identify the clinical, lesion characteristics and endoscopic factors that led tosurgical management of colorectal tumors potentially amenable to endoscopic resection(ER). Secondary aims included assessing the accuracy of optical characterization, theimpact of reassessment in an expert center, as the outcomes of non-curative ER of pT1adenocarcinoma (ADC).

retrospective analysis of all patients submitted to surgery between 2020-2024with surgical specimen showing dysplasia, pT1 ADC, or no dysplasia/ADC after ER.Patients with polyposis syndromes, inflammatory bowel disease and neoadjuvantchemoradiotherapy were excluded. Data on index colonoscopy, reassessment atreferral center, ER attempt, surgical indication, final histopathology and post-resectionoutcomes were collected.

Results:113 patients with 113 lesions were included (61% male; 71±8 years-old). Indexcolonoscopy was performed mainly in private clinics (55%); 24% were reassessed at areferral center. Median lesion size was 35 mm (IQR 20), most frequently located in theleft colon (28%), ascending colon (24%), or caecum (22%).53% were described as neoplastic, and 28% underwent virtual chromoendoscopyevaluation. When enhanced imaging was used, there was a correlation between opticaldiagnosis and the surgical specimen histopathology: NICE/JNET 3 showed 64% of ADC(p=0.050). Describing a lesion as “neoplastic” on white-light endoscopy showed noassociation with final ADC (p=0.177).ER was attempted in 23% of cases. Piecemeal mucosectomy (42%), polypectomy(35%) and en-bloc mucosectomy (8%) were the most common techniques. Amongpatients operated for non-curative ER (12%) - defined as positive margins,lymphovascular invasion, poor differentiation or deep invasion - 77% had no dysplasia/ADC on the surgical specimen.Indications for surgery were suspicion of deep invasion (61%), non-curative ER (12%)and ER failure (11%). Predictors of “advanced lesion suggestive of deep invasion” were:description as neoplastic (OR 19.4, p<0.001), NICE/JNET 3 (OR 11.7, p=0.005), andbiopsy suggesting ADC (OR 2.56, p=0.062). Reassessment at a referral center wasprotective (OR 0.17, p=0.005). Lesion size did not reach significance (p=0.081).Final histology revealed 41% HGD and 48% ADC. Submucosal invasion was quantifiedin 13.5% of the ADC specimens. Excluding deep submucosal invasion as a risk factor,7% presented any high-risk criteria for lymph-node metastasis (LNM) (50% was deepsubmucosal invasion), with one confirmed LNM. Postoperative morbidity occurred in27%, 55% with dysplasia. The median length of hospital stay was 6 days (IQR 3).

surgical management of colorectal dysplasia and T1 cancer was largelydriven by initial endoscopic impression on index colonoscopy, rather than by structuredand comprehensive lesion characterization. Limited use of virtual chromoendoscopy,low rates of expert reassessment, and underutilization of ER contributed to escalationtoward surgery, often without high-risk histological features. These findings highlightsubstantial overtreatment potential and underscore the need to strengthen opticaldiagnosis, standardized referral pathways, and reassessment in expert centers.