Modelling the impact of spatial repellents, odour-baited traps and push-pull systems on reducing residual malaria transmission and clinical disease

Measures reducing mosquito abundances and mosquito-human contacts (vector control) form the most efficient malaria prevention strategies and have averted over 600 million malaria cases between 2000 - 2015 in Africa. However, progress in reducing malaria case incidence has stalled since 2015 in many African countries despite intensive use of the two front line vector control interventions: insecticide treated nets (ITN) and indoor residual spraying (IRS). It became evident that residual transmission of malaria would persist even if coverage, use and formulation of ITNs and IRS could be improved further. Changes in vector species compositions and behavioural adaptations of the mosquito populations in response to the large-scale implementation of ITNs has shifted mosquito biting and transmission from indoors late at night to the outside areas earlier in the evening. Also, mosquito species across Africa have become increasingly resistant to the insecticides currently used in ITN and IRS formulations. New types of nets with different classes of insecticides and better strategies to counter insecticide resistance are currently being developed; yet these cannot solve the more fundamental problem of an altered transmission landscape. Therefore, new vector control tools addressing outdoor biting and small-scale heterogeneity in transmission modes need to be developed in parallel. Three candidate tools were tested in semi-field experiments in Kenya and Tanzania: A spatial repellent (transfluthrin-treated eave ribbon), a trap (odour-baited Suna trap) and a push-pull system (combination of the spatial repellent and trap). We developed a stochastic-process model of mosquito host-seeking behaviour in semi-field experiments, that is able to infer mosquito mortality from time-stratified data on mosquito-human contact (human landing catch) only. With this semi-field model we estimated the extent to which the spatial repellent reduced mosquito-human contacts by delaying host finding (repelling effect) and by stopping host seeking (either through killing or through disarming, that is, preventing host-seeking for one night). These nuanced effects allowed us to estimate both the personal and the community protection of the spatial repellent. From a small, randomised-block field trial conducted in Ahero (Kenya), we estimated for all candidate tools (i.e. spatial repellent, trap and push-pull system) the reduction of mosquito-human contacts and the change of indoor-outdoor host seeking ratios for the common An. funestus and An. arabiensis species; and for the trap we estimated mosquito removal for both species. We used a hierarchical Bayesian model for this task in order to estimate the variability of the intervention effects across different houses and over time. Based on semi-field and field estimates, and by fully accounting for their uncertainty, we simulated with the OpenMalaria model (an individual-based model for malaria transmission and within-host dynamics in humans) the impact of a large-scale implementation of all candidate tools on clinical malaria in two East African settings. We found a considerable impact of the spatial repellent on malaria incidence by improving indoor protection, especially in settings where highly anthropophilic An. funestus species dominate transmission and where transmission intensity is overall relatively low. The spatial repellent was characterised by a substantial killing or disarming effect in addition to its diversion effect from indoor biting, but provided no protection against outdoor biting. However, we found an antagonistic interaction of the spatial repellent with the ITN and thus suggest its implementation only in settings with insufficient indoor protection by ITN. The trap only showed a marginal impact on malaria case incidence in the given configuration and the push-pull system offered no reasonable benefit over the spatial repellent alone. People who are not protected by an ITN, and who might be difficult to reach with ITN campaigns, may greatly benefit from the spatial repellent. However, our simulations showed that mosquitoes may be pushed towards people who remain protected by neither the spatial repellent nor the ITN, and thus may increase malaria case incidence among this group. We developed a mosquito mark-release-recapture methodology to improve the understanding of mosquito dispersion and elucidate this potential negative community effect of the spatial repellent. In consequence, the considered spatial repellent is a promising supplementary tool to ITNs to protect people at times and locations when they are indoors but not under the net. However, the search for tools to reduce outdoor residual transmission needs to continue. Our modelling framework may be used to estimate the impact of future candidate tools in a variety of target settings based on early semi-field and field data, and thus considerably accelerate the development of new vector control interventions