Although randomized trials are preferred scientifically and for regulatory purposes, non-randomized study designs are sometimes the only reasonable option and a number of practical considerations arise. Among these are the following: 1. As regulators, what do we say with regard to choice of the comparison group? 2. As regulators, do we recommend propensity score stratification or direct use of covariates or some combination such as covariates within propensity score strata? 3. If we recommend some direct use of covariates, what should we recommend with respect to covariate selection? 4. How do the above considerations change with larger or smaller overall study sample size?