Management of T1 colorectal cancers following endoscopic resection remains challenging, with decisions between surveillance and surgery primarily driven by histopathologic risk factors. This study aimed to identify independent predictors of surgery and evaluate long-term outcomes, incorporating a combined clinicopathologic model.

A retrospective analysis was conducted on 264 T1 colorectal lesions for surgery prediction and a subgroup of 100 cases with available long-term data for five-year follow-up (FU5y) analysis. Variables included patient age, lesion size, depth of submucosal invasion (sm1–sm3), lymphatic and vascular invasion, tumor budding grade, and tumor grading (G1–G3). Logistic regression models identified independent predictors of surgery and FU5y. Model performance was assessed by area under the ROC curve (AUC). A combined clinicopathologic score (sm2, tumor budding, lymphatic invasion) was tested for enhanced discrimination.

Surgery prediction model showed good discrimination (AUC ≈0.80). Independent predictors were high-grade tumor budding, lymphatic/vascular invasion, sm3 invasion, and G3 grading. Critically, sm2 was not independently significant (OR ≈1, p ≈0.75) after adjustment, indicating its effect is mediated by coexisting aggressive features. The novel combined model outperformed single factors (higher AUC), enabling refined risk stratification. No robust FU5y predictors emerged due to sample limitations.

In this cohort, tumor aggressiveness markers rather than demographic or lesion size factors were the primary independent predictors guiding surgery after endoscopic resection of T1 colorectal cancers. Sm2 alone did not independently increase surgery risk, but when combined with budding and lymphatic invasion, it enhanced predictive accuracy. Limited long-term data highlight the need for larger samples to refine risk stratification for follow-up outcomes.