Sex work or male-with-male sexual activity and the risks of HIV and hepatitis C acquisition among people who inject drugs: a global systematic review and meta-analysis

Introduction

The sharing of injection equipment is a primary risk factor for hepatitis C virus (HCV) and HIV transmission, which are leading causes of morbidity and mortality for people who inject drugs (PWID).^1,2 Globally, PWID experience high rates of new HIV and HCV infections,^3,4 with contributing factors including suboptimal coverage of harm reduction interventions leading to heightened injecting and sexual risks and low coverage of HIV and HCV treatment.^5,6 Further, structural barriers such as unstable housing and incarceration, which are commonly experienced by PWID, also increase HIV and HCV risk.^7,8 Despite reductions in HIV and HCV incidence among PWID over the last two decades in some settings, transmission generally remains high,³ suggesting that the 2030 WHO and UNAIDS elimination targets are unlikely to be achieved among PWID without expanding prevention and treatment services, and reducing structural barriers.^3,4

Beyond PWID, men who have sex with men (MSM) and sex workers also experience a disproportionate burden of HIV and HCV.⁹ Most HIV and HCV research among these groups treats them as distinct populations,^9,10 while prevention and care interventions are organised around specific populations and behaviours.11, 12, 13 However, evidence suggests there is considerable overlap in sexual and injecting behaviours that are associated with greater risk of HIV and HCV transmission between PWID, MSM and sex workers.^5,14 According to a recent global systematic review, 15% of PWID recently engaged in sex work and 9% identified as lesbian, gay, or bisexual.⁵

These overlapping behaviours may increase HIV and HCV exposure among PWID through compounding injecting and sexual risks, and heightened vulnerability to social and structural factors (e.g. stigma, marginalisation, poverty and criminalisation).15, 16, 17 For instance, compared with sex workers who do not inject drugs, those who inject drugs are more likely to engage in street-based or survival sex work,^18,19 which is associated with limited condom negotiation, rushed or unsafe injections, more interactions with law enforcement and greater exposure to physical and sexual violence.^12,20,21 Criminalisation of both sex work and drug use in most countries also subjects sex workers who inject drugs to intensified policing, and incarceration, which can disrupt health care access and increase HIV and HCV risk.^8,22, 23, 24, 25 Gender dynamics compound these risks; globally, most sex work is conducted by women, who experience power imbalances, gender-based violence, and poverty.26, 27, 28 Similarly, MSM who use drugs are more likely to engage in condomless anal sex compared with MSM who do not use drugs^29,30 and may participate in sexualised drug use (including ‘chemsex’),³¹ which can involve multiple partners and other sexual and injection behaviours that are associated with greater HIV and HCV acquisition risk.^11,32,33 More broadly, experiences of stigma and discrimination related to sex work or sexuality34, 35, 36 can further influence these behaviours and deter access to health and harm reduction services.^37,38

Despite evidence of overlapping sexual and injecting behaviours between PWID, sex workers and MSM, the extent to which the risks of HIV and HCV acquisition may differ among PWID who engage in sex work or male-with-male sex compared to those who do not is uncertain. We conducted a global systematic review and meta-analysis of published and unpublished data to estimate the magnitude of these risks.

Methods

Results

Searches conducted as part of previous reviews and our update identified 82,105 studies published between January 1st, 2000, and June 9th, 2026, of which 27,279 were duplicates (Fig. 1). After screening titles and abstracts, 767 studies were eligible for full-text screening. Of these, 331 were excluded,^3,4,7,8 with 436 studies reporting HIV or HCV incidence among PWID retained in our database. These 436 studies were full-text screened for this review, with 62 reporting at least one association between male-with-male sex or sex work and HIV or HCV incidence. After contacting authors of 36 studies to request additional data, a further 24 studies (24/36; 67% response rate) responded and provided data, resulting in a total of 86 studies. After removing studies that reported duplicate estimates from the same study population, 57 studies contributing 164 estimates (49 published and 115 unpublished) were retained. The updated search, which covered November 2nd, 2024 to June 9th, 2026, yielded one additional eligible study.

Among included studies, 23 reported estimates of association with HCV incidence, 22 reported estimates of association with HIV incidence, and 12 reported estimates of association for both HIV and HCV incidence. Twenty-four studies provided data on associations with sex work only, 11 on associations with sex between men only and 22 on associations with both exposures (Table 1).

In studies reporting associations between male-with-male sex and HIV or HCV acquisition risk (33 studies), recent and lifetime exposure were each reported in 17 studies. For HIV, all estimates were restricted to men who inject drugs. For HCV, three studies included estimates of same-sex sexual activity among participants who were not exclusively men (73%, 83% and 96% were men in each). In random-effects meta-analyses, the pooled proportions of men who inject drugs reporting recent and lifetime sex with men were 4.8% (95% CI: 2.7–8.3%; number of estimates (N) = 12; I² = 98%) and 21.9% (95% CI: 11.9–36.8%; N = 14; I² = 99%), respectively.

For HIV, 13 studies contributed crude estimates for the association between recent male-with-male sex and HIV acquisition risk among men who inject drugs, yielding a pooled incidence rate ratio (IRR) of 2.35 (95% CI: 1.51–3.66, N = 13), with moderate between-study heterogeneity (I² = 51%; Table 2, Fig. 2). The pooled adjusted IRR was slightly higher than the crude IRR for recent male-with-male sex (IRR: 2.66; 95% CI: 1.43–4.98; N = 6; I² = 71%). Lifetime male-with-male sex had a pooled crude IRR for HIV acquisition of 2.44 (95% CI: 1.18–5.02; N = 8, I² = 87%), which was similar when using adjusted estimates (IRR: 2.32; 95% CI: 1.23–4.37; N = 8; I² = 84%) (Appendix 10).

Conversely, there was no evidence of an association between either recent (IRR: 1.28; 95% CI: 0.95–1.72; N = 8; I² = 21%) or lifetime (IRR: 0.89; 95% CI: 0.74–1.07; N = 12; I² = 0%) male-with-male sex and HCV acquisition risk when crude estimates were used (Table 2 and Appendix 10). Pooled estimates remained similar in sensitivity analyses where studies with fewer than 100% of the sample were male PWID were excluded (Appendix 13).

In studies reporting associations between sex work and HIV or HCV acquisition risk (N = 42), recent and lifetime exposure were reported in 38 and nine studies, respectively. In random-effects meta-analyses, the pooled proportions of PWID reporting recent and lifetime sex work were 10.6% (95% CI: 7.9–14.1%; N = 36; I² = 99%) and 14.6% (95% CI: 8.8–23.2%; N = 9; I² = 99%), respectively. By sex, these estimates were lower among men than women: 4.0% (95% CI: 2.5%–6.3%; N = 19; I² = 98%) in men vs 27.2% (95% CI: 17.4%–39.8%; N = 15; I² = 98%) in women for recent exposure, and 8.9% in men (95% CI: 4.5%–16.9%; N = 7; I² = 99%) vs 44.0% in women (95% CI: 22.2%–68.4%; N = 5; I² = 97%) for lifetime exposure.

Twenty studies contributed crude estimates for the association between recent sex work and HIV acquisition risk among male and female PWID combined, yielding a pooled IRR of 1.72 (95% CI: 1.30–2.28; N = 20) with moderately-high between-study heterogeneity (I² = 63%, Table 3, Fig. 3). Pooled adjusted estimates were similar (IRR: 1.73; 95% CI: 1.27–2.36; N = 11; I² = 59%). When crude estimates were stratified by sex, there was evidence of an association between recent sex work and HIV acquisition among women who inject drugs (IRR: 2.30; 95% CI: 1.44–3.68; N = 12; I² = 68%), but no evidence among men who inject drugs (IRR: 1.43; 95% CI: 0.92–2.22; N = 12; I² = 51%).

Recent sex work was also associated with HCV acquisition with a pooled IRR of 1.59 (95% CI: 1.29–1.96; N = 22) based on crude estimates, with moderately-high between-study heterogeneity (I² = 60%, Table 3, Fig. 4). The association persisted, though slightly attenuated, when using adjusted estimates (IRR: 1.43; 95% CI: 1.13–1.80; N = 12; I² = 54%). When crude estimates were stratified by sex, recent sex work was associated with HCV acquisition among men who inject drugs (IRR: 1.51; 95% CI: 1.22–1.87; N = 14; I² = 0%), but not women who inject drugs (IRR: 1.26; 95% CI: 0.90–1.77; N = 12; I² = 55%).

There were fewer estimates of the association between lifetime sex work and HIV and HCV acquisition risk. When pooling crude estimates, we found no evidence of an association between lifetime sex work and HIV (IRR: 1.17; 95% CI: 0.76–1.79; N = 6; I² = 52%) or HCV (IRR: 1.24; 95% CI: 0.95–1.63; N = 5; I² = 7%) acquisition risk. Adjusted estimates also showed no evidence of an association with HIV (IRR: 1.10; 95% CI: 0.61–1.98; N = 4; I² = 70%) or HCV (IRR: 1.21; 95% CI: 0.90–1.61; N = 4; I² = 0%) acquisition.

Select sub-group analyses for male-with-male sex as the exposure are shown in Table 2 with complete results shown in Supplementary Material (Appendix 11 and Appendix 12). Subgroup analyses resulted in estimates that were largely similar with the main results across strata. However, we noted some differences. For the association between male-with-male sex and HIV acquisition risk, the magnitude of the association appeared larger in studies conducted in 2010 or later (vs earlier studies) and in HICs, whereas there was no clear association in LMICs. For male-with-male sex and HCV risk, an association was observed in studies with a shorter median duration of injection or a lower proportion of homeless PWID, whereas no clear association was evident in the corresponding studies with higher injecting duration or higher proportion homelessness.

For sex work and HIV risk, the magnitude of the association was greater in studies with a higher proportion of women, higher levels of recent incarceration and in HICs, whereas there was no clear association in studies from LMICs. For sex work and HCV risk, stronger associations were observed in published studies, in studies conducted before 2010, in HICs, and in studies with a lower proportion of homeless PWID. However, most pooled RRs across strata were based on small numbers of estimates, resulting in wide and overlapping 95% CIs.

When assessing differences between low- and high-incidence settings, we found no evidence that pooled effects differed between male-with-male sex and HIV incidence (<2 cases/100 PYs: [RR: 2.62; 95% CI: 1.45–4.74; N = 7; I² = 66%] vs ≥2 cases/100 PY: [RR: 2.81; 95% CI: 1.36–5.82; N = 5; I² = 0%]) and HCV incidence (<15 cases/100 PYs: [RR: 1.55; 95% CI: 0.79–2.37; N = 2; I² = 70%] vs ≥15 cases/100 PY: [RR: 1.32; 95% CI: 1.01–1.73; N = 4; I² = 0%]). For sex work, we found substantially higher pooled effects in low incidence studies for HIV (<2 cases/100 PYs: [RR: 2.38; 95% CI: 1.80–3.16; N = 10; I² = 38%] vs ≥2 cases/100 PY: [RR: 1.24; 95% CI: 1.01–1.52; N = 8; I² = 0%]), and slightly elevated pooled effects in low incidence studies for HCV incidence (<15 cases/100 PYs: [RR: 1.78; 95% CI: 1.24–2.55; N = 8; I² = 79%] vs ≥15 cases/100 PY: [RR: 1.38; 95% CI: 0.98–1.95; N = 8; I² = 59%]).

Visual inspection of funnel plots and the results from Egger's tests did not provide any evidence of asymmetry bias for our pooled estimates (Appendix 14). This finding was mostly consistent when assessing published and unpublished estimates separately, except we found evidence of asymmetry for published studies assessing the association between male-with-male sexual activity and HIV acquisition (P = 0.011).

Discussion

In this global systematic review and meta-analysis, recent sex work and male-with-male sexual activity were associated with higher HIV acquisition risk among PWID. Male PWID who engaged in recent male-with-male sex had a 2.3-fold higher risk of HIV acquisition. Similarly, PWID engaged in recent sex work had a 1.7-fold higher risk of HIV acquisition, with this risk being higher among women who inject drugs (2.3-fold higher risk), but there being no evidence among men who inject drugs. In contrast, evidence regarding recent male-with-male sex or sex work and the association with HCV acquisition risk was weaker. There was no evidence that PWID reporting recent male-with-male sex had a higher risk of HCV acquisition. However, among PWID who recently engaged in sex work, there was evidence of a 1.6-fold increase in HCV acquisition risk, which persisted when considering men only (1.5-fold higher risk) but there being no evidence among women. These findings, based on crude estimates for recent exposure to male-with-male sex and sex work, were consistent when considering adjusted estimates. For lifetime exposure, findings were similar for the association between HIV acquisition risk and male-with-male sexual activity, but there was no evidence of an association for HCV acquisition risk or for exposure to sex work. Notably, the point estimates of pooled effects in which 95% confidence intervals crossed the null still suggest a higher risk of HIV and HCV acquisition among PWID engaging in male-with-male sexual activity or sex work, possibly reflecting a lack of statistical power rather than a lack of effect, given that we would expect these behaviours to elevate risk.

Previous studies have generally treated sex workers, MSM, and PWID as distinct populations when describing transmission risks and their burden of HIV and HCV.⁹ Our review is novel in quantifying the extent to which PWID who engage in male-with-male sex or sex work experience an increased risk of HIV and HCV acquisition compared to PWID without these behaviours.^3,5,44, 45, 46, 47, 48, 49, 50 While our analyses cannot directly identify the mechanisms driving these associations, our findings might be explained by higher levels of stigma, discrimination, and criminalisation among PWID who engage in sexual and injecting behaviours that increase HIV and HCV acquisition risk, though clear data supporting this remain limited.^24,51 These factors are, in turn, associated with reduced access to health and harm reduction services.^52,53 These associations should also be interpreted in the context of underlying transmission dynamics for HIV and HCV. For example, the probability of HCV transmission is higher for parenteral exposure compared with sexual exposure, which may explain the lack of an association between male-with-male sex and HCV acquisition. The association between sex work and HCV acquisition may therefore be attributable to structural risks such as unstable housing, criminalisation, and reduced access to harm reduction services which are associated with increased injecting behaviours that increase HIV and HCV acquisition risk.^8,54,55 For HIV, where sexual transmission is more efficient, the elevated risks observed for both male-with-male sex and sex work likely reflect a combination of direct sexual transmission risk, with significant a HIV burden among MSM and people engaged in sex work, and structural vulnerabilities that concurrently elevate injecting-related behaviours that increase HIV and HCV acquisition risk.

Furthermore, our study design does not allow us to distinguish between sexual and injecting transmission. Thus, a better understanding of the underlying mechanisms is needed to help inform intervention development. Initiatives to address these elevated risks should include the development of integrated, low-threshold sexual health and harm-reduction services alongside policy changes to reduce criminalisation, stigma, and barriers to care. This is particularly important given how common overlapping sexual exposures are among PWID, with our study finding that 11% of PWID reported recent sex work, while approximately 5% of male PWID reported recent male-with-male sexual activity and around 22% reported lifetime male-with-male sexual exposure. Recent evidence from global systematic reviews suggests that HIV and HCV incidence is higher for women who inject drugs compared with men,³ likely reflecting the structural vulnerabilities that women who inject drugs can experience, including gender-based violence and unequal power dynamics.⁵⁶ Importantly, our analyses also suggest that the association between sex work and HIV acquisition is strong among women who inject drugs, with no association among men who inject drugs, with a much higher proportion of women who inject drugs reporting recent sex work (27% vs 5% in men who inject drugs). This underscores the added importance of ensuring that health and harm reduction services for PWID are equipped to support the needs of women engaged in sex work, including such things as STI testing and reproductive health services.⁵⁷

A key strength of our analysis was the incorporation of both published and unpublished estimates. By including unpublished estimates, we increased statistical power and reduced the risk of publication bias.⁵⁸ We found that the pooled associations were sometimes higher in published studies than in unpublished data, with the greatest discrepancy observed for the association between recent sex work and HCV acquisition. Including these additional unpublished data has likely made our review more representative, enhancing the generalisability of the findings by allowing a wider range of study settings to be included.

Our study has limitations, some of which reflects the underlying data available for inclusion. First, most estimates are from HICs, meaning our findings may be less generalisable to PWID located in LMICs. Second, there was variation across studies in how sex work and male-with-male sex were measured. For example, in four studies, lifetime male-with-male sexual activity was defined based on sexual identity rather than reported behaviour, which could have misclassified exposure, for example by excluding men who have sex with men that identify as heterosexual. Further, the sexual behaviour exposures were based on self-report and responses to these questions may vary across different cultures and could be under-reported in some settings due to behaviours being socially stigmatised. This misclassification is likely to have been largely non-differential with respect to the outcome, thereby biasing estimates toward the null. Third, we included three estimates of same-sex sexual activity in our analyses among participants who were not exclusively men (73%, 83% and 96% of samples were men), one in our pooled analysis of recent exposure and two in our pooled analysis of lifetime exposure. These studies were excluded in our sensitivity analyses, and the results were largely unchanged. Fourth, because our main analyses were crude, we cannot rule out confounding. We opted to prioritise crude estimates due, in part, to the premise that some of the factors included in adjustment (e.g. injecting practices, sexual behaviours, housing instability) may act as mediators rather than confounders, potentially attenuating the effect. In reality, the relationship between these variables is complex. For example, housing instability could also act as a confounder. Thus, this distinction cannot be reliably disentangled in a systematic review relying on estimates with adjustment sets pre-determined by the original study authors. However, adjusted analyses generally produced similar pooled effects. Fifth, there was moderate to high heterogeneity in most of our pooled estimates, limiting their generalisability. While we conducted subgroup analyses across a range of study-level study characteristics, these variables did not explain all the observed heterogeneity, and most subgroup estimates were based on small numbers of studies, resulting in wide, overlapping confidence intervals. Unmeasured sources of heterogeneity likely contributed to differences across studies, particularly variation in how sex work, and male-with-male sex were defined and measured, which was not standardised across included studies and could not be captured by our subgroup categories. Other unmeasured factors may also include the broader social and structural context surrounding sex work or male-with-male sex (e.g. stigma, social acceptance), accessibility of HIV/HCV prevention and harm reduction services, and the epidemic phase. Sixth, transgender PWID were not considered in our stratified analyses of the association between sex work and HIV/HCV incidence, as there was insufficient data to calculate meaningful estimates. Seventh, high baseline risk among PWID may attenuate relative measures of association. As baseline incidence rises, these measures compress toward the null even when the absolute effect of an exposure remains constant. Last, although we contacted cohorts systematically, unpublished estimates did not undergo formal peer review, and we cannot eliminate risk of selection bias as cohorts with prior published data were more likely to be identified for contact, and responding cohorts may systematically differ from non-responding cohorts in ways that could bias the pooled estimate in either direction.

In this global systematic review and meta-analysis among PWID, we found recent sex work was associated with increased HIV risk, particularly among women, and, to a lesser extent, with HCV risk, while male-with-male sexual activity was associated with higher HIV risk among male PWID, but not HCV risk. These findings highlight the need for integrated prevention and treatment services that address both injection-related and sexual risks. Rather than siloed approaches, services should combine access to harm reduction measures (i.e. sterile injecting equipment and OAT, HIV and HCV testing), with sexual health care including provision of pre-exposure prophylaxis and condoms, STI testing and treatment, and reproductive health services. Further research is needed on the intersectionality among PWID and the factors associated with elevated risk among PWID belonging to more than one population, particularly in LMICs where evidence remains limited. This research could help in the development of tailored interventions to mitigate for the increased risks that these PWID experience.

Contributors

JD, JS, PV and AAA designed the study. JS and PV wrote the study protocol. JD conducted statistical analyses. JD, JS, BJN and AAA screened the articles and extracted data from the studies. AT, BK and LB accessed and verified the underlying data. JD, JW and PV contacted authors to request unpublished data. DA, MA, IA, KB, AB, JB, KDB, DDJ, PD, KD, CH, AH, JRH, KH, MH, VH, VJ, AK, LM, SM, HM, MDM, KP, NEP, LP, MP, KR, LRS, SR, RSD, AS, SAS, SSS, VS, DVS, GZ provided access to unpublished data. JD drafted the first version of the manuscript. JS, PV, AAA provided critical feedback on the first version of the manuscript. All authors read and approved the final manuscript.

Data sharing statement

Data used in this study are available from the corresponding author upon reasonable request.

Declaration of interests

KP declares support from the National Institutes of Health (NIH), New Mexico Department of Health, and New Mexico's State Opioid Response (SOR). VS's declares that their institution has received funding from Roche, AstraZeneca, and Gilead Sciences. KDB and JB declares support from AbbVie and Gilead Sciences. PD declares that their institution has received support from Navamedic. VS declares support from Roche, AstraZeneka, and Gilead Sciences, and is an unpaid member of the Scientific Committee of the European Union Drugs Agency. SM declares support from Abbott Diagnostics. All other authors have no competing interests to declare.

Appendix A. Supplementary data