Drug and Alcohol Dependence 57 (1999) 7 22 www.elsevier.

com/locate/drugalcdep

A postmarketing surveillance program to monitor Ultram (tramadol hydrochloride) abuse in the United States
Theodore J. Cicero a,*, Edgar H. Adams b, Anne Geller c, James A. Inciardi d, Alvaro Munoz e, Sidney H. Schnoll f, Edward C. Senay g, George E. Woody h
a

Washington Uni6ersity School of Medicine, 660 South Euclid A6enue, Campus Box 8027, St. Louis, MO 63110 -1093, USA b Gordon S. Black Corporation, Rochester, NY, USA c St. Lukes/Roose6elt Hospital Center and Columbia Uni6ersity College of Physicians & Surgeons, New York, NY, USA d Uni6ersity of Delaware, Newark, DE, USA e The Johns Hopkins Uni6ersity School of Public Health, Baltimore, MD, USA f Medical College of Virginia, Virginia Commonwealth Uni6ersity, Richmond, VA, USA g Pritzker School of Medicine, Uni6ersity of Chicago, Chicago, IL, USA h Uni6ersity of Pennsyl6ania, Philadelphia, PA, USA Received 24 September 1998; received in revised form 24 March 1999; accepted 15 April 1999

Abstract Tramadol HCl, marketed as Ultram in the USA, was introduced as a non-scheduled drug in April 1995 based on the assumption that the risk of abuse was sufciently low to warrant a non-scheduled status. However, approval was contingent upon the development of an innovative proactive surveillance program, to be overseen by an independent steering committee, which would detect unexpectedly high levels of abuse. The postmarketing surveillance program consisted of systematic collection and scientic evaluation of reports of suspected abuse in high-risk populations surveyed through an extensive key informant network of drug abuse specialists and all spontaneous reports of abuse received through the FDA MedWatch system. Methods to estimate the number of patients prescribed tramadol were also developed. Monthly rates of abuse were calculated as an index of the riskbenet ratio (i.e., abuse cases per 100 000 patients prescribed the drug). The data for the 3 years since the drug was introduced show that the reported rate of abuse has been low. Although a period of experimentation seemed to occur in the rst 18 months after its introductionwhich reached a peak rate of approximately two cases per 100 000 patients exposed during the 2 year period prior to June 1998, the reported rate of abuse has signicantly (P = 0.011) declined, reaching levels of less than one case per 100 000 patients in the last 18 months. The overwhelming majority of abuse cases (97%) have been found to occur among individuals with a history of substance abuse and the abuse has been conned to isolated pockets around the countrynotably none of which have signicant populations of street drug abusers. Thus, the data support the decision not to schedule tramadol and, furthermore, suggest that a proactive post-marketing surveillance program can be successfully developed to effectively monitor abuse of new medications. 1999 Elsevier Science Ireland Ltd. All rights reserved.
Keywords: Tramadol HCl; Drug abuse; Drug dependence; Ultram; Postmarketing surveillance; Controlled Substances Act

1. Introduction Tramadol HCl is a centrally active analgesic with weak m-opioid receptor agonist afnity (Raffa et al., 1992). It has been postulated that its analgesic activity may be mediated by both opioid and non-opioid (i.e.
* Corresponding author. Tel.: +1-314-362-7010; fax: + 1-314-3624856. E-mail address: cicerot@msnotes.wustl.edu (T.J. Cicero)

norepinephrine and serotonin reuptake inhibition) mechanisms (Raffa et al., 1992; Frink et al., 1996; Lai et al., 1996), but the relative contribution of these independent actions to its analgesic prole has not been fully characterized. Tramadol, which is marketed as Ultram by Ortho-McNeil Pharmaceutical (OMP) in the USA, was approved by the US Food and Drug Administration (FDA) in April 1995 as a non-scheduled drug based on the recommendation of its Drug Abuse Advisory Committee (DAAC) made in August of 1994. The

0376-8716/99/$ - see front matter 1999 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 3 7 6 - 8 7 1 6 ( 9 9 ) 0 0 0 4 1 - 1

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

DAAC concurred that tramadol was likely to have a low rate of abuse, despite its afnity for m-opioid receptors, based on a review of the preclinical (Villarreal and Seevers, 1968; Frederichs et al., 1978; Murano et al., 1978; Yanagata, 1978), clinical (Arend et al., 1978; Richter et al., 1980, 1985; Preston et al., 1991), and epidemiological data gathered for nearly two decades by Grunenthal, the original developer of tramadol (Keup, 1993). Two major epidemiological data sources existed: spontaneous reports of drug abuse problems reported to Grunenthal and structured interviews in the Substance Abuse Warning System (SAWS), a German governmentsponsored program whose objective was the systematic monitoring of abuse, including that of prescription medications (Keup, 1993). The spontaneous reports involved potential abuse by 200 300 individuals from over 20 million patients in more than 70 countries who had been prescribed tramadol, yielding a reporting rate of just 1.01.5 cases per 100 000 patients exposed. The SAWS data were drawn from interviews with 8145 drug-experienced individuals. Only 122 persons reported abuse of tramadol, representing an incidence rate of approximately 1.5%, which was far less than that observed with any other prescription drug with signicant abuse potential (Keup, 1993). Based upon this evidence of very low actual abuse and the recognition that the reluctance of physicians to prescribe scheduled analgesics (Sigler et al., 1984; Weissman et al., 1991; Joranson et al., 1992; Joranson and Gilson, 1994) contributed to the acknowledged undertreatment of pain in the USA (Portenoy, 1992; Cleeland et al., 1994; Breitbart et al., 1996), the DAAC decided that the interests of public health might best be served if tramadol were marketed as a non-scheduled drug. The committee, however, expressed concern that the experience in the USA might be different from that in other countries since tramadol would be the rst weak m-opioid agonist, with a recognized abuse potential, to be widely available as an unscheduled medication. The DAAC was particularly concerned that a signicant problem with abuse might occur and become widespread long before it was recognized by the surveillance systems currently in place to detect and monitor abuse in this country such as the Drug Abuse Warning Network (DAWN) and the FDA Spontaneous Reporting System (MedWatch) (Kessler, 1993). Both of these systems have well-recognized limitations (Faich, 1986; Piazza-Hepp and Kennedy, 1995); they are passive, retrospective, often anecdotal and, most importantly, the published data are seriously outdated by over a year and a half publication delay. To address the DAAC concerns, Ortho-McNeil pledged that: tramadol would be marketed responsibly stressing that it had the potential for abuse; a comprehensive post-marketing surveillance program would be developed and overseen by an Independent Steering

Committee (ISC) that would recommend scheduling if signicant abuse were detected; and that intervention strategies would be developed in an effort to minimize any potential abuse. On the basis of these important commitments, the DAAC was persuaded that a safety net was in place to permit a non-scheduling recommendation. The ISC developed a two-tiered approach to detect whether tramadol had unexpectedly high abuse potential. The rst tier consisted of a comprehensive, proactive surveillance program devised to detect any signal that abuse of tramadol might be emerging, particularly in populations at risk for abuse. The second tier consisted of two analytical post-marketing studies focusing on the potential abuse of tramadol in pain patients and impaired health care professionals. In this paper, only the surveillance program will be discussed; the results of the two ongoing Phase IV studies will be published elsewhere. The ISC recognized that no systematic monitoring for abuse liability had ever been conducted for any newly approved psychoactive medication and, thus, the surveillance program for tramadol was unprecedented (Brewer and Colditz, 1999; Temple, 1999). The committee concluded that it needed to incorporate the following critical elements: rst, the program needed to be proactive and timely in soliciting evidence of tramadol abuse; second, it had to be sensitive enough to detect isolated, regional clusters of abuse; third, methods needed to be developed to characterize the abuse permitting the development of intervention strategies; fourth, in contrast to the less rigorous DAWN and MedWatch programs, suspected cases of abuse should be validated utilizing scientically appropriate diagnostic criteria; and, nally, the program had to incorporate methods to estimate the number of patients exposed to tramadol and, hence, the rate of abuse to enable a riskbenet analysis (i.e. cases of abuse per 100 000 patients exposed). While any adverse event, such as abuse, is unfortunate, this risk must be balanced against the potential benets of the drug; the rate of abuse provides such a measure. The program developed to meet these goals and the abuse prole of tramadol in the 3 years since its introduction5 April 1995 through 30 June 1998are described in this paper. It should be noted that, although the past tense is used throughout this report, the surveillance program was still in place at the time of the submission of this paper.

2. Methods

2.1. Composition and general operating principles of the ISC

The ISC was appointed within 30 days of the DAAC meeting in August 1994; it consisted of the eight au-

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

thors of this paper who have a broad array of experience in the substance abuse eld: pre-clinical and clinical abuse liability assessment and treatment; epidemiology; statistics; psychiatry; addiction medicine; sociology and the criminal justice system. This diversity was required to develop this unprecedented surveillance program and to assist in the interpretation of what was expected to be a large and complex epidemiological data base. The rst author (TJC) served as the chairman of the ISC. The ISC met monthly for at least 1 day from late 1994 through 30 June 1998 initially to develop the surveillance program and subsequently to review the data, characterize the nature of any abuse observed, develop intervention strategies, prepare periodic reports to OMP and the FDA, and to determine whether the results continued to warrant a non-scheduled status for tramadol. In addition to attending these regularly scheduled meetings, members of the ISC carried out a number of site visits to obtain rst-hand information in areas where abuse seemed to be disproportionately represented or to provide educational programs for physicians and OMPs sales force. Ortho-McNeil Pharmaceutical provided the bulk of all funding for committee-related activities. To ensure the committees independence, the monthly meetings were closed sessions with only the eight members present to discuss the data and vote if required; reports were issued at periodic intervals to OMP and the FDA. Representatives from OMP were occasionally invited if their participation was required.

2.2.2. Physician awareness Ortho-McNeil also assessed the performance of the sales force on a regular basis by interviewing physicians on the thoroughness of the brieng they received from the sales representatives. If needed, individual representatives were retrained when it was documented that their physicians had not heard the complete or appropriate message. In addition, OMP enhanced its physician education programs in those areas targeted by the ISC as potentially troublesome and, in March 1996, sent a letter to over 900 000 health care professionals nationwide stressing that tramadol had the potential for abuse, particularly in those groups identied by the ISC as being especially vulnerable. 2.3. Spontaneous reports of abuse
Reports of suspected abuse of tramadol are received by OMP from patients, pharmacists, physicians, and the FDA via the MedWatch system (Kessler, 1993). All the adverse events suggestive of abuse were sent to the Chairman of the ISC for scientic evaluation as described below. The ISC provided special training to the employees of OMP who had direct contact with those reporting adverse events to ensure that as much information as possible was gathered at the time of contact. Separate questionnaires were developed by the ISC for patients, pharmacists, and physicians.

2.2. O6er6iew of tramadols promotional and marketing acti6ities

The ISC examined all of the materials used in the promotion of tramadol and in the training of the OMP sales representatives at the launch of the drug in 1995 and then all updated materials in March 1997 to ensure that the commitment to market tramadol responsibly was honored. In addition, the ISC also assisted OMP in the training of the sales force and physician education programs as described in the next two subsections.

2.4. Key informant network

The ISC also established an extensive network of drug abuse specialists to proactively seek out documented cases of abuse since it was recognized that the MedWatch system had a number of limitations as a proactive surveillance technique (Faich, 1986; PiazzaHepp and Kennedy, 1995). First, the program signicantly underestimates abuse since it relies upon spontaneous reports; second, the reports generally lack sufcient information for valid clinical assessments; and, nally, the program primarily records events occurring in patients in whom abuse is expected to be very low and, hence, probably understates the nature and scope of abuse. As a result of these limitations, the ISC developed a large national base of key informants to proactively seek evidence of abuse in individuals at risk for abuse. The term key informants, which is borrowed from the elds of social and cultural anthropology (Agar, 1980), refers to clinicians, epidemiologists, treatment counselors, and other observers who are well-recognized experts in the eld of substance abuse and who are in a position to know about new and emerging drug problems in their areas. This network was composed of 110

2.2.1. Sales force training The Chairman of the ISC addressed the full sales force of 1500 on two occasions to reinforce the message that tramadol had the potential for abuse in vulnerable populations, and to stress that physicians should be warned to use it carefully in all patients, particularly those with a known history of drug abuse. In addition to these national meetings, the Chairman also met with the regional sales force in all of the areas identied as potential sites for disproportionately high tramadol abuse to ensure that OMP representatives were not contributing to any misconception by physicians that tramadol had no abuse potential.

10

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

NIDA grantees conducting comprehensive epidemiological and treatment outcome studies of drug-abusing populations, along with 145 other drug abuse experts identied by the ISC (e.g. clinicians, treatment counselors, methadone clinic directors). The geographical distribution of the ofces of these key informants is shown in Fig. 1. Collectively, the network provided access to approximately 250 000 at risk individuals. A questionnaire, jointly developed by the ISC and NIDA, was sent quarterly to the key informants, who also agreed to contact the Chairman immediately if tramadol use emerged in their populations between quarterly surveys. In all cases when an informant indicated that a case of abuse had been detected, the Chairman obtained more information by direct contact. The Institutional Review Board (IRB) at Washington University approved the questionnaire used by the ISC; Washington University was used as the primary site for the research since the senior author handled all aspects of mailing the questionnaires, securing additional information as needed and distilling the data. The informants were offered a payment of $75 for each completed questionnaire and all followup information requested, but more than 25% refused the reimbursement offer. In the 12 quarters covered by this report, the average response rate was 48.7% (range, 45.8 54.2%). There were no trends in response rates over time. This relatively high response rate for a mailed survey may be due in part to the token payment, but the scientic interest of the participants in the program seemed to be the

stronger motivating factor and the ease with which responses to the questionnaire could be made (e-mail, fax or regular mail) probably also facilitated the high response rates.

2.5. Internet sur6eillance

The Internet has become an important medium for information exchange among drug users of various types, especially particular bulletin boards and interactive discussion groups that focus on psychoactive drugs. The ISC established a system for periodic searches of the Internet for any mention of tramadols abuse properties or use for mood-altering effects. The approach utilized keyword and text searches, lurking (i.e. monitoring discussion groups for mentions of tramadol without contributing questions), and periodic, anonymous queries regarding tramadols euphorigenic potential. In a number of cases the Chairman, using an anonymous name, was able to obtain sufcient information from the reporter to permit the generation of a case report.

2.6. E6aluation of reports

All reports, obtained from OMP and elicited by the committee, were evaluated and classied by a sub-committee of the ISC according to DSM-IV (Diagnosis and Statistical Manuel, 4th Edition) criteria for substance abuse and dependence (American Psychiatric Association, 1994). For those reports received by the ISC via its key informant network or from Internet sources, the

Fig. 1. Geographical distribution of key informants by ve digit zip code. These key informants consisted of 110 NIDA grantees and 145 other drug abuse experts.

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722 Table 1 Independent Steering Committee (ISC) ratings of 932 tramadol reports through 30 June 1998 ISC Rating Positive ISC subcategory 10 11 12 13 14 Possible 20 Description Number Percent of total 11.7 3.0 4.5 3.6 4.9 2.6 30.4 26.8 6.0 32.8 18.3 4.0 11.2 3.0 0.3 18.5 100.0

11

Satises DSM-IV criteria for drug dependence 109 Satises DSM-IV criteria for drug abuse 28 Physician reports drug abuse or dependence, not clear that DSM-IV criteria 42 were met Tramadol intentionally used to produce euphoria, euphoria occurred 34 Tramadol intentionally used to produce euphoria in combination with other 46 drugs, euphoria occurred DSM-IV criteria for drug dependence partially met, but no denitive diagnosis Subtotal 24 283 250 56 306

Withdrawal

21 22

Typical opioid-like withdrawal upon discontinuation of tramadol (no indication of abuse) Atypical opioid-like withdrawal upon discontinuation of tramadol (no indication of abuse) Subtotal

Alleged Negative

30 31 40 41 42

Suspected abuse/dependence, insufcient information to draw denitive con- 171 clusion Euphoria/high occurred after therapeutic use of tramadol, no other signs of abuse/dependence No evidence of abuse, dependence, or euphoria Drug experienced person tried tramadol to get high, reports lack of success Drug experienced person tried tramadol with other drugs to get high, reports lack of success Subtotal Total 37 104 28 3 172 932

Chairman generated a case report, based on the facts presented in the surveys and all follow-up information obtained, which was transmitted to the sub-committee for evaluation. The chairman also transmitted these reports to OMP so that they could be sent to the FDA via the MedWatch system. The reports were classied by the sub-committee as either positive, possible, alleged or negative for abuse or as withdrawal-typical or atypical (Table 1). In the latter category, those cases rated as withdrawal were exclusively those in which no other signs or symptoms of dependence or abuse occurred; if either of the latter were found, the case was rated as positive. These ve categories were further subdivided into 13 subcategories to provide as much descriptive information as possible (see Table 1). In the presentation of results in this paper, reports were grouped into four categories: 1. Positive for abuse: reports rated by the ISC as positive and possible (subcategories 10, 11, 12, 13, 14 and 20); the ISC determined that the category of positive for abuse should include both positive and possible, to be as encompassing as possible in estimating abuse rates.

2. Withdrawal alone: reports rated as sub-category 21 or 22. 3. Alleged abuse: reports rated as sub-category 30. 4. Negative for abuse: reports rated as sub-categories 31, 40, 41 or 42. When the sub-committee had rated the reports, code sheets were completed by the Chairman containing information related to a number of important domains (e.g., source of report, abuse rating, zip code, and history of alcohol or drug use); the forms were then sent to the Biostatistics/Epidemiology Center, Johns Hopkins University (AM) where the main data base was created for subsequent analyses as described below.

2.7. Estimation of patient exposure and rates of abuse

A key component of the surveillance program developed by the ISC was the method for deriving an estimate of the number of individuals who were exposed to tramadol each calendar month so that rates of abuse could be calculated. Starting from sales and inventory information on the number of tablets sold and in patients hands, the Biostatistics/Epidemiology Center at Johns Hopkins University developed an algorithm which led to the estimate of patients exposed

12

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

to tramadol in a given month. This estimate is preferable to the usual estimate of total therapies/prescriptions taken by new patients which is widely used for marketing surveillance. Hereafter we provide a general description of the algorithm used to calculate the number of individuals exposed to tramadol in a given month. To facilitate its comprehension, we have included as an example the calculation for the last month of this report: June 1998.

The rst input of the algorithm was the estimate provided by OMP sales data of the tablets sold and in patients hands in a given month (median over the 39 months from April 1995 to June 1998: 46 043 000 tablets). The second input was provided by the National Prescription Audit (IMS American, Ltd.) regarding the average number of tablets per prescription (median over the study period=47.52 tablets per prescription). The ratio of these two inputs provides the number of prescriptions sold in a given month. The third input was provided by the National Disease and Therapeutic Index database (IMS American, Ltd.) regarding the proportion of the prescriptions given to new (median over the study period =64%) and to continuing patients. Multiplying the number of prescriptions sold in a given month by this third input, we derived the number of prescriptions sold to new and to continuing patients in a given month. The fourth input was provided by the National Disease and Therapeutic Index database (IMS American, Ltd.) regarding the proportion of total prescriptions that were sold to new patients (median over the study period = 86%). Since it was expected that continuing patients received fewer free samples than new patients, the estimate for continuing patients was 75% shrank towards one from the

estimate used for new patients (i.e. 0.75 1+0.25 proportion for new patients). The ratio of the prescriptions sold to new patients and the proportions of total prescriptions sold to continuing patients gave the total number of prescriptions used by patients in a given month. Once the total number of prescriptions used by new and continuing patients were available, the nal step in the algorithm was to estimate the average number of prescriptions that new and continuing patients received in a given month. To estimate this, we made the following assumptions: (a) since a new patient in a given month could commence at any point during the month, we halved the total number of prescriptions a person could possibly receive in a given month; and (b) based on clinical experience, we used a rate of 85% for new patients and 70% for continuing patients who used all of the medication as prescribed. The ratio of the total number of prescriptions used by patients and the average number of prescriptions taken by patients yielded the estimates of the total number of new and continuing patients exposed to tramadol in a given month. Their sum provided the denominator for the calculation of rates. Once the denominator was available, the rate per month of abuse by for each of the categories shown in Table 1 was determined by dividing the actual number of cases received by the number of patients exposed. In order to provide an average rate per month for each calendar quarter, the three monthly rates for a given quarter were averaged using methods based on the Poisson distribution describing the number of cases observed in a given month. Specically, if r1, r2, and r3 are the cases observed in the 3 months of a given quarter, and n1, n2 and n3 are the total number of individual exposed to tramadol, respectively, the average rate per month is given by [((r1n1)1/2 + (r2n2)1/2 + (r3n3)1/2(n1 + n2 + n3) 1)]2. Under this averaging procedure, we calculated the corresponding 95% condence intervals for the rate per month for each quarter. Furthermore, in order to formally test for trends over time, we used Poisson regression methods (Kelsey et al., 1996) to determine whether a model with a linear trend in the log scale of the rates was signicant and whether it provided an appropriate goodness of t for the data in the last 2 years (i.e. 3rd quarter of 1996 to 2nd quarter of 1998).

2.8. Cross-tabulation of sales and case reports

Methods were also developed to detect areas in which the use of tramadol was disproportionately high compared to other prescribed analgesics and to identify zip codes with disproportionately high increases in sales from one quarter to the next. The main objective of these analyses was to determine whether areas with

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

13

high sales of tramadol, particularly relative to other analgesics, were associated with large numbers of reports of abuse and dependence.

2.9. Geographical analysis

Rates of abuse were tracked geographically by zip code in order to identify areas with disproportionately high rates of abuse or unusually sustained abuse. Maps of the USA were created each quarter with each diagnosed case of abuse depicted using the software MapLinx (MapLinx, Brunswick, ME). Geographical analysis of prescribing and prescription-lling practices, sales, and distribution of tramadol were also conducted and tracked on maps to detect any inexplicably large use of tramadol and whether there was a correlation between sales and abuse patterns. These data were provided by OMP from internal and external databases, some of which are proprietary in nature and therefore cannot be described in any detail here.

substance abuse had been prescribed tramadol by physicians who believed it to be a safe, non-addictive analgesic. In these instances, intervention strategies were implemented in an effort to reduce this inappropriate use, including sales force retraining and the enhanced physician awareness programs described earlier (Section 2.2) to ensure that the appropriate message was being delivered by the sales force and clearly understood by physicians.

2.11. Phase IV studies

Two phase IV studies, which form the second part of the two-tiered approach, were developed by the ISC to assess the abuse potential of tramadol. The results will be reported elsewhere, but briey they focus on two important populations: (1) chronic pain patients, the intended users of tramadol; and (2) impaired health care professionals, a population considered to be most likely to experiment with and to abuse tramadol at a very early stage. The study focusing on chronic pain patients will compare the rate of abuse of tramadol with that of nonsteroidal anti-inammatory drugs (NSAIDs) , known to be very low, and with that of medications containing hydrocodone, a drug known to have a relatively high rate of abuse. Impaired health care professionals were selected as the population for the second phase IV study because they are at a high risk and have very easy access to drugs. It was postulated that this population would potentially misuse a drug like tramadol at a very early stage after its introduction. Health care professionals have relatively easy access to the drug and they also have access to detailed information about its weak opiate prole, its nonscheduled status, and the fact (contained in the original label statement) that a routine urine screen was unavailable. It may also be noted that health care professionals were one of the earliest populations detected abusing pentazocine and fentanyl citrate. Programs to monitor this population already exist and, in addition, well-organized monitoring programs for impaired health care professionals with standardized monitoring systems for detecting abuse agreed to participate in the study.

2.10. Inter6ention strategies

One of the important commitments made by OMP to the DAAC was that the ISC would develop mechanisms to intervene when abuse was detected with the goal of potentially minimizing its occurrence. At its monthly meetings, the ISC reviewed and characterized all cases of abuse and determined whether there were cities in which a disproportionately high number of abuse cases had been detected. Specic strategies that were employed to characterize the abuse in cities with disproportionately high numbers of cases included: rst, the ISC contacted all key informants, and any additional contacts in the region, to gather more information about the potential abuse and any proactive steps they believed could be taken to reduce its incidence; second, NIDA grantees, if available, were requested to expand their studies to include a more focused analysis of whether tramadol was being abused in their study populations, the reasons for its use and suggestions for intervention; and, nally, site visits a total of eightwere carried out if the ISC felt it was possible to obtain rst hand information, particularly if individuals found to have abused tramadol were available to be interviewed. If individuals were available for interviews, they were assured of condentiality. All written material excluded their names and other condential information. Based upon these data, the ISC found in most cases that the apparent abuse was transient, often unsuccessful experimentation; the best intervention in these cases was judged to be enhanced monitoring by informants to determine if the abuse dissipated or had become a more substantial problem requiring further attention. In some instances, however, it was determined that individuals with a history of

2.12. Data presentation and terminology

As evident in the rating categories, the ISC clearly recognized the distinction between the terms abuse and dependence (Table 1). However, the term drug abuse will be used throughout the text to characterize abuse and/or dependence of tramadol as dened by DSM-IV criteria, unless there is a reason to distinguish the two. The ISC has pooled the reports from OMP (spontaneous reports) and those generated by the committee in all Tables and Figures in this report. Since spontaneous reports underestimate the magnitude of abuse and

14

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

mainly reect rare cases of abuse in patients, the ISC concluded that by combining spontaneous reports with the actively elicited reports from at risk populations, a closer approximation of the actual abuse rate of tramadol could be made. The ISC was fully aware, however, that the numerator would then consist of spontaneous reports and actively elicited cases detected in several hundred thousand at-risk individuals, whereas the denominator would consist of only the number of patients exposed to tramadol. Thus, the resulting rates would be higher using this method than if spontaneous reports alone were used. In an ideal situation, it would have been desirable to obtain two rates: spontaneous reports per 100 000 patients exposed and actively elicited cases per 100 000 at risk individuals screened. Since only patient exposures were known with some condence, we opted to use the most conservative estimate of the rate of abuse: combining spontaneous and actively-elicited responses as the numerator and using as the single denominator the number of patients exposed. It was also recognized that comparisons to other drugs, in which only spontaneous reports served as the source for the MedWatch report, could be inappropriate and misleading if it was not clearly recognized that the ISC utilized both spontaneous and actively solicited cases of abuse. In an effort to avoid this confusion the Committee generated two curves in all graphs: one for all abuse reports (committee-generated and spontaneous reports from OMP) and a second restricting the reports to those received as spontaneous reports. The difference between the two curves provides a direct comparison of the contribution of the proactive surveillance program to the total incidence of abuse and provides a valid comparison between spontaneous reports of abuse of other drugs and tramadol.

The most frequent report was that of typical and atypical withdrawal upon discontinuation of tramadol (32.8% of the total) with no indication of abuse, followed by positive and possible for abuse (30.4%), negative (18.5%) and alleged (18.3%). Using DSMIV criteria, only 11.7% and 3.0% of all cases were rated as positive for drug dependence and abuse, respectively.

3.2. Source of reports

The distribution of all reports by the source OMP or the ISCis shown in Table 2. About half of the total cases judged to be positive for abuse came from each source for the 3 years covered by this report. As shown in Table 2, the ISC informants were less likely to generate alleged or negative reports. The negative events reported by the informants generally reected the failure of the drug to produce euphoria in drug experienced individuals. Of particular note in Table 2 is that the informant network generated only eight withdrawal cases, since these drug abuse experts recognized that withdrawal alone was not a predictor of abuse.

3.3. Estimated use of tramadol

The monthly estimates of the total number of new and continuing patients exposed to tramadol from the date of launch, 5 April 1995 until 30 June 1998, are shown in Fig. 2. Monthly patient exposures increased in a linear fashion for the rst year reaching approximately 700 000 patients per month. The rate of increase decreased substantially in the subsequent 2 years, but nevertheless has continued to rise to nearly 900 000 patients per month in the second quarter of 1998.

3. Results

3.4. Rate of abuse 3.1. E6aluation of all reports suggesti6e of abuse

Table 1 shows the distribution of all reports rated as positive, alleged or negative for abuse or as withdrawal.
Table 2 Source of 923 reports through 30 June 1998 Source of report Abuse/dependence N Ortho-McNeil ISC sourcesb Total 137 146 283 % 48.4 51.6 100 Withdrawal N 298 8 306 % 97.4 2.6 100 Alleged N 134 37 171 % 78.4 21.6 100 Negativea N 119 53 172 % 69.2 30.8 100

Fig. 3 presents a plot of the average rates per month of abuse (rated as positive or possible, see Table 1), for the total cases for each quarter during the period

a Most of the negatives arising from the ISC sources represent unsuccessful experimentation with Ultram, alone or in combination, for its euphorigenic properties. b Includes NIDA grantees, key informants and Internet sources.

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

15

Fig. 2. Estimated number of patients prescribed tramadol per month from the date of launch, 5 April 1995 through 30 June 1998.

covering 5 April 1995 through June 30 1998. The absolute number of diagnosed cases of abuse by quarter is also shown. During the rst two quarters after tramadols introduction, there was no evidence of abuse. Thereafter, reports of abuse increased, reaching a peak in the rst three quarters of 1996 of approximately two cases per month per 100 000 patients. Subsequently, despite continuously active surveillance efforts, the rate had decreased signicantly (P = 0.011)as shown by the best-t line in Fig. 3 to approximately 0.5 1.0 case per month for 100 000 patients over the year ending 30 June 1998. It is apparent in this gure that the use of both the key informant network and spontaneous reports generated more than twice the number of abuse cases than the traditional reliance on spontaneous reports alone.

abuse in a strict diagnostic sense, pooling these cases with those positive for abuse would provide the worst-case scenario; these composite data are shown in Fig. 4. It is evident that, in this worst-case scenario, the combined rate of all cases of positive, possible or alleged abuse displayed the same overall pattern as that shown in Fig. 3 for diagnosed cases of abuse. From a peak of 23 cases per 100 000 patients in mid-1996, the rate has signicantly (P= 0.022) declined over the 2 years ending 30 June 1998 to an average of around 1 case per 100 000 patients.

3.7. Estimated rate of withdrawal

Fig. 5 gives the absolute number of reported cases of withdrawal by quarter and presents a plot of the estimated rates per month of withdrawal in each quarter. There is little difference between the total number of reports (spontaneous and committee-generated reports) and the number of spontaneous reports since, as discussed above, the informants would not report simple cases of withdrawal as abuse or dependence unless there were evidence for abuse. It is apparent that withdrawal was reported almost immediately after the launch of tramadol, and reached a peak of approximately 23 cases per month for 100 000 patients in the second quarter of 1996. However, the rate of withdrawal reports decreased gradually and has remained constant at around 1 case per month per 100 000 for the 24 months ending 30 June 1998. The sharp increase in the second quarter of 1996 coincides with the distribution of a letter by Ortho-McNeil to 900 000 health care

3.5. Characterization of abuse

Although data were collected on a number of important domains, such as gender and length of exposure to tramadol, there was insufcient information available in most instances to draw any meaningful conclusions. However, the histories of drug/alcohol abuse were available in 77% of all abuse/dependence cases. In 97.3% of these, there was a history of opiate, alcohol, or other drug abuse.

3.6. Estimated rate of abuse/dependence and alleged abuse

While the ISC believes that many of those cases classied as alleged for abuse should not be considered

16

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

professionals reafrming that tramadol had abuse potential, especially in vulnerable populations. This letter appeared to have sparked a number of reports of withdrawal.

where heroin and other drug use is prevalent: the northeastern corridor of the United States (Washington to Boston), South Florida, Detroit, Chicago, and Los Angeles.

3.8. Geographical analysis

The data shown in Fig. 6 illustrate that the abuse of tramadol was conned to isolated pockets across the country and, most signicantly, was transient in nature: abuse cases appeared rather quickly in clusters in a particular city and dissipated just as rapidly. It is important to note that cases of continuing abuse in a given region were virtually non-existent. Specically, our informants were asked in each questionnaire whether any reports they observed and reported were new or continuing cases. There were no continuing cases reported in the 13 quarters covered by this report, further reinforcing the conclusion that these events reected brief periods of unsuccessful experimentation. One of the more striking features of the patterns of abuse of tramadol depicted in Fig. 6, is that high levels of abuse were not detected in areas

3.9. Internet discussion of tramadol

Internet searches indicated that within 2 months of its release, there was extensive discussion of tramadol on the Internet, primarily by individuals requesting information about its mood-altering effects. The ISC logged in excess of 150 mentions of tramadols euphorigenic properties on the Internet during a period of approximately 6 months. While a very small number of reports (B12) indicated that tramadol could be used to alter mood or enhance the effects of other drugs, over 90% of the mentions indicated that tramadol was devoid of any benecial euphorigenic effects. In the third quarter of 1997, the ISC posted a request for information on tramadols use as a mood altering agent or its potential abuse on 12 Internet bulletin boards for recreational drug users. This posting failed to elicit a single response in 60 days.

Fig. 3. Average rates per month of tramadol abuse or dependence (rated as positive or possible, Table 1) by calendar quarter from 5 April 1995 until 30 June 1998. The total number of reports in each quarter are listed as integers across the top of the graph. The rates, expressed as the average total number of abuse cases per month for 100 000 individuals exposed (derived from Fig. 2, see Section 2), are shown in the line tracings together with the 95% condence limits. The solid line with closed circles represents all reports (spontaneous reports and committee generated), whereas the solid line with closed triangles represents spontaneous reports alone. Thus, the difference between the two reects the contribution of the ISC elicited reports to overall reporting. Also shown is the curve that models the rates (log scale) as a linear function of time in the last eight quarters preceding June 1998. The t was appropriate (P= 0.768) and the downward trend was statistically signicant (P= 0.011).

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

17

Fig. 4. Average rates per month of tramadol abuse or dependence (rated as positive, possible or alleged) by calendar quarter from 5 April 1995 until 30 June 1998. The downward trend over the last eight quarters was statistically signicant (P= 0.022). See the legend to Fig. 3 for additional details and description of the plot.

3.10. Inter6ention strategies

Although widespread abuse of tramadol was not observed (Fig. 6), areas of some concern did emerge in which there appeared to be an over-representation of abuse cases. The following discussion is intended to provide, and exemplify for Atlanta, the general characteristics of the abuse that was observed and the types of intervention strategies employed by the ISC. In general, the ISC determined that tramadol abuse was observed in two populations in disproportionately high numbers in several locations: street drug abusers (e.g. Atlanta, Phoenix, Madison, New Orleans, Baltimore, Indianapolis and Rockford, IL); and individuals with a past history of substance abuse who received tramadol as a non-addictive analgesic without appropriate warnings about its potential abuse liability (e.g. in Atlanta and Phoenix). The general features of these types of abuse and the nature of the interventions employed in Atlanta are provided in the next section; nearly identical abuse characteristics were found, and corresponding intervention strategies were employed, in all the other cities mentioned above.

3.10.1. Tramadol abuse in Atlanta and inter6ention strategies employed In the rst quarter of 1996, the ISC discovered through two informants that six to eight impaired physicians enrolled in treatment programs were abusing

tramadol and had been readmitted for treatment. The ISC sent a team of four members to Atlanta for a 2-day site visit to characterize the nature of the problem; the informants and three of the impaired physicians were interviewed. The ISC found that all three physicians were either prescribed tramadol by a physician who thought it was safe to use in this population (i.e. no addiction potential) or they took samples for pain relief on the assumption that it was safe to take. Only one physician appeared to have abused tramadol exclusively for a long-term period (several months); the others used multiple drugs. Interestingly, all of the physicians viewed tramadol as a very poor alternative to their drugs of choice but used it because it was available and they believed it could not be detected in urine. Since several of the impaired health care professionals indicated that they were prescribed the drug for pain by a physician who was well aware of their drug abuse history, the ISC determined that the message that tramadol had abuse potential in vulnerable populations was not being conveyed clearly to some physicians in Atlanta. Thus, the enhanced physician awareness and sales force retraining programs described in Section 2 were initiated. This intervention may have been successful since there have been no new cases of this type reported in Atlanta in the year ending 30 June 1998. In the third quarter of 1996, the ISC also received a report from one key informant in Atlanta that a number of heroin addicts had used tramadol to take the

18

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

edge off withdrawal and as a substitute for heroin when it was not readily available. The ISC contacted the informant who reported that this occurrence represented some minor experimentation that seemed to have dissipated very rapidly. The committee requested that two NIDA grantees in the Atlanta area expand their general surveillance programs in populations at risk to include questions about tramadol abuse. There have been no further indications that tramadol has been used as a substitute for heroin in Atlanta, suggesting that this event was an example of isolated, unsuccessful experimentation. Finally, in April 1997 the ISC received information from another informant in Atlanta that six to eight heroin users, enrolled in a methadone program, had solubilized cocaine and tramadol for intravenous injections. The ISC immediately contacted the informant with a request for more information and suggested the possibility of a site visit by committee members. The informanta NIDA grantee stated that there was no more information available but that markedly enhanced surveillance efforts would be initiated as part of the study. There have been no additional reports of such use through 30 June 1998.

had abuse potential and would not use its non-scheduled status as a marketing device. The ISC was reassured by an examination of the training and promotional materials and by the committees attendance at several training sessions conducted by OMP that tramadol was being promoted appropriately.

4. Discussion At the 3 August 1994 meeting of the DAAC of the FDA, Ortho-McNeil, with the assistance of a number of members of the ISC, argued that tramadol HCl, should be marketed in the USA as a non-scheduled drug because of its already documented low abuse potential world-wide (Keup, 1993). The decision to recommend not scheduling tramadol was based on a riskbenet analysis in which the risk of a very low level abuse was counter-balanced by the benet of making an apparently quite efcacious analgesic readily available. The DAAC concurred that scheduling under the Controlled Substance Act contributed to the acknowledged undertreatment of pain in this country (Portenoy, 1992; Cleeland et al., 1994; Breitbart et al., 1996) due to the reluctance of physicians to use scheduled drugs (Sigler, et al., 1984; Weissman et al., 1991; Joranson, et al., 1992; Joranson and Gilson, 1994). The reluctance to use scheduled medications despite their efcacy appears to be related to two factors: rst, an unwarranted fear of addiction by physicians (Portenoy,

3.11. Promotion and marketing acti6ities

As discussed in Section 1, the FDA/DAAC were concerned that tramadol be marketed appropriately: specically, that OMP would emphasize that the drug

Fig. 5. Average rates per month of tramadol withdrawal cases by calendar quarter from 5 April 1995 until 30 June 1998. See the legend to Fig. 3 for additional details and description of the plot.

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

19

Fig. 6. Cases of abuse/dependence that meet the DSM-IV criteria plotted by ve digit zip code by quarter from the third quarter of 1995 through the second quarter of 1998. The number of cases depicted in these gures will not necessarily sum to the total shown in Table 1 or Fig. 3 since zip codes were unavailable in approximately 5% of the total cases received. In addition, to illustrate the number of cases in a given city, the software MAPLINX program required scattering the points to avoid overlap; this invariably results in a slight dispersion of the cases around specic geographical loci.

1992; Ward et al., 1993); and, second, the increased regulatory burdens and oversight by state regulatory groups associated with scheduling (Weintraub et al., 1991; Joranson and Gilson, 1994). Thus, the DAAC concluded that the interests of public health might best be served by allowing tramadol to be marketed as a non-scheduled drug. However, the advisory group also recognized that this decision carried some risk that abuse in this country might be greater than that observed worldwide and, thus, recommended that a safety net be established: an extensive post-marketing surveillance program, to be developed and overseen by an independent steering committee, which would provide an early warning of unexpectedly high abuse. The ISC concludes that the unprecedented surveillance program (Brewer and Colditz, 1999) described in this paper has met the objectives of the safety net the DAAC required in making its recommendation: it has been proactive and timely in soliciting and reporting cases of abuse with sufcient sensitivity to stratify abuse by geographic location; and methods were developed to characterize abuse which permitted the implementation of apparently successful intervention strategies to reduce abuse. On the basis of the data gathered through this extensive program, the ISC concludes that the rates of abuse of tramadol have been low and consistent with prior pre-clinical (Villarreal and Seevers, 1968; Frederichs et al., 1978; Murano et al., 1978; Yanagata, 1978), clinical (Arend et al., 1978; Richter et al., 1980, 1985; Preston et al., 1991), and epidemiological experience with this drug world-wide

(Keup, 1993). Moreover, it can be argued that the non-scheduled status of the drug combined with this surveillance program has made possible the appropriate prescribing of an analgesic with low abuse potential, thus beneting public health. Although a period of experimentation and subsequent abuse of tramadol seems to have occurred 618 months after its introduction, with the rate peaking at two to three cases per 100 000 patients exposed, the abuse rate declined signicantly reaching a relatively low level of 0.51.0 cases per 100 000 patients over the 18 months ending 30 June 1998. This decline cannot be attributed to a decrease in reporting, which is a common feature associated with the MedWatch reliance on spontaneous reports (Faich, 1986; Piazza-Hepp and Kennedy, 1995), since the ISC actively solicited cases of abuse through its informant network and the informant response rates remained constant. Thus, it appears that the decline in abuse is real and is probably not an artifact of under-reporting. Given these considerations, the principal conclusion from the surveillance data is that the abuse of tramadol has not become a substantial problem in this country, and that the current rate is very low. An important issue in this study, and others to come, is the denition of what is a low rate of abuse potential? Making direct comparisons of the abuse of newly launched drugs to other scheduled and non-scheduled drugs is difcult at best since all of the necessary information needed is generally not available: number of abuse cases immediately after launch of the drug;

20

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

and, most importantly, an accurate estimation of the number of individuals exposed to the medication. Most importantly, a comparison of the number of tramadol abuse cases to, for example, existing drugs with signicant abuse potential using MedWatch, at this point in time, would be improper. It is well established (Faich, 1986; Piazza-Hepp and Kennedy, 1995) that spontaneous reports markedly decrease over time. Thus, comparisons between a newly released drug, particularly one for which both spontaneous and actively solicited cases of abuse were obtained, and more established drugs of abuse for which spontaneous reports alone were obtained clearly serve as an inappropriate index of relative abuse potential (Faich, 1986; Piazza-Hepp and Kennedy, 1995). However, there are several data sources that can be used to document that the rates of tramadol abuse are, in fact, low, relative to other drugs. Keup (1993), in an extensive analysis of the abuse of tramadol in Europe, concluded that the abuse of tramadol was very low compared to most other drugs of abuse. The current results document that virtually identical rates of tramadol abuse were found in this country. More direct evidence of tramadols low relative abuse potential comes from preliminary results of the two Phase IV studies which are nearly completed (the complete results will be published elsewhere). Specically, in the pain patient study, the abuse of tramadol was found to be much lower than that for hydrocodone and only marginally higher than that for NSAIDS. In addition, the preliminary results from the impaired health care professional study demonstrate a remarkably low rate of abuse of tramadol in this highly vulnerable population compared to other schedule 2, 3 and 4 drugs. Consequently, on the basis of these considerations, we conclude that the abuse of tramadol is indeed very low. Clearly, however, it would be desirable to have relevant data on other drugs as they are marketed (i.e. the numerator and denominator) to permit more direct population comparisons. The ISC acknowledges that the rates of abuse reported in this paper do not provide a completely accurate estimate of the total abuse rate for tramadol since the techniques employed could not survey the entire population and rely to a greater or lesser extent on self-reports. The degree of underestimation of the actual abuse rate is difcult to ascertain, but if it is assumed that spontaneous reports represent about 10% of all cases of adverse events (Faich, 1986; Weintraub et al., 1991; Piazza-Hepp and Kennedy, 1995), then it follows that, using the proactive surveillance techniques employed by the ISC, the under-estimation of abuse is probably considerably smaller. Thus, it is clear that the program developed by the ISC provides a much more accurate assessment of the magnitude of the tramadol abuse problem than the historical reliance on passive

and retrospective procedures, such as DAWN and MedWatch, and, even with this aggressive program, the rate of abuse of tramadol is nevertheless as low as originally predicted. The abuse of tramadol detected by the ISC was found almost exclusively in individuals with a past history of substance abuse/dependence. In cases for which information was available, over 97% of the individuals who abused tramadol had a past history of alcohol/drug abuse or dependence. These observations underscore that tramadol must be used carefully in individuals with a history of substance abuse but, of perhaps more signicance, indicate that the therapeutic use of tramadol has not led to signicant problems of abuse in pain patients without abuse histories. This observation should not be surprising in view of an extensive literature suggesting that pain patients rarely abuse their prescribed analgesic medications (Porter and Jick, 1980; Perry and Heindrich, 1982; Foley, 1985), although fears of drug addiction remain a major concern of physicians that may inappropriately limit the use of analgesics (Portenoy, 1992; Ward et al., 1993). Perhaps, the current set of observations that the abuse of tramadol has been low, despite very signicant patient exposures of over 17 million, will help reinforce that pain can and should be treated adequately without an unnecessary and unwarranted fear of drug abuse. The most striking feature of the data gathered by the ISC was the very restricted pattern of abuse observed with tramadol. Nearly all of the cases could be traced to 810 locations nationwide. The ISC was struck by the almost total absence of abuse/dependence in the countrys largest cities with substantial populations of street addicts, such as New York, Philadelphia, Chicago, Detroit, Miami, and Los Angeles. This lack of reports is not due to a lack of ISC informants in these cities; indeed as shown in Fig. 1, there were very high densities of informants in these areas. Since these areas also had some of the highest sales of tramadol and number of patient exposures, the relative availability of tramadol cannot explain this phenomenon; rather, the ISC postulates, based upon information from the key informant network, that the absence of wide-spread street abuse in these areas is due to the low euphorigenic properties of tramadol in comparison to other drugs that are as readily available. The reasons for the apparent over-representation of abuse cases in such cities as Madison (WI), Atlanta, Phoenix and New Orleans are not clear. However, according to the key informant network and information gathered by the ISC at site visits, this may be due to occasional limitations in the availability of more attractive alternatives and/or experimentation with unsatisfactory results. Nevertheless, even in these regions, it is clear that the experimentation was transient suggesting that tramadol has little sustained appeal for the

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722

21

production of euphoria. Based on the pattern of abuse observed by the ISC (Fig. 6) it can be concluded that tramadols abuse liability is very different than that observed with more potent euphorigenic agents, such as heroin or cocaine. An important commitment to the DAAC was that intervention strategies to reduce abuse would be developed, if at all possible. The ISC found that in several locations a number of individuals with a history of drug abuse had been prescribed tramadol on the mistaken assumption that it had no abuse potential. In these regions the ISC and OMP implemented an apparently successful intervention strategy: enhanced physician awareness and sales force retraining programs. In the cities targeted by the ISC, there have been relatively few additional cases of this type of abuse suggesting that the proactive, educational programs developed by the ISC were effective intervention strategies to help reduce abuse in this population. However, the ISC concludes that much of the abuse observed with tramadol street use by heroin and other drug users was not in reality amenable to intervention since most of this abuse spontaneously dissipated with no active intervention presumably because of unsatisfactory experimentation. The Controlled Substance Act requires that physical dependence be monitored as evidence of abuse and the FDA therefore considers withdrawal to be a dening feature of dependence. Since it is clear that withdrawal, in the absence of any other signs of drug-seeking behavior, does not constitute abuse or dependence and that withdrawal occurs upon abrupt discontinuation after chronic use of many drugs (e.g. steroids (Yesalis et al., 1990), some antipsychotics (Gilbert et al., 1995) and antidepressants (Lejoyeux and Ades, 1997), it is clearly inappropriate to consider withdrawal as evidence of abuse or dependence. Indeed, it must be underscored that, although the literature suggests that neuroadaptation and its expression as a withdrawal syndrome is often associated with opiate abuse (Wikler, 1980), there is now substantial scientic evidence that withdrawal plays a relatively minor role in the initiation or perpetuation of drug abuse and dependence (Foley, 1985; Everitt, 1997; Gardner, 1997; Nutt, 1997; Johnson and Fleming, 1989; Koob and Bloom, 1998). An interesting question is whether the surveillance program described in this paper had any impact on the rates of abuse of tramadol that have been found. Specically, it can be argued that intervention strategies as opposed to any intrinsic properties of the drug led to lower rates of abuse than would have been observed had this program not been in place. Conversely, it is equally plausible to argue that the nationwide surveys may have actually stimulated experimentation that might otherwise never have occurred. It is virtually impossible at this time to ascertain whether the program had any net effect on the abuse rates

reported. Additionally, the ISC, can see no practical or theoretical way in which this question can be answered, but, at the least, future post-marketing surveillance programs should incorporate a consideration of this conundrum. The ISC believes that the proactive surveillance program described in this paper has conrmed not only that unexpectedly high abuse of tramadol did not occur in the 3 years since its introduction in this country, but also that a program of this type can serve as a prototype for future postmarketing assessments of newly launched drugs with predicted low abuse potential. Indeed, as discussed elsewhere (Brewer and Colditz, 1999; Temple, 1999) such a program might be a vastly preferable alternative to a simple scheduling decision. Specically, the ISC postulates that, for drugs with a suspected very low rate of abuse, public health might best be served by a non-scheduled decision coupled with a strong post-marketing surveillance and intervention program. Safe and efcacious drugs would be appropriately available to patients who may otherwise not benet from them because of the unwillingness of their physician to prescribe scheduled drugs, and the risk of abuse can be minimized by monitoring abuse and implementing intervention strategies if necessary.

Acknowledgements Funding support was provided by Ortho-McNeil Pharmaceutical, Raritan, NJ. The authors are grateful to the National Institute on Drug Abuse (NIDA) for their generous help in identifying NIDA grantees and assisting in the development of the questionnaires used.

References
Agar, M.H., 1980. The Professional Stranger: An Informal Introduction to Ethnography. Academic Press, Orlando. American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, 4th edn. American Psychiatric Association, Washington, DC. Arend, L., Arnim, B., Nussen, J., Scheele, J., Flohe, L., 1978. Tramadol und Pentazocin im klinischen Doppelblind-CrossoverVergleich. Arzneim. Forsch. 28, 199 208. Breitbart, W., Rosenfeld, B.D., Passik, S.D., 1996. The undertreatment of pain in ambulatory AIDS patients. Pain 65, 239 249. Brewer, T., Colditz, G.A., 1999. Postmarketing surveillance and adverse drug reactions. J. Am. Med. Assoc. 281, 824 829. Cleeland, C.S., Gonin, R., Hateld, A.., 1994. Pain and its treatment in outpatient with metastatic cancer. New Engl. J. Med. 330, 592 596. Everitt, B., 1997. Craving cocaine cues: cognitive neuroscience meets drug addiction research. Trends Cogn. Neurosci. 1, 198 199. Faich, G.A., 1986. Adverse-drug-reaction monitoring. New Engl. J. Med. 314, 1589 1592. Foley, K., 1985. The treatment of cancer pain. New Eng. J. Med. 313, 84 95.

22

T.J. Cicero et al. / Drug and Alcohol Dependence 57 (1999) 722 Portenoy, R.K., 1992. Inadequate outcome of cancer pain treatment: inuences on patient and clinician behavior. In: Patt, R.B. (Ed.), Problems in Cancer Pain Management. JB Lippincott, New York, pp. 119 128. Porter, J., Jick, H., 1980. Addiction rare in patients treated with narcotics. New Eng. J. Med. 302, 123. Preston, K.L., Jasinski, D.R., Testa, M., 1991. Abuse potential and pharmacological comparison of Tramadol and morphine. Drug Alcohol Depend. 27, 7 18. Raffa, R.B., Friderichs, E., Reiman, W., Shank, R.P., Codd, E.E., Vaught, J.L., 1992. Opioid and nonopioid components independently contribute to the mechanism of action of Tramadol, an atypical opioid analgesic. J. Pharmacol. Exp. Ther. 260, 275 285. Richter, W., Barth, H., Flohe, L., Giertz, H., 1985. Clinical investigation of the development of dependence during oral therapy with Tramadol. Arzneimittelforsch. (Drug Res.) 35, 1742 1744. Richter, W., Flohe, L., Giertz, H., 1980. Clinical evaluation of dependence-liability of Tramadol. Naunyn Schmeidebergs Arch. Pharmacol. 313, 62. Sigler, K.A., Guersney, B.G., Ingrim, N.B., 1984. Effect of a triplicate prescription law on prescribing of Schedule II drugs. Am. J. Hosp. Pharm. 41, 108 111. Temple, R., 1999. Meta-analysis and epidemiologic studies in drug development and postmarketing surveillance. J. Am. Med. Assoc. 281, 841 844. Villarreal, J.E., Seevers, M.H., 1968 Evaluation of new compounds for morphine-like physician dependence in the rhesus monkey. Paper presented at a meeting of the National Academy of Sciences-National Research Council Committee on Problems of Drug Dependence, 30th Meeting, pp. 1 15. Ward, S.E., Goldberg, N., Miller-McCauley, V., 1993. Patient barriers to management of cancer pain. Pain 52, 319 324. Weintraub, M., Singh, S., Byrne, L., Maharaj, K., Guttmacher, L., 1991. Consequences of the 1989 New York State triplicate benzodiazepine prescription regulations. J. Am. Med. Assoc. 266, 2392 2397. Weissman, D.E., Joranson, D.E., Hopwood, M.B., 1991. Wisconsin physicians knowledge and attitudes about opioid analgesic regulations. Wisconsin Med. J., December. Wikler, A., 1980. Opioid Dependence. Plenum, New York. Yanagata, T., 1978. Drug dependence potential of 1-methesyphenyl2-dimethylamimethyl-eyelohean-1-ol hydrochloride Tramadol tested in monkeys. Arzneim. Forsch. (Drug Res.) 28, 158 163. Yesalisy, C.E., Vicary, J.R., Buckley, W.E., Strait, A.L., Katz, D.L., Wright, J.E., 1990. Indications of psychological dependence among anabolic-androgenic abusers. In: Lin, G., Erinoff, L. (Eds.), Anabolic Steroid Abuse. National Institute on Drug Abuse, Rockville MD, pp. 196 214.

Frederichs, E., Felgenhauser, E., Joagschaap, P., Osterloh, G., 1978. Pharmacological investigations on analgesia and the development of dependence and tolerance with Tramadol, a strongly acting analgesia. Arzneim Forsch Drug Res. 28, 122132. Frink, M.Ch., Hennies, H.-H., Englberger, W., Haurand, M., Wilffert, B., 1996. Inuence of Tramadol on neurotransmitter systems of the rat brain. Arzneim.-Forsch. 46 (11), 10291036. Gardner, E., 1997. Brain reward mechanisms. In: Lowinson, J., Luis, P., Millman, R., Langrod, J. (Eds.), Substance Abuse. Williams and Wilkens. Gilbert, P.L., Harris, M.J., McAdams, L.A., Jeste, D.V., 1995. Neuroleptic withdrawal in schizophrenic patients. Arch. Gen. Psychiatry 52, 173 188. Johnson, S.M., Fleming, W.W., 1989. Mechanisms of cellular adaptive sensitivity changes: applications to opioid tolerance and dependence. Pharmacol. Rev. 41, 435488. Joranson, D., Gilson, A., 1994. Controlled substances, medical practice and the law. In: Schwartz, H.I. (Ed.), Psychiatric Practice Under Fire. American Psychiatric Press, Washington, DC, pp. 173 194. Joranson, D.E., Cleeland, C.S., Weissman, D.E., Gilson, A.M., 1992. Opioids for chronic cancer and non-cancer pain: a survey of state medical board members. Fed. Bull. 79, 1549. Kelsey, J.L., Whittmore, A.S, Evans, A.S., Thompson, W.D., 1996. Methods in Observational Epidemiology. Oxford University Press, Oxford, England. Kessler, D.A., 1993. Introducing MEDWatch. A new approach to reporting medication and device adverse effects and product problems. J. Am. Med. Assoc. 269, 27652768. Keup, W., 1993. Missbrauchsmuster bei abhangigkeit von alkohol, medikamenten und drogen: Fruhwarnsystem-daten fur die bun desrepublik Duetschland 19761990. Lambertus, Freiburg im Breisgau. Koob, G.F., Bloom, F.E., 1998. Cellular and molecular mechanisms of drug dependence. Science 242, 715723. Lai, J., Ma, S.-W., Porreca, F., Raffa, R.B., 1996. Tramadol, M1 metabolite and enantiomer afnities for cloned human opioid receptors expressed in transfected HN9.10 neuroblastoma cells. Eur. J. Pharamcol. 316, 369372. Lejoyeux, M., Ades, J., 1997. Antidepressant discontinuation: a review of the literature. J. Clin. Psychatry 58 (Suppl 7), 1115. Murano, T., Yamamoto, E., Endo, N., Okada, N., Massuda, Y, Yano, I., 1978. Studies of dependence on Tramadol in rats. Arzneim Forsch. (Drug Res.) 28, 125158. Nutt, D., 1997. The neurochemistry of addiction. Hum. Psychopharm. 12, S53 S58. Perry, S., Heindrich, G., 1982. Management of pain during debridement: a survey of U.S. burn units. Pain 13, 267280. Piazza-Hepp, T.D., Kennedy, D.L., 1995. Reporting of adverse events to Medwatch. Am. J. Health Syst. Pharm. 2, 14361439.