and 3-year DMFS increment (1/0). Baseline DMFS could distinguish between individuals with at least 1 vs. no new DMFS within 3 years, with a sensitivity of 82% and a specificity of 60%. The false positive rate was 40% and the false negative rate 18%, implying that the accuracy would not have been acceptable for clinical decision-making. In both baseline DMFS groups, there were individuals with a high increment and individuals with no increment. The average increment, however, was more than 3 times higher in the group that had at least 1 DMF surface at baseline compared to the group with no baseline DMFS (fig. 2). It is possible to form
Baseline DMFS 1 =0 Total
3-year DMFS Increment 1 230 50 282 0 41 61 102
Total
271 113 384
TABLE 2.
Cross-classification of a group1 of initially 12-year-old Finnish children according to baseline DMFS and subsequent 3-year DMFS increment.
69
HAUSEN
FIG. 2.
Distribution of the children in table 2 according to 3-year DMFS increment.
70
CARIES RISK
groups of people with a high and a low average caries risk by using simple screening measures. Baseline Risk in Caries Trials If the intention is to give a potential caries-control measure the best chance to show its effect, as is the case in proof-of-principle trials, the strategy for conducting a trial among high-risk individuals compared to low-caries populations seems to be somewhat problematic, at least with regard to schoolchildren. There are several examples of trials covering different approaches to caries control, wherein this course has been applied, that have failed to observe both a clinically and statistically significant effect [Johnston and Lewis, 1995; Forgie et al., 2000; Zimmer et al., 2001; Kallestal, 2005]. The reasons for no effect showing up probably vary from setting to setting, but there are certain general factors that could have contributed to the failures. First, in low-risk areas, even the high-risk individuals may already have been exposed to different caries-control measures, such as those included in their background dental care, which may have diluted the effect of the study regimen. Second, it is not easy to organize proper supervision for maintaining a satisfactory compliance to the studied regimen among high-risk individuals who are scattered amidst their low-risk classmates who comprise the majority of the age group. From the behavioral point of view, it may also be unreasonable to expect the high-risk study subjects to conduct themselves differently from their classmates. Among adolescents, peer pressure against dissimilarities is high. For the high-risk adolescents in low-risk areas, caries can be a heavy burden. Consequently, there is still a need for developing more effective and feasible ways for controlling caries among them. One way is to conduct trials with the aim of finding out whether the policy of applying regimens that include caries-control measures already known to be efficacious in ideal conditions could be effective among this problematic target group. This was the motive for the studies of Seppa et al. [1981] and Hausen et al. [2000]. A low-risk comparison group was included in their design (fig. 3) in order to find out how close to the average caries increment among the low-risk majority it would be possible to get by the intensive prevention that was given to the high-risk study group. In both studies, the subjects were schoolchildren in their early teens. The effect of intensive prevention regimens, including multiple caries-control measures
HAUSEN
FIG. 3.
Outline of the design used in the studies by Seppa et al. [1981] and Hausen et al. [2000].
known to be efficacious, was evaluated. In both instances, no significant effect was observed. The main result of the latter study can be seen in fig. 4. The average 3-year DMFS increment was 4.4 in the high-risk study group that had received intensive prevention, 5.1 in the high-risk control group having received basic prevention, and 2.0 in the low-risk comparison group that had received similar basic prevention as the high-risk control group. The result supported the notion that it is difficult to control caries among high-risk adolescents in low-risk areas. New approaches need to be developed for this target group. Some preventive regimens have proven effective even among high-risk individuals in low-risk areas when the subjects have been younger or older than schoolchildren. Pienihakkinen and Jokela [2002] studied the effect of risk-based management of dental caries in comparison with routine prevention among initially 2 year-old Finnish children who were followed up for 3 years. The pro72 INDIANA CONFERENCE 2005
CARIES RISK
FIG. 4.
The main results of the study by Hausen et al. [2000]. HR and LR stand for high and low risk.
portion of children with cavitated caries or fillings was significantly lower in the risk-based group than in the routine prevention group. Interestingly, a significant difference between the groups could still be seen in a follow-up study conducted 7 years after conclusion of the original trial [Pienihakkinen et al., 2005]. The study by Rask et al. [1988] is an example of a successful effort to control caries among high-risk older adults in a low-caries country. In this case, however, the gain that had been achieved during the 1-year experimental program had almost been lost 4 years later. Today, risk-based caries trials are often conducted in high-risk areas [Zimmer et al., 1999; Kowash et al., 2000; Lo et al., 2001; Curnow et al., 2002; Davies et al., 2002; Jackson et al., 2005; Toumba and Curzon, 2005] or highrisk countries [Chesters et al., 2002; Stookey et al., 2004] rather than among high-risk individuals from low-risk areas. This strategy seems to give a cariescontrol measure a better chance to show its efficacy. Additional studies may be needed, however, to find out whether an intervention that has proven efficacious
HAUSEN in a high-risk setting is effective even among populations with a low average caries risk. A relevant aim for a caries trial is to find out whether the effect of a regimen varies according to the baseline level of risk. The randomized clinical trial by Moberg et al. [2005] evaluated the effect of a school-based fluoride varnish program on approximal caries among 13 to 16-year-olds in high, medium, and low caries risk areas on the Swedish west coast. Prevented fraction for fluoride varnish treatment twice a year at 6-month intervals was 69% in high, 66% in medium, and 20% in the low risk area. Therefore, the level of risk clearly influenced the effect, with the children in the low-risk area benefiting least from the program. In general, one would expect that subjects at a higher level of risk would benefit more from a preventive intervention, but there are examples of reverse results. Olivier et al. [1992] evaluated the efficacy of bi-annual APF gel applications without previous prophylaxis in reducing dental caries among high-risk children living in non-fluoridated communities. The subjects were initially 6 year-old children with at least 3 cavities on proximal surfaces of their primary teeth. The control group received a placebo. In the analyses, the children were stratified into those with 3-14 and 15+ def surfaces at baseline. A beneficial effect was only found among the children with the lower defs level. Conclusions If the aim is to give a caries-control measure the best possible chance to show its effect, trials in high-risk areas seem to be preferable to trials among high-risk individuals from low-risk areas. However, caries trials are still needed, even in the low-risk areas, to find out whether interventions that have proven efficacious in high-risk settings are effective even among populations with a low average caries risk. Even though low-risk individuals may only get a small number of lesions per person, together, they develop many cavities, since they comprise the majority of the populations in the contemporary, industrialized countries. Part of these lesions might be controllable by new approaches. Caries can be a heavy burden for high-risk individuals among the average low-risk populations. Currently, our ability to protect the high-risk minority against the development of cavities is modest. New ways of caries control need to be developed
74
CARIES RISK
for them. Preschool children make an important target population for caries trials, since it is possible that long-term effects can be achieved by interventions among them. Caries trials should also be conducted among older people, some of whom may develop a significant amount of cavities. Prevention regimens developed for children and adolescents are not necessarily ideal for them.
References
Alaluusua S, Kleemola-Kujala E, Gronroos L, Evalahti M: Salivary caries-related tests as predictors of future caries increment in teenagers. A three-year longitudinal study. Oral Microbiol Immunol 1990;5:77-81. Alaluusua S, Malmivirta R: Early plaque accumulationa sign for caries risk in young children. Community Dent Oral Epidemiol 1999;27:406-412. Anonymous: Caries diagnosis and risk assessment. A review of preventive strategies and management. J Am Dent Assoc 1995;126 (supplement):1S-24S. Anonymous: Review of methods of identification of high caries risk groups and individuals. Federation Dentaire Internationale Technical Report No. 31. Int Dent J 1988;38:177-189. Bader JD (ed): Risk assessment in dentistry. Chapel Hill: University of North Carolina Dental Ecology, 1990. Banting DW: Diagnosis and prediction of root caries. Adv Dent Res 1993;7:80-86. Beck JD, Kohout F, Hunt RJ: Identification of high caries risk adults: attitudes, social factors and diseases. Int Dent J 1988;38:231-238. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson JR, Nicholson JA, Gendvilyte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002:637-640. Curnow MM, Pine CM, Burnside G, Nicholson JA, Chesters RK, Huntington E: A randomised controlled trial of the efficacy of supervised toothbrushing in high-caries-risk children. Caries Res 2002;36:294-300. Davies GM, Worthington HV, Ellwood RP, Bentley EM, Blinkhorn AS, Taylor GO, Davies RM: A randomised controlled trial of the effectiveness of providing free fluoride toothpaste from the age of 12 months on reducing caries in 5-6 year old children. Community Dent Health 2002;19:131-136. Demers M, Brodeur J-M, Simard PL, Mouton C, Veilleux G, Frechette S: Caries predictors suitable for massscreenings in children: a literature review. Community Dent Health 1990;7:11-21. Disney JA, Graves RC, Stamm JW, Bohannan HM, Abernathy JR, Zack DD: The University of North Carolina Caries Risk Assessment study: further developments in caries risk prediction. Community Dent Oral Epidemiol 1992;20:64-75. Eriksen HM, Bjertness E: Concepts of health and disease and caries prediction: a literature review. Scand J Dent Res 1991;99:476-483. Forgie AH, Paterson M, Pine CM, Pitts NB, Nugent ZJ: A randomised controlled trial of the caries-preventive efficacy of a chlorhexidine-containing varnish in high-caries-risk adolescents. Caries Res 2000;34:432439.
75
HAUSEN
Giannoni M, DAmario M, Gatto R, Barone A: Some tools for the identification of high caries risk individuals. A review. Minerva Stomatol 2005;54:111-127. Hausen H: Caries prediction; in Fejerskov 0, Kidd E (eds): Dental caries. The disease and its clinical management. Oxford, Blackwell Munksgaard, 2003, pp. 327-341. Hausen H: Caries predictionstate of the art. Community Dent Oral Epidemiol 1997;25:87-96. Hausen H, Karkkainen S, Seppa L: Application of the high-risk strategy to control dental caries. Community Dent Oral Epidemiol 2000;28:26-34. Jackson RJ, Newman HN, Smart GJ, Stokes E, Hogan JI, Brown C, Seres J: The effects of a supervised toothbrushing programme on the caries increment of primary school children, initially aged 5-6 years. Caries Res. 2005;39:108-115. Johnson NW (ed): Dental caries. Markers of high and low risk groups and individuals. Cambridge, Cambridge University Press, 1991. Johnston DW, Lewis DW: Three-year randomized trial of professionally applied topical fluoride gel comparing annual and biannual applications with/without prior prophylaxis. Caries Res 1995;29:331-336. Kallestal C: The effect of five years implementation of caries-preventive methods in Swedish high-risk adolescents. Caries Res 2005;39:20-26. Kaste LM, Selwitz RH, Oldakowski RJ, Brunelle JA, Winn DM, Brown LJ: Coronal caries in the primary and permanent dentition of children and adolescents 1-17 years of age: United States, 1988-1991. J Dent Res 1996;75(special number):631-641. Kidd EA: Assessment of caries risk. Dent Update 1998;25:385-390. Kingman A: Statistical issues in risk models for caries; in: Bader JD (ed): Risk assessment in dentistry. Chapel Hill: University of North Carolina Dental Ecology, 1990:193-200. Kowash MB, Pinfield A, Smith J, Curzon ME: Effectiveness on oral health of a long-term health education programme for mothers with young children. Br Dent J 2000;188:201-205. Krasse B: Biological factors as indicators of future caries. Int Dent J 1988;38:219-225. Lo EC, Chu CH, Lin HC: A community-based caries control program for pre-school children using topical fluorides: 18-month results. J Dent Res 2001;80:2071-2074. Messer LB: Assessing caries risk in children. Aust Dent J 2000;45:10-16. Moberg Skold U, Petersson LG, Lith A, Birkhed D: Effect of school-based fluoride varnish programmes on approximal caries in adolescents from different caries risk areas. Caries Res 2005;39:273-279. Moss ME, Zero DT: An overview of caries risk assessment, and its potential utility. J Dent Educ 1995;59:932940. Olivier M, Brodeur JM, Simard PL: Efficacy of APF treatments without prior toothcleaning targeted to highrisk children. Community Dent Oral Epidemiol 1992;20:38-42. Pienihakkinen K, Jokela J: Clinical outcomes of risk-based caries prevention in preschool-aged children. Community Dent Oral Epidemiol 2002;30:143-150.
76
CARIES RISK
Pienihakkinen K, Jokela J, Alanen P: Assessment of caries risk in preschool children. Caries Res 2004;38:156-162. Pienihakkinen K, Jokela J, Alanen P: Risk-based early prevention in comparison with routine prevention of dental caries: a 7-year follow-up of a controlled clinical trial; clinical and economic aspects. BMC Oral Health. 2005;5(1):2. Pitts NB: Risk assessment and caries prediction. J Dent Educ 1998;62:762-770. Powell LV: Caries prediction: a review of the literature. Community Dent Oral Epidemiol 1998a;26:361-371. Powell LV: Caries risk assessment: relevance to the practitioner. J Am Dent Assoc 1998b;129:349-353. Rask PI, Emilson CG, Krasse B, Sundberg H: Effect of preventive measures in 50-60-year-olds with a high risk of dental caries. Scand J Dent Res 1988;96:500-504. Reich E, Lussi A, Newbrun E: Caries-risk assessment. Int Dent J 1999;49:15-26. Scheinin A, Pienihakkinen K, Tiekso J, Holmberg S, Fukuda M, Suzuki A: Multifactorial modeling for root caries prediction: 3-year follow-up results. Community Dent Oral Epidemiol 1994;22:126-129. Seppa L, Hausen H: Frequency of initial caries lesions as predictor of future caries increment in children. Scand J Dent Res 1988;96:9-13. Seppa L, Hausen H, Pollanen L, Karkkainen S, Helasharju K: Effect of intensified caries prevention on approximal caries in adolescents with high caries risk. Caries Res 1991;25:392-395. Seppa L, Karkkainen S, Hausen H: Caries trends 1992-1998 in two low-fluoride Finnish towns formerly with and without fluoridation. Caries Res 2000;34:462-468. Sderholm G, Birkhed D: Caries predicting factors in adult patients participating in a dental health program. Community Dent Oral Epidemiol 1988;16:374-377. Stamm JW, Stewart PW, Bohannan HM, Disney JA, Graves RC, Abernathy JR: Risk assessment for oral diseases. Adv Dent Res 1991;5:4-17. Stookey GK, Mau MS, Isaacs RL, Gonzlez-Gierbolini C, Bartizek RD, Biesbrock AR: The relative anticaries effectiveness of three fluoride-containing dentifrices in Puerto Rico. Caries Res 2004;38:542-550. Thibodeau EA, OSullivan DM: Salivary mutans streptococci and caries development in the primary and mixed dentitions of children. Community Dent Oral Epidemiol 1999;27:406-412. Tinanoff N: Dental caries risk assessment and prevention. Dent Clin North Am 1995;39:709-719. Toumba KJ, Curzon ME: A clinical trial of a slow-releasing fluoride device in children. Caries Res 2005;39:195-200. van Houte J: Microbiological predictors of caries risk. Adv Dent Res 1993;7:87-96. van Palenstein Helderman WH, ter Pelkwijk L, van Dijk JW: Caries in fissures of permanent first molars as a predictor for caries increment. Community Dent Oral Epidemiol 1989;17:282-284. van Palenstein Helderman WH, vant Hof MA, van Loveren C: Prognosis of caries increment with past caries experience variables. Caries Res 2001;35:186-192.
77
HAUSEN
Vanderas AP: Bacteriologic and nonbacteriologic criteria for identifying individuals at high risk of developing dental caries: a review. J Public Health Dent 1986;46:106-113. Wilson RF, Ashley FP: Identification of caries risk in schoolchildren: salivary buffering capacity and bacterial counts, sugar intake and caries experience as predictors of 2-year and 3-year caries increment. Br Dent J 1989;167:99-102. Winter GB: Prediction of high caries riskdiet, hygiene and medication. Int Dent J 1988;38:227-230. Zero D, Fontana M, Lennon AM: Clinical applications and outcomes of using indicators of risk in caries management. J Dent Educ 2001;65:1126-1132. Zimmer S, Bizhang M, Seemann R, Witzke S, Roulet JF: The effect of a preventive program, including the application of low-concentration fluoride varnish, on caries control in high-risk children. Clin Oral Investig 2001;5:40-44. Zimmer S, Robke FJ, Roulet JF: Caries prevention with fluoride varnish in a socially deprived community. Community Dent Oral Epidemiol 1999;27:103-108.
78
Acceptance Criteria for Clinical Caries Models

Albert Kingman
Biostatistics Core, Division of Clinical Research and Health Promotion, National Institute of Dental and Craniofacial Research, Bethesda, MD, USA
Background Dental caries is a disease that is characterized as a dynamic, complex interaction among susceptible tooth surfaces, microflora, and substrate that involves repeated fluctuations between demineralization and remineralization of enamel structure. In some cases, cavitation with dentinal involvement, an irreversible condition, is the consequence of progressive net demineralization of tooth structure. The D3MFS caries increment has served as the traditional caries outcome in caries randomized clinical trials (RCT). The D3 component is based on a crude visual or visual-tactile assessment of lesions that are cavitated into dentin [Radike, 1972]. Although it is possible to assess various aspects of enamel, such as white or brown spots and incipient lesions, historically, such non-cavitated lesions have been considered as sound surfaces when deriving the D3MFS value for a subject, primarily due to the lack of examiner reliability in diagnosing non-cavitated lesions. In Europe, the WHO criteria have been used [WHO, 1979], but investigators counted only lesions scored at the D3 and D4 level (with or without cavitation) as disease. The strongest evidence regarding the efficacy or effectiveness of a caries preventive product is derived from a 2- or 3-year randomized clinical (RCT) trial. Although technically, the results obtained from RCTs apply only to subjects who participate in the trial, the external validity for many large caries clinical trials has been accepted. The acceptance is based on some mild assumptions and demonstrated comparability data provided by the investigators. For years, the Food and Drug Administration (FDA) has required 2 independent pivotal studies to approve a caries preventive drug product. This meant that large, Phase 3 RCTs demonstrating the test product had a clinically meaningful and statistically significant caries preventive effect. Such trials had the fol-
79
KINGMAN
lowing characteristics: 1) randomization of participants to study groups in a parallel design, 2) 24- or 36-month D3MFS increment at the cavitation level was the primary outcome, 3) proper masking (blinding) of participants to test products, and 4) inclusion of an appropriate reference group. Many pivotal studies included the use of radiographs, and some utilized a supervised treatment study design. In 1995, the FDA issued a final monograph [Federal Register, 1995] that reclassified Over-the-Counter (OTC) Anti-caries Drug Products into groups of products that were covered by the monograph and those not covered by the monograph. The approval process was streamlined for products covered by the monograph. Fluoride-based products covered by this monograph include any product whose F concentration is within the FDA specified accepted dose range for that fluoride formulation. Monograph products are required to pass a Laboratory Testing Profile (LTP). The LTP includes: 1) animal caries reduction and one of the following tests, 2) enamel solubility reduction, or 3) fluoride enamel uptake. The tests were originally developed by industry and appear in detail in a document called Biological Testing Procedures for Fluoride Dentifrices [FDA]. The FDA continues to require 2 pivotal 2- or 3-year Phase 3 RCTs to demonstrate effectiveness for drug products not covered by the 1995 final monograph. New Scoring and Diagnostic Methods In recent years, several high-technology-based methods have been developed to assist in diagnosing aspects of the caries process at the non-cavitated state. Two methods, Quantitative Light Fluorescence (QLF) and DIAGNOdent, have been cleared by the Devices Division of the FDA as effective diagnostic devices when used as adjuncts to the clinical assessment of dental caries. Electrical conductivity measurement (ECM), fiber optic transillumination (FOTI), and digital imaging FOTI (DIFOTI) technologies are other high-technology-based methods that have been developed and tested in clinical settings. Much research has been conducted to enhance and refine the WHO clinical-visual examination [Ismail, 2004]. The Dundee Selectable Threshold Method (DSTM), International Caries Disease Assessment System (ICDAS), and the Nyvad (N) scoring systems are hybrids of the clinical-visual approach. The Dundee Selectable Threshold Method (DSTM) scores dental caries using white and brown spots and various levels of non-cavitated lesions [Fyffe et al.,
80
ACCEPTANCE CRITERIA
FOR
CLINICAL CARIES MODELS
2000]. The Nyvad index [Nyvad et al., 1999] is a 10-level diagnostic system that scores active and inactive lesions, non-cavitated and cavitated, and recurrent decay. The International Caries Diagnostic Assessment System (ICDAS) is a 2-tiered scoring system [Pitts, 2004]. One tier involves a 7-level visual assessment that includes an assessment of different levels of non-cavitated lesions and cavitation. The other tier scores caries associated with restorations and sealants. All of the clinical-visual scoring systems require complex clinical evaluation involving sustained concentration and precise calibration of personnel. The high-technology-based methods also involve complex imaging and evaluation systems. QLF [Stookey et al., 2004] and DIAGNOdent [Lussi et al., 2004] involve laser or light enhancement, computer plots, patch selection, and highly trained personnel. Ultrasonic sound, Electrical Conductivity Measures, fiber-optic transillumination (FOTI), and digital imaging fiber-optic transillumination (DIFOTI) are other approaches that have been developed. FOTI and DIFOTI have been used in several clinical trials as adjuncts to the clinical visual assessment of dental caries. Most of these newer diagnostic and scoring systems are under development or refinement and currently are not ready for general or stand-alone use. Any new scoring system or diagnostic method should satisfy the following criteria before it can be considered for routine use in a definitive trial of a new drug product: 1) it should be easy to use, 2) it should have adequate reliability for trained examiners, 3) it should have content validity, 4) it should have predictive validity, and 5) its associated primary outcome variable should represent evidence of meaningful changes in the carious process (mineral loss in enamel). PracticalityEase of Use The Radike criteria have been used for more than 3 decades in the USA. The criteria were clearly described but posed a degree of difficulty for some examiners because they involve some subjectivity. The criteria required evidence of enamel opacity, softness of the surface, and resistance to the probe when exploring pit and fissure surfaces. Over the years, examiner reliability studies were conducted in which examiners recorded non-cavitated lesions, including white and brown spots. It was determined that it was difficult for multiple examiners to agree on caries diagnoses at this level.
81
KINGMAN
European caries researchers generally used the clinical-visual WHO criteria. These criteria involved 4 levels, labeled D1, D2, D3, and D4 lesions and are described in the WHO document [WHO, 1979]. The D3 level lesions are comparable to the Radike criteria. In the past few years, use of the tactile probe has been eliminated by European caries researchers. Several refinements of the WHO scoring system have been developed and used. A few expert examiners were able to successfully use this index, but, in general, the experience with this index paralleled that of the Radike index. As a result, D3 level lesions are used to define the presence of dental caries in clinical trials incorporating the WHO criterion. The DSTM, Nyvad, and ICDAS clinical-visual methods include an assessment of caries activity, scoring surfaces wet and dry, recording the presence of plaque, and cavitated and non-cavitated status. The new visual scoring systems require intensive training, consistent attention to detail, and endurance by the examiners in a clinical trial. Although these signs and conditions may be theoretically important and provide useful information to potentially increase efficiency in clinical trials, the burden of proof lies with the investigators to demonstrate that such scoring systems can be used in a practical setting. This minimally requires a demonstration of the reliability of the examiners in use of the indices. Reliability Concerns The first test of practicality of a scoring system is that of examiner reliability. Can a trained examiner reproduce the scores for the new system (intraexaminer reliability), and can multiple examiners agree on diagnoses using the system (inter-examiner reliability)? The DSTM, Nyvad, and ICDAS methods have undergone varying degrees of reliability testing. Generally, investigators report very good or excellent values for kappa (weighted or unweighted) for the method they have developed, but data for inter-examiner agreement or reliability are unavailable or unreported. Often, reported kappa values are difficult to interpret because: 1) the prevalence distributions are not included and 2) the analytical techniques used are not adequately described. Given the difficulty experienced investigators have had in scoring caries at the D1 or D2 levels (WHO criteria), a clear demonstration of the examiners ability to reliably score disease with these more complicated diagnostic meth-
82
ACCEPTANCE CRITERIA
FOR
ods is needed. Nyvad [Nyvad et al., 1999] reported kappa values for intra-and inter-examiner reliability for a 10-level scoring system using data from a 3-year trial of a 1500 ppm F dentifrice. The authors reported kappas for 2 x 2 tables, focusing on the examiners ability to distinguish between sound and diseased surfaces, active and inactive lesions, and cavitation, separately. However, no overall assessment of the scoring system was included. Intra-examiner reliability data from Table 5 of their report will be used to illustrate one method and are presented in Table 1. These data represent pooled intra-examiner replicates for 1 examiner that were performed each year during the trial. The distinction of active and inactive states is ignored, and a 5-level modification of their 10-level diagnostic system is derived to focus on examiner reliability to distinguish cavitated and non-cavitated lesions overall. The 5-level modification includes sound surfaces, non-cavitated (NC), cavitated (C), filled (F), and a combination DF of filled and decayed (cavitated or non-cavitated) lesions. In this form, the system can be viewed as a quasi ordinal scaled diagnostic system. As such, weighted kappa statistics are appropriate to evaluate the level of examiner agreement. However, one should first test the marginal distributions for homogeneity to check for examiner drift. The marginal distributions for this examiner are quite similar. The weighted kappa (linear weights) statistic for these data is = 0.89. Kappa statistics (exact) for prevalence of disease (caries) at the noncavitated and cavitated lesion level are D1 = 0.86 and D3 = 0.93, respectively. The results given in their report [Nyvad et al., 1999] for active versus inactive diagnoses are difficult to interpret, because they pool these differences across several trait categories. However, their data certainly suggest this examiner was internally consistent (intra-examiner reliability). The authors present interexaminer data showing the 2 examiners maintained the same level of consistency (reproducibility) over the 3-year clinical trial [Nyvad et al., 1999, Table 4].
S NC C F DF Total
S 14486 242 21 14 1 86%
NC 271 1006 26 4 4 8%
C 7 23 245 3 2 2%
F 22 7 8 556 36 4%
DF 0 6 3 14 157 1%
Total 86% 7% 2% 3% 1% 17164
Wgt kappa = 0.89; Kappa (NC) = 0.86; Kappa (C) = 93
TABLE 1.
Intra-examiner reliability data. 83
KINGMAN
B/G 0 1 2 3 4 5 6 Total 0 1226 17 21 1 0 0 4 81% 1 30 32 16 0 0 0 0 5% 2 38 8 80 0 2 1 0 8% 3 1 1 2 1 0 0 0 0% 4 5 2 17 0 11 0 0 2% 5 1 1 2 2 1 2 3 1% 6 3 0 0 0 1 2 41 3% Total 83% 4% 9% 0% 1% 0% 3% 1576
Wgt kappa = 0.79; Kappa (PCA) = 0.74; Kappa (OCA) = 0.74
TABLE 2.
Inter-examiner agreement for ICDAS.
B/G S PCA OCA TOTALS
S 1226 39 4 81%
PCA 69 140 3 13%
OCA 10 24 61 6%
Totals 83% 13% 4% 1576
Wgt kappa = 0.79; Kappa (PCA) = 0.74; Kappa (OCA) = 0.74
TABLE 3.
Inter-examiner agreement for ICDAS.
The question is whether or not such a complex scoring system can be replicated for a large group of examiners. The ICDAS diagnostic system is fairly recent and still undergoing revisions. Very little clinical data are available for this scoring system. Excellent values for weighted kappa were reported at the 2005 IADR meeting [Ismail, 2005] for some inter-examiner pairs, and adequate values for others were restricted to the untreated disease component. This reliability dataset was obtained from the author and re-analyzed to demonstrate how the weighted kappa statistics can be used to assess examiner agreement. The paired data for examiner B and G, the trainer, are presented in Table 2. The marginal distributions are somewhat different, reflecting the fact that examiner G was scoring slightly higher than examiner B. However, the magnitude of examiner bias detected was minimal. The associated weighted kappa statistic (using linear weights) was = 0.83. The associated reduced paired dataset, where categories 1-3 (preventive care advised) and categories 4-6 (operative care advised) were merged, is presented in Table 3. Weighted kappa for the 3-level scale is D3 = 0.79. Kappa statistics (exact) for the prevalence of disease at the PCA level was
84
ACCEPTANCE CRITERIA
FOR
PCA = 0.74 and at the C level was OCA = 0.74 for examiner B and G. These values represent very good agreement. However, examiners participating in a study reported by Eggertsson [Eggertsson et al., 2005] had considerable difficulty in reproducing diagnoses made on occlusal surfaces and buccal pits and lingual grooves of molars. As more experience with this scoring system is obtained, evidence regarding the reproducibility among examiners will become available. It may be necessary to permute the categories to obtain an ordinal-scaled version of the ICDAS, which would enhance its internal validity. Examiner reliability reported for the continuous scale measures for the QLF and DIAGNOdent methods looks quite good. Good reliability for QLF was reported by Tranus [Tranus et al., 2002] for assessing lesion area, average, and maximum changes in lesion fluorescence. Intraclass correlations were in the 0.93 to 0.99 range for both intra- and inter-examiner reliability. However, investigators do not present the distributions for their outcome measures, and it is more difficult to assess how well these ICCs reflect good examiner reliability. Another factor influencing the reproducibility of these high-technology methods is the 2-phase nature of the process. In some reports, individual, highly-trained investigators can demonstrate high reliability, but it is not always evident whether or not the results are for image selection, image interpretation, or both. In reports of findings from studies involving multiple examiners, care must be given to present findings for reliability of image selection and image interpretation. Validity Concerns The validity issue is of paramount importance. Caries researchers are somewhat handicapped by their predilection to view validity issues through fluoride lenses. So much information has been published regarding the efficacy and effectiveness of fluoride products that there is a tendency to make assumptions regarding issues of validity that may or may not be applicable to new active systems. Two important aspects of validity need clarification. Content Validity The first is content validity. By this is meant, do diagnosed lesions (white or brown spots) or non-cavitated lesions correspond to histological evidence of
85
KINGMAN
dental caries? For example, when evaluating white spot diagnoses, investigators are careful to report that tooth surfaces need to be examined wet and dry to ensure the white spot lesions are specific to caries (as opposed to fluorosis, enamel opacity, or defect). Various methods (histology, polarized light, TMR) have been accepted as gold standards for cavitated lesions. Are there gold standards for white and brown spot lesions? What is the appropriate gold standard to determine whether a lesion is active or inactive? A rigorous assessment of diagnostic accuracy is unavailable for most of these methods. Many investigators report accuracy corresponding to a single sensitivity and specificity pair, whether they are reporting for an ordinal-based clinical-visual scale or for a continuous measure generated by the QLF or DIAGNOdent methods. Very few investigators perform a Receiver-OperatingCharacteristic (ROC) analysis based on several cutpoints (sensitivity, specificity pairs) for disease, and even fewer report diagnostic test accuracy based on the area under the ROC curve (AUC). The ROC curve is a method of describing the accuracy of a diagnostic test that incorporates a range of decision thresholds. It is a plot of the sensitivity of the test against the complement of its specificity (1 SP). Using the area under the curve (AUC) for a ROC curve (or a partial area in the curve) is often a useful means of describing the accuracy of the test or diagnostic procedure [Zhou et al., 2002]. The content validity question regarding active versus inactive lesions may be more difficult to assess, because it depends on what is used as a gold standard. Some investigators require the presence of visible plaque on a portion of the surface, while others base their diagnosis on aspects such as the texture of the surface as felt by the probe. More research is needed to convince caries researchers that lesion activity can be accurately measured. Content validity for QLF and DIAGNOdent has been demonstrated to a degree, but via a more indirect route. Two studies have reported evidence that QLF or DIAGNOdent are able to separate S, D1, D2, and D3 lesions. One study [Aljehani et al., 2004] showed great separation between sound (S), outer half enamel (OHE), and inner half enamel (IHE) lesion depth for these methods: DIAGNOdent values ranged between (0,1) for sound, (2,7) for OHE, and (11,1)] for IHE surfaces; QLF values ranged between (5,7) for sound, (13,20) for OHE, and (23,27) for IHE surfaces.
86
ACCEPTANCE CRITERIA
FOR
In a second study, a 2-year longitudinal study reported at the 2005 ORCA meeting [Katz et al., 2005] that QLF produced good separation between sound and carious lesions, but the abstract does not include ranges of values for gradations of carious lesions. The authors pointed out that the results for validity were dependent on the depth or lesion. If the level of separation shown by Aljehani can be replicated, it would provide stronger evidence of the internal validity for the DIAGNOdent and CLF diagnostic techniques. However, one needs to rely on the direct validity comparison for the clinical-visual scale to correctly conclude that the QLF or DIAGNOdent measures are valid. Again, rigorous analyses of accuracy for these laser methods are not routinely reported. A quality evaluation of the DIAGNOdent method was presented at the 6th Indianapolis Conference in 2003 [Eakle et al., 2003]. The authors concluded from an ROC analysis that the optimal sensitivity/specificity tradeoff was achieved to distinguish between sound and carious lesions (any level) for a DIAGNOdent cutoff value between 15 and 20. They also concluded that DIAGNOdent was significantly better than radiographs at detecting lesions in the inner-half of the enamel. In another study, investigators evaluated 71 non-cavitated approximal surfaces on 40 extracted premolars by DIAGNOdent and QLF [Shi et al., 2001]. The gold standard consisted of recording lesion depth with a 5-point scale using histopathology and microradiography on sectioned teeth. The QLF method produced 94% sensitivity and 100% specificity, using a cutoff point of 20% of fluorescence loss; DIAGNOdent produced values of 75% sensitivity and 96% specificity, using a cut-off point of 9. It would be more informative if investigators report the corresponding pairs for other cutoffs for fluorescence loss as well. This would allow for a more direct comparison with the findings published by Eakle [Eakle et al., 2003]. Predictive Validity The second aspect is predictive validity. At least 3 criteria should be mentioned. The first criterion is that the new scoring method should be able to statistically distinguish fluoride products having established efficacy in the usual 2to 3-year clinical trial. For example, it should be possible to separate a 1100 ppm F dentifrice from a 2800 ppm F dentifrice for preventing dental caries assessed by the new scoring system.
KINGMAN
7 6
DMFS increment
5 4 3 2 1 0 500 1100 2800
Examiner A Examiner B
product
FIG. 1.
Puerto Rico caries clinical trial24-month D2MFS increments.
A 2-year clinical trial conducted in Puerto Rico was reported by Stookey [Stookey et al., 2004] comparing a 500 ppm F, 1100 ppm F, and an 1800 ppm fluoride dentifrice. The caries scoring criteria corresponded to D2 through D4 criteria as reported by Pitts [Pitts, 2001]. White spots without tactile change were not included. The authors reported 24-month D2MFS increments of 6.24, 6.27, and 5.45 surfaces for the 500 ppm NaF, 1100 ppm NaF, and 2800 ppm NaF dentifrices, respectively, for examiner A. The corresponding increments for examiner B were 4.42, 4.95, and 3.80 surfaces, respectively. The pattern among groups for examiner A and B is illustrated in fig. 1. Significant differences between the 1100 ppm F and 1800 ppm F dentifrices were shown by examiner A and B, separately. Thus, the study satisfies the first predictive validity criterion. The Puerto Rico study also illustrates the problem examiners have using the D2MFS scoring system. Each examiner examined all of the subjects in this trial. Therefore, the observed differences in caries are due to examiner, not subjects. Therefore, even though the examiners were trained at the beginning of the study, there were systematic differences (parallel profiles among study groups means) in interpretation of the diagnostic criteria evidenced in the 24-month increments. The investigators reported the D2MFS increments but do not present or discuss the 2-year D3MFS increments for the study groups. The second criterion is whether one demonstrates that changes in these diagnostic measures predict the onset of D3 lesions in human populations. Are the probabilities of developing D3 lesions increased for non-cavitated lesions
ACCEPTANCE CRITERIA
Untreated Group
0 36 S INC ANC C or F U S 88 18 22 3 84 INC 4 42 22 2 4 ANC 5 18 28 0 9
FOR
C or F 3 22 27 94 3
Total 7668 347 372 419 1328
1500 ppm F Dentifrice

0 36 S INC ANC F U S 90 25 36 5 90 INC 4 53 28 3 4 ANC 4 6 19 2 4 F 2 16 16 89 2 Total 17016 1616 758 931 4004
TABLE 4.
Predictive validity. Three-year surface transition percentages.
compared to sound surfaces? Is there a trend in these probabilities for the ordered levels of non-cavitated lesions? Is the time needed for an untreated D1 lesion to progress to a D3 lesion shorter than for an untreated S surface of the same type? The findings from a 3-year RCT of a 1500 ppm F dentifrice and a placebo dentifrice reported by Nyvad [Nyvad et al., 2003] clearly demonstrate the first predictive validity criterion of their scoring system (focusing on active non-cavitated lesions). In Tables 4a and 4b, the 3-year estimates are presented for transitions between inactive and active non-cavitated lesions. There is a positive trend in probabilities for developing D3 level lesions from sound to activenon-cavitated lesions in the placebo group (3%, 22%, and 27%) and a 1500 ppm F dentifrice group (2%, 16%, and 16%), respectively. The transition probabilities to cavitated lesions are smaller for the 1500 ppm F group. These differences were more pronounced for occlusal surfaces with transition probabilities of 8%, 14%, and 52% in the placebo dentifrice group and 5%, 27%, and 34% in the 1500 ppm F dentifrice group, respectively (not shown here). Furthermore, the results show that regression or arrestment of inactive and active non-cavitated lesions is possible and is enhanced in the fluoride dentifrice group compared to the controls. It should be pointed out that these are estimates that are based on a 3-year time window. It would be interesting to see what the results are using a survival analysis of these data. How quickly do these surface groups progress
KINGMAN
to D3 lesions and to what extent is the survival time lengthened by regular use of a 1500 ppm F dentifrice? A third criterion is that the new method should have predictive validity relative to the traditional 2- or 3-year clinical trial. If a new system has higher test performance, it may be able to separate test products more efficiently. This could manifest itself either by: 1) requiring fewer subjects in the usual 2- or 3year clinical trial or 2) by demonstrating a statistical effect in a shorter term trial, say after 6- or 12-months. In either case, the 2- or 3-year traditional trial should produce the parallel finding obtained by the new method, i.e., the new method correctly predicts the finding that would have occurred using the traditional trial design. For the Chesters method, caries is assessed by an integration of the clinical visual assessment (CVA-simplified version of Dundee Selectable Threshold MethodDSTM), bitewing radiography, and fiber optic transillumination (FOTI). Changes in disease status are classified using pre-prepared matrices as follows: 0 (unchanged), 1initiation or progression, or -1 regression. All surface codes are summed for each subject to yield a subject-based hybrid D1MFS 12 month increment score. The Vilnius study [Chesters et al., 2002] was used to test the predictive validity of this model. This 24-month RCT included a 12month examination and compared a 1000- and 2500-ppm F dentifrice for efficacy using a hybrid D1MFS and the usual D3MFS increments. The results are presented in Table 5. Significant differences in 12-month D1MFS increments were detected for the 1000-ppm F and 2500-ppm F dentifrices. These differences were confirmed for the 24-month D1MFS increments as well as the traditional D3MFS increment analysis (CVA and FOTI at the D3 level + radiography). This study provides an example of predictive validity by the third criterion in 2 respects. The first is that related to proposing a new scoring system and conducting the trial over a shorter time-period, which was the stated objective of the authors. The second perspective is that of running the trial for the usual 24-month time-period, but using an alternative scoring system. The authors were able to statistically separate treatment groups using the new D1MFS scoring system as well as by the traditional D3MFS increment analysis. Therefore, this study would support the third criterion for predictive validity for either respect.
90
ACCEPTANCE CRITERIA
12 Months D1MFS D3MFS D1MFS D3MFS 1000 9.72 3.10 24 Months 15.69 5.46 14.82 4.96 2500 8.91 2.98
FOR
%R 8.3 3.7 5.6 9.3
TABLE 5.
Vilnius 2-year caries trial. 24-month D1MFS and D3MFS increments.
Meaningful Clinical Changes More than 40 years ago, Backer-Dirks [Backer-Dirks, 1966] reported the findings of a study on the progression of white spot lesions in 8 year-old children. He found that the vast majority (87%) of these lesions either regressed or remained as white spot lesions, while 13% of the white spot lesions progressed to D3 level lesions over a 7-year period. At the 2005 ORCA Congress, Katz [Katz et al., 2005] reported on a study in which 1117 surfaces were followed for 2-years. The authors reported that QLF showed good separation between sound and carious lesions. Of the 248 sound surfaces followed that showed progression, 198 (80%) had a white spot visible only after drying. Twenty-one (8%) of the sites progressed further, but 13 (62%) of those went from sound to beyond a white spot in 6 months. Progression using visual examination was rare among sites that showed a white spot at baseline. QLF was unable to identify sites that progressed. Progression beyond a white spot was uncommon and, when it occurred, was seen more often without the diagnosis of an intermediate white spot. These results provide corroboration of the earlier finding by Backer-Dirks, i.e., that very few white spot lesions (indicative of early stages of the caries process) progress beyond that stage. Further evidence is provided by the findings from a 6-month trial evaluating the effects of a fluoride varnish [Tranus, 2001] in a Swedish teenaged population. This trial involved 18 subjects (30 white spot lesions) given a professional cleaning and 13 subjects (32 white spot lesions) given a professional cleaning followed by a fluoride varnish. All treated premolars and molars were evaluated using the QLF system 5 times, using 6-week intervals. There was a significant change (statistically) in both lesion area and average fluorescence in the cleaning plus fluoride varnish group during the study, but no significant
91
KINGMAN
changes were detected for the cleaning-only group. There was a significant difference (statistically) between the changes for the cleaning-plus-fluoride varnish and cleaning-only group. No clinical changes in the lesions were detected by visual inspection. The average changes in fluorescence were very small. This study was not followed up over a 2- or 3-year period to investigate whether or not significant changes in caries status could be detected between test groups. A second clinical trial using the QLF method was conducted in a Japanese population over a 12-month period [Kambara et al., 2003]. Participants used either a placebo (n = 65) or a 950 ppm fluoride dentifrice (n = 67) for 1-year. Averages of 3 replicate fluorescence images of white spot lesions were taken at 0, 3, 6, and 12 months. Changes from baseline for the 3 follow-up visits were calculated. Significant differences between the changes in the placebo and fluoride dentifrice groups were obtained for avg F, Q, maxF and lesion area. In contrast to the Swedish study, the changes in lesion area with the fluorescence measures in the Japanese population were substantial (4 to 5 times as large as those observed in the Swedish population) for the fluoride group and also the difference between test groups. The observed findings parallel those that one would predict for DMFS increment if this trial were extended to 36 months. However, the study was not continued to determine whether or not this occurred. These studies raise a major concern regarding clinically meaningful change in tooth status. The high-technology methods, specifically QLF or DIAGNOdent, generate continuous outcome measures with high discrimination potential. The Japanese and Swedish studies suggest that the continuous outcome measures reported for the QLF method are capable of separating fluoride products of known efficacy in short periods of time. Thus, they would qualify as satisfying the first criterion for predictive validity. However, they involve changes in clinical status that are statistically significant but are of questionable clinical relevance. The clinical relevance of changes in white spot lesions is at issue for several reasons. The first is that fluorescence is an indirect measure of the caries process. As such, one could consider it as a surrogate for caries or one could demonstrate that it provides a complete separation of values for a valid ordinalscaled caries process scoring system. It may not qualify as a surrogate if many carious lesions develop without making the transition through the white spot stage [Fleming, 2005; Burzykowski et al., 2005]. A second reason is that very
ACCEPTANCE CRITERIA
FOR
small, clinically unimportant changes in fluorescence can be judged as statistically significant [Tranus et al., 2001], even for small groups of subjects. The fact that these changes are unimportant is evident, because white spot lesions have been shown to have a very low propensity for progression to a cavitated lesion. Finally, it will be very difficult to validate the relevance of such changes, because the median time spent in the white spot state is longer than 36-months [Backer Dirks, 1966], which is the gold standard for caries clinical trial duration. Therefore, 1- or 2-year changes in fluorescence for white spots, either within a test product group or between changes in test product groups, if statistically significant, must be examined for clinical importance. Similar arguments may be given for the inclusion of other types of noncavitated lesions in the scoring system. The importance of retaining outcome measures that are of some clinical relevance to the subject is an important criterion Summary In summary, a set of necessary, but not necessarily sufficient issues regarding the acceptance of changes in study design or scoring system criteria in caries clinical trials include the following. The scoring system or diagnostic method should be practical and simple to use. That is, it should to be able to be used by a widely diverse group of clinical examiners, including multi-site locales having sophisticated equipment, as well as in more practical settings akin to those often encountered in a school setting or community clinics. The scoring system or diagnostic method must be demonstrated as being reliable and have content or construct validity among trained investigators. It must be able to statistically distinguish products of known efficacy from one another. It must have predictive validity involving the demonstration that the ordered categories for scoring non-cavitated lesions have higher probability of becoming cavitated during a 2- or 3-year period than sound surfaces and that there is a trend among these probabilities for the ordered categories of non-cavitated lesions.
93
KINGMAN
It must have demonstrable predictive validity relative to the traditional 2- or 3-year caries clinical trial. This could be manifested in 1 of 2 ways: using a shorter time-period for the trial or using fewer subjects for the usual term of the trial. If product A and product B are statistically distinguishable at 6 or 12 months, they should be statistically distinguishable at 24- or 36-months by following the same cohorts. Likewise, if A and B are statistically distinguishable at 24- or 36-months by the new method, they should be statistically distinguishable for the traditional method. The primary outcome measure should represent meaningful changes in the carious process that represent mineral loss in enamel or comprise a bona fide surrogate for such changes. External Validity The use of an RCT design rather than an alternative study design greatly enhances the internal validity of the study findings. However, adherence to these conditions does not guarantee the external validity or relevance of the study findings to the intended user population. Strictly speaking, the results of the study apply only to the set of participants who were randomized in the study. The prevalence of the disease in the tested population, eligibility criteria used for entry into the trial, clinical criterion for diagnosing caries, monitoring techniques, degree of compliance with treatment, and attrition rate are factors affecting external validity. We have focused on the assessment methods for diagnosing caries. However, changes in study design such as: 1) modifying the eligibility criterion by requiring substantial numbers of higher-risk subjects in the test groups, 2) altering the traditional full-mouth caries examination and incorporating a partial mouth examination (scoring only higher-risk tooth surfaces) within subjects, or 3) incorporating supervision of treatment use into the trial are alternative avenues that could be investigated. In the past, many clinical trials of fluoride dentifrice were conducted utilizing supervised brushing as part of the trial design. In recent years, some clinical trials have been conducted in higher-risk populations, e.g., including only subjects who have evidence of caries at the baseline (say DMFS >0 or DMFS >1). Another possibility would be to focus on higher-risk tooth surfaces in the study design, e.g., using a partial mouth examination approach could consist of scoring only molar surfaces in the trial.
ACCEPTANCE CRITERIA
FOR
The traditional D3MFS increment-based trials have served us well as a venue for demonstrating the caries preventive benefits of fluoride-based products. The wide acceptance and broad-based utilization of these effective products and procedures has resulted in a substantial reduction in dental caries in the U.S. and in many countries throughout the world. One consequence of this decrease in dental caries has been an exponential increase in the cost of conducting RCTs using the traditional method. The traditional method is based on a crude index of caries, and, therefore, is relatively inefficient. Clearly, in todays environment, newer, more efficient methods are needed to test new drug products as they become available. Newer methodologies will evolve and be adopted as a result of increased knowledge about the caries process, improved diagnostic systems for scoring disease status, and statistical methods that can use more efficient models based on an accepted ordinal-based caries outcome in RCTs.
References
Aljehani A, Tranus S, Forsberg, C-M, Angmar-Mnsson B: In vitro quantification of white spot enamel lesions adjacent to fixed orthodontic appliances using quantitative light-induced fluorescence and DIAGNOdent. Acta Odontol Scand 2004;62:313-318. Backer Dirks O: Posteruptive changes in dental enamel. J Dent Res 1966;45:503-511. Burzykowski T, Molenberghs G, Buyse M: The evaluation of surrogate endpoints. Springer, 2005. Nyvad B, Machiulskiene V, Baelum V: Reliability of a new caries diagnostic system differentiating between active and inactive caries lesions. Caries Res 1999;33:252-260. Chesters RK, Ellwood RP, Biesbrock AR, Smith SR: Potential modern alternative designs for caries clinical trials (CCTs) and how these can be validated against the conventional model. J Dent Res 83(Spec. Issue C) 2004;C122-C124. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Mateson JR, Nicholson JA, Gendvilyte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002;81(9):637-640. Eakle WS, Gansky SA, Zhan L, Featherstone JDB: Clinical evaluation of the DIAGNOdent device. In: Early Detection of Dental Caries: Proceedings of the 6th Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 2003. Eggertsson H, Papas A, Ciancio SG, Ferreira-Zandon A, Singh M, Shilby O, Fabiano J, Eckert GJ, Zero DT: Clinical calibration by five examiners using ICDAS on occlusal surfaces, buccal pits and lingual grooves. 2005 ORCA Abstract #37. Caries Res 2005;39:299.
95
KINGMAN
FDA, biological testing procedures for fluoride dentifrices, Docket No. 80N-0042, Dockets Management Branch (HFA-305), Food and Drug Administration, Rockville, MD. Federal Register, Anticaries Drug Products for Over-the-Counter Human Use: Final Monograph; Final Rule, FDA, DHHS, 1995 Oct 6; 60(194):52474-52510. Fleming T: Surrogate endpoints and FDAs accelerated approval process. Health Affairs 2005;24:67-78. Fyffe HE, Deery C, Nugent ZJ, Nuttall NM, Pitts NB: Effect of diagnostic threshold on the validity and reliability of epidemiological caries diagnosis using the Dundee Selectable Threshold Method for caries diagnosis (DSTM). Community Dent Oral Epidemiol 2000;28:42-51. Fyffe HE, Deery C, Nugent ZJ, Nuttall NM, Pitts NB: In vitro validity and reliability of the Dundee Selectable Threshold Method for caries diagnosis (DSTM). Community Dent Oral Epidemiol 2000;28:52-58. Ismail AI: Visual and visuo-tactile detection of dental caries. J Dent Res 83(Spec. Issue C) 2004;C56-C66. Kambara M, Uemura M, Miyake T, Doi T, Nakashima S, Eckert GJ, Stookey GK: Results of clinical trial of fluoride dentifrice using QLF. In: Early Detection of Dental Caries: Proceedings of the 6th Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 2003, pp 229-235. Katz BP, Ofner S, Eckert GJ, Ferreira-Zandon AG, Ando A, Eggertsson H, Mau MS, Kelly SA, Doi T, Lukantsova, Wefel JS, Stookey GK: Measuring the caries process over time. 2005 ORCA Abstract #93. Caries Res 2005;39:319. Ismail AI: Visual and visuo-tactile detection of dental caries. J Dent Res (Spec. Issue C) 2004;83:C56-C66. Ismail AI: Reliability of the international caries detection and assessment system. J Dent Res 2005;84:Abstr #141. Lussi A, Hibst R, Paulus R: DIAGNOdent: An optical method for caries detection. J Dent Res 83(Spec. Issue C) 2004;C80-C83. Nyvad B, Machiulskiene V, Baelum V: Construct and predictive validity of clinical caries diagnostic criteria assessing lesion activity. J Dent Res 82(2);2003:117-122. Pitts NB: Clinical diagnosis of dental caries: a European perspective. J Dent Educ 2001;65:972-978. Pitts NB: ICDASan international system for caries detection and assessment being developed to facilitate caries epidemiology, research and appropriate clinical management. Community Dent Health 2004;21:193198. Radike AW: Criteria for diagnosing dental caries. In: Proceedings of the Conference on the Clinical Testing of Cariostatic Agents. Chicago: American Dental Association, 1972:87-88. Shi XQ, Tranus S, Angmar-Mnsson B: Comparison of QLF and DIAGNOdent for quantification of smooth surface caries. Caries Res 2001;35:21-26. Stookey GK, Mau MS, Isaacs RL, Gonzles-Gierbolini, Bartizek RD, Biesbrock AR: The relative anticaries effectiveness of three fluoride containing dentifrices in Puerto Rico. Caries Res 2004;38:542-550. Stookey GK: Optical methodsquantitative light fluorescence. J Dent Res 83(Spec. Issue C) 2004;C84-C88.
96
ACCEPTANCE CRITERIA
FOR
Tranus S, Al Khateeb S, Bjrkman S, Twetman S, Angmar-Mnsson B: Application of quantitative lightinduced fluorescence to monitor incipient lesions in caries-active children. A comparative study of remineralisation by fluoride varnish and professional cleaning. Eur J Oral Sci 2001;109:71-75. Tranus S, Shi XQ, Lindgren LE, Trollsas K, Angmar-Mnsson B: In vivo repeatability and Reproducibility of the quantitative light-induced fluorescence method. Caries Res 2002;36:3-9. World Health Organization: A guide to oral health epidemiological investigations. Geneva: World Health Organization, 1979. Zhou XH, Obuchowski NA, McClish DK: Statistical Methods in Diagnostic Medicine, 2002 Wiley New York.
97
98
The Nyvad Criteria for Assessment of Caries Lesion Activity

Bente Nyvada, V. Machiulskieneb, V. Baeluma
a b
Faculty of Health Sciences, University of Aarhus, Denmark Faculty of Odontology, Kaunas University of Medicine, Lithuania
Introduction Over the past few decades, we have seen an accelerated development in new diagnostic technologies. This is reflected in a substantial expansion of research on diagnostic tests. According to a MEDLINE search, the number of such publications increased from about 2,000 in the period 1966-1970 to about 17,000 in the period 1996-2000 [Knotterus and Weel, 2002]. It is unfortunate, however, that the methodology for evaluating diagnostic techniques lags far behind that of research into the effectiveness of therapeutic interventions. This is also true for caries diagnostic methods that are often introduced for clinical use without prior critical appraisal. Like other research disciplines, diagnostic research has its specific principles and rules. In an attempt to improve methodological development of the evidence base of clinical diagnosis in medicine, Knotterus and co-workers [2002] have recently published a textbook, The Evidence Base of Clinical Diagnosis, which presents a framework for investigators who want to conduct diagnostic research. The overall goal of diagnostic testing is the collection of additional information with the intention to (further) clarifying the character and prognosis of the patients condition such that the ultimate objective of the diagnostic phase is to optimise the patients prognosis by enabling the clinician to choose an adequate therapeutic strategy [Knotterus and Muris, 2002]. When assessing a test, it is therefore not sufficient to look at the test itself; it is also important to consider the specific question the test is supposed to answer. Hence, the performance of diagnostic tests must be evaluated in accordance
NYVAD
ET AL.
with their intended objectives. Key objectives include [Knotterus and Weel, 2002]: Detecting or excluding disease Assessing prognosis Contributing to the decision-making process with regard to further diagnostic and therapeutic management Monitoring the clinical course Choosing the appropriate study design for diagnostic testing depends on the research question. The most important designs are the cross-sectional study, used to determine the accuracy and added value of diagnostic procedures, and the randomized controlled trial, used to evaluate the clinical impact of testing. When preparing and reporting the results of a clinical diagnostic study, it is imperative to consider generalizability, as this will have obvious implications for clinical applicability. Studies may be designed to focus on maximising the health perspectives of the individual patient (the physicians aim) or to focus on the best possible cost effectiveness (the economists perspective). To make the step from research to practice, clinical decision analysis, cost-effectiveness studies, and quality of care research will also have to be considered [Knotterus and Weel, 2002]. Up until now, caries diagnostic research has rarely been concerned with a systematic approach to the assessment of new diagnostic technologies. The aim of this presentation is to demonstrate a systematic evaluation of the clinical effectiveness of a caries diagnostic methodthe Nyvad method for assessment of caries lesion activitybased on the principles outlined above, with the main focus on maximizing the health outcome perspective. Prior to the evaluation, the pathobiological rationale and the technology of the method will be described. Finally, the benefits of using this diagnostic method in epidemiology and in clinical trials will be highlighted. Pathobiological Rationale The pathobiological rationale of caries activity assessment rests on the well-known empirical observation that the clinical status of lesions may change over time, in particular, if lesions are followed at the non-cavitated level of diagnosis [for review; Nyvad and Fejerskov, 1997]. This concept was already described a century ago by Black [1914], who noted that, Even in cases of
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
CARIES LESION ACTIVITY
marked whitening of the enamel in several teeth, my experience shows plainly that the decay of the enamel can be effectively checked in any case in which the enamel has not been penetrated. The brush and water are all that are needed, but these must be correctly used to be effective. Similar, but considerably more systematic observations were made half a century later by Backer-Dirks [1966]. He studied the development of white spot lesions on the buccal surfaces of maxillary molars in children and found that 36% of such lesions became arrested and half of the lesions regressed to sound over a 7-year period. Backer-Dirks interpreted the observations to be mainly a result of altered environmental conditions due to the masticatory cleansing effect associated with the fully erupted teeth, which promoted the natural removal of bacterial accumulations and, hence, lesion arrest. Regrettably, Backer-Dirks original observations were not followed up, possibly because of the common perception among some influential sections of the scientific community that it is not possible to make a reliable diagnosis of non-cavitated caries lesions, cf. WHO [1997]. It was not until the 1980s, when Thylstrup and co-workers, in a series of elegant in situ studies, explored the enamel reactions to undisturbed plaque, that the issue of caries activity assessment again became a topic of clinical interest [for review; Thylstrup et al., 1994]. These authors found that caries lesion activity is reflected in the surface reflection and texture of the teeth; chalky and rough lesions being active and shiny, and smooth lesions being inactive/arrested [Holmen et al., 1987a; Holmen et al., 1987b]. A few years later, such criteria for diagnosing non-cavitated caries lesions were officially adopted by the Danish National Board of Health [1988]. These diagnostic criteria were also taught at the dental schools of Denmark; however, without having passed a formal evaluation. This was the background for the studies that were initiated in 1994 and aimed at a formal critical evaluation of caries diagnostic criteria, which discriminate between active and inactive lesions. Caries Activity Assessment According to Nyvad et al. Caries activity assessment was primarily intended as a tool for diagnosis and treatment planning for the individual patient in dental surgery. However, at the same time, it should be possible to apply the criteria for caries epidemiological studies and clinical trials. It was therefore decided to develop and test the criteria in a population of adolescents with relatively high caries prevalence and incidence for which it was known that all the diagnostic categories would
NYVAD
ET AL.
be represented [Machiulskiene et al., 1998]. Moreover, the criteria were designed to meet the recommendation of including enamel and dentine lesions at both the non-cavitated and cavitated levels [Ismail, 2004] in order to allow for flexibility with respect to prospective clinical and scientific applications. The description of the diagnostic criteria and the reliability of the criteria were presented some years ago [Nyvad et al., 1999]. In the original version of the criteria (Table 1), active and inactive lesions were detected at 3 different levels of severity; intact surfaces, surfaces with localized discontinuity (microcavity in enamel), and surfaces with overt cavity. In addition, scores were reserved for fillings and fillings with active and inactive caries, respectively. In later applications, we have incidentally reduced the number of diagnostic categories by collapsing 2 of the scores, localized discontinuity and overt cavity, into
Score 0 1
Category Sound Active caries (intact surface)
Active caries (surface discontinuity) Active caries (cavity) Inactive caries (intact surface)
Inactive caries (surface discontinuity) Inactive caries (cavity) Filling (sound surface) Filling + active caries Filling + inactive caries
Criteria Normal enamel translucency and texture (slight staining allowed in other wise sound fissure). Surface of enamel is whitish/yellowish opaque with loss of luster; feels rough when the tip of the probe is moved gently across the surface; generally covered with plaque. No clinically detectable loss of substance. Smooth surface: caries lesion typically located close to gingival margin. Fissure/pit: Intact fissure morphology; lesion extending along the walls of the fissure. Same criteria as score 1. Localized surface defect (microcavity) in enamel only. No unermined enamel or softened floor detectable with the explorer. Enamel/dentin cavity esily visible with the naked eye; surface of cavity feels soft or leathery on gently probing. There may or may not be pulpal involvement. Surface of enamel is whitish, brownish, or black. Enamel may be shiny and feels hard and smooth when the tip of the probe is moved gently across the surface. No clinically detectable loss of substance. Smooth surface: Caries lesion typically located at some distance from gingival margin. Fissure/pit: Intact fissure morphology; lesion extending along the walls of the fissure. Same criteria as score 4. Localized surface defect (microactivity) in enamel only. No undermined enamel or softened floor detectable with the explorer. Enamel/dentin cavity easily visible with the naked eye; surface of cavity may be shiny and feels hard on probing with gentle pressure. No pulpal involvement. Caries lesion may be activated or non-cavitated. Caries lesion may be cavitated or non-cavitated.
7 8 9
TABLE 1.
The Nyvad caries diagnostic criteria for assessment of the stage and activity of lesions [Nyvad et al., 1999].
102
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
1 diagnostic category, cavity, because of a relatively low number of surfaces presenting with localized discontinuity. Assessment of the activity status of a lesion is determined on the basis of combined visual-tactile criteria after drying of the teeth. The method focuses on characterizing the surface characteristics of lesions, primarily integrity, and surface texture. It is important to stress that use of a sharp probe is an integral part of the diagnostic process when diagnosing carious lesions according to the Nyvad method. However, the probe is not used to poke vigorously into the tissues (as most Americans tend to believe) but rather is used as a highly refined tactile instrument (the Scandinavian approach). In fact, poking with the probe necessitates a firm grip, which effectively precludes use of the probe in the highly tactile mode necessary with the Nyvad method. Use of the sharp probe basically serves 2 purposes during the diagnostic process; first, to check for signs of demineralization and surface break beneath the dental plaque by scraping the plaque away, and second, to feel the surface roughness of a lesion, as sensed through vibrations of the instrument by the supporting fingers. To use the probe in this way, of course, requires a gentle grip on the probe. The typical characteristics of an active enamel caries lesion are those of a whitish/yellowish opaque surface with loss of luster; the surface feels rough when the tip of the probe is moved gently across the surface. Active non-cavitated enamel lesions are normally covered by tacky dental plaque and removal of this adherent material (using the side of the probe) is part of the diagnostic process. In fact, we have observed that professional removal of plaque prior to diagnosis makes it more difficult to ascertain an active lesion. This may also be a complicating factor with repeated measurements, such as in calibration studies where plaque has often been removed by the first examiners. Inactive enamel caries lesions do not present the same problem, as they are usually not covered by dental plaque. Inactive enamel lesions are generally shiny and feel smooth upon gentle probing. The colour of an inactive enamel lesion may vary from whitish to brownish or black. However, colour is not a central differential caries diagnostic characteristic for such lesions. For cavitated lesions, the criteria mimic those applied for root caries [Nyvad and Fejerskov, 1986], active lesions being soft or leathery, while inactive lesions may be shiny and feel hard and smooth upon gentle probing. When applying activity assessment, it is important to realize that the clinical presentation of a lesion always represents the cumulated result of numerous
NYVAD
ET AL.
de- and remineralization episodes that have occurred over a longer period of time [Fejerskov, 1997]. For some lesions, demineralization has predominated, for others, remineralization has predominated. In some cases, de- and remineralization is going on intermittently to the same extent. Because of the dynamic nature of caries, the clinical appearance of a lesion may not always match the typical description of lesion activity, as outlined above. For example, non-cavitated inactive lesions may develop localized microfractures of the outermost enamel as a result of wear and tear in the oral cavity [rtun and Thylstrup, 1986]. Such lesions may therefore present with smoothed out surface irregularities that have a distinctly different topography compared to the surface roughness characteristic of active non-cavitated lesions. In cases of mixed lesions, the lesions may contain components of both active and inactive caries. In such cases, we recommend that lesions be scored as active. When adopting such decision rules, the criteria system has shown a high reliability when used under epidemiological conditions by 2 well-calibrated examiners. Repeated measurements over 3 years showed that the kappa values ranged between 0.74 and 0.85 for intra-examiner agreement in caries diagnosis and between 0.78 and 0.80 for inter-examiner agreement [Nyvad et al., 1999]. Most of the recorded misclassifications (about 80%) involved disagreement between sound surfaces and noncavitated caries lesions. Disagreement between sound surfaces and non-cavitated active or non-cavitated inactive lesions (31% and 32%, respectively) was more common than disagreement between non-cavitated active and non-cavitated inactive lesions (11%). More importantly, there was no evidence of systematic deviations in the distribution of specific categories of diagnoses. Despite the inherent shortcomings of kappa statistics, we have therefore concluded that the reliability of caries activity assessment is in the same order of magnitude as reported in studies that have included non-cavitated caries diagnoses into the criteria system [Pitts and Fyffe, 1988; Manji et al., 1989; Ismail et al., 1992] Detecting or Excluding Disease A crucial feature of a diagnostic test is its ability to discriminate between specific diagnostic categories. Most often, the discriminatory power of a test is achieved by comparing the test result with a diagnostic standard, the so-called gold standard (concurrent criterion validity). However, a gold standard pro-
104
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
viding full certainty on the disease status rarely exists. This is true for the caries process per se, as well as for caries activity status. According to the pathobiological definition referred to above, the concept of caries activity is a surfacebased phenomenon that cannot be imitated by the commonly applied histological caries gold standard, which merely reflects depth of caries lesion penetration. In the absence of a suitable concurrent reference standard, there are, however, other ways of appraising validity, such as construct validity and predictive criterion validity. In construct validity, the measurement reflects theoretical concepts regarding the phenomenon to be studied [Last, 1995]. Hence, criteria may be said to have construct validity, if they can reflect the well-known caries-controlling effect of fluoride that is brought about by inhibition of demineralization and enhancement of remineralization [Fejerskov et al., 1981; ten Cate and Featherstone, 1996]. We evaluated the construct validity by a re-analysis of the data from a 3-year clinical trial of the effect of daily supervised brushing with fluoride toothpaste [Machiulskiene et al., 2002], which also included a control group in which no professional intervention was performed. The analyses showed that the relative risks for caries lesion transitions in the fluoride group relative to the control group imitated the anticipated effect of fluoride on caries lesion dynamics. Hence, at all stages of lesion formation, supervised brushing with fluoride toothpaste inhibited the development or progression of lesions, while at the same time enhancing lesion arrest and regression [Nyvad et al., 2003] (fig. 1). These effects were most pronounced for active non-cavitated lesions, which supports the notion that fluoride exerts its predominant effect on the active caries process [Fejerskov et al., 1981]). When interpreting, the results it should be appreciated that up to 70% of participants in the control group may have used fluoride toothpaste at their own initiative, as there was no intention to interfere with already established caries preventive routines. This may have reduced the contrast of the outcome between the 2 experimental groups. Nevertheless, it was concluded that the diagnostic criteria had construct validity for assessment of caries lesion activity. Predictive criterion validity was assessed in the same study [Nyvad et al., 2003] by contrasting the risks that different lesion types (active non-cavitated, inactive non-cavitated, or sound) would progress to the cavitation stage over a 3-year period. It was predicted that active non-cavitated lesions would have a higher risk of progression than inactive non-cavitated lesions, which, in turn,
NYVAD
ET AL.
Sound
Active non-cavitated
Inactive non-cavitated
Cavity, filling or extracted
: :
RR < 1 RR > 1
FIG. 1.
The relative risk that a fluoride-exposed surface versus a control surface undergoes lesion transition, as observed in a 3-year caries preventive trial of daily supervised brushing with fluoride toothpaste in young teenagers. RR<1 indicates inhibition of lesion development/progression; RR >1 indicates promotion of lesion arrest/regression. Data modified from Nyvad et al. [2003].
would have a higher risk of progression than sound surfaces. It was also anticipated that the contrasts involving active non-cavitated lesions would be less marked for the fluoride toothpaste intervention group, owing to a selective effect of a combination of fluoride and mechanical effects on the active noncavitated lesions. Indeed, results showed that active non-cavitated lesions in the no-intervention group had a 24% greater risk of progression than did the inactive non-cavitated lesions, and this contrast was less marked when fluoride toothpaste was used on a daily basis. Active non-cavitated lesions in the nointervention group had a 10-fold increased risk of progression than did sound surfaces, and this was again less marked in the fluoride intervention group. Based on these results, it was concluded that these caries activity assessment criteria had predictive criterion validity. Detecting or Excluding Disease A crucial feature of a diagnostic test is its ability to discriminate between specific diagnostic categories. Most often, the discriminatory power of a test is
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
achieved by comparing the test result with a diagnostic standard, the so-called gold standard (concurrent criterion validity). However, a gold standard providing full certainty on the disease status rarely exists. This is true for the caries process per se, as well as for caries activity status. According to the pathobiological definition referred to above, the concept of caries activity is a surfacebased phenomenon that cannot be imitated by the commonly applied histological caries gold standard, which merely reflects depth of caries lesion penetration. In the absence of a suitable concurrent reference standard, there are, however, other ways of appraising validity, such as construct validity and predictive criterion validity. In construct validity, the measurement reflects theoretical concepts regarding the phenomenon to be studied [Last, 1995]. Hence, criteria may be said to have construct validity, if they can reflect the well-known caries-controlling effect of fluoride that is brought about by inhibition of demineralization and enhancement of remineralization [Fejerskov et al., 1981; ten Cate and Featherstone, 1996]. We evaluated the construct validity by a re-analysis of the data from a 3-year clinical trial of the effect of daily supervised brushing with fluoride toothpaste [Machiulskiene et al., 2002], which also included a control group in which no professional intervention was performed. The analyses showed that the relative risks for caries lesion transitions in the fluoride group relative to the control group imitated the anticipated effect of fluoride on caries lesion dynamics. Hence, at all stages of lesion formation, supervised brushing with fluoride toothpaste inhibited the development or progression of lesions, while at the same time enhancing lesion arrest and regression [Nyvad et al., 2003] (fig. 1). These effects were most pronounced for active non-cavitated lesions, which supports the notion that fluoride exerts its predominant effect on the active caries process [Fejerskov et al., 1981]). When interpreting, the results it should be appreciated that up to 70% of participants in the control group may have used fluoride toothpaste at their own initiative, as there was no intention to interfere with already established caries preventive routines. This may have reduced the contrast of the outcome between the 2 experimental groups. Nevertheless, it was concluded that the diagnostic criteria had construct validity for assessment of caries lesion activity. Predictive criterion validity was assessed in the same study [Nyvad et al., 2003] by contrasting the risks that different lesion types (active non-cavitated, inactive non-cavitated, or sound) would progress to the cavitation stage over a
NYVAD
ET AL.
3-year period. It was predicted that active non-cavitated lesions would have a higher risk of progression than inactive non-cavitated lesions, which, in turn, would have a higher risk of progression than sound surfaces. It was also anticipated that the contrasts involving active non-cavitated lesions would be less marked for the fluoride toothpaste intervention group, owing to a selective effect of a combination of fluoride and mechanical effects on the active noncavitated lesions. Indeed, results showed that active non-cavitated lesions in the no-intervention group had a 24% greater risk of progression than did the inactive non-cavitated lesions, and this contrast was less marked when fluoride toothpaste was used on a daily basis. Active non-cavitated lesions in the nointervention group had a 10-fold increased risk of progression than did sound surfaces, and this was again less marked in the fluoride intervention group. Based on these results, it was concluded that these caries activity assessment criteria had predictive criterion validity. Assessing Prognosis For scientific purposes, it may be important to know whether or not a result from a medical test corresponds to the truth, as expressed by sensitivity and specificity. From the patients perspective, however, such knowledge is relatively uninteresting. What matters to the patient is the future course of his/her condition [Wulff, 1979]. Patients will only benefit from diagnostic tests if the information generated by the tests is correctly used in subsequent treatment decisions [Lijmer and Bossuyt, 2002]. Therefore, the relevance of diagnostic test information is closely related to prognosis. When assessing the prognostic value of caries activity assessment (acknowledging the value of diagnosing lesions at the non-cavitated level), the central question is whether inactive lesions fare better than active lesions. We addressed this question in the aforementioned clinical trial and found that active non-cavitated lesions ran a greater risk of progressing to a cavity than did inactive non-cavitated lesions [Nyvad et al., 2003]. As expected, the risk of developing into a cavity over a 3-year period was 24% higher for an active non-cavitated lesion compared to an inactive lesion in the no-intervention group. In the intervention group, which received daily supervised toothbrushing with fluoride toothpaste, this contrast was less marked, owing to the fact that toothbrushing exerted a selective fluoride and mechanical effect on the active, smooth surface
108
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
lesions, which, therefore, tended to disappear or become inactive. Although these differences may appear modest, the observations should be seen in the perspective of a short 3-year trial conducted in a high-caries population in which lesion development is quite fast amounting to 3-year DFMS increments in the order of magnitude of 10 surfaces [Baelum et al., 2003]. Contributing to the Decision-making Process with Regard to Further Diagnostic and Therapeutic Management As we have argued, there may be clear prognostic benefits to assessing caries lesion activity at the lesion/site level. This observation implies that caries activity assessments may be used to decide on the subsequent course of treatment. We have previously proposed a decision-making tree for dental caries that includes activity assessment as a key factor in the decision process [Nyvad and Fejerskov, 1997] (fig. 2). According to this model, the first level of the decision process is to determine the state of the tooth surface, whether sound, having a lesion, or being filled. If it is decided that there is a lesion, one may proceed to level 2, which is to assess the activity state/diagnosis of the lesion: inactive (non-progressing) or active (progressing). The third step is to generate a proper treatment decision. Inactive lesions, because of their slowly or non-progressing nature, do not require professional treatment. Active lesions, on the other hand, because of their progressive character, demand professional intervention. The professional treatment of active lesions will depend on their accessibility to plaque removal. If plaque control is likely to be insufficient, such as in cavitated lesions, the lesions should be filled. However, for non-cavitated active lesions, non-operative preventive intervention is to be preferred. This line of thinking illustrates how caries activity assessment could serve as an important diagnostic tool for making evidence-based treatment decisions. Monitoring the Clinical Course The predictive validity of the criteria, which was demonstrated above, has direct implications also for monitoring the clinical course of lesions. This is a particularly important point as it provides the clinician with a inexpensive, easy tool for controlling lesion activity in response to preventive interventions. There are several examples in the literature, in addition to the original observations by Black [1914], which show that the surface features and, hence, the activity of
109
NYVAD
ET AL.
The status of each tooth surface
Level 1:
Sound
Lesion
Filling
Level 2: (Diagnosis)
Inactive (non-progressing)
Active (progressing)
No defect
Defect
Level 3: (Treatment decision)
NOTAL (NO Treatment At alL
NOT (Non-Operative Treatment)
OT (Operative Treatment)
NOREP (NO REPlacement)
REP (REPlacement)
FIG. 2:
Decision-making tree for dental caries according to Nyvad and Fejerskov [1997]. The flow chart does not take into account factors operating at the level of the individual.
lesions may change in response to targeted oral hygiene procedures, including the use of fluoride toothpaste, both as far as enamel [rtun and Thylstrup, 1986] and root/dentin surfaces [Nyvad and Fejerskov, 1986] are concerned. Similarly, it has been demonstrated in a clinical trial that improved oral hygiene may help to arrest dentin lesions in preschool children [Lo et al., 1998]. Advantages of Applying Caries Activity Assessments According to the Nyvad Criteria Since the Nyvad criteria for assessment of caries lesion activity was first introduced, we have obtained considerable information about the clinical performance of the criteria, both from epidemiological studies and from clinical trials. A major benefit to using activity assessment in epidemiological studies relates to the fact that the results not only serve to demonstrate the prevalence of disease but also indicate the immediate need for different types of treatment (see arguments presented in previous section). Estimates of the treatment need may be of importance to clinical practitioners in dealing with the individual patient as well as for professional health care planners when allocating resources for preventive and operative interventions. For example, in a highcaries population, mean values of 2.9 cavitated and 4.9 non-cavitated surfaces with active lesions, respectively, suggest that there is a clear need for both oper-
110
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
ative and preventive interventions [Machiulskiene et al., 1998]. In contrast, mean values of 0.3 cavitated and 1.5 non-cavitated surfaces with active lesions in a low-caries population imply that interventions should primarily be focused at preventive treatments [Nyvad and Heidmann, 1999]. No other caries diagnostic method has the ability to provide estimates of the current caries status while directing the future need for treatment. It has also been demonstrated in a prospective epidemiological study that caries activity assessment may improve the prediction of caries in a low-caries population, over and above that predicted by DMFS alone [Nyvad et al., 2004]. Even if improvement of the positive predictive value of combining DMFS scores and activity assessment in a regression model was marginal and exclusively confined to high-risk groups comprising 15-30 % of the population (e.g., from 71% to 79% at a high risk group size of 16%), this observation may nevertheless be of clinical importance when designing high-risk strategies for caries control in such populations. Hence, in a subsequent theoretical study based on observed caries lesion transitions in the same population [Nyvad et al., 2005], it was proposed that identification of high-risk individuals on the basis of DMFS scores and activity assessment may be instrumental for implementing cost-effective caries preventive programs including topical fluorides [Nyvad, 2005]. The likely benefit of such a program was improved health (fewer cavities and fewer active non-cavitated lesions) rather than economic savings. A very promising perspective of applying caries lesion activity assessment relates to the possibility of generating new knowledge about the dynamics of the caries process and factors that may influence these processes. For this purpose, 2 types of caries lesion transition ratescaries incidence and caries recovery were calculated on the basis of clinical trial data on the caries-preventive effect of sugar-substituted chewing gums and fluoride toothpaste [Machiulskiene et al., 2001; Machiulskiene et al., 2002]. Using an exponential survival-time regression model, the hazard ratios for lesion transitions were estimated for the covariates experimental group (control, sugar substituted chewing gum, fluoride tooth paste) and post-eruptive surface age, gender, and surface type [Baelum et al., 2003]. When the caries transitions considered were those from active to inactive (recovery) and from inactive to active (incidence), respectively, the incidence hazard rate ratios were decreased by 18-49% in the fluoride group compared to the control group, while the recovery hazard rate ratios were increased by 51-85%, depending on the type of surface examined (Table 2).
NYVAD
ET AL.
Inactive Active Inactive (Recovery) Active (Incidence) Hazard Ratio Decrease Gum Fluoride Gum vs. Ref Fluoride vs. Ref 0.77 0.73 23% 27% 0.89 0.86 15% 18% 1.97 1.09 7% 49%
A Smooth Approximal Occlusal Ref 1 1.05 2.12
B Smooth Approximal Occlusal Ref 1 1.08 1.17
Active Inactive (Recovery) Hazard Ratio Gum Fluoride 1.03 1.60 1.16 1.63 1.28 2.17
Increase Gum vs. Ref Fluoride vs. Ref 3% 60% 7% 51% 9% 85%
TABLE 2.
The effect of daily use of sugar-substituted chewing gum or daily, supervised toothbrushing with fluoride toothpaste on the occurrence of: A) caries lesion transitions from inactive to active stages (incidence) and B) caries lesion transitions from active to inactive stages (recovery). Data from Baelum et al. [2003].
These observations suggest that that the therapeutic effect of fluoride by far exceeds the preventive effect, an observation that could explain the marked retardation of caries progression beyond the stage of enamel caries in populations subjected to fluoride in the water supplies [Groeneveld, 1985]. The observations also support the contention that fluoride predominantly exerts its effect in periods of demineralisation, i.e., when lesions are active [Fejerskov et al., 1981]. The pattern of lesion transitions observed in the fluoride group was distinctly different from the pattern obtained when using sugar-substituted chewing gums. For sugar-substituted chewing gums, the preventive effect (the incidence hazard rate ratios) decreased by 7-23% and was more pronounced than the therapeutic effect (the recovery hazard rate ratios), which was increased by 3-9% compared to the control group. Hence, inhibition of lesion development in the gum group was not as marked as that observed in the fluoride group, and there was hardly any therapeutic effect to the gum use. Collectively, these observations imply that the pattern of lesion transitions reflect the nature of the preventive program. Furthermore, it is obvious that, when applying survival analysis to caries data, based on activity assessment, it may be possible to extract far more information from clinical trials than when using the traditional DMFS-based analyses.
112
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
Concluding Remarks According to the mission of this conference, the overall goal is to develop a general consensus among the scientific community regarding the acceptability of new, more efficient models for clinical trials designed to assess the impact of caries-preventive measures on remineralization, precavitation, and caries. We have severe reservations with respect to these goals. First, it seems inappropriate to develop consensus about particular models that should be applicable for all purposes of caries clinical trials. To consolidate such a single standard may not only be biologically wrong but would tend to withhold the development of better, more discriminative models in the future. Moreover, there are several ways in which the discriminative potential of a clinical trial could be increased. Our assignment has focussed specifically on improving diagnostic effectiveness and relevance for caries management. At the end of the day, all our activities are meant to benefit our patients and the populations we serve. We believe that, for caries clinical trials and for caries recording in daily clinical practice, it is crucial that the diagnostic criteria reflect current knowledge about the disease process, that they have prognostic value, and that they are relevant for therapeutic decision making. While it may, for some purposes, be quite sufficient to record caries at the non-cavitated level without including lesion activity assessment, it should be borne in mind that reduction of a diagnostic system is likely to restrict the information that may be derived in subsequent analyses. This is amply illustrated by our alternative approaches to the analysis of data from a caries preventive trial of daily supervised brushing with fluoride toothpaste, 1 analysis being based on traditional DMFS calculations at the cavitated and non-cavitated levels [Machiulskiene et al., 2002], and the other, more informative analysis, being based on estimates of the hazard rate ratios for lesion transitions between different stages of caries activity [Baelum et al., 2003]. We are also concerned that modifications in the design of a trial for the purpose of cost minimization [Chesters et al., 2002], such as the application of rigorous exclusion criteria, may severely compromise the transferability of the methods and results to other settings and populations. We have applied an evidence-based approach to demonstrate the effectiveness of a new clinical method for the recording of caries lesion activity, and we have shown that such criteria have construct and predictive criterion validity for lesion activity assessment. The criteria have proved to produce highly
113
NYVAD
ET AL.
reproducible recordings when the examiners are thoroughly trained. Moreover, the lesion activity assessment criteria have been shown to have prognostic value for lesion progression into cavitation stages. To the best of our knowledge, no other studies have tested the long-term outcomes of caries diagnostic procedures, in spite of the fact that clinical outcomes in terms of the appropriateness of a treatment provided in response to a given diagnosis is of principle interest for patients and clinicians alike. In view of the positive results obtained, we have therefore no hesitation to recommend the Nyvad criteria for assessing caries lesion activity in daily clinical practice and in caries epidemiological studies and clinical trials.
Acknowledgment We gratefully acknowledge the support of the Colgate-Palmolive Company (USA) for allowing us to synthesize this review
References
rtun J, Thylstrup A: Clinical and scanning electron microscopic study of surface changes of incipient enamel caries lesions after debonding. Scand J Dent Res 1986;94:193-210. Backer-Dirks O: Posteruptive changes in dental enamel. J Dent Res 1966;104:480-485. Baelum V, Machiulskiene V, Nyvad B, Richards A, Vth M: Application of survival analysis to carious lesion transitions in intervention trials. Community Dent Oral Epidemiol 2003;31:252-260. Black GV: Operative Dentistry. Chicago: Medico-Dental Publishing Co., 1914;1:188-190. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson JR, Nicholson JA, Gendvilyte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: J Dent Res 2002;81:637-640. Danish National Board of Health. Sundhedsstyrelsen: Brugervejledning til Sundhedsstyrelsens Centrale Odontologiske Register. Copenhagen, 1988 (in Danish). Fejerskov O: Concepts of dental caries for understanding the disease. Community Dent Oral Epidemiol 1997;25:5-12. Fejerskov O, Thylstrup A, Larsen MJ et al: Rational use of fluorides in caries prevention: A concept base don possible cariostatic mechanisms. Acta Odontol Scand 1981;39:241-249.
114
THE NYVAD CRITERIA
FOR
ASSESSMENT
OF
Groeneveld A: A longitudinal study of prevalence of enamel lesions in a fluoridated and non-fluoridated area. Community Dent Oral Epidemiol 1985;13:159-163. Holmen L, Thylstrup A, rtun J: Clinical and histological features observed during arrestment of active enamel carious lesions in vivo. Caries Res 1987a;21:546-554. Holmen L, Thylstrup A, rtun J: Surface changes during the arrest of active enamel carious lesions in vivo. A scanning electron microscope study. Acta Odontol Scand 1987b;45:383-390. Ismail A: Diagnostic levels in dental public health planning. Caries Res 2004;38:199-203. Ismail AI, Brodeur J-M, Gagnon P, Payette M, Picard D, Hamalian T, Oliver M, Eastwood BJ: Prevalence of non-cavitated and cavitated carious lesions in a random sample of 7-9-year-old schoolchildren in Montreal, Quebec. Community Dent Oral Epidemiol 1992;20:250-255. Knotterus JA: Preface. In: Knotterus AJ ed, The Evidence Base of Clinical Diagnosis. London, BMJ Books, 2002. Knotterus JA, Muris JW: Assessment of the accuracy of diagnostic tests: the cross sectional study. In: Knotterus AJ ed, The Evidence Base of Clinical Diagnosis. London, BMJ Books, 2002, 39-59. Knotterus JA, van Weel C: General introduction: evaluation of diagnostic procedures. In: Knotterus AJ ed, The Evidence Base of Clinical Diagnosis. London, BMJ Books, 2002, 1-17. Last JM: A Dictionary of Epidemiology, ed 3. Oxford, Oxford University Press, 1995. Lijmer JG, Bossuyt PM: Diagnostic testing and prognosis: the randomised controlled trial. In: Knotterus AJ ed, The Evidence Base of Clinical Diagnosis. London, BMJ Books, 2002, 61-80. Lo EC, Schwarz E, Wong MC: Arresting dentine caries in Chinese preschool children. Int J Pediatr Dent 1998;8:253-260. Machiulskiene V, Nyvad B, Baelum V: Prevalence and severity of dental caries in 12-year-old children in Kaunas, Lithuania 1995. Caries Res 1998;32:175-180. Machiulskiene V, Nyvad B, Baelum V: Caries-preventive effect of sugar-substituted chewing gum. Community Dent Oral Epidemiol 2001;29:278-288. Machiulskiene V, Richards, A, Nyvad B, Baelum V: Prospective study of the effect of post-brushing rinsing behaviour on dental caries. Caries Res 2002;36:301-307. Manji F, Fejerskov O, Baelum V, Luan W-M, Chen X: The epidemiological features of dental caries in African and Chinese populations; implications for risk assessment. In: Johnson NW, ed. Risk markers for oral diseases. Vol. 1. Dental caries. Markers of high and low risk groups and individuals. Cambridge: Cambridge University Press, 1991;62-99. Nyvad B: Udvikling og evaluering af en screeningsmetode til identifikation af cariesaktive individer. MPH Thesis. Faculty of Health Sciences, rhus University, 2005 (in Danish). Nyvad B, Fejerskov O: Active root surface caries converted into inactive caries as a response to oral hygiene. Scand J Dent Res:1986;94:281-284. Nyvad B, Fejerskov O: Assessing the stage of caries lesion activity on the basis of clinical and microbiological examination. Community Dent Oral Epidemiol 1997;25:69-75.
115
NYVAD
ET AL.
Nyvad B, Heidmann: Caries decline revisited. Caries Res 1999;33:281 (abstract). Nyvad B, Heidmann J, Vaeth M: Caries predictioneffect of lesion activity assessment. Caries Res 2004;38:381 (abstract). Nyvad B, Heidmann J, Vaeth M: Caries lesion transitions in a low-caries population of adolescents. Caries Res 2005;39:293 (abstract). Nyvad B, Machiulskiene V, Baelum V: Reliability of a new caries diagnostic system differentiating between active and inactive caries lesions. Caries Res 1999;33:252-260. Nyvad B, Machiulskiene V, Baelum V: Construct and predictive validity of caries diagnostic criteria assessing lesion activity. J Dent Res 2003;82:117-122. Pitts NB, Fyffe HE: The effect of varying diagnostic thresholds upon clinical caries data for a low prevalence group. J Dent Res 1988;67:591-596. ten Cate JM, Featherstone JDB: Physicochemical aspects of fluoride-enamel interactions. In: Fejerskov O, Ekstrand J, Burt BA, eds. Fluoride in dentistry. Copenhagen, Munksgaard, 1996, 252-272. Thylstrup A, Bruun C, Holmen L: In vivo caries modelsmechanisms for caries initiation and arrestment. Adv Dent Res 1994;8:144-157. WHO: Oral Health Surveys: Basic methods, ed 4. Geneva, World Health Organization, 1997. Wulff HR: What is understood by a disease entity? J R Coll Physicians Lond 1979;13:219-220.
116
The Chesters Model Using the Simplified DSTM

Richard K. Chesters1,4, I. Balciuniene3, G. Matuliene3, N.B. Pitts2, J.R. Matheson1
1
Unilever Dental Research, Port Sunlight, Bebington, UK University of Dundee, Dundee, Scotland Vilnius University, Vilnius, Lithuania Colgate Palmolive Europe, Geneva, Switzerland
Abstract The Chesters Model for caries clinical trials was developed in response to the need for more efficient ways of assessing the anti-caries activity of oral care products, including dentifrices. This paper reviews the principles and rationale behind the model, its use of the clinical visual diagnostic approach as developed in the Dundee Selectable Threshold Method, and summarises the validation clinical trial. The principles behind the approach include: no compromise to the standard of proof of efficacy, the method must use natural human teeth tested in vivo, and finally, the model must give results compatible with those obtained in conventional caries clinical trials for regimens previously shown to have differing anti-caries activity. The rationale behind the approach was that it should take account of all clinically visually detectable aspects of the caries process, including lesion development, progression, arrest, and reversal. Consequently, the clinical examination must assess enamel lesions (including both cavitated and non-cavitated lesions) and dentinal caries lesions. A randomised, double-blind, 2-year validation trial was designed to confirm that the Chesters Model was able to demonstrate differences in a 12-month period in anti-caries activity between 2 fluoride dentifrices (1000 and 2500 ppm F) with proven differing efficacy. The study population at the outset comprised 2,387 Lithuanian schoolchildren aged 1114 (mean 13.1) years. Caries was assessed via clinical visual assessment using a simplified version of the Dundee Selectable Threshold Method (DSTM) at the D1 (including enamel lesions) and
CHESTERS
ET AL.
D3 (excluding enamel lesions) levels, fibre optic transillumination (FOTI), and digital bitewing radiography. A conventional caries increment analysis and an events analysis, incorporating caries lesion initiation, progression, and regression were performed. Product discriminatory ability after 12 months was assessed against the conventional 24-month caries increment. A total of 2,141 and 2,011 subjects were re-examined using DSTM at the 12- and 24-month examinations, respectively. The trial showed that, after 12 months, it was possible to distinguish products use between 2 dentifrices of differing anti-caries efficacy using only clinical visual assessment at the D1 threshold; whereas, with the conventional approach, this took 24-months. In conclusion, the Vilnius clinical trial supports the validity of the Chesters Model and similar models that determine anti-caries efficacy by measuring changes in the caries continuum detected by clinical visual assessment of all caries lesions, including non-cavitated enamel caries lesions. The Chesters Model allows the assessment of anti-caries efficacy to be made in a shorter time period than the conventional caries clinical trial using a dentine (usually cavitation) threshold. Finally, the approach has a number of advantages, including that it is simple, robust, and in line with contemporary philosophy of care and disease management.
Introduction Conventional caries clinical trials [Stephen et al., 1988; National Institutes of Health, 1991] have traditionally been performed at the dentine diagnostic threshold (D3) [e.g., Radike, 1972] and extend over periods of 2 to 4 years [Clarkson et al., 1993; Stookey et al., 1993; Kingman & Selwitz, 1997; Ismail, 1997]. In recent years, the conduct and interpretation of such trials has been complicated by the decline in the prevalence and rate of progression of caries and the increase in levels of fluoride exposure [Carlos & Wolfe, 1988; Newbrun, 1992; Marthaler et al., 1996; Vehkalati et al., 1997]. Against this background, the number of subjects required for the detection of statistically significant differences between anti-caries products has increased substantially. As a result, the cost of conducting such efficacy trials is becoming ever more prohibitive.
118
CHESTERS MODEL
Caries clinical trial design needs to be reviewed to keep pace with our improving knowledge of caries as a continuum. Recent improvements in caries diagnostic techniques, coupled with the use of visual criteria employing the D1 (enamel and dentine) diagnostic threshold, which covers all demineralisation stages clinically detectable by the naked eye, offer the prospect of more sensitive measurement of early caries lesions [Pitts & Fyffe, 1988; Pitts, 1992; Angmar-Mnsson, 2001; Reich, 2001]. More importantly, the detection of noncavitated lesions is crucial in order to prevent the progression of dental caries before the occurrence of cavitation [Kingman and Selwitz, 1997; Ismail, 1997; Pitts, 1997]. Assessment at the D3 level fails to take into consideration the earlystage disease processes that precede dentinal involvement. In the past, shortened (~12-month) protocols have been successfully employed in caries clinical trials in populations with high incidences of caries [Muhler et al., 1955a; Muhler et al., 1955b; Goaz et al., 1963; Muhler, 1970; Lu et al., 1980; Lu et al., 1985]. However, these protocols have not been widely adopted, partly because conventional trials conducted at the D3 threshold were believed to yield more clinically reliable results [Stookey et al., 1993]. Evidence provided from studies using the D1 diagnostic threshold would, nevertheless, suggest that this method offers acceptable reproducibility [Pitts & Fyffe, 1988; Deery et al., 2000]. Measurement at the D1 diagnostic threshold also aligns caries clinical trials more closely with the contemporary philosophy of care and disease management [Pitts, 2004]. It is thought that use of the full clinically observable caries continuum could also potentially improve the efficiency of caries clinical trials, allowing efficacy to be demonstrated over a shorter time period and possibly with a smaller sample size. This paper reviews the principles and rationale behind the model, its use of the clinical visual diagnostic approach as developed in the Dundee Selectable Threshold Method (DSTM), and summarises the validation clinical trial previously reported by Chesters et al. [2002]. Attention has been taken to ensure that the paper follows the CONSORT Statement recommendations for improving the quality of reports on parallel-group randomised trials [Moher et al., 2001]. Basic Principles The basic principles for any new method to be used to assess anti-caries efficacy were reported at the International Consensus Conference on Caries
CHESTERS
ET AL.
Clinical Trials [Chesters et al., 2004]. These principles are in line with the philosophy behind the Chesters model and are set out below: Any novel design must not compromise the standard of proof of efficacy. Proof of clinical efficacy of caries-preventive products must be conducted in vivo using natural human teeth. Validation of any new trial design must involve comparisons with regimens previously shown in conventional caries clinical trials to have different anti-caries efficacy. Rationale Behind Chesters Model The rationale was that the design should take into account all of the visually detectable aspects of the caries process, since caries is a dynamic process marked not only by cycles of active disease and arrest of lesion progression, but also, in its early (pre-cavitation) stage, by remineralisation and repair [Marthaler, 1984; Mellberg, 1988; Featherstone, 2000; Fyffe et al., 2000a; Fyffe et al., 2000b]. Clearly, traditional caries assessment at the D3 level fails to detect the early disease processes that precede dentinal cavitation [Ismail 1997; Pitts, 1997]. Therefore, the Chesters Model included the measurement of caries lesion development at the enamel D1 level, including white spot and brown spot enamel caries lesions. This allowed the assessment of reversals and transitions [Pitts, 1985] between caries states of different lesion severity to be assessed. Vilnius Validation Caries Clinical Trial Trial Objectives The objectives of the Vilnius validation caries clinical trial were to: 1. Demonstrate a statistically significant difference in the conventional 24-month caries increment (CVA at the D3 threshold + Fibre Optic Transillumination + radiography) and, hence, provide a benchmark measure of efficacy for assessing the discriminatory ability of other caries diagnostic techniques. 2.Determine whether a statistically significant difference in efficacy could be demonstrated between the standard (1000 ppm F) and the high-dose (2500 ppm F) fluoride dentifrice after 12 months product use.
CHESTERS MODEL
Pre-study Preparation A key element in the successful execution of any clinical trial is preparation prior to the start of the study. This can involve the conduct of cross-sectional epidemiology surveys at potential locations. Such survey data are essential for estimating the caries increment for a specific population. The estimated increment is then used, together with an estimate of the variance, to determine the cell size necessary to give the appropriate power. Another essential pre-study element, following the location of the trial site and the test population, is to train potential clinical examiners in the clinical indices to be used in the trial. In the case of the Vilnius validation trial, training was conducted by an experienced examiner over a period of 4 days, at the end of which the clinical examiners were selected based on factors including their reproducibility using the clinical indices. Trial Population The trial was conducted among school children (both boys and girls) aged 1114 years from 28 of the larger schools in Vilnius, Lithuania. These larger schools were selected simply for reasons of logistics. For inclusion in the trial, subjects were required to have at least 1 erupted second permanent molar and a clinical visual D3MFS score of between 2 and 24. Subjects were excluded from participation in the trial if their parent/guardian was unwilling to disclose the childs medical and dental history, if they did not brush their teeth regularly, if they had received an intraoral x-ray for diagnosis of caries within the previous 6 months, or if they presented with more than 2 occlusal surfaces of second permanent molars that were restored and/or clinically cavitated and/or sealed. In addition, subjects were excluded if they displayed unacceptable levels of stress during the initial oral screening examination; if they had existing medical or dental conditions or were receiving treatments that were likely to affect caries development (e.g., long-term antibiotic therapy); if they had a heart condition or were receiving cancer treatment; or if they presented with a fixed orthodontic appliance that made it impractical to assess all surfaces of erupted teeth.
121
CHESTERS
ET AL.
Trial Design and Methodology The Vilnius validation trial adopted a single-centre, double-blind, parallelgroup design in keeping with established practice in caries clinical trials comparing dentifrice efficacies [Edlund and Koch, 1977; De Paola et al., 1993]. The trial was conducted between January 1999 and May 2001. Subjects underwent dental screening, which included an assessment of supra-gingival calculus (presence or absence on lingual surfaces of mandibular anterior teeth) and caries prevalence at the D3 dentine threshold. Subjects also completed an oral habits questionnaire, and those meeting the trial selection criteria were then subjected to a more detailed baseline dental examination before being assigned by stratified random allocation to 1 of 2 silica-based dentifrices containing either 0.76% sodium monofluorophosphate (equivalent to 1000 ppm F) or 1.90% sodium monofluorophosphate (equivalent to 2500 ppm F). These 2 fluoride toothpastes were chosen, because their anti-caries efficacy has been shown to be significantly different on 2 occasions in conventional caries clinical trials [Stephen et al., 1988; Marks et al., 1994]. Each dentifrice (1000 and 2500 ppm F) was assigned 2 codes, giving 4 unique alphanumeric product codes. The products were identical except for fluoride level and different coloured packaging for each product code. The subjects, clinical examiners, and those administering the test products were not aware of the product identities at any time during the trial. The investigators were supplied with sealed, opaque, code break envelopes that could have been opened in an emergency. This was not required, and the integrity of the product code was confirmed with regular GCP monitoring and independent audit. Subjects were allocated to strata automatically using an SAS program generated for this purpose. Stratification factors were gender, D1MFS score (2-16, 17-26, >26), and second molars with occlusal surfaces still at risk (low = 1 or 2 surfaces classified visually as sound or with potentially reversible caries lesions; high = 3 or 4 such surfaces). Subjects were then allocated to a product group using a pre-prepared computer generated list of randomised blocks (2 x 4 product codes per block; with preparation date used as the random seed; and list generated by Trial Statistician) for each stratum on an ongoing basis by a member of the trial team not associated with the clinical assessments. This was carried out in the Site Office away from any of the school trial sites. Siblings were assigned the same product in order to ensure that only 1 product was supplied to each household. Subjects were provided with a free supply of dentifrice and toothbrushes during the trial and were instructed to
CHESTERS MODEL
brush their teeth twice daily and to attend daily school toothbrushing sessions (involving brushing for 1 minute with the assigned dentifrice once per day) during term-time. Every effort was made to re-examine subjects accepted into the trial after 12 and 24 months of dentifrice use. The oral hygiene habits of individuals were also collected using questionnaires at the 12- and 24-month examinations. During the course of the trial, subjects were instructed to use only their assigned dentifrice and to avoid fluoride-containing mouthwashes and other fluoride supplements, such as fluoride tablets or fluoridated salt. There was no interference with the dental treatment of subjects in the trial; however, subjects who were wearing fixed orthodontic appliances were not examined. Compliance was checked on an ongoing basis by school toothbrushing supervisors, and subjects were also questioned at the 12- and 24-month examinations to verify that they were brushing their teeth at home at least once daily. At the baseline, 12-month and 24-month examinations, caries status was assessed using the following methods: Clinical Visual Assessment (CVA) using a simplified version of the Dundee Selectable Threshold Method for caries detection and measurement adapted for use in clinical trials (Table 1.1), Fibre Optic Transillumination (FOTI) (Table 1.2), and bitewing radiography (Table 1.3). A single investigator conducted the CVA and FOTI examinations, and a second investigator assessed the bitewing radiographs. Permanent teeth were assessed visually with a mirror under standard conditionsi.e., with the subject in the prone position, using good lighting on clean, dry teeth (compressed air was used to dry the teeth). The approximal and occlusal surfaces of posterior teeth were then examined using FOTI [Mitropoulos, 1985]. Bitewing radiographs were taken of the approximal and occlusal surfaces of posterior teeth from the distal surface of the second permanent molar to the mesial surface of the first premolar using the DEXIS digital system and the Oralix Gendex AC xray tube (64 kV; typically 0.120.14 seconds). The Dexis film holder was used to standardise projection geometry. These digital images were scored radiographically using a monitor display with contrast and brightness adjustment chosen by the assessor for each image. The images were coded using criteria compatible with transition matrices (Table 1.3) [Hollender and Koch, 1969; Pitts, 1985; Kingman and Selwitz, 1997]. Repeat CVA and FOTI examinations were performed during the baseline, 12- and 24-month examinations in 5% to 10% of subjects in order to determine
CHESTERS
ET AL.
General Tooth Codes Code
Diagnostic Criteria
U X Y Z R Caries Surface Codes Code
Unerupted or missing Extracted due to caries involvement Missing due to orthodontic treatment Missing due to trauma or accident Retained roots
Diagnostic Criteria
S W K E H C P F T A $
Diagnostic Criteria Sound/partially erupted White spot caries Brown spot caries Cavity restricted to enamel Non-cavitated lesion into dentin Cavity into dentin Cavity into pulp Filled Traumatised Unreadable surface Sealed
TABLE 1.1
Diagnostic codes and categories for the simplified DSTM system [Fyffe et al., 2000a].
Caries Surface Codes Code
Diagnostic Criteria
0 1 2 3
No evidence of caries Evidence of a caries lesion possibly restricted to enamel Strong shadow probably involving dentin Surface not scored with FOTI (a)
Note (a) This code was first introduced at 12-month examinations.
TABLE 1.2
Diagnostic codes and categories for Fibre Optic Transillumination [OMullane et al., 1997].
124
CHESTERS MODEL
General Tooth Codes Code Diagnostic Criteria
U 8
Tooth present but below the oral mucosa Tooth absent from radiograph Caries surface codes for occlusal surfaces Code Diagnostic Criteria 0 No radiolucency or restoration visible
Caries surface codes for approximal surfaces Code Diagnostic Criteria 0 1 No radiolucency or restoration visible Zone of increased radiolucency confined to outer 1/2 of enamel Zone of increased radiolucency involving both inner and outer 1/2s of enamel including lesions extending up to but not beyond the ADJ Zone of radiolucency penetrating enamel but confined to the outer 1/2 of dentine Zone of radiolucency penetrating into inner
1/2 of dentine with or without apparent pulpal
Zone of increased radiolucency confined to enamel including lesions extending up to but not beyond the ADJ Zone of radiolucency penetrating enamel but confined to the outer 1/2 of dentine Zone of radiolucency penetrating into inner 1/2 of dentine with or without apparent pulpal involvement
involvement 5 Overlapped surface. Overlap of more than 1/2 of enamel but not beyond the ADJ. No zone of increased radiolucency in dentine Zone of increased radiolucency associated with a filled surface Radiographic appearance consistent with a restoration Unerupted, extracted or missing from radiographic image Overlap of less than 1/2 the enamel, zone of increased radiolucency in inner 1/2 of enamel including lesions extending up to but not behond the ADJ Partial overlap if less than 1/2 enamel thickness, no zone of radiolucency 6
Zone of increased radiolucency associated with a filled surface Radiographic appearance consistent with a restoration Unerupted, extracted or missing from radographic image
TABLE 1.3
Diagnostic codes and categories for Bitewing Radiography [Pitts, 1984].
125
CHESTERS
ET AL.
the intra-examiner reliability of caries measurement. For radiography, the baseline and 12- and 24-month radiographs were re-assessed for 5% to 10% of subjects. In addition, the trial evaluated the potential of several new methods, including Electrical Caries Monitor and Laser Fluorescence Device at these time points; however, the results for these methods are not reported in this paper. Statistical Aspects A recruitment target of 2,000-3,000 subjects (i.e., 1,000-1,500 per dentifrice group) was accepted for the Vilnius validation caries clinical trial using the data collected in a cross-sectional prestudy survey. The bottom end of this range was based upon the following assumptions: (i) a coefficient of variation of 0.9 and an inter-group difference of 10% for the D1MFS incremental data at 12 months, (ii) a coefficient of variation of 1.0 and an inter-group difference of 12.5% for the D3MFS incremental data at 24 months, (iii) an annual attrition rate of 10%, and (iv) a trial power of 80% at the 5% significance level (1-sided test). The conventional caries increment [OMullane, 1997] used data from CVA and FOTI at the D3 threshold and bitewing radiography (all radiolucencies) to classify each surface as sound or decayed/missing/filled at each examination using the most severe score to define the status of the surface at that time point. The increment for each subject is calculated by summing the changes (& +) for all surfaces at the relevant examinations. A mean increment score for each dentifrice group was then calculated using the individual subject increments. The statistical assessment of the effect of fluoride on caries development was primarily performed using an analysis of variance procedure on the mean increment at the relevant examinations. Paired comparison between the mean group event score for the 2500 and the 1000 ppm F dentifrice groups was performed using the z-test, with the residual standard deviation of the analysis of variance and applying two-tailed criterion Another analysis was carried out that takes account of caries lesion progression, lesion initiation, and reversal in an event analysis, where an event was defined as a clinically observable transition in the caries status of a surface. The Events Analysis used transition matrices to record caries initiation, progression, and regression relative to baseline, as previously proposed by Pitts [1985] and later by Kingman and Selwitz [1997].
CHESTERS MODEL
Transition Matrix CTR2 lesion penetration transition score Basic principle: S(0) W/K/E (1) H/C/F (2) P/X/R (3) U S W K H E C P /*U*/ zero zero plus plus plus plus plus plus /*S*/ void zero plus plus plus plus plus plus /*W*/ void mnus zero zero plus zero plus plus /*K*/ void mnus zero zero plus zero plus plus /*H*/ void mnus mnus mnus zero mnus zero plus /*E*/ void void void void void zero plus plus /*C*/ void void void void void void zero plus /*P*/ void void void void void void void zero /*F*/ void void void void void void void void /*T*/ lost lost lost lost lost lost lost lost /*A*/ lost lost lost lost lost lost lost lost /*$*/ lost lost lost lost lost lost lost lost /*X*/ void void void void void void void void /*R*/ void void void void void void void void /*Y*/ lost lost lost lost lost lost lost lost /*Z*/ lost lost lost lost lost lost lost lost
F plus plus plus plus zero plus zero zero zero lost lost lost void void lost lost
T lost lost lost lost lost lost lost lost lost lost lost lost vdls vdls lost lost
A lost lost lost lost lost lost lost lost lost lost lost lost vdls vdls lost lost
$ lost lost lost lost lost lost lost lost lost lost lost lost vdls vdls lost lost
X plus plus plus plus plus plus plus zero zero lost lost lost zero zero lost lost
R plus plus plus plus plus plus plus zero zero lost lost lost void zero lost lost
Y lost lost lost lost lost lost lost lost lost lost lost lost lost lost lost lost
Z lost lost lost lost lost lost lost lost lost lost lost lost lost lost lost lost
TABLE 2.1
Clinical visual assessment scoresLesion depth transition matrix.
0 /*0*/ /*1*/ /*2*/ zero mnus mnus
1 plus zero mnus
2 plus plus zero
3 void void void
TABLE 2.2
FOTI transition matrix.
Two transition matrices for the CVA scores were created prior to the start of the analyses, with 1 being based on lesion depth (see Table 2.1) and the second on surface continuity, which is not reported here. Transitions could be: a) initiation/progression towards a more severe caries surface state (positive values), or b) regression towards a less severe caries surface state (negative values), or c) remain unchanged (zero value). Additionally, transitions that were declared clinically impossible were considered attributable to the clinician or recorder error (e.g., transitions from a cavity into dentine to white spot or filled to sound). These void transitions were to be excluded from the prime analysis, provided that there was not a statistically significant difference between their incidence in the 2 treatment groups, as indicated by the chi-square test. Similar transition matrices were created for FOTI (Table 2.2) and for radiography (Table 2.3)
127
CHESTERS
ET AL.
Approximal Surfaces U 0 /*U*/ zero free /*0*/ void free /*1*/ void mnus /*2*/ void mnus /*3*/ void mnus /*4*/ void mnus /*5*/ void free /*6*/ void void /*7*/ void void /*8*/ 1ost lost /*9*/ void mnus /*A*/ void zero Occlusal Surfaces U /*U*/ zero /*0*/ void /*2*/ void /*3*/ void /*4*/ void /*6*/ void /*7*/ void /*8*/ lost
1 plus plus zero mnus mnus mnus zero void void lost mnus zero
2 plus plus plus zero mnus mnus zero void void lost zero plus
3 plus plus plus plus zero mnus plus zero zero lost plus plus
4 plus plus plus plus plus zero plus zero zero lost plus plus
5 free free zero zero mnus mnus zero void void lost zero zero
6 plus plus plus plus zero zero plus zero zero lost plus plus
7 plus plus plus plus zero zero plus zero zero lost plus plus
8 lost lost lost lost lost lost lost lost lost lost lost lost
9 plus plus plus zero mnus mnus zero zero zero lost zero plus
A free free zero mnus mnus mnus zero void void lost mnus zero
0 free free mnus mnus mnus void void lost
1 void void void void void void void void
2 plus plus zero mnus mnus void void lost
3 plus plus plus zero mnus zero zero lost
4 plus plus plus plus zero zero zero lost
6 plus plus plus zero zero zero zero lost
7 plus plus plus zero zero zero zero lost
8 lost lost lost lost lost lost lost lost
A void void void void void void void void
lost refers to those transitions where at either or both exams surface not scoreable. mnus refers to those transitions where caries regression observed. plus refers to those transitions where caries progression observed. void refers to those transitions deemed invalid. zero refers to those transitions where no change occurred.
TABLE 2.3
Radiography transition matrix.
Ethical and Safety Aspects The Vilnius Caries Clinical Trial was conducted in accordance with the Declaration of Helsinki [1964] and its subsequent amendments [Council for International Organisations of Medical Sciences, 1993]. The trial adhered to the standards of good clinical practice. Approval to conduct the trial was obtained from the Lithuanian National Committee on Biomedical Ethics, the Ministry of Public Education and Science, and the head teachers at the participating schools. Prior to entry into the trial, written informed consent was obtained from each subject and his/her parent or guardian. Subjects were free to withdraw from the trial of their own volition at any time. Where known, the investigators documented the reason for withdrawal. In
128
CHESTERS MODEL
addition, the investigators had the discretion to withdraw subjects because of serious adverse events or non-compliance with the trial protocol. Any adverse events arising during the trial were followed-up by the investigators, and a safety evaluation describing the adverse event and its likely cause was submitted to the trial sponsor. As part of the safety assessment, the oral soft tissues were assessed at each annual examination for any signs of oral pathology. Overall responsibility for the trial rested with the sponsor, Unilever Dental Research. However, the responsibility for supervising the fieldwork in Lithuania, distributing the test products to the brushing supervisors, and monitoring the various trial centres was delegated to Scope International, Mannheim, Germany, following training of their staff by Unilever Dental Research personnel. Interpretation of the data was the responsibility of the sponsor Results Trial Demographics and Attrition The Vilnius caries clinical trial was completed on schedule in May 2001 and a summary of trial demographics as a flow diagram is shown in Figure 1. The number of subjects examined using CVA, FOTI, and radiography are summarised in Table 3.1. Of note, subjects who, during the course of the trial, received an oral x-ray within the 6-month period preceding follow-up, were not subjected, as per protocol, to repeat radiographic examination; this accounts for the lower number of subjects examined radiographically at 12 and 24 months. The 1- and 2-year attrition rates at 8% and 14%, respectively, were well within the 10% and 20% rates that were allowed for in the protocol. Table 3.2 shows a breakdown of the reasons for the attrition at the end of the trial. No data were excluded due to non-compliance with the trial protocol. Adverse Events
Diagnostic Method DSTM FOTI Radiography Baseline 2387 2387 2387 12 Months 2141* 2141* 1857 24 Months 2011 2011 1793
*1072 and 1069 subjects in the 1000 and 2500ppm F groups, respectively. 994 and 1017 subjects in the 1000 and 2500ppm F groups, respectively.
TABLE 3.1
Summary of the number of subjects examined by each diagnostic method. 129
CHESTERS
ET AL.
Invited to participate (n=8207) Responded (n = 4283)
Assessed for eligibility (n=3924) January - May 1999
Excluded Failed to attend screening (n=359) Not meeting inclusion criteria (n=1537 )
Randomised (n= 2387)
Allocation to 1000 pp F group (n = 1193; male:female 528 : 665) Received toothpaste (n = 1193) Did not receive toothpaste (n = 0)
Baseline examinations
Allocation to 2500 pp F group (n = 1194; male:female 529 : 665) Received toothpaste (n = 1194) Did not receive toothpaste (n = 0)
1000 ppm F toothpaste group
Assessed between Jan. & May 2000 (n=1072; male:female 471 : 601) Mean baseline age 13.0 years Absent (n=22) Withdrawn (n = 99) No. included in main analysis (n=1072)
Assessed between Jan. & May 2000 (n=1069; male:female 468 : 601) Mean baseline age 13.1 years Absent (n=27) Withdrawn (n = 98) No. included in main analysis (n=1069)
Assessed between Jan. & May 2001 (n = 994; male:female 431 : 563) Mean baseline age 13.0 years Absent (n=24) Total withdrawn (n=175) No. included in main analysis (n = 994)
24-month examinations Absent n= 41
Assessed between Jan. & May 2001 (n = 1017; male:female 439 : 578) Mean baseline age 13.1 years Absent (n=17) Total withdrawn (n=160) No. included in main analysis (n =1017)
Note: Main analysis defined as the analysis of the clinical visual assessment data.
FIG. 1:
Fibre optic confocal microscope.
Overall attrition rate after 24 months Did not wish to continue on trial Moved school Adverse events Protocol violation or non-compliance Excluded, because did not meet continuation criteria
14.3% 7.4% 2.6% 2.0% 1.8% 0.5%
TABLE 3.2.
Breakdown of causes of subject attrition at the end of 24-month trial.
130
2500 ppm F toothpaste group
12-month examinations Absent n=49
CHESTERS MODEL
None of the 303 serious adverse events (as defined by Good Clinical Practice) reported in the trial were considered to be product-related. However, 26 minor adverse events, including tooth hypersensitivity (n = 3), sore gums (n = 3), and gingivitis (n = 2) were reported as possibly product-related, with equal numbers in each product group. There was no significant difference in the incidence of adverse events between the 1000 ppm and the 2500 ppm F toothpaste groups. Reproducibility For baseline measurements, repeat data were obtained from 194 subjects (8%) for CVA/FOTI assessment and 226 subjects (9.5%) for radiographic assessment. For 12-month measurements, repeat data were obtained from 140 (6.5%) subjects for CVA/FOTI and 154 (8.3%) subjects for radiography. The corresponding numbers for 24-month measurements were 124 (6.2%) and 155 (8.6%) subjects. Kappa values for baseline, 12-month, and 24-month measurements at both the D1 and D3 thresholds are presented in Table 4. Baseline Caries Balance At baseline, the D1MFS prevalence (mean SE) in the 12-month examination population was slightly higher for the 1000 ppm F group than for the 2500 ppm F group (see Table 5), although this difference was not statistically significant. Similarly, at baseline, the D1MFS prevalence in the 24-month examination population tended to be slightly higher for the 1000 ppm F group than for the 2500 ppm F group (Table 5). There were no statistically significant differences between the 2 product groups with regard to D3MFS prevalence at baseline for the 12-month or 24-month populations (Table 5). Similarly, there was no statistically significant difference for the baseline FOTI and radiography prevalence scores for the 12- or 24-month populations (Table 6).
Method Threshold Baseline 12 months 24 months
DSTM D1 0.840 0.855 0.840
DSTM D3 0.941 0.957 0.952
FOTI D1 0.850 0.902 0.881
FOTI D3 0.909 0.964 0.938
Radiography D1 0.821 0.849 0.845
Radiography D3 0.826 0.851 0.858
TABLE 4.
Reproducibility results: Kappa values.
131
CHESTERS
ET AL.
Mean D1MFS (Standard Error) 1000 ppm F group 2500 ppm F group 32.89 (0.40) 31.99 (0.37) 32.95 (0.42) 32.01 (0.37) Mean D3MFS (Standard Error) 1000 ppm F group 2500 ppm F group 9.04 (0.19) 9.00 (0.18) 9.12 (0.20) 8.99 (0.18)
12-m population 24-m population
TABLE 5.
Baseline CVA (DSTM) assessment scores for 12- and 24-month study populations.
12-m population 24-m population
Mean number lesions (DS) detected with FOTI (Standard Error) 1000 ppm F group 2500 ppm F group 13.10 (0.16) 12.83 (0.16) 13.07 (0.17) 12.81 (0.16)
Mean number lesions (DS) detected radiographically (Standard Error) 1000 ppm F group 2500 ppm F group 19.47 (0.19) 19.06 (0.19) 19.45 (0.20) 18.98 (0.20)
TABLE 6.
Baseline DS scores for FOTI and radiography for 12- and 24-month study populations.
Conventional Increment Analysis Results from increment analyses for the simplified CVA at 12- and 24months are shown in Table 7. As expected, there was no statistically significant difference between the 2 product groups after 12 months using the D3 increment analysis of CVA + FOTI + radiography. As predicted in the design of the trial, a significant product effect (p<0.03) was observed after 24 months, using this benchmark efficacy assessment. Similar results were obtained for the increment analysis on the combination of CVA and radiography data. In contrast, CVA, using DSTM alone, showed a statistically significant product effect at both 12 and 24 months using D1 increment analysis. Events Analysis for the Simplified DSTM The primary results from the events analyses based on lesion depth transition matrices using a simplified version of the DSTM at 12 and 24 months are shown in Table 8. A statistically significant difference was demonstrated between the 2 dentifrices using DSTM at the D1 threshold after 12 months product use (p<0.005) (Table 8). No statistically significant difference was seen between the 2 product groups at the DSTM D3 threshold at this time, although the trend was the same as for D1 threshold results. Similarly, after 24 months of product use, a statistically significant difference was observed for the DSTM events at the D1 but not at the D3 threshold.
CHESTERS MODEL
Threshold DSTM at D1[1] DSTM at D2[2] DSTM and FOTI at D3 DSTM and FOTI at D3 + all radiographic lesions DSTM and all radiographic lesions
12 months
2500 F group mean increment score 1000 F group mean increment score p-value % effect
8.911 (0.200)
2.090 (0.079)
1.960 (0.080)
2.983 (0.159)
2.656 (0.111)
9.715 (0.216)
2.208 (0.085)
2.104 (0.081)
3.099 (0.160)
2.863 (0.114)
<0.0063 8.3 14.817
ns 5.4 3.253
ns 6.8 3.975
ns 3.7 4.955
ns 7.2 3.636
24-months
2500 F group mean increment score 1000 F group mean increment score p-value % effect
[1] [2] [3] [4] [5]
(0.247)
(0.107)
(0.113)
(0.158)
(0.128)
15.691 (0.271)
3.508 (0.120)
4.290 (0.125)
5.466 (0.161)
4.102 (0.134)
0.0165 5.6
ns 7.3
0.0613[3] 7.3
0.0233[4] 9.3
0.0119[5] 11.3
Including all carious lesions. Conventional increment including dentin cavities (W, K, and H treated as sound). Using conventional ANOVA adjusting for stratification variables then p<0.04. Using conventional ANOVA adjusting for stratification variables then p<0.03. Using conventional ANOVA adjusting for stratification variables then p<0.02.
TABLE 7.
Traditional 12- and 24 -month increment analyses on the simplified DSTM plus FOTI and the simplified DSTM + FOTI + radiography: mean increment score (standard error).
Events Analysis by Surface Type Events analysis by surface type showed statistically significant differences between the 2 dentifrice groups at 12 months for approximal and buccal/lingual
CHESTERS
ET AL.
Threshold 12-months 2500 F group mean event score 1000 F group mean event score p value * % effect
DSTM at D1[1]
DSTM at D3[2]
10.412 (0.212) 11.351 (0.228) 0.0026 8.3
2.090 (0.079) 2.208 (0.085) ns 5.4
24-months
2500 F group mean event score 1000 F group mean event score p value * % effect
Abbreviations: ns = not significant (p>0.1).
17.434 (0.273) 18.470 (0.294) 0.0098 5.6
3.253 (0.107) 3.508 (0.120) ns 7.3
* Significant inter-group difference, two-tailed t-test.

[1] [2]
including all D1STM lesions. conventional including dentin cavities, hence W, K, E, H treated as sound.
TABLE 8.
Events analyses for the 12- and 24-month examinations: mean event scores (standard error).
surfaces, but not for occlusal surfaces (Table 9). Discussion Conventional caries clinical trials last up to 4 years and demand co-operation from a considerable number of subjects [Stookey et al., 1993; Kingman and Selwitz, 1997]. These clinical trials are expensive in terms of manpower and materials and, therefore, call for maximum use to be made of available knowledge of the disease process and diagnostic techniques [Rugg-Gunn et al., 1975]. Recent improvements in detection and assessment methods, particularly in those applicable during the early stages of the caries process, potentially
134
CHESTERS MODEL
All Surfaces 12-months Approximal Surfaces Buccal/Lingual Durfaces Occlusal Surfaces
2500 F group mean event score (n = 1069) 1000 F group mean event score (n = 1072) p value * % effect 24-months
10.412 (0.212)
4.961 (0.140)
2.369 (0.091)
3.082 (0.070)
11.351 (0.228)
5.451 (0.147)
2.734 (0.101)
3.165 (0.072)
0.0026 8.3
0.0158 9.0
0.0073 13.4
ns 2.6
2500 F group mean event score (n = 1017) 1000 F group mean event score (n = 994) p value * % effect
17.434 (0.273)
8.782 (0.176)
4.055 (0.115)
4.597 (0.085)
18.470 (0.294)
9.366 (0.188)
4.472 (0.125)
4.632 (0.084)
0.0098 5.6
0.0235 6.2
0.0141 9.3
ns 0.8
Abbreviations: ns = not significant (p>0.1). * Significant inter-group difference, 2-tailed t-test.
TABLE 9.
DSTM 12- and 24-months events analysis by surface type: mean event score (standard error).
enable dentifrices to be differentiated in terms of their anticariogenic efficacy over a shorter time period. The Vilnius validation caries clinical trial set out to establish the validity of an abbreviated design, where the efficacy of the test products was already known. This approach overcomes any uncertainty that the absence of significant differences is due to the products rather than the study design. The trial assessed the validity of adopting such an abbreviated (12-month) protocol at a location (Vilnius, Lithuania) with a comparable background prevalence of caries (as
135
CHESTERS
ET AL.
measured at the D3 threshold) to that reported in a caries clinical trial conducted in western Europe in the 1980s [Stephen et al., 1988]. This study demonstrated that the reproducibility of the three diagnostic techniques was excellent with Kappa values of over 0.8 (Table 4) [Fleiss, 1981]. In the past, concern has been expressed about the reproducibility of caries diagnosis of non-cavitated enamel caries lesions. The study showed that although Kappa values decreased slightly (Table 4) on switching from the D3 to the D1 threshold, this small loss in precision is greatly outweighed by the increase in product discrimination at the D1 threshold. With the growing emphasis on caries prevention in dental practice, the ability to detect non-cavitated lesions at the D1 threshold has become increasingly important [Pitts, 1987; Alfano et al., 2001]. As previously reported in conventional caries clinical trials [Stookey et al., 1993], the traditional combination of CVA and FOTI at the D3 threshold, plus bitewing radiography, were not capable of differentiating between the 2 dentifrices (of proven differing anticariogenic efficacy) after 12 months use. In contrast, CVA using the simplified DSTM alone at the D1 threshold allowed statistically significant differences to be demonstrated between these products after 12 months, using either a conventional increment or an events analysis. This ability to differentiate products was found to be both surface-specific, with significant differences being demonstrated for approximal and buccal/lingual surfaces, but not for occlusal surfaces. The absence of discriminatory ability on occlusal surfaces is consistent with the results of previous studies [Chesters et al., 2002] and with data on the mode of action of water fluoridation [Kalsbeek et al., 1993]. Imagine that the caries continuum can be thought of in terms of distance, where sound is the starting point of a journey and white spots, brown spots, enamel cavities, and dentine cavities are all distances along a road. The speed of travel along this road will depend upon the intrinsic caries resistance of the surface. Different forms of transport can then be used to represent the different surface types. For example, a walker could be used to represent buccal and lingual; smooth, approximal surfaces by a cyclist, and occlusal surfaces by a motorcyclist. Clearly, in a fixed time, walkers, cyclists, and motorcyclists would be expected to travel different distances along the road. The different caries diagnostic methods can then be thought of as a device that measures whether the person has travelled more or less than a specific distance or breakpoint (fig. 2). The effect of the higher fluoride concentration can then be thought of
CHESTERS MODEL
D1 Threshold
D3 Threshold Buccal & lingual
2500
1000
Approximal
2500 1000
Occlusal
2500 1000
Increasing caries severity

FIG. 2: Rate of caries lesion progression.
as slowing the speed of the traveller along the road. The explanation for the absence of a significant effect on occlusal surfaces can then be explained as follows. Although fluoride is having an effect on all surface types, the effect will only become apparent if the mean distance travelled is near to the breakpoint of the diagnostic method. This situation occurs with both free smooth and approximal surfaces for DSTM data at the D1 threshold. However, for occlusal surfaces, the suggestion is that the mean distance travelled exceeds the breakpoint for D1 threshold, but it has not reached the point for the D3 cavitation threshold as measured by DSTM. Supplementation of the DSTM data with the FOTI plus radiographic findings failed to produce significant product differentiation at the D3 threshold after 12 months, but, as expected, resulted in statistically significant differences between the products at 24 months (Table 7). These results confirm the validity of the present trial as per protocol, and the observed effect is consistent with the findings of previous caries clinical trials testing 1000 and 2500 ppm F toothpastes [Stephen et al., 1988; Marks et al., 1994].
137
CHESTERS
ET AL.
Transition matrices based on the integrity of the tooth surface (with and without cavitation) were also tested in this trial. These were developed from those devised for monitoring the behaviour of approximal caries on radiographic images [Pitts, 1985]. The surface discontinuity matrix differed from the lesion depth matrix only in the relative severity of 2 caries surface codes. In the surface discontinuity matrix, a cavity restricted to enamel was classified as more severe than a non-cavitated dentine caries lesion; whereas, the opposite was the case for the lesion depth transition matrix. The surface discontinuity results are not reported here, since both transition matrices gave very similar results. This is because of the relatively low incidence of these 2 types of caries lesions in this population (cavitated enamel lesions = 1718 [0.7%] and non-cavitated lesions = 961 [0.4%]).
Conclusions The Chesters Model involves using all visually detectable caries lesions to improve the efficiency of caries clinical trials and, specifically, to allow anticaries product efficacy to be demonstrated over a shorter time period. The results of the Vilnius validation caries clinical trial show that it is possible to distinguish, after just 12 months products use, between 2 dentifrices of differing anticaries efficacy using only clinical visual assessment at the D1 (enamel and dentine) threshold; whereas, this took 24 months with the conventional D3 (dentine cavitation) approach. The Chesters Model and the conventional approaches gave similar results in terms of the observed product difference (8.3% & 9.3%, respectively). Finally, the Chesters Model proved to be both simple and robust and is in line with contemporary philosophy of care and disease management.
Acknowledgment We would like to thank Dr. C. Deery, formerly of the University of Dundee, for help in training the clinical and radiography sub-investigators; the team from Scope International, for supervising the fieldwork, distributing the test products to the brushing supervisors, and monitoring the trial; Paul Jones for preparation of the SAS software; John Cahill, for help in installing the com138 INDIANA CONFERENCE 2005
CHESTERS MODEL
puter systems in Lithuania; Egidijus Suslavicius, for his assistance in the initial setting up of the clinical trial, and all the support team in Lithuania.
References
Aleksejuniene J, Arneberg P, Eriksen HM: Caries prevalence and oral hygiene in Lithuanian children and adolescents. Acta Odontol Scand 1996;54:75-80. Alfano MC, Coulter ID, Gerety MB, Hart TC, Imrey PB, LeResche L, Levy J, Luepker RV, Lurie AG, Page RC, Rye LA, Smith L, Walker CB: Diagnosis and Management of Dental Caries throughout Life. J Am Dent Assoc 2001;132,1153-1161. Angmar-Mnsson B: How to measure the effects of fluoride treatments in clinical trials. Assessment: modern versus traditional methods. Caries Res 2001;35(Suppl. 1)30-33. Carlos JP, Wolfe MD: Dental caries: historic and current perspectives. Compend Contin Educ Dent 1988; 11 (Suppl.): S356-S364. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson JR, Nicholson JA, Gendvilyte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002;81:637-640. Chesters RK, Ellwood RP, Biesbrock AR, Smith SR: Potential modern alternative designs for caries clinical trials (CCTs) and how these can be validated against the conventional model. J Dent Res 2004 (Spec. Issue C):C122-124.) Clarkson JE, Ellwood RP, Chandler RE: A comprehensive summary of fluoride dentifrice caries clinical trials. Am J Dent 1993;6:S59-106. Council for International Organisations of Medical Sciences: Guidelines for biomedical research involving human subjects. ISBN 92 9036 0569 Geneva 1993. Deery C, Care R, Chesters R, Huntington E, Stelmachonoka, S, Gudkina Y: Prevalence of Dental Caries in Latvian 11- to 15-year old children and the enhanced diagnostic yield of temporary tooth separation, FOTI and Electronic Caries Measurement. Caries Res 2000;34:2-7. DePaola PF, Soparkar PM, Triol C, Volpe AR, Garcia L, Duffy J, Vaughan, B: The relative anticaries effectiveness of sodium monofluorophosphate and sodium fluoride as contained in currently available dentifrice formulations. Am J Dent 1993;6 Spec. Issue:S7S12. Edlund K, Koch G: Effect on caries of daily supervised toothbrushing with sodium monofluorophosphate and sodium fluoride dentifrices after 3 years. Scand J Dent Res 1977;85:4145. Featherstone JDB: The science and practice of caries prevention. J Am Dent Assoc 2000;13: 887-899. Fleiss JL: Statistical methods for rates and proportions. John Wiley and Sons 1981. Fyffe HE, Deery C, Nugent ZJ, Nuttall NM, Pitts NB: Effect of diagnostic threshold on the validity and reliability of epidemiological caries diagnosis using the Dundee Selectable Threshold Method for caries diagnosis (DSTM). Community Dent Oral Epidemiol 2000a;28:4251.
139
CHESTERS
ET AL.
Fyffe HE, Deery C, Nugent ZJ, Nuttall NM, Pitts NB: In vitro validity of the Dundee Selectable Threshold Method for caries diagnosis (DSTM). Community Dent Oral Epidemiol 2000b;28:5258. Goaz PW, McElwaine LP, Biswell HA, White WE: Effect of daily applications of sodium monofluorophosphate solution on caries rate in children. J Dent Res 1963;42:965972. Hollender L, Koch G: Influence of topical applications of fluoride on rate of progress of carious lesions in children. A long-term roentgenographic follow-up. Odontol Rev 1969;20:3741. Ismail AI: Clinical diagnosis of precavitated carious lesions. Community Dent Oral Epidemiol 1997;25:1323. Kalsbeek H, Kwant GW, Groeneveld A, Backer Dirks O, van Eck AAMJ, Theuns HM: Caries experience of 15-year-old children in The Netherlands after discontinuation of water fluoridation. Caries Res 1993;27:201205. Kingman A, Selwitz RH: Proposed methods for improving the efficacy of the DMFS index in assessing initiation and progression of dental caries. Community Dent Oral Epidemiol 1997;1:6068. Lu KH, Hanna JD, Peterson JK: Effect on dental caries of a stannous fluoride-calcium pyrophosphate dentifrice on an adult population; one-year results. Pharmacol Ther Dent 1980;5:1116. Lu KH, Yen DJC, Zacherl WA, Ruhlman CD, Sturzenberger OP, Lehnhoff RW: The effect of a fluoride dentifrice containing an anticalculus agent on dental caries in children. J Dent Child 1985;52:449451. Marks RG, Conti AJ, Moorhead JE, Cancro L, DAgostino R: Results from a three-year caries clinical trial comparing NaF and SMFP fluoride formulations. Int Dent J 1994;44:275285. Marthaler TM: Improvement of diagnostic methods in clinical caries trials. J Dent Res 1984;63:746750. Marthaler TM, OMullane DM, Vrbic V: The prevalence of dental caries in Europe 1990-1995. Caries Res 1996;30:237-255. Mellberg JR: Remineralization: a status report for the American Journal of Dentistry. Part II. Am J Dent 1988;1:85-89. Moher D, Schulz KF, Altman DG: The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomised trials. Lancet 2001;357:1191-1194. Muhler JC, Radike AW, Nebergall WH, Day HG: A comparison between the anticariogenic effects of dentifrices containing stannous fluoride and sodium fluoride. J Am Dent Assoc 1955a;51:556559. Muhler JC, Radike AW, Nebergall WH, Day HG: Effect of a stannous fluoride-containing dentifrice on caries reduction in children: II. Caries experience after one year. J Am Dent Assoc 1955b;50:163166. Muhler JC: A clinical comparison of fluoride and antienzyme dentifrices. J Dent Child 1970;37:501513. National Institutes of Health, National Institute of Dental Research: Oral health surveys of the National Institute of Dental Research, diagnostic criteria and procedures. Publication No. 91-2870, 1991. Newbrun E: Preventing dental caries: current and prospective strategies. J Am Dent Assoc 1992;123:68-73.
140
CHESTERS MODEL
OMullane DM, Kavanagh D, Ellwood RP, Chesters RK, Schafer F, Huntington E: A three-year clinical trial of a combination of trimetaphosphate and sodium fluoride in silica toothpastes. J Dent Res 1997;76:17761781. Pitts NB: Score system for monitoring the behaviour of radiologically diagnosed approximal carious lesions. Community Dent Oral Epidemiol 1985;13:268-272. Pitts NB, Fyffe HE: The effect of varying diagnostic thresholds upon clinical caries data for a low prevalence group. J Dent Res 1988;67:592-596. Pitts NB: The diagnosis of dental caries. 3. Rationale and overview of present and future techniques. Dental Update 1992;19:32-38. Pitts NB: Diagnostic tools and measurements-impact on appropriate care. Community Dent Oral Epidemiol 1997;25:24-35. Pitts NB. Are we ready to move from operative to non-operative/preventive treatment of dental caries in clinical practice? Caries Res 2004;38:294-304 Radike AW: Examiner error and reversals in diagnosis; in ADA Proceedings of the Conference on the Clinical Testing of Cariostatic Agents, American Dental Association, Chicago, USA, 1972. Reich E: How to measure the effects of fluoride treatments in clinical trials? The role of caries prevalence and caries assessment. Caries Res 2001;35(Suppl. 1):34-39. Rugg-Gunn AJ, Holloway PJ, Davies TGH: Partial recording in caries incremental studies in English schoolchildren. Community Dent Oral Epidemiol 1975;3:1119. Stephen KW, Creanor SL, Russell JI, Burchell CK, Huntington E, Downie CFA: A 3-year oral health doseresponse study of sodium monofluorophosphate dentifrices with and without zinc citrate: anti-caries results. Community Dent Oral Epidemiol 1988;16:321-325. Stookey GK, DePaola PF, Featherstone JDB, Fejerskov O, Mller IJ, Rotberg S, Stephen KW, Wefel JS: A critical review of the relative anticaries efficacy of sodium fluoride and sodium monofluorophosphate dentifrices. Caries Res 1993;27:337360. Vehkalati M, Tarkonkonen L, Varsjo S, Heikkila P: Decrease in and polarization of dental caries occurrence among child and youth populations, 1976-1993. Caries Res 1997;31:161-165.
141
142
The Biesbrock-Bartizek Caries Clinical Model

Aaron R. Biesbrock, and R.D. Bartizek Procter & Gamble Mason, OH, USA
Traditional Caries Models Over the past 40 years, large, randomized controlled clinical studies of 2 to 3 years duration have been accepted as the standard for proof of efficacy of anti-caries products. The clinical endpoint in the vast majority of these studies has been the presence or absence of caries on a tooth surface at the diagnostic threshold of a lesion with surface cavitation that extends into dentin, as detected by visual-tactile examination supplemented with radiographs. This model evolved from the 1968 conference proceedings of the American Dental Association-sponsored Conference on Clinical Testing of Cariostatic Agents [Radike, 1972]. The utility of this basic design as a tool to differentiate products based on cariostatic activity has been remarkably long-lived, as the current ADA guidelines for caries testing still recognize a conventional 2-year study as a necessity for proof of anti-caries efficacy. Current FDI guidelines also recognize conventional 2-year studies as the standard, but also discuss provisions for shorter studies [Reich, 1999]. It is important to recognize that the primary objective of 2- to 3-year randomized, controlled caries trials is to test a precisely framed hypothesis [OMullane, 1976]. In the case of caries studies, that hypothesis is usually based on the determination of an experimental products efficacy relative to a control product. The external validity of conventional 2- to 3-year caries clinical studies, with respect to its ability as a tool to differentiate cariostatic products, is well established, with fluoride dose response effects repeatedly demonstrated with different product forms, by different examiners, and in diverse populations throughout the world. In addition, the model design is robust enough to support subtle variations in study design. For example, variations in diagnostic examination method exist with an expanded visual examiCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 143
BIESBROCK
ET AL.
50 45 40 35 30 25 20 15 10 5 0 1960-69 1970-79 1980-89 1990-99
FIG. 1:
Number of industry-sponsored caries clinical trials: frequency by decade.
nation without the tactile component successfully employed to replace a visual tactile exam. Furthermore, fiber optic transillumination examination for diagnosis of interproximal surfaces has been successfully substituted to replace radiographic examination in some studies. However, the traditional design of caries clinical trials is not without limitations. It has increasingly been viewed as inefficient with respect to cost, resource utilization, and harmony with the underlying biology. These traditional clinical studies typically require very large sample sizes (n>1000/treatment group), which, in combination with the long study duration, result in very high study costs. The rising costs of caries clinical trials have, in fact, acted as a hindrance on caries innovation in industry. In one case study of industry sponsored caries clinical work, it is readily apparent that the number of product efficacy studies longer than 1 year in duration, conducted in each decade over the last 40 years, has dropped precipitously (fig. 1). Furthermore, there is a growing sentiment in the scientific community that more attention should be given to measuring the disease process as opposed to the terminal sequela of the disease process. The endpoint of a carious lesion with loss of enamel integrity (cavitation), which is most frequently used in traditional caries studies, focuses on one
THE BIESBROCK-BARTIZEK CARIES CLINICAL MODEL
end of the caries progression continuum at the expense of early caries initiation and progression. Early tooth surface demineralization can be reliably identified through methodical visual techniques at earlier stages in lesion development [Ekstrand, 2004]. These precavitated lesions are below the threshold of detection in traditional indices that are based on lesion cavitation as the point of differentiation between sound tooth surface and caries.
Sources of Variability in Caries Clinical Studies Traditional caries trials have generally required large sample sizes, often ranging from 500-2000 subjects per treatment group [Zacherl, 1981; Lu et al., 1987; Biesbrock et al., 2001; Bartizek et al., 2001; Marks et al., 1992; Hawley et al., 1995; Feller et al., 1996; Mann et al., 2001]. The size and duration of traditional caries clinical studies is, in part, driven by their uncontrolled nature. In many respects, conventional caries trials are quite similar in concept to large, simple trials, as subjects are generally seen at infrequent intervals, usually once a year and evaluated with fairly crude measurement techniques (presence or absence of cavitated tooth surface). Over the last 40 years, sample size requirements for these traditional caries clinical studies have become increasingly larger, driven by lower caries rates in the population and the need to differentiate less robust (10-20%) treatment differences. The changing epidemiology in the fluoride era has exacerbated this situation, as new clinical trials participants were even less likely to experience new lesions over a defined study period. A number of key factors drive variability in these types of traditional caries studies. Study subject populations are variable with respect to age, caries activity, and caries risk. Historically, factors such as age, baseline caries level, gender, and number of erupted permanent teeth at risk have been included in statistical analysis models [Grainger et al., 1984] as a way of controlling for the variability they introduce. Other factors, such as carbohydrate intake patterns, socioeconomic status, differential access to dental care, and previous disease experience (both individual and by siblings [Gerlach et al., 2001]), all are sources of population variation that typically are not well controlled. Product use in caries clinical studies is generally ad libitum in children and teens, leading to less than optimal compliance and product exposure. Examination of product usage patterns from historical studies supports the tenant that a minority of the population is fully compliant with product instructions, while a substantial
BIESBROCK
ET AL.
portion of the population is partially to wholly non-compliant. High dropout rates that are generally unrelated to product use are also common. An additional significant confounding factor is the level of uncontrolled background restorative dentistry that takes place over the course of the study. This restorative dentistry directly affects the observed caries score (DMFSDecayed, Missing, and Filled Surfaces) and may have a profound influence on perceived caries experience. In a large study of 3 years duration spanning multiple counties, it is not uncommon to have dozens of dentists delivering restorative care to subjects in the study. Usually, these multiple dentists have different backgrounds in educational training, treatment planning decision-making, and practice philosophy. In addition, examiner calibration and diagnostic threshold also can effect overall study variability. Collectively, such factors as population heterogeneity, variable product exposure, high dropout rates, and the uncontrolled placement of dental fillings by multiple dentists lead to large studies over long durations in order to compensate for these confounding factors
Development of a Novel Model: Approaches to Control Variability in Caries Clinical Study A novel clinical model derived from the traditional caries clinical trial standard was designed based on controlling for the most prominent sources of variability. It was hypothesized that, by controlling typically uncontrolled variables, product effects could be observed in shorter time periods with smaller sample sizes. Other published research suggests that, under the right conditions, anti-cavity treatments can be differentiated in short time periods. Product effects have been observed in as short as 12 months when comparing placebo and 1100 ppm F as sodium fluoride in studies with sample sizes of greater than 500 subjects per treatment group [Koch et al., 1990; Lu et al., 1985]. In addition, smaller treatment differences of 15-20% (250 ppm F vs 1100 ppm F) have been observed with relatively small sample sizes (N = 250/group) over 3-year treatment periods [Jensen and Kohout, 1988]. A number of controls were built into the new study design to provide greater model sensitivity in the differentiation of product efficacy. The first design element of this model was that the population should be relatively consistent with respect to age and disease presentation. A series of 1 dozen natural history caries epidemiology studies were conducted globally to help identify

Location Solola, Guatemala Coban, Guatemala Chiquimula, Guatemala Guatemala City Estanzuela, Guatemala Camuy/Cidra, Puerto Rico San Antonio, Texas Derry, Northern Ireland Limerick, Republic of Ireland
a b c
Na 230
6 years 2.71
7 years 4.39
8 years 5.35
9 years 6.38
10 years 8.93
11 years 11.76
12 years 15.73
13 years n/ab
227
3.00
4.97
5.14
8.45
9.17
11.85
16.52
n/ab
228
2.50
3.21
4.49
6.53
7.75
9.69
12.00
n/ab
231
3.27
2.98
5.45
5.74
8.71
8.97
11.26
n/ab
229
2.04
2.61
3.36
4.45
6.24
6.48
7.33
n/ab
268
1.68
2.60
2.93
4.31
5.00
6.10
7.29
n/ab
430
1.71
2.79
3.17
3.05
4.50
5.20
n.e.c
n/ab
103
n/ab
n/ab
n/ab
n/ab
n/ab
n/ab
8.09
9.97
108
n/ab
n/ab
n/ab
n/ab
n/ab
n/ab
4.89
4.18
Cumulative sample size across all ages examined. Subjects this age not included in the sample at this location. No reliable estimate for this age (n = 5 subjects in the sample).
TABLE 1.
Identification of caries prone populations: average DMFS scores by age of subjects.
caries prone populations throughout the world. Caries DMFS patterns were determined by examination of average DMFS scores for each 1-year increment in subject age (Table 1). Based on this data, populations with similar ranges of caries experience were selected for further research. Caries prone populations were identified based on age and geography. The second design element of this novel model was that product usage was supervised twice a day with standardized oral hygiene instruction in schools in addition to ad libitum product use outside the schools. Tooth brushing was supervised twice a day in the school setting by the teachers or by tooth brushing instructors during school hours. A controlled brushing pattern that included buccal, lingual, and occlusal surfaces was demonstrated at each session. Subjects brushed for a 1-minute duration. Outside of normal school hours, including evenings, weekends, and holidays, brushing was performed on an ad
BIESBROCK
ET AL.
libitum basis. This rigorous supervision was done to ensure compliance with the intended product usage instructions and to avoid the lack of compliance that often plagues traditional caries clinical trials. The third design element of this novel model was to control background access to care in order to reduce the variation with which background restorative dentistry was administered. Control was gained through selection of communities with limited access to care or selection of communities with generally controlled access to care (e.g., a limited number of dentists who trained at the same school in a similar timeframe). An additional mechanism to control background restorative dentistry in a clinical study would be to administer the restorative dentistry in a controlled fashion as part of the study. It has been proposed that new caries methods and models need to be validated relative to traditional caries clinical studies with respect to their ability to differentiate products that have previously been shown to have different efficacy profiles, such as fluoride dentifrice dose responses [Biesbrock et al., 2004; Chesters et al., 2004]. Three dose response model validation studies were conducted to examine the utility of a clinical design, which followed the tenants listed above. For each of the studies, a diagnostic threshold of D2 was employed, a lesion thought to be limited to enamel with tactile surface roughness or softness. Caries lesions were scored as either enamel or frank lesions, with an enamel lesion (D2) call based on visual evidence of demineralization accompanied by tactile surface softness, while a frank lesion (D3 and D4) call was based on a more pronounced loss of enamel integrity and the involvement of underlying dentin. In accordance, white spot lesions with no tactile component were not scored as caries. The lesions scored in this study correspond to D2 through D4, as reported by Pitts [2001] (Table 2).
Diagnostic Level Pre-D D1 D1 D3 D4 Clinical Description Sub-clinical lesions detectable only by additional diagnostic aids. Clinically detectable enamel lesion with intact surface (white spot). Clinically detectable enamel lesion with surface roughness/softness (incipient lesion). Clinically detectable lesion into enamel with detectable cavitation (frank lesion). Clinically detectable lesion into pulp (terminal lesion).
TABLE 2.
Diagnostic threshold in caries lesion diagnosis [adapted from Pitts, 2001]. INDIANA CONFERENCE 2005
148
Visual-tactile Exam Study 1 (Examiner AP) Repeatability weighted kappa 0.77
Radiographic Exam Sensitivity 98% Specificity 91% Sensitivity 99% Specificity 89% Sensitivity 100% Specificity 93%
Study 2 (Examiner JR)
Repeatability weighted kappa 0.96
Study 3 (Examiner RI)
Repeatability weighted kappa 0.90
TABLE 3.
Summary of examiner calibration for the 3 validation studies.
In addition, comprehensive examiner calibration of both visual-tactile examination and radiographic examination was completed for all examiners who participated in the 3-dose response validation studies. The examiners participated in 2 calibration sessions prior to the start of the clinical studies. The first calibration session was a visual-tactile repeatability exercise that involved 50 subjects between the ages of 8-12. At baseline, the subjects were examined by using a visual-tactile examination method performed with the aid of fiberoptic illumination, mouth mirror, compressed air, and a dental explorer employing the modified Radike criteria (D2 threshold). One day later, these exams were repeated. Repeatability was determined on a subject basis, using a weighted kappa analysis [Fleiss and Cohen, 1973]. The second calibration session was a radiographic diagnosis exercise based on a test set of radiographs from a previous caries study. Sets of 4 bitewing films for 20 subjects (ages 9-12) from both baseline and 1 year were randomized and blinded. This test set, which included 160 films, was evaluated by a panel of 3 expert clinicians for the presence of interproximal carious lesions. This panel of clinicians agreed on a consensus of 97 lesions (a total of 113 lesions were identified by at least 1 clinician). Each of the 3 examiners diagnosed interproximal caries on this test set of radiographs. Sensitivity and specificity by the examiners in this study were analyzed relative to the consensus findings (97 lesions) of the expert panel (Table 3).
Dose Response Validation Studies Study 1: A controlled fluoride (F) dose-response study was conducted in concurrence with a supervised school oral hygiene regimen to assess whether dentifrices with increasing fluoride levels could be differentiated with small
149
BIESBROCK
ET AL.
sample sizes in short timeframes [Biesbrock et al., 2003a]. The study was a randomized, double-blind study conducted for 21 months. Subjects (N = 657 with approximately 219 subjects per treatment group) were randomized to placebo dentifrice, 500 ppm F- dentifrice, or 1450 ppm F- dentifrice treatments for the first 9 months of the study. Subjects in the placebo group were then re-randomized to either 500 ppm F- or 1450 ppm F- dentifrice for the remainder of the study, while subjects in the fluoride groups continued with their original treatment assignments. A calibrated examiner measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 9 months, and 21 months for each subject. The mean caries increments at 9 months were 0.35 surfaces, 0.34 surfaces, and 1.28 surfaces for the 1450 ppm F-, 500 ppm F-, and placebo groups, respectively (Table 4). The mean caries increments at 21 months were 0.21 surfaces, 0.26 surfaces, 1.75 surfaces, and 1.90 surfaces for the 1450 ppm F-, 500 ppm F-, placebo/1450 ppm F-, and placebo/500 ppm F- groups, respectively. The 500 ppm F- and 1450 ppm F- fluA. 9-month increment. DMFS (Integrated) Treatment Placebo 500 ppm 1450 ppm N 210 198 199 Mean* 1.28 0.34 0.35
a
DMFS (Visual-Tactile) Mean* 0.87 0.41 0.40

a
SEM 0.23 0.24 0.24
% Reduction -------73.4% 72.7%
SEM 0.18 0.18 0.18
% Reduction ------52.9% 54.0%
*adjusted means. a p<0.05, compared to Placebo group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA).
B. 21-month increment. DMFS (Integrated) Treatment Placebo500 ppm Placebo14500 ppm 500 ppm 1450 ppm N 97 Mean* 1.90 SEM 0.40 % Reduction -------Mean* 0.44 DMFS (Visual-Tactile) SEM 0.31 % Reduction -------
88
1.75
0.43
7.9%
0.53
0.32
-20.5%
169 180
0.26a,b 0.21
a,b
0.31 0.30
86.3% 88.9%
-0.15b -0.17
b
0.23 0.22
n/a n/a
*adjusted means. a p<0.05, compared to placebo/500 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA). b p<0.05, compared to placebo/1450 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA).
TABLE 4. 150
Dose response caries validation study #1. INDIANA CONFERENCE 2005
oride dentifrices delivered statistically significantly (p<0.05) lower DMFS scores than the placebo control dentifrice at 9 months, while at 21 months, the 500 ppm F- and 1450 ppm F- fluoride dentifrices delivered statistically significantly lower DMFS scores as compared to the both the placebo/500 ppm F- and the placebo/1450 ppm F- dentifrice groups. There was no evidence of a dose response (1450 ppm F- <500 ppm F-) at 9 months or at 21 months. Study 2: This was a randomized, double-blind study conducted for 21 months with a supervised school oral hygiene regimen to assess whether dentifrices with increasing fluoride levels could be differentiated with small sample sizes in short timeframes [Biesbrock et al., 2003b]. Subjects (N = 644, with approximately 215 subjects per treatment group) with a mean age of 10.4 years (9-12 y.o.) used a placebo dentifrice, an 1100 ppm F dentifrice, or a 2800 ppm F dentifrice for the first 9 months of the study. Subjects in the placebo group were then re-randomized to either 1100 ppm or 2800 ppm F dentifrice for the remainder of the study, while subjects in the fluoride groups continued with their original treatment assignments. Three calibrated examiners measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 9 months, and 21 months for each subject. The results of this study were consistent with the previous results reported for sodium fluoride dentifrices. For all examiners, the 1100 ppm and 2800 ppm fluoride dentifrices delivered statistically significantly (p<0.05) lower DMFS scores than the placebo control dentifrice at 9 months, while at 21 months, the 1100 ppm and 2800 ppm fluoride dentifrices delivered statistically significantly lower DMFS scores compared to both the placebo/1100 ppm and the placebo/2800 ppm dentifrice groups (Table 5). In addition, 1 of the 3 examiners observed a directional (p=0.11) dose response (2800 ppm F<1100 ppm F) at 9 months, while at 21 months, all 3 examiners observed evidence of a dose response, with 1 examiner observing a statistically significant difference between 1100 ppm and 2800 ppm F. Caries scores on occlusal surfaces provided the strongest evidence of an 1100 ppm F vs. 2800 ppm F difference. In this study, the effectiveness of the 2 fluoride dentifrices was observed at 9 months, and these outcomes were still present at 21 months, confirming that caries benefits can be observed in timeframes as short as 9 months, with approximately 200 subjects per treatment group. Study 3: The objective of this research was to evaluate the anti-caries effectiveness of a low dose (500 ppm F-) sodium fluoride dentifrice, a high dose
BIESBROCK
ET AL.
A. 9-month increment. DMFS (Integrated) Treatment Placebo 1100 ppm 2800 ppm N 203 201 196 Mean* 3.47 2.00a 1.57a SEM 0.24 0.24 0.25 % Reduction -------42.4% 54.8% Mean* 1.73 0.81a 0.43a DMFS (Visual-Tactile) SEM 0.17 0.18 0.18 % Reduction ------53.2% 75.1%
*adjusted means. a p<0.05, compared to placebo group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA). b p<0.05, compared to 1100 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA).
B. 21-month increment. DMFS (Integrated) Treatment Placebo1100 ppm Placebo2800 ppm 1100 ppm 2800 ppm N 83 Mean* 2.08 SEM 0.36 % Reduction -------Mean* 1.04 DMFS (Visual-Tactile) SEM 0.26 % Reduction -------
90
2.56
0.37
-23.1%
1.16
0.26
-11.5%
168 153
1.04a,b -0.06a,b,c
0.26 0.28
50.0% 102.9%
0.14a,b -0.45a,b,c
0.19 0.20
86.5% 143.3%
*adjusted means. a p<0.05, compared to placebo/1100 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA). b p<0.05, compared to placebo/2800 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA). c p<0.05, compared to 1100 ppm F group via Least Significant Difference (LSD) test following Analysis of Covariance (ANCOVA).
TABLE 5.
Dose response caries validation study #2 [Biesbrock et al., 2003].
(2800 ppm F-) sodium fluoride dentifrice, and an experimental 0.454% stabilized stannous fluoride (1100 ppm F) with sodium hexametaphosphate dentifrice, each relative to a standard 1100 ppm F- sodium fluoride positive control dentifrice [Stookey et al., 2004]. Subjects (N = 955) with a mean age of 10.6 years were randomly assigned to 1 of 4 dentifrice treatments, with about 239 subjects per treatment group. Two calibrated examiners measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 12 months, and 24 months for each subject. At the 24-month examination, Examiner A observed statistically significantly less caries increment in the 2800 ppm dentifrice group (mean DMFS increment of 5.45) relative to the positive control 1100 ppm treatment group (mean DMFS increment of 6.27), with p=0.045 in an analysis based on all subjects with 24-month data. In an analysis based on the subset of evaluable subjects (subjects who attended at least 60% of supervised toothbrushing sessions), Examiner A also observed sta152 INDIANA CONFERENCE 2005

A. 12-month increment. Adjusted Mean DMFS (All subjects completing BL and year 1 exams) Treatment 500 ppm 1100 ppm 2800 ppm SnF2/Na Hexc
a
N 202 203 208 183
Meana 1.67 1.85 1.25 1.18
SEM 0.182 0.182 0.179 0.191
% Reduction 10.2% -------32.8% 36.4%
p-value 0.768 -------0.009 0.005
Adjusted means from analysis of covariance. b % Reduction = 100% (1100 ppm mean minus treatment mean) divided by 1100 ppm mean. c 0.454% stannous fluoride with sodium hexametaphosphate (1100 ppm F-).
B. 24-month increment. Adjusted Mean DMFS (All subjects completing BL and year 1 exams) Treatment 500 ppm 1100 ppm 2800 ppm SnF2/Na Hexc
a
N 168 174 180 161
Meana 4.42 4.95 3.80 3.72
SEM 0.299 0.293 0.288 0.306
% Reduction 10.7% -------23.2% 24.8%
p-value 0.896 -------0.003 0.002
Adjusted means from analysis of covariance. b % Reduction = 100% (1100 ppm mean minus treatment mean) divided by 1100 ppm mean. c 0.454% stannous fluoride with sodium hexametaphosphate (1100 ppm F-).
TABLE 6.
Dose response caries validation study #3 [Stookey et al., 2004].
tistically significantly less caries in the 2800 ppm dentifrice group (mean DMFS increment of 5.38) relative to the 1100 ppm treatment group (mean DMFS increment of 6.21) with p=0.043. In addition, Examiner A observed directionally less caries in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group (mean DMFS increment of 5.52) relative to the 1100 ppm treatment group (mean DMFS increment of 6.27) with p=0.065 in the analysis based on all subjects with 24-month data. Examiner A also observed statistically significantly less caries in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group (mean DMFS increment of 5.16) relative to the 1100 ppm treatment group (mean DMFS increment of 6.21) with p=0.019 in an analysis of the evaluable subjects. The results of this study for Examiner B are presented in Table 6. At both the 12- and 24-month examinations, Examiner B observed statistically significantly less caries in the 2800 ppm dentifrice group relative to the 1100 ppm
BIESBROCK
ET AL.
treatment group, with p0.009 in analyses based on all subjects. Examiner B also observed statistically significantly less caries in the 2800 ppm dentifrice group relative to the 1100 ppm treatment group at both 12 and 24 months (p0.011) in the analysis of evaluable subjects. In addition, Examiner B observed statistically significantly less caries at both 12 and 24 months in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group relative to the 1100 ppm treatment group in both the analysis based on all subjects (p0.011) and the analysis based on evaluable subjects (p0.010). Neither examiner observed statistically significant differences between the mean caries increments in the 500 ppm and 1100 ppm treatment groups. Collectively, these results demonstrated that the 2800 ppm F- sodium fluoride and the 0.454% stabilized stannous fluoride (1100 ppm F) with sodium hexametaphosphate dentifrices delivered statistically significantly greater reductions in caries than the positive 1100 ppm F- sodium fluoride dentifrice. The 500 ppm F- sodium fluoride dentifrices delivered similar caries efficacy relative to the positive 1100 ppm F- sodium fluoride dentifrice. Study 1: A controlled fluoride (F) dose-response study was conducted in concurrence with a supervised school oral hygiene regimen to assess whether dentifrices with increasing fluoride levels could be differentiated with small sample sizes in short timeframes [Biesbrock et al., 2003a]. The study was a randomized, double-blind study conducted for 21 months. Subjects (N = 657 with approximately 219 subjects per treatment group) were randomized to placebo dentifrice, 500 ppm F- dentifrice, or 1450 ppm F- dentifrice treatments for the first 9 months of the study. Subjects in the placebo group were then re-randomized to either 500 ppm F- or 1450 ppm F- dentifrice for the remainder of the study, while subjects in the fluoride groups continued with their original treatment assignments. A calibrated examiner measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 9 months, and 21 months for each subject. The mean caries increments at 9 months were 0.35 surfaces, 0.34 surfaces, and 1.28 surfaces for the 1450 ppm F-, 500 ppm F-, and placebo groups, respectively (Table 4). The mean caries increments at 21 months were 0.21 surfaces, 0.26 surfaces, 1.75 surfaces, and 1.90 surfaces for the 1450 ppm F-, 500 ppm F-, placebo/1450 ppm F-, and placebo/500 ppm F- groups, respectively. The 500 ppm F- and 1450 ppm F- fluoride dentifrices delivered statistically significantly (p<0.05) lower DMFS scores than the placebo control dentifrice at 9 months, while at 21 months, the
500 ppm F- and 1450 ppm F- fluoride dentifrices delivered statistically significantly lower DMFS scores as compared to the both the placebo/500 ppm F- and the placebo/1450 ppm F- dentifrice groups. There was no evidence of a dose response (1450 ppm F- <500 ppm F-) at 9 months or at 21 months. Study 2: This was a randomized, double-blind study conducted for 21 months with a supervised school oral hygiene regimen to assess whether dentifrices with increasing fluoride levels could be differentiated with small sample sizes in short timeframes [Biesbrock et al., 2003b]. Subjects (N = 644, with approximately 215 subjects per treatment group) with a mean age of 10.4 years (9-12 y.o.) used a placebo dentifrice, an 1100 ppm F dentifrice, or a 2800 ppm F dentifrice for the first 9 months of the study. Subjects in the placebo group were then re-randomized to either 1100 ppm or 2800 ppm F dentifrice for the remainder of the study, while subjects in the fluoride groups continued with their original treatment assignments. Three calibrated examiners measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 9 months, and 21 months for each subject. The results of this study were consistent with the previous results reported for sodium fluoride dentifrices. For all examiners, the 1100 ppm and 2800 ppm fluoride dentifrices delivered statistically significantly (p<0.05) lower DMFS scores than the placebo control dentifrice at 9 months, while at 21 months, the 1100 ppm and 2800 ppm fluoride dentifrices delivered statistically significantly lower DMFS scores compared to both the placebo/1100 ppm and the placebo/2800 ppm dentifrice groups (Table 5). In addition, 1 of the 3 examiners observed a directional (p=0.11) dose response (2800 ppm F<1100 ppm F) at 9 months, while at 21 months, all 3 examiners observed evidence of a dose response, with 1 examiner observing a statistically significant difference between 1100 ppm and 2800 ppm F. Caries scores on occlusal surfaces provided the strongest evidence of an 1100 ppm F vs. 2800 ppm F difference. In this study, the effectiveness of the 2 fluoride dentifrices was observed at 9 months, and these outcomes were still present at 21 months, confirming that caries benefits can be observed in timeframes as short as 9 months, with approximately 200 subjects per treatment group. Study 3: The objective of this research was to evaluate the anti-caries effectiveness of a low dose (500 ppm F-) sodium fluoride dentifrice, a high dose (2800 ppm F-) sodium fluoride dentifrice, and an experimental 0.454% stabilized stannous fluoride (1100 ppm F) with sodium hexametaphosphate dentiCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 155
BIESBROCK
ET AL.
frice, each relative to a standard 1100 ppm F- sodium fluoride positive control dentifrice [Stookey et al., 2004]. Subjects (N = 955) with a mean age of 10.6 years were randomly assigned to 1 of 4 dentifrice treatments, with about 239 subjects per treatment group. Two calibrated examiners measured visual-tactile caries as DMFS that was supplemented with a radiographic examination at baseline, 12 months, and 24 months for each subject. At the 24-month examination, Examiner A observed statistically significantly less caries increment in the 2800 ppm dentifrice group (mean DMFS increment of 5.45) relative to the positive control 1100 ppm treatment group (mean DMFS increment of 6.27), with p=0.045 in an analysis based on all subjects with 24-month data. In an analysis based on the subset of evaluable subjects (subjects who attended at least 60% of supervised toothbrushing sessions), Examiner A also observed statistically significantly less caries in the 2800 ppm dentifrice group (mean DMFS increment of 5.38) relative to the 1100 ppm treatment group (mean DMFS increment of 6.21) with p=0.043. In addition, Examiner A observed directionally less caries in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group (mean DMFS increment of 5.52) relative to the 1100 ppm treatment group (mean DMFS increment of 6.27) with p=0.065 in the analysis based on all subjects with 24-month data. Examiner A also observed statistically significantly less caries in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group (mean DMFS increment of 5.16) relative to the 1100 ppm treatment group (mean DMFS increment of 6.21) with p=0.019 in an analysis of the evaluable subjects. The results of this study for Examiner B are presented in Table 6. At both the 12- and 24-month examinations, Examiner B observed statistically significantly less caries in the 2800 ppm dentifrice group relative to the 1100 ppm treatment group, with p0.009 in analyses based on all subjects. Examiner B also observed statistically significantly less caries in the 2800 ppm dentifrice group relative to the 1100 ppm treatment group at both 12 and 24 months (p0.011) in the analysis of evaluable subjects. In addition, Examiner B observed statistically significantly less caries at both 12 and 24 months in the 0.454% stabilized stannous fluoride with sodium hexametaphosphate group relative to the 1100 ppm treatment group in both the analysis based on all subjects (p0.011) and the analysis based on evaluable subjects (p0.010). Neither examiner observed statistically significant differences between the mean caries increments in the 500 ppm and 1100 ppm treatment groups. Collectively, these
results demonstrated that the 2800 ppm F- sodium fluoride and the 0.454% stabilized stannous fluoride (1100 ppm F) with sodium hexametaphosphate dentifrices delivered statistically significantly greater reductions in caries than the positive 1100 ppm F- sodium fluoride dentifrice. The 500 ppm F- sodium fluoride dentifrices delivered similar caries efficacy relative to the positive 1100 ppm F- sodium fluoride dentifrice.
Summary Traditional caries models, as they have been practiced for the last 40 years, constitute a serviceable tool that has been used to assess the efficacy and safety of products and, to a lesser extent, the generalizability of safety and efficacy to the broader population. However, the study designs are relatively inefficient and suffer from a fundamental problem of competing objectives that compromise a clean, unambiguous interpretation on any single objective. Traditional study designs skew toward answering the efficacy question at the expense of optimally addressing safety, and efficacy in the broader population is not directly addressed at all. A more targeted approach would let a given study be designed to answer a single primary objective, and that goal would not be confounded by the collection of additional data. Furthermore, the competing objectives of safety, efficacy, and generalizability would be best answered separately, using individual studies of optimal design for the stated objective. The research community should be moving to a model where smaller directed studies are used to assess efficacy of an agent in a population with the disease that uses the product as directed. Safety should be assessed in larger populations using the product as directed, but also examining the effects of overuse of product. In contrast to traditional caries studies, safety should be examined at more frequent intervals, with more detailed oral soft tissue examinations. Finally, generalizability of effectiveness to the population at large is optimally answered through large, community-based studies across wide age ranges in geographically disperse areas. Caries prone subsets (active caries at baseline) should be broken out and compared to results of the efficacy studies. However, it should be pointed out that, in a caries prone population like those who participated in the Guatemala studies, anti-caries interventions such as fluoride in the drinking water have been reported to offer anti-caries community effectiveness consistent with that observed in U.S. general population [Archila et al., 2003].
BIESBROCK
ET AL.
Collectively, the research summarized in this paper supports that, wellcontrolled, supervised caries clinical trials using a D2D4 diagnostic criteria can be used to differentiate anti-caries products using relatively small sample sizes (n = 150-200 completed subjects per treatment group) after 9 to 12 months and 21 to 24 months of treatment. Three independent, randomized, controlled, dose response clinical trials were conducted in which the positive control fluoride dentifrice delivered statistically significantly less caries increment relative to the control dentifrices at both the 9-to-12 month and the 21-to-24 month observation time periods. The fundamental clinical model design is based on rigid control of sources of variation. Key tenets of the model include: 1) the use of caries prone populations with a similar range of caries activity, 2) supervised use of product to ensure compliance with intended product usage, and 3) control of background access to care to reduce variation induced by background restorative dentistry. Recently, a novel caries trial design was reported, where clinical visual assessment (CVA) in the absence of supplemental diagnostics was found to differentiate 1000 ppm F- as MFP from 2500 ppm F- as MFP in 12 months [Chesters et al., 2002]. That trial design shares similarities with the 3 trials summarized in this paper, in that both models rely on testing in caries prone populations, using expanded visual diagnostic thresholds relative to the traditional D3 level. The CVA method grades white spot lesions (D1) as caries, where the grading scale in the Biesbrock-Bartizek model considered white spot lesions as sound for diagnostic purposes. However, lesions that correspond to D2 through D4 are considered as caries under both grading systems. Importantly, the key guidance in selection of the diagnostic threshold is to understand the underlying implications of the selected diagnostic threshold on signal to noise ratio (ratio of true positives to false positives). Ultimately, the diagnostic threshold (method) selected must be guided by a number of factors, including the relative disease activity in the population (increment rate) and the signal-to-noise ratio of the diagnostic.
References
Archila LA, Bartizek RD, Gerlach RW, Jacobs SA, Biesbrock AR: Incidence and prevalence of dental caries in five Guatemalan communities in children ages 6 to 12. J Clin Dent 2003;14:53-58.
158

Bartizek RD, Gerlach RW, Faller RV, Jacobs SA, Bollmer BW, Biesbrock AR: Reduction in dental caries with four concentrations of sodium fluoride in a dentifrice: A meta-analysis evaluation. J Clin Dent 2001;12:5762. Biesbrock AR, Chesters RK, Ellwood RP, Smith SR: The challenges of validating diagnostic methods relative to a conventional two-year caries clinical trial. J Dent Res 2004;83 (Spec. Issue C):C53-C56. Biesbrock AR, Bartizek RD, Gerlach RW, Jacobs SA, Archila L: Effect of three concentrations of sodium fluoride dentifrices on clinical caries. Am J Dent 2003a;16:99-104. Biesbrock AR, Bartizek RD, Gerlach RW, Jacobs SA, Archila LA: Dose response efficacy of sodium fluoride dentifrice at 9 and 21 months with a supervised brushing regimen. Am J Dent 2003b;16:305-312. Biesbrock AR, Bartizek RD, Faller RV, Jacobs SA, Gerlach RW: The effects of 1100, 1700, 2200 and 2800 ppm fluoride ion as sodium fluoride in a dentifrice: A one-year dose response caries study. Community Dent Oral Epidemiol 2001;29:382-389. Chesters RK, Ellwood RP, Biesbrock AR, Smith SR: Potential modern alternative designs for caries clinical trials (CCTs) and how these can be validated against the conventional model. J Dent Res 2004;83(Spec. Issue C):C122-C124. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson JR, Nicholson JA, Gendvilte A, Sabalaite R, Ramanauskiene J, Savage D, Mileriene J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002;81:637-640. Ekstrand KR: Improving clinical visual detection-potential for caries clinical trials. J Dent Res 2004;83:C67C71. Feller, RP, Kiger RD, Triol CW, Sintes, JL, Garcia L, Petrone ME, Volpe AR, Proskin HM: Comparison of the clinical anticaries efficacy of an 1100 NaF silica-based dentifrice containing triclosan and a copolymer to an 1100 NaF silica-based dentifrice without those additional agents: a study on adults in California. J Clin Dent 1996;7:85-89. Fleiss JF, Cohen J: The equivalence of weighted kappa and the interclass correlation coefficient as measures of reliability. Educ Psychol Meas 1973;33:613-619. Gerlach RW, Bartizek RD, Jacobs SA, Biesbrock AR: Sibling caries: evidence from four geographicallydiverse cross-sectional studies (abs). J Dent Res 2001;80:36. Grainger RM, Lehnhoff RW, Bollmer BW, Zacherl WA: Analysis of covariance in dental caries clinical trials. J Dent Res 1984;63(Spec. Issue):766-771. Hawley GM, Hamilton FA, Worthington HV, Davies RM, Holloway PJ, Davies TGH, Blinkhorn AS: A 30month study investigating the effect of adding triclosan/copolymer to a fluoride dentifrice. Caries Res 1995;29:163-167. Jensen ME, Kohout F: The effect of a fluoridated dentifrice on root and coronal caries in an older adult population. J Am Dent Assoc 1988;117:829-832. Koch G, Bergmann-Arnadottir I, Bjarnson S, et al.: Caries preventive effect of fluoride dentifrices with and without anticalculus agents: A 3-year controlled clinical trial. Caries Res 1990;24:72-79.
159
BIESBROCK
ET AL.
Lu KH, Ruhlman CD, Chung KL, Sturtzenberger OP, Lenhoff RW: A three-year clinical comparison of a sodium monofluorophosphate dentifrice with sodium fluoride dentifrices in dental caries on children. J Dent Child 1987;54:241-245. Lu KH, Yen DJC, Zacherl WA, et al.: The effect of fluoride dentifrice containing anticalculus agent on dental caries in children. ASDC J Dent Child 1985;52:449-451. Mann J, Vered Y, Babayof I, Sintes J, Petrone ME, Volpe AR, Stewart B, DeVizio W, McCool JJ, Proskin HM: The comparative anti-caries efficacy of a dentifrice containing 0.3% triclosan and 2.0% copolymer in a 0.243% sodium fluoride/silica base and a dentifrice containing 0.243% sodium fluoride/silica base: a twoyear coronal caries clinical trial on adults in Israel. J Clin Dent 2001;12:71-76. Marks RG, DAgostino R, Moorhead JE, Conti AJ, Cancro L: A fluoride dose-response evaluation in an anticaries clinical trial. J Dent Res 1992; 71:1286-1291. OMullane DM: Efficiency in clinical trials of caries preventative agents and methods. Community Dent Oral Epidemiol 1976;4:190-194. Pitts NB: Clinical diagnosis of dental caries: A European perspective. J Dent Educ 2001;65:972-978. Radike AW: Criteria for the diagnosis of dental caries. Proc Conf on the Clin Testing of Cariostatic Agents. Oct 1968. Chicago, American Dental Association, 1972, pp. 87-88. Reich E: Acceptance and application of new caries methods. In: Early Detection of Dental Caries II. Stookey GK, editor. Indianapolis: Indiana University, 1999, pp. 415-420. Stookey GK, Mau M, Isaacs RI, Gonzlez-Gierbolini C, Bartizek RD, Biesbrock AR: The anti-caries effectiveness of three fluoride containing dentifrices relative to an 1100 ppm sodium fluoride positive control. Caries Res 2004;38:542-550. Zacherl WA: A three-year caries clinical evaluation of the effect of sodium fluoride-silica abrasive dentifrice. Pharmacol Therapy Dent 1981;6:1-7.
160
Rationale and Evidence for the International Caries Detection and Assessment System (ICDAS II)
Amid I. Ismaila (co-chair) and the International Caries Detection and Assessment System Coordinating Committee1
1
D. Banting, H. Eggertsson, K.R. Ekstrand, A. Ferreira-Zandon, C. Longbottom, N.B. Pitts (co-chair), E. Reich, D. Ricketts, R. Selwitz, W. Sohn, G.V. Topping (coordinator), D. Zero
a
Department of Cariology, Restorative Sciences, and Endodontics, School of Dentistry, University of Michigan, Ann Arbor, MI, USA
The International Caries Detection and Assessment System (ICDAS) presents a new paradigm for the measurement of dental caries that was developed based upon insights gained from a systematic review of the literature on the clinical caries detection system [Ismail, 2004a] and other sources [Chesters et al., 2002; Ekstrand et al., 1997; Fyffe et al., 2000; Ekstrand et al., 2001; Ekstrand et al., 2005; Ricketts et al., 2002]. That review found that, while new caries detection criteria measured different stages of the caries process, there were inconsistencies in how the caries process was measured. The review also found that there is a gulf between European and American systems for caries detection, and there were inconsistencies among the research criteria for measuring dental caries. By and large, especially in the USA, dental caries has been synonymous with the presence of cavitation. In Europe, at least among the research community, the understanding of dental caries appears to be more advanced than the dichotomous approach used in the American criteria for measuring caries in that the clinical stages of the disease process, which precede cavitation, are acknowledged and often recorded. The future of research, practice, and education in cariology requires the development of an integrated definition of dental caries and uniform systems for measuring the caries process. The systematic review concluded that there is an urgent need to address the answers to the following questions: 1) what stage of the caries process should be measured; 2) what are the definitions for each selected stage; 3) what is the best clinical approach to detect each stage on difCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 161
ISMAIL
ET AL.
ferent tooth surfaces; and 4) what protocols of examiners training can provide the highest degree of examiner reliability? These were the initial questions that initiated the discussion of the ICDAS process. In this paper, we will describe the philosophy of the ICDAS system, how the ICDAS answers each of the 4 questions, and whether ICDAS can serve as a basis and benchmark for clinical and epidemiological research and inform dental undergraduate and postgraduate teaching in cariology. At the outset, it is important to define an important guiding principle of ICDAS. Several times, members of the coordinating committee have attempted to include the greatest input of the cariology community into the process of developing integrated criteria. The ICDAS committee was expanded to include a larger group of participants. Invitations were mailed to cariologists from Europe and the USA. The following document summarizes the discussions that took place during the Baltimore ICDAS II workshop, which was held in March 2005. It should also be understood that the ICDAS committee explicitly acknowledges that the ICDAS approach is built on a foundation of evidence that dates back over the last hundred years to G.V. Black in the USA and to many of the founding fathers of ORCA (the European Organisation for Caries Research) in Europe.
Historical Perspectives and the Need for an Integrated System Developments in Epidemiological Caries Measures More than a decade ago, concern was expressed about how the quality and comparability of caries data could best be safeguarded in order to achieve valid assessments of disease status at a time when significant service developments were accompanied by changes in both the pattern and distribution of dental caries [Pitts, 1993]. These issues are even more relevant today. There is a danger that key information and concepts are not being disseminated sufficiently well [Pitts, 1994], and many of the clear and established issues and challenges in this area are still not recognized in dental public health. There is, therefore, a need to continue work to bring together the evidence base from research in the field of cariology (which is very robust in some areas, but more deficient in others) on the one hand, and the national and international dental epidemiology, dental public health, and dental practice communities on the other. Many of the concepts debated, at least since the 1980s in cariology, are still seen as new or as radical by many working in other fields. There are,
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
SIMPLER TERMS
DENTISTS TERMS
Traditional indicators are now termed:

severe decay Pulpal decay
obvious decay into dentin
established decay
Visible dentin decay Unseen dentin decay Unseen enamel decay
Proportion with obvious decay experience or Proportion with no obvious decay experience
early stage decay
Visible enamel decay
+ specific indication of when diagnostic threshold includes enamel lesions
very early stage decay
Sub-clinical decay
FIG. 1:
Updated caries terminology recommended to allow continuity with traditional measures, while also reflecting the current evidence from cariology [Pitts 2004].
however, some encouraging signs in the UK [Drugan, 2004], the European Association for Dental Public Health (EADPH), the American Dental Association (which supported the Baltimore workshop), and the Federation Dentaire Internationale. The new emphasis on caries measurement and management may indicate that the dental community worldwide has started to recognize that we need new approaches in caries detection, assessment, and management. The ICDAS coordinating committee has been guided by the model depicted in fig. 1, which illustrates graphically the type of updated caries terminology now being recommended. This allows more clarity for lay and non-dental audiences and continuity with traditional measures, while also reflecting the current research evidence from cariology. Key changes are to carefully avoid the use of the misleading and widely misunderstood term caries free and to explicitly acknowledge whether or not initial lesions clinically confined to the enamel are included or excluded in examinations. Developments in Caries Measures for Clinical Research There have been several conferences held during the last 5 years that focus on caries detection and assessment. A recent issue of Caries Research, reportCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 163
ISMAIL
ET AL.
ing the peer-reviewed proceedings of the 50th Anniversary European Organisation for Caries Research (ORCA) Congress on Cariology in the 21st Century, is a good start [Nyvad et al., 2004]. The series of published proceedings from the Indiana Conferences on Early Detection of Dental Caries organised by Professor George Stookey and published by Indiana University also contain a wealth of details of the work in this area [Stookey, 1996; 2000; 2004]. In the field of randomized clinical trials of caries preventive agents, it has now been shown that, by using clinical visual diagnostic criteria that include enamel lesions, it is possible to detect differences in treatment effect over a shorter period than by using criteria that only relies on the later stage caries changes extending into dentin [Chesters et al., 2002]. An International Consensus Workshop on Caries Clinical Trials (ICWCCT) was held in 2002 involving 95 participants from 23 countries. The final Consensus Statements represent international agreement on where the evidence leads in the field of caries clinical trials [Pitts and Stamm, 2004]. The final agreed text includes: There is some confusion with the terminology employed in the literature around caries diagnosis (which should imply a human professional summation of all available data), lesion detection (which implies some objective method of determining whether or not disease is present), and lesion assessment (which aims to characterise or monitor a lesion, once it has been detected). The understanding of the caries process has progressed far beyond the point of restricting the evidence for dental caries to the D2 (caries in enamel only) or D3 (caries in enamel and dentin) levels of cavitation. For future clinical trials, recording only cavitated lesions as an outcome measure is becoming outmoded. The workshop participants also recommended that, in light of the evidence reviewed, both here and elsewhere, pertaining to modern caries definitions and measurement conceptsthe participants supported a statement recommending that, in future controlled clinical trials, caries measurement methods are employed which: 1. Are capable of accurately capturing at any given point in time the manifestations of the caries process in dental hard tissues (enamel and dentin).
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
2. When applied sequentially, can monitor definitive changes in manifestations of the caries process over time, over and above any background noise from normal levels of deand re-mineralization, or from variations attributable to the caries detection system(s) employed. 3. When applied sequentially, can differentiate actual product effects in terms of group differences in lesion initiation and lesion behavior (progression, arrest and/or regression). Immediately following the ICW-CCT workshop (April, 2002), the ad hoc ICDAS coordinating committee was formed by Drs. Pitts and Ismail. The goal of that committee has been to develop an integrated clinical detection and assessment system of dental caries for research and clinical practice. The development of new technologies and applications has the potential to supplement clinical caries detection, but these assessments will have to be clinically meaningful by providing measurements over and above the noise of arrested initial and sub-clinical lesions (Pitts and Stamm, 2004]. A major challenge in synthesising the developing evidence in the partially overlapping fields of caries epidemiology, clinical caries research, and clinical caries management is the incompatibility of the terminology, criteria, and grading systems currently used across these 3 fields. This challenge, together with a number of recommendations from the NIH Consensus Development Conference [2001] and the ICW meeting on Clinical Caries Trials, led an ad hoc group to start the development of the International Caries Detection and Assessment SystemICDAS. ICDAS: The CommitteeICDAS activities have been carried out under the supervision of and on behalf of an informal, unfunded, ad hoc and voluntary committee that was assembled in an attempt to advance some of the key recommendations in the area of caries detection and assessment criteria. After the first meeting in Dundee, Scotland, an invitation was mailed to cariologists from Europe and the USA to attend a development workshop in Ann Arbor, Michigan, USA. No attempt was made to exclude any researcher or individual. The founding committee comprised: from the Dental Health Services Research Unit, University of Dundee (DHSRU): Nigel Pitts, Christopher Longbottom, Gail Topping, David Ricketts; from the University of Michigan: Amid Ismail; from Indiana University: Domenick Zero; from Copenhagen University: Kim Ekstrand; from the International Dental Federation (FDI): Elmar Reich; and
ISMAIL
ET AL.
from NIH/NIDCR: Rob Selwitz. At the first meeting, there was helpful input from Andrew Forgie (Dundee) and Chris Deery (now Edinburgh). From the second meeting, the Committee was joined by David Banting (Ontario), Hafsteinn Eggertsson (Indiana), and Woosung Sohn (Michigan), and the third meeting onward, the committee was joined by Andra Ferreira-Zandon (Indiana). To this group, an additional 10 individuals participated in the Ann Arbor workshop. This group comprised the ICDAS development committee in 2002. ICDAS: Philosophythe philosophy upon which this truly collaborative initiative is based is one where the methodology from caries epidemiology meets that from caries clinical trials and practice, and the whole is conducted according to the values of evidence based dentistry (EBD). There have been many systems devised over the years for grading dental caries, which have been visually based and included non-cavitated lesions in enamel and all are fully acknowledged. The driving principles of the ICDAS committee are: integration, scientific validation, and utility of the criteria in different research and practice settings. ICDAS: Development MeetingsBefore the ICDAS II workshop, 4 development meetings were heldDundee, Scotland in April 2002; Ann Arbor, Michigan, USA in August 2002, where the ICDAS I criteria were developed; Indianapolis, Indiana, USA in May 2003; and Bornholm, Denmark in April 2004. The ICDAS II workshop was held in Baltimore, MD, USA to share progress in the ICDAS criteria and to seek input of wider international expertise. Invitations were mailed to large group of experts, and those who accepted the invitation convened to review, revise as necessary, and agree on the ICDAS II version of the criteria. The invitations were mailed to more than 60 cariologists and researchers in the field. ICDAS: ConceptsThe use of a standardized system based on best evidence should lead to better quality information to inform decisions about appropriate diagnosis, prognosis, and clinical management of dental caries at both the individual and public health levels. A wardrobe of validated tools should allow users to select the best criteria and conventions for a specific use. Adoption of the system should, in the longer term, also facilitate the work of those who subsequently seek to systematically review published evidence in the 3 fields referred to above. The concept is that the system will be an open one, maintained on the World Wide Web, and subject to peer review. Users of the system will have to: 1) specifically acknowledge the version of the system they employ and 2) specify which parts of the ICDAS wardrobe are being used.
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Figure 2 summarises the following key features of ICDAS: The ICDAS caries detection criteria have been piloted in various guises in Dundee, Detroit, Indiana, Copenhagen, Columbia, Mexico, and Iceland. They are now ready for wider use and have been further peer reviewed in 2005. The ICDAS caries activity criteria are still part of an expanding research agenda. Preliminary caries activity assessment criteria have been developed, using the ICDAS approach of relying on visual assessment and the WHO/PSR probe. Further research is planned to validate the proposed criteria. The ICDAS caries system provides a vital step forward in giving a coherent framework of comparison against which the potential benefits and performance of existing and new aids to caries detection and diagnosis can be assessed against the optimised clinical visual method. Previous systematic reviews and consensus conferences have had considerable difficulty with the heterogeneous methodology and reporting in this area. Caries diagnosis is an important part of the dentists daily work. Caries diagnosis is a process that can be considered a 3-step procedure: detection of the lesion, followed by an assessment of the severity of the lesion, which, again, is
ICDAS
International Caries Detection & Assessment System

For use on coronal and root surfaces, as well as caries adjacent to restorations and sealants These unifying, predominantly visual, criteria code a range of the characteristics of clean, dry teeth in a consistent way that promotes the valid comparison of results between studies, settings & locations ICDAS criteria record both enamel and dentine caries and explore the measurement of caries activity in all of the domains below
Core ICDAS Criteria 2004
Education
ICDAS Clinical Visual Criteria
Epidemiology / Public Health
Clinical Research
Clinical Practice
The ICDAS Detection codes are in use now and are recommended The ICDAS Assessment codes are part of a developing research agenda The ICDAS System provides an evidence based framework to validate and explore the impact of existing and new-technology aids to caries diagnosis
FIG. 2:
Overview of the development of the International Caries Detection and Assessment System ICDAS. 167
ISMAIL
ET AL.
followed by an assessment of the activity of the lesion [Ekstrand et al., 2001]. Caries risk assessment, on the other hand, is assessment of the risk of getting new lesions in the near future [Bratthall et al., 1997]. Early in the discussions of the ICDAS coordinating committee, it was recognised that lesion detection without assessment was of little clinical relevance. However, in reviewing the literature, it was found that there was insufficient current replicated evidence on the visual signs and symptoms of lesion activity to present an evidence-based system within the ICDAS criteria for lesion assessment. The limited evidence available and previous criteria systems were synthesised and used to develop draft criteria for assessing caries activity, and during the ICDAS II workshop, participants reviewed and modified these. These proposed criteria will be investigated and further revised, if necessary.
Coronal Primary Caries Detection Criteria Principles Used to Develop the Criteria for Coronal Primary Caries Dental caries is a dynamic process, with cycles of demineralization followed by remineralization. The balance between the 2 cycles determines the stage of the disease as depicted in fig. 1. It is hard to categorize a complex disease like dental caries into a scale, because the process is continuous and could be measured, if feasible, as stages representing minute loss of tooth structure that is currently not detectable using the current technology available for in vivo use. Clinically, we rely on visual signs (change in color, cavitation), which represent manifestations of a relatively advanced caries process. To understand the measurement of caries, it is important to review basic concepts. Sound enamel is translucent and microporous. After repeated demineralization challenges, microporosity of the subsurface enamel increases. The increase in microporosity leads to a change in the refractive index of enamel. The first sign of carious change, hence, is a change in the translucency and light refraction of enamel after it is dried for a short period. If demineralization continues and enamel microporosity and surface loss increases, further reduction occurs in the refractive index of enamel. As a result, early carious lesions are seen even when the surface is covered with saliva. This is a more advanced stage of dental caries. Ekstrand et al. [1995] has correlated between the severity of carious lesions and their histological depth. White spot lesions, which require air-dry168 INDIANA CONFERENCE 2005
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
ing, are most likely to be limited to the outer half of enamel. The depth of a white or brown spot lesion, which is obvious without air-drying, is located some place between the inner half of the enamel and the outer third of the dentin. Localized enamel breakdown due to caries, with no visible dentin, indicates that the lesion extends to the middle third of the dentin. In addition, a greyish, brownish, or bluish shadow of the dentin shining through apparently intact enamel also indicates a lesion extending to the middle third of dentin. Frank cavities with visible dentin indicate that a lesion has been extended to the inner third of dentin. The ICDAS I and II criteria (Appendix) incorporate concepts from the research conducted by Ekstrand et al. [1995, 1997] and other caries detection systems described in the systematic review conducted by Ismail [2004]. These systems indicate that measurement of non-cavitated carious lesions in enamel or dentin can be based on visual topography at the surface level. While such systems are not perfectly accurate, they have both content and correlational validity with histological depth of carious lesions. As stated before, ICDAS was developed to provide an international system for caries detection that would allow for comparison of data collected in different locations and at different points in time. The ICDAS system was developed to bring forward the current understanding of the process of initiation and the progression of dental caries to the fields of epidemiological and clinical research. The coordinating committee also took into consideration developing a system that has wider utility for dental practitioners. If dental caries is classified using agreed upon criteria and systems, then a comparison of findings by epidemiologists and clinicians from different countries would be feasible. The ICDAS measures the surface changes and potential histological depth of carious lesions by relying on surface characteristics. The coordinating committees have discussed at great length the concept of measuring caries activity and have tested different clinical criteria systems [Ekstrand et al., 2005]. At the ICDAS II workshop in Baltimore, existing activity criteria were modified to fit with the ICDAS approach for clinical detection of dental caries. The proposed caries activity system will be evaluated in future research projects. The primary requirement for applying the ICDAS system is the examination of clean, dry teeth. The ICDAS examination is visual, aided by a ball-ended explorer that is used to remove any remaining plaque and debris and to check
ISMAIL
ET AL.
for surface contour, minor cavitation, or sealants. It is highly advisable that the teeth be cleaned with a toothbrush or a prophylaxis head/cup before the examination. The use of a sharp explorer is not necessary, because it does not add to the accuracy of detection, and it may damage the enamel surface covering early carious lesions [Ekstrand et al., 1987; Bergmen and Lindn, 1969]. The ICDAS criteria for coronal caries are described in the criteria document attached to this paper.
Caries Adjacent to Restorations and Sealants (CARS) Rational and Terminology When a restoration is placed in a tooth, the adjacent tooth tissue, which is vulnerable to caries, can be considered in 2 planes. There is the surface enamel and the enamel and dentin of the cavity wall. Secondary caries has classically been described as occurring in 2 ways: an outer lesion and a wall lesion. The chemical and histological processes involved in outer lesions are the same as primary caries, and it has been suggested that they occur as the result of a new, primary, attack on the surface of the tooth adjacent to the restoration. A number of researchers have suggested that secondary caries is quite likely to be primary caries adjacent to restorations [Ozer, 1997; Kidd and Beighton, 1996]. Additionally, however, given the appropriate conditions, a wall lesion may start on the wall of a cavity in the presence of leakage or microleakage. Thus, these lesions only occur secondary to the presence of a restoration. The definitions given to caries found in association with a restoration, or on a restored tooth, vary greatly in the literature. Secondary caries, recurrent caries, and residual caries are some of the terms commonly used. However, the same terms are used to describe different conditions by different investigators. Many of the definitions focus on the spread of caries at the enamel-dentin junction (EDJ) in a restored tooth. Other definitions include the failure to remove all diseased tissues in the deep part and/or at the margin of cavity preparations; however, this is more commonly considered to be residual caries. In an epidemiological survey, the presence or absence of caries adjacent to restorations is recorded without differentiating between new and residual caries. The terminology used should reflect this, and it is suggested that the term caries associated with restorations and sealants (CARS), may be suitable.
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Principles Used to Develop the Criteria for CARS Since outer carious lesions adjacent to restorations are thought to be analogous with primary caries, the broad principles applied to the criteria for primary caries are also applied to CARS, where relevant. However, it should be noted that the scientific basis for doing so has not been established, and the literature in the area of secondary caries is far more limited than for primary coronal caries. Much of the work, which has been conducted, has been done under ideal conditions within the laboratory setting, and even then, most have found poor correlations between visual signs and the histological findings. Although caries associated with restorations is histologically similar to primary caries, its features cause certain diagnostic problems, including difficulties in the differentiation among restoration margin discrepancies (marginal integrity, discoloration of the tooth at restoration margin), secondary caries, and residual caries [Mjr and Toffenetti, 2000]. Sharp probing for signs of secondary caries has all of the limitations and drawbacks associated with its use for primary caries detection. In addition, probing restored teeth can be misleading, as a probe may become impacted in a margin discrepancy that is not, in fact, carious. It has been demonstrated that discoloration at the restoration margin is difficult to evaluate, as shown by a moderate inter-examiner agreement (kappa of 0.49) [Tobi et al., 1999]. In part, this is due to the variety of causes of discoloration found next to amalgam, in particular. It is not always predictive of secondary caries, as a large amalgam restoration or its corrosion products may discolor the tooth grey or blue without caries being present. It has also been suggested that slowly progressing lesions are darkly stained [Miller and Massler, 1962], probably from exogenous dietary sources such as tea or coffee. It is possible that lesions that are most obvious clinically, because of their color, may be the ones that are inactive, arrested, or slowly progressing [Kidd, 1989a and 1989b]. Although corrosion products are known to form around amalgam fillings and are dark colored, Kidd and coworkers [Kidd et al., 1994] found similar levels of staining around amalgam fillings compared to that found around tooth colored restorations. A number of studies have been conducted to investigate the association between shadowing or grey discoloration and the presence or absence of caries, some concluding that there is a statistically significant association [Kidd et al., 1994; Rudolphy et al., 1995; Topping, 2001], while others found no such correlation [Kidd et al., 1995; Rudolphy et al., 1996].
ISMAIL
ET AL.
In conclusion, therefore, it should be noted that while many studies have shown that grey discoloration or shadowing at the margins of restorations is statistically significantly associated with caries, recording this as non-cavitated dentinal caries is likely to result in an overestimation of the amount of disease. The confounding of shadowing due to restoration color means that there are likely to be more false positives than that found in unrestored teeth if discoloration or shadowing alone is used to predict the presence of caries. Non-carious Changes and CARS A number of features of restored teeth, which are not necessarily associated with the presence of caries, may be worthy of recording. Although some of these categories may ultimately be counted as sound, it may be of some importance to be able to differentiate some states from that defined above. Such non-carious changes seen in restored teeth include discrepancies in the integrity of the tooth-restoration interface (marginal ditching) and fractured restorations (e.g., isthmus fractures as opposed to marginal discrepancies). If any features were present, concurrently with signs of caries, then the appropriate caries code would take precedence over any non-carious change code. Many studies have concluded that secondary caries is poorly related to marginal discrepancy [Kidd et al., 1992; Kidd, 1989a and 1989b; Kidd et al., 1994; Elderton, 1989; Kidd and OHara, 1990; Boyd and Richardson, 1985; Hamilton et al., 1993; Topping, 2001; Ando et al., 2004]. Some studies, however, have reported that the wider the gap at faulty margins, the greater the likelihood of caries [Goldberg et al., 1981; Goldberg, 1990; Jorgensen and Wakumoto, 1968]. It may, therefore, be important in an epidemiological study to record the presence of marginal ditching as an indication of teeth with an increased caries risk. The extent of marginal deficiencies will range from those barely perceptible on visual examination alone to those that will readily admit a ball-ended probe. Since an increased width of the marginal deficiency may be a risk factor for the likelihood of developing caries or not, it may be important to have a threshold at which the deficiency is recorded as present or absent. If a ballended probe is part of the examination kit, 2 categories of ditching could be recorded according to whether or not the probe can be admitted into the gap between the tooth and restoration.
172
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
The ICDAS criteria for CARS are described in the criteria document attached to this paper (Appendix). Root Caries A recent systematic review commissioned for the National Institutes of Health (NIH) Consensus Development Conference on Dental Caries Diagnosis and Management Throughout Life concluded that there is insufficient evidence on the validity of clinical diagnostic systems for root caries [Bader et al., 2001]. However, the review only included clinical studies that used histology to validate the clinical caries diagnosis. This inclusion criterion excluded the vast majority of the literature on root caries. Surveys describing the clinical appearance of root caries began to appear in the literature in the early 1970s, and many surveys and longitudinal studies on root caries were reported over the next 2 decades. Since the early 1990s, however, very few clinical studies on root caries have been conducted. These clinical studies primarily used diagnostic criteria proposed by Sumney et al. [1973]; Hix and OLeary [1976]; Banting et al. [1980]; Katz [1984]; and the U.S Department of Health and Human Resources [1987]. Generally, root caries lesions have been described as having a distinct outline and present with a discolored appearance in relation to the surrounding non-carious root. Many root caries lesions are cavitated, although this is not necessarily the case with early lesions. The base of the cavitated area can be soft, leathery, or hard to probing. Probing of root caries lesions with a sharp explorer using controlled, modest pressure, however, may create surface defects that prevent complete remineralization of the lesion [Warren et al., 2003]. Therefore, for detection and classification of root caries utilizing ICDAS criteria, examiners are directed to use a Community Periodontal Index (CPI) probe [World Health Organization, 1997]. Root caries frequently is observed near the cemento-enamel junction, although lesions can appear anywhere on the root surface. Lesions usually occur near (within 2 mm) the crest of the gingival margin. The distinction between an active and an arrested lesion further complicates clinical detection of root caries. The color of root lesions has been used as an indication of lesion activity. Active lesions have been described as yellowish or light brown in color, whereas, arrested lesions appear darkly stained. However, color subsequently has been shown not to be a reliable indicator of caries activity [Hellyer et al., 1990; Lynch and Beighton, 1994].
ISMAIL
ET AL.
Since the clinical signs of lesions are considered to be different for active versus arrested root caries and the clinical signs associated with lesion activity have yet to be validated, the criteria proposed within the ICDAS incorporate all of the reported clinical signs and, therefore, consider both lesion detection and assessment unlike the criteria for coronal caries. The presence of cavitation (loss of surface integrity) associated with a root caries lesion does not necessarily imply lesion activity. Non-cavitated (early) root caries lesions almost universally are considered to be active. A cavitated lesion, however, may be either active or arrested. Lesion activity has been linked to lesion depth [Billings et al., 1985], but this clinical observation has not been verified. The texture of a root caries lesion also has been linked to lesion activity. Active lesions have been described as soft or leathery, compared to arrested lesions that have a hard texture. There is supporting laboratory evidence from a study that used microbiological indicators for lesion activity that soft or leathery lesions on root surfaces are more heavily infected with bacteria than hard root surfaces [Lynch and Beighton, 1994]. Root caries lesions that occur closely adjacent to (within 2 mm) the crest of the gingival crest are considered to be active; whereas, lesions that occur on the root surface more distant from the gingival crest are more likely to be arrested. There is microbiological evidence to support this clinical observation [Beighton et al., 1993]. The determination of root caries activity probably is more closely related to decisions regarding treatment or management than to determination of the presence of caries on the tooth root. However, because the clinical signs of lesions are considered to be different for active versus arrested root caries and the clinical signs related with lesion activity have yet to be validated, the criteria proposed incorporate all of the reported clinical signs. Published reports on the clinical measurement of root caries were consulted in developing the ICDAS criteria [Hellyer and Lynch, 1991; Banting, 1993; Banting, 2001; Leake, 2001]. Given the paucity and generally low level of the scientific evidence, the ICDAS Coordinating Committee, recommends that the following clinical criteria be used for the detection and classification of root caries: 1. Color (light/dark brown, black); 2. Texture (smooth, rough); 3. Appearance (shiny or glossy, matte, or non-glossy);
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
4. Perception on gentle probing (soft, leathery, hard); and 5. Cavitation (loss of anatomical contour). Additionally, the outline of the lesion and its location on the root surface are useful in detecting root caries lesions. Root caries appears as a distinct, clearly demarcated circular or linear discoloration at the cemento-enamel junction (CEJ) or wholly on the root surface. The ICDAS criteria for root caries are described in the criteria document attached to this paper. Principles Used to Develop the Criteria for Caries Lesion Activity Assessment While detecting a caries lesion is important, it only represents part of the diagnostic process necessary to properly assess the caries disease status. A long sought after goal in cariology is to be able to accurately and reliably characterize the caries activity status of lesions. Is the lesion progressing, arrested, or regressing? Two approaches have been considered. The first approach involves monitoring over multiple clinical examinations changes in the physical and/or optical properties of caries lesions. For this approach, ICDAS severity scoring can be applied. The second approach involves attempting to characterize caries lesion activity during a single clinical examination in real-time, and this subject is under consideration here. The modern understanding of the dynamic nature of the caries process, where lesion progression can be arrested at any stage of the process, supports the importance of clinically assessing caries activity status [Nyvad and Fejerskov, 1997]. This is particularly important for non-cavitated lesions, because they may self-arrest as part of the natural history of the disease or become arrested due to changes in the local environment [Backer Dirk, 1966]. In older adults, arrested, non-cavitated lesions may be scars from disease activity occurring years or even decades earlier; however, these scars do not provide useful information about the current disease status of an individual unless they reflect a recently documented change from an active lesion status. Some consider activity assessment to be the Holy Grail of cariology, because it can provide chairside evidence of the caries disease process in real-time. Furthermore, it may also prove to be the best way to determine caries risk status and identify patients who require intensive preventive intervention [Zero et al., 2001].
175
ISMAIL
ET AL.
Clinical research is the other arena where caries lesion activity can and should play an important role. Caries assessment is necessary for identifying subjects who have teeth with active disease for studies designed to test treatments intended to arrest or reverse caries. Caries inactive lesions (scars) have a very low probability of progressing or regressing and, thus, mitigate the possibility of showing a treatment effect. The assessment of caries activity status of early lesions is currently very challenging, as it relies on the clinicians ability to identify subtle changes in enamel by visual and tactile inspection. Clinical criteria for caries lesion activity assessment have been developed [Ekstrand et al., 1997; Nyvad et al., 1999]. The criteria are generally based on the physical properties of surface reflection and texture of early lesions, with chalky, rough surfaces being active, and smooth, shiny surfaces being inactive. The color of the lesion can also be used to make the distinction between arrested and active, with arrested lesions acquiring internal brown pigmentation and surface stain, while active lesions retain their white appearance. The Nyvad criteria, which combine severity scoring with lesion activity assessment, have been recently validated [Nyvad et al., 2003]. The validation was based on a 3-year longitudinal study involving daily supervised brushing with fluoride dentifrice (n = 193) vs. control (n = 80). The study evaluated the relative risk of transitions (progression or regression) of the test group in relation to the control group. They found that the activity criteria were capable of reflecting their hypothesized fluoride effect (inhibition of lesion progression/enhancing lesion regression) and, thus, established construct validity. Predictive validity was established based on the finding that active, noncavitated lesions had a higher risk of progressing to a cavity than did inactive, non-cavitated lesions. A recent study [Ekstrand et al., 2005] pointed out the difficulty of trying to differentiate active lesions from inactive lesions in a single appointment without specific training or calibration. As with all clinical indices, a certain measure of uncertainty must be expected. Given the highly site-specific nature of caries, it is possible to have areas that are arrested and active on the same tooth surface. There is also the possibility of lesions being in a transitional stage, either going from active to inactive or inactive to active. The future holds promise for the development of clinically useful tools to assist dentists and researchers in making decisions about the activity status of caries lesions. Currently available technology, such as Quantitative Light Fluorescence (QLF)
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
and DIAGNOdent, may be useful for monitoring changes in lesion activity over time. QLF has the added potential for real-time assessment of caries activity status by measuring the pattern of fluorescence radiance change during dehydration [Ando et al., 2001; al-Khateeb et al., 2002]. The development of userfriendly technology to assist clinicians and researchers in real-time assessment of the activity of early lesions should be given the highest research priority. Criteria for the ICDAS Caries Lesion Activity Assessment are largely based on the Nyvad et al. [1999] system for differentiating between active and inactive caries lesions both at the non-cavitated and cavitated levels. However, the ICDAS version differs in a number of ways from the original criteria: 1) in the original system [Nyvad et al., 1999], lesion severity and active status are determined as 1 combined score; whereas, the ICDAS severity score and activity assessment are provided as 2 separate scores; 2) the Nyvad criteria are applied to initially plaque-covered teeth, while ICDAS exams are initiated on cleaned teeth, which is why ICDAS includes plaque stagnation areas as a surrogate for the presence of plaque; and 3) the Nyvad criteria for texture are determined using a sharp probe, while for the ICDAS approach, the use of a ballended probe is recommended to avoid unnecessary damage. ICDAS I and Histological Validation During development of the ICDAS I criteria in August 2002, workshop participants examined the occlusal surfaces of 57 extracted teeth. The consensus of all participants was used to define the clinical status of the occlusal surfaces. The teeth were stored in moist containers and sectioned and examined under a magnifying lens (10x). Each designated area was scored using the scale of Ricketts et al. [2002] into: 0 = No enamel demineralization 1 = Enamel demineralization limited to the outer 50% of the enamel surface 2 = Demineralisation (brown discoloration) involving between 50% of the enamel and 1/3 of the dentin 3 = Demineralization (brown discoloration) involving the middle third of the dentin 4 = Demineralization (brown discoloration) involving the inner third of the dentin
177
ISMAIL
ET AL.
The histological scoring was carried out concurrently by 2 examiners. The 2 examiners re-scored 10 teeth and agreed on 8 of the 10 scores at the second scoring. The percentage of tooth surfaces classified clinically with codes 0, 1, 2, 3, 4, and 5+6 are seen, upon sectioning, to extend into dentin and are presented in Table 1. These data support the decision of the ICDAS II workshop to switch the original codes 3 and 4 (ICDAS I) to portray a sequential progression of dental caries. The likelihood of ratios that a tooth classified with codes 2, 3, 4, or 5+6 had dental caries into dentin, relative to a tooth classified with codes 0 or 1, are presented in Table 2. These ratios show that the ICDAS II (with codes 3 and 4 switched) have an ordinal sequence in terms of histological extension into dentin. These ratios are relatively high [Goodman, 1989] compared with the likelihood ratio (LR) of standard medical signs and symptoms. For example, in relation to heart attacks, an elevation in the ST segmentation on an electrocardiogram (ECG) has an LR of 11.2; while radiating pain to both arms has an LR of 7.1 [Panju et al., 1998].
Clinical Code 0 1 2 3 4 5+6 Total
Number of Teeth 2 11 18 8 13 5 57
Percentage in Dentin 0% 9% 50% 88% 77% 100%
TABLE 1.
Percentage of tooth surfaces classified using the ICDAS by histological caries status.
Histological 0 1 2 3 4 Total LR [0-1)
0 1 1
1 0 10 1
11
Clinical [ICDAS I) 2 3 2 0 7 1 8 3 1 1 3 18 8 6.5 11.4
4 0 3 7 1 2 13 10.0
5 0 0 1 1 3 5 13.0
Number 3 22 19 4 9 57
TABLE 2.
Likelihood ratios of a ICDAS-classified teeth having caries in dentin. INDIANA CONFERENCE 2005
178
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Ekstrand and colleagues also investigated the relationship between the ICDAS I 7-point classification system when applied to the occlusal, free smooth, and approximal surfaces of extracted posterior teeth. The results, using the ICDAS I system, are cross-tabulated with the original histological scoring system (Ekstrand, et al., 1997). A strong relationship was found between the 2 variables for occlusal, free smooth, and approximal surfaces (Spearman correlation coefficients = 0.93, 0.95 and 0.94, respectively). Similarly, for the second examiner, the correlation coefficients were 0.87, 0.96, and 0.92, respectively. The LR+ ratio (positive likelihood ration) that an approximal lesion classified with ICDAS codes 3-6 is in dentin is around 18. Reliability of the ICDAS: Coronal Caries Ismail et al. [2005] has collected data on the training of examiners in the Detroit Dental Health Project. The study found good to very good inter-examiner reliability among dentists who were trained over a period of 1 week. The kappa coefficients for inter-examiner agreement ranged between 0.74 and 0.88. The intra-examiner kappa coefficients for the 2 main examiners were around 0.78. One secondary examiner had an intra-examiner reliability of 0.77, and a fourth secondary examiner, who worked only on Saturdays, had an intra-examiner kappa of 0.50. Details on reliability analysis using log-linear modeling are presented in a separate paper [Ismail et al., 2005]. For CARS, the inter-examiner kappa coefficient ranged between 0.33 for 1 examiner and more than 0.80 for the 2 main examiners. The main examiners had an intra-examiner reliability of 0.80. Ekstrand [unpublished] reported that intra-examiner kappa coefficients, when examining extracted teeth using the ICDAS I, was substantial (Kappa = 0.87). The inter-examiner reliability was around 0.80. Data from various studies at Indiana University have shown the ICDAS criteria to be a reliable and effective tool for various applications. It has been successfully applied in different types of studies, in vitro studies and clinical studies (validation study, secondary caries, epidemiology, study on caries risk factors, and clinical trial), in different dentitions (primary and permanent teeth), in different age groups (children, teenagers, young adults, adults), and by multiple examiners with different backgrounds as well as previous exposure and experience with the criteria.
179
ISMAIL
Study
ET AL.
Intra-examiner Agreement Kappa WK 0.620.90 Inter-examiner Agreement Kappa 0.500.62 WK 0.640.74 Activity, Intra- ex Kappa n/a 2 days; didactic, extracted teeth, subjects 1 day; didactic, subjects Training
Calibration of occlusal caries, 5 examiners Additional calibration of the same group, 4 examiners Calibration of 3 examiners, internal at OHRI Validation study, (primary caries in primary teeth) Secondary caries Dose-response study (occlusal surfaces) In-vitro calibration of 30 faculty, graduate students, & students, 60 teeth*
0.620.81
0.640.84
0.780.91
0.510.71
0.650.81
0.710.89
0.580.69
0.650.77
0.400.45
0.550.58
n/a
1 1/2 days; didactic, subjects
0.72
0.81
n/a
n/a n/a
n/a n/a n/a n/a
n/a n/a n/a n/a
0.73 0.61
0.76 0.75
n/a n/a
n/a n/a
0.610.69
0.800.84
0.340.38**
0.550.70**
0.620.65
3 hours; didactic, extracted teeth
*Scores are averages for each of three groups; WK = weighted kappa. **Agreement of participants with histological evaluation.
TABLE 3.
Reliability of ICDAS in various studies at Indiana University.
Several training and calibration studies have been conducted in Indiana and co-operative sites. Reliability from the various studies is presented in Table 3. The ICDAS criteria were used in a project in Mexico, where caries risk factors and indicators measured in 5 rural village populations, were correlated with caries prevalence. Intra-examiner reliability gave weighted kappa of 0.93 [Cook et al., unpublished data]. ICDAS and the International Context As evidence underpinning caries detection and activity is international, so has the ICDAS Committee ensured that its search for evidence and its outlook is also international. In Europe, as part of the Community Action Programme on Health Monitoring of the European Commission, a project on Health
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Surveillance in Europe ran from 2002-2005. This European Oral Health Global Indicators Development Project has adopted ICDAS criteria for its proposed European-wide indicators of caries severity. The move recognizes that, as the focus of public health planning embraces evidence-based healthcare moves away from providing only restorative interventions (fillings) and moves towards the delivery and evaluation of preventive programs and services, oral health indicators are needed which can be used to document the need for and the degree of success achieved in controlling early stage decay through prevention and meeting a continuing need to assess the pattern of restorative care that is provided for decay, which has progressed to the more severe stages of the disease process. The recommended indicator for dental caries severity now provides the necessary flexibility to record at different stages of the caries process, according to public health and clinical need. DMFT can, therefore, now be recorded at the early clinical stage of decay, enamel and dentine caries (at the D1 level, as in the ICDAS Method), or, where it is sensible to collect data which records only the later stages of decay, this is done using dentin-only clinical caries (at the D3 level as in WHO Basic Methods). It should be appreciated that data collected at the D1 threshold can be reported at either the D1 or the D3 level. The ICDAS criteria are also supported by the Epidemiology Special Interest Group of the European Association for Dental Public Health (EADPH) and have been discussed by the Council of European Chief Dental Officers (CECDO). At the dental practice level, discussions are ongoing with the Federation Dentaire Internationale (FDI). ICDAS I has already been piloted in a number of countries in addition to the USA and the UK. These include Copenhagen, Columbia, Mexico, and Iceland (as part of a National Survey of child dental health). Currently, there are formal requests to use the ICDAS caries detection criteria from Germany, Portugal, Italy, Thailand, Peru, and Austria. As the problems of communicating information about caries between epidemiology, research, clinical practice, and education interests are truly global, the Committee hopes that the ICDAS methodology may find widespread applications. The Future of ICDAS The mission of ICDAS is to provide a foundation for inclusion of other social and biological measures of dental caries. This foundation allows
181
ISMAIL
ET AL.
Adaptation of WHO Stepwise approach to Surveillance of Non Communicable Diseases for use with Oral Health Indicators
Future Tech & Caries Activity Assessments
ree
Biochemical Mea surements + Diagnostic Aids

QoL
Th Step
Phys
Dentine & Enamel
ical
D1
Me a
Optional
Qu est
OHIP
sure me n
Pain
Caries in Dentine D3
ts
On e
Step
Two
Expanded
ion na ir
eD ata
p Ste
Core
FIG. 3:
Adaptation of the World Health Organization (WHO) stepwise approach to surveillance of non-communicable diseases to enable an appropriate selection and integration of oral health indicators.
Adapted from Pitts NB: ICDASan international system for caries detection and assessment being developed to facilitate caries epidemiology, research, and appropriate clinical management. Community Dent Health 2004;21:193-198.
researchers and clinicians to choose the stage of disease and characteristics for assessment. Using the World Health Organisation (WHO) Stepwise or STEPS approach, we have identified indicators of dental caries that may be measured now or in the future (fig. 3). The STEPS approach allows a logical organization of the different and often disparate indicators used into a series of core indicators that can be used at STEP 1, 2, or 3, depending on the circumstances and local needs, preferences, and resources. This approach also documents how each STEP can be supplemented into an expanded form, when needed, and also identifies a series of standardised optional indicators that could be added as needed, when needed, or when they can be afforded. This philosophy is entirely consistent with the wardrobe approach of ICDAS, and its use would result in improved comparability of data collected nationally and internationally and, thereby, facilitates systematic reviews in the area. It would seem wise that, in moving forward in the area of dental caries, the 3 elements discussed above relating to epidemiology/dental public health, to clinical research, and to clinical practice should not be seen as competitive with each other or with continuing public health initiatives being mounted upstream from individual patients at the population level. The future should be better informed by improved com182 INDIANA CONFERENCE 2005
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
munication among these related fields of activity and by sound planning and evaluations based on valid epidemiological data. In parallel, systematic reviews of high-quality clinical research should inform appropriate, evidence based, clinical practice and preventive care delivered to well informed and involved patients. The future of ICDAS depends on acceptance of the concepts of integration and utility within a caries detection and assessment system. We cannot, as a community and a scientific discipline, rely on ad hoc and unreplicated methods of testing measurement systems. We cannot solely rely on clinically irrelevant gold standards, such as histological validation. Hence, the future of ICDAS depends on adaptive confidence of the cariology community in critically researching and modifying a common system for measurement and on further research. The participants at the ICDAS II workshop in Baltimore identified the following research areas for the field of detection and the assessment of dental caries: 1. Conduct multi-center studies to evaluate the validity and reliability of ICDAS, caries activity indicators, and other diagnostic tools. 2. Test the feasibility and reliability of using ICDAS in detecting caries on primary teeth. 3. Investigate different methods for effectively cleaning and drying teeth and their impact on the usability of ICDAS. 4. Develop and test new explorers to allow for the detection of surface roughness or tackiness of root surfaces without causing damage to the surface. 5. Define the appropriate time required to dry teeth to identify the first visible signs of dental caries. 6. Validate the decision tables for clinical, radiographic, and other detection tools that were developed at the ICDAS Workshop (fig. 4). 7. Define and validate the treatment decisions table (fig. 4) defined by the workshop participants. 8. Develop and test clinical and other measures to assess caries activity. For root caries, the workshop recommends the following research agenda: 1. Demonstrate in vitro validity of the root caries criteria used in the ICDAS system.
183
ISMAIL
ET AL.
FIG. 4:
Decision table for the ICDAS system.

Key: p = progressing, a = arresting, r = remineralizing; h = high risk, m = medium risk, I = low risk; PCA = Preventive Care Advised; OCA = Operative Care Advised.
2. Demonstrate in vivo reliability of the root caries criteria used in the ICDAS system. 3. Investigate the feasibility and reproducibility of dental examiners using a rounded probe to detect root caries and assess root caries activity. Compare the use of a rounded probe for root caries detection and assessmentcontrast this with the gentle use of a sharp probe or the application of visual detection and assessment methods only. 4. Establish the optimal time and method to be used for drying a tooth surface, while preserving lesion characteristics in the detection and assessment of root caries using ICDAS criteria. Additionally, the participants identified the need for the following supporting resources: 1. A library of images to depict the different codes and conditions related to ICDAS.
184
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
2. Statistical protocols for analysis of reliability data and analysis of the ICDAS system in clinical and epidemiological studies. 3. Standardized protocols and online simulations to train examiners to use ICDAS. Finally, the ICDAS coordinating committee wishes that, soon the dental community would be able to detect, assess, and decide on caries diagnosis and management using the most current scientific evidence. Figure 4 depicts the future integration of a wider range of detection and analysis systems. We hope that evidence will become available to complete the matrix. Acknowledgement The ICDAS II Baltimore Workshop was sponsored by the National Institute of Dental and Craniofacial Research and the American Dental Association. The participants in the workshop contributed to the ideas and concepts described in this paper. Their names are listed in the Appendix of the Criteria Manual.
References
al-Khateeb S, Exterkate RAM, de Josselin de Jong E, Angmar-Mnsson B, ten Cate JM: Light-induced fluorescence studies on dehydration of incipient enamel lesions. Caries Res 2002;36:25-30. Ando M, Zero DT, Eckert GJ, Stookey GK: Pattern of fluorescence intensity during dehydration as determined by Quantitative Light-induced Fluorescence in vitro. Caries Res 2001; 35:270 Abstr 16. Ando M, Gonzlez-Cabezas C, Isaacs RL, Eckert GJ, Stookey GK: Evaluation of several techniques for the detection of secondary caries adjacent to amalgam restorations. Caries Res 2005;38:350-56. Backer Dirks O: The clinical testing of agents for the prevention of dental caries. Advances in Fluorine Res Dent Caries Prevention. 1966;4:1-2. Bader JD, Shugars DA, Bonito AJ: Systematic review of selected dental caries diagnostic and management methods. J Dent Educ 2001;65:960-968. Banting DW: Diagnosis and prediction of root caries. Adv Dent Res 1993;7:80-86. Banting DW: The diagnosis of root caries. J Dent Educ 2001;65:991-996. Beighton D, Lynch E, Heath MR: A microbiological study of primary root caries lesions with different treatment needs. J Dent Res 1993;63:623-639. Bergman G, Lindn LA: The action of the explorer on incipient caries. Svensk Tandlkare Tidsskrift 1969;62:629-634.
185
ISMAIL
ET AL.
Billings RJ, Brown LR, Kaster AG: Contemporary treatment strategies for root surface dental caries. Gerodontics 1985;1:20-27. Boyd MA, Richardson AS: Frequency of amalgam replacement in general dental practice. J Can Dent Asso 1985;10:763-766. Bratthall D: A Streptococcus mutans Safari. J Dent Res 1997;76:1332-1336. Chesters RK, Pitts NB, Matuliene G, Kvedariene A, Huntington E, Bendinskaite R, Balciuniene I, Matheson J, Savage D, Milerience J: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002; 81:637640. Drugan C: Report of a BASCD Dental Epidemiology Workshop, London May 16th 2003: Design of future BASCD coordinated oral health surveys to meet the information needs of the modernised NHS. Community Dent Health 2004;21:5457. Ekstrand K, Qvist V, Thylstrup A: Light microscope study of the effect of probing in occlusal surfaces. Caries Res 1987;21:363-374 Ekstrand KR, Kuzmina I, Bjorndal L, Thyrlstrup A: Relationship between external and histologic features of progressive stages of caries in the occlusal fossa. Caries Res 1995;29:243-250. Ekstrand KR, Ricketts DN, Kidd EA: Reproducibility and accuracy of three methods for assessment of demineralization depth of the occlusal surface: an in vitro examination. Caries Res 1997;31:224-231. Ekstrand KR, Ricketts DN, Kidd EA: Occlusal caries: pathology, diagnosis and logical management. Dent Update 2001;28:380-387. Ekstrand KR, Ricketts DNJ, Longbottom C, Pitts NB: Visual and tactile assessment of arrested initial enamel carious lesions: an in vivo pilot study. Caries Res 2005;39:173-77. Elderton RJ: Variability in the decision making process and implications for change towards a preventive philosophy. In Anusavice KJ (ed): Quality Evaluation of Dental Restorations: Criteria for Placement and Replacement. Quintessence Publishing Co. 211-219, 1989. Fyffe HE, Deery CH, Nugent, ZJ, Nuttall NM, Pitts NB: Effect of diagnostic threshold on the validity and reliability of epidemiological caries diagnosis using the Dundee Selectable Threshold Method for caries diagnosis (DSTM). Community Dent Oral Epidemiol 2000;28: 42-51. Goldberg J, Tanzer J, Munster E, Amara J, Thal F, Birkhed D: Cross-sectional clinical evaluation of recurrent enamel caries, restoration of marginal integrity, and oral hygiene status. J Am Dent Assoc 1981;102:635-641. Goldberg AJ: Deterioration of restorative materials and the risk for secondary caries. Adv Dent Res 1990;4:14-18. Goodman SN: Meta-analysis and evidence. Controlled Clinical Trials 1989;10:188-204. Hamilton JC, Moffa JP, Ellison JA, Jenkins WA: Marginal fracture not a predictor of longevity for two dental amalgam alloys: a ten year study. J Prost Dent 1983;50:200-202. Hellyer PH, Beighton D, Heath MR, Lynch EJR: Root caries in older people attending a general practice in East Sussex. Brit Dent J 1990;169:201-206. Hellyer P, Lynch E: The diagnosis of root caries. Gerodontol 1991;9:95-102.
186
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Hix JO, OLeary TJ: The relationship between cemental caries, oral hygiene status and fermentable carbohydrate intake. J Periodontal 1976;47:394-404. Ismail AI, Tellez M, Sohn W, Sen A: Reliability of the International Caries Detection and Assessment System (ICDAS). Community Dent Oral Epidemiol, 2005. Ismail AI: Visual and Visuo-tactile Detection of Dental Caries. J Dent Res 2004a;83(Spec. Issue C):C56-C66. Ismail AI: Diagnostic levels in dental public health planning. Caries Res 2004b;38:99203. Jorgensen KD, Wakumoto S: Occlusal amalgam fillings: marginal defects and secondary caries. Odontol Tidsk 1968;76:43-54. Katz RV: Development of an index for the prevalence of root caries. J Dent Res 1984;63:814-818. Kidd EAM: Caries diagnosis within restored teeth. Oper Dent 1989a;14:149-158. Kidd EAM: Caries diagnosis within restored teeth, in Anusavice KJ (ed): Quality evaluation of dental restorations: Criteria for placement and replacement. Chicago, Quintessence Publishing Co Inc:111-123, 1989b. Kidd EAM: Caries diagnosis within restored teeth. Adv Dent Res 1990;4:10-13. Kidd EAM, Beighton D: Prediction of secondary caries around tooth-colored restorations: a clinical and microbiological study. J Dent Res 1996;75:1942-1946. Kidd EAM, OHara JW: The caries status of occlusal amalgam restorations with marginal defects. J Dent Res 1990;69:1275-1277. Kidd EAM, Toffenetti F, Mjr IA: Secondary caries. Int Dent J 1992;42:127-138. Kidd EAM, Joyston BS, Beighton D: Diagnosis of secondary caries: a laboratory study. Br Dent J 1994;176:135-139. Kidd EAM, Joyston BS, Beighton D: Marginal ditching and staining as a predictor of secondary caries around amalgam restorations: a clinical and microbiological study. J Dent Res 1995;74:1206-1211. Leake JL: Clinical decision-making for caries management in root caries. J Dent Educ 2001;65:1147-1153. Lynch E, Beighton D: A comparison of primary root caries lesions classified according to color. Caries Res 1994;28:233-239. Miller WA, Massler M: Permeability and staining of active and arrested lesions in dentine. Br Dent J 1962;112:187-197. Mjr IA, Toffenetti F: Secondary caries: a literature review with case reports. Quintessence Int 2000;31:165179. National Institutes of Health: The diagnosis and management of dental caries throughout life. National Institutes of Health Consensus Development Conference, Washington DC, March 26th28 th 2001. J Dent Educ 63, No. 10. Nyvad B, Fejerskov O: Assessing the stage of caries lesion activity on the basis of clinical and microbiological examination. Community Dent Oral Epidemiol 1997;25:69-75.
187
ISMAIL
ET AL.
Nyvad B, ten Cate JM, Robinson C: Cariology in the 21st centurystate of the art and future perspectives. Caries Res 2004;38,167329. Nyvad B, Machiulskiene V, Baelum V: Reliability of a new caries diagnostic system differentiating between active and inactive caries lesions. Caries Res 1999;33:252-260. Nyvad B, Machiulskiene V, Baelum V: Construct and predictive validity of clinical caries diagnostic criteria assessing lesion activity. J Dent Res 2003;82:117-122. Ozer L: The relationship between gap size, microbial accumulation and the structural features of natural caries in extracted teeth with Class II amalgam restorations. 1997. University of Copenhagen. Panju AA, Hemmelgarn BR, Guyatt GH, Simel DL: The rational clinical examination. Is this patient having a myocardial infarction? JAMA 1998;280:1256-1263. Pitts NB: Safeguarding the quality of epidemiological caries data at a time of changing disease patterns and evolving dental services. Community Dent Health 1993;10:19. Pitts NB: Discovering Dental Public Health: from Fisher to the future. Community Dent Health 1994;11:172178. Pitts NB: Modern concepts of caries measurement. J Dent Res 2004a;83:4347. Pitts NB: Are we ready to move from operative to non-operative/preventive treatment of dental caries in clinical practice? Caries Res 2004;38:294304. Pitts NB, Stamm J: International Consensus Workshop on Caries Clinical Trials (ICW-CCT) Final Consensus Statements: Agreeing Where the Evidence Leads. J Dent Res 2004;83:125128. Ricketts DNJ, Ekstrand KR, Kidd EAM, Larsen T: Relating visual and radiographic ranked scoring systems for occlusal caries detection to histological and microbiological evidence. Operative Dent 2002;27:231-237. Rudolph MP, van Amerongen JP, Penning C, ten Cate JM: Grey discoloration and marginal fracture for the diagnosis of secondary caries in molars with occlusal amalgam restorations: an in vitro study. Caries Res 1995;29:371-376. Rudolphy MP, van Loveren C, van Amerongen JP: Grey discoloration for the diagnosis of secondary caries in teeth with Class II amalgam restorations: an in vitro study. Caries Res 1996;30:189-193. Russell AL: The differential diagnosis of fluoride and nonfluoride enamel opacities. J Public Health Dent 1961;21:143-146. Stookey G: (Ed.). Early Detection of Dental Caries: Proceedings of the 1st Annual Indiana Conference, Indianapolis, Indiana University School of Dentistry, 1996. Stookey G: (Ed.). Early Detection of Dental Caries: Proceedings of the 2nd Annual Indiana Conference, Indianapolis, Indiana University School of Dentistry, 2000. Stookey G: (Ed.) Early Detection of Dental Caries: Proceedings of the 3rd Annual Indiana Conference, Indianapolis, Indiana University School of Dentistry, 2004. Tobi H, Kreulen CM, Vondeling H, van Amerongen WE: Cost-effectiveness of composite resins and amalgam in the replacement of amalgam Class II restorations. Community Dent Oral Epidemiol 1999;27:137-143.
188
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Topping GVA: Secondary caries misdiagnosis: an in vitro study in premolar and molar teeth restored with amalgam and conjoint analysis of patients and dentists preferences for attributes of a caries diagnosis device. University of Dundee, 2001. U.S. Department of Health and Human Services, Oral Health of United States Adults: NIH Publication No. 87-2868, 1987. Warren JJ, Levy SM, Wefel JS: Explorer probing of root caries lesions. Spec Care Dent 2003;23:18-21. World Health Organization: Oral health surveys: basic methods. 4th ed. Geneva: World Health Organization, 1997. Zero D, Fontana M, Lennon M: Clinical applications and outcomes of using indicators of risk in caries management. J Dent Educ 2001;65:1126-1132.
189
ISMAIL
ET AL.
Appendix I Criteria Manual

International Caries Detection and Assessment System (ICDAS II) Workshop held in Baltimore, Maryland, March 12-14, 2005 Sponsored by: The National Institute of Dental and Craniofacial Research, the American Dental Association, and the International Association for Dental Research *Author: International Caries Detection and Assessment System Coordinating Committee *Authorship of this report should be cited as follows: International Caries Detection and Assessment System (ICDAS) Coordinating Committee. Committee Members: D. Banting K.R. Ekstrand A.I. Ismail (co-chair) N.B. Pitts (co-chair) D. Ricketts W. Sohn D. Zero H. Eggertsson A. Ferreira Zandon C. Longbottom E. Reich R. Selwitz G.V. Topping (coordinator)
This report summarizes the key decisions and clinical criteria discussed by participants (Appendix) in ICDAS II, which was held in Baltimore, Maryland, USA, on March 12 through 14, 2005. The workshop was funded by the National Institute of Dental and Craniofacial Research (NIDCR) and the American Dental Association (ADA). The International Association for Dental Research (IADR) provided administrative support for the workshop. The objective of the workshop was to develop consensus on clinical caries detection criteria among experts in cariology, clinical research, restorative dentistry, pediatric dentistry, public health, biological sciences, and dental organizations. This goal was achieved by the end of the workshop. Additionally, the
190
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
participants: (1) defined the stages of the caries process that can portray the concept of demineralization at the non-cavitated stage and the caries process, overall; and (2) defined clinically relevant validation methods and a research agenda for the newly developed detection system. No definitive conclusion was reached regarding how to measure caries activity, and research of this important concept will continue. The final outcome of the workshop was the revision of the ICDAS criteria developed in 2002. The new criteria for the detection and assessment of dental caries will be referred to as ICDAS II. The workshop participants concluded their deliberation by recognizing that the ICDAS system will continue to evolve as new information and tools are developed and validated. ICDAS II presents a foundation upon which new caries assessment tools can be embedded to aid in making more accurate decisions for clinical practice as well as for clinical and epidemiological research. The ICDAS II system strives to achieve integration and coordination of the emerging field of caries assessment.
Trial Demographics and Attrition Overview The ICDAS detection codes for coronal caries range from 0 to 6, depending on the severity of the lesion. There are minor variations between the visual signs associated with each code, depending on a number of factors, including the surface characteristics (pits and fissures versus free smooth surfaces), whether there are adjacent teeth present (mesial and distal surfaces), and whether or not the caries is associated with a restoration or sealant. Therefore, a detailed description of each of the codes is given under the following headings to assist in training examiners in the use of ICDAS: Pits and fissures; smooth surface (mesial or distal); free smooth surfaces; and caries associated with restorations and sealants (CARS). However, the basis of the codes is essentially the same throughout. Code Description 0 Sound 1 First Visual Change in Enamel (seen only after prolonged air drying or restricted to within the confines of a pit or fissure)
191
ISMAIL
ET AL.
Code Description (cont.) 2 Distinct Visual Change in Enamel 3 Localized Enamel Breakdown (without clinical visual signs of dentinal involvement) 4 Underlying Dark Shadow from Dentin 5 Distinct Cavity with Visible Dentin 6 Extensive Distinct Cavity with Visible Dentin Coronal Primary Caries Codes Pits and Fissures Sound tooth surface: Code 0 There should be no evidence of caries (either none or a questionable change in enamel translucency after prolonged air drying [suggested drying time 5 seconds]). Surfaces with developmental defects, such as enamel hypoplasias; fluorosis; tooth wear (attrition, abrasion, and erosion), and extrinsic or intrinsic stains, will be recorded as sound. The examiner should also score as sound a surface with multiple stained fissures, if such a condition is seen in other pits and fissures, a condition which is consistent with non-carious habits (e.g., frequent tea drinking). Table 1 provides a useful guide for differential diagnosis for carious opacities versus other opacities. First visual change in enamel: Code 1 Code 1 is assigned for the following pits and fissures: When seen wet, there is no evidence of any change in color attributable to carious activity but, after prolonged air drying (approximately 5 seconds is suggested to adequately dehydrate a carious lesion in enamel), a carious opacity or discoloration (white or brown lesion) is visible, which is not consistent with the clinical appearance of sound enamel OR When there is a change of color due to caries, which is not consistent with the clinical appearance of sound enamel and is limited to the confines of the pit and fissure area (whether seen wet or dry). The appearance of these carious areas is not consistent with that of stained pits and fissures as defined in Code 0. Distinct visual change in enamel: Code 2 The tooth must be viewed wet. When wet, there is an: (a) carious opacity (white spot lesion) and/or (b) brown carious discoloration, which is wider than the natural fissure/fossa that is not consistent with the clinical appearance of
RATIONALE
Characteristic Area affected Milder Forms of Fluorosis Usually seen on or near tips of cusps or incisal edges. Resembles line shading in pencil sketch; lines follow incremental lines in enamel, forms irregular caps on cusps. Shades off imperceptibly into surrounding normal enamel. Slightly more opaque than normal enamel; paper-white. Incisal edges, tips of cusps may have frosted appearance. Does not show stain at time of eruption (in these milder degrees, rarely at any time). Most frequently on teeth that calcify slowly (cuspids, bicuspids, second, and third molars). Rare on lower incisors. Usually seen on 6 or 8 homologous teeth. Extremely rare in deciduous teeth. None. Pitting of enamel does not occur in the milder forms. Enamel surface has glazed appearance, is smooth to point of explorer. Often invisible under strong light; most easily detected by line of sight tangential to tooth crown.
AND
EVIDENCE
FOR THE
ICDAS
Non-fluoride Enamel Opacities Usually centered in smooth surface; may affect entire crown. Often round or oval.
Shape of lesion
Demarcation
Clearly differentiated from adjacent normal enamel. Usually pigmented at time of eruption often creamy-yellow to dark reddishorange.
Color
Teeth Affected
Any tooth may be affected. Frequently on labial surfaces of lower incisors. May occur singly. Usually, 1 to 3 teeth affected. Common in deciduous teeth. Absent to severe. Enamel surface may seem etched, be rough to explorer.
Gross hypoplasia
Detection
Seen most easily under strong light on line of sight perpendicular to tooth surface.
Russell AL: The differential diagnosis of fluoride and non-fluoride enamel opacities. J Public Health Dent 1961;21:143-146.
TABLE 1.
Differential diagnosis between milder forms of dental fluorosis (questionable, very mild, and mild) and non-fluoride opacities of enamel.
sound enamel. (Note: the lesion must still be visible when dry). Localized enamel breakdown due to caries with no visible dentin or underlying shadow: Code 3 The tooth viewed wet may have a clear carious opacity (white spot lesion) and/or brown carious discoloration that is wider than the natural fissure/fossa, which is not consistent with the clinical appearance of sound enamel. Once dried for approximately 5 seconds, there is carious loss of tooth structure at the entrance to, or within, the pit or fissure/fossa. This will be seen visually as evidence of demineralization (opaque [white,] brown, or dark brown walls) at the entrance to or within the fissure or pit and, although the pit or fissure may appear substantially and unnaturally wider than normal, the dentin is NOT visible in the walls or base of the cavity/discontinuity.
ISMAIL
ET AL.
If in doubt, or to confirm the visual assessment, the WHO/CPI/PSR probe can be used gently across a tooth surface to confirm the presence of a cavity apparently confined to the enamel. This is achieved by sliding the ball end along the suspect pit or fissure, and a limited discontinuity is detected if the ball drops into the surface of the enamel cavity/discontinuity. Underlying dark shadow from dentin with or without localized enamel breakdown: Code 4 This lesion appears as a shadow of discolored dentin visible through an apparently intact enamel surface that may or may not show signs of localized breakdown (loss of continuity of the surface that is not showing the dentin). The shadow appearance is often seen more easily when the tooth is wet. The darkened area is an intrinsic shadow, which may appear as grey, blue, or brown in color. The shadow must clearly represent caries that started on the tooth surface being evaluated. If, in the opinion of the examiner, the carious lesion started on an adjacent surface, and there no evidence of any caries on the surface being scored, then the surface should be coded 0. Codes 3 and 4, histologically, may vary in depth, with one being deeper than the other, and vice versa. This will depend on the population and properties of the enamel. For example, more translucent and thinner enamel in primary teeth may allow the undermining discoloration of the dentin to be seen before localized breakdown of enamel. However, in most cases, Code 4 is likely to be deeper into dentin than Code 3. Distinct cavity with visible dentin: Code 5 Cavitation in opaque or discolored enamel, exposing the dentin beneath. The tooth viewed wet may have darkening of the dentin visible through the enamel. Once dried for 5 seconds, there is visual evidence of loss of tooth structure at the entrance to or within the pit or fissurefrank cavitation. There is visual evidence of demineralization (opaque [white,] brown, or dark brown walls) at the entrance to, or within, the pit or fissure and, in the examiners judgment, dentin is exposed. The WHO/CPI/PSR probe can be used to confirm the presence of a cavity apparently in dentin. This is achieved by sliding the ball end along the suspect pit or fissure, and a dentin cavity is detected if the ball enters the opening of the cavity and, in the opinion of the examiner, the base is in dentin. (In pits or fissures, the thickness of the enamel is between 0.5 and 1.0 mm. Note the deep pulpal dentin should not be probed.)
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Extensive distinct cavity with visible dentin: Code 6 Obvious loss of tooth structure, the cavity is both deep and wide, and dentin is clearly visible on the walls and at the base. An extensive cavity involves at least half of a tooth surface or possibly reaches the pulp. Smooth surface (mesial and distal) This requires visual inspection from the occlusal, buccal, and lingual directions. Sound tooth surface: Code 0 There should be no evidence of caries (either no or questionable change in enamel translucency after prolonged air drying [suggested drying time 5 seconds]). Surfaces with developmental defects, such as enamel hypoplasias, fluorosis, tooth wear (attrition, abrasion, and erosion), and extrinsic or intrinsic stains, will be recorded as sound. First visual change in enamel: Code 1 When seen wet, there is no evidence of any change in color attributable to carious activity but, after prolonged air drying, a carious opacity (white or brown lesion) is visible, which is not consistent with the clinical appearance of sound enamel. This will be seen from the buccal or lingual surface. Distinct visual change in enamel when viewed wet: Code 2 There is a carious opacity or discoloration (white or brown lesion) that is not consistent with the clinical appearance of sound enamel (Note: the lesion is still visible when dry). This lesion may be seen directly, when viewed from the buccal or lingual direction. In addition, when viewed from the occlusal direction, this opacity or discoloration may be seen as a shadow confined to enamel, seen through the marginal ridge. Initial breakdown in enamel due to caries with no visible dentin: Code 3 Once dried for approximately 5 seconds, there is a distinct loss of enamel integrity, as viewed from the buccal or lingual direction. If in doubt, or to confirm the visual assessment, the CPI probe can be used gently across the surface to confirm the loss of surface integrity. Underlying dark shadow from dentin with or without localized enamel breakdown: Code 4 This lesion appears as a shadow of discolored dentin visible through an apparently intact marginal ridge, buccal, or lingual walls of enamel. This appearance is often seen more easily when the tooth is wet. The darkened area
ISMAIL
ET AL.
is an intrinsic shadow, which may appear as grey, blue, or brown in color. Distinct cavity with visible dentin: Code 5 Cavitation in opaque or discolored enamel (white or brown), with exposed dentin, in the examiners judgment. If in doubt, or to confirm the visual assessment, the CPI probe can be used to confirm the presence of a cavity apparently in dentin. This is achieved by sliding the ball end along the surface, and a dentin cavity is detected if the ball enters the opening of the cavity and, in the opinion of the examiner, the base is in dentin. Extensive distinct cavity with visible dentin: Code 6 Obvious loss of tooth structure, the extensive cavity may be deep or wide, and dentin is clearly visible on both the walls and at the base. The marginal ridge may or may not be present. An extensive cavity involves at least half of a tooth surface or possibly reaches the pulp. Free smooth surface (buccal and lingual, and direct examination of mesial and distal surfaces [with no adjacent teeth]) Sound tooth surface: Code 0 There should be no evidence of caries (either no or questionable change in enamel translucency after prolonged air drying [approximately 5 seconds]). Surfaces with developmental defects, such as enamel hypoplasias, fluorosis, tooth wear (attrition, abrasion, and erosion), and extrinsic or intrinsic stains, will be recorded as sound. First visual change in enamel: Code 1 When seen wet, there is no evidence of any change in color that is attributable to carious activity but, after prolonged air drying, a carious opacity is visible, which is not consistent with the clinical appearance of sound enamel. Distinct visual change in enamel when viewed wet: Code 2 There is a carious opacity or discoloration that is not consistent with the clinical appearance of sound enamel (Note: the lesion is still visible when dry). The lesion is located in close proximity (in touch or within 1 mm) of the gingival margin. Localized enamel breakdown due to caries with no visible dentin: Code 3 Once dried for 5 seconds, there is carious loss of surface integrity without visible dentin.
196
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
If in doubt, or to confirm the visual assessment, the CPI probe can be used with NO digital pressure to confirm the loss of surface integrity. Underlying dark shadow from dentin with or without localized enamel breakdown: Code 4 This lesion appears as a shadow of discolored dentin visible through the enamel surface beyond the white or brown spot lesion, which may or may not show signs of localized breakdown. This appearance is often seen more easily when the tooth is wet and is a darkening and intrinsic shadow, which may be grey, blue, or brown in color. Distinct cavity with visible dentin: Code 5 Cavitation in opaque or discolored enamel exposing the dentin beneath. If in doubt, or to confirm the visual assessment, the CPI probe can be used with NO digital pressure to confirm the presence of a cavity apparently in dentin. This is achieved by sliding the ball end along the surface, and a dentin cavity is detected if the ball enters the opening of the cavity and, in the opinion of the examiner, the base is in dentin. Extensive distinct cavity with visible dentin: Code 6 Obvious loss of tooth structure, the cavity is both deep and wide, and dentin is clearly visible on the walls and at the base. An extensive cavity involves at least half of a tooth surface or possibly reaches the pulp. Figure 1 depicts a simple decision tree for applying the 7-code for classifying coronal tooth surfaces following the ICDAS criteria. Caries-Associated with Restorations and Sealants (CARS) Detection Criteria Caries Associated with Restorations and Sealants Codes Sound tooth surface with restoration or sealant: Code 0 A sound tooth surface adjacent to a restoration/sealant margin. There should be no evidence of caries (either no or questionable change in enamel translucency after prolonged air drying for 5 seconds). Surfaces with marginal defects less than 0.5 mm in width (i.e., will not admit the ball end of the CPI Probe), developmental defects, such as enamel hypoplasias, fluorosis, tooth wear (attrition, abrasion, and erosion), and extrinsic or intrinsic stains, will be recorded as sound. Stained margins consistent with non-carious habits (e.g., frequent tea drinking) and, which do not exhibit signs consistent with demineralization, should be scored as sound.
ISMAIL
ET AL.
YES More than total surface involved?
Could be code 0 or 1
Could be codes 1-6
Any carious discoloration seen when wet?

YES Is dentin exposed?
Could be codes 3.5.6
Yes Code = 6
Could be codes 1-6
YES Is there cavitation?
NO Any carious lesions when dried?
YES Is there cavitation?
Yes Code = 1
No Code = 3
No Code = 5
No Code = 0
Could be codes 1, 2, 4
Could be codes 3.5.6
Any carious discoloration seen when wet?
NO Is there shadowing?
Could be codes 1 or 2
NO Any carious lesions when dried?
Yes Code = 4
NO Is there shadowing?
Could be codes 1, 2, 4
YES Is dentin exposed?
FIG. 1:
Decision table for the ICDAS system.

Key: p = progressing, a = arresting, r = remineralizing; h = high risk, m = medium risk, I = low risk; PCA = Preventive Care Advised; OCA = Operative Care Advised.
First visual change in enamel: Code 1 When seen wet, there is no evidence of any change in color attributable to carious activity but, after prolonged air drying (for approximately 5 seconds), an opacity or discoloration consistent with demineralisation is visible, which is not consistent with the clinical appearance of sound enamel. Distinct visual change in enamel/dentin adjacent to a restoration/sealant margin: Code 2 If the restoration margin is placed on enamel, the tooth must be viewed wet. When wet, there is an opacity consistent with demineralisation or discoloration that is not consistent with the clinical appearance of sound enamel (Note: the lesion is still visible when dry). If the restoration margin is placed on dentin: Code 2 applies to discoloration that is not consistent with the clinical appearance of sound dentin or cementum.
No Code = 0
Yes Code = 1
No Code = 1
No Code = 1
Yes Code = 2
Could be codes 1 or 2
NO Extends beyond pit/fissure?
Could be code 0 or 1
Yes Code = 2
NO Extends beyond pit/fissure?
Yes Code = 4
No Code = 3
YES More than total surface involved?
No Code = 5
Yes Code = 6
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Carious defects of <0.5 mm with the signs of Code 2: Code 3 Cavitation at the margin of the restoration/sealant less than 0.5 mm, in addition to either an opacity or discoloration consistent with demineralisation, which is not consistent with the clinical appearance of sound enamel or with a shadow of discolored dentin. Marginal caries in enamel/dentin/cementum adjacent to restoration/sealant with underlying dark shadow from dentin: Code 4 The tooth surface may have characteristics of Code 2 and has a shadow of discolored dentin, which is visible through an apparently intact enamel surface, or with localized breakdown in enamel, but no visible dentin. This appearance is often seen more easily when the tooth is wet and is a darkening and intrinsic shadow that may be grey, blue, orange, or brown in color. Note: view tooth wet and then dry. This lesion should be distinguished from amalgam shadows. Distinct cavity adjacent to restoration/sealant: Code 5 Distinct cavity adjacent to restoration/sealant, with visible dentin in the interfacial space, with signs of caries as described in Code 4, in addition to a gap > 0.5 mm in width. OR In those instances where margins are not visible, there is evidence of discontinuity at the margin of the restoration/sealant and tooth substance of the dentin as detected by a 0.5 mm ball-ended probe running along the restoration/sealant margin. Extensive distinct cavity with visible dentin: Code 6 Obvious loss of tooth structure, the extensive cavity may be deep or wide, and dentin is clearly visible on both the walls and at the base. ICDAS 2-digit Coding Method A 2-number coding system is suggested to identify restorations/sealants, with the first digit, followed by the appropriate caries code; for example, a tooth restored with amalgam, which also exhibited an extensive distinct cavity with visible dentin, would be coded 4 (for an amalgam restoration), 6 (distinct cavity), an unrestored tooth with a distinct cavity would be 06. The suggested restoration/sealant coding system is as follows: 0 = Sound: i.e., surface not restored or sealed (use with the codes for primary caries) 1 = Sealant, partial
ISMAIL
ET AL.
2 3 4 5 6 7 8 9
= = = = = = = =
Sealant, full Tooth colored restoration Amalgam restoration Stainless steel crown Porcelain or gold or PFM crown or veneer Lost or broken restoration Temporary restoration Used for the following conditions: 96 = Tooth surface cannot be examined: surface excluded 97 = Tooth missing, because of caries (tooth surfaces will be coded 97) 98 = Tooth missing for reasons other than caries (all tooth surfaces will be coded 98) 99 = Unerupted (tooth surfaces coded 99)
Special Considerations 1. In case of doubt, the examiner should score low. 2. It may be necessary to distinguish among unerupted teeth, teeth extracted because of caries, and those extracted or missing for other reasons. 3. Non-vital teeth should be scored in the same manner as vital teeth. 4. Banded or bracketed teeth. All visible surfaces should be examined as well as possible and scored in the usual manner. When a surface is completely covered by a band or bracket and there is no evidence of caries, the tooth status code is 0. 5. In the case of supernumerary teeth, the examiner should decide which tooth is the legitimate occupant of the space. Only that tooth should be scored. 6. When both a primary and permanent tooth occupy the same space, only the permanent tooth is coded. 7. All surfaces restored with full coverage should be coded as crowned. If a tooth has been restored with anything less than full coverage, the surfaces involved in the restoration will be scored separately. 8. If part of a restoration is lost on a surface, the surface should be coded as 7 (first number), even when not all the restoration is missing. 9. It is important that there is a code to record the instances where there are non-carious cavities; i.e., where a restoration has been lost. It could
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
be argued that such cases are analogous with temporary restorations, although it is the convention, in some epidemiological studies, to record these in a way that means that they are recorded within the filled, rather than the decayed element of the study findings. 10. Where more than one carious lesion exists on a surface, the worst lesion should be scored, though scoring pits and fissures separately to free smooth surfaces is an option. 11. If a pit or fissure on an occlusal surface is not included in a distinct shadow originating from the mesial or distal surface, then the occlusal surface should be scored as sound. However, in all other instances, the examiner should not determine the surface origin of a carious lesion, and each tooth surface should be scored separately as it appears. A tooth surface is bounded by the line angle, when viewed in a perpendicular direction. 12. For determining whether there is an enamel cavity (Code 3), the ball point of the CPI probe should detect a ditch on a tooth surface that partially covers the ball end of the probe. If all of the ball end of the probe can enter the ditch, then the area should coded as 5, unless the examiner concludes that the lesion is in enamel, then the code is a 3. 13. A shadow underneath a marginal ridge or surrounding a pit or fissure must be distinct and colored grey before it is classified with Code 4. 14. Whenever both the coronal and root surface are affected by a single carious lesion that extends at least 1 mm or more past the CEJ in both the cervical-incisal and cervical-apical directions, both surfaces should be scored separately. For a lesion affecting both crown and root surfaces, with extension from the CEJ of less than 1 mm, only that surface of tooth with the greater portion (more than 50%) of the lesion involvement should be scored. When it is impossible to invoke the 50% rule (i.e., when both the coronal and root surfaces appear equally affected), both surfaces should be scored as carious. 15. A root surface adjacent to a crown margin that is free of decay should be scored sound. 16. If more than one lesion is present on the same root surface, the most severe lesion is scored. 17. All tooth surfaces of retained roots should be scored as (06).
201
ISMAIL
ET AL.
Root Caries Criteria Codes for the detection and classification of carious lesions on the root surfaces One score will be assigned per root surface. The facial, mesial, distal, and lingual root surfaces of each tooth should be classified as follows: Code E If the root surface cannot be visualized directly as a result of gingival recession or by gentle air-drying, then it is excluded. Surfaces covered entirely by calculus can be excluded or, preferably, the calculus can be removed prior to determining the status of the surface. Removal of calculus is recommended for clinical trials and longitudinal studies. Code 0 The root surface does not exhibit any unusual discoloration that distinguishes it from the surrounding or adjacent root areas, nor does it exhibit a surface defect either at the cemento-enamel junction or wholly on the root surface. The root surface has a natural anatomical contour, OR The root surface may exhibit a definite loss of surface continuity or anatomical contour, which is not consistent with the dental caries process. This loss of surface integrity usually is associated with dietary influences or habits, such as abrasion or erosion. These conditions usually occur on the facial surface. These areas typically are smooth, shiny, and hard. Abrasion is character-
Can the root surface be visualized directly?
No
Yes
After 5 seconds of air-drying, is there a color change present? (discoloration = light/dark brown, black)
No
Yes
no caries
Is cavitation present? (loss of anatomical contour => 0.5 mm)
No
Yes
non-cavitated root caries
cavitated root caries
FIG. 2: 202
Decision tree for primary caries on the root surface. INDIANA CONFERENCE 2005
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
ized by a clearly defined outline, with a sharp border; whereas, erosion has a more diffuse border. Neither condition shows discoloration. Code 1 There is a clearly demarcated area on the root surface or at the cementoenamel junction (cej) that is discoloured (light/dark brown, black), but there is no cavitation (loss of anatomical contour 0.5 mm) present. Code 2 There is a clearly demarcated area on the root surface or at the cementoenamel junction (cej) that is discoloured (light/dark brown, black), and there is cavitation (loss of anatomical contour ? 0.5 mm) present. Figure 2) serves as a useful prompt for examiners in deciding on appropriate coding of root caries. Caries Associated with Root Restorations When a root surface is filled and there is caries adjacent to the restoration, the surface is scored as caries. The criteria for caries associated with restorations on the roots of teeth are the same as those for caries on non-restored root surfaces. Figure 3 will assist the examiner in deciding on the appropriate coding of caries adjacent to restorations on root surfaces. Root Caries Activity The characteristics of the base of the discolored area on the root surface can be used to determine whether or not the root caries lesion is active. These characteristics include texture (smooth, rough), appearance (shiny or glossy, matte or non-glossy), and perception on gentle probing (soft, leathery, hard). Active root caries lesions are usually located within 2 mm of the crest of the gingival margin. Figure 4 will be helpful in making a determination regarding the activity of root caries. Special Considerations Whenever both a coronal and root surface are affected by a single carious lesion that extends at least 1 mm past the CEJ in both the incisal and apical directions, both surfaces should be scored as caries. However, for a lesion affecting both crown and root surfaces that does not meet the 1 mm or greater extent of involvement, only the coronal or root surface that involves the greater
ISMAIL
ET AL.ISMAIL ET AL.
Is there a color change adjacent to the root restoration? (discoloration = light/dark brown, black)
No
Yes
Is there cavitation present? (loss of anatomical contour => 0.5mm)
No
Yes
FIG. 3:
Decision tree for caries associated with root restorations.
For both Code
1 and Code 2
What is the texture and appearance of the base of the discolored area?
ARRESTED
Smooth, Shiny
Rough, Matte
What is the sensation on gentle probing?
QUIESCENT
FIG. 4: Decision tree for root caries activity.
Leathery
ACTIVE
Soft
portion (more than 50%) of the lesion should be scored as caries. When it is impossible to invoke the 50% rule (i.e., when both coronal and root surfaces appear equally affected), both surfaces should be scored as caries. When a carious lesion on a root surface extends beyond the line angle of the root to involve at least 1/3 of the distance across the adjacent surface, that adjacent surface also should also be scored as caries. If more than one lesion is present on the same root surface, the most severe lesion is scored. Non-vital teeth are scored the same as vital teeth.
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Preliminary Plan for ICDAS Caries Lesion Activity Assessment Working Definitions An Active Lesion is considered to have a greater likelihood of transition (progress, arrest, or regress) than an inactive lesion. An Inactive (arrested) Lesion is considered to have a lesser likelihood of transition than an active lesion. Clinical observations to be taken into consideration for assessing enamel lesion activity are based on a modification of the Nyvad et al. [1999] caries lesion activity assessment criteria and include visual appearance, tactile feeling, and potential for plaque accumulation. Recommendations for Examiner Training The ICDAS Committee recommends the following training program: 1. One half-day of slide presentations and discussions of the ICDAS codes and protocol for examination. 2. At least 2 days of examiner training, which will include examination of a set of subjects providing balanced numbers of tooth surfaces with ICDAS Codes 1-5. The examination findings of all examiners should be reviewed to identify differences in interpretation. Examinations are to be repeated until agreement is reached among the examiners. This exercise should be conducted by a senior examiner. Part of the exercise may involve using extracted teeth; however, the training exercise must include examining live subjects.
Characteristics of Lesion ICDAS Code Active Lesion Surface of enamel is whitish/yellowish opaque with loss of luster; feels rough when the tip of the probe is moved gently across the surface. Lesion is in a plaque stagnation area, i.e.: pits and fissures, near the gingival and approximal surface below the contact point. Probably active. Cavity feels soft or leathery on gently probing the dentin. Cavity may be shiny and feels hard on gently probing the dentin. Inactive Lesion Surface of enamel is whitish, brown, or black. Enamel may be shiny and feels hard and smooth when the tip of the probe is moved gently across the surface. For smooth surfaces, the caries lesion is gingival margin.
1, 2, or 3
4 5 or 6
205
ISMAIL
ET AL.
3. Two days of reliability assessment, using live subjects presenting with carious lesions with severity ranging between 1 and 5 (ICDAS). At least 20 patients should be examined per examiner and the senior examiner. 4. A senior examiner is a dentist with experience in using the ICDAS, who has a high degree (Kappa = 0.75+) of intra-examiner reliability, and has been calibrated and is reliable with another experienced ICDAS examiner. In some studies, a senior examiner may work concurrently with the other examiners to reach a final decision. The term senior examiner is used to refer to the standard that will be used to compare with the findings of the examiners in a study. The report of a study should provide details on the calibration exercise and the senior examiners(s). Statistical Consideration for Analysis of the Reliability of ICDAS The conventional method of analysis of reliability data has been to present aggregate kappa coefficients for either each examiner or all examiners participating in a study. Kappa coefficients have the following advantages over simple percent agreements: 1) they account for agreement by chance alone for binary and nominal ratings [Maclure and Willett, 1987], and 2) there are standards for evaluating the strength of the agreement using this method. However, kappa analyses have disadvantages as well. Kappa is more a measure of exact agreement instead of being a measure of the degree of approximate agreement [Maclure and Willett, 1987]. A simple kappa coefficient does not distinguish between the different sources and magnitudes of disagreement. This measure of agreement tends to treat all the cases of disagreement alike, however large or small they might be [Maclure and Willett, 1987]. In other words, kappa does not consider the degree of disagreement between observers. Kappa may not be comparable across different studies, as the statistic is influenced by trait prevalence or distribution and disease categories [Spitznagel et al., 1985; Thompson et al., 1988a and 1988b; Feinstein et al., 1990]. The presence of bias between observers and the variations in the distribution of data across the categories may cause computational and interpretation problems in a kappa analysis [Byrt et al., 1993]. When continuous data are categorized to form ordinal categories, kappa becomes arbitrary and virtually meaningless [Maclure and Willet, 1987].
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Sometimes, the examiners may be consistent, but the kappa statistic may not display this agreement due to large number of categories, lack of marginal homogeneity, or marginal distribution of the data. In such cases, other flexible approaches, such as statistical modeling, may have to be used [Uebersax, 1987a and 1987b]. In order to account for the degree of disagreement between observers and also to distinguish the disagreements, weighted kappa may be used. This statistic incorporates the factor of agreement by chance alone and also has a feature of weighted proportional agreement. This is obviously an improved measure over the simple Cohens kappa, but the use of standard weights makes the new statistic of weighted kappa equivalent to intraclass correlation coefficient [Fleiss et al., 1973]. One important requirement for testing whether the kappa coefficients are statistically accurate is to test for marginal homogeneity of the distribution of codes for each examiner. Marginal homogeneity [Barlow, 1998; Bishop et al., 1975] means that the marginal frequencies or proportions of one or more categories are the same for both examiners. The Stuart-Maxwell (SM) statistic tests the homogeneity of marginal frequencies and is interpreted like a chi-squared test [Uebersax, 2005]. If the marginal distributions are not homogenous, then the kappa coefficients may not be accurate and may lead to erroneous conclusions. In such case, we recommend using other methods for analysis of reliability data. Log-linear modeling provides another approach for the analysis of examiners reliability [Uebersax, 1993; Kingman, 1986]. This approach is quite flexible in its assumptions of the distribution of the codes assigned by the examiners to tooth surfaces. Further, the general framework allows for simultaneous incorporation of multiple (more than 2) examiners, each rating an arbitrary number of categories. Hence, the symmetry of categories required for computing the kappa coefficients is not required for log linear models [Tanner, 1985]. The users of ICDAS should provide the following reliability statistics: 1. Kappa coefficients for comparisons between the senior examiner and each examiner separately. 2. Kappa coefficients for intra-examiner reliability for each examiner. 3. Rows X Columns table should be included for all comparisons. If possible, it is recommended that SM tests also be performed. However, computing the SM tests requires some advanced programming skills.
ISMAIL
ET AL.
References
Barlow W: modeling of categorical agreement. In: Armitage P, Colton T (eds): The Encyclopedia of Biostatistics (pp. 541-545). New York: Wiley, 1998 Bishop Y, Fienberg S, Holland P: Discrete multivariate analysis: theory and practice. Cambridge, Massachusetts: MIT Oress; 1975. Byrt T, Bishop J, Carlin JB: Bias, prevalence and kappa. J Clin Epidemiol 1993;423-429. Feinstein AR, Cicchetti DV: High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol 1990;43;543-549. Fleiss JL, Cohen, J: The equivalence of weighted kappa and the intraclass correlation coefficient as measures of reliability. Educ Psych Measurement 1973;33:613-619. Kingman A: A procedure for evaluating the reliability of a gingivitis index. J Clin Periodontol 1986;13:385391. Maclure M, Willett WC: Misinterpretation and misuse of the kappa statistic. Am J Epidemiol 1987;126:161169. Nyvad B, Machiulskiene V, Baelum V: Reliability of a new caries diagnostic system differentiating between active and inactive caries lesions. Caries Res 1999;33:252-260. Spitznagel EL, Helzer JE: A proposed solution to the base rate problem in the kappa statistic. Arch Gen Psychiat 1985; 42:725-728. Tanner MA, Young MA: modeling agreement among raters. J Am Stat Assoc 1985;80:175-180. Thompson WD, Walter SD: A reappraisal of the kappa coefficient. J Clinical Epidemiol 1988a;41:949-958. Thompson WD, Walter SD: Kappa and the concept of independent errors. J Clin Epidemiol 1988b;41:969-970. Uebersax JS: Measuring diagnostic reliability: Reply to Spitznagel and Helzer (letter). Arch Gen Psychiat 1987a;44:193-194. Uebersax, JS: Diversity of decision-making models and the measurement of interrater agreement. Psych Bulletin 1987b;101:140-146. Uebersax JS: Statistical modeling of expert ratings on medical treatment appropriateness. J Am Stat Assoc 1993;88:421-427. Uebersax JS: Statistical methods for rater agreement: The tetrachoric and polychoric correlation coefficients. http://ourworld.compuserve.com/homepages/jsuebersax/tetra.htm (accessed June 24, 2005).
208
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Appendix II Participants in the ICDAS Baltimore Workshop*

*(Upon the request of Federal Officials and 1 participant, their names were deleted)
Coordinators Amid Ismail Professor School of Dentistry, University of Michigan Ann Arbor, MI 48109-1078 USA Telephone: 001 734 647 9190 Fax: 001 734 936 1597 ismailai@umich.edu Gail Topping Director of Dental Caries Control Programme/ Honorary Consultant in Dental Public Health Dental Health Services Research Unit University of Dundee Dundee, Scotland Telephone: +44(0)1382 420050 (Secretary: Hazel Braid) Mobile phone: +44(0)7962 211219 Fax: +44(0)1382 420051 g.topping@chs.dundee.ac.uk Students Khalifa Sulaiman Al-Khalifa (University of Michigan) Adjunct Clinical Assistant Professor, Cariology, Restorative Sciences, and Endodontics 2361 Dent Ann Arbor, MI 48109-1078 USA Telephone: 001 734 647 4182 khalifaa@umich.edu Fang Gu (University of California) fanggu@ucla.edu
209
ISMAIL
ET AL.
Sonia Kumari Makhija (University of Alabama) Department of Diagnostic Sciences Post-doctoral Fellow and Clinical Instructor at the University of Alabama at Birmingham School of Dentistry Birmingham, AL USA drsmak2003@yahoo.com Stefania Martingnon (Universidad El Bosque, Colombia) PhD Candidate University of Copenhagen Copenhagen, Denmark smartignon@yahoo.com Participants Jim Bader Research Professor Dept. of Operative Dentistry & Senior Fellow Sheps Center for Health Services Research University of North Carolina 725 Airport Rd. Chapel Hill, NC 27514 USA Telephone: 001 919 966 5727 Fax: 001 919 966 3811 jim_bader@unc.edu David Banting School of Dentistry Faculty of Medicine & Dentistry University of Western Ontario London, ON N6A 5C1 Canada Telephone: 001 519 661 2111 x86130 Fax: 001 519 661 3875 dbanting@uwo.ca
210
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Richard Chesters Director, Oral Care Professional Relations Colgate-Palmolive Europe 13-15, Cours de Rive 1204 Geneva Switzerland Telephone: +41 22 722 0784 Fax: +41 22 722 0703 Mobile: +41 79 596 3956 Richard_Chesters@colpal.com Chris Deery Consultant in Paediatric Dentistry Edinburgh Dental Institute Lauriston Building 3 Lauriston Place Edinburgh EH3 9YW Scotland Telephone: 44 (0)131 536 4994 Fax: 44 (0)131 536 4908 Chris.deery@lpct.scot.nhs.uk Elbert de Josselin de Jong Chief Research Inspektor Research Systems Quellijnstraat 92 1072 XX Amsterdam The Netherlands Telephone: +31 20 676 4988 Fax: +31 20 679 3183 e.dejosselindejong@inspektor.nl Kenneth A. Eaton United Kingdom Telephone/Fax: ++ 44 1233 813585 K.Eaton@eastman.ucl.ac.uk or keaton@rcseng.ac.uk
211
ISMAIL
ET AL.
Hafsteinn Eggertsson Assistant Professor Indiana University School of Dentistry Oral Health Research Institute 415 Lansing Street Indianapolis, IN 46202-2876 USA Telephone: 001 317 278 3457 Fax: 001 317 274 5425 heggerts@iupui.edu Frederick Eichmiller ADAF Paffenbarger Research Center 100 Bureau Drive, MS 8546 Gaithersburg, MD 20899-8546 USA Telephone: 001 301 975 6813 Fax: 001 301 963 9143 Fred Eichmiller fred.eichmiller@nist.gov Kim Ekstrand Associate Professor of Cariology and Endodontics Department of Cariology and Endodontics, School of Dentistry Faculty of Health Sciences University of Copenhagen 20 Noerre All DK-2200 Copenhagen N Denmark Telephone: 45 35326813 Fax: 45 35326505 kim@odont.ku.dk Augusto R. Elias-Boneta University of Puerto Rico School of Dentistry PO Box 365067 San Juan, PR 00936-5067 Telephone: 001 787 765-3379 Fax: 001 787 763-4868 aelias@rcm.upr.edu
212
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Roger P. Ellwood Dental Health Unit Skelton House Lloyd St North Manchester M15 6SH England UK Telephone: 0161-232-4705 Fax: 0161-232-4700 roger.ellwood@manchester.ac.uk John D.B. Featherstone Professor and Chair Department of Preventive and Restorative Dental Sciences University of California San Francisco P.O. Box 0758 707 Parnassus Ave. San Francisco, CA 94143-0758 USA Telephone: 001 415 476 0456 Fax: 001 415 476 0858 jdbf@itsa.ucsf.edu Andrea G. Ferreira-Zandon Assistant Professor Oral Health Research Institute Indiana University School of Dentistry Department of Preventive and Community Dentistry 415 Lansing Street, Room 129 Indianapolis, IN 46202 USA Telephone: 001 317 274 3409 Fax: 001 317 274 5425 azandona@iupui.edu
213
ISMAIL
ET AL.
Rainer Haak University of Cologne Centre of Dental Medicine Dept. of Operative Dentistry and Periodontology Kerpener Str. 32 D-50931 Kln Germany Telephone: +49 (0)221 478-4124; -4710 (Secretary: Herbert Stecher) Fax: +49 (0)221 478 6405 Rainer.Haak@medizin.uni-koeln.de <http://www.medizin.uni-koeln.de/kliniken/zahn/erhalt/> Andrew Hall Senior Lecturer in Restorative Dentistry Glasgow University Dental School 378 Sauchiehall Street Glasgow G2 3JZ Scotland UK Telephone: 44 141 211 9778 Fax: 44 141 331 2798 a.hall@dental.gla.ac.uk Marie-Charlotte Huysmans Division of Conservative Dentistry Dept. of Dentistry and Dental Hygiene GUMC University of Groningen A. Deusinglaan 1 NL-9713 AV Groningen The Netherlands Telephone: 31 50 363 3203 Fax: 31 50 363 2696 m.c.d.n.j.m.huysmans@med.rug.nl Edwina Kidd (married name Littleton) Retired Professor, University of London Justine Kolker University of Iowa
214
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Jessica Y. Lee Assistant Professor Depts. of Pediatric Dentistry and Health Policy Analysis 228 Brauer Hall, CB #7450 University of North Carolina at Chapel Hill Chapel Hill, NC 27599-7450 USA Telephone: 001 919 966 2739 Fax: 001 919 966 7992 Jessica_Lee@dentistry.unc.edu Steven Levy Professor, College of Dentistry University of Iowa N330 DSB Iowa City, IA 52242 USA Telephone: 001 319 335 7185 Fax: 001 319 335 7187 steven-levy@uiowa.edu Chris Longbottom Senior Lecturer in Preventive & Childrens Dentistry Dundee Dental School Programme Methodologist Dental Health Services Research Unit University of Dundee 9th Floor University of Dundee Dental School Park Place Dundee, DD1 4HN UK Telephone: 44 (0)1382 425 759 Fax: 44 (0)1382 206 321 c.longbottom@dundee.ac.uk
215
ISMAIL
ET AL.
Daniel Meyer Associate Executive Director Director, Division of Science American Dental Association 211 E. Chicago Ave. Chicago, IL 60611-2678 USA Telephone: 001 312 440 2543 meyerd@ada.org Athena S. Papas Johansen Professor of Dental Research Tufts School of Dental Medicine 1 Kneeland St. Boston, MA 02111 USA Telephone: 001 617 636 3932 Fax: 001 617 636 4083 Athena.Papas@tufts.edu Deok-Young Park Associate Professor Department of Preventive and Public Health Dentistry College of Dentistry, Kangnung National University 123 Jibyeon-dong, Kangnung-shi Kangwon-do 210-702 South Korea Telephone: +82-640-3185 Cell Phone: +82-10-8988-7542 Fax: +82-640-3103 jguitar@kangnung.ac.kr Neil Pender Senior Lecturer/Consultant in Orthodontics Dept. of Clinical Sciences The University of Liverpool Liverpool, L69 3BX UK n.pender@liv.ac.uk
216
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Mathilde C. Peters Professor School of Dentistry, University of Michigan Ann Arbor, MI 48109-1078 USA Telephone: 001 734 763 3366 mcpete@umich.edu Klaus Pieper Professor Medizinisches Zentrum ZMK Georg-Voigt-Str. 3-5 35033 Marburg Germany Telephone: +49 6421 2863224 Fax: +49 6421 2866691 pieper@med.uni-marburg.de Nigel B. Pitts Professor Director, Dental Health Services Research Unit Honorary Consultant in Dental Public Health Dental Health Services Research Unit University of Dundee The Mackenzie Building Kirsty Semple Way, Ninewells Hospital Dundee, DD2 4BF Scotland Telephone: 44 (0)1382 420 058 Fax: 44 (0)1382 420 051 Iain A. Pretty Research Fellow Dental Health Unit University of Manchester Unit 3A, Skelton House Manchester Science Park Manchester M15 6SH UK Telephone: 0161 226 1211 Fax: 0161 226 1244 iainbds1@tiscali.co.uk
ISMAIL
ET AL.
Elmar Reich Rolf-Keller-Platz 1 88400 Biberach Germany ereich@t-online.de (EReich@t-online.de) David Ricketts Senior Lecturer/Hon. Consultant in Restorative Dentistry Dundee Dental School Park Place Dundee DD1 4HR Scotland UK Telephone: 01382 660111, Ext. 35820 Fax: 01382 635984 d.n.j.ricketts@dundee.ac.uk Robert H. Selwitz (ICDAS Committee member: former official representative of NIDCR) Formerly: Chief, Population Research and Health Promotion Branch Director, Residency Program in Dental Public Health Division of Clinical Research and Health Promotion Natcher Building, Room 4As-37J 45 Center Drive MSC 6401 Bethesda, MD 20892-6401 USA Telephone: 001 301 594 3977 Fax: 001 301 480 8322 robert.selwitz@nih.gov Xie-Qi Shi (Birgit Angmar-Mnsson) Department of Cariology and Endodontology Institute of Odontology Karolinska Institutet Box 4064 SE 141 04 Huddinge Sweden Telephone: + 46 8 524 88184 Fax: +46 8 711 8343 xie.qi.shi@ofa.ki.se
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Dan Shugars Professor, School of Dentistry University of North Carolina, CB# 7450 Chapel Hill, NC 27599-7450 USA Telephone: 001 919 966 1214 dan_shugars@dentistry.unc.edu Richard J. Simonsen Associate Dean and Professor of Restorative Dentistry Arizona School of Dentistry & Oral Health 5850 East Still Circle Mesa, AZ 85206 USA Telephone: 001 480 219 6082 Cell Phone: 001 480 203 9195 Fax: 001 480 219 6180 rsimonsen@atsu.edu Woosung Sohn Assistant Professor School of Dentistry, University of Michigan Ann Arbor, MI 48109-1078 USA Telephone: 001 734 615 6622 Fax: 001 734 936 1597 woosung@umich.edu George W. Taylor Associate Professor University of Michigan School of Dentistry Dept. of Cariology, Restorative Sciences & Endo 1011 N. University Ann Arbor, MI 48109 USA Telephone: 001 734 764 1737 Fax: 001 734 936 1597 gwt@umich.edu
219
ISMAIL
ET AL.
Marisol Tellez Former Research Fellow University of Michigan School of Dentistry Dept. of Cariology, Restorative Sciences & Endo 1011 N. University Ann Arbor, MI 48109 USA Telephone: 001 734 615 7186 Fax: 001 734 936 1597 mtellez@umich.edu Van P. Thompson Biomaterials & Biomimetics NYU College of Dentistry 345 E. 24th St., 804S New York, NY 10010 USA Telephone: 001 212 998 9638 Fax: 001 212 995 4244 van.thompson@nyu.edu Norman Tinanoff Department of Health Promotion and Policy University of Maryland Dental School 666 W. Baltimore St. Baltimore, MD 21201 USA Telephone: 001 410 706 7970 NTinanoff@dental.umaryland.edu Monique H. van der Veen Senior Science Officer Inspektor Research Systems Quellijnstraat 92 1072 XX Amsterdam The Netherlands Telephone: +31 20 676 4988 Fax: +31 20 679 3183 m.vd.veen@inspektor.nl
220
RATIONALE
AND
EVIDENCE
FOR THE
ICDAS
Helen Whelton Director, Oral Health Services Research Centre Senior Lecturer in Dental Public Health and Preventive Dentistry Department of Oral Health and Development University Dental School and Hospital Wilton Cork Ireland Telephone: +353 21 4901212 Fax: +353 21 4545391 H.Whelton@ucc.ie Domenick T. Zero Associate Dean for Research Professor and Chair Department of Preventive and Community Dentistry Director, Oral Health Research Institute Indiana University School of Dentistry Oral Health Research Institute 415 Lansing Street Indianapolis, IN 46202-2876 USA Telephone: 001 317 274 8822 Fax: 001 317 274 5425 Mobile: 001 317 402 4607 dzero@iupui.edu
221
222
Quantitative Light Induced Fluorescence in Caries Clinical Trials

Susan M. Highama, I.A. Prettyb, and P.W. Smitha
Cariology and Toothwear Research Group, School of Dental Studies, The University of Liverpool, Liverpool, UK Dental Health Unit, The School of Dentistry, The University of Manchester, Manchester, UK
Introduction For centuries, the pain and morbidity associated with tooth decay have focused considerable interest in the aetiology, diagnosis, and management of dental caries [Miller, 1890; Marcus et al., 1996]. As our understanding of caries has evolved, the thrust of caries management has been directed towards the prevention of early enamel caries. Our present day understanding of dental caries has been achieved by a variety of methods, which have used in vitro, intra-oral, in situ, and animal models. The level of information provided by these different methods is variable. Criticism has been directed at in vitro and animal models as being far removed from the processes leading to caries in humans. Although in situ or intra-oral model systems have addressed much of this criticism, they also have disadvantages. The main drawbacks are that they attempt to model the carious process and that quantification of de- and re-mineralisation requires the use of destructive techniques that lead to loss of the sample under study, thus making longitudinal monitoring of mineral changes impossible. Undoubtedly, given unlimited time and funds, most researchers would wish to conduct well designed clinical trials to test the efficacy of caries preventive and management strategies. Although randomized, controlled clinical
HIGHAM
ET AL.
trials are held as the counsel of perfection, these are expensive, require a large number of subjects, raise significant ethical and logistical problems, and often are lengthy. They may not be feasible for practical reasons and are difficult to justify [Higham et al., 2005]. Chesters et al. [2004] reported that it was becoming commercially prohibitive to conduct new caries clinical trials. Clearly, it would be advantageous for all concerned if more discriminatory, quicker, and more cost effective trials could be conducted using fewer subjects, which provided a high quality alternative to traditional, lengthy, caries trials. Quantitative light induced fluorescence (QLF) is one technique that may offer an appropriate and viable means to reduce the perceived burden of the full clinical trial. QLF Background QLF is founded on the principle that caries brings about a change in enamel autofluorescence [Benedict, 1929]. The reduction in fluorescence, which is observed when enamel demineralises, was later described by Bjelkhagen et al. [1982]. It has been proposed that this loss of fluorescence results from increased porosity seen in carious enamel compared with sound enamel. The uptake of water in pores reduces the refractive index of enamel; this, in turn, increases the scattering of light, which ultimately produces a decrease in auto-fluorescence [Angmar-Mnsson and ten Bosch, 1987]. Techniques using changes in auto-fluorescence were developed in subsequent years and, in 1989, Sundstrm and coworkers reported that early caries lesions had been detected in vitro. The development of software that is able to quantify changes in enamel fluorescence into mineral content [de Josselin de Jong et al., 1995] was a major advance that has led to the further development of QLF for use in clinical trials. Further developments in hardware technology led to the use of a blue visible light source and filters that allow for the production of more compact systems capable of being used in the clinical setting [al-Khateeb et al., 1977a]. Is QLF Valid? With any new technology, it is important to evaluate whether it is able to reliably measure what it purports to measure. A number of validation studies have been performed relating to QLF, which have been reviewed by ten Bosch
QUANTITATIVE LIGHT INDUCED FLUORESCENCE
IN
CARIES CLINICAL TRIALS
[1999] and, more recently, by Pretty, Edgar and Higham [2003a]. Validation studies have compared the quantification of mineral changes measured by QLF with the current gold standard measure of mineral content, transverse microradiography (TMR). By necessity, these studies have been conducted in carefully controlled in vitro experiments, which have demonstrated that QLF is a valid, reliable technique for measuring changes in enamel mineralisation. Several studies by al-Khateeb and co-workers [1998; 1997a; 1997b] have shown a good correlation between QLF and TMR (r=0.84), confirming that QLF is a valid means for evaluating mineral change in enamel caries. Is QLF Repeatable and Reliable? As already outlined above, QLF relies on several steps when used in vivo, relating to image capture and analysis. Tranus et al. [2001] demonstrated in an in vivo clinical trial that for image capture stage of QLF, the intra class correlation coefficient (r) was between 0.95 and 0.98. For the analytical stage, the same authors showed a value for r between 0.93 and 0.99, and they concluded that QLF had excellent reliability and repeatability. In a separate study, Pretty et al. [2002] showed that QLF analysis of artificially produced caries lesions was also highly repeatable and reliable, both between examiners and for multiple attempts by the same examiner. It may be concluded that for studies carried out to date, QLF is a potentially valid, reliable, and repeatable technique. It is also helpful to evaluate the ability of QLF to detect carious lesions that affect different tooth surfaces and sites, as this may be important in the assessment of the effectiveness of caries prevention and management. Smooth Surface Enamel Caries QLF has been shown to be a promising and useful device in the detection of early demineralisation during the development of smooth surface caries (particularly during orthodontic treatment), root caries, secondary caries, and occlusal caries. Most of the studies monitoring smooth surface de- and re-mineralisation have involved patients undergoing treatment with fixed orthodontic appliances; al- Khateeb [1998] concluded that QLF was useful for monitoring the effectiveness of preventive regimes in caries susceptible individuals. In another study involving orthodontic patients, Boersma [2005] showed that QLF was useful in detecting carious lesions at an earlier stage. The usefulness of
HIGHAM
ET AL.
QLF in detecting and monitoring early enamel caries around orthodontic brackets has also been confirmed by in vitro studies reported by Benson et al. [2003a & 2003b] and Pretty et al. [2003b]. Smooth surface caries in non-orthodontic studies has also been subjected to QLF analysis [Tranus et al., 2001; Kambara, 2003], and these studies demonstrated that QLF was a sensitive method for measuring changes in incipient enamel caries, in addition to the re-mineralising effects of fluoride being quantified using a relatively small panel of subjects in a shorter time than conventional clinical trials. Occlusal Caries There has been relatively little reported on the use of QLF for the detection and quantification of caries affecting occlusal surfaces. Studies in Indiana [Ferreira-Zandon et al., 1998; 1999] have reported that it is possible to use QLF technology for occlusal caries and for the detection of incremental caries on occlusal surfaces. This is an important finding, especially bearing in mind the significant number of occlusal surfaces observed to be carious in the 8-15 year age group [Ripa et al., 1988]. Improvements in oral health, especially the widespread use of fluoride containing dentifrices, have led to a reduction in frank occlusal cavities. This, in turn, has increased the complexity of diagnosis of occlusal caries [Kidd et al., 1993]. Fissures in the occlusal surface are prone to staining, which may hinder accurate diagnosis. In addition, the poor performance of radiographs in the detection of early occlusal caries further confounds diagnosis [Thomas et al., 2001]. Occlusal caries generally becomes more readily visualised in radiographs when it extends into dentine, and the extent of demineralisation is often more extensive than indicated by radiographs [Ricketts et al., 1995]. Given that the diagnosis of occlusal caries by traditional means is problematic, Higham et al. [2003] have reported the development of an occlusal caries index using QLF, the intention being that, with further development, this may be useful in guiding clinical decisions regarding caries management at an individual patient level. This study involved the in vitro evaluation of 75 extracted human premolars and molars. The teeth were examined visually, radiographically, histologically, and were scored according to the criteria proposed by Ekstrand et al. (1997); the same teeth were then examined using QLF. Good agreement was found between the QLF parameter F at the 15% threshold level with the histological gold standard. Further development of the QLF technique
IN
for use in this application in vivo would be a major advance, but from these early findings, it is apparent that QLF holds promise in the detection and longitudinal monitoring of occlusal caries. Interproximal Caries The challenge in detecting interproximal caries results from the inaccessibility of these surfaces to direct visualisation. In an attempt to utilise QLF for these surfaces, Buchalla and co-workers [2002] have attempted to determine the optimal illumination and camera angulations for the use of QLF to evaluate interproximal caries. It is unlikely that QLF would be able to detect interproximal sites at as early a stage as is possible with other tooth sites. This is an area that requires further development and study in order to reap the potentially significant clinical benefits. Deciduous Enamel It is also important that QLF can measure changes in the enamel of deciduous teeth. This has been evaluated in vitro and also in in vivo pilot studies [Ando et al., 2001; Pretty et al., 2002], and the findings are that it is a method that is able to determine mineral changes as readily and accurately in deciduous enamel as in permanent teeth. Root Caries Although initially QLF analysis has focussed on enamel caries, the growing epidemiological problem of root caries [Galen & Lynch, 1993] has caused attention to be directed toward the evaluation of root caries [Smith et al., 2005a; 2005b; Katz et al., 1995]. The differing optical properties of dentine compared with enamel make the evaluation of root caries more problematical [Gonzlez-Cabezas et al., 2001; Amaechi and Higham, 2002]. However, certain dyes can be used to stain carious dentine [Wilkinson et al., 1997] and provide the possibility of evaluating root caries in a quantifiable and archivable way using QLF. The use of fuorescein dye in conjunction with QLF for evaluating root caries in vitro has been shown to be effective in monitoring de- and remineralisation [Pretty et al., 2003c]. These authors have suggested that it might be possible to develop QLF as a method for longitudinal and objective monitoring of root caries in vivo, thereby reducing the difficulties traditionally encountered with the study of root caries.
227
HIGHAM
ET AL.
Secondary Caries So far, our attention has been concentrated on the need to be able to diagnose primary caries and evaluate established and novel preventive caries strategies. Clearly, the ability to achieve the same levels of detection and evaluation for caries occurring around dental restorations is equally important. It has been suggested that traditional diagnostic techniques are more useful in the diagnosis of secondary caries when significant amounts of tissue loss have already occurred [Gonzlez-Cabezas et al., 2003]. A number of studies have evaluated QLF in relation to the detection and monitoring of secondary caries. Pretty et al. [2003d] demonstrated in an in vitro study that QLF was able to longitudinally monitor demineralisation of enamel adjacent to a number of provisional and definitive restorative materials. Ando and co workers [2004] were able to demonstrate that QLF was able to detect naturally occurring secondary caries adjacent to amalgam restorations in extracted teeth. QLF also showed the greatest sensitivity among the other techniques studied for the detection of early secondary caries. This advantage falls for more advanced carious lesions (deeper than 400m), underlining the importance of QLF as a technique for detecting early secondary caries. Is There a Role for QLF in Clinical Trials? Historically, caries diagnosis in clinical trials has been predominantly based on the interpretation of information provided by visual inspection, bitewing radiographs, and fibre optic transillumination (FOTI). Often, these methods use different examiners who, despite training and calibration, may lack objectivity and reliability and, unless the trials are carefully designed and controlled, may be subject to bias. In addition, changes in carious lesions are difficult to quantify and do not allow for longitudinal monitoring of mineral changes in lesions [Angmar-Mnsson and ten Bosch, 2001]. In addition, most indices used for recording caries lesions do so only at the cavitation stage, which may give a relatively crude measure of caries increment. It would be advantageous if there were more quantitative techniques that could be used to detect and monitor early pre-cavitation mineral changes in enamel lesions [Ferreira Zandor et al., 2003], as this is thought to be where preventive strategies may be more effective. Supporting evidence for the use of QLF in clinical caries trials comes from several sources:
228
IN
Published clinical trials Unpublished clinical studies Ongoing clinical studies Other sources, e.g., observational, in vitro Published Clinical Studies Using QLF In response to the emerging need for sensitive clinical methods to detect and quantify mineral changes in caries lesions, several studies have reported the use of QLF as a technique for monitoring the effectiveness of caries preventive strategies in clinical trials. Tranus et al. [2001] compared the use of fluoride varnish and professional tooth cleaning as a means of white spot remineralisation, and they were able to conclude that QLF was a sensitive method for longitudinal quantification of incipient caries lesions on the smooth surfaces of childrens teeth. Boersma et al. [2005] reported on the use of QLF compared with visual inspection and photographic techniques as a means of monitoring incipient caries lesions in children undergoing treatment with fixed orthodontic appliances. It is well known that such caries lesions may develop with alarming speed, sometimes becoming clinically apparent after only 1 month [Ogaard et al., 1988]. Boersmas study concluded that QLF was more sensitive than clinical visual inspection, allowing for the earlier detection of caries. Ferreira-Zandon et al. [2003] conducted a clinical validation study that aimed to investigate the abilities of newer, emerging technologies, including QLF, to detect primary caries. This longitudinal study was conducted in Connersville (Indiana, USA) and involved 152 children aged 8-11 years. The data reported indicates that QLF has high examiner repeatability and is also able to detect enamel lesions at an earlier stage and in greater numbers when compared with visual examination. Kambara [2003] reported a study involving 132 subjects aged from 9-48 years (mean 20+/-8,4 years). This study aimed to assess changes to white spot caries lesions using QLF and, additionally, to determine the effects of a fluoride dentifrice in a clinical trial structure using QLF. This study clearly demonstrated the ability of QLF to monitor and quantify the changes to white spot lesions. Since the impact of fluoride dentifrice has been demonstrated clinically on numerous occasions using more conventional clinical trial designs, QLF was
229
HIGHAM
ET AL.
shown to quantify the remineralising effects using a relatively small number of subjects in a shorter time. In a clinical study involving 107 subjects, QLF and ICDAS measures were used to evaluate the effects of chlorhexidine gel and fluoride varnish on caries [Eckert et al., 2005]. One important objective was to evaluate the statistical power of the 2 caries assessment measures. The data demonstrate that QLF is able to detect a 20% change from baseline measures, with sample sizes of less than 40 subjects per group. It can be concluded that the evidence from published clinical studies shows that QLF is a valid, reliable method for evaluating enamel caries. It seems to be able to detect lesions at an earlier stage than currently used conventional methods, such as clinical visual inspection. The QLF method is also sufficiently sensitive, reliable, and repeatable to allow the use of fewer subjects in the evaluation of some currently used caries preventive and management strategies. It is important to note, however, that, in most diagnostic systems, as sensitivity increases, so does the signal/noise ratio. One solution to this problem, using the QLF technique, is to set the fluorescence threshold to higher values (e.g., from 5% to 15%), but, as this decreases noise, the tradeoff is that the smaller, earlier lesions are no longer detected. Unpublished Clinical Studies Currently, QLF has been used in various clinical situations where the evaluation of changes in enamel mineralisation status would be advantageous. Several research groups are active in these areas and, as yet, some of the work is unpublished. Van der Veen et al. [2005 in press] studied naturally occurring white spot lesions that occurred around fixed orthodontic brackets. QLF was useful in monitoring the lesions longitudinally; over time, the caries lesions regressed when preventive measures were introduced, but without these, the chance of natural regression was small. Higham and co-workers [unpublished data] have also longitudinally monitored the remineralisation of white spot lesions in orthodontic patients. Pretty and Elwood [unpublished data] have demonstrated the ability of QLF to monitor mineral changes in naturally occurring caries lesions on free smooth surfaces, buccal pits, and occlusal fissures. They have proposed that studies which are able to assess the construct validity of QLF, i.e., by assessing a known change in a population group, are important in this field of study. An
IN
example of this would be the discrimination of a fluoride dose response in dentifrice products or the separation of caries levels in populations both with and without water fluoridation.. Ongoing Clinical Studies and Observations To date, QLF has been shown to have potential for use in paediatric dental patients, patients suffering from xerostomia [Pretty et al., unpublished data]; the extent to which the QLF technique may be used to evaluate preventive management strategies is promising but has yet to be confirmed. There is currently a move towards the development of clinical care pathways following a definitive diagnosis by a clinician. Diagnoses that are assisted by QLF may offer the possibility of true preventive practice, as early, remineralisable lesions are detected and quantified. It is important to encourage dental practitioners to be involved in research that aims to evaluate the clinical use of QLF in the dental practice setting. Emerging Issues with the Use of QLF in Caries Evaluation This review has sought to demonstrate the evidence that currently exists for the use of QLF as a method of detecting, quantifying, and longitudinally monitoring mineral changes in enamel caries and whether this modality can be applied to clinical trials involving caries preventive products and strategies. There is clear evidence that QLF is a valid and reliable means of quantifying mineral loss in early enamel caries. It is able to do so before more conventional visual inspection, more reliably, and with fewer implications for health effects than some other methods, such as radiography. There are unresolved issues relating to QLF analysis, including which parameters are most appropriate to use when evaluating enamel caries, for example, change in fluorescence versus change in lesion area, this is relevant to the debate surrounding the presence of staining, which is a confounding factor in QLF analysis. Similarly, the selection of a fluorescence threshold level to optimise sensitivity but to minimise the signal/noise ratio requires further investigation. QLF has demonstrated a clear ability to measure mineral changes in relation to the use of fluoride dentifrices, although, to date, no in vivo studies have been published that show dose responses. Such studies are required if the validity of QLF is to gain universal acceptance.
HIGHAM
ET AL.
There is also some debate over the issue of whether the buccal caries model (natural or orthodontic) is appropriate for use in caries studies, and also whether further development is possible to allow the use of QLF in the evaluation of occlusal and approximal caries. Conclusions QLF has the potential to provide a viable and appropriate abbreviated clinical trial, replacing or augmenting the more traditional visual clinical inspection. QLF is a valid method for detecting and quantifying enamel caries and may offer the possibility of effective determination of lesion change over time. The methods sensitivity allows for the use of fewer subjects than has traditionally been the case, and evaluation can be conducted over shorter periods of time, producing quantifiable data that is archived by dedicated software.
References
al-Khateeb S, Forsberg C-M, de Josselin de Jong E, Angmar-Mnsson B, Sundstrm G, Exterkate RAM: A longitudinal laser fluorescence study of white spot lesions in orthodontic patients. Am J Orthod Dentofacial Orthop 1998;113:595-602. al-Khateeb S, Oliveby A, de Josselin de Jong E, Angmar-Mnsson B: Laser fluorescence quantification of remineralisation in situ of incipient enamel lesions: influence of fluoride supplement. Caries Res 1997b;31:132-140. al-Khateeb S, ten Cate JM, Angmar-Mnsson B, de Josselin de Jong E, Exterkate RAM, Sundstrm G, Oliveby A: Quantification of lesion formation and remineralisation with a new portable fluorescence device. Adv Dent Res 1997a;11:502-506. Amaechi BT, Higham SM: Detection and quantification of root caries using quantitative light-induced fluorescence. J Dent Res 2002;81:A-99. Ando M, Gonzlez-Cabezas C, Isaacs RL, Eckert GJ, Stookey GK: Evaluation of several techniques for the detection of secondary caries adjacent to amalgam restorations. Caries Res 2004;38:350-356. Ando M, van der Veen MH, Schemehorn BR, Stookey GK: Comparative study to quantify demineralized enamel in deciduous and permanent teeth using laser- and light-induced fluorescence techniques. Caries Res 2001;35;464-470. Angmar-Mnsson B, ten Bosch JJ: Optical methods for the detection and quantification of caries. Adv Dent Res 1987;1:14-20. Angmar-Mnsson B, ten Bosch JJ: Quantitative light-induced fluorescence (QLF): A method for assessment of incipient caries lesions. Dentomaxillofacial Radiology 2001:30,298-307.
232
IN
Benedict HC: The fluorescence of teeth as another method of attack on the problem of dental caries. J Dent Res 1929;9:274-275. Benson PE, Pender N, Higham SM: Quantifying enamel demineralization from teeth into orthodontic brackets a comparison of two methods. Part 2: Validity. Eur J Orthod 2003b;25:159-165. Benson PE, Pender N, Higham SM: Quantifying enamel demineralization from teeth with orthodontic bracketsa comparison of two methods Part I: Repeatability and agreement. Eur J Orthod 2003a;25:149-158. Bjelkhagan H, Sundstrm F, Angmar-Mnsson B, Ryder H: Early detection of enamel caries by the luminescence excited by visible laser light. Swed Dent J 1982;6:1-7. Boersma JG, van der Veen MH, Lagerweij MD, Bokhout B, Prahl-Andersen: Caries prevalence measured with QLF after treatment with fixed orthodontic appliances: Influencing factors. Caries Res 2005;39:4147. Buchalla W, Lennon AM, van der Veen MH, Stookey GK: Optimal camera and illumination angulations for detection of interproximal caries using quantitative light-induced fluorescence. Caries Res 2002;36:320-326. Chesters RK, Ellwood RP, Biesbrock AR, Smith SR: Potential modern alternative designs for caries clinical trials (CCTs) and how these can be validated against the conventional model. J Dent Res 2004;83(Spec. Issue C):C 122-C124. De Josselin de Jong E, Sundstrm F, Westerling H, Tranus S, ten Bosch JJ, Angmar-Mnsson B: A new method for in vivo quantification of changes in initial enamel caries with laser fluorescence. Caries Res 1995;29:2-7. Eckert GJ, Katz BP, Ofner S, Ferreira-Zandon AG, Eggertson H, Doi T, Stookey GK, Wefel JS: Analysis of QLF and ICDAS measurements in a clinical trial. Caries Res 2005;294-295 (Abstract). Ekstrand KR, Ricketts DN, Kidd EAM: Responsibility and accuracy of three methods for assessment of demineralization depth of the occlusal surface: an in vitro examination. Caries Res 1997;31:224-231. Ferreira-Zandon AG, Analoui M, Schemehorn BR, Eckert GJ, Stookey GK: Laser fluorescence detection of demineralization in artificial occlusal fissures. Caries Res 1998;32:31-40. Ferreira-Zandon AG, Isaacs RL, van der Veen M, Stookey GK: Indiana pilot clinical study of quantitative light fluorescence In: Early Detection of Dental Caries: Proceedings of the 2nd Annual Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 1999: pp 219-230. Ferreria-Zandon AG, Stookey GK, Eggertsson H, Wefel J, Katz B, Ofner S, Eckert G: Clinical validation study of QLF at Indiana. In: Early Detection of Dental Caries: Proceedings of the 3rd Annual Indiana Conference, GK Stookey, editor. Indianapolis, Indiana School of Dentistry, 2003: pp 237-252. Galan D, Lynch E: Epidemiology of root caries. Geriodontology 1993;10:59-71. Gonzlez-Cabezas C, Dunne E, Ando M, Eggertsson H, Eckert GJ, Fontana M, Stookey GK: Detection of small caries lesions on root surfaces of extracted teeth. Caries Res 2001;35:282. Gonzlez-Cabezas C, Fontana M, Gomes-Moosbauer D, Stookey GK: Early detection of secondary caries using quantitative light induced fluorescence. Op Dent 2003;28-4:415-422. Higham SM, Pretty IA, Edgar WM, Smith PW: The use of in situ models and QLF for the study of coronal caries. J Dent 2005;33:235-241.
233
HIGHAM
ET AL.
Higham SM, Smith PW, Pretty IA: Development of an occlusal caries index for quantitative light-induced fluorescence (QLF) In: Early Detection of Dental Caries: Proceedings of the 3rd Annual Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 2003:pp 195-212. Kambara M, Vemura M, Miyake T, Doi T, Nakashima S, Eckert GJ, Stookey GK: Results of clinical trail of fluoride dentifrice using QLF. In: Early Detection of Dental Caries: Proceedings of the 3rd Annual Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 2003:pp. 229-235. Katz, RV: The clinical diagnosis of root caries: issues for the clinician and the researcher. Amer Jour Dent 1995;8:335-341. Kidd EAM, Ricketts DN, Pitts NB: Occlusal caries diagnosis, a changing challenge for clinicians and epidemiologists. J Dent 1993;21:323-331. Marcus SE, Dury TF, Brown LJ, Zion GR: Tooth retention and tooth loss in the permanent dentition of adultsUnited States 1988-1991. J Dent Res 1996;75:684-695. Miller WD: The microorganisms of the human mouth. Philadelphia, PA/Basel: The SS White Dental Mfg Co./Karger;1890. Ogaard B, Rolla G, Arends J: Orthodontic appliance and enamel demineralisation. Part I: Lesion development. Am J Orthod Dentofacial Orthop 1988;94:68-73. Pretty IA, Edgar WM, Higham SM: A review of the effectiveness of Quantitative Light-induced Fluorescence (QLF) to detect early caries. In: Early Detection of Dental Caries: Proceedings of the 3rd Annual Indiana Conference, Stookey, GK, editor, Indianapolis, Indiana University School of Dentistry, 2003a:pp 253-289. Pretty IA, Edgar WM, Higham SM: Detection of in vitro demineralization of primary teeth using quantitative light-induced fluorescence (QLF). Int J Paed Dent 2002b;12:158-167. Pretty IA, Hall AF, Smith PW, Edgar WM, Higham SM: The intra and inter-examiner reliability of QLF analyses. Br Dent J 2002a;193:105-109. Pretty IA, Ingram GS, Agalamanyi EA, Edgar WM, Higham SM: The use of fluorescein-enhanced quantitative light-induced fluorescence to monitor de- and remineralisation of in vitro root caries. J Oral Rehab 2003c;30:1151 1156. Pretty IA, Pender N, Edgar WM, Higham SM: The in vitro detection of early enamel de- and remineralization adjacent to bonded orthodontic cleats using quantitative light-induced fluorescence. European J Ortho 2003b,25:217-233. Pretty IA, Smith PW, Edgar WM, Higham SM: Detection of an in vitro dematerialization adjacent to restorations using quantitative light induced fluorescence (QLF). Dent Mat 2003d;19:368374. Ricketts DN, Kidd EAM, Smith BG, Wilson RF: Clinical and radiographic diagnosis of occlusal caries a study in vitro. J Oral Rehab 1995;22:15-20. Ripa LW, Leske GS, Varma AO: Longitudinal study of the caries susceptibility of occlusal and proximal surfaces of first permanent molars. J Public Health Dent 1988;48:8-13. ten Bosch JJ: Summary of research of quantitative light-induced fluorescence. In: Early Detection of Dental Caries: Proceedings of the 2nd Annual Indiana Conference, Stookey, GK, editor, Indianapolis, Indiana University School of Dentistry 1999;pp261-277.
234
IN
Smith PW, Preston KP, Higham SM: Development of an in situ root caries model. A. In vitro investigation J Dent 2005a;33;253-267. Smith PW, Preston KP, Higham SM: Development of an in situ root caries model. B. In situ investigation J Dent 2005b;33:269-273. Sundstrm F, Hafsstrom-Bjrkmann U, Strom J, Angmar-Mnsson B, Frostell G, Takazoe I: Evaluation of a model for short term clinical testing of cariogenicity. J Biol Bucccale 1989;17:115-120. Thomas MF, Ricketts DW, Wilson RF: Occlusal caries diagnosis in molar teeth from bitewing and panoramic radiographs. Prim Dent Care 2001;8:63-69. Tranus S, al-Khateeb S, Bjrkman S, Twetman S, Angmar-Mnsson B: Application of quantitative lightinduced fluorescence to monitor incipient lesions in caries-active children. A comparative study of remineralisation by fluoride varnish and professional cleaning. Eur J Oral Sci 2001:109:71-75. van der Veen MH, ten Bosch JJ: An in vitro evaluation of fluorescein penetration into natural root surface carious lesions. Caries Res 1993;27:258261. Wilkinson SC, Higham SM, Ingram GS, Edgar WM: Visualization of root caries lesions by means of a diazorium dye. Advances in Dental Research 1997,11:515-522.
235
236
Other Instrumental Detection Methods for Caries Clinical Trials

James S. Wefel1, Masatoshi Ando2, and Jeffrey D. Harless1
1
Dows Institute for Dental Research, College of Dentistry, University of Iowa, Iowa City, IA, USA Department of Preventive and Community Dentistry, Oral Health Research Institute, Indiana University School of Dentistry, Indianapolis, IN, USA
Introduction I would first like to congratulate the organizers for once again putting together an International cast of presenters and workshop discussants for this unique conference. Although agreement/consensus can often be difficult to obtain, the workshop venue offered here has one of the best lineups of experts and interested scientists to discuss acceptance of new, more efficient models for caries clinical trials. The International Consensus Workshop on Caries Clinical Trials (ICW-CCT) held in Scotland in 2002 and published in JDR 83, Special Issue C, 2004 emphasized that more efficient models for caries clinical trials needed to be developed. Similarly, the Indiana Conference held in 2003 provided a forum for presentations on the development of state of the art detection methods. Certainly, detection of the caries process at an early point in time would allow for the opportunity to also intervene at an earlier point. The detection of caries at the pre-cavitation phase will involve the measurement of demineralization and remineralization. The ability to monitor progression, arrestment, or reversals needs to be part of the new methodologies. As such, the NIHNIDCR sponsored a conference dealing with the use of surrogate markers and other strategies to improve the efficiency of clinical trials. In spite of the obvious interest and conference/workshops held in the past several years, no clear consensus exists on new methodologies to replace the traditional caries clinical trials. This conference may be able to take advantage of past findings, combine them with current data from the new methods examined, and possibly redefine the detection and measurement of dental caries in the future.
WEFEL
ET AL.
This paper presents results on other instrumental detection techniques that have been validated in the clinical setting and the laboratory. Their use and acceptance in caries clinical trials will be noted. I must bring attention to the recent Symposium at IADR on Early Detection of Dental Decay. This symposium summarized many of the new and established detection techniques and will be published in the near future. The current paper will provide a quick update of several of these methods, then describe our program project grant on Early Caries Detection and Intervention. The use of Laser Fluorescence DIAGNOdent (LF), Digital Imaging Fiber Optic Transillumination (DIFOTI), and Fiber Optic Transillumination (FOTI) in a trial on primary caries development are compared to Polarized Light Microscopy on exfoliated teeth as the gold standard. The final validation studies will be presented at the ORCA, 2005 meeting [Ferreira-Zandon et al., 2005]. Other Instrumental Detection Methods It is generally accepted that caries prevalence has dramatically declined in the past several decades, with smooth surface caries being most affected. Acceptance of the medical and/or preservation model(s) of caries recognizes that the pattern of disease has changed and that we need to have more refined measures of decay other than surface loss or holes in teeth. This would require an earlier detection of decay, such as reductions in mineral content within the tissue. The loss of mineral may result in electrical conductivity changes, changes in x-ray absorption, scattering of light (fluorescence) throughout the tissue, and other changes in properties that are currently being evaluated. It is these changes in the tooths properties that may prove valuable in early caries detection. Electronic Conductivity Measurements The use of electrical measurements for early caries detection was summarized at the 2003 Indiana Conference by Huysmans and was most recently reviewed by Longbottom and Huysmans [2004] at the workshop in Scotland (ICW-CCT). It stated that only limited data are available, and the current instrument being used is the Electronic Caries Monitor (ECM), which is presently limited in scope to occlusal surfaces. The authors pointed out a number of factors that effect the measurements, e.g., porosity, surface area, thickness of the tissue, hydration, and temperature. Results have shown the
EARLY DETECTION METHODS
ability to distinguish between different treatment groups [Ashley et al., 1997; 2001] in one study, but the inability to show product differences with ECM in a second study conducted in Vilnius, Lithuania [Bedinskaite et al., 2001]. The Ashely study was also said to show the inherent scientific invalidity of using an insensitive technique (Clinical Visual) as the gold standard for demonstrating product effects when these are compared with new detection systems. The only study on root surface caries [Baysan et al., 2001] tried to distinguish product effects between 5000 ppm and 1100 ppm F dentifrice. Fluoride was effective in hardening many of the root surfaces, and ECM readings between the 2 groups were significantly different. This may help in identifying the activity of root surface caries as much as detection. A confounding variable was the amount of plaque found in the two groups and its possible effect on hardening. In summary, Longbottom and Huysmans stated that the limited data available is both mixed and tantalizing and, while ECM measurements have higher sensitivity but lower specificity than Clinical Visual methods, the data is limited to occlusal surfaces. This could pose a problem in relation to interpreting and comparing caries clinical trial (CCT) results in which subjects have differing relative occlusal/approximal caries prevalence and/or incidence ratios. It was recommended that more studies are needed before a definitive conclusion can be reached on the extent of applicability and usefulness of electrical measurements in CCTs. As such, the authors concluded that electrical measurements are showing early promise as a technique for the clinical detection of caries in the context of caries clinical trials. Further studies are needed in this area before this method can be adopted in caries clinical trials Laser Fluorescence-DIAGNOdent Another technique that seems to have been developed due to the greater prevalence of occlusal caries, in light of the declining caries rate overall, is DIAGNOdent. Normally, the use of traditional methods for caries diagnosis gives very good values for specificity. This implies that few false positives are detected and mistaken operative interventions are kept to a minimum; therefore, more teeth may remain non-restored and can be treated with appropriate preventive methods. If one could combine this good specificity with a technique that would have a higher sensitivity, then caries diagnosis overall could improve. An improved sensitivity would mean the ability to detect the presence
WEFEL
ET AL.
of caries correctly has also improved. The use of DIAGNOdent was also reviewed at the Clinical Caries Trials workshop in Scotland [Lussi et al., 2004], and it was pointed out that the exact nature of the fluorescence is still uncertain. It may involve porphyrins of bacterial origin, which would need to diffuse into the tooth as demineralization is increased. Thus, one would not expect to have greater fluorescence in early in vivo lesions or in artificially created lesions unless bacteria are present. A table of sensitivity and specificity values was given, using in vitro data but not clinical caries trial data. There were 4 clinical articles [Lussi et al., 2001; Pinelli et al., 2002; Sheehy et al., 2001; HeinrichWeltzien et al., 2002] and a host of in vitro studies reviewed in this paper. Collectively, the clinical trials showed that DIAGNOdent would be appropriate as a second opinion device and gave good intra-examiner reproducibility. Only 1 study was found that looked at inter-examiner reproducibility on smooth surfaces and showed a substantial kappa value. The gold standard in these clinical trials is problematic but was taken as the extension of caries after the opening of the fissure in vivo. This is one of the limitations of this type of clinical trial. In spite of this limitation, DIAGNOdent has been shown to have good sensitivity and specificity, particularly at the more advanced D2 and D3 levels [Pitts, 2004] and is most appropriate for occlusal and accessible smooth surfaces. It was stated that, for longitudinal monitoring of the caries process, the standardized orientation of the DIAGNOdent stylus is a prerequisite. Finally, a caution as to false positives due to various fluorescent materials was noted. As mentioned on the DIAGNOdent website, for example, these false positives include organic plugs, stain, composites, calculus, some sealants and pastes, as well as impacted food. The measurement of remineralized tissue may produce fluorescence and could therefore be incorrectly interpreted as well. Since the above review, several other studies have been published on LF and clinical usage. Anttonen, et al. [2003] reported on the use of LF as part of routine dental check-ups for the detection of occlusal caries in children. Primary molars were found to have lower LF values than permanent molars for sound surfaces but similar values when dentinal caries was detected. Higher visual scores corresponded with higher fluorescence values, but the variation in these values was large. In conclusion, the use of DIAGNOdent was said to be a useful adjunct to visual examination and could be done without prior cleaning in children with moderately good oral hygiene. In a follow-up to this study, the authors attempted to measure the ability of the device to monitor caries longi240 INDIANA CONFERENCE 2005
tudinally by re-examining the children 1.19 yrs. later [Anttonen, et al., 2004]. In permanent teeth, if the visual score went from sound to caries, the mean LF value increased from 24 to 37, while the primary value went from 8 to 40. The mean LF value at baseline was higher for teeth that became carious 1 year later. A visual reversal to sound or inactive caries resulted in a change from 36 to 24 in the LF value. The authors concluded that DIAGNOdent is useful in monitoring occlusal caries in both permanent and primary molars. In an in vivo study to evaluate how different cutoff levels effected the detection of occlusal caries, kappa values and ROC curves were used to determine optimal levels [Heinrich-Weltzien et al., 2003]. Permanent molars of 19yr. olds were measured after professional tooth cleaning, and the tooth fissures were subsequently opened to determine the extent of caries. Differences existed between the manufacturers cutoff values and those based on clinical trials and in vitro studies. The final conclusion was that the agreement between the extent of validated caries and laser fluorescence value is still unsatisfactory. Until more clinical data are available, it was stated that the results should be interpreted carefully. At the more recent Indiana Conference2003, Eakle, et al. showed good reproducibility among examiners. The authors stated that the LF has the ability to help detect hidden caries but could not determine depth. It was determined to be a good negative predictor but again it was stated that one had to be aware of false positives. A clinical study by Tranus et al. [2004] also showed poor correlation between lesion depth and the DIAGNOdent readings of two separate devices. Systematic differences between the devices were also noted and required individual cut-off values of dentinal lesion to be used in order to get good qualitative agreement. There has also been some debate on the need for professional cleaning prior to the measurement for caries. Two articles on this topic have recently been published in Caries Research [Lussi, et al., 2005; Anttonen, et al., 2005]. Cleaned and uncleaned teeth gave differences in values in both studies. Both studies conclude that professional cleaning is helpful in reducing variability of the measurements. Anttonen said that cleaning should be done in teeth with visible plaque and in cases where LF readings approach threshold levels for operative intervention. Similarly, Lussi stated that LF was developed for use as a second opinion device in the detection of occlusal caries and to use LF for monitoring enamel lesions, teeth should be cleaned and dried prior to measurements.
WEFEL
ET AL.
This should not be a problem since most visual inspection systems also require clean and dry surfaces for the most meticulous examination for caries detection. FOTIDIFOTIDIAGNOdent Comparison A different approach to validating caries detection devices has been used in our program project grant. This model involves the use of children, whose teeth were assessed during the clinical phase of the trial, using various detection devices. After these teeth naturally exfoliated, they were then sectioned and evaluated by polarized light microscopy as the gold standard. Cavitated teeth were excluded from the analysis, along with those that had restorations or sealants. One hundred and nineteen children 8-12 yrs. of age were recruited for this study, and they provided a total of 188 exfoliated teeth for evaluation. The FOTI scoring system was 0 = no caries; 1 = small caries; 2 = large caries, while the DIFOTI system was 0 = no caries; 1 = probably no caries; 2 = not sure; 3 = probably caries present; 4 = caries present. The criteria chosen for identifying caries were: FOTI > 0, DIFOTI > 3 or 4, and DIAGNOdent > 14. The LF system is a continuous one, and a cutoff value for caries was taken as greater than 14. This should enable detection of enamel lesions that shows demineralization, although no surface breakdown is evident. The teeth to be measured were limited to upper first and second primary molars and lower first and second primary molars. Specific sites (fig. 1) on each molar were chosen for measurements by all 3 techniques, which were subsequently compared to the histological results of the polarized light microscope (PLM). A classification system was devised to segregate the caries severity into 4 main groups, progressing from early enamel involvement to extreme dentinal involvement and cavitation. The scoring system is shown in fig. 2, with examples of PLM sections. It should be noted that E1 lesions are rarely detected by radiography and that little surface breakdown is noted prior to a D2 lesion [Ando et al., 2005]. Thus, this scale is intended to study early caries detection with the goal of early intervention and the prevention of further demineralization. The clinical exams were done in a classroom, using portable equipment at 6, 12, 18, and 24 months. The visual exam was used to determine all measurable sites and was followed by FOTI, DIFOTI, and LF conducted at different stations by separate examiners.
242
FIG. 1:
Individual examination sites for each molar.
Fiber Optic Transillumination (FOTI) Each site was examined visually, first with Denlite (USA), then with Fiber-optic Transillumination (FOTI, Schott KCL 1500 LCD light, settings C4/3000K, Doncaster, UK). The FOTI instrument had a halogen lamp (150 W) and a probe with a 0.5-mm tip, and it was set to the maximum intensity. Both the visual and FOTI scores were recorded and, for the final call in the event discrepancy, the higher score was used for the analysis. Both the visual and FOTI examinations were performed with no ambient light in the room.sis software (Media Cybernetics, Silver Spring, MD). Digital Imaging Fiber Optic Transillumination (DIFOTI) The images were observed and later acquired in a darkened room after air-drying with Digital Imaging Fiber-optic Transillumination (DIFOTI, Electro-Optical Sciences Inc., NY, USA). DIFOTI incorporates a low-voltage 50 W arc lamp to transilluminate the tooth and an image relay mirror and CCD camera to display the image on the computer monitor and acquire the image. There are 2 illuminating systems available, one for the smooth/approximal surfaces and the other for the occlusal surface. For smooth surfaces, the light-emitting component is placed on the opposite side of the tooth surface to
WEFEL
ET AL.
FIG. 2: 244
Lesion severity classification. INDIANA CONFERENCE 2005
be examined. For occlusal surfaces, the light-emitting component is placed on the buccal and lingual surfaces of the tooth, illuminating the tooth at angles through both the buccal and lingual surfaces, allowing the image to be observed from the occlusal surface. Each site was scored by observing the computer monitor. Laser Fluorescence (LF) The infrared laser fluorescence (LF, DIAGNOdent 2095, KaVo, Germany) examination was performed in accordance with the manufacturers instructions. A tapered fiber-optic tip (type A) was used. After being calibrated by measuring a region of sound tissue on each patient (subject), the probe was placed on the examination site. The peak value was recorded (scale from 00 to 99). The children were given vials that hold the exfoliated deciduous molars, then the molars were mailed to the examiners. White light images and teeth were sent to Iowa, along with the list of sites examined on each surface. At the University of Iowa, the teeth were sectioned and evaluated with polarized light microscopy for mineral loss. The results of each exam were matched with the polarized light microscopy evaluations to produce sensitivity and specificity values for comparisons. These are shown in Table 1 for all surfaces, occlusal only and smooth only.. An easier comparison between surfaces can be made by using the bar graphs in fig. 3 for all surfaces and fig. 4 for individual surfaces. These data show that the specificity remained high, while varying sensitivities were displayed by the different detection devices. One would expect the FOTI and DIFOTI devices to do better on smooth surfaces, but it is surprising as to how well they performed on occlusal surfaces as well. The data may also be evaluated according to severity as determined by the PLM evaluations (Table 2). The bar graph in fig. 5 shows the individual sensitivities of each detection technique, with the severity of carious lesions present. This revealed that, as lesion severity increased, the ability of DIAGNOdent to detect those lesions also increased. This is in agreement with previous authors in terms of detecting more severe demineralization with LF. It would again demonstrate its usefulness as a second opinion device [Lussi, et al., 2004]. Another interesting finding was the sensitivities of the transillumination devices. The best detection occurred at the E2 level and did not improve with further lesion severity. This
WEFEL
ET AL.
TABLE 1:
FIG. 3: 246
Sensitivity and specificity for all surfaces. INDIANA CONFERENCE 2005
TABLE 2:
FIG. 4:
Sensitivity and specificity by surface. 247
WEFEL
ET AL.
FIG. 5:
Sensitivities by lesion classification.
implies that the devices were capable of detecting lesions once they reached the E2 level of demineralization but did not improve in detection with increasing severity. This is in agreement with in vitro studies that stated DIFOTI could not be used to determine lesion depth but merely detection (surface phenomenon). Conclusions In conclusion, it could be stated that all 3 techniques gave reasonable specificities, but that the sensitivities were disappointing on smooth surfaces. Better sensitivities were noted on the occlusal surface with DIFOTI and FOTI but neither device could be used to monitor lesion progression once it was detected. The only change that could be expected to be identified was a reversal back to the E1 level of demineralization. All of these techniques may prove to be a valuable adjunct to diagnostic discrimination, but none should be used alone and, thus, they should not replace a visual exam. The use of DIAGNOdent seems to improve with the severity of lesions as is seen in the permanent teeth clinical trials previously summarized. Depending on the type of caries clinical trial, certain techniques may be useful in increasing sensitivity of the diagnosis.
No one detection method is a panacea for all trials and should be used only in appropriate situations.
References
Ando M, Krushinski C, Ferreira-Zandon AG, Eggertsson H, Stookey GK, Ofner S, Harless J, Wefel J: Evaluation of Digital Imaging Fiber Optic Trans-Illumination (DIFOTI) and FOTI. J Dent Res 2005;84(Spec. Issue A):#2060. Anttonen V, Seppa L, Hausen H: Clinical study of the use of the laser fluorescence device DIAGNOdent for detection of occlusal caries in children. Caries Res 2003;37:17-23. Anttonen V, Seppa L, Hausen H: A follow-up study of the use of DIAGNOdent for monitoring fissure caries in children. Community Dent Oral Epidemiol 2004;32:312-318. Anttonen V, Seppa L, Hausen H: Clinical study on the effect of professional tooth cleaning of occlusal tooth surfaces on laser fluorescence measurements. Caries Res 2005;39:280-283. Ashley PF: The use of electrical impedance measurements for the diagnosis of posterior occlusal caries in clinical trials (PhD thesis). Manchester UK: University of Manchester. 1997. Ashley PF, Elwood RP, Worthington HV, Davies RM: Predicting occlusal caries using the Electronic Caries Monitor. Caries Res 2003;34:201-203. Baysan A, Lynch E, Elwood R, Davies R, Petersson L, Borsboom P: Reversal of primary root caries using dentifrices containing 5,000 and 1,100 ppm fluoride. Caries Res 2001;35:41-46. Bedinskaite R, Sabatalaite R, Gendvilyte A, Huntington E, Matuliene G, Balciuniene I, Nicholsen J, Matheson J, Pitts N, Chesters R: Comparisons of DIAGNOdent and ECM techniques with radiography for the assessment of caries. Caries Res 2001;35:279-280. Eakle WS, Gansky SA, Zhan L, Featherstone JDB: Clinical evaluation of the DIAGNOdent device. In: Early Detection of Dental Caries: Proceedings of the 6th Indiana Conference, Stookey, GK, editor. Indianapolis, Indiana University School of Dentistry, 2003, pp 179-194. Ekstrand KR, Ricketts DN, Kidd EAM, Qvist V, Schou S: Detection, diagnosing, monitoring and logical treatment of occlusal caries in relation to lesion activity and severity: An in vivo examination with histological validation. Caries Res 1998;32:247-254. Ferreira-Zandon AG, Ando M, Eggertsson H, Katz B, Eckert GJ, Ofner S, Mau M, Kelly S, Doi T, Lukantsova L, Wefel JS, Stookey GK: Clinical validation of caries detection methodologies: Final report. Caries Res 2005;39:Abst. #55. Heinrich-Weltzien R, Kuhnisch J, Oehme T, Ziehe A, Stosser L, Garca-Godoy F: Comparison of different DIAGNOdent cut-off limits for in vivo detection of occlusal caries. Op Dent 2003;28-6:672-680. Huysmans MCDNJM: Electrical measurements for early caries detection. In: Early Detection of Dental Caries: Proceedings of the 4th Annual Indiana Conference, Stookey, GK editor. Indianapolis, Indiana University School of Dentistry, 2001, pp 123-142.
249
WEFEL
ET AL.
Longbottom C, Huysmans MC: Electrical measurements for use in caries clinical trials. J Dent Res 2004;83:C76-C79. Lussi A, Megert B, Longbottom C, Reich E, Francescut P: Clinical performance of a laser fluorescence device for detection of occlusal caries lesions. Eur J Oral Sci 2001;109:14-19. Lussi A, Hibst R, Paulus R: DIAGNOdent: An optical method for caries detection. J Dent Res 2004;83:C80C83. Lussi A, Longbottom C, Gygax M, Braig F: Influence of professional cleaning and drying of occlusal surfaces on laser fluorescence in vivo. Caries Res 2005;39:284-286. Pinelli C, Serra MC, Loffredo L: Validity and reproducibility of a laser fluorescence system for detecting the activity of white-spot lesions on free smooth surfaces in vivo. Caries Res 2002;36:19-24. Sheehy EC, Brailsford SR, Kidd EAM, Beighton D, Zoitopoulos L: Comparison between visual examination and a laser fluorescence system for in vivo diagnosis of occlusal caries. Caries Res 2001;35:421-426. Tranus S, Lindgren LE, Karlsson L, Angmar-Mnsson B: In vivo validity and reliability of IR fluorescence measurements for caries detection and quantification. Swed Dent J 2004;28:173-182.
250
Workshop Group on Statistical Modeling Strategies

Albert Kingman, Chairman Division of Clinical Research and Health Promotion National Institute of Dental and Health Promotion National Institute of Dental and Craniofacial Research Bethesda, MD, USA
The comments of the Workshop Group on Statistical Modeling Strategies are predicated on the comprehensive definition of caries adopted at the Loch Lomond ICW-CCT and are generally also reflected in the approach of the Consensus Conference on the Diagnosis and Management of Caries Throughout Life. This definition focuses on net demineralization rather than cavitation as the defining caries process. However, it is recognized that the natural history and clinical course of caries are not fully understood, and the new measurement methods and technologies discussed at this conference offer exciting opportunities to improve our understanding of the natural caries process through clinical epidemiology and to further refine this definition. Consequently, all recommendations are contingent on the use of clinical trials measuring the continuum of caries stages to study in detail the progression from early to later stages of caries, and the extent to which both progression towards and treatment effects on cavitation are paralleled by progression and treatment effects in early stages. The Workshop Group on Statistical Modeling Strategies recommends, based on the evidence presented, that: 1. The development of a visual (tactile) system of caries detection be continued. This system should include: a. Non-cavitated and cavitated stages of caries. b. A measure of lesion severity on an ordinal scale. 2. Further research into definition and validation of activity for both noncavitated and cavitated caries lesions should occur: a. At a single point in time. b. By progression/regression over time.
KINGMAN
ET AL.
The definition of dental caries as a multi-factorial disease capable of inducing progressive net demineralization of tooth structure and cavitation with dentinal involvement simply as an outcome of this process has direct implications for therapeutic evaluation. Under this definition, therapies that sufficiently prevent, retard, or reverse demineralization must be beneficial against cavitation provided that: i) therapeutic activity extends to demineralization across many levels of progression, and ii) no countervailing side effect or safety issue intervenes. Under these circumstances, measurement scales that are sufficiently reliable and account for development and evolution of non-cavitated lesions can bring additional information into clinical trial outcome assessments and, hence, contribute to more efficient trial designs. However, it will be essential that such designs include patients at sufficiently high risk of cavitation that treatment effects on cavitation and non-cavitated lesions be observable in the trial, even if statistical significance is assessed only on their combination. The ICDAS II and Nyvad measurement systems constitute different but related approaches for incorporating information from visible, non-cavitated lesions into clinical trials. Evidence is accumulating to support the intra-observer reliability and inter-observer agreement of sufficiently trained examiners using each of these approaches. Each has been shown at least once to produce effective treatment discrimination in trials of known active agents. We are optimistic that one or more of these measurement systems, or fully ordinal variants thereof, will prove useful in future trials, particularly if the information in these scales is exploited beyond dichotomization. However, evidence presented to date does not allow a full and impartial review and assessment of the validity and range of applicability of these methods from available publications and, as yet, unpublished data. It would be helpful, in understanding these measurement systems, to see reliability and agreement assessments restricted to the range of non-cavitated lesions in which additional information from the scale, and possibly new clinical guidance, are to be extracted. For instance, for the Nyvad measurement system, it would be useful to know meas252 INDIANA CONFERENCE 2005
STATISTICAL MODELING STRATEGIES
ures of agreement restricted to teeth that observers agree are non-cavitated, but on which at least one observer has detected a non-cavitated lesion. The use of hierarchical kappa statistics or a similarly directed approach could address this issue. 3. Efforts should be continued to refine study design to improve the efficiency of clinical trials of monograph products. The Biesbrock-Bartizek Model is an instance of a clinical trial design optimized by several choicespatient population, selected mode of delivery of the experimental therapeutic agent, choice of comparative control intervention, and other featuresto maximize signal-to-noise ratio and statistical power of the trial, while maintaining internal validity. Trials of this nature are sometimes called proof-of-principle trials, because they allow recognition of the maximum possible therapeutic benefit of the experimental intervention under optimized conditions. Such trials generally require smaller samples to demonstrate treatment effect with specified power than do trials under conditions more relevant to general usage. Although the internal validity of proof-of-principle trials tends to be high, their generalizability to other populations or circumstances, i.e., their external validity, tends to be limited. Proof-of-principle trials may play an important role in the evaluation of future therapeutic interventions. However, they should never wholly replace trials conducted in the broader patient population for which an experimental intervention is intended and under circumstances akin to those in which the intervention is likely to be used. 4. The development and testing of technology to monitor the activity and severity of individual carious lesions should continue. The use of diagnostic tests (LF, FOTI, DIFOTI, QLF, ECM, DIAGNOdent, digital radiography, etc.) as adjuncts to, rather than a substitute for visual-tactile caries detection, is encouraged. New technologies for caries assessment, including Quantitative LightInduced Fluorescence (QLF), Electrical Conductivity Measurement (ECM), digital radiography, FiberOptic Transillumination (FOTI) and Digital FiberOptic Transillumination (DIFOTI), Diagnodent, and other methods of enhancing visualization of the tooth surface offer promise. However, their utility for use in clinical trials is not fully established. Some of the following components
KINGMAN
ET AL.
of true validation appear to be missing for these technologies, at least in the context of general application: a. Reliability and agreement assessments, including all components of measurement variability in the process of data acquisition and interpretation, e.g., image extraction as well as image interpretation. b. Assessments and comparisons of sensitivities and specificities at multiple measurement cut-points, so that sensitivities can be compared at comparable specificities and specificities compared at comparable sensitivities, i.e., ROC analyses. c. Assessment of the added value provided by the technology over and above that from trained and calibrated examiners without the technology. Trial designs such as that used by Chesters et al. can be used more extensively to closely analyze the parallel and additional information provided by multiple technologies and help provide such assessments. 1. Developing the evidence: a. Modeling the performance of new methods for detecting caries lesions and disease activity, and for predicting cavitation. ROC performance models. b. What should the criterion be for success of a new method? 2. Are there special considerations for modeling new caries outcome measures, e.g., DIFOTI, contrast radiography, etc.? a. Highly skewed data b. Ordinal modeling c. Misclassification error models 3. Given the high dropout in many caries trials and the general perception that dropout is random w.r.t. outcome, the Loch Lomond workshop took a more tolerant position towards per protocol analyses than is usually accepted by clinical trialists in other areas. How should modern missing data methods be applied in caries clinical trials? Should multiple imputation, selection models, and/or pattern mixture models be used in primary analyses? If so, are there particular approaches to modeling and imputation that can be recommended based on past experience and knowledge of the subject matter?
254
STATISTICAL MODELING STRATEGIES
Nyvad Need a standard of activity, e.g., microbiological. Need reliability, reproducibility measures specifically for visualizing noncavitated lesions, and for differentiating active and inactive lesions. Need ROCs for comparing various quantitative procedures. Caution about studies based on orthodontic model. Most of these remineralize on their own linearly and do not reflect caries biology.
255
256
The Nyvad Criteria Breakout Group Report

Members: Bente Nyvad (Chair), Ken Anusavice, Jaap ten Bosch, Margherita Fontana, Ana Rita Duarte Guimares, ine Lennon, Steven Levy, Cor van Loveren, Vita Machiulskiene, Peter Rechmann, Chris Sissons, Steven Steinberg, Livia Tenulta, Domenick Zero The following represents the consensus position of the group. Discuss the various study designs. The design of the Nyvad criteria is based on sound principles of clinical experience and fundamentals of cariology. Discuss supporting validation data. It was agreed: That the criteria were validated (construct and predictive validity) for use by the group by which they were designed (Aarhus/Kaunas). Determine the strengths and limitations of the model. Strengths: Nyvads criteria have pioneered the use of caries activity assessment. They are simple, easily applicable, and reliable when used by trained/calibrated clinicians and researchers. The relationship between surface roughness and lesion activity is biologically plausible. The criteria have the potential for improving oral health care (diagnosis, risk assessment, and treatment planning), because of their prognostic value. Limitations: The criteria have not been widely tested for their generalizability in different populations and by other examiners. However, it should be noted that published epidemiological studies have successfully applied the Nyvad criteria in Brazil, Columbia, Denmark, and Lithuania in age groups 9-17 years of age.
257
NYVAD
ET AL.
The criteria are not designed to quantify lesion progression (increase in size and depth) within a severity code. As with all clinical indices, there is subjectivity in the interpretation of the clinical criteria, which requires training and understanding of the caries process. Studies to determine the relative importance of the different components of the clinical criteria (tactile feel, texture, presence of plaque) would expand our understanding. Define the most appropriate use for future research of innovative cariespreventive measures. The Nyvad criteria, which includes caries lesion activity assessment, has potentially broad application for: RCT caries intervention trials including all stages of caries disease process. In situ caries model studies.
258
2005 Indiana Conference Workshop Session: Chesters Model

Brian Clarkson, Stu Fischman, Marisa Maltz, Carlos Gonzlez-Cabezas, Mibu Uemura, Andreas Schulte, Joana De Carvalho, Ana Maria Acevedo, Fred Schafer, Richard Chesters, Roger Ellwood (part-time)
Introduction The Chesters abbreviated caries clinical trial design is based primarily on the results of an RCT that was published in the Journal of Dental Research 8 (9):637-640, 2002. The principle features of the trial are summarized below: Proof of principle approach tested products previously shown to have different anticaries activity. Double blind, well-controlled, randomised clinical trial conducted in a high caries prevalence population. Subjects age chosen to maximize caries incidence (i.e., period following eruption of molar). Subjects selected on the basis of caries status, dental age, and brushing habits. Vilnius Clinical TrialKey Elements Caries assessment includes non-cavitated enamel lesions (D1 threshold) using a well established, robust, reproducible caries index. Caries progress measured as events (+ve and -ve) for all surfaces, but could be restricted to molars or even use the proportion of surfaces remaining sound. Validate trial by separating products using traditional caries increment analysis after normal duration. Consensus Feedback 1. Are adequate data available to support the use of this model for controlled clinical trials of caries-preventive agents?
CHESTERS
ET AL.
2.
3.
4.
5.
6.
7.
8.
Yes. The data from Vilnius support the results from other studies, such as the 3 reported by Biesbrock, where non-cavitated enamel lesions are included as decay. Do any diagnostic systems or models that incorporate an ordinal or continuous outcome improve the efficiency of conducting caries clinical trials compared to the traditional D3MFS based method? Yes. Which scoring or diagnostic system has the greatest potential for use among a board range of examiners and study populations? The basic principle should be to keep the system as simple as possible. Is there a realistic potential for conducting valid caries clinical trials in shorter time periods? Yes, as demonstrated by the Vilnius Proof-of-principle caries clinical trial and the 3 Biesbrock clinical trials. Can one obtain valid findings for an anticaries drug by assessing only molars and premolars in a caries clinical trial? Yes, posterior teeth can demonstrate anti-caries efficacy as shown by the Vilnius Proof-of-principle caries clinical trial. The group commented that the inclusion of premolars adds little, if any, value beyond use of just molars. The study reported by Chesters et al. [2002] demonstrated the effect of the anti-caries agent on the caries process. Can researchers achieve a consensus as to how much mineral loss in enamel is clinically important? Any caries lesion that is visually detected should be considered clinically important, taking account, if possible, of its activity. If you found adequate data available, what modifications of the current guidelines should be recommended? The guidelines should allow for shorter time periods. Acceptance of non-cavitated enamel caries lesions in the criteria. The group does NOT consider non-cavitated enamel caries to be a surrogate, but rather a part of the caries continuum. If you did not find adequate data available, what additional data is required for the acceptance of this model for such clinical trials?
260
CHESTERS MODEL
We found the data to be adequate; however, it may be possible to increase the discriminatory power of the approach by using other indices and measures of activity. Key Features of the Recommended Chesters Approach for Assessing Relative Anti-caries Efficacy Double blind, controlled, randomised clinical trial of 1-years duration conducted in a high caries prevalence population. Age of trial subjects where caries is most likely to develop. Promote subject compliance of test product use. Subjects inclusion criteria include caries status, stage of tooth eruption, and, if appropriate, daily toothbrushing. Caries assessment to include non-cavitated enamel lesions on clean, dry teeth using a well established, robust, reproducible caries detection index. Assess all teeth and surfaces, but have the primary outcome variable restricted to molars, then confirm the same trend observed for all teeth. Finally, the group supported the position that the Chesters approach was suitable for comparing the efficacy of 2 products. The recommendation was that it would better to assess product safety in a custom designed study rather than using a trial designed to test efficacy. Similarly, if the objective is to measure the effectiveness of a product, this should also be tested in a custom designed study.
261
262
2005 Indiana Conference Workshop Session: The Biesbrock-Bartizek Model

Workshop participants: Jerry Vogel, Kevin Donly, Roger Ellwood, Athena Papas, Sukim Hong, Stan Heifetz, Luis Archila, Christine Garbass, Van Thompson, Franklin Garca-Godoy, Aaron Biesbrock
Collectively, the research summarized in this presentation supports that well-controlled, supervised caries clinical trials using a D2-D4 diagnostic criteria can be used to differentiate anti-caries products using relatively small sample sizes (n=150-200 completed subjects per treatment group) at 9 to12 months and 21 to 24 months of treatment. Three independent, randomized, controlled, dose response clinical trials were conducted in which the positive control fluoride dentifrice delivered statistically significantly less caries relative to the control dentifrices at both the 9 to 12 and the 21 to 24 month examination times. The fundamental clinical model design is based on rigid control of all sources of variation combined with testing in caries prone populations. Key tenets of the model include: 1) the use of caries prone populations with a similar range of caries activity, 2) the supervised use of product to ensure compliance with intended product usage, and 3) control of background access to care to reduce variation induced by background restorative dentistry. The workshop panel concluded that adequate data was available to support the use of this model for controlled clinical trials testing the efficacy of cariespreventive agents. Existing guidelines governing caries clinical testing should be modified to allow for shorter term models, where evidence supports, as in this case. The panel also identified potential limitations of this model. Shorterterm studies of 9 months duration may not be adequate to assess long-term product safety in some circumstances. In addition, this model, while ideally designed to answer the specific question of relative product efficacy, is not designed for addressing the issue of generalizability of the treatment effect to the broad community. Generalizability is best addressed through large community-based studies across wide age ranges in geographically disperse areas. Additional areas in which this model could potentially be improved are attemptCLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES 263
BIESBROCK
ET AL.
ing to take control of post-brushing rinsing, utilizing microbial testing/salivary buffering capacity for identifying at-risk populations, and using primary evaluable subset identified apriori as the high caries-prone subgroup (high caries vs. low caries breakouts).
264
Indiana Conference ICDAS Workshop Group Report*

N.B. Pitts1-2
1 2
Dental Health Services Research Unit Centre for Clinical Innovations, University of Dundee, Dundee, Scotland, UK
*Based on a presentation made at the Indiana ConferenceClinical Models Workshop: Remin-Demin, Precavitation, Caries.
Abstract At the Indiana ConferenceClinical Models Workshop: Remin-Demin, Precavitation, Cariesdelegates were presented with updates of a range of systems and models that have become available in the area. Working Groups with wide participation were then formed to discuss, challenge, and debate a series of key questions. At the end of the Workshop, each group then reported back to a plenary session. The International Caries Detection and Assessment (ICDAS) Groups key answers and comments were as follows: 1. Are adequate data available to support the use of this model for controlled clinical trials of caries-preventive agents? For ICDAS Coronal Carious Lesion Detection criteriaYes; For ICDAS Root Carious Lesion Detection and Caries Activity CriteriaNot Yet. 2. If so, what modifications of the current ADA and FDA guidelines should be made or recommended? a. Convene discussions with regulators on what is required for guidelines suited for modern methods of caries prevention and control; b. Allow inclusion of ICDAS Codes 1-4 as part of diagnostic criteria; c. Recognise reversals as remineralizations and outcomes; d. Activity (at one time point) is desirable to be included when sufficient generalisability is demonstrated; d) a facility for appropriate microbiological and biochemical tests should be available. 3. If adequate data are NOT available to support the use of this part of the ICDAS model for controlled clinical trials of caries preventive
265
PITTS
agentsWhat additional data are required for acceptance of the remaining elements of this model for such clinical trials? a. For Root CariesRefine codes further, validate the root caries system, reliability data among examiners in different settings, duplicate some of the earlier studies; b. For the Caries Activity part of ICDAS, we need to validate the system and demonstrate reliability between examiners in different settings; c. For all systems, we need more data in different age groups and communities with low and high Fluoride levels, training aids for different groups, and more help for new examiners regarding the differential diagnosis of early caries from Fluorosis. A further series of questions was also answered by the Group.
Introduction At the Indiana ConferenceClinical Models Workshop: Remin-Demin, Precavitation, Cariesdelegates were presented with updates on a range of systems and models that have become available in the area. Working Groups with wide participation were then formed to discuss, challenge, and debate a series of key questions. At the end of the Workshop, each group then reported back to a plenary session. The International Caries Detection and Assessment (ICDAS) Groups key comments and answers are featured below. The Group debated, challenged, and refined the ICDAS 2005 model, which was supported by both NIDCR and the American Dental Association (ICDAS, 2005) and which had already been peer-reviewed, refined, and updated in March 2005, at a Workshop of 65 content experts from the USA and Europe in Baltimore. Discussion Key areas debated and made clear in the group were that: There were different configurations of ICDAS available for use in clinical research, depending upon the specific research question being studied. There was a good fit between the ICDAS wardrobe of methodological choices and the World Health Organization (WHO) STEPS model for non-communicable diseases.
ICDAS WORKSHOP GROUP REPORT
Building on the ICW-CCT Consensus Statements, it was important that as wide a group of stakeholders as possible understood the dynamics of early caries and the potential to prevent cavities from ever forming. More widespread use of ICDAS codes for caries detection (equivalent to the WHO advanced methods codes of the 1970s) in clinical research would build the evidence base in this area and drive innovation towards new methods of caries prevention and control. Use of the ICDAS criteria in dental practice would facilitate PracticeBased Research and the implementation of research results in practice. Use of the ICDAS criteria in epidemiology, public health, and dental education would also help to achieve a more consistent understanding of current evidence in cariology and its application in both policy and practice. The ideal concept is for optimal clinical visual assessment of caries, which ultimately would be supplemented by additional new tools that provide better and more complete information on both caries detection and caries activity. Caries activity measures (seen as compatible with biochemical estimates in cardiovascular risk) are seen as a holy grail and a key part of a continuing ICDAS research agenda. ICDAS provides a framework, giving a foundation of optimal clinical visual assessment the ability to complement this with lesion detection aids and, then, care planning aids.
Answers to the Questions Posed Three key questions were given to each Workshop Group. These questions were then supplemented by an additional 5 that arose from Dr. Kingmans presentation. 1. Are adequate data available to support the use of this model for controlled clinical trials of caries preventive agents? For ICDAS Coronal Carious Lesion Detection criteriaYes (see ICDAS 2005a). For ICDAS Root Carious Lesion Detection and Caries Activity CriteriaNot Yet.
267
PITTS
2. If there are adequate data available to support the use of this model for controlled clinical trials of caries preventive agents, what modifications of the current guidelines should be recommended? Convene discussions with regulators on what is required for guidelines suited for modern methods of caries prevention and control. - Toothpastes + including, for example, sealing initial lesions, risk assessment, new therapeutic interventions, etc. Allow for the inclusion of ICDAS Codes 1-4 as part of diagnostic criteria. - That is, allow the inclusion of clinical enamel and non-cavitated caries. Recognise reversals as remineralizations and outcomes. - Allow transitions in both directions (remin and demin). Activity (at one time point) is desirable to be included when sufficient generalisability is demonstrated. A facility for appropriate microbiological and biochemical tests should be available. 3. If there are NOT adequate data available to support the use of this part of the ICDAS model for controlled clinical trials of caries preventive agents, what additional data are required for acceptance of the remaining elements of this model for such clinical trials? For Root Caries: - Refine codes further (ongoing debate regarding activity estimates). - Validate the root caries system (may rely on surrogates, possibly via acid assessment and may use new technologies to show changes). - Reliability among examiners in different settings. - Duplicate some of the earlier studies. For the Caries Activity part of ICDAS, we need to: - Validate the system (may rely on surrogates). Define valid biological markers of caries activity. - Demonstrate reliability between examiners in different settings.
268
For all systems, we: - Need more data in different age groups and communities with low and high fluoride levels. - Need training aids for different groups. - Need more help for new examiners regarding the differential diagnosis of early caries vs. fluorosis. Dr Kingmans additional questions: 1. Do any of the diagnostic systems or models that incorporate an ordinal or continuous outcome improve the efficiency of conducting CCTs compared to the traditional D3MFS based method? When we include non-cavitated lesions (ICDAS Code 1 and 2), there is evidence of increased efficiency. We believe that examples from StookeyICDAS method [Stookey, 2005] and Chesters and co-workersDSTM method [Chesters et al., 2005] demonstrate that this is possible, but associations for high disease populations may be different, and further studies are needed for these groups. (Note change in F component may mean that historical validity comparisons are challenged in the future). 2. Which scoring or diagnostic system has the greatest potential for use among a broad range of examiners and study populations? We believe that the ICDAS caries detection criteria have the greatest potential for use among a broad range of examiners and study populations as: - it is based on methods that have already been used in many settings. - it has been designed to be used by multiple examiners in multiple settings. - it has shown feasibility, ease of use, and validity in many settings, with many examiners, in a number of countries. - ICDAS, coupled with newer technologies and risk assessments, may enhance utility further. - the framework has the scope to deal with CARS (Caries Associated with Restorations and Sealants) and Root surface caries, as well being relevant to modern clinical care. 3. Is there a realistic potential for conducting valid CCTs in shorter time periods?
PITTS
With ICDAS caries detection criteria, YES, we believe that there is a realistic potential. 4. Can one obtain valid findings for an anti-caries drug by assessing only molars and premolars in a CCT? Probably for coronal caries in individuals older than 5 years. 5. Can caries researchers achieve a consensus as to how much mineral loss in enamel is clinically important? - Yes: The equivalent for an ICDAS Code 1 = 5% increase in pore volume. - The Group wants to prevent and control disease, so all mineral loss that cannot be physiologically maintained is clinically important. - The Group believes that net velocity of mineral loss over time is important. - This question cannot be answered fully without having more specification regarding the setting. - Note: not all mineral is the same to lose, some can be eliminated, some might want to be keep. Comment from the Group on an Issue Raised in a Morning Presentation Caries as a Surrogate? The group felt that early caries lesions/white spots are NOT a surrogate; they are an end point. Different thresholds: - In clinical trials of caries around orthodontic brackets, the patient can see white spots, which are the end point of the trial. - More people need to be aware of the continuum of the caries process. - Researchers need to state relevant endpoints for studies (under the outline given in Baltimore, the traditional Radike model is a surrogate, as the ultimate outcome given is tooth mortality). Modern Caries Management: aims and objectives from Group members: - in California (as an example), minimally invasive care is relevant NOW. Its aim is to maintain health and reduce the prevalence of restorations.
- in Columbia, groups are working on preventive management NOW. - in Europe and the UK, this style of practice is seen as the norm. - The Director of the NIH maintains that detecting pre-clinical stages of chronic diseases are relevant aims for comprehensive disease management. - White spots and brown spots are now well established as part of the caries process. Conclusions The International Caries Detection and Assessment System (ICDAS) is a structured, evidence based, clinical visual examination system that includes both the early and later clinical stages of the disease process. It is an evolving standard, which seeks to define the best way of using this old technology to act as a link between clinical research/epidemiology/clinical practice and dental education. It is reproducible, in terms of caries detection and feasibility for use in clinical trials, epidemiological, and practice settings. The ICDAS is also designed to act as a caries activity system, and it provides an essential clinical reference standard for the newer, emerging caries detection technologies. Acknowledgements The author gratefully acknowledges all of the members in the Indiana Workshop Group, contributions from the ICDAS Group (including colleagues from the universities of Copenhagen, Dundee, Indiana, and Michigan) and, specifically, from colleagues at the Dental Health Services Research Unit and the Centre for Clinical Innovations in Dundee. The views expressed in this paper are personal and do not necessarily reflect the views of any UK Health Department, any research funder, or sponsor.
References
Chesters RK, Pitts NB, Matuliene G, et al.: An abbreviated caries clinical trial design validated over 24 months. J Dent Res 2002:81; 637-640. International Caries Detection and Assessment System (ICDAS) Coordinating Committee (2005a): Rationale and evidence for the International Caries Detection and Assessment System (ICDAS II). (paper 1 submitted to G Stookey by Amid).
271
PITTS
International Caries Detection and Assessment System (ICDAS) Coordinating Committee (2005b): Criteria Manual, International Caries Detection and Assessment System (ICDAS II). (paper 2 submitted to G Stookey by Amid). Stookey G: (Study referred to by G Stookey & Aaron Biesbrock at the 2005 Indiana Workshop) 2005.
272
The Quantitative Light Fluorescence (QLF) Model

Susan M. Higham, Chairman
Cariology and Toothwear Research Group, School of Dental Studies, The University of Liverpool, Liverpool, UK
This was a popular workshop, which had 24 participants who contributed in an excellent discussion, and the workshop was chaired by Professor Susan Higham. In addition, several other interested delegates listened in on the discussion from time to time over the 2 sessions. First, the Group discussed the study designs that were thought to be applicable to the QLF technique. These were accessible, smooth surfaces, pits and fissure. The consensus was that the smooth surface model, while being useful as a screening system, could not be extrapolated to all populations. Early work has suggested that QLF has the potential for use in the study of root caries, and this is an area that warrants further investigation and development. It was accepted that, currently, QLF cannot be accepted as a feasible technique for the study of early approximal lesions. Supporting evidence for this validation was considered to be excellent in vitro for single endpoints in time for both smooth and occlusal surfaces. There have also been several in vivo studies comparing QLF with radiography and visual assessment [Radike, Ekstrand and ICDAS]. Future work is now needed to establish dose responses to validate changes in lesions. The Group knew of at least 5 studies that are in progress and are due to be completed within the next couple of years. The strengths of QLF were considered to be numerous. One of the most important factors was the ability to longitudinally monitor very early caries with quantification of mineral loss and archive the data as well as being able to provide continuous data. QLF is very useful for the screening of new products, using smooth surface models that also allow the combination of in situ data with in vivo data, which is an important consideration for some studies.
273
HIGHAM
ET AL.
Remote analysis of lesions is also possible; this has several advantages, including reducing the need to train busy clinicians to perform this aspect of the technique and reducing the time required. Indeed, the use of QLF is not restricted to clinicians alone and can be used by other members of the dental team: for example, dental therapists, hygienists, nurses, and dental researchers. In due course, it is anticipated that a large database of lesions and information on how they change will be available to clinicians and dental researchers across the world. QLF was considered to be a very reliable technique, provided that a few simple rules are followed, with the recent introduction of video repositioning (Vidrep) software, which is an important advancement. The major advantage of the QLF technique for use in clinical trials is the fact that it allows lower numbers of subjects to be used and shorter time periods, which clearly have major cost implications. The Group also accepted, as with any method, that QLF had limitations. Accessibility of proximal and subgingival surfaces, as with many other techniques, is a problem. Teeth need to be clean; otherwise, plaque is imaged rather than the tooth tissue. The teeth also need to be dry, and QLF is particularly sensitive to this, more so, perhaps, than other techniques; but this is only likely to be a problem if operators do not adhere to simple guidelines. If teeth are not dry when imaged, lesions will not be visible and, therefore, incorrect interpretations would be made. Fogging of the mirror has been a problem for some investigators, which leads to a reduction in lesion quality, although this problem has been addressed by the manufacturers, and the majority of users no longer consider this to be a problem. Similarly, ambient light was an issue with early systems, and imaging had to take place in darkened rooms away from fluorescent lighting. New ambient light shields that cover the camera and handpiece have addressed this problem. Restorations of teeth are not a limitation of QLF; indeed, it is possible to use QLF to monitor secondary caries around restorations. The group did feel, however, that consideration should be given as to how restored teeth lost during trials should be dealt with statistically. As with much dental equipment, the initial acquisition cost of QLF is fairly high, but may be offset by savings in trial time. The group also recognised that, once purchased, the running costs were extremely low. Stain was considered to be a confounding factor that caused most concern, since when it occurs, QLF sees the stain as a dark area on the enamel, similar to demineralization, which appears dark due
THE QUANTITATIVE LIGHT FLUORESCENCE (QLF) MODEL
to a loss of fluorescence. This is an area that is receiving attention; however, it should be recognised that any diagnostic technique, including QLF, must only be an aid and not a replacement for clinical judgement. The Group then went on to the brainstorming activities that had been set: 1. Is adequate data available to support the use of QLF for controlled clinical trials of caries-preventive agents? Not yet, but it was thought that this will be possible once confirmatory dose response studies have been completed to validate lesion change. 2. If you did not find adequate data available, what additional data is required for the acceptance of this model for such clinical trials? The need for a fluoride dose response with a visual comparison was thought to be essential. A minority of the group felt that at least 3 levels of fluoride should be included. In order for comparisons across validation studies completed in different centres to be compared, it was considered important that the same QLF metrics should be reported. This, in fact, should not be a problem even if, initially, the same metrics are not reported, since, because of the archivable nature of the QLF technique, all data can be reanalysed to provide the required information. Additional work that the Group considered to be advantageous included: 1. A repository of QLF images and information should be set up. 2. Clinically relevant endpoints should be identified; eg., children requiring hospitalisation for extractions. 3. Agreement on the issue of stain. Kingman Question 1 Measurement systems that are ordinal or continuous. How do we deal with surfaces that become filled during a clinical trial? The Group thought that there are diagnostic systems that allow continuous data and that this will improve the efficiency in comparison with D3MFS, but more evidence is required for QLF. Kingman Question 2 Which system has the greatest potential?
275
HIGHAM
ET AL.
The Group thought that QLF had excellent potential for application in caries clinical trials and, particularly, for the potential to reduce subject numbers, duration, and increase diagnostic efficiency. Kingman Question 3 Is there a realistic potential for conducting caries clinical trials in shorter time periods? The consensus was that the evidence supports QLF as one of the methods that will enable caries clinical trials to be conducted in shorter time periods. Kingman Question 4 Can one obtain valid findings for an anti-caries drug by assessing only molar and premolar teeth in a caries clinical trial? Although the findings reported by Chesters at this conference indicated that the results observed on molars reflected the treatment effects observed for the entire dentition, adequate documentation of this relationship and the validity of using only premolars and molars to assess treatment effects still need to be demonstrated. Kingman Question 5 Can caries researchers agree as to what degree of mineral loss is clinically relevant? The consensus was that it is unlikely that researchers will ever agree about what degree of mineral loss is clinically relevant, but it was felt that QLF has the potential to provide one part of the story, but in conjunction with other relevant factors, such as salivary composition and flow, oral hygiene, plaque composition, retention, etc.
276
Other Instrumental Detection Methods for Clinical Caries Trials

James S. Wefel Dows Institute for Dental Research, College of Dentistry, University of Iowa, Iowa City, IA, USA
First, I would like to thank the group members who so willingly contributed to our frank, open discussions and the writing of this report. The diverse nature of the group and the expertise present in the room made my job as moderator an easy one. As a starting point, it was felt that we should reaffirm our understanding of the caries process and its dynamic nature. Dental caries is a multifactorial disease process involving the main factors of acidogenic bacteria, host tissues, and fermentable carbohydrates, which are influenced by the salivary fluids found in the oral cavity. Many more factors have been shown to also play a role in this disease process. The health or disease status of teeth depends on the interplay of the protective versus pathologic factors. The result of this interplay is an equilibrium between demineralization and remineralization, which can be upset when imbalances occur. The imbalance towards caries will cause tissue changes that cover the continuum of demineralization from the atomic, through the clinically visible white spot lesion, to dentinal involvement, and eventual cavitation. Cavitation has been considered as a point of no return; but, prior to that time, reversal of the disease is possible. Thus, mineral can also be gained and not just lost by the tissue. It is important to remember that remineralization or repair/reversal of the disease process can occur on a daily basis. With our ability to quantify and detect demineralization at many points along this continuum, we are able to measure and follow the course of the disease process. The end point of cavitation is not the only measure of dental caries, and early detection is a valid measure of the disease process and not a surrogate. Thus, trials may be designed using endpoints other than cavitation, which may be appropriate for evaluating intervention methods in much shorter timeframes and with less numbers. (See previous model summaries).
277
WEFEL
As we turned our attention to assessing the applicability of using Other Instrumental Detection Methods for Clinical Caries Trials, it became obvious that we did not have a magic bullet that would fill all the needs of the many types of clinical caries trials available. It was also felt that each CCT should be designed to fulfill the desired outcomes. This may be an epidemiological survey, a trial to distinguish between products with known mechanisms, and/or trials designed to look at specific parts of the caries process, e.g., bacterial load, enhance remineralization, improve salivary factors, decrease demineralization episodes, etc. These various trials designs will require measuring changes in mineral content on different surfaces, as well. Therefore, a detection method that is sensitive and accurate on only selected surfaces could be inappropriate, costly, and/or time intensive in some trials. In order to make the most of the assembled expertise and experiences in using these techniques, a compilation of data into a grid or table was thought to be most useful. It was felt that one part of the chart would need to distinguish which surfaces can be measured accurately with a given techniques. Table 1 was designed to show which techniques could supplement the detection of caries on a given surface (+) or whether insufficient data is available to draw that conclusion (-). NA would imply that the detection method is not suited for measuring a given surface for dental caries. In general, it was assumed that surfaces are clean and dry prior to use of the technique, since this is also needed for an appropriate visual inspection. A summary of the usefulness of each technique follows. The use of ECM was most recently summarized by Longbottom and Huysmans at the Loch Lomond meeting [JDR 83; C76-C80, 2004]. One of the
Method Occlusal Occlusal ECM
Surface
Surface Free Smooth Free Smooth Root Root
Approximal Approximal
+ (Surface Specific
2nd molar)
+ (E & D) + +?
-? +
+ (Surface Specific
Activity)
FOTI DIFOTI DIAGNOdent
+ (D) + (E2 & D)
All point measurements will have repositioning issues.

E2=Deep enamel lesion; E= Detection of enamel lesion; D=Detection of dentin lesion; ?=Needs more data; -=Cannot detect lesions;
+ = Can detect lesions.
TABLE 1.
Current detection methods.
278
OTHER INSTRUMENTAL DETECTION METHODS
FOR
problems noted was a limited number of studies using ECM. It was felt that ECM could be used effectively with occlusal surfaces, but not on smooth surfaces at this time and, only one study was available on root surfaces, which was positive in terms of detection (more data needed on this use). The review also mentioned many of the limitations and variables when using ECM. The correlation between examiners and instruments was deemed poor, which implied poor reproducibility, as well. Little, if any, data is available on the use of ECM in the primary dentition. As mentioned, this technique has mixed reviews at this time but still offers tantalizing possibilities. The use of FOTI in CCTS was also one technique in which early data was quite promising but few recent studies are available. The methodology is limited to detection from light scattering, as is DIFOTI. The use of optical methods for caries detection avoids the disadvantage of ionizing radiation, similar to what one gets with x-rays. FOTI was thought to be most useful for proximal surfaces but not occlusal and smooth surfaces. Enamel to dentine changes are the easiest to detect, while any correlation to further depth is limited. This was shown in the study from Indiana on primary teeth, as well. It was felt that reversals could be detected in enamel. Studies, in which monitoring over time was part of the protocol, would need repositioning aid and a careful control of measuring parameters. A similar detection method is used with DIFOTI, but the handpiece and measurement technique limit placement of the light. Both in vitro and in vivo studies showed that, once a lesion was detected, no further information on increasing lesion depth or mineral loss could be detected. Few clinical trials are available in which the detection method has been used. A soon-to-be-published trial was said to show that DIFOTI improved the detection of approximal lesions. Use of DIFOTI to detect lesions on occlusal and free smooth surfaces is limited, and there is no data to support use in CCTS, which needs to evaluate these surfaces. The one advantage of this caries detection technique is the capture of a digital image, but further progression of the lesion is not easily detectable. The use of digital radiography and subtraction radiography offers some advantages over traditional x-rays. One of the obvious drawbacks is the need for ionizing radiation. The most appropriate surfaces are the approximal, since other techniques are limited in their ability to get access to these surfaces. Subtraction radiography requires the monitoring of lesions over time and may
WEFEL
be used to determine activity and progression. Both approximal and occlusal lesions into dentine may be detected. There is little evidence that automatic radiographic screening will benefit populations with low caries experience. Use of digital radiography may reduce the extent of exposure. As summarized at the Loch Lomond Conference, The evidence for dental caries is not restricted to the levels of surface cavitation, and radiography can add information about many of those clinical stages of the caries process at approximal surfaces and the more advanced stages on occlusal surfaces. Recent studies provided far more literature for discussion on DIAGNOdent. From the studies summarized by Lussi at the Loch Lomond Conference, it was determined that DIAGNOdent was best suited as an adjunct for caries detection on occlusal surfaces. These lesions needed to involve most of the enamel or be into dentin in order to be detected with the best sensitivity. It was shown that many clinical trials used the opening of fissures as the gold standard for lesion detection. In contrast, the use of exfoliated primary teeth as a gold standard looked at early lesion detection and found that DIAGNOdent was least sensitive in this situation. The ability to detect lesions increased as the severity of the lesion increased with the DIAGNOdent device, unlike FOTI and DIFOTI. The original device is not suited for approximal caries detection, but a new tip development was presented at the ORCA meeting, which may prove helpful on these hard-to-reach surfaces. This new development was termed the DIAGNOdent pen and should be added to table 2. Monitoring in CCTs will
Novel Methods Occlusal Occlusal Near Infrared
Surface
Surfaces Free Smooth Free Smooth Root Root
Approximal Approximal
+
(Qualitative)
+ + -
+ ? + ?
? (Possible) ? (Possible) -
ps-OCT
+
(Potential)
Ultrasound Raman and Optical Spectroscopy Terahertz
? (Enamel) -
?=Needs more data; -=Cannot detect lesions; + = Can detect lesions.
TABLE 2.
Developing detection methods.
280
OTHER INSTRUMENTAL DETECTION METHODS
FOR
require a positioning device to get reproducible results with these laser fluorescence devices. The ongoing development of other caries detection techniques resulted in significant discussions as to future usefulness. More data is needed in all cases sited, and some needs to adapt the equipment for use in the oral cavity. Other new devices need to be modified to give easier, less time-consuming procedures (table 2).
References
Proceedings from the International Consensus Workshop on Caries Clinical Trials, Ed by Pitts and Stamm, J Dent Res 83, Special Issue C, 2004.
281
282
Concluding Remarks
by George K. Stookey, Ph.D. Let me close this workshop by first thanking all of the presenters and discussion leaders for their outstanding effortsall of the presentations were excellent and reflected the preparations of each presenter. Likewise, the discussion leaders maintained outstanding critical reviews of each topic and developed significant summary reports. And a special thanks is due to all invited scientists and participating experts in cariology who made the discussions very productive. The scientific presentations on the various clinical caries models have been quite illuminating and have clearly indicated a number of potential approaches for designing future clinical caries trials of caries-preventive agents. Some of these approaches have been investigated in clinical caries trials and, as noted in some of the workshop discussion reports, have been adequately validated to justify their use. Most certainly, the results of this workshop reflect a significant step forward in the identification of more efficient models for clinical caries trials. As in the past, we will be publishing the papers from the entire group of presenters and the reports from each of the discussion groups. Copies of the hard cover publication will be sent to all attendees. Thank you for making this conference/workshop scientifically a success.
283

7th Proceedings Indiana Conference

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

7th Proceedings Indiana Conference

Transféré par

Droits d'auteur :

Formats disponibles

George K.

Clinical Models Workshop: Remin-Demin, Precavitation, Caries

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

Introductory Remarks & Workshop Goals

INDIANA CONFERENCE 2005

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

Traditional Models for Clinical Caries Trials

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

CLINICAL CARIES TRIALS

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

coronal and root

Characteristic of the traditional caries detection models. INDIANA CONFERENCE 2005

CLINICAL CARIES TRIALS

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

Requirements Non-cavitated lesion coronal root Cavitated lesion coronal root

yes yes yes yes

Severity (depth) Activity (progresses/arrests) Treatment options

medical and surgical

CLINICAL CARIES TRIALS

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL CARIES TRIALS

What is Going On?

S, D1w, D1d, D1w

S, D1w, D1d, D1w, D2

S, D1w, D1d, D1w, D2, D1w

D1d, S D1d, S, D1d

D1d, S, D1d, S D1d, S, D1d, S, D2 D1d, S, D1d, S, D2,D2

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL CARIES TRIALS

Radikes Criteria (1968)

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CLINICAL CARIES TRIALS

WHO Basic Methods (1971)

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

NIDR Criteria (1987) Coronal Caries

INDIANA CONFERENCE 2005

CLINICAL CARIES TRIALS

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

Conclusions from the 2002 Scotland Consensus Conference (ICW-CCT)*

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

INDIANA CONFERENCE 2005

CONCLUSIONS FROM THE 2002 SCOTLAND CONSENSUS CONFERENCE (ICW-CCT)

INDIANA CONFERENCE 2005

CONCLUSIONS FROM THE 2002 SCOTLAND CONSENSUS CONFERENCE (ICW-CCT)

Three types of modern caries measurement.

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

CONCLUSIONS FROM THE 2002 SCOTLAND CONSENSUS CONFERENCE (ICW-CCT)

CLINICAL MODELS WORKSHOP: REMIN-DEMIN, PRECAVITATION, CARIES

CONCLUSIONS FROM THE 2002 SCOTLAND CONSENSUS CONFERENCE (ICW-CCT)