HIV prevalence and risk behaviour among female injecting drug users in London, 1990 to 1996

Objective: To describe trends in HIV prevalence among female injecting drug users (IDU) in London between 1990 and 1996.

Design: HIV prevalence and risk behaviour were measured yearly between 1990 and 1993, and in 1996, in point prevalence HIV surveys of IDU recruited from both drug-treatment and community-based settings within Greater London. Sample sizes were 173 in 1990, 111 in 1991, 128 in 1992, 146 in 1993 and 200 in 1996.

Methods: Each survey used structured questionnaires and common sampling and interview strategies. Oral fluid specimens were collected for testing for antibodies to HIV (anti-HIV). Multiple logistic regression was used to assess the trend in HIV prevalence.

Results: The percentage of female IDU testing positive for antibodies to HIV showed a marked decline over the study period, from 15.0% in 1990 to 1.0% in 1996 (P<0.001). This trend was independent of all other variables examined. Each year, higher HIV prevalences were found among IDU recruited from community settings compared with treatment agencies.

Conclusions: These results concur with those of IDU recruited from treatment sites, although the yearly estimates in this study are higher. London benefits from low prevalence of HIV infection among IDU, coupled with behaviour change facilitated by early intervention. Continued surveillance of injectors recruited from both community and treatment settings is necessary in order properly to assess HIV prevalence among IDU.

Introduction

HIV infection among injecting drug users (IDU) is a major public health problem in many parts of the world [¹]. However, in the UK the anticipated major spread of HIV-1 infection among IDU did not occur. HIV prevalence among injectors in most parts of the UK has remained at low levels throughout the 1990s, with London prevalences being slightly higher [^2,3].

Few studies have specifically examined female injectors and their risk of HIV infection, and in most studies of IDU, females are under-represented. Although there may be lower numbers of female IDU than males [^4-6], they may be at greater potential risk of HIV infection than male IDU. With regard to injection-related risk, Barnard (1993) has noted that women tend to have dyadic sharing relationships (they are more likely to share exclusively with their sexual partners) but are secondarily linked to extended networks (their partners are more likely to borrow from strangers) [⁷]. Women who share used injecting equipment often inject second or may be injected by their male partners once equipment has been used [⁸].

Methods

The methodology of these surveys has been described elsewhere [^3,9,10]. Briefly, female IDU were recruited from drug treatment sites (e.g. drug dependency clinics) and non-treatment community settings (e.g. street settings) in London between 1990 and 1993, and in 1996. Injectors were eligible for participation if they had injected drugs in the 2 months preceding interview, and if they had not previously been interviewed for the study in the current calendar year (validated by post-interview demographic matching). Community respondents were recruited through social network sampling by trained indigenous interviewers [^9,11]. Interviews were verified by randomized spot checks [¹²]. The number and characteristics of refusals were not collected. A questionnaire, designed by a World Health Organization working group, was used for the 1990 to 1993 surveys, with some modifications made for the 1996 survey. The role of sexual transmission of HIV infection among IDU may be less important than the parenteral route [^13,14] and was not investigated here because of revision in the sexual questions in 1996.

Oral fluid specimens were collected using the Salivette device (Starstedt Ltd, Leicester, UK) with informed consent [^15,16]. Specimens were screened for anti-HIV by HIV-1/2 GACELISA (Murex Biotech Ltd, Dartford, UK), with independent confirmatory testing for positive specimens. From 1992, specimens were also tested for antibodies to hepatitis B virus (HBV) core antigen (anti-HBc) by GACRIA. Test results were not reported back to participants.

Multiple logistic regression was performed in a hierarchical manner, and all independent factors were subsequently controlled for when any factor of interest was considered. Statistical significance was assessed using the likelihood ratio statistic [¹⁷]. Factors with effects that were close to statistical significance and that were considered important epidemiologically were included in the final regression model. The statistical software package Stata 5.0 (Stata Corp, College Station, Texas, USA) was used for all analyses.

Results

In 1990, 173 female IDU were tested for anti-HIV, as were 111 in 1991, 128 in 1992, 146 in 1993 and 200 in 1996. Of all respondents, 4% (34/833) refused to provide a specimen, and 41 specimens proved inadequate for analysis.

Each year, the majority of female IDU lived in central-west and north-west London, and three quarters were recruited from community settings. Most (89.3%) were white, and 78.1% were born in the United Kingdom. An average of 76% of injectors each year reported having received treatment related to their drug use. Of those recruited from community sites, 30.4% (169/556) reported that they had never received drug treatment.

Median age at interview increased with year of interview, rising from 26 years in 1990 to 29 years in 1993 and 30 years in 1996. Respondents with drug treatment experience were older than respondents with no treatment experience (medians 29.5 and 27.0 years, respectively).

The distribution of age at first injection was similar in every year (median ≊19 years). Duration of injecting, following the trend of increasing age of interviewee, showed an increasing trend with year of interview, from a median of 7 years (1990) to one of 10 years (1996). The proportion of respondents injecting before 1986 decreased, from 71.2% (1990) to 48.7% (1996). The proportion of respondents reporting daily injecting fell from 70.0% (1990) to an average of 52% between 1992 and 1996. In all years, at least 80% of respondents had injected heroin in the previous 6 months.

Around 25-35% of respondents had injected with used needles and syringes (‚sharing‚) in the last 6 months. Among these sharers, the proportion that injected only with used equipment given by their regular sexual partner increased from 38.7% (1990) to 67.2% (1996). Similarly, the proportion who reported sharing with only one person in the last 6 months increased from 51.7% (1990) to 82.0% (1996).

The percentage of female IDU whose oral fluid contained anti-HIV declined markedly over the study period, from 15.0% in 1990 to 1.0% in 1996 (c² trend 32.48; P<0.001; Table 1). The prevalence of anti-HBc decreased from 46.5% in 1992 to 41.1% in 1993 and 18.6% in 1996, which equates to an average decline of 29% per year [coefficient for linear trend (e^b) 0.710; 95% confidence interval (CI) 0.630-0.800; P<0.0001]. There was no discernible trend in HIV prevalence among those who were anti-HBc positive (6.8% in 1992, 6.7% in 1993 and 5.4% in 1996), but there was a marked decrease in HIV prevalence from 10.3% (1992) to 5.8% (1993) and 0.0% (1996) among those who were anti-HBc negative (c² trend 17.67; P<0.001).

Table 1

Women recruited in the community showed consistently higher prevalences than those recruited from treatment centres, but both groups showed a marked decline in HIV prevalence over the time of the survey, from 15.8% (1990) to 1.2% (1996) for community-recruited respondents, and from 12.5% (1990) to 0.0% (1996) for treatment-recruited respondents.

In a multivariate model, factors associated with HIV prevalence were year of interview, country of birth, treatment history, year of first injection and frequency of injecting. After adjustment for these factors, the relationship between HIV prevalence and year of interview remained highly significant (Table 1). The results suggest that female IDU interviewed in 1990 had a 15 times higher odds of HIV infection than their 1996 counterparts. The adjusted prevalence of HIV declined by an average of 31% per year (e^b 0.693; 95% CI 0.585-0.820; P<0.0001).

In earlier survey years, respondents born outside the UK had a higher prevalence of HIV than those born within the UK [odds ratio (OR) 3.077; P<0.001 after adjustment; Table 1].

Female injectors who reported never receiving treatment for their drug use had higher prevalences each year and higher odds, of HIV infection than those currently or previously receiving treatment (OR 2.019; P=0.038 after adjustment; Table 1). There was weak evidence for a relationship between age and HIV prevalence (P= 0.086 after adjustment; Table 1).

Women who commenced injecting before 1986 had nearly twice the odds of HIV infection than those starting injecting from 1986 onwards (P=0.055 after adjustment; Table 1). Similarly, women who reported injecting daily were almost twice as likely to be HIV positive as those who injected less frequently (adjusted OR 1.961; P=0.028; Table 1).

Discussion

There was a strong trend of declining HIV prevalence over the survey years, from 15.0% in 1990 to 1.0% in 1996. This trend was highly statistically significant and was not weakened after adjustment for all other independent factors available for analysis (year of interview, country of birth, treatment history, year first injected and frequency of injecting). There was a similarly highly significant decline in anti-HBc prevalence. These findings are consistent with results from routine surveys of IDU attending treatment services in London [^2,18].

This study has several limitations. How representative the samples are of the wider population of IDU is unknown, although sampling strategies may have helped to minimize biases [¹¹]. The increasing median age at interview in the survey years might suggest an ‚ageing cohort‚ effect, although we also observed a similar trend for female opiate injectors presenting to treatment services in North London in the same time period (unpublished data).

Back-calculation estimates suggest that HIV incidence among UK injectors peaked in 1985 and then declined [^19,20], and other evidence indicates that in the 1990s the number of new infections fell [²¹]. The differential effects of mortality, behaviour change, cessation of drug-injecting behaviour and the migration of IDU from drug treatment to specialist HIV-treatment services would all act to reduce the prevalence of HIV among female injectors found in this study.

The vast majority of respondents were highly selective about who they accepted previously used needles and syringes from in the preceding 6 months. Increasing numbers shared only with one other person, and decreasing numbers reported injecting on a daily basis over the survey period. These changes in injecting risk behaviour are consistent with findings elsewhere which suggest that since 1986 IDU have changed their injecting behaviour in response to the risk of HIV, and that harm reduction interventions have had some effect in facilitating behaviour change [^9,22,23]. However, recent findings suggest that IDU share many paraphernalia in addition to needles and syringes [²⁴]. Such findings have important implications with regard to the future incidence of HIV infection and other blood-borne diseases, especially hepatitis C. Prevention and harm reduction interventions need to focus increasingly on discouraging indirect sharing and on the prevention of transitions towards injecting.

In this study, lifetime lack of experience of treatment for drug use was related to higher risk of HIV infection. Current routine HIV surveillance of IDU in England and Wales relies on the recruitment of IDU from treatment sites only, and survey results suggest that such a strategy underestimates the true prevalence of HIV among IDU [^2,3,25]. Future surveillance needs to accommodate the simultaneous recruitment of samples of IDU from various treatment and community settings in order to assess the prevalence of HIV and other infections among IDU properly.

Acknowledgements

We would like to thank our fieldworkers, staff at collaborating fieldwork agencies and Alison Mahoney, Julie Newham and Kim Lewis at the VRD for laboratory support. We thank Tim Rhodes for his useful comments on an earlier draft of this paper. We would also like to thank the women who gave up their time to be interviewed for these studies and colleagues previously involved in the studies.