The Role of Different Risk Groups and Treatment Assignment Probabilities in Subgroup Analysis of Randomized Trials

Retrospective subgroup analyses can introduce significant bias in the estimation of hazard ratios (HRs), particularly when patient distributions across treatment arms are imbalanced. Such disparities can distort the validity of HR outcomes, especially in the presence of unequal risk group compositions and varying treatment assignment probabilities. These factors may artificially shift HR estimates across different risk populations, leading to misleading correlations between subgroup classifications and treatment effects. To quantify this phenomenon, we conducted Monte Carlo simulations across 1000 trials, demonstrating how hazard ratios vary systematically with changes in the underlying risk group population. These findings underscore the need for caution in interpreting HRs from subgroup analyses and highlight the importance of robust trial design to mitigate bias.