Académique Documents
Professionnel Documents
Culture Documents
net/publication/12955957
CITATIONS READS
382 1,538
3 authors:
Vibeke Baelum
Aarhus University
196 PUBLICATIONS 7,126 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Vita Machiulskiene on 04 May 2015.
all misclassifications involved non-cavitated caries le- caries progression [Glass et al., 1983]. Furthermore, it has
Eur.Org.for Caries Research
requires a diagnostic system which reflects the dynamic na- ed for extracted teeth and unerupted teeth/surfaces. Tooth surfaces
Eur.Org.for Caries Research
0 Sound Normal enamel translucency and texture (slight staining allowed in otherwise
sound fissure).
1 Active caries Surface of enamel is whitish/yellowish opaque with loss of luster;
(intact surface) 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.
2 Active caries Same criteria as score 1. Localized surface defect (microcavity) in enamel only.
(surface discontinuity) No undermined enamel or softened floor detectable with the explorer.
3 Active caries Enamel/dentin cavity easily visible with the naked eye; surface of cavity feels soft
(cavity) or leathery on gentle probing. There may or may not be pulpal involvement.
4 Inactive caries Surface of enamel is whitish, brownish or black. Enamel may be shiny and feels
(intact surface) 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.
5 Inactive caries Same criteria as score 4. Localized surface defect (microcavity) in enamel only.
(surface discontinuity) No undermined enamel or softened floor detectable with the explorer.
6 Inactive caries Enamel/dentin cavity easily visible with the naked eye; surface of cavity may be
(cavity) shiny and feels hard on probing with gentle pressure. No pulpal involvement.
7 Filling (sound surface)
8 Filling + active caries Caries lesion may be cavitated or non-cavitated.
9 Filling + inactive caries Caries lesion may be cavitated or non-cavitated.
with two or more lesions were classified on the basis of the most Results
severe of these lesions (active lesion c inactive lesion; manifest cav-
ity c surface discontinuity c non-cavitated lesion). Intra-examiner
The percentage of agreement and the kappa values for
examinations were repeated with an interval of 1–2 weeks. The time
spent per examination of each child was approximately 5–8 min. the different intra- and inter-examiner examinations are
shown in tables 3 and 4, respectively. When applying the
Assessment of Reliability new diagnostic criteria (ten diagnostic categories) the per-
The intra- and inter-examiner reliability of the caries diagnostic centage agreement of the caries diagnoses varied between
criteria was assessed at the tooth surface level using five different di-
94.2 and 96.2%. The kappa values ranged between 0.74 and
agnostic thresholds, as outlined in table 2. The cut-off points were the
following: sound versus diseased (diseased including all visible signs 0.85 for intra-examiner examinations and between 0.78 and
of caries); active versus not active (cavitated and non-cavitated le- 0.80 for inter-examiner examinations. The high level of
sions pooled); cavity level (non-cavitated caries and caries with dis- reliability was maintained over the 3-year period (kappa =
continuity recorded as sound); discontinuity level (non-cavitated 0.74–0.84, table 3; examiner V.M.).
caries recorded as sound), and cavity level according to WHO (non-
When kappa values were calculated on the basis of the
cavitated caries and cavities with hard floors recorded as sound). The
results were expressed as percentage agreement and Cohen’s kappa. same data sets using the alternative diagnostic thresholds
The probability of confirming a sound or non-cavitated active or non- (two diagnostic categories; table 2), the kappa values varied
cavitated inactive caries diagnosis at repeated examination was as- somewhat more than would be expected on the basis of the
sessed by examining the distribution of pooled diagnoses (1994– percentage agreement, which remained high at all diagnos-
1996) from the intra- and inter-examiner examinations. The group of
tic thresholds (tables 3, 4). The most dramatic change in
disagreements was analysed separately, and the percentage distribu-
tion of misclassified non-cavitated active and inactive caries lesions kappa was observed when the diagnostic threshold was
was calculated. changed from ‘sound versus diseased’ to ‘active versus in-
193.219.162.172 - 11/13/2013 9:32:03 AM
Table 5 shows the distribution of the intra-examiner data ments between active and inactive non-cavitated lesions.
for examiner V.M., pooled over the 3 years (1994–1996). The remaining 221 misclassifications involved other com-
Given that a tooth surface was recorded as sound at one binations (tables 5, 6). Table 6 shows that the distribution
examination, the probability of diagnosing the same surface of the misclassifications was rather similar for the pooled
as sound at the other examination was 98.0% inter-examiner data and for the intra-examiner data for ex-
[=100 E 2E14,486/(14,764+14,786)]. The corresponding aminer B.N.
figures for non-cavitated active lesions and non-cavitated
inactive lesions were 68.7 and 72.5%, respectively. The
probability of reconfirming a non-cavitated caries diagnosis Discussion
was moderately higher for occlusal surfaces than for
smooth surfaces (data not shown). The results for the inter- It is a fundamental premise for both clinical and epide-
examiner data and for the intra-examiner data for examiner miological studies of dental conditions that diagnostic
B.N. were essentially similar (data not shown). methods exist which may provide consistent and standard-
The total number of misclassifications in the pooled data ized expressions of the condition in question. This require-
set (table 5) amounted to 821 (= 17,164–14,486–398–521– ment places emphasis on the issue of reliability of the diag-
232–9–556–73–68); 670 of the misclassifications (81.6%) nostic methods available. Thus, the purpose of the present
involved the two non-cavitated diagnoses (tables 5, 6). A study was to evaluate the reliability a new set of clinical
total of 257 misclassifications (31.3%) concerned disagree- caries diagnostic criteria which provides information about
ment between sound surfaces and non-cavitated active le- the activity of caries lesions in addition to distinguishing
sions, 256 misclassifications (31.2%) concerned disagree- between cavitated and non-cavitated stages of caries lesion
ment between sound surfaces and non-cavitated inactive formation.
193.219.162.172 - 11/13/2013 9:32:03 AM
Intra-examiner V.M. 1994 Intra-examiner V.M. 1995 Intra-examiner V.M. 1996 Intra-examiner B.N. 1996
(ns = 5,510) (ns = 5,450) (ns = 6,204) (ns = 6,118)
New diagnostic criteria % agreement 95.5 % agreement 94.2 % agreement 95.6 % agreement 96.2
(ten categories) kappa 0.84 kappa 0.74 kappa 0.83 kappa 0.85
Diagnostic threshold – + – + – + – +
Sound vs. diseased – 4,594 74 4,685 107 5,207 119 5,190 73
(two categories) + , 86 756 ,116 542 , 76 802 , 77 778
% agreement 97.1 % agreement 95.9 % agreement 96.9 % agreement 97.5
kappa 0.89 kappa 0.81 kappa 0.87 kappa 0.90
Active vs. inactive – 5,065 45 5,109 75 5,812 92 5,714 58
(two categories) + , 95 305 , 86 180 , 61 239 , 71 275
% agreement 97.5 % agreement 97.0 % agreement 97.5 % agreement 97.9
kappa 0.80 kappa 0.68 kappa 0.74 kappa 0.80
Discontinuity level – 5,362 15 5,258 22 6,095 14 5,897 16
(two categories) + , 22 111 , 47 123 , 6 89 , 42 163
% agreement 99.3 % agreement 98.7 % agreement 99.7 % agreement 99.1
kappa 0.85 kappa 0.77 kappa 0.90 kappa 0.84
Cavity level – 5,406 14 5,341 10 6,154 6 6,006 11
(two categories) + , 17 73 , 25 74 , 3 41 , 16 85
% agreement 99.4 % agreement 99.4 % agreement 99.9 % agreement 99.6
kappa 0.82 kappa 0.81 kappa 0.90 kappa 0.86
WHO cavity level – 5,369 16 5,263 23 6,098 13 5,908 15
(two categories) + , 19 106 , 46 118 , 7 86 , 42 153
% agreement 99.4 % agreement 98.7 % agreement 99.7 % agreement 99.1
kappa 0.86 kappa 0.77 kappa 0.89 kappa 0.84
The data are given for the new diagnostic system (ten categories) as well as for different diagnostic thresholds collapsing the ten categories
into two. – and + represent negative and positive diagnoses, respectively.
The results of the present study demonstrate that when agreement that is expected as a result of chance. This said,
the new criteria are applied to the site-specific diagnosis of it is also clear that the kappa statistics has several draw-
caries lesions under real-life conditions in a clinical trial backs. One of these is that the kappa value depends on the
they can be reproduced by the same examiner or by another underlying true, but unknown, prevalence of the condition
examiner to the extent of almost perfect agreement. This being diagnosed [Altman, 1991]. This may be illustrated by
conclusion applies both when evaluated in terms of the per- the results in tables 3 and 4 which show that with approxi-
centage agreement of diagnoses, which always exceeded mately the same proportional agreement (95.9–99.9%), the
94%, and in terms of the value of the kappa statistics, which kappa values varied considerably more (0.68–0.90). The
essentially always exceeded 0.70, and usually exceeded consequence of this prevalence dependence is that poten-
0.80, indicating almost perfect agreement [Landis and tially misleading conclusions may be drawn if comparisons
Koch, 1977]. Moreover, the kappa values reported here are are made of the kappa values obtained in different studies
in the same order of magnitude as those reported by other where the caries prevalence rates are likely to be different.
researchers who have included non-cavitated caries diag- For the same reason it may be misleading to make use of a
noses into the criteria system [Pitts and Fyffe, 1988; Manji comparison of the different kappa values obtained for dif-
et al., 1989; Ismail et al., 1992]. ferent caries diagnostic thresholds within a given study to
Use of the kappa statistics to evaluate the reliability of a determine which cut-point yields the highest reliability
given diagnostic method has the advantage over the value [Pitts and Fyffe, 1988]. Therefore, the only tenable conclu-
193.219.162.172 - 11/13/2013 9:32:03 AM
for percentage agreement in that it takes into account the sion that may be drawn on the basis of the results presented
Eur.Org.for Caries Research
in tables 3 and 4 is that the present study provides no evi- The third drawback of the kappa statistics as a measure
dence that the reliability of a simple caries recording system of diagnostic reliability is that the kappa value depends on
in which only cavitated caries lesions are included would be the number of categories in the diagnostic system. Tables 3
substantially higher than the reliability of more elaborate and 4 thus illustrate that the kappa values for the original
and sensitive caries diagnostic systems. This conclusion diagnostic system with ten categories were generally lower
corroborates the observations by Pitts and Fyffe [1988] and – although remaining within the range of good to excellent
Ismail et al. [1992]. agreement – than the kappa values for the same data sets
A second drawback of the use of the kappa statistics to obtained when collapsing the ten categories into only two
evaluate diagnostic reliability is that the kappa statistic is (‘sound versus diseased’).
rather insensitive to the type of examiner disagreements, i.e. In spite of these shortcomings, we may, in concert with
whether they are systematic or not. However, such know- Altman [1991], conclude that the kappa statistic remains the
ledge could be important, since systematic deviations may best approach to assessing the reliability of caries diagnos-
indicate a diagnostic slide (in case of systematic intra-ex- tic criteria, when the scope is to evaluate the reliability of
aminer disagreements) or the use of different diagnostic the site-specific caries diagnoses. However, it should be
thresholds (in case of systematic inter-examiner disagree- recognized that no single value of kappa exists which can be
ments). In the present study, there were no indications of regarded universally as indicating good agreement.
such systematic inter- or intra-examiner deviations. As previously noted the kappa value cannot be used to
193.219.162.172 - 11/13/2013 9:32:03 AM
Exam 1 Exam 2
Sound non-cavitated cavitated fillings fillings total
active inactive active inactive active inactive
Table 6. Total numbers and percentage distribution of misclassifications recorded at the seven intra- and inter-examiner examinations
1994–1996 (pooled data)
Intra-examiner V.M. 17,164 821 670 (81.6%)a 257 (31.3%) 256 (31.2%) 87 (10.6%) 221 (26.9%)
Intra-examiner B.N. 6,118 235 175 (74.4%) 55 (23.4%) 69 (29.4%) 25 (10.6%) 86 (36.6%)
Inter-examiner
V.M./B.N. 16,272 783 653 (83.4%) 194 (24.8%) 289 (36.9%) 63 (8.0%) 237 (30.3%)
specific categories of diagnoses. A closer look at the data ly, these findings contradict the general opinion [WHO,
sets from the present study revealed that most of the mis- 1997] that the reproducibility of diagnostic systems includ-
classifications (approximately 80%) involved disagreement ing non-cavitated caries lesions is poor in clinical settings.
between sound surfaces and non-cavitated caries lesions Disagreements between sound surfaces and non-cavitat-
(active or inactive). Therefore, a detailed analysis of the dis- ed active lesions and sound surfaces and non-cavitated
tribution of misclassifications within this particular catego- inactive lesions occurred with about the same frequency
ry was performed (table 5). Although most of these misclas- (about 30%). Only about 10% involved disagreement
sifications concerned the distinction between sound and between non-cavitated active and non-cavitated inactive
non-cavitated active lesions or sound and non-cavitated lesions. This pattern of distribution shows that the new
inactive lesions, the probability of confirming the diagnosis diagnostic criteria are effective in separating active lesions
‘sound’, ‘non-cavitated active’, or ‘non-cavitated inactive’ from inactive lesions. However, non-cavitated caries le-
was 98, 69 and 73%, respectively. These figures are in the sions, being active or inactive, are difficult to differentiate
same order of magnitude as those reported by Ismail et al. from sound surfaces. Ismail et al. [1992] have previously
193.219.162.172 - 11/13/2013 9:32:03 AM
[1992] for incipient lesions in pits and fissures. Collective- drawn attention to this diagnostic problem after observing
Eur.Org.for Caries Research
References
Altman DG: Practical Statistics for Medical Re- Glass RL, Peterson JK, Bixler D: The effects of Manji F, Fejerskov O, Baelum V: Pattern of dental
search. London, Chapman & Hall, 1991, pp changing caries prevalence and diagnostic cri- caries in an adult rural population. Caries Res
396–409. teria on clinical caries trials. Caries Res 1983; 1989;23:55–62.
Backer-Dirks O: Posteruptive changes in dental 17:145–151. Marthaler TM: Caries status in Europe and predic-
enamel. J Dent Res 1966;45:503–511. Holmen L, Thylstrup A, Årtun J: Surface changes tions of future trends. Caries Res 1990;24:
Beighton D, Lynch E, Heath MR: A microbiologi- during the arrest of active enamel caries le- 381–396.
cal study of primary root-caries lesions with sions in vivo. A scanning electron microscope Marthaler TM, O’Mullane DM, Vrbic V: The
different treatment needs. J Dent Res 1993;72: study. Acta Odontol Scand 1987;45:383–390. prevalence of dental caries in Europe 1990–
623–629. Ismail AI, Brodeur J-M, Gagnon P, Payette M, Pi- 1995. Caries Res 1996;30:237–255.
Bjarnason S, Finnbogason SY, Holbrook P, Köhler card D, Hamalian T, Oliver M, Eastwood BJ: Neilson A, Pitts NB: The clinical behavior of free
B: Caries experience in Icelandic 12-year-old Prevalence of non-cavitated and cavitated car- smooth surface carious lesions monitored over
urban children between 1984 and 1991. Com- ious lesions in a random sample of 7–9-year- 2 years in a group of Scottish children. Br Dent
munity Dent Oral Epidemiol 1993;21:194– old schoolchildren in Montreal, Quebec. Com- J 1991;171:313–318.
197. munity Dent Oral Epidemiol 1992;20:250– Nyvad B, Fejerskov O: Active root surface caries
Brunelle JA, Carlos JP: Recent trends in dental 255. converted into inactive caries as a response to
caries in US children and the effect of water Kuzmina IN, Kuzmina E, Ekstrand KR: Dental oral hygiene. Scand J Dent Res 1986;94:281–
fluoridation. J Dent Res 1990;69(special is- caries among children from Solntsevsky – a 284.
sue):723–727. district in Moscow, 1993. Community Dent Nyvad B, Fejerskov O: Assessing the stage of
Fejerskov O: Concepts of dental caries and their Oral Epidemiol 1995;23:266–270. caries lesion activity on the basis of clinical
consequences for understanding the disease. Landis JR, Koch GG: The measurement of observ- and microbiological examination. Community
Community Dent Oral Epidemiol 1997;25: er agreement for categorical data. Biometrics Dent Oral Epidemiol 1997;25:69–75.
5–12. 1977;33:159–174. Pitts NB, Fyffe HE: The effect of varying diagnos-
Fejerskov O, Clarkson BH: Dynamics of caries le- Larsen MJ, Bruun C: Caries chemistry and fluo- tic thresholds upon clinical caries data for a
sion formation; in Fejerskov O, Ekstrand J, ride – mechanisms of action; in Thylstrup A, low prevalence group. J Dent Res 1988;67:
Burt BA (eds): Fluoride in Dentistry, ed 2. Fejerskov O (eds): Textbook of Clinical Cari- 592–596.
Copenhagen, Munksgaard, 1996, pp 187–213. ology, ed 2. Copenhagen, Munksgaard, 1994, Sköld UM, Klock B, Rasmusson C-G, Torstensson
Fejerskov O, Manji F: Risk assessment in dental pp 231–257. T: Is caries prevalence underestimated in to-
caries; in Bader JD (ed): Risk Assessment in Machiulskiene V, Nyvad B, Baelum V: Prevalence day’s caries examination? Swed Dent J 1995;
Dentistry. Chapel Hill, University of North and severity of dental caries in 12-year-old 19:213–217.
Carolina Dental Ecology, 1990, pp 215–217. children in Kaunas, Lithuania 1995. Caries WHO: Oral Health Surveys: Basic Methods, ed 4.
Glass RL: The First International Conference on Res 1998;32:175–180. Geneva, World Health Organization, 1997.
the Declining Prevalence of Dental Caries. J
Dent Res 1982;61(special issue):1301–1388. 193.219.162.172 - 11/13/2013 9:32:03 AM
Eur.Org.for Caries Research