Methods We installed solar-powered odour-baited mosquito trapping systems (SMoTS) to households on Rusinga Island, Lake Victoria, western Kenya (mean population 24 879), in a stepped-wedge cluster-randomised trial. All residents in the completed health and demographic surveillance system were eligible to participate. We used the travelling salesman algorithm to assign all households to a cluster (50 or 51 geographically contiguous households); nine contiguous clusters formed a metacluster. Initially, no cluster had SMoTS (non-intervened). During the course of the intervention roll-out SMoTS were gradually installed cluster by cluster until all clusters had SMoTS installed (intervened). We generated 27 cluster randomisations, with the cluster as unit of randomisation, to establish the order to install the traps in the clusters until all had a SMoTS installed. Field workers and participants were not masked to group allocation. The primary outcome of clinical malaria was monitored through repeated Household visits covering the entire population, once before roll-out (baseline) and five times throughout the 2-year roll-out. We measured clinical malaria as fever plus a positive result with a rapid diagnostic test. The SolarMal project was registered on the Dutch Trial Register (NTR 3496).
Findings We enrolled 34 041 participants between April 25, 2012, and March 23, 2015, to 81 clusters and nine metaclusters. 4358 households were provided with SMoTS during roll-out between June 3, 2013, and May 16, 2015. 23 clinical malaria episodes were recorded in intervened clusters and 33 episodes in non-intervened clusters (adjusted effectiveness 40·8% [95% CI –172·8 to 87·1], p=0·5) during the roll-out. Malaria prevalence measured by rapid diagnostic test was 29·8% (95% CI 20·9–38·0) lower in SMoTS clusters (prevalence 23·7%; 1552 of 6550 people) than in non-intervened clusters (prevalence 34·5%; 2002 of 5795 people).
Interpretation The unexpectedly low clinical incidence of malaria during roll-out led to an imprecise estimate of effectiveness from the clinical incidence data. The substantial effect on malaria prevalence is explained by reduction in densities of Anopheles funestus. Odour-baited traps might be an effective malaria intervention.