Characterisation of between-cluster heterogeneity in malaria cluster randomised trials to inform future sample size calculations

Cluster randomised trials (CRTs) are important tools for evaluating the community-wide effect of malaria interventions. During the design stage, CRT sample sizes need to be inflated to account for the cluster-heterogeneity in measured outcomes. One such measure of heterogeneity, the coefficient of variation ( k ), is typically used in malaria CRTs yet is often estimated without prior data. Underestimation of k undermines study power and increases the probability of CRTs generating null results. We conducted a meta-analysis of cluster-summary data from 24 malaria CRTs and calculated true k values for prevalence and incidence outcomes using methods-of-moments and regression modelling approaches. Using random effects regression modelling we investigated the impact of empirical k values on original trial power, effect size uncertainty and explored associated factors. Results revealed empirical estimates of k often exceeded those used in sample size calculations which heavily contributed to compromised study power and effect size precision. Increased between-cluster heterogeneity of outcomes was associated with outcome measures (i.e. incidence or prevalence), lower endemicity, seasonality of surveys and uneven intervention coverage across clusters. Study findings can be used to inform future malaria CRT sample size calculations and trial design to help ensure malaria interventions are effectively and feasibly evaluated.