FINDINGS AT SURVEILLANCE COLONOSCOPY IN HIGH-RISK PATIENTS WITH PRIOR COMPUTED-ASSISTED DETECTION DURING INDEX COLONOSCOPY: A PROPENSITY SCORE MATCHING ANALYSIS

Introduction: Computer-assisted detection (CADe) has been shown to significantly improve the adenoma detection rate during index colonoscopy. However, the real impact of CADe on subsequent surveillance colonoscopy findings, particularly among patients who were previously found to have high-risk lesions, remains uncertain. This study aims to compare the findings at surveillance colonoscopy of patients who were previously found to have high risk lesions at baseline colonoscopy with or without the use of CADe.
Method: This study first identified patients who underwent colonoscopy (with or with CADe) and were found to have high-risk lesions at baseline in our hospital. High risk lesion was defined by the presence of an advanced adenoma, advanced serrated lesion, or the detection of three or more adenomas at baseline. For this analysis, patients who were found to have high-risk lesion at baseline and had surveillance colonoscopy were divided into two groups according to use of CADe (CADe) or no use of CADe (conventional colonoscopy) at baseline. Propensity score matching (PSM) was performed in 1:5 ratio according to patient’s demographics, comorbid conditions, baseline colonoscopy findings, and surveillance intervals. The primary outcome was advanced adenoma (AA) detection rate during surveillance colonoscopy, and secondary outcomes included adenoma (ADR) and sessile serrated lesion (SSL) detection rates.
Results: Between February 2016 and June 2021, a total of 611 patients had baseline colonoscopy procedure with high-risk lesions detected and had received a surveillance colonoscopy. Among them, 35 patients had their initial colonoscopy performed with the CADe assistance and 576 patients had no use of CADe at baseline. The mean surveillance interval for the CADe group of patients was 894 days, and no AA were detected during their surveillance colonoscopies. After PSM, both the AA rate (0% vs 25.7%, p<0.01) and the mean number of AA per colonoscopy of the CADe group (0 vs 0.38, p<0.01) were significantly lower than the conventional colonoscopy group. Similar trend was observed for ADR (34.3% vs 65.1%, p<0.01) and SSL detection rate (8.6% vs 32.5%, p<0.01) as well as the mean number of colonic lesions detected (adenoma: 0.3 vs 1.7, p<0.01; SSL: 0.1 vs 0.7, p =0.02) between CADe and conventional colonoscopy group (Table 1). The probability of a patient with metachronous adenomas or SSLs detected at surveillance colonoscopy is also significantly lower in the CADe group than conventional colonoscopy group (log rank test p<0.01).
Conclusion: These early results suggested that patients who had baseline use of CADe during index colonoscopy had significantly lower detection rates of metachronous colonic lesions, including AA, during subsequent surveillance.