Objective: To estimate the effects of alcohol consumption and number of unprotected receptive anal intercourse partners on HIV seroconversion while appropriately accounting for time-varying confounding.
Design: Prospective cohort of 3725 HIV-seronegative men in the Multicenter AIDS Cohort Study between 1984 and 2008.
Methods: Marginal structural models were used to estimate the joint effects of alcohol consumption and number of unprotected receptive anal intercourse partners on HIV seroconversion.
Results: Baseline self-reported alcohol consumption was a median 8 drinks/week (quartiles: 2, 16), and 30% of participants reported multiple unprotected receptive anal intercourse partners in the prior 2 years. Five hundred and twenty-nine HIV seroconversions occurred over 35 870 person-years of follow-up. After accounting for several measured confounders using a joint marginal structural Cox proportional hazards model, the hazard ratio for seroconversion associated with moderate drinking (1–14 drinks/week) compared with abstention was 1.10 [95% confidence limits: 0.78, 1.54] and for heavy drinking (>14 drinks/week) was 1.61 (95% confidence limits: 1.12, 2.29) (P for trend <0.001). The hazard ratios for heavy drinking compared with abstention for participants with 0–1 or more than 1 unprotected receptive anal intercourse partner were 1.37 (95% confidence limits: 0.88, 2.16) and 1.96 (95% confidence limits: 1.03, 3.72), respectively (P for interaction = 0.42).
Conclusion: These findings suggest that alcohol interventions to reduce heavy drinking among men who have sex with men should be integrated into existing HIV prevention activities.