Surveillance findings in high-risk patients after baseline computer-assisted detection colonoscopy: a propensity score matching analysis

Abstract

Background and Aims: The impact of computer-assisted detection (CADe) on surveillance colonoscopy findings remains unknown. We compare surveillance colonoscopy findings in patients with high-risk lesions detected at baseline, with or without the application of CADe. Methods: We included patients with high-risk baseline lesions who had undergone surveillance colonoscopy. High-risk baseline lesions included advanced adenoma, advanced serrated lesion, or 3 or more adenomas at baseline colonoscopy. Patients were divided into 2 groups according to baseline use of CADe (vs conventional detection), and matched by propensity score matching (PSM) in a 1:2 ratio for surveillance interval, baseline characteristics, and index colonoscopy findings. The primary outcome was metachronous advanced adenoma (AA) rate. Results: Of 403 patients with high-risk baseline lesions and surveillance colonoscopy, 162 were included after PSM. In patients with baseline use of CADe, both the advanced adenoma detection rate (11.1% vs 24.1%; P = .05) and the number of advanced adenomas per colonoscopy (0.1 vs 0.3; P = .01) were significantly lower than in those with conventional colonoscopy. Similar pattern was observed for the adenoma detection rate (44.4% vs 63.0%; P = .02) and the number of adenomas detected per colonoscopy (0.6 vs 1.2; P = .01). The cumulative incidence of patients with any metachronous adenoma or serrated lesion was significantly lower in patients with baseline use of CADe than in those with conventional colonoscopy (log rank P = .05). Conclusions: Patients who had CADe use at baseline colonoscopy had significantly lower rates of metachronous lesions, including advanced adenomas, on surveillance colonoscopy.