Introduction: Inflammatory bowel disease clinical trial recruitment faces complex challenges; recent research has show that enrollment has decreased over time. We present the results of a machine-learning supported retrospective chart review recruitment tool in a real-world setting for a Crohn’s disease study with biologic-naive criteria. Retrospective analysis examines the outcomes of the recruitment tool and the recruitment population.
Methods: Five GI-focused endoscopy sites, mixed academic and private, used the tool between July 2023 and February 2024. Sites recorded videos across all installed endoscopy scopes with Virgo devices. Recordings were automatically screened by the vision-transformer machine learning tool. Likely endoscopically-relevant referrals were sent to respective site clinical research coordinators for review on a next-day cadence. Decisions to outreach about trial enrollment were made at the site level by each site primary investigator. Three sites completed prospective and retrospective procedure review from June 1 onward; the fourth site joined in August 2023 without retrospective review, and the fifth in December 2023 with November retrospective review.
Results: In total, 60,146 endoscopy videos were recorded and screened; an estimated 70% (42,102) were colonoscopies. 182 videos were flagged by the machine learning tool as likely-endoscopically relevant; CRCs reviewed 152. Corresponding diagnosis was recorded by CRCs for 98 referrals: 43 referrals corresponded to a patient with a Crohn’s Disease diagnosis. Of these, 4 were biologic-naive and reviewed for further participation, yielding 1 randomization. Of the remaining 39, 16 were excluded only for previous biologic exposure, 10 for additional diagnoses, 3 for other clinical trial enrollment, 1 for age, 1 for unlisted medication, and 5 excluded for unlisted reasons. 2 referrals were undergoing additional review when recruitment closed.
Discussion: Initial results are presented for a small cohort of sites involved in a recruitment support project for a biologic-naive Crohn’s Disease study. Of reviewed referrals, 43.88% (43 of 98) corresponded to a patient with a diagnosis of the disease state of interest, with one resulting biologic-naive randomization in the nine month period. Future investigations should compare such chart reviews to other methods of recruitment over longer time intervals.<br><br>Uzzan M, et al. "Declining Enrolment and Other Challenges in IBD Clinical Trials: Causes and Potential Solutions." J Crohns Colitis. 2023 Jul 5;17(7):1066-1078.
Disclosures:
Safia Speer: Virgo – Employee.
Manisha Cole: Virgo – Employee.
Patrick Dermyer: Virgo – Employee.
Angad Singh: Virgo Surgical Video Solutions – Employee.
Vassilis Mastorostergios: Virgo Surgical Video Solutions, Inc – Employee.
Matthew Schwartz: Virgo – Employee.
Safia Speer, MPH1, Manisha Cole, BS, MS2, Patrick Dermyer, 3, Angad Singh, 4, Vassilis Mastorostergios, 5, Matthew Schwartz, 6. P1519 - Recruitment Support Project Results for Vision-Transformer Model Supported Retrospective Chart Reviews, ACG 2024 Annual Scientific Meeting Abstracts. Philadelphia, PA: American College of Gastroenterology.