Discussion of: Statistical and Regulatory Issues with the Application of Propensity Score Analysis to Nonrandomized Medical Device Clinical Studies by Lilly Q. Yue

摘要

We thank the author for a nice review and discussion on the current application of the propensity score analysis on medical device clinical studies. Nonrandomized clinical studies are commonly seen in the medical device area since randomization may not be feasible for either ethical or practical reasons. Propensity score analysis provides an attractive alternative for group comparison and has several advantages over other methodologies such as conventional multiple linear regression (Rubin, 1997). A typical application of this approach is to construct propensity scores using logistic regression and then to perform group comparison using either matching, stratification (subclassification) or regression (covariate) adjustment (D'Agostino, 1998).