Sex Differences in Risk Factors for HIV Seroconversion Among Injection Drug Users: A 10-Year Perspective

Background  Injection drug use directly or indirectly accounts for nearly half the annual human immunodeficiency virus (HIV) infections in the United States. Prospective studies that investigate both sexual and parenteral HIV risks among injection drug users (IDUs) are needed. We studied factors for HIV seroconversion among male and female IDUs in Baltimore, Md.

Methods  The HIV-negative IDUs (1447 male and 427 female) were recruited into a prospective study from 1988 to 1989 or in 1994. Participants underwent semiannual HIV tests and surveys through December 1998. Poisson regression was used to identify risk factors for HIV seroconversion, stratified by sex. Behaviors were treated as time-dependent covariates that varied at each semiannual period.

Results  Subjects were primarily African American (91%), and median age at enrollment was 35 years. Incidence of HIV was 3.14 per 100 person years (95% confidence interval, 2.78-3.53) and did not significantly differ by sex. Younger age independently predicted HIV seroconversion for both men and women. Among men, factors that independently predicted HIV seroconversion were the following: less than a high school education, recent needle sharing with multiple partners, daily injection, and shooting-gallery attendance. The incidence of HIV was double for men recently engaging in homosexual activity and cocaine injection. Among women, the incidence of HIV was more than double for those recently reporting sexually transmitted diseases.

Conclusions  The incidence of HIV remained high among IDUs in Baltimore over the past decade. Risk factors for HIV seroconversion differed markedly by sex. Predominant risks among men included needle sharing and homosexual activity; among women, factors consistent with high-risk heterosexual activity were more significant than drug-related risks. Human immunodeficiency virus interventions aimed at IDUs should be sex-specific and incorporate sexual risks.

SINCE THE human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) epidemic began in the early 1980s, the proportion of HIV/AIDS cases attributed to men having sex with men has significantly declined, with subsequent increases among heterosexuals and injection drug users (IDUs).^1-3 Taking into account direct transmission of HIV infection through sharing contaminated injection equipment and indirect transmission to sexual partners and offspring, injection drug use accounts for nearly half the annual total number of HIV cases in the United States.³

The changing dynamic of the HIV/AIDS epidemic has had a significant impact on women and minorities. Women accounted for 23% of all reported adult AIDS cases in the United States in 1999,¹ which represents a doubling over the past decade. Injection drug use accounts for 42% of all reported AIDS cases among women and adolescent girls.¹ The proportion of HIV/AIDS cases diagnosed among African Americans and Hispanics has significantly increased.^1,2 These ethnic groups accounted for 55% of the cumulative total number of AIDS cases and 77% of AIDS cases in women and girls reported through December 1999.¹

Rates of IDU-associated AIDS cases have been consistently higher in the northeastern region of the United States, where a large population of IDUs (primarily heroin injectors who began injecting in the 1960s) already existed prior to the HIV epidemic.^4,5 Compared with the United States overall, injection drug use accounts for a significantly higher proportion of AIDS cases in Maryland (54% vs 36%, respectively, in 1999).^1,6 In Baltimore, Md, the prevalence of HIV infection in a community-based cohort of IDUs at enrollment in 1988 was 24%,⁷ which has given rise to persistently elevated HIV incidence rates.

Because early studies of IDUs suggested that most HIV infections were due to needle sharing,^8,9 it has generally been assumed that sexual transmission was negligible or was overshadowed by parenteral routes.^10-12 Although some studies suggested an important role for heterosexual transmission,^13-15 few published reports have prospectively examined both drug-related and sexual risk factors for HIV seroconversion among male and female IDUs.

We examined both drug-related and sexual risk factors for HIV transmission in a cohort of more than 1800 IDUs for a 10-year period. A large study sample and extended follow-up enabled us to study potential sex differences in HIV risk factors, taking into account changing behaviors over time. Such data are critical for developing appropriate interventions to reduce HIV incidence among marginalized IDU populations, including women and ethnic minorities.

From 1988 to 1989, a total of 2946 persons who were at least 18 years old were recruited into a longitudinal study of the natural history of HIV infection in IDUs in Baltimore. Potential participants were recruited through extensive community outreach, as described elsewhere.^7,16 All participants were free of an AIDS-defining illness at enrollment and had a history of illicit injection drug use within the previous 10 years. After providing informed consent (approved by the Committee on Human Research, Johns Hopkins School of Hygiene and Public Health, Baltimore), participants underwent interviewer-administered questionnaires, physical examinations, and blood specimen collection.

Of 2946 subjects at enrollment, 708 (24%) were HIV-seropositive and 2238 (76%) were seronegative. Of the HIV-negative subjects, 1560 (70%) returned for at least 1 follow-up visit, at which time they were invited to enroll in a study of risk factors for HIV infection. A detailed analysis comparing participants who returned with those who did not has been reported.¹⁷ To replenish cohort numbers because of losses to follow-up, HIV seroconversion (n = 218), and deaths (n = 261), 363 HIV-seronegative current IDUs were additionally recruited in 1994, 338 (93%) of whom returned for at least 1 follow-up visit.

To be eligible for analysis, participants were required to be HIV-seronegative at enrollment and to have returned for at least 1 follow-up visit. Human immunodeficiency virus seroconverters for whom there was more than a 3-year interval between their last HIV-negative and first HIV-positive test were also excluded (24 of 301 seroconverters). Therefore, a total of 1874 persons were included in the present analysis.

Semiannual interviews collected data on drug use history, sociodemographics, and drug use and sexual behavior within the last 6 months. At each study visit, HIV seroconverters were identified by commercial HIV antibody enzyme-linked immunosorbent assay. Repeatedly, seropositive specimens were confirmed by Western blot. All participants received their HIV test results, including pretest and posttest HIV counseling, administered by trained interviewers.

Incidence density of HIV infection was calculated using person-time methods, taking into account varying lengths of follow-up between December 1988 and December 1998. The date of seroconversion was considered to be the midpoint between the last seronegative and first documented seropositive HIV test result. Incidence rates were examined by exposure variables of interest (eg, sex and age). All behavioral data were obtained through self-reported survey data with the exception of data pertaining to needle exchange program (NEP) attendance, which was obtained from the Baltimore NEP using a reverse identifier based on portions of select information (eg, last 4 digits of the participant's social security number).

Poisson regression was used to calculate relative incidence for exposure variables per semiannual period, stratified by sex. The mean was proportional to semiannual periods and the antilog of the linear combination of covariates, whose weights represented the regression coefficients.¹⁸ Behavioral data and variables reflecting use of specific services within the last 6 months (eg, attendance at NEP and methadone maintenance programs) were treated as time-dependent covariates that were subsequently updated for each semiannual period. Since the Baltimore NEP was introduced in August 1994, variables pertaining to NEP attendance were coded as 0 for all study visits prior to this date. For each exposure variable, potential sex differences were examined by assessing interaction terms. In cases of missing data, values for the preceding semiannual visit were assumed reliable for a maximum of 1 year, after which time values were treated as missing.

To determine the independent effect of covariates of interest on the outcome of HIV seroconversion, multivariate Poisson models were generated. To take into account temporal trends in HIV seroconversion, we adjusted for calendar year. Goodness of fit was assessed by assessing the deviance of a model from the saturated model, based on differences in degrees of freedom. Statistical significance was determined by comparing nested models using the likelihood ratio test. The relative contribution of each exposure variable, adjusting for the simultaneous effects of all covariates in the final models, was expressed in terms of adjusted relative incidence rates (ie, rate ratios).

Of 1874 eligible participants, 1447 (77%) were male and 427 (23%) were female. Most subjects (91%) were African American. The median age and duration of injection drug use at enrollment were 35 years (interquartile range [IQR], 30-40 years) and 14 years (IQR, 6-24 years), respectively. In comparing cohort members who returned for at least 1 visit from 1988 to 1989 with subjects additionally recruited in 1994 there were significant differences in age at enrollment (34.7 vs 37.1 years; P < .001) and in the proportions who were male (80.3% vs 63.0%; P < .001) and African American (89.78% vs 95.27%; P = .002).

In the 6 months prior to enrollment, 91% of our subjects reported having injected drugs. Of a total of 17 021 study visits, the median number of visits was 7.8 per person (IQR, 3-12); the median interval between study visits was 6.5 months (IQR, 5.5-6.9 months). Among a total of 8826.45 person years (PYs) of observation, 277 documented HIV seroconversions occurred for a global HIV incidence rate of 3.14 per 100 PYs (95% confidence interval [CI], 2.78-3.53). Human immunodeficiency virus incidence declined significantly over time from 4.58 per 100 PYs in 1988 to 1.29 per 100 PYs in 1998; this translated to a 10% decrease per year (rate ratio, 0.90 per year). Despite our earlier finding that female IDUs had significantly higher HIV incidence compared with male IDUs,¹⁷ during the extended 10-year follow-up period HIV incidence remained only slightly higher among women (3.23 per 100 PYs) compared with men (3.11 per 100 PYs). The rate ratio for HIV seroconversion among women vs men was 1.04 (95% CI, 0.8-1.4).

Sex-specific HIV incidence rates associated with specific demographic and behavioral characteristics are summarized in Table 1. Higher rates of HIV seroconversion occurred among both men and women aged 30 years or younger at enrollment. In addition, significantly higher HIV incidence rates occurred among those who injected cocaine alone or in combination with other drugs, injected daily or more, shared injection paraphernalia (eg, cotton, cookers, and water), had multiple needle-sharing partners, and had a sexually transmitted disease (STD). Among men, annual prevalence rates for gonorrhea and syphilis varied between 0.52% and 4.63% and 0.30% and 1.72%, respectively; for female IDUs these rates were generally higher, varying annually from 1.47% and 7.14% and 0.61% and 4.50%, respectively. Human immunodeficiency virus incidence for persons who injected cocaine alone was 7.20 per 100 PYs among women and 5.05 per 100 PYs among men.

To assess whether HIV incidence and specific risk factors differed for male and female IDUs, Poisson regression models were constructed, stratifying by sex (Table 2). Among men, higher annual income and higher levels of education were inversely associated with HIV seroconversion, but this was not the case for women. Homelessness, number of needle sharing partners, and shooting-gallery attendance (locations where IDUs gather to buy or rent syringes and/or inject drugs in groups) were associated with an increased risk of HIV seroconversion, but only among men. Recent enrollment in a methadone maintenance program was inversely associated with HIV seroconversion for women (rate ratio, 0.40; 95% CI, 0.19-0.84). Among women, HIV incidence was elevated for those who reported using coccaine; a similar association was not observed for men.

Sex differences were more striking in terms of sexual risks (Table 2). Men who reported recently engaging in homosexual activity were 4 times more likely to become infected with HIV (rate ratio, 4.04; 95% CI, 1.5-10.9). Among female IDUs, HIV incidence was more than double among those who reported recently having sex with another IDU (rate ratio, 2.27; 95% CI, 1.3-3.8). Among women, condom use was significantly associated with an increased risk of HIV seroconversion, which is not surprising because condom use is highly correlated with high-risk sexual behaviors (eg, sex trade). Rate ratios for male IDUs compared with women differed significantly for reporting sex with an injection drug–using partner, numbers of heterosexual partners, and homosexual activity.

After simultaneously taking into account multiple risk factors for HIV infection that were permitted to vary over time, younger age was the only risk factor that remained independently associated with the risk of HIV seroconversion for both men and women (Table 3). Injection drug users who were younger than 30 years at enrollment were more than twice as likely to experience HIV seroconversion. Among men, independent predictors of HIV seroconversion included sociodemographic, drug-related, and sexual risks. Having achieved at least a high school education was inversely associated with HIV seroconversion, whereas injecting daily or more, injecting cocaine, and attending shooting galleries were associated with an increased risk of HIV transmission. Male IDUs recently engaging in homosexual activity were more than twice as likely to seroconvert compared with those who reported no recent homosexual contact, whereas men reporting 1 or more heterosexual partners were almost half as likely to seroconvert compared with those reporting no recent heterosexual partners.

Among female IDUs, none of the drug-related variables associated with HIV incidence in univariate models retained statistical significance after taking into account other risk factors. After adjusting for the effects of age and time, female IDUs having a recent STD increased the risk of HIV seroconversion more than 2-fold (rate ratio, 2.52; 95% CI, 1.39-4.58).

Since our study sample comprised subjects who were recruited from 1988 to 1989 and in 1994, we repeated the above analyses, restricting to the former sample. Independent predictors of HIV seroconversion were identical and the magnitude of observed associations did not appreciably differ.

The major findings of this study were 2-fold: first, after accounting for parenteral risks related to HIV seroconversion among IDUs, sexual risks played a significant role, and second, HIV risk factors among IDUs differed markedly by sex. Whereas drug-related risk behaviors and homosexual activity were the most important predictors of HIV seroconversion among men, factors consistent with high-risk heterosexual activities were the main predictors among women. Because this study was based on longitudinal data from more than 1800 IDUs for a 10-year period, we were able to uncover associations that may not have been observed by others because of low statistical power or lack of a prospective design. Our results offer new directions for HIV prevention among highly marginalized injection drug–using populations that remain at high risk for infection.

The only common predictor of HIV seroconversion that we observed among male and female IDUs was younger age. Injection drug users aged 30 years or younger at enrollment were more than twice as likely to seroconvert compared with those aged 40 years or older. This is consistent with several reports that indicate that younger IDUs are more likely to engage in needle-sharing and other behaviors that place them at higher risk of acquiring HIV and hepatitis B or C viruses.^14,16,19-23 Compared with older drug users, younger IDUs tend to have initiated injection more recently; specific factors related to initiation into injection (eg, number of "trainers" who teach new initiates how to inject) may be associated with elevated HIV transmission risks.²³ Young drug users should be a focus for intensive interventions aimed at preventing substance abuse and reducing the spread of blood-borne pathogens.

Not unexpectedly, we observed a high risk of HIV infection associated with injection-related behaviors that are associated with needle-sharing behaviors (eg, cocaine injection and shooting-gallery attendance), which is consistent with early reports.^8,9 Injection-related risks were more pronounced among men, in whom cocaine injection and shooting-gallery attendance were independent risk factors for HIV seroconversion. The finding that cocaine injection is strongly associated with HIV seroconversion supports several previous studies.^9-11 Compared with heroin injectors, cocaine injectors lead more chaotic lives, inject and share needles more frequently, and are more likely to engage in the sex trade, elevating their risk of HIV infection.

Shooting galleries were identified as strong correlates of HIV infection earlier in the epidemic, especially in the Northeast.^4,15,24 Shooting galleries have exerted an independent effect on the risk of HIV seroconversion for a decade, which suggests that they may be useful venues for targeted interventions. A few studies have documented needle-sharing behaviors among shooting-gallery attendees^24,25; however, published reports of interventions in these settings are lacking.²⁶

We demonstrated only a modest protective effect of methadone maintenance and NEP attendance on HIV seroconversion, in contrast with earlier reports.^27-32 Incidence of HIV was 60% lower among men enrolled in methadone maintenance compared with men who were not, but this association did not persist after adjusting for other factors. A possible interpretation for our findings is that in the presence of sexual risks, measurable effects of interventions focusing on parenteral transmission will be underestimated. Low power may have also limited our ability to draw inferences, since 15% of the cohort members were enrolled in methadone maintenance at any given time, and NEP was introduced in Baltimore more than halfway through the follow-up period.

We observed that male IDUs who recently engaged in homosexual activity were more than twice as likely to seroconvert as those who did not. Among the 69 men who had sex with another man during the study period, HIV incidence was 10.2% per year. Despite 2 decades of HIV/AIDS research, men who have sex with men who also inject drugs remain a hidden population of which relatively little is known.^33,34 These men may experience a dual risk of HIV infection through the sharing of injection equipment and unprotected anal sex, both of which may be more common among men trading sex for money or drugs.^33,35 Being a gay male drug user has been closely associated with shooting-gallery attendance, needle sharing, and HIV positivity.^24,35-37 Our study confirms the elevated risk of HIV seroconversion among male IDUs who have sex with men and underscores the urgent need to develop effective interventions for this vulnerable subgroup.

Interestingly, male IDUs who reported not having a heterosexual partner were more likely to acquire HIV infection, a finding that was not explained by homosexual/bisexual activity. Similar associations have occurred elsewhere, which may suggest that high-risk injection practices are more common in the absence of a partner who can provide social support.^14,38 Lack of heterosexual partnership may also be a marker for more severe drug dependence.

In contrast, indicators of high-risk heterosexual activity were more significant than needle-sharing behaviors as independent predictors of HIV seroconversion among female IDUs. Incidence of HIV was double among women who reported an STD in the prior 6 months compared with those who did not. This supports earlier cross-sectional studies, most of which focused on male-to-female HIV transmission.^16,39-41 Apart from being an indicator of unprotected sex, several common STDs are cofactors of HIV transmission,^40,42,43 suggesting that NEPs and drug treatment programs, for example, should place greater emphasis on STD screening and treatment for IDUs. Interventions are needed to reduce sexual risks among women, eg, couple counseling or promoting woman-controlled barriers (including microbicides and female condoms). Prevention programs should take into account the fact that women tend to have greater overlap between their sex and drug networks than men.^23,44

We did not observe an independent effect of trading sex with HIV seropositivity after accounting for other sexual risks, which is unlike the observations in other studies.^23,45,46 A possible explanation is that a higher proportion of commercial sex acts may have been protected through condom use compared with the sex acts of noncommercial partners.

Interpretation of our study findings should take into account a number of limitations. In any cohort, differential losses to follow-up can bias associations of interest. During the study period, follow-up rates were excellent among participants who returned at least once, but those who did not return were significantly younger and more likely to have been homeless,¹⁹ suggesting that incidence could have been underestimated. However, compared with dropouts, subjects who returned were more likely to report frequent injection and needle sharing at enrollment.¹⁹ Therefore, it is difficult to predict whether our HIV incidence estimates could be biased upwards or downwards. Caution should be exercised in generalizing our data to the IDU population in Baltimore or other settings.

With the exception of data on NEP attendance, we relied on self-reports, which are prone to socially desirable responding.⁴⁷ Comparing interviewer-administered questionnaire data with audio computer-assisted self-interviews (ACASI), we recently found that men were more likely to overreport heterosexual activity to an interviewer, which could explain the inverse association we observed between HIV seroconversion and number of heterosexual partners.⁴⁸ Smaller numbers of women may have masked other associations. On the other hand, because this was a prospective study there is greater confidence that the drug-related and sexual risk factors we identified were causally related to HIV seroconversion.

In this cohort of more than 1800 IDUs, HIV incidence declined over time but remained unacceptably high for a 10-year period. We documented a number of risk factors for HIV seroconversion that differed significantly by sex. Although parenteral risks remained important risk factors for HIV seroconversion among men, those who engaged in homosexual activity were at increased risk of infection, and sexual risks predominated among women. The extent to which drug-related and sexual risks contribute to HIV incidence likely depends on local factors such as social networks, the nature and intensity of interventions, and the background prevalence of HIV infection among various subpopulations.^13,49,50 Nevertheless, the HIV risk factors we identified are similar to those described in a recent report from San Francisco, Calif, where HIV prevalence and incidence among IDUs has remained low.⁴⁶ As HIV incidence attributable to injection drug use decreases (which is the case in New York City⁵⁰), it can be expected that the relative contribution of sexual HIV risks will increase. In light of these findings, HIV prevention programs aimed at IDUs should be sex specific and take into account both sexual and parenteral risks.

Accepted for publication November 11, 2000.

This study was supported by grant DA12568 from the National Institute on Drug Abuse, Bethesda, Md.

Presented in part at the 13th International AIDS Conference, Durban, South Africa, July 9-14, 2000.

We are indebted to the study participants for their continued participation in the AIDS Link to Intravenous Experience (ALIVE) study. The authors also thank study staff, especially Grace Macalino, PhD, and the Baltimore Needle Exchange Program and evaluation staff for technical support; Roel Coutinho, MD, PhD, Alex Kral, PhD, and Samuel Friedman, PhD, for critical appraisal of the manuscript; and Ms Hazel Hamond-Terry for assistance with manuscript preparation.

Reprints: Steffanie A. Strathdee, PhD, Department of Epidemiology, Johns Hopkins School of Public Health, 615 N Wolfe St, Room E6010, Baltimore, MD 21205 (e-mail: sstrathd@jhsph.edu).