The method of propensity score matching has been widely used in observational studies to balance the observed covariates thus to control for bias. In a randomized clinical trial, randomization is the key to remove the potential bias of allocating subjects to different interventions. However, when interests arise to look at the endpoints within a specific subgroup classfied by on-study outcomes, randomization does not carry over to this subgroup analysis. In this case, matching methods based on estimated propensity scores may provide more comparable tests between subgroups. How do we preserve randomization ratio after matching approach, such as by nearest neighbor matching, if ratio is not 1:1? What are ways to validate the effectiveness of such matching? Or can one trial data be sufficient to validate the matching? What are the risks associated with subgroup matching by propensity scores?