Representativeness in cancer treatment trials in community and academic settings

e18550 Background: Patients with cancer who participate in clinical trials may not be representative of the broader patient population, which can impact the generalizability of results. However, it is unknown whether practice setting influences representativeness of trial participants. Here we evaluated (1) whether demographic characteristics of patients enrolled in cancer treatment trials at academic sites differ from those in the community and (2) the degree to which trial patient characteristics are representative of the overall patient population seen in each practice setting. Methods: This retrospective study used the nationwide Flatiron Health electronic health record-derived de-identified database, originating from ~280 US cancer clinics (~800 sites of care). We included patients with a cancer diagnosis and clinical visits between 11/1/17 - 11/1/22. Clinical study drug administration was used as a proxy for trial participation. We applied multivariate logistic regression to assess the association of age, sex, race/ethnicity, insurance status and neighborhood socioeconomic status [SES] quintile with practice setting among trial participants, and Pearson 𝛘2 test for independence to assess the association between distribution of patient characteristics in trial participants versus trial non-participants. Results: Of 28,235 trial participants, 13,297 (47%) were treated at academic and 14,938 (53%) at community sites. Among patients who participated in clinical trials, the odds of enrollment at a community versus academic site were significantly higher among patients who were older (75+ odds ratio [OR] 1.91, 95% confidence interval [CI] 1.73-2.11), Female (OR 1.22, CI 1.14-1.31), Black (OR 1.32, CI 1.16-1.50), had Medicaid (OR 1.22, CI 1.07-1.39 ) or were uninsured/unknown (OR 2.05, CI 1.82-2.31) and had the lowest quintile SES (OR 1.78, CI 1.58-2.00), and were significantly lower among patients who were Latinx (OR 0.56, CI 0.48-0.65) or higher ECOG Performance Status (2+ OR 0.75, CI 0.65-0.87). Significant differences in each patient characteristic category existed between trial participants and trial non-participants at both academic and community sites (p < 0.001). Trial participants and trial non-participants were more similar (𝛘2 values less discordant) in sex and race/ethnicity at academic sites, and more similar in insurance and SES categories in the community. Conclusions: When compared to academic centers, community sites had proportionately greater enrollment of older, Black, Medicaid/uninsured and low SES patients. Significant differences existed between trial participants and trial non-participants in both settings for each demographic category. These findings highlight how community site involvement may improve trial representativeness, and underscores the need for targeted solutions to optimize trial representativeness in both academic and community practices.