Clinical Trials Logo

Clinical Trial Summary

We hypothesize that AI-assisted colonoscopy can reduce post-colonoscopy neoplasia incidence after 3 years, over standard colonoscopy. Moreover, this protective effect may allow surveillance intervals to be lengthened, by modifying long-term outcome of high-risk subgroup.


Clinical Trial Description

Recently, rapid development in artificial intelligence (AI) and automatic computer-aided polyp detection (CADe) systems have revolutionized the medical field. Multiple clinical trials reported significant benefits in colorectal neoplasia detection with the use of CADe over standard colonoscopy. The overall ADR and number of adenomas detected per colonoscopy were consistently higher. The performance of CADe was consistent in different levels of endoscopist experience. With the promising data, position statements from World Endoscopy Organization (WEO) and European Society of Gastrointestinal Endoscopy (ESGE) have supported the acceptance and introduction of CADe into clinical practice. However, most clinical trials only demonstrated a net benefit in the detection of small-to-medium size, non-advanced adenomas. A recent large-scale randomized trial failed to show any difference in advanced neoplasia detection rate between CADe and control groups. Furthermore, a meta-analysis of 21 randomized trials showed that CADe only improved adenoma but not advanced adenoma detection. It remains questionable whether AI can overcome all pitfalls and challenges in standard colonoscopies. One of the major concerns before universal implementation of AI-assisted colonoscopy is the lack of real-world long-term data on the clinical efficacy of PCCRC prevention. It remains uncertain whether the detection and removal of small non-advanced adenomas could be translated into absolute clinical benefit. On one hand, patients with more non-advanced adenomas would be classified into higher risk groups according to the US Multi-Society Task Force (USMSTF) guideline, leading to more intensive surveillance colonoscopies, heavier service burden and higher healthcare costs. On the other hand, a high-quality index colonoscopy with fewer missed lesions and low PCCRC risk, could allow patients and clinicians to have higher confidence in lengthening surveillance intervals, leading to an ultimate reduction in the service demand and expenditure in the long run. Unfortunately, most current literature only focused on short-term outcomes and did not address this unsolved problem. Therefore, a prospective real-world cohort to confirm the long-term effectiveness of AI-assisted colonoscopy is urgently warranted. Between April 2021 and July 2022, our group completed a parallel-group, randomized controlled trial in Hong Kong. [ENDOAID-TRAIN study; NCT04838951] 856 subjects undergoing colonoscopies were randomized 1:1 to receive colonoscopies with CADe (ENDO-AID, Olympus Co., Japan) or standard colonoscopies (control). Our study proved that AI-assisted colonoscopies could increase the overall ADR, especially small-to-medium size adenomas. It remains questionable whether the increased detection and removal of these non-advanced adenomas can be translated into any sustained long-term benefit. The impact of this AI-driven intensive surveillance on general population and healthcare system is also largely unknown. In this research project, we aim to assess the long-term effectiveness of AI-assisted colonoscopy on adenoma recurrence and PCCRC prevention, by conducting a real-world, prospective study with longitudinal extension from a randomized trial. ;


Study Design


Related Conditions & MeSH terms


NCT number NCT06251700
Study type Interventional
Source Chinese University of Hong Kong
Contact Felix Sia
Phone 26370428
Email felixsia@cuhk.edu.hk
Status Recruiting
Phase N/A
Start date January 24, 2024
Completion date April 30, 2027