Chapter 9 188 research in an understandable manner, and we develop a conceptually simple and easy-to-use combination of the modified power-prior with propensity score weighting. In addition we give a detailed interpretation of the entirety of ’hybrid’ approaches in the framework of causal estimands. Finally, we evaluate our methods through simulation and a case study by jointly analyzing the trial and expanded access program of vemurafenib in the treatment of metastatic melanoma. The majority of aforementioned applications focus on the integration of external (historical) controls with current trial controls. Limited attention has been devoted to research on augmenting current treatment arms with external treatment arms. This lack of research may in part be attributed to the focus on trial design for regulatory product approval. After all, it may be difficult to find an external data set on active treatment usage before the product is readily on the market. Nonetheless, these data may be available through expanded access programs. In expanded access (also known as compassionate use or early access), patients who are ineligible for registered treatment options and ongoing trials may be granted access to active, unlicensed treatments prior to regulatory approval. Expanded access pathways have become increasingly popular in recent years, and data generated through expanded access form a substantial and increasing area of academic literature - especially due to the COVID-19 pandemic. Moreover, the analyses of such access programs have been integrated into regulatory and cost-effectiveness decision-making.37,38 However, the statistical literature has not yet focused on models designed for the analysis of these types of programs and through this paper, we aim to make a first contribution to this area. The remainder of this paper is organized as follows. Section 2 discusses the background of propensity scores, dynamic borrowing, and hybrid methods. Section 3 details our new proposed method. Section 4 evaluates our method with a simulation study, and Section 5 illustrates our method with a real-life expanded access program and trial. Finally, Section 6 concludes with a discussion.
RkJQdWJsaXNoZXIy MTk4NDMw