SA55 Considerations for Common Exclusion Criteria in Real-World (RW) Retrospective Observational Studies in Oncology

Abstract

A key strength of well-designed RW research studies is that they have high potential to include larger and more representative patient populations compared to randomized-controlled trials. While each study generally has unique inclusion and exclusion criteria, two commonly applied exclusion criteria across RW oncology studies involve (1) patient participation in clinical trials or (2) diagnosis with concomitant primary cancers. These two criteria are often specified in an effort to maximize a study’s internal validity (minimize bias), which is prioritized against external validity (generalizability and transportability). However, these criteria are heterogeneously defined across studies, and there may be opportunities to examine their necessity and operational definitions when designing fit-for-purpose research. Examining a series of RW studies based on oncology-specific electronic medical records, we identified key dimensions across which these two common exclusion criteria have been differently defined and examined their impact on patient disqualification. Recognizing a trade-off between sample size, bias reduction, and operational efficiency related to the implementation of these criteria, we propose a conceptual framework for the application of these criteria in future studies. The application of these criteria may have a substantial impact on the internal and external validity of RW studies. Therefore, these criteria should be thoroughly and systematically assessed for each study to ensure selection of the appropriate patient population for the research question being addressed. Customization of eligibility criteria to reduce bias while maximizing sample size and representativeness is a crucially important element of RW study design. Complexity of this task can be reduced for investigators by having a clear and cogent framework for the considerations at play when developing criteria related to trial participation and concomitant primary cancer diagnoses. We encourage investigators to standardize their decision process rather than defaulting to these prevalent, but sometimes restrictive, criteria.