Objective: Limited diversity of patients in oncology trials hinders equity of access and affects result generalizability. This study aims to 1) quantify eligibility-based representativeness, with a focus on Black patients and 2) identify factors affecting eligibility-based representativeness.

Method: We assessed how representative eligibility criteria were in terms of race for 10 phase III non-small cell lung cancer trials using log disparity (LD), a machine learning fairness metric. We empirically evaluated individual eligibility criteria to formulate hypotheses to increase representativeness.

Results: Using data from Flatiron Health’s nationwide electronic health record (EHR)-derived de-identified database, we mapped trial eligibility and therapy indication criteria to define trial eligible (n=160–27,888) and target indication populations (n=3,379–63,637), respectively for each trial. We compared the distribution of Black patients between the trial eligible and target indication populations. Most trials had a LD <100% (ie parity) suggesting that enrolling representative study samples of the target population could be a challenge. For all trials, the LD for Black patients was ≥80% (median:92.8%; range:84.0–143.5%; n [trials]:10). If the representativeness benchmark is set to 80% (a criterion used by the US Equal Employment Opportunity Commission), each trial eligible population would be considered representative of the distribution of Black patients in the target indication population. With a more meaningful criteria of LD ≥95% 3 of 10 trials would be considered representative. To explore factors affecting trial eligibility for enrolling Black patients, we removed eligibility criteria one-at-a-time and evaluated the percent change in LD. Our analysis revealed that removing criteria related to hemoglobin (median:2.7%; range:1.4–6.2%; n:9), creatinine/creatinine clearance (median:1.3%; range:0.3–10.5%; n:9), and ECOG Performance Status (median:1.3%; range:-0.1–4.8%; n:10) increased representativeness of the trial eligible population the most. However, removal of other criteria related to progression on select treatments (median:-3.0%; range:-7.1–2.0%; n:6), CNS metastases (median:-0.4%; range:-1.3–0.0%; n:9), and platelets (median:-0.4%; range:-1.7–0.5%; n:9) resulted in the largest decrease in representativeness. These results highlight the balance needed between optimizing for representativeness vs. patient counts; removing eligibility criteria will increase the number of eligible patients but may not increase eligibility-based representativeness in all situations.

Conclusion: The degree to which clinical trials are representative directly impacts the applicability of their outcomes to target indication populations. Appropriate scoring systems are essential to quantify the degree of representativeness and identify opportunities for improvement of trial diversity. Our analysis found that eligibility-based representativeness is consistently below parity for Black patients in the 10 selected cross-sponsor trials. This result indicates that inequities in eligibility criteria exist providing clinical trial scientists and operations an opportunity to undertake these assessments at the time of protocol development, planned or unplanned interim analysis points, or protocol amendments in an effort to maintain equitable eligibility. Our results identified eligibility criteria that could potentially be modified to increase eligibility-based representativeness, suggesting that selected eligibility criteria disproportionately affect Black patients. This analysis could be extended to identify similar criteria for other patients from historically marginalized groups. Thereby it is imperative that trial design approaches the implementation of eligibility criteria from a health equity lens. Reported Black patient enrollment for these evaluated trials obtained from public sources ranged from 0-3%, substantially below the anticipated rate of ~10%, indicating that the disparities in enrolled populations cannot be fully explained by eligibility criteria alone. This discrepancy underscores the urgent need for innovative strategies to address enrollment diversity as well as the need to optimize trial eligible populations to be enriched for patients from historically marginalized groups in order to reach equitable participation. Making clinical trials more equitable can improve their broad applicability as well as provide more robust evidence for payers and regulators.

**Objective**: Limited diversity of patients in oncology trials hinders equity of access and affects result generalizability. This study aims to 1) quantify eligibility-based representativeness, with a focus on Black patients and 2) identify factors affecting eligibility-based representativeness.

**Method:** We assessed how representative eligibility criteria were in terms of race for 10 phase III non-small cell lung cancer trials using log disparity (LD), a machine learning fairness metric. We empirically evaluated individual eligibility criteria to formulate hypotheses to increase representativeness.

**Results:** Using data from Flatiron Health’s nationwide electronic health record (EHR)-derived de-identified database, we mapped trial eligibility and therapy indication criteria to define trial eligible (n=160–27,888) and target indication populations (n=3,379–63,637), respectively for each trial. We compared the distribution of Black patients between the trial eligible and target indication populations. Most trials had a LD <100% (ie parity) suggesting that enrolling representative study samples of the target population could be a challenge. For all trials, the LD for Black patients was ≥80% (median:92.8%; range:84.0–143.5%; n [trials]:10). If the representativeness benchmark is set to 80% (a criterion used by the US Equal Employment Opportunity Commission), each trial eligible population would be considered representative of the distribution of Black patients in the target indication population. With a more meaningful criteria of LD ≥95% 3 of 10 trials would be considered representative. To explore factors affecting trial eligibility for enrolling Black patients, we removed eligibility criteria one-at-a-time and evaluated the percent change in LD. Our analysis revealed that removing criteria related to hemoglobin (median:2.7%; range:1.4–6.2%; n:9), creatinine/creatinine clearance (median:1.3%; range:0.3–10.5%; n:9), and ECOG Performance Status (median:1.3%; range:-0.1–4.8%; n:10) increased representativeness of the trial eligible population the most. However, removal of other criteria related to progression on select treatments (median:-3.0%; range:-7.1–2.0%; n:6), CNS metastases (median:-0.4%; range:-1.3–0.0%; n:9), and platelets (median:-0.4%; range:-1.7–0.5%; n:9) resulted in the largest decrease in representativeness. These results highlight the balance needed between optimizing for representativeness vs. patient counts; removing eligibility criteria will increase the number of eligible patients but may not increase eligibility-based representativeness in all situations.

**Conclusion:** The degree to which clinical trials are representative directly impacts the applicability of their outcomes to target indication populations. Appropriate scoring systems are essential to quantify the degree of representativeness and identify opportunities for improvement of trial diversity. Our analysis found that eligibility-based representativeness is consistently below parity for Black patients in the 10 selected cross-sponsor trials. This result indicates that inequities in eligibility criteria exist providing clinical trial scientists and operations an opportunity to undertake these assessments at the time of protocol development, planned or unplanned interim analysis points, or protocol amendments in an effort to maintain equitable eligibility. Our results identified eligibility criteria that could potentially be modified to increase eligibility-based representativeness, suggesting that selected eligibility criteria disproportionately affect Black patients. This analysis could be extended to identify similar criteria for other patients from historically marginalized groups. Thereby it is imperative that trial design approaches the implementation of eligibility criteria from a health equity lens. Reported Black patient enrollment for these evaluated trials obtained from public sources ranged from 0-3%, substantially below the anticipated rate of ~10%, indicating that the disparities in enrolled populations cannot be fully explained by eligibility criteria alone. This discrepancy underscores the urgent need for innovative strategies to address enrollment diversity as well as the need to optimize trial eligible populations to be enriched for patients from historically marginalized groups in order to reach equitable participation. Making clinical trials more equitable can improve their broad applicability as well as provide more robust evidence for payers and regulators.

Towards Equitable Participation in Oncology Trials: Evaluating Eligibility-Based Representativeness for Black Patients

DIA Europe 2024 will focus on sustainability, innovation and collaboration – three key pillars for shaping the future of healthcare. From the environmental impact of pharmaceuticals and of the drug development lifecycle to sustainable partnerships and sustainable healthcare systems, DIA Europe 2024 will be the prime opportunity to discuss how collaboration and innovation can support sustainability in healthcare and each stakeholder’s role in making it happen.

Join us in the vibrant city of Brussels for an unforgettable experience, where the power of collaboration will be on full display to inspire actionable outcomes and mobilise for solutions the life sciences’ community. Witness first-hand the profound impact of united efforts, gain valuable insights from across the drug development lifecycle, and play an active role in shaping the future of the pharmaceutical arena.

DIA Europe 2024

A global, member-driven non-profit organization mobilizing life sciences and healthcare. DIA fosters innovation to improve health and well-being worldwide. Professionals from 80 countries continue to engage with DIA through our unparalleled membership network, educational offerings, and professional development opportunities.