According to global reports, colorectal cancer is the third most common cancer in the United States and a leading cause of cancer-related mortality. The main strategy to reduce colorectal cancer (CRC) risk includes identifying and collecting precancerous polyps through colonoscopy. The variability in colonoscopy execution quality determines their performance outcome. The Adenoma Detection Rate (ADR) is a key quality indicator because it shows that higher detection rates lead to lower post-exam cancer development. Studies demonstrate that colonoscopies fail to detect approximately 25% of adenomas because it remains challenging to locate both subtle and flat lesions.
The analysis of artificial intelligence (AI)-assisted technology through Computer-Aided Detection (CADe) systems has been extensively studied to improve colonoscopy detection accuracy. Systems based on deep learning technology have drawn interest because they can boost ADR results and decrease missed adenoma diagnoses. Overdiagnosis and expense, combined with resource consumption, have become subjects of increasing study. This research investigates the effect of CADe on colonoscopy by evaluating its advantages and obstacles through systematic reviews and guidelines.
The American Gastroenterological Association (AGA) collaborated with the BMJ Rapid Recommendation series to create guidelines through the application of the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) methodology. The group formed their recommendations using systematic reviews that evaluated three key areas: (a) the benefits and risks of using CADe for colonoscopy, (b) a long-term microsimulation of patient outcomes under CADe, and (c) an assessment of patient preferences toward colonoscopy evaluation. Additionally, the group studied (d) healthcare providers’ trust in AI applications for gastroenterology.
The methodology incorporated here is for the findings from systematic reviews, randomized controlled trials (RCTs), and other research regarding the advantages and disadvantages of CADe-assisted colonoscopy.
The assessment involved the collection of feedback from healthcare providers about their opinions on using AI technology in their normal work routines. These guidelines share evidence from a meta-analysis that incorporated 44 RCTs with more than 30,000 participants in the study. The research assessed how CADe technology affected the measurement of polyp identification rates, ADR, advanced colon cancer detection, and the identification of non-cancerous polyps, along with other outcomes. A systematic weighing of CADe effects occurred using the Evidence to Decision framework, which examined favorable and unfavorable aspects of patient outcomes together with resource consumption and healthcare provider methods.
The meta-analysis proved that CADe-aided colonoscopy significantly improved polyp detection rates and ADR. The main findings were:
The polyp detection rate was higher in CADe-assisted colonoscopy (56.1%) than in standard colonoscopy (47.9%), with a relative risk (RR) of 1.22 (95% CI, 1.15-1.28).
Adenoma Detection Rate (ADR): The ADR with CADe improved to 44.8%, from 37.4% with standard colonoscopy, with an RR of 1.22 (95% CI, 1.16-1.29).
Adenoma Miss Rate: CADe had a much lower adenoma miss rate (16.1% missed in the CADe group compared with 35.3% in the control colonoscopy group), with an RR of 0.47 (95% CI, 0.36-0.60).
Advanced Colorectal Neoplasia Detection: CADe-assisted colonoscopy showed a higher detection rate of advanced colorectal neoplasia, with 12.7% of cases detected compared to 11.5% in standard colonoscopy. The relative risk (RR) was 1.16 (95% CI, 1.02-1.32).
Non-neoplastic Polyps: The detection rate of non-neoplastic polyps increased by 11% with CADe, with 34.0% detected in the CADe group versus 28.8% in the standard group. The relative risk (RR) was 1.11 (95% CI, 1.04–1.19).
Withdrawal Time: CADe led to a small increase in total withdrawal time (9.17 minutes vs. 8.60 minutes) and inspection time (8.34 minutes vs. 7.95 minutes). The mean differences were 0.57 minutes (95% CI, 0.31-0.83) for withdrawal time and 0.31 minutes (95% CI, 0.14-0.48) for inspection time.
According to systematic reviews about healthcare provider views, the majority (85%) showed interest in AI colonoscopy, while 74% believed using AI could boost adenoma detection. The findings also included healthcare provider concerns related to higher costs (43%), accountability for mishaps in diagnosis (40%), and concerns regarding medicolegal issues (23%). The available research confirms that CADe-assisted colonoscopy enables higher identification rates of polyps and adenomas while reducing missed adenomas and identifying more cases of advanced colorectal cancer. The advantage of using automated detection technology is that it results in additional screenings of small potential adenomas that present minimal risks for cancer development.
The impact of CADe extends to potential CRC incidence reduction, but the specific benefit for CRC-related mortality remains unclear. The use of CADe primarily presents no major safety concerns but requires some patients to undergo an increased frequency of follow-up colonoscopic procedures. Most healthcare providers support the use of AI applications in colonoscopy procedures, yet they continue to raise concerns regarding cost expenses as well as medical liability complications.
AI systems and guidelines must progress with expanding data availability to achieve proper risk management between diagnosis benefits and diagnosis overreach, along with growing healthcare expenses. The advantages of CADe in improving colonoscopy quality must be measured against possible costs and their impact on patient outcomes alongside resource requirements.
References: Sultan S, Shung DL, Kolb JM, et al. AGA Living Clinical Practice Guideline on Computer-Aided Detection–Assisted Colonoscopy. Gastroenterology. 2025;168(4):691-700. doi:10.1053/j.gastro.2025.01.002
