No Effect of Computer-Aided Diagnosis on Colonoscopic Adenoma Detection in a Large Pragmatic Multicenter Randomized Study

INTRODUCTION: Multiple randomized studies have demonstrated that computer-assisted detection (CADe) improves the colonoscopic adenoma detection rate (ADR); however, these improvements have not been consistently reproduced in some more recent studies. METHODS: Patients aged 50 years and older undergoing outpatient colonoscopy in 12 German institutions were randomized to either the CADe group or the control group. The primary outcome parameter was the ADR, while secondary outcomes included the adenomas per colonoscopy rate, detection rates of different histologic subgroups, and adverse events. The median examiner-specific observed rates and examiner-adjusted rates were compared between treatment groups. RESULTS: One thousand six hundred twenty-seven patients were included (mean age 63.1 years, 39.8% female; 71% screening indications). The observed median examiner ADR was 40.0% (interquartile range 20.0%-57.1%) in the study vs 37.5% (inter-quartile range 20.0%-50.0%) in the control group. Similarly, adjusted overall ADR was 40.6% (95% confidence interval [CI]: 35.8%, 45.5%) in the study and 38.3% (95% CI: 33.5%, 43.1%) in the control group. No differences were observed in adenomas per colonoscopy or any of the adenoma subgroups with regard to size, morphology, location, and histology as well as colonoscopy indication. The examiner had a large influence on ADR (adjusted median odds ratio [OR] 1.32), similar to patient age (OR per 10-year increase 1.40, 95% CI: 1.23-1.59) and sex (OR 1.62, 95% CI: 1.30-2.04). On the other hand, more hyperplastic polyps were found by CADe (OR for adjusted hyperplastic polyp detection rate = 1.29; 95% CI: 1.02-1.63; P = 0.033). DISCUSSION: In patients older than 50 years with mixed colonoscopy indications, CADe did not increase the ADR. Further studies should define the requirements for selective CADe use in routine clinical practice, particularly in relation to the examiner's baseline ADR.