Challenges in Estimating Insecticide Selection Pressures from Mosquito Field Data

Insecticide resistance has the potential to compromise the enormous effort put into the control of dengue and malaria vector populations. It is therefore important to quantify the amount of selection acting on resistance alleles, their contributions to fitness in heterozygotes (dominance) and their initial frequencies, as a means to predict the rate of spread of resistance in natural populations. We investigate practical problems of obtaining such estimates, with particular emphasis on Mexican populations of the dengue vector Aedes aegypti. Selection and dominance coefficients can be estimated by fitting genetic models to field data using maximum likelihood (ML) methodology. This methodology, although widely used, makes many assumptions so we investigated how well such models perform when data are sparse or when spatial and temporal heterogeneity occur. As expected, ML methodologies reliably estimated selection and dominance coefficients under idealised conditions but it was difficult to recover the true values when datasets were sparse during the time that resistance alleles increased in frequency, or when spatial and temporal heterogeneity occurred. We analysed published data on pyrethroid resistance in Mexico that consists of the frequency of a Ile1,016 mutation. The estimates for selection coefficient and initial allele frequency on the field dataset were in the expected range, dominance coefficient points to incomplete dominance as observed in the laboratory, although these estimates are accompanied by strong caveats about possible impact of spatial and temporal heterogeneity in selection. Author Summary: The emergence and spread of insecticide resistance compromise the control of mosquito borne diseases such as dengue or malaria, which are responsible for millions of deaths every year in tropical and subtropical areas. There are currently no easily implemented methodologies to quantify the strength of selection for resistance occurring in nature. Using field data from Mexico on the frequency of an allele mutation conferring resistance in the mosquito Aedes aegypti we use maximum likelihood (ML) to estimate the selection and dominance coefficients driving the evolution of resistance. We explored the impact of poor data collection, data that combine information from different locations and the consequences of selection and dominance coefficients varying over the sampling time period. The ML method can accurately estimate these parameters with simulated data in ideal sampling situations but it is difficult to recover true values when spatial and temporal heterogeneity occurs. The analysis highlighted factors relevant to field work such as the need for frequent surveillance in discrete sentinel sites.