A framework for extrapolating evidence for targeted therapies in rare biomarker-defined cancers

Abstract

This thesis aimed to develop an approach to assess the clinical benefit of targeted therapies for rare cancers. Despite claims that targeted therapies are superior based on strong biological rationale, a review of randomised controlled trials (RCTs) showed the clinical benefit of targeted therapies is only modest when compared with chemotherapy (Chapter 2). However, conducting adequately powered RCTs using endpoints that capture net clinical benefit may be infeasible in rare cancers. Using molecular profiling, even common cancers are being reclassified into rare biomarker-defined diseases. Many rare cancer trials have single-arm design and benchmark outcomes against historical controls. There are limitations of this approach, as shown by an example in lung cancer, where significant heterogeneity of response rates was observed with docetaxel in the control arms of RCTs (Chapter 3). Extrapolation of evidence from common cancers may inform estimates of the clinical benefit of targeted therapy in rare cancers sharing the same biomarker. A literature review did not identify an explicit guidance for data extrapolation. However, this review identified components essential for extrapolation (Chapter 4). A framework to assess suitability for data extrapolation was developed by outlining key requirements for each component (Chapter 5). Based on this assessment, the definition of disease, traditionally based on histopathological criteria, may shift to group cancer types by an actionable biomarker profile. Natural history data for the biomarker-defined rare cancer is essential for benchmarking outcomes, particularly when the treatment arm was not compared with a concurrent control. The test used to detect the biomarker should be analytically validated in each cancer type, with discussions of issues where a “gold” standard does not exist. Treatment benefit in a rare cancer cohort could be extrapolated from a common cancer provided net clinical benefit has been demonstrated in the latter and signals of efficacy on validated surrogate measures are similar between the cancer types. Whenever possible, randomised evidence should still be sought, using trial designs such as basket studies of multiple rare cancers sharing same biomarker. Signals of efficacy of targeted therapy could be evaluated against chemotherapy based on validated surrogate endpoints. The safety profile of the targeted therapy may differ between cancer types and strategies to augment safety data for rare cancers are proposed. Practical application of this framework was demonstrated using the example of trastuzumab in HER2-positive rare cancers (Chapter 6). This work could promote standardized decision making for regulatory approvals and clinical recommendations. It will serve as a foundation for future work, help identify research priorities to address knowledge gaps and stimulate ongoing discussion between key stakeholders.