(079) Synthesized Learnings and Considerations from Global Historical Cohort Studies via Retrospective Chart Review in Oncology

Background: With the number of available treatments across oncology indications, real-world (RW) studies are needed to understand natural history and treatment outcomes. RW oncology studies frequently utilize secondary data sources, however sources may lack clinical outcomes, safety events, lines of treatment and reasons for discontinuation. Retrospective chart review has potential to capture these outcomes.

Objectives: To describe recently undertaken oncology chart review studies to inform on appropriate use-cases, design considerations, and share synthesized metrics and learnings.

Methods: A review of 8 oncology site-based or direct-to-physician historical cohort studies via retrospective chart review (5 completed, 3 ongoing) was undertaken to evaluate objectives, population, outcome measures/variables and data availability.

Results: Of the 8 studies, 4 were US-based, 1 was China-based, 1 was UK-based and 2 were multi-country. Study objectives varied: effectiveness (8 studies), treatment (tx) patterns (8 studies) and safety (4 studies). Oncology indications differed among studies, 1 study was comparative, and 7/8 included adult patients. Prior to study conduct, sample size and/or a qualitative assessment of data completeness in charts was verified through feasibility. Data collected included patient characteristics, medical history, tx patterns, clinical outcomes (e.g., response, progression-free survival [PFS], overall survival [OS]) and adverse events (AEs). Among the 5 completed studies, sample sizes ranged from approximately 100 to 400 patients, and follow-up was (min-max) 3 to 72 months. Data on age and sex had no missingness, data on race/ethnicity had low missingness ( < 5%), data on medical history (i.e., diagnoses, staging, genetic mutation, etc.) had missingness ≤19%. Biomarkers (when used as clinical characteristics or outcomes) had highly varied missingness. Data on tx patterns (i.e., tx lines, tx modifications and reasons) had < 6% missingness. Clinical outcomes data availability varied from 50-100%, and assessment methods varied among sites/physicians within studies (maximum of 12 different methods/study), and commonly included imaging, biochemical assessments and clinical judgement. While the Common Terminology Criteria for AEs (CTCAE) was captured in one study, AE type and severity was still captured in other studies when CTCAE was not available. Due to limitations in methodology, it is unclear if any AEs were not reported.

Conclusions: Chart review studies in oncology populations generated RWE to meet diverse objectives, with a generally high level of data availability. High-quality chart review results are dependent on an understanding of key considerations and limitations.