Intergroup comparability is of paramount importance in clinical research since it is impossible to draw conclusions on a treatment if populations with different characteristics are compared. While an adequate randomization process in randomized controlled trials (RCTs) ensures a balanced distribution of subjects between groups, the distribution in observational prospective and retrospective studies may be influenced by many confounders.
Propensity score (PS) is a statistical technique that was developed more than 30 years ago with the purpose of estimating the probability to be assigned to a group. Once evaluated, the PS could be used to adjust and balance the groups using different methods such as matching, stratification, covariate adjustment, and weighting. The validity of PS is strictly related to the confounders used in the model, and confounders that are either not identified or not available will produce biases in the results. RCTs will therefore continue to provide the highest quality of evidence, but PS allows fine adjustments on otherwise unbalanced groups, which will increase the strength and quality of observational studies.