Sample Size Effect in Two-Arm Non-Randomized Trials
*Shiowjen Lee, FDA 


In a non-randomized two-arm clinical trial, sample size calculation is usually based on the formula that as if it was a randomized two-arm clinical trial. For a randomized clinical trial, randomization usually balances patients’ baseline characteristics (observed and unobserved) between the two treatment groups. That is not the case, however, in a non-randomized clinical trial. Therefore, propensity score methodology has been used widely to adjust baseline covariates prior to the treatment comparison. The use of propensity score method has its limitation and depends upon the degree of similarity of baseline covariates between groups. Because of covariate adjustment prior to treatment comparison, sample size determination based on the formula of a randomized two-arm trial may be incorrect in a non-randomized clinical study setting.