Abstract
Malaria control uses insecticides to kill Anopheles mosquitoes. Recent successes in malaria control are threatened by increasing levels of insecticide resistance (IR), requiring insecticide resistance management (IRM) strategies to mitigate this problem. Field trials of IRM strategies are usually prohibitively expensive with long timeframes, and mathematical modeling is often used to evaluate alternative options. Previous IRM models in the context of malaria control assumed IR to have a simple (monogenic) basis, whereas in natural populations, IR will often be a complex polygenic trait determined by multiple genetic variants. A quantitative genetics model was developed to model IR as a polygenic trait. The model allows insecticides to be deployed as sequences (continuous deployment until a defined withdrawal threshold, termed "insecticide lifespan", as indicated by resistance diagnosis in bioassays), rotations (periodic switching of insecticides), or full-dose mixtures (two insecticides in one formulation). These IRM strategies were compared based on their "strategy lifespan" (capped at 500 generations). Partial rank correlation and generalized linear modeling was used to identify and quantify parameters driving the evolution of resistance. Random forest models were used to identify parameters offering predictive value for decision-making. Deploying single insecticides as sequences or rotations usually made little overall difference to their "strategy lifespan", though rotations displayed lower mean and peak resistances. Deploying two insecticides in a full-dose mixture formulation was found to extend the "strategy lifespan" when compared to deploying each in sequence or rotation. This pattern was observed regardless of the level of cross resistance between the insecticides or the starting level of resistance. Statistical analysis highlighted the importance of insecticide coverage, cross resistance, heritability, and fitness costs for selecting an appropriate IRM strategy. Full-dose mixtures appear the most promising of the strategies evaluated, with the longest "strategy lifespans". These conclusions broadly corroborate previous results from monogenic models.