P1749 - Impact of Artificial Intelligence Enhanced Colonoscopy in a Clinical Practice Setting

Maryam Alahmad, MBBS¹, Youstina A.. Ghoris, BSc¹, Colleen T. Ball, MS², Mutaz I. Atieh, ¹, Heba Abualkas, MBBS¹, Darshan Nagaraju, MS³, Mohammed Karajeh, MBChB¹, Michael B. Wallace, MD, MPH^4
1Sheikh Shakbout Medical City, Abu Dhabi, Abu Dhabi, United Arab Emirates; ²Mayo Clinic, Jacksonville, FL; ³Sheikh Shakbout Medical City, Abudhabi, United Arab Emirates/Mayo Clinic, Rochester, MN, Abu Dhabi, Abu Dhabi, United Arab Emirates; ⁴Sheikh Shakbout Medical City, Abudhabi, United Arab Emirates/Mayo Clinic, Jacksonville, FL, Abu Dhabi, Abu Dhabi, United Arab Emirates

Introduction: Colonoscopy reduces the incidence and death from colorectal cancer. There is a strong association between quality, as measured by adenoma detection rate (ADR) or per colonoscopy (APR) and preventive benefit. Randomized controlled trials of Artificial intelligence (AI) tools demonstrate improved ADR, APC and reduce adenoma miss rates (Wallace et al. Gastroenterology; 2022). Little data is available about the effect of AI systems in clinical practice but some evidence suggests lower benefit (Uri Ladabaum, et al, Gastroenterology, 2023). The objective of this study is to demonstrate the effectiveness of colonoscopy with AI compared to colonoscopy without AI in a clinical practice setting in a region with a screening age of 40 years and higher.

Methods: We conducted a retrospective case control registry study of all patients who had colonoscopy with AI immediately after implementation in our practice in 2022 at Sheikh Shakhbout Medical City (SSMC) in Abu Dhabi, a tertiary care hospital based in the United Arab Emirates. One of four endoscopy rooms were outfitted with AI, and assignment to AI vs non-AI rooms were pseudo-random, based on schedule and space factors. Cases performed with AI were matched, based on endoscopist, gender, and indication, with approximately 4 control patients who had colonoscopy without AI during the same period. The primary outcomes were APC and ADR comparing colonoscopy with and without AI.

Results: We included 511 patients who had screening or surveillance or diagnostic/screening colonoscopy with or without AI. We excluded 69 patients who had inadequate bowel preparation (Boston Bowel Prep < 6). Our final analysis included 442 patients (97 with AI, 345 without AI).

The median age was 49 years (IQR 42 to 62 years) with AI and 48 years (IQR 42 to 60 years) without AI. A 54 (55.7%) were male with AI and 167 (48.4%) without AI.

The mean APC was 0.57 +/-1.18 with AI and 0.30 +/-0.83 without AI (unadjusted risk ratio: 1.88, 95% CI 2.35-2.60, p=0.001). Likewise, the ADR with AI was also 27.8% vs. 17.7% without AI; RR 1.57 91.04-2.30, p 0.023). After adjusting for age, gender, and indication the adjusted risk ratio for APC was of borderline significant 1.39 (95% CI 0.97-1.98, p=0.069).

Discussion: In a clinical practice setting, AI enhanced colonoscopy increased ADR and APC although with a borderline significant impact when controlling for endoscopist, patient gender and indication.


Disclosures:


Maryam Alahmad, MBBS¹, Youstina A.. Ghoris, BSc¹, Colleen T. Ball, MS², Mutaz I. Atieh, ¹, Heba Abualkas, MBBS¹, Darshan Nagaraju, MS³, Mohammed Karajeh, MBChB¹, Michael B. Wallace, MD, MPH⁴. P1749 - Impact of Artificial Intelligence Enhanced Colonoscopy in a Clinical Practice Setting, ACG 2023 Annual Scientific Meeting Abstracts. Vancouver, BC, Canada: American College of Gastroenterology.