The systematic review assembled 94 studies representing 186,613 female sex workers across 39 countries. The study team reported higher odds of prevalent HIV infection among FSWs who experienced recent violence (past year) and among those who reported lifetime violence. The pooled estimate for recent violence was elevated (pooled odds ratio 1.31), and the lifetime exposure pooled estimate was similar (pooled odds ratio 1.35). Heterogeneity varied between pooled analyses, reflecting differences in study design, setting, and measurement of both violence and HIV outcomes. The authors preferred adjusted estimates where available; some included studies contributed multiple estimates when they reported more than one outcome or exposure.
Short, direct point: exposure to violence correlated with higher odds of prevalent HIV in pooled analyses. Longer detail: these associations persisted across analyses stratified by recall period and were observed in both low- and middle-income country settings and higher-income settings included in the review, although the paper notes varying effect sizes and statistical heterogeneity across those subgroups.
When investigators stratified by type of violence, sexual violence showed the strongest pooled associations with prevalent HIV. For recent sexual violence the pooled odds ratio was higher than for aggregated forms of violence, and lifetime sexual violence also yielded a notable pooled association. The study team reported that sexual violence had larger effect sizes than physical, emotional/psychological, or financial violence in the pooled prevalence analyses. This pattern suggests that the mechanism linking violence to HIV prevalence may be partly driven by sexual exposure and by associated behaviour changes that increase acquisition risk.
The authors made no claims beyond the reported pooled effect sizes. They also noted measurement complexity: many primary studies relied on self-report, different recall windows, and heterogeneous violence definitions, which complicates direct comparisons between violence types.
Pooled analyses addressed several steps of the HIV treatment cascade. Recently experienced violence was associated with reduced odds of ART use (pooled OR 0.78). Lifetime exposure to violence was associated with reduced odds of viral suppression (pooled OR 0.88). The review therefore linked recent and lifetime violence with poorer programmatic outcomes at key cascade points: initiation or continuation of ART and achieving viral suppression.
The study team observed that pooled estimates for ART-related outcomes came from fewer studies than the prevalence analyses, which reduced precision and increased sensitivity to between-study differences. The included reports varied in how they measured ART use and viral suppression — for example, some used clinical records while others used self-report — and the authors noted they preferentially pooled adjusted estimates when available.
The meta-analysis found no evidence of pooled associations between violence and HIV incidence, HIV testing uptake, or ART adherence. The investigators cautioned that these null findings should be interpreted with care: relatively few studies reported incidence data and cohort designs were uncommon, which limits statistical power to detect associations with incident infection. Similarly, outcomes such as testing uptake and adherence were measured inconsistently across studies, reducing the ability to synthesise findings.
In short: pooled prevalence and some treatment outcomes showed associations, but incidence, testing, and adherence did not show consistent pooled links in the available literature.
The authors identified three main limitations in the evidence base. First, most included studies were cross-sectional, which constrains causal inference between experiences of violence and HIV outcomes. Second, pooled estimates exhibited heterogeneity driven by variation in study populations, geographic contexts, violence definitions, recall windows, and outcome measurement methods. Third, measuring experiences of violence reliably proved difficult across primary studies; the review documented variation in instruments and reporting that may bias estimates toward the null or inflate heterogeneity.
The paper also highlighted that fewer studies contributed data on ART use, viral suppression, incidence, testing, and adherence than contributed to prevalence estimates, which affects the robustness of pooled estimates for those outcomes. The authors reported that data and analysis code are publicly available at the cited repository, enabling further scrutiny and secondary analyses.
Overall, the investigators reported consistent associations between experiences of violence and increased odds of prevalent HIV, with the largest effects for sexual violence, and evidence that violence relates to poorer ART uptake and viral suppression. The authors concluded by noting the need for HIV programmes for FSWs to integrate violence prevention and response, and they identified the predominance of cross-sectional studies and measurement heterogeneity as priority gaps for future research.