Abstract:
|
To obtain marketing approval of a therapeutic product, the statutes require manufactures to demonstrate substantial evidence of effectiveness through the conduct of adequate and well-controlled studies. What constitutes adequate and well-controlled studies is usually interpreted as the conduct of randomized controlled trials (RCTs). However, these trials are sometimes unfeasible due to their size, duration, cost, or ethical constraints. In this case, data derived from external control can be useful to complement information provided by RCTs. Propensity score methods can be employed to balance groups by matching the new treatment and control units based on a set of measured covariates. In this talk, different data augmentation schemes based on propensity scores are explored. These techniques are applied to real clinical data examples to augment data in a trial where the randomization is disproportionate, utilizing historical active controls information. Explorations on how much data augmentation is also considered. The simulation experiment results show that data augmentation improves the balance of covariates and the accuracy of estimation of the average treatment effects.
|