Application of principal stratification in crossover and parallel clinical trial designs

Estimating causal effects is crucial in clinical research for assessing treatment efficacy across various study designs. This paper explores causal effect estimation using the crossover and parallel models widely used in clinical trials. We investigate the application of principal stratification within both designs, focusing on key assumptions such as ignorability and no interference between units. Using a real-world example of basal insulin regimens, we compare treatment effect estimates derived from principal stratification with those obtained without these assumptions. Our findings illustrate that these methodologies yield reliable causal effect estimates, improving the accuracy of treatment effect evaluation in clinical trials.