Background Men who have sex with men (MSM) are the population most affected by HIV in Latin America (LA). Youth MSM (YMSM) are of particular interest given potentially greater levels of risk behaviour and as a barometer of recent epidemic dynamics. Yet single MSM studies lack sufficient sample size to characterise YMSM so that HIV prevalence and variation among YMSM within LA is unknown.

Methods We developed a multilevel statistical framework to borrow strength across recent (2006–2009) respondent-driven sampling (RDS) studies with similar survey methodologies to permit characterisation of HIV prevalence and risk factors among YMSM in LA and comparisons across cities. Data from participants aged 18–24 years tested for HIV in MSM studies in nine high-population cities in Argentina, Bolivia, Costa Rica, El Salvador and Honduras (N=1410) were pooled. Weights incorporating personal network and population sizes adjusted for selection probability. Logistic, multilevel models identified correlates of HIV infection, including city-level fixed effects and accounted for intraclass correlation within recruitment chains.

Results HIV prevalence was 4.6% (95% CI: 3.2% to 6.5%) in the pooled sample and varied significantly across cities from 2.7% in Buenos Aires, Argentina to 9.3% in San Pedro Sula, Honduras. Samples varied (p value<0.05) in terms of completion of secondary education (69.6%–100.0%), gay (range 2.5%–77.5%) and bisexual self-identity (18.8%–97.6%), past-year drug use (5.9%–64.8%), condom use at last anal sex (26.6%–72.7%) and other risk behaviours, knowledge of HIV prevention and transmission (26.4%–71.9%) and lifetime HIV testing (28.7%–59.2%). In multivariate analysis, syphilis infection (adjusted OR [AOR]=2.5), aged 23–24 years (vs 18–22) (AOR=2.2), past-year crack/cocaine use (AOR=2.2) and bisexual self-identity (AOR=0.5) were associated with HIV infection. Significant city effects for Salvador, El Salvador and Santa Cruz, Bolivia persisted net of individual-level differences.

Conclusions Formal comparisons of HIV burden and risk differences among cities can be achieved through multilevel analysis. HIV prevalence, drug and sexual risk behaviours, and low HIV testing among YMSM in LA are considerable and vary substantially within the region. The need for prevention among YMSM in LA is urgent and must address substance abuse, STI and be tailored to local context. Prevention should target younger YMSM as probability of infection rises rapidly with age.