Has racial and ethnic representation in oncology clinical trials changed?: A 30-year systematic analysis using artificial intelligence.

11079 Background: Despite societal emphasis on racial and ethnic representation in oncology trials, it remains unclear whether participant diversity has meaningfully changed over time. Recent advances in large language models (LLMs) enable scalable extraction of data from PDF documents, facilitating high-throughput research. Methods: We identified all oncology clinical trials published in The Lancet and The New England Journal of Medicine (NEJM) between 1995 and 2025, and retrieved full-text PDFs from the journals' websites. Using an LLM (Gemini-2.5-Flash), structured variables, including trial phase, cancer type, enrollment countries, and participant race and/or ethnicity, were automatically extracted from texts, tables, and figures. Trials were stratified by decade (1995–2005, 2006–2015, and 2016–2025) to evaluate temporal trends. Results: We identified 1,385 oncology trials between 1995 and 2025, including 334, 448, and 603 for the three decades, respectively. The proportion of phase III trials increased steadily over time (68%, 72%, and 76%). Trials also became more globally distributed, with multi-continental studies rising from 27% to 46% and 58%. Most notably, Asian countries were increasingly involved among trial enrollment sites, rising from 14% to 25% and 45%. Reporting of race/ethnicity increased over time, from 16% to 29% and 47% of publications across decades. Among studies that reported race, non-White enrollment increased, with the median percentage per study rising from 10% in both earlier decades to 21% in 2016–2025. However, after excluding Asian participants, the median non-White percentage declined across decades (7.5%, 5.3%, and 4.3%). In pooled analyses across all trials, non-White patients comprised 15%, 11%, and 15% of total enrollment, respectively; excluding Asian patients, this decreased from 13.2% to 6.2% and 6.9%. Black enrollment declined over time, with the median percentage falling from 6.8% to 2.4% and 1.5%. Among trials reporting ethnicity, Hispanic enrollment remained consistently low (5.4%, 4.0%, and 4.6%). Conclusions: Globalization and race/ethnicity reporting in oncology trials improved, but the observed increases in diversity were largely attributable to the expansion of trial enrollment in Asia. In contrast, Black and Hispanic representation remained low or declined across decades. 1995-2005, n (%) 2006-2015, n (%) 2016-2025, n (%) Total Studies 334 448 603 Clinical Trial Phase Phase 1 78 (23) 86 (19) 55 (9) Phase 2 25 (7) 38 (8) 87 (14) Phase 3 228 (68) 323 (72) 460 (76) Major Cancer Types Breast 55 (17) 64 (14) 78 (13) Hematologic 54 (16) 77 (17) 114 (19) Colorectal 49 (15) 40 (9) 29 (5) Lung 17 (5) 37 (8) 66 (11) Continents Africa 16 (5) 29 (6) 16 (3) Asia 47 (14) 113 (25) 274 (45) Europe 158 (47) 272 (61) 403 (67) North America 102 (31) 219 (49) 361 (60) South America 0 (0) 5 (1) <jats:td colspan=