Post written by Chae Min Michelle Lee, MD, MEng, from the Division of Gastroenterology and Hepatology, Department of Medicine, University Health Network, and Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada.

Numerous trials have been conducted to examine the role of artificial intelligence (AI) assistance in polyp detection during colonoscopy. Available AI detection tools rely on a convolutional neural network—based algorithm for image recognition, and their use generally demonstrates improved adenoma detection rates. Different methodologies, namely parallel and tandem designs, have been used to evaluate the efficacy of AI-assisted colonoscopy in these randomized controlled trials (RCTs).

Our study explored whether study design characteristics moderate between-trial differences in reported outcomes when evaluating the effectiveness of AI-assisted colon polyp detection.

Systematic reviews of the literature have researched pooled results of the performance of AI in polyp detection. Our study uniquely focused on the implications of trial design characteristics, which had yet to be investigated. AI continues to have increasing applications in the field of endoscopy, so we felt that it is important to shed light on how research methods may impact trial outcomes.

Our study included 24 RCTs, with 18 parallel RCTs and 6 tandem trials. We found that regardless of the study design, AI-assisted polyp detection tools helped improve adenoma detection rate. This suggests that tandem studies may be just as good as parallel studies, which are typically more resource-intensive, when evaluating adenoma detection rate.

Worth further noting are some of our other outcome measures, such as the prolongation of withdrawal time seen with AI assistance only in parallel studies. In tandem studies, withdrawal time did not change with the application of AI technology. One possible explanation for these data is that endoscopists in tandem studies are more motivated to closely inspect the mucosa when scoping without AI assistance in anticipation of comparison with AI.

If this were indeed the case, it raises the question: How are endoscopists affected when incorporating a novel technology into their practice, especially in trials unamenable to participant blinding? Asking these types of questions highlights the importance of considering study methodology when designing trials and encourages critically viewing pooled results of trials when making broader conclusions.

Selection of studies from databases. ADR, Adenoma detection rate; AI, artificial intelligence; df, degrees of freedom; IEEE, Institute of Electrical and Electronics Engineers; RCT, randomized controlled trial; WoS, Web of Science.

Read the full article online.

The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.