Quantifying the impact of experimental hut design on intervention evaluation outcomes and predicted reductions in vectorial capacity

Résumé fourni par la source

Abstract Experimental hut trials (EHTs) are WHO-recommended for the entomological evaluation of insecticide-treated nets (ITNs), but several hut designs are in operational use, and structural differences between them may confound efficacy predictions and limit cross-site comparability. We developed a Bayesian hierarchical framework comprising a host-seeking model, which jointly estimates biting deterrence and preprandial mortality while accounting for night-to-night variation and overdispersion, and a postprandial mortality model, which expresses hut and net effects as hazard ratios through a complementary log-log link. We applied it to a comparative trial of four hut designs (East African, West African, Ifakara and Rapley) conducted at a single site in Tanzania, evaluating eight ITNs when new and after twenty washes. Posterior estimates parameterise a vectorial capacity framework to predict reductions in transmission potential. Hut design influenced baseline mosquito behaviour and all three modes of action. Relative to the Rapley reference, baseline feeding rates were substantially lower in the East African and West African huts and closer to Rapley in the Ifakara hut. Preprandial mortality was amplified in the Ifakara hut. Comparing to previous analysis provides evidence that combining mortality before and after feeding into a single endpoint does not reliably reflect impact, supporting the decomposition of entomological outcomes into separate modes of action. Expressing modes of action as mechanism-specific parameters allows the estimates to be carried directly into transmission models. For every net, the predicted reduction in vectorial capacity was greatest in the Ifakara hut and smallest in the West African and Rapley huts. The effect of hut design on predicted impact exceeded that of washing the nets twenty times. Results indicate that the hut design under which trial data were collected should be considered when forecasting population-level effect.