Towards a more systematic evaluation of cancer endpoints: A method for synthesis and analysis of NSCLC regulatory filings and clinical evidence.

Abstract

e18681 Background: Ongoing debate surrounds the use of surrogate and proxy endpoints in oncology such as Progression Free Survival (PFS) and Overall Response Rate (ORR); the oncology community has debated the utility, meaningfulness, validity, and evidentiary definitions of such outcomes. Literature across disease states points to potential inconsistencies in the reporting of, and definition of, clinical trial endpoints more broadly. No data source currently exists to robustly analyze how endpoints are defined, deployed, and applied across a number of trials within a particular tumor type, let alone across oncologic diseases more broadly. This study presents a systematic method to build such a dataset using regulatory filings and clinical evidence publications. Methods: This study developed a five-step method to synthesize and analyze data in a systematic manner. Given the number of new therapeutic approvals in NSCLC, it was selected as a disease area of focus to test the method. For each step of the method, data was extracted and recorded in a Google Sheets database. Step 1 identified therapy FDA approvals from 2016-2020 via systematic search of the FDA website. Step 2 searched the Drugs@FDA database to identify approval packages. Step 3 synthesized the clinical evidence supporting approvals through systematic analysis of FDA Review Documents. Step 4 identified the published protocol for each study from Step 3 using ClinicalTrials.gov. Step 5 identified publications reporting results from Step 4 using the National Clinical Trial identifier numbers and keywords. Results: A systematic search and synthesis of FDA and clinical trial documentation in this manner yielded a dataset of six first approvals for NSCLC and 27 line extensions for NSCLC indications between 2016 and 2020. Conclusions: Informed debate around the use of proxy and surrogate endpoints is difficult to achieve without an evidentiary basis for the discussion. At present, no data source exists for researchers to answer questions such as: “How frequently was PFS used as the primary endpoint in registrational trials for lung cancer over the past ten years?” By developing and deploying a systematic method as described above, researchers can build such a dataset, and then apply the same method across additional disease states and tumor types to understand a more comprehensive picture of how endpoints are defined, measured, and applied in clinical trials. Not only can such a dataset enable research that informs clinical trial design, but it can also enable research and policy development regarding endpoint fidelity, utility, and meaningfulness.