RESUMO
BACKGROUND: Poor retention can reduce study power and thwart randomization, possibly resulting in biased estimates of effect. Some HIV prevention trials conducted in developing countries have been challenged by high loss to follow-up. Identifying factors associated with non-retention could lead to recruitment of women more likely to remain in the trial, potentially yielding greater efficiency and validity. METHODS: We summarized retention rates and, using Cox regression, evaluated factors associated with non-retention in four trials of two candidate vaginal microbicides (1% C31G or SAVVY® and 6% cellulose sulfate or CS) conducted in multiple sub-Saharan African countries. We defined retention as completion of the trial, including those with an HIV outcome. Non-retention comprised participants randomized to a study arm who were either lost to follow-up or discontinued prior to infection with HIV. RESULTS: 7,367 women were enrolled and randomized in the four trials; 7,086 are included in this analysis. 1,514 (21.4%) participants were either lost to follow-up or had early discontinuation. In the final Cox model, the following baseline factors were associated with non-retention: younger age (hazard ratio [HR] = 0.95); less education (HR = 0.97); condom use at last sex (HR = 1.18); larger number of sex acts in a typical week (HR = 1.01); and baseline candidiasis or bacterial vaginosis (HR = 1.12). CONCLUSIONS: Younger and less educated women were more difficult to retain in these microbicide trials. But these same traits may be associated with higher HIV infection rates. Enhanced retention methods focused on those at highest risk of non-retention and possibly infection will optimize study efficiency and validity.