Requirements for designing cluster randomised control trials to detect suppression of malaria vector population densities

Abstract Background Novel interventions for mosquito-borne disease control which release modified mosquitoes that are sterilised or genetically modified to cause offspring inviability are progressing towards field applications. Cluster randomised control trials (CRCTs) could provide robust assessment of intervention efficacy in suppressing mosquito populations in field environments, but guidance on designing CRCTs to detect mosquito suppression impacts is limited. Results We developed statistical models to simulate CRCTs, informed by a 5-year time series measuring densities of malaria vector species from the Anopheles gambiae complex in four villages in western Burkina Faso. We estimated requirements for parallel and step wedge designs, varying the targeted vector species, the suppression effect and the monitoring regime. For a suppression effect of 50%, 21–22 clusters were required to detect suppression with 90% power when all An. gambiae complex species were targeted, while 24–26 clusters were required when only An. coluzzii was targeted and 60–66 clusters were required when only An. gambiae was targeted. For stronger suppression effects, required trial sizes depended less on target species, with 9–10 clusters being sufficient to detect a 90% suppression effect. We investigated how reducing sampling effort, by sampling fewer houses and restricting sampling to rainy season months, affected statistical power. Conclusions Our results provide empirically based guidance for designing CRCTs to evaluate interventions aiming to suppress malaria vector populations.