Table 2. Study strengths and potential limitations.

Citation	1st author, publication year	Study strengths	Study limitations
[27]	Degener C.M. 2014	• Clusters paired on baseline mosquito density with intervention randomly assigned to one cluster per pair. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • Active case finding.	• Lack of baseline dengue seroprevalence data. • Reduced household intervention participation rates toward end of study (decreased from 60.5% to 36%). • Low dengue transmission during the study period. • The low rate of dengue and high Culex catch rates may have led to false reassurance with subsequent relaxing of anti-dengue measures. • Unclear how many households used BG-traps continuously; trapped mosquitoes may have been lost during power cuts and eaten by ants entering the catch bag. • Contamination risk from migrating mosquitoes; minimum distance between clusters of 250 metres.
[8]	Andersson N. 2015	• High community engagement achieved by involving community members, creating community-led campaigns. • Randomisation used to assign intervention. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • Active case finding.	• Non-participation bias among wealthier people. • Security issues reduced intervention participants’ engagement with researchers. • Sharing the baseline results with participants in the control and intervention arms may have mobilised both groups to perform anti-dengue control measures. • Entomology evaluators were not blinded. • Intensive government anti-dengue campaigns reduced the difference between study arms. • Two non-participating clusters were included in the intervention arm in the data analysis. • Contamination risk from migrating mosquitoes.
[22]	Syafruddin D. 2014	• Randomisation used to assign intervention. • Outcomes measured clearly and reliably. • Active case finding. • Placebo controlled.	• Four burning spatial repellent coils per house each night was not practical to implement. • The two participating villages (each with an intervention and control arm) had very different baseline malaria rates (heterogeneity). • Few clusters (N = 4 in total). • Possible contamination from mosquitoes in the intervention cluster being diverted to the control cluster in each village, no buffer zones.
[28]	Degener C.M. 2015	• Clusters paired on baseline mosquito density with intervention randomly assigned to one cluster per pair. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • Intervention and control arms were similar at baseline. • Active case finding.	• No buffer zones, small cluster size (possible contamination). • Possibly low mosquito trapping efficiency. • Post intervention increase in mosquito numbers observed with a decrease in mosquito numbers observed in control clusters. • Low participation rate among intervention cluster households. • Possibly too few traps used per intervention house. • Intervention participants may have relaxed anti-dengue measures having felt reassured by the MosquiTRAP traps (false reassurance). • Fewer control participants in serological survey. • Dengue virus IgM not evaluated at baseline. • The study was performed during a time of low dengue transmission.
[23]	Yapabandara AM. 2001	• 1.5km buffer zones between clusters. • Passive case surveillance but access to primary care enhanced. • Clusters stratified on baseline malaria incidence with intervention randomly assigned to half the clusters in each strata. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • High community engagement.	• Contamination risk from migrating mosquitoes and people–highly mobile human population. • Area is not representative of other parts of Sri Lanka.
[24]	Yapabandara AM. 2004	• 1.5km buffer zones between clusters. • Passive case surveillance but access to primary care enhanced. • Clusters stratified on baseline malaria incidence with intervention randomly assigned to half the clusters in each strata. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • High community engagement. • Large study area which was representative of other parts of Sri Lanka.	• Residual house spraying with lambdacyhalothrin occurred in both study arms during the study period (in June and November each year) as part of a government campaign, and this may have diluted the intervention effect. • Pyriproxyfen was not applied to paddy fields. • Contamination risk from migrating mosquitoes and people–highly mobile human population.
[25]	Sluydts V. 2016	• Clusters stratified by malaria endemicity and population size at baseline with intervention randomly assigned to half the clusters in each strata. • High community engagement. • Active case finding. • Primary outcomes measured clearly and reliably.	• Insufficient statistical power to show an effect from the intervention. • Medical treatment for malaria cases may have reduced case numbers detected at future survey points. • Possible poor compliance with topical repellent use (maybe as low as 15% compliance) in the intervention arm with no direct confirmation of entomological endpoint (reduced blood feeding). • Contamination risk from migrating mosquitoes.
[26]	Hill N. 2007	• Active case finding. • High compliance (98.5%) with the study intervention in the placebo arm and the intervention arm. • Similar loss to follow up across both study arms. • Randomisation used to assign intervention to households. • Outcomes measured clearly and reliably. • Trial design and statistical methods appear appropriate. • High community engagement. • Placebo controlled.	• An unexpected round of outdoor fogging with lambdacyhalothrin was performed by some health districts governments mid-way through the trial, affecting some clusters in both study arms. • Low incidence of P. falciparum observed during the study period. • Contamination risk from migrating mosquitoes. • No direct assessment of repellent effectiveness (such as through reduced feeding by mosquitoes)