A national, patient-centered AI framework for equitable oncology clinical trial access: Early outcomes from the first 10 months of ACS ACTS.

Abstract

e13573 Background: Less than 10% of eligible cancer patients enroll in clinical trials (CT) due to fragmented information, logistical barriers, and inequities in trial awareness and navigation. To address these challenges, the American Cancer Society (ACS) launched ACS ACTS (Access to Clinical Trials & Support), a national, patient-centered program combining clinical trial education, health-related social needs (HRSN) screening, navigation, and artificial intelligence (AI)–enabled trial matching to reduce friction between trial interest and trial action. Methods: We conducted a descriptive analysis of the first 10 months of ACS ACTS, from launch on 2/23/25 to 12/31/25. Eligible participants (EP) included individuals of any age with any cancer type across all U.S. states & territories, referred by patients, caregivers, or providers. Program components included trial education, HRSN screening, supportive services, AI-generated personalized trial matching (via Massive Bio–sourced interventional trials), and centralized prescreening hubs (CPH) for eligibility and site coordination. The AI platform employs a neurosymbolic, multi-agent, explainable architecture integrating rule-based eligibility reasoning with machine learning–based semantic extraction of clinical data to align patient characteristics with protocol-level inclusion & exclusion criteria. Outcomes included reach, HRSN burden, AI matching throughput, trial matches, and downstream trial referral metrics. Results: 1,479 EP across over 30 cancer types and 48 states participated in the program, with representation from medically underserved and rural communities. 2,713 HRSNs were reported with 54% of EP reporting at least one unmet HRSN, most commonly financial, lodging, and emotional concerns. 66.7% of EP received CT education and HRSN support, and among them 75.0% elected to pursue CT matching, receiving AI-generated personalized trial lists with rapid turnaround. 1,591 partial or exact matches to trials were identified, with 99.9% of patients receiving at least one match. CPH completed for most matched EP and efficient triage to investigative sites. Early downstream outcomes demonstrated meaningful progression from trial awareness to site referral, with subsets advancing to trial screening and enrollment where confirmable. Conclusions: In its early national implementation, ACS ACTS demonstrates the feasibility of a scalable, patient-centered AI framework that integrates education, social needs support, AI-enabled trial matching, and centralized prescreening to reduce barriers to oncology CT participation. Early outcomes highlight substantial unmet social needs alongside strong demand for trial navigation and matching, particularly directly from patients. Ongoing analyses will examine impact on trial participation, equity, and patient engagement.