Regulatory Considerations in the Design of Comparative Observational Studies Using Propensity Scores

Abstract

In the evaluation of medical products, including drugs, biological products, and medical devices, comparative observational studies could play an important role when properly conducted randomized, well-controlled clinical trials are infeasible due to ethical or practical reasons. However, various biases could be introduced at every stage and into every aspect of the observational study, and consequently the interpretation of the resulting statistical inference would be of concern. While there do exist statistical techniques for addressing some of the challenging issues, often based on propensity score methodology, these statistical tools probably have not been as widely employed in prospectively designing observational studies as they should be. There are also times when they are implemented in an unscientific manner, such as performing propensity score model selection for a dataset involving outcome data in the same dataset, so that the integrity of observational study design and the interpretability of outcome analysis results could be compromised. In this paper, regulatory considerations on prospective study design using propensity scores are shared and illustrated with hypothetical examples.

Key Words:

• Causal inference
• Observational study
• Propensity scores
• Regulatory perspective

ACKNOWLEDGMENTS

The author thanks Drs. Greg Campbell, Gerry Gray, Heng Li, and Yunling Xu for their valuable comments during the preparation of this paper. The author is also grateful to the referees for their great suggestions and comments, which helped to improve the article.

Notes

This article is not subject to US copyright law.

No official support or endorsement by the Food and Drug Administration of this presentation is intended or should be inferred.