Secondary Logo

Journal Logo

ARTICLE

Maximum impact of HIV prevention measures targeted at injecting drug users

van Ameijden, Erik J.C.1,2; Coutinho, Roel A.1

Author Information
  • Free

Abstract

Introduction

Amongst injecting drug users (IDU), HIV and hepatitis B and C viruses are mainly transmitted via the sharing of needles and syringes, although sexual transmission of these viruses also occurs [1]. Previous studies have showed that IDU are capable of reducing injecting risk behaviours, such as borrowing and lending of used needles [2,3]. In a previous review [3], it was concluded that various combinations of HIV prevention measures can be effective in reducing risk behaviour amongst IDU, although a complete elimination of risk behaviour has never been reported.

Several studies in Amsterdam originating from a comprehensive cohort study among drug users have showed a large reduction but no elimination in HIV incidence [4], injecting [5] and sexual risk behaviours [6,7]. In Amsterdam, a city with an estimated 6000 drug users, including 2500 IDU [8], HIV prevention measures have been implemented within the concept of harm reduction [9,10]. In 1984, large-scale needle-exchange services became available with no limits to the number of needles exchanged; there are now 14 exchange locations in Amsterdam. Needles and syringes can easily be bought at shops in the red light district and pharmacies, for example. Methadone programmes were started before the beginning of the HIV epidemic. Methadone is dispensed at various thresholds from these sites, the lowest being at the Drug Department of the Municipal Health Service (there is no waiting list, and continued drug use or injecting is allowed) [11,12]. An estimated 70% of Amsterdam IDU are in contact with one of the methadone programmes. HIV counselling and testing services have been available in Amsterdam since 1986/1987, but drug users are not actively approached for testing [13]. Other interventions include outreach activities, targeted at drug users both in and out of treatment, and specific interventions for drug-using prostitutes [e.g., special sexually transmitted disease (STD) clinics, free condoms].

The main study question was whether there has been an ongoing reduction in injecting risk behaviours in Amsterdam in recent years, or whether there has been a stabilization of HIV risk as indicated by HIV incidence and injecting risk behaviours given the extensive preventive efforts. The analyses of these behavioural trends is an extension of a previous study in the Amsterdam cohort in which only intake (first) visits were analysed until 1992 [5]; long-term follow-up of IDU is now available (1986–1997), and novel analytical methods allow analysis of both intake and follow-up visits, resulting in a major increase in statistical power and opportunities for detailed data analyses. For planning and evaluation of interventions it is important to gain insight into the time it can take before programmes have their maximum effect. There may be a strongly delayed effect of the interventions (e.g., long-term influence on attitudes and norms against needle-sharing), or a stabilization or relapse of HIV risk, indicative of some level of risk that may be difficult to prevent. Behavioural trends have also been stratified by frequency of attending needle-exchange and methadone programmes, where a decline in risk behaviour in all strata will indicate an intended community-wide impact of these interventions.

Given some level of residual risk, the second aim is to study the determinants of borrowing and lending amongst HIV-positive and HIV-negative IDU in order to find clues for improvement of the interventions. Prevention may be expanded to high-risk groups and insight may be gained into the causal determinants of borrowing and lending. In general, there are very few studies in which risk factors for lending have been determined, lending being as crucial as borrowing for HIV transmission. Lending amongst HIV-negative IDU constitutes a public health risk because hepatitis B and C viruses are also prevalent in this group [1]. Borrowing amongst HIV-positive IDU may be relevant given the possibility of reinfection with HIV or hepatitis virus subtypes that are more virulent or therapy-resistant.

Methods

In December 1985, an open cohort study (continuous recruitment of new participants) amongst IDU was started in Amsterdam [14]. Participants were required to be free of AIDS at intake. Recruitment was via methadone programmes, an STD clinic for drug-using prostitutes (until June 1996), and by self-referral. Participation was voluntary with informed consent. Participants are asked to return every 4 months for follow-up visits. At every visit, blood was drawn and a standardized questionnaire was administered by specially trained nurses. At intake, questions regarding current behaviour referred to the period of 6 months preceding the intake, whereas at follow-up visits these questions referred to the period between the present and preceding visits. Blood specimens were tested for HIV by enzyme-linked immunosorbent assay, and positive specimens were confirmed by immunoblot.

Participants were counselled at every study visit and received their HIV test results and post-test counselling by visiting the study centre 3 weeks or more after a visit, or at maximum at the following study visit. There was a gradual increase in the proportion attending intake visits who had already received a prior HIV test result (from 29% in 1989 to 59% in 1994) [13].

For the purpose of the present study, only cohort visits in which current injecting was reported were selected. The study period was from January 1986 to June 1997. The current injecting behaviours that were used as outcome variables were borrowing (defined as using a needle or syringe that had already been used by somebody else), lending (giving a needle or syringe that had been used by the participant to somebody else), multiple needle use (mainly injecting with one's own needle or syringe more than once), and frequent injecting (injecting two times per day or more). Trends are presented stratified according to HIV serostatus, because declining trends over time only become apparent after stratification, as seen in a previous study in this cohort [7].

Data analysis

Trends in HIV incidence (expressed as new HIV infections per 100 person-years of follow-up) are presented. For statistical testing, additive Poisson regression models were used (see below).

To determine trends in the four injecting risk behaviour groups, study visit was the unit of analysis. Because observations within one individual were not independent but positively interrelated, generalized estimating equations (GEE) were used for data analyses [15,16]. Exploration of the data revealed that an exchangeable autocorrelation structure gave the best fit (same correlation between behaviour at all cohort visits for one person), and almost identical results were obtained using other correlation structures.

To answer the main study question, the percentage reporting an injecting risk behaviour was first modelled as a linear function of calendar year (additive model). If adding the quadratic term of calendar year significantly improves the fit of the model, and if its effect estimate is positive, the rate of behaviour change slows down and one can calculate the estimated rate of behaviour change in a specific calendar year (differential equation). An additive model rather than a logistic model was chosen because (i) the main effect of calendar year in a logistic model inherently assumes a non-linear behavioural trend (i.e., a declining rate of behavioural change), and (ii) results can be more easily interpreted. To implement this in GEE, the dichotomous outcomes were modelled using an identity (linear) link function with a binomial mean-variance relationship. Because percentages can be estimated in such models to be < 0 and > 100%, special attention was paid to the fit of the model by comparison of (i) observed percentages, (ii) predicted percentages using model coefficients, and (iii) predicted values based on GEE output (no important deviations were present). Testing was two-sided and P values < 0.05 were considered significant.

Behavioural trends were adjusted for the number of the cohort visit, which was used to indicate a protective effect of the HIV testing and counselling component in the cohort, and increasing unwillingness to admit the same risk behaviours at subsequent cohort visits (socially desirable answering). As a result, trends become more generalizable, but the absolute levels of self-reported risk behaviour at follow-up visits may remain underestimated. The number of the cohort visit was coded from 0 to 5 (0, visit 1; 1, visits 2–3; 2, visits 4–6; 3, visits 7–12; 4, visits 13–20; 5, visits > 20), and was initially treated as a categorical variable. The basic model was defined as always containing calendar year and the number of the cohort visit; the quadratic effect of year was only included if significant.

Changes in the population structure may explain behavioural trends. To examine this, the basic models were extended with the following sociodemographic variables: gender, age (four categories; quartiles), nationality (Dutch, German, other), ethnicity (Western European, Surinamese/Antillean, other), time since onset of injecting (four categories; quartiles), and current prostitution. Adjustment for changes in current behaviours other than the outcomes can be over-adjustment as preventive measures may change both injecting risk behaviour and current behaviours (e.g., frequency of drug use), but may indicate causes of changes in injecting risk behaviour. Current behaviour variables were as follows: frequency of alcohol use (none, 1–4, > 4 units per day) and tranquillizer use (none, less than daily, daily), frequency of non-parenteral heroin and cocaine use (none, less than daily, daily), type of drug usually injected (heroin, cocaine, speedball, other), frequency of injecting (less than daily, daily, more than daily), multiple needle use (none, 2–4, > 4 times injecting per needle), injecting with others (0, 1–99, 100%), main injecting place (at home, not at home), percentage of new needles obtained via the needle-exchange programmes (0, 1–99, 100%), and frequency of methadone use (none, less than daily, daily; this mainly concerns methadone obtained via the methadone dispensing circuit, and was a valid surrogate for frequency of attending methadone programmes).

To explore possible determinants of borrowing and lending, we investigated which of the above variables could significantly improve the fit of the basic models. If the relationship between a covariate and an outcome did not differ between the covariate categories shown above, these categories were collapsed. Only variables that significantly (P < 0.05) improved the basic models were available for further multivariate analysis to determine independent predictors. To check whether trends in borrowing and lending were different according to strata of frequency of attending methadone and needle-exchange programmes, we evaluated whether the interaction terms between attendance of a programme and calendar year (linear and quadratic term) could improve the basic model that also contained the main effect of programme attendance. Selection of cohort visits with an interval between visits of less than 6 months gave similar results.

Results

A total of 1234 IDU were recruited between January 1986 and June 1997. Current injecting was reported at 6645 visits by 879 participants, corresponding to 2966 person-years of observation. The mean time period between visits was 5.3 months; for 11% of visits this period exceeded 6 months, and for only 3% of visits this period exceeded 1 year. For 760 IDU, the first selected visit was the first (intake) visit in the cohort; the remaining 119 (14%) IDU (re)started injecting during follow-up. Fifty-two per cent of IDU had their first selected visit in 1988 or earlier.

Of the 879 IDU, 42% were women, 65% were Dutch, 19% were German, and 91% were Caucasian. At the first selected visit, 88% were currently attending a methadone programme, 30% were HIV-positive, mean age was 30.5 years (SD, 6.3), and mean time since first injection was 9.1 years (SD, 6.2); 17% of IDU had their first injection ≤ 2 years previously. The main type of drug injected was heroin in 24%, cocaine in 12%, speedball in 56%, and other drugs in 7% (mainly amphetamines and methadone).

Trends in calendar time

Of 460 HIV-negative IDU who reported current injecting, 76 seroconverted during 1637 person-years of follow-up. HIV incidence declined from 8 to 4 per 100 person-years during 1986–1991 (P = 0.02), although thereafter no further changes were observed (Fig. 1). In an additive Poisson regression model containing the linear and quadratic effects of calendar year, the latter variable was of borderline significance (P = 0.092).

Fig. 1
Fig. 1:
. Trends in injecting risk behaviours by HIV serostatus. (▪), Frequent injecting; (□), multiple needle use; (*), borrowing; (•), lending; (♦), HIV incidence [per 100 person-years (PY); HIV-negative only].

Similarly, there were major risk reductions for all four outcome variables, amongst both HIV-seropositive and HIV-negative IDU (Fig. 1). However, with regard to borrowing, lending and multiple needle use, there was no further decline in recent years; only the decline in frequent injecting was ongoing. In 1986–1987, HIV-positive IDU already reported a substantially lower level of borrowing than HIV-negative IDU (25 versus 43%), whereas for the other outcomes there were no major initial differences. In 1996–1997, the percentage of lending and multiple needle use was lower amongst HIV-positive IDU.

Table 1 shows the basic models for the behavioural outcomes (see Methods). At HIV-negative visits, the rate of decline of borrowing, lending and multiple needle use decreased as the quadratic effects of calendar year were significant. According to model coefficients no further risk reduction occurred in more recent calendar years. For instance, the rate of decline of borrowing was estimated at −4.2% per year in 1986 (year 0), −1.7% per year in 1991 (year 5; the differential equation of −4.2 × year and +0.25 × year2 is −4.2% + 2 × 0.25 × year), and −0.2% per year in 1994 (year 8); note that there was no evidence for relapse because the estimated increase in borrowing from 1995 onwards was rather small and the 95% confidence intervals included 0%. At HIV-positive visits, the rate of decline became significantly smaller for multiple needle use and lending. However, because at these visits the percentage lending injecting equipment dropped below 10% from 1988 onwards (Fig. 1), cautious interpretation was needed because there was little room for further risk reduction (percentages could not drop below 0%).

Table 1
Table 1:
Modelling trends in injecting risk behaviour using 6645 cohort visits of 879 injecting drug users in the Amsterdam cohort study, 1986–1997.

With regard to the outcome of frequent injecting, the quadratic effect of year was not significant (P > 0.50), and no major differences were observed between HIV-negative and positive IDU with regard to intercept and rate of decline (Table 1).

Table 1 also shows that at follow-up visits, compared with intake visits, much lower percentages of all four outcomes were observed. With one exception, there was no further variation in injecting risk behaviour within follow-up visits (data not shown); only for the outcome of multiple needle use amongst HIV-negative IDU was there a linear effect of the number of the cohort visit. Adjustment for the number of the cohort visit as a continuous (absolute and log-transformed) or categorical (six categories; see Methods) variable did not change the effect estimates of the linear and quadratic effect of calendar year. Not adjusting for the number of the cohort visit would have resulted in approximately twofold higher estimates of the rate of decline of borrowing and lending, and a 1–1.5-fold higher rate of decline of multiple needle use and frequent injecting (i.e., the number of the cohort visit was a strong confounder).

The considerably lower risk estimates at follow-up visits limits the possibility for further risk reduction. Indeed, for borrowing and lending amongst HIV-negative IDU and for borrowing amongst HIV-positive IDU, the interaction term between calendar year and the number of the cohort visit significantly improved the basic models. In general, declining trends were more pronounced in the subset of intake visits than in follow-up visits. However, for all strata the finding holds that no further risk reduction occurred since 1991 (Fig. 2). Amongst HIV-negative IDU, borrowing stabilized at about 30% at intake visits and at 15% at follow-up visits (the quadratic term of year was significant at both intake and follow-up visits). Patterns were comparable for lending amongst HIV-negative IDU and borrowing amongst HIV-positive IDU: at intake visits, risk did not further decrease from 1990/1991 onwards, whereas at follow-up visits, levels of risk were stable and low during the entire study period. There was no firm evidence for relapse because the apparent increase in borrowing among both HIV-positive and negative IDU at intake visits from 1992 onwards (Fig. 2) was not significant.

Fig. 2
Fig. 2:
. Trends in borrowing and lending at intake and follow-up visits. (—), borrowing among HIV-negative injecting drug users (IDU); (- -), lending among HIV-negative IDU; (…), borrowing among HIV-positive IDU; (•), intake visits; (▪), follow-up visits.

Trends may be explained by changes in the study population, as described previously [7]. The most important changes were an increase in age and duration of injecting, and a decrease in German or Western European nationality, current prostitution and current amphetamine injecting. These changes were not responsible for the observed trends; if these covariates were added in the basic models (Table 1), the linear and quadratic effect estimates of year did not substantially change and remained statistically significant. Trends were also determined in subgroups based on strata of the sociodemographic variables to check generalizability of trends; no important deviations of the overall patterns were observed (data not shown).

Determinants of borrowing and lending

Borrowing and lending appeared to be strongly interrelated within study visits: HIV-negative IDU who reported borrowing were 14.6 times more likely to lend, and amongst HIV-positive IDU this odds ratio was 9.4 (P < 0.0001, using GEE with a logistic link function). Table 2 shows the variables (out of those described in Methods) that were significantly related to either borrowing or lending, either amongst HIV-negative or positive IDU (adjusted for the variables in the basic model). The proportion of the study group expressing a determinant in recent years (1996–1997) is also presented.

Table 2
Table 2:
Determinants of borrowing and lending using 6645 cohort visits of 879 injecting drug users in the Amsterdam cohort study, 1986–1997.

Further multivariate analyses were performed because several of the determinants were interrelated. Amongst HIV-negative IDU, independent determinants of borrowing were heroin use (+4.5%), tranquillizer use (less than daily, +2.9%; daily, +4.3%; reference, no use), multiple needle use (+7.8%), injecting with others (+11.9%), and irregular use of needle-exchange programmes (+9.0%; reference, 100% exchange use).

Independent predictors for lending amongst HIV-negative IDU were multiple needle use (+3.1%), injecting with others (+5.7%), frequent injecting (+4.2%) and irregular use of needle-exchange programmes (+4.5%; reference, 100% exchange use). Independent determinants of borrowing amongst HIV-positive IDU were a subset of those identified amongst HIV-negative IDU: tranquillizer use (less than daily, +4.2%; reference, no use), multiple needle use (+6.8%), injecting with others (+4.7%), and irregular use of needle-exchange programmes (+6.9%; reference, 100% exchange use). With regard to lending amongst HIV-positive IDU, both multiple needle use (+2.6%) and irregular use of needle exchange programmes (+3.5%; reference, 100% exchange use) remained significant.

As determinants may change over time, interactions between the independent determinants and calendar year (linear and quadratic) were checked, and the multivariate analyses were repeated only using data from 1994 onwards. With one exception, none of the interaction terms was significant and the effect estimates did not change substantially. Only with regard to borrowing amongst HIV-negative IDU was there a significant interaction between frequency of attending needle-exchange programmes and calendar year. Similar levels of borrowing (about 50%) were reported in 1986–1987 among non- and irregular needle-exchange users, compared with only 27% among 100% exchangers (Fig. 3). However, the decline in borrowing among nonexchangers was stronger than among 100% exchangers, resulting in similar levels of borrowing from 1992 onwards, to a low of 13% in 1996–1997. Amongst irregular exchangers, borrowing also decreased initially but relapse occurred since 1990–1991, with borrowing reaching a level of 42% in 1996–1997. With regard to lending amongst HIV-negative IDU, and borrowing and lending amongst HIV-positive IDU, similar but less pronounced and non-significant interaction patterns were observed (data not shown). There was no difference in trends in borrowing and lending in subgroups based on frequency of attending methadone programmes.

Fig. 3
Fig. 3:
. Trends in borrowing among HIV-negative injecting drug users stratified by percentage of new needles obtained via needle-exchange programmes.

Discussion

The main finding of the present study was that, after a large initial decline in injecting risk behaviours (1986–1991), the rate of behavioural change significantly decreased and that no substantial further risk reduction occurred through 1997. This holds for borrowing, lending and multiple needle use amongst HIV-negative participants in the Amsterdam cohort, and for lending and multiple needle use amongst HIV-positive participants. Accordingly, the HIV incidence among current IDU initially declined from 8% in 1986 to 4% in 1991, and thereafter remained stable through 1997. Only the decline in the frequency of injecting appeared ongoing. At intake visits, borrowing and lending tended to increase from 1992 onwards, although not significantly; the statistical power was too low to provide evidence for relapse. These trends should be carefully monitored in the future, especially due to recent availability of potent antiretroviral combination treatments, which may reduce motivation for safe injection practices.

A previous study within our cohort showed that underreporting of risk behaviour may be a problem at follow-up visits in particular [17], because participants are repeatedly counselled and may become more reluctant to admit the same risks at consecutive visits. Indeed, reported risk at follow-up visits was much lower. However, both at intake and follow-up visits, injecting risk behaviour was stable (or non-significantly increased) after 1991 (Fig. 2). Another point of concern may be the crude measurement of outcome measures: there may be an ongoing decrease in frequency of borrowing or lending, a decrease in the number of persons borrowed from or lent to, or a decrease in risky mixing (e.g., HIV-negative IDU stop borrowing from HIV-positive IDU). However, available cohort data (1989 onwards) indicated a stable frequency of borrowing amongst HIV-negative IDU, and no major changes in the type of person(s) borrowed from (data not shown). Moreover, another study in our cohort showed that borrowing was mainly risky, as only 36% borrowed exclusively from IDU known to be HIV-negative, and used equipment was almost never adequately disinfected (unpublished data); this is corroborated by the high and stable HIV incidence in recent years.

For several reasons we think there is some minimum level of injecting risk that is difficult to prevent. First, harm-reduction efforts in Amsterdam have already included a wide variety of extensive HIV prevention measures for many years (see Introduction). Implementation of new interventions may only have a small additional effect, and cost-effectiveness is expected to be low. Second, another study within our cohort showed that borrowing of needles is mainly a deliberate behaviour (opposed to accidental behaviour), and access to injecting equipment and withdrawal symptoms seem to play a minor role (unpublished data). In addition, the present study indicated that psychopathology (tranquillizer use) and social factors (injecting with others) determined borrowing, these factors being hard to influence. In general, the lifestyles of IDU are often hectic and disorganized, and for some IDU obtaining and using drugs have a higher priority than preventing viral infectious diseases.

Therefore, an alternative intervention, which may be more cost-effective, is prevention of injecting and promotion of cessation of injecting [18]. This will not only reduce spread of HIV infection, but may also reduce the incidence of other highly prevalent diseases and causes of death specific for IDU (e.g., abscess, endocarditis, sepsis, drug overdose). Especially as the prevalence of hepatitis B and C infection can be over 50% within 2 years of onset of injecting [19], prevention of injecting seems warranted because it does not seem feasible to promote safe injecting practices among drug users who have not (yet) started injecting. However, previous studies have suggested that prevention of injecting may be difficult [20,21], so we started a study among young IDU and non-IDU to elucidate this. In Amsterdam, not many young drug users appear to initiate injecting, as indicated by the ageing of the population of current IDU [22], but if they do start there is a large potential for further spread of HIV because young IDU in particular are at high risk of HIV infection [23]. Although a nationwide campaign has started to promote the change from injecting to non-injecting use of heroin base (‘chasing the dragon’), no specific interventions exist to target prevention of initiation of injecting.

It will take a considerable amount of time before HIV prevalence substantially decreases, mainly because of this risk residue and the perceived difficulties for a further reduction. Therefore, prevention measures will have to be maintained for many years. Relatively low levels of risk behaviour still give rise to a high HIV incidence (3–4% per year in the cohort) due to the high and stable background HIV prevalence in Amsterdam of about 30% in the previous 12 years. Although this high HIV incidence almost balances mortality due to AIDS [24], HIV prevalence is expected to decline slowly because the number of pre-AIDS and AIDS deaths combined is exceeding the number of seroconversions [25]. However, as initiation and cessation of drug injecting, migration, and effective treatment of HIV infection also determine HIV prevalence trends, future predictions remain uncertain.

Although a substantial further risk reduction appears unlikely, this study provides some indications of how effectiveness of existing prevention measures may be improved.

First, in addition to other Amsterdam studies [5–7,13], there is support for an active HIV testing and counselling policy (compared with passive availability of services). Levels of injecting risk of all four outcome variables are much lower at follow-up visits than at intake visits, and HIV-positive IDU reduce risk behaviour to a larger extent than HIV-negative IDU (Fig. 1). It is unlikely that this behavioural change can be fully attributed to an increase in socially desirable answering. In particular, relative underreporting of borrowing at follow-up visits was estimated to be 42% (unpublished data), which can explain only a fraction (< 30%) of the large risk reduction from visit 1 to 2. Multiple needle use and frequent injecting are expected to be less subject to socially desirable answering than borrowing. Another compelling reason to promote testing to drug users is the recent availability of antiretroviral combination treatments.

Second, several high-risk groups have been identified to which prevention can be targeted (Table 2), such as young IDU, non-injecting heroin and cocaine users, frequent tranquillizer users (which may be a proxy for psychopathology), and IDU who use needles more than once and who inject with others. Because multiple needle use may be a causal factor for borrowing and lending, and is amenable for intervention itself, the prevention message to use new injecting equipment for every injection (‘one set, one shot’) should be reinforced, together with stressing the importance of destruction/safe disposal of used needles/syringes so that they cannot be used anymore.

This study points towards several general methodological issues with regard to evaluation of prevention measures in observational studies. Because it can take a considerable amount of time before interventions have their maximum effect (here up to 6 years), long-term evaluation is required, preferably by use of a serial cross-sectional study design [3]. Risk reduction rather than risk elimination appears to be a realistic goal. Because in high HIV prevalence areas a decline in prevalence can take a long time, and because the HIV incidence may stabilize at relatively high rates, self-reported risk behaviour should (also) be used as an outcome of interest.

There are major methodological drawbacks in studies evaluating specific interventions using observational (cohort) data, and causal relationships between intervention and outcome cannot be established [3,26]. In particular, in a comparison of attenders and non-attenders of specific programmes, there may be contamination of the control group (e.g., needle exchangers may exchange for other IDU who do not attend an exchange programme, attenders may disseminate safe injecting guidelines) and selection bias (e.g., programmes that attract high or low-risk IDU). This study provides further evidence of such biases. For instance, in the late 1980s, obtaining all new needles via the exchange programme was associated with lower levels of borrowing, whereas in recent years, 100 and 0% exchangers have had the same levels of borrowing (Fig. 3). Instead of concluding that needle exchange was more effective in early than in later years, it is more likely that in early years these programmes attracted IDU who were highly motivated to reduce HIV risk, whereas later on exchanging become just one of several ways to obtain sterile injecting equipment (e.g., pharmacies, shops in the red light district). In a previous study, we found a significantly different relationship between exchanging and HIV incidence over time; frequent use of the needle exchange was more protective in early years [4]. The finding that exchanging is strongly associated with high frequency of injecting also indicates self-selection bias. A complicating factor is that the presented trends do not necessarily reflect individual behavioural changes. The increase in borrowing amongst irregular exchangers may indicate that a group of IDU of decreasing size remains who have not (yet) adopted stable injecting habits (the others have started exchanging 0 and 100%), leaving those IDU with highest risk.

In conclusion, IDU can substantially decrease injecting risk behaviour, but there is a residual risk that seems hard to prevent. As the HIV epidemic is only slowly dwindling in Amsterdam, prevention measures should be maintained. Meanwhile, long-term measures targeted at reducing injecting behaviour seem to be warranted to prevent problems in the future. Monitoring trends in the numbers of drug users and IDU, HIV prevalence, and injecting risk behaviour are important to guide prevention policy.

Acknowledgements

The authors thank J. Bax, A. Snuverink, H. Fennema and I. Spijkerman for data collection, and H.J.A. van Haastrecht for methodological advice. This study was performed as part of the Amsterdam cohort studies on AIDS, a collaboration between the Municipal Health Service, the Academic Medical Centre and the Central Laboratory of the Netherlands Red Cross Blood Transfusion Service, Amsterdam, The Netherlands.

References

1. van Ameijden EJC, van den Hoek JAR, Mientjes GHC, Coutinho RA: A longitudinal study on the incidence and transmission patterns of HIV, HBV and HCV infection among drug users in Amsterdam. Eur J Epidemiol 1993, 9:255–262.
2. Des Jarlais DC, Friedman SR, Choopanya K, Vanichseni S, Ward TP: International epidemiology of HIV and AIDS among injecting drug users. AIDS 1992, 6:1053–1068.
3. van Ameijden EJC, Watters JK, van den Hoek JAR, Coutinho RA: Interventions among injecting drug users: do they work?AIDS 1995, 9(suppl A):S75–S84.
4. van Ameijden EJC, van den Hoek JAR, van Haastrecht HJA, Coutinho RA: The harm reduction approach and risk factors for HIV seroconversion in injecting drug users, Amsterdam. Am J Epidemiol 1992, 136:236–243.
5. van Ameijden EJ, van den Hoek JAR, Coutinho RA: Injecting risk behavior among drug users in Amsterdam, 1986 to 1992, and its relationship to AIDS prevention programs. Am J Public Health 1994, 84:275–281.
6. van Ameijden EJ, van den Hoek JAR, van Haastrecht HJ, Coutinho RA: Trends in sexual behaviour and the incidence of sexually transmitted diseases and HIV among drug using prostitutes, Amsterdam 1986–1992. AIDS 1994, 8:213–221.
7. van Ameijden EJC, van den Hoek JAR, Coutinho RA: Large declines in sexual risk behavior with non-commercial partners among heterosexual injection drug users in Amsterdam, 1989–1995. Am J Epidemiol 1996, 144:772–781.
8. van Brussel GHA, Buster MCA, van der Woude DH: Dovend vuur: Jaarbericht van de drugsafdeling 1994/1995. Amsterdam: Municipal Health Service (GG & GD), Sector GGZ; 1996.
9. Brettle RP: HIV and harm reduction for injecting drug users. AIDS 1991, 5:125–136.
10. Des Jarlais D, Friedman SR, Ward TP: Harm reduction: a public health response to the AIDS epidemic among injecting drug users. Annu Rev Public Health 1993, 14:413–450.
11. Buning EC, van Brussel GHA, van Santen G: The methadone by bus project in Amsterdam. Br J Addict 1990, 85:1247–1250.
12. Langendam MW, van Haastrecht HJA, van Brussel GHA, van den Hoek JAR, Coutinho RA, van Ameijden EJC: Differentiation in the Amsterdam methadone dispensing circuit: determinants of methadone dose and location of dispensation. Addiction 1998, 93:61–72.
13. Langendam MW, van Ameijden EJC, van den Hoek JAR: HIV-testen en -counselen in Amsterdam: kenmerken van niet eerder op HIV geteste injecterende druggebruikers. Tijdschr Soc Gezondheidsz 1995, 73:354–359.
14. van den Hoek JAR, Coutinho RA, van Haastrecht HJA, van Zadelhoff AW, Goudsmit J: Prevalence and risk factors of HIV infections among drug users and drug-using prostitutes in Amsterdam. AIDS 1988, 2:55–60.
15. Liang KY, Zeger SL: Longitudinal data analysis using generalized linear models. Biometrika 1986, 73:13–22.
16. Zeger SL, Liang KY: Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986, 42:121–130.
17. Fennema JSA, van Ameijden EJC, Coutinho RA, van den Hoek JAR: Validity of self-reported sexually transmitted diseases in a cohort of drug-using prostitutes in Amsterdam; trends from 1986 to 1992. Int J Epidemiol 1995, 24:1034–1041.
18. Stimson GV: The global diffusion of injecting drug use: implications for HIV infection. Bull Narc 1993, 45:3–17.
19. Garfein RS, Vlahov D, Galai N, Doherty MC, Nelson KE: Viral infections in short-term injection drug users: the prevalence of the hepatitis C, hepatitis B, human immunodeficiency, and human T-lymphotropic viruses. Am J Public Health 1996, 86:655–661.
20. Des Jarlais DC, Casriel C, Friedman SR, et al.: AIDS and the transition to illicit drug injection: results of a randomized trial prevention program. Br J Addict 1992, 87:493–498.
21. van Ameijden EJC, van den Hoek JAR, Hartgers C, Coutinho RA: Risk factors for the transition from non-injection to injection drug use and accompanying AIDS risk behavior in a cohort of drug users. Am J Epidemiol 1994, 139:1153–1163.
22. Fennema JSA: HIV infection among drug users and the potential for heterosexual spread. PhD Thesis, University of Amsterdam, 1997.
23. Fennema JSA, van Ameijden EJC, van den Hoek JAR, Coutinho RA: The impact of young and recent-onset injection drug users on the HIV epidemic in Amsterdam, 1986–1995. Addiction 1997, 92:1457–1465.
24. van Haastrecht HJA, van Ameijden EJC, van den Hoek JAR, Mientjes GHC, Bax JS, Coutinho RA: Predictors of mortality in the Amsterdam cohort of human immunodeficiency virus (HIV)-positive and HIV-negative drug users. Am J Epidemiol 1996, 143:380–391.
25. van Haastrecht HJA, Bindels PJE, van den Hoek JAR, Coutinho RA: Estimating the size of the HIV epidemic among injecting drug users in Amsterdam. Eur J Epidemiol 1997, 13:161–165.
26. Booth RE, Watters JK: How effective are risk-reduction interventions targeting injecting drug users?AIDS 1994, 8:1515–1524.
Keywords:

HIV infection/transmission; HIV infection/prevention and control; Amsterdam; The Netherlands; risk factors; injecting drug use; surveillance; cohort studies; needle sharing

© Lippincott-Raven Publishers.