An AI-powered, human-in-the-loop, real-time registry platform to enable clinical trials, pragmatic studies, and prospective external control arms (ProECA) at scale in US community practices.

Abstract

11058 Background: Drug development is requiring novel approaches such as prospective external controls to reduce development timelines and de-risk phase transitions. The majority of oncology patients in the US are seen at community oncology centers, many of which have limited clinical research infrastructure. The Kaleido registry was initiated to address these challenges through a hybrid operational technology model combining on-site staff, clinical note preparation, pre-visit abstraction, and automated workflows to facilitate pre-screening, data capture and research operations. Methods: The Kaleido registry enrolls patients at US community oncology practices with variable research infrastructure and is actively enabling seven studies. Real-time abstraction is performed prior to each patient visit across an expanding set of indications, including non-small cell lung cancer (NSCLC), colorectal cancer (CRC), prostate cancer, and myeloproliferative neoplasms, enabling standardized data collection and systematic/unbiased trial prescreening. An automated prescreening tool was developed and validated against manual chart review. Standardized draft notes are generated to reduce clinical documentation burden. A patient questionnaire captures social determinants of health (SDOH) and self-reported performance status. Results: As of 1/1 2026 more than 16,000 patients (60% female, median age 70 years, range 19-105) from 28 sites have been consented. Most common diagnoses are breast (29%), colorectal (9%), prostate (6%) and lung cancer (6%). 99% of approached patients filled out the patient questionnaire and missingness of race/ethnicity/education/marital status was <4%. Comprehensive biomarker testing exceeded 85% for patients with metastatic NSCLC. Time from specimen collection to NGS results (median 18 days, range 5-45) or PD-L1 (median 21 days, range 4-45) was a key driver of delays in evaluating trial-eligibility. Patients were consistently identified prior to starting next line of therapy. The model identified additional trial-eligible patients beyond standard workflows even at high-performing sites. In 2025, more than 500 patients were included in trials, ProECAs and pragmatic studies with enhanced data depth within the registry. Conclusions: The Kaleido registry demonstrates a scalable, integrated infrastructure for real time patient identification and comprehensive prospective data collection in the community setting. By facilitating community trial enrollment, pragmatic studies and prospective external controls while reducing operational burden of clinical teams, Kaleido has the potential to significantly accelerate drug development.