Manicaland PrEP Uptake Through Interactive Counselling Trial

With 6 clusters per arm, mean cluster size of 48 individuals, with co-efficient of variation in outcomes between 0.15 to 0.55, the trial will have 80% power to detect difference of 5 - 9 percentage points when baseline PrEP uptake is assumed to be 1%. Primary analysis: A two-stage intention-to-treat (ITT) analysis of cluster level summaries will be used. The parameter of interest is the adjusted risk difference in primary and secondary outcomes. In the first stage, Logistic regression will be used to adjust for confounding variables at the individual level. The model will be fitted using data from both trial arms. The regression model will include terms for the covariates of interest and pair but not trial arm. Adjustment variables will include age group (\<20, 20-24), highest level of education completed, marital status and baseline measures of outcomes. In the second stage, a formal statistical comparison for differences in observed versus expected uptake by arm will be conducted using a two-sided paired significance test with α=0.05. Secondary analysis: In secondary analysis, the complier average causal effect (CACE) will be estimated using an instrumental variables approach. The parameter of interest is the adjusted risk difference in primary and secondary outcomes. Heterogeneity by age and baseline levels of risk and time preferences will also be estimated.