Practical considerations of utilizing propensity score methods in clinical development using real-world and historical data

Abstract

In recent years, with the rapid increase in the volume and accessibility of Real-World-Data (RWD) and Real-World-Evidence (RWE), we have seen the unprecedented opportunities for their use in drug clinical development and life-cycle management. RWD and RWE have demonstrated the significant potential to improve the design, planning, and execution of clinical development. Furthermore, they can feature in the designs as either a substitute or compliment to traditional clinical trials. However, to utilize RWD and RWE appropriately and wisely, it is critical to apply rigorous statistical methodologies that enable the robustness of results to be characterized and ascertained. Several statistical methodologies including exact matching, propensity score methods, matching-adjusted indirect comparisons and meta-analysis have been proposed for analyzing RWD. Among them, propensity score method is one of the most commonly used methods for non-randomized trials with indirect comparison. Although massive methodologies and examples have been published and discussed since propensity score methods were introduced, systematic review and discussion of how to rigorously use propensity score methods in the practical clinical development is still deficient. This paper introduces commonly used and emerging propensity score methods with detailed discussions of their pros and cons. Three different case studies are presented to illustrate the practical considerations of utilizing propensity score methods in the study design and evaluation using real-world and historical data. Additional considerations including selection of patient populations, endpoints, baseline covariates, propensity score methods, sensitivity analysis and practical implementation flow in clinical development will be discussed.

Keywords: Drug development; External control; Historical data; Natural history study; Propensity score; Real-world-data (RWD); Real-world-evidence (RWE).