Comparison of HIV incidence estimated in clinical trial and observational cohort settings in a high risk fishing population in Uganda: Implications for sample size estimates

Background: Clinical trial participants may differ from the source population due to the demands of trial participation and self-selection, inadvertent selection of a lower-risk group, or both. We investigated the HIV risk status of volunteers in a Simulated Vaccine Efficacy Trial (SiVET) nested within a prospective observational cohort of fisher folks in South Western Uganda. Methods: Volunteers aged 18-49 years, at high risk for HIV from fishing communities in Masaka district were recruited into an observational cohort and followed quarterly. High risk was defined as a self-report, of at least one of the following in the past three months; sexually transmitted infections, unprotected sex with >1 partner or a new sexual partner, use of recreational drugs, weekly alcohol use, and/or frequent travel. Volunteers who had at least three months of follow-up in the observational cohort were consecutively enrolled in SiVET, administered Hepatitis B vaccine at months (0, 1, 6) and followed-up three days post vaccinations to mimic a vaccine trial schedule. HIV incidence over the next 12 months was compared between SiVET and the observational cohort studies. Results: Between January 2012 and February 2013, 575 individuals were enrolled in the observational cohort, of whom 282 were enrolled in SiVET between July 2012 and February 2013. Despite similar pattern of reported risk behaviour in both studies, HIV incidence was higher in observational cohort, 11.4 cases/100 PYO [95% CI: 7.4-17.7] compared to 3.8 [95% CI: 2.0-7.0] in SiVET (p < 0.01). SiVET volunteers tended to be men, having some education and longer-term residents, all factors that are also associated with lower HIV risk. Conclusion: We observed a lower HIV incidence in SiVET than in the observational cohort. The two populations differed significantly in demographics but not in reported risk. HIV incidence estimates from observational cohorts must be used with caution to estimate the trial study size.