Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Joseph S. Rossi


Men who have sex with men (MSM) in the United States continue to experience disproportionate HIV disease burden. Historically, most HIV prevention research has focused on reducing sexual risk behaviors through individual-level behavior change interventions. To date, behavioral interventions have not reduced HIV incidence among MSM and combination prevention approaches that package behavioral interventions with additive biomedical or structural interventions are now recommended. The last meta-analysis of HIV behavioral interventions for MSM was conducted in 2008. Since then sixteen new rigorous trials have been identified. New evidence and new recommendations justify an updated meta-analysis to identify the most promising and relevant features of behavioral interventions to be used in new combination approaches. This study aimed to calculate an updated effect size for MSM-specific HIV behavioral interventions, identify moderators of effect size, examine cumulative effect sizes over time, and describe trials that addressed more than one behavioral outcome (“integrated interventions”). Systematic review and meta-analysis evaluated effects of 34 randomized controlled trials for 17,872 US MSM conducted between 1989 and 2014. Behavioral interventions reduced the odds of sexual risk behavior by 14 percent (OR=.859, 95% CI [0.790, 0.933], p<.001). Findings suggest behavioral interventions are still somewhat effective to reduce sexual risk behavior, but the effect size was smaller than effect sizes observed in earlier meta-analyses. Cumulative meta-analysis further demonstrated that intervention effects gradually declined over time. From 1991 to 2014, the magnitude of the effect size decreased by 19.5 percent (OR=.719-.859). Reasons for effect size decline are not clear, but HIV prevention fatigue, inclusion criteria that focus on very high-risk MSM, choice of comparison condition, and underpowered primary trials likely contribute to effect size shrinkage. Statistical homogeneity restricted this study’s objective to reliably detect moderators of effect [Q(33)=39.35, p=.207; I2=16.14]. All moderators hypothesized a priori were not significant. Post-hoc moderator analyses found intervention effects to be moderated by age (p<.001), peer delivery (p=.002), community-level interventions (p=.032), HIV status (p=.019), education (p=.023), evidence-level (p=.076), retention (p=.09), and MSM subgroup (p=.09). Nine trials were identified that addressed at least one additional problem behavior other sexual risk behavior; six trials addressed substance use and three trials addressed HIV testing. In conclusion, this study provides new evidence that behavioral interventions have become less effective over time. Development of new combination prevention packages presents an opportune time to improve and update behavioral interventions. HIV prevention research would benefit from frequent research synthesis to monitor effect sizes, identify the most effective intervention components, retire outdated intervention components, and identify gaps in current research. To our knowledge, this is the first meta-analysis of HIV behavioral interventions to demonstrate effect size shrinkage for HIV behavioral interventions using cumulative meta-analysis. Routine cumulative meta-analysis should be included in research synthesis protocols to examine and explain effect size shift as new evidence is accumulated.