Maximising power with the minimum number of mosquitoes: Designing robust sample sizes for the WHO Cone Bioassay

Power calculations are an essential component of experimental design when evaluating vector control tools. Determining appropriate sample sizes for robustly detecting a difference between treatment groups in a bioassay (or any comparative experiment) is complicated by multiple sources of variation. While modern simulation-based methods exist to account for compounding sources of variation, uptake is slow due to limited availability of training and hardware. Here we present an accessible, user-friendly framework for performing sample size calculations for World Health Organisation (WHO) cone bioassays. Additionally, we conduct a literature review of studies published between 1998 and 2024 to identify sources of variability in WHO cone bioassay methodologies.

We use simulation-based methods to assess power in the WHO cone bioassay, utilising the 2013 WHO guidance for phase I laboratory testing of ‘long-lasting’ insecticidal nets as an illustrative example of how sample size impacts detection of differences in mosquito mortality between treatments. Futhermore, we establish plausible variability assumptions across three levels of bioassay variability for the testing of insecticide-treated nets: between nets of the same product, between net pieces from the same net and between replicates on a net piece.

We demonstrate that the biggest factor in determining the number of samples, and therefore mosquitoes, needed in an experiment is the effect size to be detected (mortality difference between treatments). Larger mortality differences (e.g. a 20% difference) are readily detected with the phase I guidance yet detecting a 10% difference requires more than triple this sample size. Here, we present a user-friendly browser application to allow researchers to easily design robust WHO cone bioassay experiments (link: Cone Bioassay Sample Size app).