Mosquito biting behavior: statistical power and sources of variation in toxicity and repellent bioassays

Compounds thought to be effective against mosquitoes as either 'insecticides' or 'repellents' have recently been shown to contain properties of both, or possess other behavior-modifying actions. Prompted in part by these reports, we conducted posterior analyses of our data to examine some interrelated statistical issues inherent in our bioassay system. Using a modified K&D module, the responses of over 25,000 adult Aedes aegypti (L.) females exposed to either alphacypermethrin or DEET were compared with the responses of mosquitoes exposed to untreated controls for toxicity and biting (alphacypermethrin) or biting alone (DEET). Our analyses indicated that; (1) our bioassay system has more statistical power to determine a compound's toxicity than its repellent qualities, (2) day-to-day variability is large and needs to be accommodated in analyses; there are other, potentially even larger sources of variability (e.g., mosquito heterogeneity) which invalidate statistical tests that are based on the assumption of binomially or multinomially distributed data (e.g., chi2 tests), and (3) unlike biting mosquitoes exposed to DEET, the proportions of biting mosquitoes exposed to alphacypermethrin are unrelated to the proportions of concurrently tested biting controls, even after adjusting for daily variation in toxicity. Thus, there is a clear behavioral indicator in this bioassay system that the 'repellency' of DEET (a presumed repellent) differs in a fundamental way from that of alphacypermethrin (a presumed toxicant), which may allow the differentiation between classes of compounds based on biting behavior alone.