Pre-Resection Endoscopic Features in T1 Colorectal Cancer Selected for Salvage Surgery After ESD/EID: An Exploratory Comparison of SM1 and SM2–3 Invasion

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background/Objectives: Accurate assessment of submucosal invasion is important when selecting patients with early colorectal cancer for organ-preserving treatment. Endoscopic submucosal dissection (ESD) and endoscopic intermuscular dissection (EID) enable en bloc resection and detailed histopathological assessment. This study explored whether routinely recorded pre-resection endoscopic characteristics differed between SM1 and SM2–3 lesions within a selected cohort of patients who underwent salvage surgery after non-curative endoscopic resection. Methods: We retrospectively analyzed 43 patients with T1 colorectal cancer who underwent ESD or EID and subsequently underwent salvage surgery because of non-curative histopathological findings. Among the 43 included patients, 37 underwent ESD and 6 underwent EID. Pre-resection Paris morphology and pit-pattern classification were compared according to final histopathological depth of submucosal invasion. Exploratory univariable logistic regression was used to examine associations with SM2–3 invasion. Results: Five patients had SM1 invasion and 38 had SM2–3 invasion. No statistically significant differences in the evaluated baseline or endoscopic variables were observed between the two groups. Exploratory regression estimates were imprecise, with wide confidence intervals, reflecting the small number of SM1 cases. Conclusions: In this selected cohort of patients undergoing salvage surgery after non-curative ESD/EID, Paris morphology and pit-pattern classification did not show statistically significant differences between SM1 and SM2–3 lesions. Because curatively treated patients were not included and only five patients had SM1 invasion, these findings do not estimate the diagnostic performance of these classifications in the broader pre-resection population and should be considered hypothesis-generating.

This publication has 33 references indexed in Scilit: