Physician- and Practice-Level Variation in Clinical Trial Participation Among Patients With Advanced Cancer.

Abstract

Importance: Clinical trial enrollment is essential to advancing cancer care, yet participation among patients with advanced cancer remains low. While patient-level barriers are frequently cited, the role of health care delivery factors-specifically physician- and practice-level variation-has been poorly quantified.

Objective: To quantify variation in clinical trial participation and identify associated delivery system factors among patients with advanced cancer.

Design, setting, and participants: This cohort study included adults with advanced or metastatic cancers diagnosed and treated between January 1, 2011, and December 31, 2024, by 1664 physicians across 107 trial-available oncology practices across the US. Data were obtained from the Flatiron Health Research Database. Analyses were conducted between January and April 2025.

Exposures: Patient-level (eg, demographic characteristics, cancer type, and performance status), physician-level (eg, patient volume, subspecialization, and prior trial enrollment activity), and practice-level (eg, size, research engagement, and patient composition) factors.

Main outcomes and measures: The primary outcome was clinical trial participation, defined as receipt of a clinical study drug. Mixed-effects logistic regression models were used to decompose the proportion of variation attributable to patient-, physician-, and practice-level factors. Multivariable logistic regression was used to identify characteristics associated with trial participation.

Results: Among the 205 743 patients with advanced or metastatic cancer (mean [SD] age at advanced cancer diagnosis, 67.5 [11.3]; 107 370 males [52.2%]), 10 785 (5.2%) participated in a clinical trial over the study period, increasing from 235 of 5964 (3.9%) in 2011 to 638 of 9336 (6.8%) in 2024. Participation remained highly concentrated: of 1664 physicians, 389 (23.4%) accounted for 9018 (83.6%) of all enrollments (10 785), while 21 of 107 practices (19.6%) accounted for 8686 (80.5%) of all enrolled patients. Additionally, 494 of 1664 physicians (29.7%) enrolled no patients. Physician-level (12.3%) and practice-level (15.1%) factors explained a greater share of variation in trial participation than observable patient-level characteristics (11.4%), with physician- and practice-level contributions increasing over time. Physician's subspecialization (treated 1-2 cancer groups: odds ratio [OR], 4.11 [95% CI, 2.53-6.68], P = .002; treated 3-4 cancer groups: OR, 2.38 [95% CI, 1.37-4.14], P < .001), physician's prior trial enrollment activity (enrolled patients in the 2 previous years: OR, 1.58; 95% CI, 1.43-1.76; P < .001), and practice's research engagement (OR, 2.68; 95% CI, 2.12-3.39; P < .001) were associated with higher odds of trial participation.

Conclusions and relevance: In this cohort study of patients with advanced cancer, clinical trial participation increased moderately but remained low over the study period, with physician- and practice-level factors consistently accounting for a substantial share of participation variation. Improving trial accrual requires interventions targeting structural factors in enrollment, including research infrastructure and physician engagement.