This tutorial provides guidance on making non-inferiority assessments from experimental hut trial (EHT) data. Code is provided in both R and STATA. You should begin by downloading the files, unzipping the folder, and moving the folder to your desired location on your computer. To use the R code, you should first download R and Rstudio. We recommend opening the R Project file 'WHO_NI_tutorial.RProj' first: this means that your R session in RStudio should begin with the working directory set to the folder containing the relevant files.
This work supports the WHO Report Technical consultation to assess comparative efficacy of vector control products: meeting report, 5 and 9 June 2023.
The R script 'power_calculator_user_script.R' contains a simulation-based method for estimating study power. The script contains a lengthy explanation of how a study should be characterised, which should be read carefully. The methods here have been adapted from earlier work. This work built on a study by Johnston et al.. The main difference between this work and the earlier work is that the simulated datasets are being analysed with generalised linear regression models that contain only fixed effects (no random effects). This is consistent with the tutorial hosted on this repository. Additionally, the methodology used here allows for a non-inferiority margin that varies with the performance of the active comparator product.
Before power can be calculated, the design of the EHT must be specified. As outlined here, the number of mosquitoes entering the hut and the level of variability present in the assay will both influence power (as will the trial duration). The script 'power_calculator_user_script.R' provides details on how the scenario in question should be specified. The functions that this script uses are contained in 'power_calculator_functions_FE.R'.