Role of Covariates in Case Control Studies with Skewed Exposure: Evidence from Monte Carlo Simulations

Abstract

Case-control studies, a widely used observational study design, are essential for investigating the association between exposure and outcomes. In such studies, logistic regression is commonly employed to analyse the relationship between binary outcome and exposure, accounting for covariates, confounders, and effect modifiers. However, skewed exposure distributions, where the exposure is disproportionately distributed among cases and controls, pose significant challenges.
This study aims to address these challenges by conducting a series of Monte Carlo simulation experiments to assess the impact of skewed exposure on the power of the Wald test and the bias in estimated logistic regression coefficients. The simulations focus on the role of continuous covariates in producing reliable estimates of exposure effects. The study highlights the importance of preliminary knowledge of exposure and covariate effects, as these factors play a crucial role in selecting an appropriate sample size. These simulations, which required significant computational time, highlight the robustness of the estimates with larger sample sizes and a greater number of covariates, eliminating the potential bias introduced by skewed exposure.