Summary measures of adherence using pill counts in two HIV prevention trials: the need for standardisation in reporting.

For trials of user-dependent HIV prevention products, accurate adherence measurements are essential to interpret and compare results across trials. We used pill count data from two recent HIV prevention trials of herpes simplex virus type 2 (HSV-2) suppression, to show that estimates of adherence vary substantially depending on assumptions that are made in analysing pill count data. We associate calculated adherence with biological markers of anti-HSV-2 activity. In both trials, calculated adherence varied considerably, depending on the summary measure used, and the handling of intervals with apparent 'over-adherence' (fewer pills returned than expected), and unreturned pills. Intervals of apparent over-adherence were associated with reduced antiviral effects on biological markers of herpes reactivation, indicating these are likely to represent periods of non-adherence. Our results demonstrate the clear need for standardisation in reporting of adherence data that are based on pill counts.