Abstract

In the present study, mortality rates and prevalence of abstinence from illicit drugs among persons with a history of addiction to heroin, cocaine, and/or amphetamines were estimated along the drug-using career time scale. Follow-up data on drug use and vital status were analyzed for participants in the Amsterdam Cohort Study among Drug Users (n = 899; 1985–2002). Participants in the study were primarily recruited at low-threshold methadone outposts. It was estimated that at least 27% of drug users had died within 20 years after starting regular drug use; for half, death had been due to causes unrelated to human immunodeficiency virus. A favorable trend towards abstinence with increasing time since initiation of regular use was observed. However, among those alive, the estimated prevalence of abstinence for at least 4 months from the above drugs and methadone was only 27% at 20 years since initiation. A higher age at initiation, a calendar year of initiation before 1980, and a Western European ethnic origin were associated with higher prevalence of abstinence. These results indicate that the concept of “maturing out” to a drug-free state does not apply to the majority of drug users. Further studies on determinants of individual transitions in drug use are important in order to establish evidence-based intervention strategies.

Received for publication May 10, 2004; accepted for publication August 24, 2004.

Addiction to illicit drugs is often regarded as a persistent affliction associated with severe consequences (14). In the course of an individual’s drug-using career, episodes of frequent use often alternate with episodes of abstinence (2, 5). Because of high relapse rates, short-term studies on drug-use patterns may present a bleak outlook (6). Long-term studies, however, may suggest higher percentages of abstinence at later stages of the drug-using career (614).

Generalization of research findings on long-term outcome is hampered by differences in drug-using patterns, in drug-related policy across countries, and in definitions of abstinence. Moreover, most long-term studies recruit participants through abstinence-oriented treatment centers or criminal justice systems. In Amsterdam, the Netherlands, approximately 60–70 percent of heroin users annually receive some type of methadone treatment, mainly in programs that aim not to achieve abstinence but to minimize the harm associated with use of illicit drugs (15). Most participants included in the present study were recruited from these easy-access (“low-threshold”) programs (15). However, their study entry and follow-up were independent of any treatment involvement. In the current paper, we describe long-term outcomes regarding death and abstinence among drug users recruited in such a setting of harm reduction.

MATERIALS AND METHODS

Study group

The Amsterdam Cohort Study (ACS) of HIV and AIDS is an open cohort study that was started in 1984 among homosexual men, followed shortly thereafter by the ACS among Drug Users in 1985. The study group for the present analysis consisted of participants in the ACS among Drug Users. The ACS among Drug Users is now referred to as the ACS. Persons with either past or current use of illegal drugs (heroin, cocaine, and/or amphetamines) were eligible; they were recruited mainly through low-threshold methadone programs (including a weekly sexually transmitted disease clinic for drug-using prostitutes) and by word of mouth. The study was approved by the institutional review board of the Academic Medical Center in Amsterdam.

Participation in the ACS is voluntary, and written informed consent is obtained prior to data collection. Participation involves returning every 4 months for follow-up visits. At every visit, a standardized questionnaire on drug use and risk behavior is administered, and a blood sample is drawn for human immunodeficiency virus (HIV) testing. In cases of loss to follow-up (e.g., because of leaving Amsterdam or refusal), ACS staff check the participant’s vital status at regular intervals with the population register in the participant’s town of residency. In cases of death, the cause of death is ascertained by examining hospital records or obtaining information from the coroner’s office.

Of the 1,487 ACS participants enrolled during or after 1985 but before January 2001, we selected 1,237 who were still in follow-up on March 1, 1989, or entered the ACS thereafter. From that date, the questionnaire asked about the calendar year of initiation of regular (at least three times per week) use of illicit drugs. Of these persons, we only included 899 Dutch nationals, because vital status could not be validly determined for foreigners. Further analyses demonstrated comparatively high mortality rates and low abstinence rates among the excluded Dutch nationals for whom ACS follow-up was terminated before March 1, 1989 (n = 145). This implies that, in the present analysis, long-term outcome was probably biased in a favorable direction.

Assessment of drug use, type, and frequency

At ACS entry, frequency of use of heroin, cocaine, and/or amphetamines for the preceding 6 months was recorded; at follow-up visits, frequency of drug use was recorded for the time since the preceding visit. At every visit, drug use was categorized as none (i.e., fully abstinent), monthly (i.e., three times per month or less often), weekly (i.e., at least once per week but not every day), or daily. July 1 of the reported calendar year of starting regular use was regarded as the starting point of a participant’s drug-using career. In a former ACS analysis, it was found that drug users give valid self-reports in a setting where no client–health-care-provider relationship is present (16).

Study dropouts

To adjust for the possible bias associated with selective loss to follow-up, we traced a number of participants (188 out of 899) who had dropped out of the study for at least 1 year on June 30, 2001, or September 30, 2001. These participants had not explicitly refused future contact at the time of their withdrawal (“traceable dropouts”). Tracing was done by means of the population register. A questionnaire on current frequency of drug use was sent by mail; if necessary, a reminder was sent. The number of dropout questionnaires that were filled out and returned was 43. Another group of ACS dropouts (58 out of 899) explicitly refused any future contact at the time of their withdrawal (“untraceable dropouts”). All data on the drug-use patterns of dropouts collected during their ACS follow-up were included in the analysis.

Statistical analysis

Mortality

Mortality rates over time since initiating regular use of illicit drugs were calculated by means of Kaplan-Meier product-limit survival estimates for left-truncated data. Survival times were left-truncated at the time of ACS enrollment, since participants entered the ACS after a time interval following their initiation of drug use. The survival times of participants who had been recruited before March 1989 were left-truncated at March 1, 1989, since these participants were not observable at risk of death between their ACS entry and March 1989. However, their data on drug-use patterns collected between study entry and March 1989 were included in the analysis on abstinence (see below). Among HIV-infected participants, deaths were classified as being related to HIV or acquired immunodeficiency syndrome (AIDS) when they did not result from an injury such as overdose, violence, or suicide. Deaths due to such causes and all deaths among non-HIV-infected participants were classified as not related to HIV/AIDS. To avoid a downward biasing of mortality rates due to reporting delay, ACS data files up to January 2003 were used, but the survival times of those alive and in follow-up were censored at the end of July 2002. The survival analysis was performed with EGRET, version 2.0.1 for Windows (Cytel Software Corporation, Cambridge, Massachusetts).

Abstinence

For the present analysis, we divided the individual drug-using-career time scale into 4-month intervals since initiation of regular drug use. The individual drug-using career began in July 1 of the reported calendar year of starting regular use of illicit drugs and ended at death or July 2002. The frequency of drug use, as reported during an ACS visit, was assigned to the end of a predefined 4-month interval that immediately preceded the ACS visit. This could be done for those predefined 4-month intervals to which the data collected in the ACS referred—that is, from 6 months before ACS enrollment to the last ACS follow-up visit or July 2002. In case of a longer interval between two consecutive ACS visits (e.g., 1 year), the data reported at the first visit following such an interval were assigned to all of our predefined 4-month intervals between those two ACS visits.

We estimated the point prevalence of abstinence at the end of all predefined 4-month intervals along the drug-using-career time scale, using three definitions: 1) no use of illicit drugs and no use of methadone since the last visit; 2) no use of illicit drugs, leaving use of methadone out of consideration; and 3) no use of illicit drugs or occasional use only, leaving both monthly use of illicit drugs (or less) and use of methadone out of consideration. For the present analysis, an episode of abstinence was defined as being abstinent minimally since the last visit (i.e., ≥4 months). A participant who relapsed after an episode of abstinence was reclassified as nonabstinent after that episode. Only participants who were alive and in ACS follow-up at a certain time point in their drug-using career contributed to the estimation of the point prevalence of abstinence for that time point along the career time scale. This resulted in a series of cross-sectional snapshots of the study group based on a varying number of contributing participants. Analyses on abstinence were performed with SAS, version 6.12 for Windows (SAS Institute, Inc., Cary, North Carolina).

We also estimated the point prevalence of abstinence according to definition 1, 2, or 3 with the same data entered into a logistic regression model for binomially distributed data, using the method of generalized estimating equations (GEE) to take into account the correlation between data on the same participant over the course of follow-up (17, 18). For any temporal distance between two of our predefined 4-month intervals (i.e., a distance of 4 months, 8 months, 12 months, etc.), a separate correlation coefficient was estimated. This resulted in a covariance structure with minimal assumptions (stationary M-dependent). Besides time since initiation of drug use and duration of drug use at ACS entry, age at starting regular drug use, gender, calendar year of initiation, and ethnicity were added to the model. A linear relation between time since initiation and the logit of the point prevalence of abstinence was assumed.

Potential sources of bias

Late entry into the study. Partici-pants entered the ACS with various time intervals since initiation of drug use. Because ACS entry was probably associated with an episode of more intense drug use and a lower likelihood of abstinence at that time, and recruitment of participants took place continuously and at various stages since initiation, the prevalence of abstinence might have been underestimated at any point along the drug-using-career time scale. In the descriptive analysis, we took into account this possible source of bias due to continuous selective inflow by stratifying the data according to the participant’s number of years since drug initiation at the time of his/her ACS entry. We reasoned that the data from participants enrolled comparatively soon after their initiation of drug use permitted the most valid estimation of the prevalence of abstinence at any time point. In the GEE model, however, data from participants who entered the ACS after longer intervals since initiation might still have contributed to the estimation of a time trend in the prevalence of abstinence along the drug-using-career time scale. Additionally, in the GEE model, we included duration of drug use at ACS entry as a covariate in order to meet the biasing effect of continuous selective inflow of drug users during an episode of intense use. Including this covariate in the GEE model made it possible to predict the abstinence rates of a “virtual” cohort of drug users ideally recruited at their time of drug-use initiation.

In addition, we excluded data on frequency of current drug use during the 6-month interval preceding the ACS entry visit from all estimations, in order to further attenuate the possible biasing influence of more intense drug use associated with ACS entry.

Loss to follow-up. The estimated prevalence of abstinence at any time point was adjusted for the possible effect of selective dropping out. This was done by re-allocating the number of dropouts at a certain time point along the drug-using-career time scale over the categories “abstinent” and “nonabstinent.” Re-allocation was conducted according to the frequency distribution with respect to current drug use as observed among the dropouts who were traced and returned the dropout questionnaire (n = 43; see table 2). In addition, sensitivity analyses were performed with assumptions of extreme prevalence (i.e., 0 and 100 percent) among all dropouts for whom no dropout questionnaire was available (n = 203). This was done to assess the maximum influence of loss to follow-up on the results.

RESULTS

Description

Characteristics of the study population are given in table 1. The majority of participants reported using a single drug in the calendar year of initiation but using more than one illicit drug at study entry (e.g., 64.8 percent used both heroin and cocaine). Most participants were using methadone at study entry (84 percent) and were recruited primarily (84 percent) from a low-threshold treatment program (including the sexually transmitted disease clinic). A small number of participants were abstinent from illicit drug use at entry (4.3 percent). The median number of ACS visits (baseline and follow-up) was 15; 5 percent of participants had one visit or less and 5 percent had more than 40 visits (5–95 percent range: 1–40). The median number of months between successive visits was 4.1 (5–95 percent range: 3.5–9.2).

Dropouts

The frequency of current drug use among the responding dropouts (n = 43) is shown in table 2. According to definitions 1, 2, and 3, the prevalence of abstinence was 37.2 percent, 41.9 percent, and 48.8 percent, respectively. At the time they completed the dropout questionnaire, the dropouts’ median number of years since initiation of regular drug use was 22.5 years (5–95 percent range: 9.5–29.5). No significant association could be established between the prevalence of abstinence and the number of years since starting illicit drug use at the time of filling out the dropout questionnaire, but this might have been due to the small number of responding dropouts.

Among the untraceable dropouts (n = 58), a significantly higher prevalence of abstinence at their last ACS visit before withdrawal was found, as compared with the prevalence among the traceable dropouts at their last visit (n = 188). Thus, because the prevalence of abstinence among ACS dropouts (table 2) was possibly underestimated and the influence of duration of drug use on abstinence rates among dropouts could not be taken into account, a sensitivity analysis was required (see table 3).

Mortality

The number of observed deaths was 183, of which 89 were unrelated to HIV/AIDS. After 20 years since drug initiation, the estimated survival was 73 percent (i.e., 27 percent mortality; 95 percent confidence interval: 67, 78) when all causes of death were taken into account (figure 1). After censoring of the survival times of those who died from HIV/AIDS, this figure was 84 percent (i.e., 16 percent mortality; 95 percent confidence interval: 78, 88).

Prevalence of abstinence

The observed prevalence of abstinence, according to definition 2, among persons who entered the ACS comparatively early in their drug-using careers (i.e., duration of drug use at study entry <6.3 years (lowest quintile)) was 29 percent at 20 years since initiation (figure 2). This figure rose to 36 percent after adjustment for the higher prevalence of abstinence among the responding dropouts. A lower prevalence of abstinence at 20 years was observed when the more stringent definition of abstinence (definition 1) was used (27 percent, adjusted for the effect of dropouts), whereas using the more flexible definition (definition 3) resulted in a higher observed prevalence (46 percent, adjusted). The much lower prevalence of abstinence (definition 2) at 20 years observed among participants who entered the ACS at a more advanced stage of their drug-using career (duration at ACS entry >17.6 years) is indicative of the bias associated with “late study entry” (figure 2; other three intermediate quintiles not shown). When number of years since initiation at ACS entry was omitted from the analysis, no increasing trend towards abstinence along the drug-using-career time axis was observed. This indicates that the unadjusted overall population trend gives a deceptive picture because of the continuous selective inflow along the career time scale (see Materials and Methods). Of the participants who were abstinent at 20 years according to definition 1, 69 percent were abstinent for at least 1 year. This percentage was 49 percent when definition 2 was used and 60 percent when definition 3 was used.

The GEE-modeled prevalence of abstinence at 20 years for an imaginary group of drug users who were recruited at the time of their initiation of drug use is shown in table 3. In general, the GEE-predicted figures were slightly higher than the figures for observed prevalence among those recruited comparatively early in their drug-using careers (<6.3 years; see figure 2). The range of estimated values associated with extreme assumptions concerning the prevalence of abstinence among dropouts is shown in table 3 as well, and it indicates the maximum uncertainty associated with the influence of study dropouts on the results. The increase in the prevalence of abstinence along the drug-using-career time scale was not significantly modified by the number of years since initiation at ACS entry (i.e., there was no interaction between time and duration of drug use at entry (p > 0.35)). This finding indicates a monotonic trend towards abstinence that is not influenced by the stage of the drug-using career at ACS entry.

Determinants of abstinence

A lower age at initiation, a recent calendar year, and a non-Western-European ethnic origin (mainly Surinam and Netherlands Antilles) were unfavorably associated with a lower estimated prevalence of abstinence (table 4).

DISCUSSION

Among participants in the ACS, we estimated mortality rates and the prevalence of abstinence along the drug-using-career time scale. Although high mortality rates are probably associated with lower abstinence rates, the interrelation between both outcome measures was met by estimating a point prevalence of abstinence among persons who were still alive at a certain time point along the career time scale—that is, by estimating a series of conditional probabilities. Combining the two independently estimated rates, the percentage of drug users who were alive and fully abstinent from illicit drugs and methadone for at least 4 months was 20 percent at 20 years after initiation of regular drug use (73 percent alive × 27 percent abstinent). This figure was 28 percent when use of methadone was left out of consideration. Even when we excluded mortality due to HIV/AIDS, defined abstinence to allow both occasional use of illicit drugs and use of methadone, and assumed an unlikely high prevalence of abstinence among ACS dropouts, this percentage was still less than 50 (84 percent alive × 56 percent abstinent = 47 percent).

For a sound interpretation of the presented data, sources of potential bias must be acknowledged. First, the majority of participants entered the ACS a considerable number of years after their initiation of drug use. Thus, mortality rates might have been underestimated, since participation was conditional on being alive at the time of study entry (or being in follow-up in March 1989). We addressed this bias by left-truncating the individual survival times, leaving out of consideration the time during which the participant was not observable at risk of death. However, because hardly any participant was enrolled immediately after the initiation of illicit drug use and no deaths were registered during the first 5 years of the drug-using time scale, the estimated mortality rates may still be regarded as being biased downwards. When assuming that the mortality rates as observed from 5 years onwards are similar to the (probably) unobserved mortality rates during the very early stages of the drug-using career, the time axis of the survival graph (figure 1) may be shifted to the right by relating the figures for the 25-year time point to the 20-year time point. This probably yields a more reliable estimate of mortality (38 percent instead of 27 percent at 20 years). Second, “late entry into the study” also implied that the prevalence of abstinence at any time point was underestimated. Therefore, number of years since initiation at ACS entry was included as a covariate in all analyses (see Materials and Methods). Third, study withdrawal in association with high rates of abstinence may cause a further underestimation of the prevalence of abstinence. We addressed this objection by reallocating all dropouts on the basis of current drug use reported by a number of traceable ACS dropouts (table 2). Furthermore, we performed a sensitivity analysis to assess the maximum uncertainty range around the estimations associated with loss to follow-up (table 3).

In a meta-analysis by Cramer and Schippers, data from a large number of follow-up studies on drug-using career outcome were summarized (6, 9, 11). They indicated that approximately 20 percent of the subjects had died after 20 years; among those who were alive, 50 percent had become abstinent and the other 50 percent were still using drugs either on a daily basis or in a less problematic way. The studies were all conducted before the outbreak of HIV, which had a major impact on mortality among injecting drug users. The 16 percent rate of mortality due to HIV/AIDS-nonrelated causes after 20 years that we found in the present study may be regarded as comparatively low but was probably biased downwards (see above). Right-shifting of the time axis by 5 years may yield a mortality rate comparable to that of Cramer and Schippers (21 percent at 20 years; see above). In other Dutch studies conducted before or after the spread of HIV, mortality rates found among drug users in the Netherlands were low in comparison with other countries (19, 20). Since the original aim of the ACS was to monitor the HIV epidemic, participants in the ACS may represent a sample of drug users who experience a higher risk of death in association with a high-risk lifestyle than persons in other drug-using populations in the Netherlands.

The estimated prevalence of abstinence found in our study seems to be less favorable than that found in other studies (6, 911). One possible explanation is that the majority of studies conducted in the 1970s and 1980s involved heroin addicts, while later studies involved users of multiple drugs (12). Cocaine use has increased during the past several decades, and it has been reported that cocaine use is related to a poor treatment outcome, a high relapse rate, and continued use of heroin (2123). We found a trend towards a lower prevalence of abstinence with increasing calendar year of initiation. In-depth analyses, in which the types of drugs used and other drug-related characteristics such as intravenous administration are handled as time-dependent variables, are required in order to obtain insight into important determinants of abstinence and relapse and associated calendar-time trends.

We found a higher prevalence of abstinence associated with a higher age at initiation of drug use, especially when the more stringent definitions of abstinence were used (table 4). This finding in itself supports Winick’s theory of the “natural history” of addiction, which is closely related to the life cycle of the addict and the problems associated with adolescence (8). However, the high mortality and low abstinence rates found long-term in the present study do not suggest that “maturing out” is operative among the majority of persons who become severely addicted, as was suggested by Winick’s study. Other studies, preferably studies including qualitative interviews, are needed in order to better understand individual motives for continuation of drug use at older ages and whether depraved living conditions may hamper a naturally occurring “maturing out” and associated rehabilitation. On the other hand, the present findings do not rule out a tendency towards less problematic or better “controlled” use in the course of an addiction career (24). This would require analyses not in which drug-use patterns were dichotomized in terms of use versus abstinence but in which a more refined categorization from intense use towards moderate use and abstinence was employed.

Some limitations of our study require consideration. First, our results primarily reflect the addiction careers of persons with opiate addiction and full access to low-threshold methadone treatment programs. Second, the ACS data did not permit us to make a distinction between enforced abstinence during imprisonment or hospitalization and internally motivated abstinence. Precise coincidence of a prison or hospital episode with the interval between two ACS visits is unlikely, and high rates of (noninjecting) use of heroin and cocaine have been reported in Dutch prisons (25). Thus, the influence of episodes of enforced abstinence on our estimated prevalence of abstinence may have been limited. Next, estimations of point prevalence along the drug-using-career time scale do not provide insight into the individual conversion between intense use and abstinence (2). In-depth analyses of the ACS data for the study of individual transitions in frequency of drug use and their associated determinants are currently under way. Finally, the present analysis does not provide insight into how persons with a history of severe addiction became addicted following their first use of illicit drugs. Evidence indicates that initial use of illicit drugs does not necessarily degenerate into loss of control and a marginal lifestyle (26). Further studies are required in order to explore the drug-using careers of persons who become severely addicted versus the careers of those who manage to keep their use within bounds.

In conclusion, we found a favorable trend towards abstinence in this Dutch cohort of illicit drug users. However, the high mortality rates and the low prevalence of abstinence among those who remained alive over the long term indicate that the concept of natural recovery or “maturing out” to a drug-free state does not apply to a substantial portion of the addict population. An older group of drug users who are addicted primarily to opiates may increasingly attract attention from public health service providers in the Netherlands (27). Studies on the primary prevention of addiction and the prevention of relapse deserve high priority.

ACKNOWLEDGMENTS

This work was supported by the Addiction Research Program of the Netherlands Organization for Health Research and Development (grant 31000049). The Amsterdam Cohort Study (ACS) among Drug Users is sponsored by the Netherlands Organization for Health Research and Development; the Ministry of Health, Welfare, and Sport; and the AIDS Fonds Netherlands (grant 4141).

The authors thank Dr. Ronald B. Geskus and Prof. Dr. Roel A. Coutinho for critically reviewing an earlier version of this article. The authors are indebted to Else te Brake for designing the questionnaire and organizing data collection for the dropout study and to research nurses Maja Totté, Ans Snuverink, and Joke Bax for daily administration of the ACS. The authors also thank Lucy Phillips for final editorial review.

FIGURE 1. Survival over time since the start of regular use of illicit drugs, Amsterdam Cohort Study among Drug Users, 1985–2002. The graph shows Kaplan-Meier product-limit estimates for left-truncated data, including deaths from all causes, deaths related to human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) only, and HIV/AIDS-nonrelated deaths only.

FIGURE 1. Survival over time since the start of regular use of illicit drugs, Amsterdam Cohort Study among Drug Users, 1985–2002. The graph shows Kaplan-Meier product-limit estimates for left-truncated data, including deaths from all causes, deaths related to human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) only, and HIV/AIDS-nonrelated deaths only.

FIGURE 2. Prevalence of abstinence (≥4 months) from illicit drug use among those still alive (according to definition 2: no use of illicit drugs, leaving methadone use out of consideration), by 4-month interval since initiation of illicit drug use, Amsterdam Cohort Study among Drug Users, 1985–2002. Prevalence was calculated according to the categorized number of years since initiation at study entry (lowest and highest quintiles are shown). The prevalence modeled by means of generalized estimating equations (GEE) was estimated by combining all data in one model and including the number of years since initiation (YSI) at study entry as a covariate (see Materials and Methods).

FIGURE 2. Prevalence of abstinence (≥4 months) from illicit drug use among those still alive (according to definition 2: no use of illicit drugs, leaving methadone use out of consideration), by 4-month interval since initiation of illicit drug use, Amsterdam Cohort Study among Drug Users, 1985–2002. Prevalence was calculated according to the categorized number of years since initiation at study entry (lowest and highest quintiles are shown). The prevalence modeled by means of generalized estimating equations (GEE) was estimated by combining all data in one model and including the number of years since initiation (YSI) at study entry as a covariate (see Materials and Methods).

TABLE 1.

Characteristics of the Dutch participants at the time of study entry (n = 899), Amsterdam Cohort Study among Drug Users, 1985–2002

Characteristic Mean or median 5%–95% range No. 
Mean age (years) 32.1 21.5–44.4   
Male gender   615 68.4 
Ethnic origin     
Western Europe   721 80.2 
Surinam/Netherlands Antilles   132 14.7 
Other   46 5.1 
Median month and year of study entry October 1991 March 1986–October 2000   
Mean time (years) between starting regular use of illicit drugs and study entry 12.2 2.4–23.5   
Mean duration of follow-up (years) 8.8 1.5–15.9   
Use of illicit drugs (heroin, cocaine, and/or amphetamines) and methadone     
Frequency of use     
Daily     
Plus methadone   483 53.7 
Without methadone   57 6.3 
Weekly     
Plus methadone   181 20.1 
Without methadone   48 5.3 
Monthly or less often     
Plus methadone   73 8.1 
Without methadone   15 1.7 
Use of methadone only   19 2.1 
Fully abstinent from both illicit drugs and methadone   20 2.2 
Type of illicit drug(s) used*     
No use of heroin, cocaine, or amphetamines   39 4.3 
Use of:     
Heroin only   102 11.3 
Cocaine only   59 6.6 
Amphetamines only   10 1.1 
Heroin and cocaine   583 64.8 
Heroin, cocaine, and amphetamines   75 8.3 
Other combinations   31 3.4 
Type of drug(s) reportedly used in the calendar year of initiation     
Heroin only   393 43.7 
Cocaine only   93 10.3 
Amphetamines only   139 15.5 
Heroin and cocaine   181 20.1 
Heroin, cocaine, and amphetamines   28 3.1 
Other combinations   65 7.2 
Characteristic Mean or median 5%–95% range No. 
Mean age (years) 32.1 21.5–44.4   
Male gender   615 68.4 
Ethnic origin     
Western Europe   721 80.2 
Surinam/Netherlands Antilles   132 14.7 
Other   46 5.1 
Median month and year of study entry October 1991 March 1986–October 2000   
Mean time (years) between starting regular use of illicit drugs and study entry 12.2 2.4–23.5   
Mean duration of follow-up (years) 8.8 1.5–15.9   
Use of illicit drugs (heroin, cocaine, and/or amphetamines) and methadone     
Frequency of use     
Daily     
Plus methadone   483 53.7 
Without methadone   57 6.3 
Weekly     
Plus methadone   181 20.1 
Without methadone   48 5.3 
Monthly or less often     
Plus methadone   73 8.1 
Without methadone   15 1.7 
Use of methadone only   19 2.1 
Fully abstinent from both illicit drugs and methadone   20 2.2 
Type of illicit drug(s) used*     
No use of heroin, cocaine, or amphetamines   39 4.3 
Use of:     
Heroin only   102 11.3 
Cocaine only   59 6.6 
Amphetamines only   10 1.1 
Heroin and cocaine   583 64.8 
Heroin, cocaine, and amphetamines   75 8.3 
Other combinations   31 3.4 
Type of drug(s) reportedly used in the calendar year of initiation     
Heroin only   393 43.7 
Cocaine only   93 10.3 
Amphetamines only   139 15.5 
Heroin and cocaine   181 20.1 
Heroin, cocaine, and amphetamines   28 3.1 
Other combinations   65 7.2 

* Methadone was excluded because it was not illicit.

TABLE 2.

Use of illicit drugs (heroin, cocaine, and/or amphetamines) and methadone among study dropouts who completed the dropout questionnaire (n = 43), Amsterdam Cohort Study among Drug Users, 1985–2002

 No. 
Frequency of use   
Daily   
Plus methadone 16 37.2 
Without methadone 4.7 
Weekly   
Plus methadone 9.3 
Without methadone 
Monthly or less often   
Plus methadone 4.7 
Without methadone 2.3 
Use of methadone only 4.7 
Fully abstinent from both illicit drugs and methadone 16 37.2 
 No. 
Frequency of use   
Daily   
Plus methadone 16 37.2 
Without methadone 4.7 
Weekly   
Plus methadone 9.3 
Without methadone 
Monthly or less often   
Plus methadone 4.7 
Without methadone 2.3 
Use of methadone only 4.7 
Fully abstinent from both illicit drugs and methadone 16 37.2 
TABLE 3.

Modeled prevalence of abstinence (≥4 months) at 20 years since initiation of regular drug use and maximum influence of dropouts on the results, Amsterdam Cohort Study among Drug Users, 1985–2002

Definition of abstinence Estimated prevalence without adjustment for prevalence among dropouts  Estimated prevalence (%) after adjustment for prevalence among dropouts 
 Prevalence (%) 95% confidence interval  According to observations in the sample of dropouts (n = 43) Assuming that other dropouts (n = 203*) were all nonabstinent Assuming that other dropouts (n = 203*) were all abstinent 
1—No use of illicit drugs and no use of methadone 24 16, 35  27 20 39 
2—No use of illicit drugs† 38 29, 48  39 31 50 
3—No use of illicit drugs or monthly use at most† 45 37, 53  46 36 56 
Definition of abstinence Estimated prevalence without adjustment for prevalence among dropouts  Estimated prevalence (%) after adjustment for prevalence among dropouts 
 Prevalence (%) 95% confidence interval  According to observations in the sample of dropouts (n = 43) Assuming that other dropouts (n = 203*) were all nonabstinent Assuming that other dropouts (n = 203*) were all abstinent 
1—No use of illicit drugs and no use of methadone 24 16, 35  27 20 39 
2—No use of illicit drugs† 38 29, 48  39 31 50 
3—No use of illicit drugs or monthly use at most† 45 37, 53  46 36 56 

* A total of 145 traceable dropouts without dropout questionnaires and 58 dropouts who refused any future contacting at the time of study withdrawal.

† In this definition, use of methadone was left out of consideration.

TABLE 4.

Odds ratios* for abstinence (≥4 months) from illicit drug use according to three different definitions, Amsterdam Cohort Study among Drug Users, 1985–2002

 Definition of abstinence 
 1—No use of illicit drugs and no use of methadone  2—No use of illicit drugs†  3—No use of illicit drugs or monthly use at most† 
 Adjusted OR‡  95% CI‡  Adjusted OR 95% CI  Adjusted OR  95% CI 
Time since initiation of drug use (per year) 1.06 1.03, 1.08  1.08 1.06, 1.10  1.03 1.02, 1.05 
p value§ <0.0001   <0.0001   <0.0001  
Duration of drug use at study entry (years) 0.88 0.84, 0.93  0.90 0.87, 0.94  0.94 0.91, 0.96 
p value <0.0001   <0.0001   <0.0001  
Age at starting drug use (per year) 1.04 1.01, 1.07  1.03  1.01, 1.07  1.01 0.99, 1.03 
p value 0.009   0.004   0.18  
Male vs. female gender 1.56 0.95, 2.55  1.05 0.76, 1.44  1.02 0.81, 1.29 
p value 0.076   0.74   0.81  
Date of starting drug use (before July 1980 vs. July 1980 or later) 1.71 1.01, 2.89  1.48 1.03, 2.12  1.45 1.90, 1.11 
p value 0.044   0.031   0.005  
Western European ethnicity vs. other 1.40 0.85, 2.30  1.54 1.07, 2.21  1.58 1.19, 2.10 
p value 0.18   0.019   0.0014  
 Definition of abstinence 
 1—No use of illicit drugs and no use of methadone  2—No use of illicit drugs†  3—No use of illicit drugs or monthly use at most† 
 Adjusted OR‡  95% CI‡  Adjusted OR 95% CI  Adjusted OR  95% CI 
Time since initiation of drug use (per year) 1.06 1.03, 1.08  1.08 1.06, 1.10  1.03 1.02, 1.05 
p value§ <0.0001   <0.0001   <0.0001  
Duration of drug use at study entry (years) 0.88 0.84, 0.93  0.90 0.87, 0.94  0.94 0.91, 0.96 
p value <0.0001   <0.0001   <0.0001  
Age at starting drug use (per year) 1.04 1.01, 1.07  1.03  1.01, 1.07  1.01 0.99, 1.03 
p value 0.009   0.004   0.18  
Male vs. female gender 1.56 0.95, 2.55  1.05 0.76, 1.44  1.02 0.81, 1.29 
p value 0.076   0.74   0.81  
Date of starting drug use (before July 1980 vs. July 1980 or later) 1.71 1.01, 2.89  1.48 1.03, 2.12  1.45 1.90, 1.11 
p value 0.044   0.031   0.005  
Western European ethnicity vs. other 1.40 0.85, 2.30  1.54 1.07, 2.21  1.58 1.19, 2.10 
p value 0.18   0.019   0.0014  

* Odds ratios from multivariate logistic regression analysis, adjusted for number of years since initiation of drug use at the time of study entry.

† In this definition, use of methadone was left out of consideration.

‡ OR, odds ratio; CI, confidence interval.

§ Two-sided p value from χ2 test.

Correspondence to Dr. Fabian Termorshuizen, Municipal Health Service Amsterdam, Nieuwe Achtergracht 100, 1018 WT Amsterdam, the Netherlands (e-mail: ftermorshuizen@gggd.amsterdam.nl or FTermorshuizen@cs.com).

References

1.
McLellan AT, Lewis DC, O’Brien CP, et al. Drug dependence, a chronic medical illness: implications for treatment, insurance, and outcomes evaluation.
JAMA
 
2000
;
284
:
1689
–95.
2.
Galai N, Safaeian M, Vlahov D, et al. Longitudinal patterns of drug injection behavior in the ALIVE Study cohort, 1988–2000: description and determinants.
Am J Epidemiol
 
2003
;
158
:
695
–704.
3.
Buning EC, Coutinho RA, van Brussel GHA, et al. Preventing AIDS in drug addicts in Amsterdam. (Letter).
Lancet
 
1986
;
1
:
1435
.
4.
Robertson JR, Bucknall ABV, Skidmore CA, et al. Remission and relapse in heroin users and implications for management: treatment control or risk reduction.
Int J Addict
 
1989
;
24
:
229
–46.
5.
Hser YI, Anglin MD, Grella C, et al. Drug treatment careers: a conceptual framework and existing research findings.
J Subst Abuse Treat
 
1997
;
14
:
543
–58.
6.
Cramer EASM, Schippers GM. Visies op en feiten over ontwenning. In: Cramer EASM, Schippers GM, eds. Zelfcontrole en ontwenning van harddrugs: eindrapport van een onderzoek naar de ontwikkeling en evaluatie van een zelfcontrole-programma voor druggebruikers. (In Dutch). Nijmegen, the Netherlands: University of Nijmegen Research Group on Addictive Behaviours, 1994:21–50.
7.
Winick C. Maturing out of narcotic addiction.
Bull Narc
 
1962
;
14
:
1
–7.
8.
Winick C. The life cycle of the narcotic addict and of addiction.
Bull Narc
 
1964
;
16
:
1
–11.
9.
Maddux JF, Desmond DP. New light on the maturing out hypothesis in opioid dependence.
Bull Narc
 
1980
;
32
:
15
–25.
10.
Swierstra K. Heroïneverslaving: levenslang of gaat het vanzelf over? Een kwart eeuw buitenlandse follow-up studies.
Tijdschr Alcohol Drugs
 
1987
;
13
:
78
–92.
11.
Onstein EJ. Heroïneverslavingscarrières. Hoe is het beloop in de tijd? Heeft therapie effect? Zijn er prognostisch gunstige factoren?
Tijdschr Psychiatrie
 
1987
;
29
:
516
–25.
12.
Tims FM, Fletcher BW, Hubbard RL. Treatment outcomes for drug abuse clients.
NIDA Res Monogr
 
1991
;
106
:
93
–113.
13.
Valliant GE. A 20-year follow-up of New York narcotic addicts.
Arch Gen Psychiatry
 
1973
;
29
:
237
–41.
14.
Hser YI, Hoffman V, Grella CE, et al. A 33-year follow-up of narcotics addicts.
Arch Gen Psychiatry
 
2001
;
58
:
503
–8.
15.
Langendam MW, van Haastrecht HJA, van Brussel GHA, et al. Differentiation in the Amsterdam methadone dispensing circuit: determinants of methadone dosage and site of methadone prescription.
Addiction
 
1998
;
93
:
61
–72.
16.
Langendam MW, van Haastrecht HJA, van Ameijden EJC. The validity of drug users’ self-reports in a non-treatment setting: prevalence and predictors of incorrect reporting methadone treatment modalities.
Int J Epidemiol
 
1999
;
28
:
514
–20.
17.
van Ameijden EJC, Coutinho RA. Large decline in injecting drug use in Amsterdam, 1986–1998: explanatory mechanisms and determinants of injecting transitions.
J Epidemiol Community Health
 
2001
;
55
:
356
–63.
18.
Zeger S, Liang KY. Longitudinal data analysis for discrete and continuous outcomes.
Biometrics
 
1986
;
42
:
121
–30.
19.
Brinkman N. Over dood en dosis: een literatuuronderzoek naar sterfte onder heroïnegebruikers. (NcGv series no. 80). Utrecht, the Netherlands: Nederlands Centrum Geestelijke Volksgezondheid, 1985.
20.
Buster MCA, Sperati A, Bargagli AMM, et al. Reference rates in mortality studies among opiate users. In: Buster MCA, ed. Prevalence, morbidity, and mortality among heroin users and methadone patients. (Doctoral thesis). Amsterdam, the Netherlands: University of Amsterdam, 2003:119–33.
21.
Craddock SG, Rounds-Bryant JL, Flynn PM, et al. Characteristics and pretreatment behaviors of clients entering drug abuse treatment: 1969 to 1993.
Am J Drug Alcohol Abuse
 
1997
;
23
:
43
–59.
22.
Pérez DLC, Trujols J, Ribalta E, et al. Cocaine use immediately prior to entry in an inpatient heroin detoxification unit as a predictor of discharges against medical advice.
Am J Drug Alcohol Abuse
 
1997
;
23
:
267
–79.
23.
Hartel DM, Schoenbaum EE, Selwyn PA, et al. Heroin use during methadone maintenance treatment: the importance of methadone dose and cocaine use.
Am J Public Health
 
1995
;
85
:
83
–8.
24.
Harding WM, Zinberg NE, Stelmack SM, et al. Formerly-addicted-now-controlled opiate users.
Int J Addict
 
1980
;
15
:
47
–60.
25.
Van Haastrecht HJA, Bax JS, van den Hoek JAR. High rates of drug use, but low rates of HIV risk behaviours among injecting drug users during incarceration in Dutch prisons.
Addiction
 
1998
;
93
:
1417
–25.
26.
Cohen P. Shifting the main purpose of drug control: from suppression to regulation of use.
Int J Drug Policy
 
1999
;
10
:
223
–34.
27.
Vitale SG, Meerkerk GJ. Oudere druggebruikers in Rotterdam en hun behoefte aan zorg.
Tijdschr Soc Geneeskd
 
2003
;
81
:
135
–40.