Evaluating The Role Of Real-World Evidence For Oncology Product Registration In The United States: A Systematic Review Of Approvals By The Food And Drug Administration From 2015-2019

Use of real-world evidence (RWE) to support clinical development and serve as historical controls for trials has increased in oncology regulatory filings. Although frameworks exist describing the role of RWE, concrete examples are dearth. This study aimed to systematically review FDA approvals for oncology products to identify use cases of RWE. We systematically reviewed FDA’s oncology new drug application (NDA) and biologic license application (BLA) approvals, stratified by 2015-2016 (pre-Cures Act) vs. 2017-2019 (post-Cures Act), to identify those involving RWE. FDA’s critique of methodology and RWE-specific information (data source, comparability of trial and real-world populations, endpoint definitions, and study design) were extracted and synthesized. 93 NDA/BLA approvals were identified; 0 included RWE in 2015-2016 while 4 included RWE (all retrospective analyses) in 2017-2019, in support of efficacy claims (review is ongoing). Mean time from investigational new drug submission to approval was 6.0 years; among the 4 including RWE, mean was 4.5 years and all 4 had priority review designation. Avelumab (2017) used McKesson iKnowMed data with endpoints on overall response rate (ORR) and duration of response. Axicabtagene ciloleucel (2017) pooled data from trials and observational studies with meta-analysis to assess ORR, complete response, and overall survival (OS). Selinexor (2019) analyzed Flatiron Health Analytic Database to evaluate OS. Erdafitinib (2019) assessed OS, rwORR, and rw disease-control rate in Flatiron-Foundation Medicine Clinico-Genomic Database. Beyond contextualizing trial findings, RWE compared selinexor and erdafitinib trial patients to real-world patients. FDA critiques of RWE included different practice settings/countries in data sources, incomparable biased populations to trial, missing key clinical variables, and uncontrolled confounding, but acknowledged that limitations inherent to RWE are challenging to overcome. Successful examples of approvals involving RWE involve a priori protocol development, early engagement with FDA on design components, and careful research design to minimize sources of bias.