To accurately select patients with malignant colorectal polyps who are at high risk of adverse outcome, we examined the predictive value of clinicopathological factors, with special attention paid to the histology at the invasive margin. We examined 75 submucosal carcinomas from 75 patients, initially resected by polypectomy, including endoscopic, trans-anal, trans-sacral, and trans-sphincteric local excision. The associations between clinicopathological features such as sex and age; tumor size, location, shape, depth of submucosal invasion, vascular invasion, histology at the central part, and histology at the invasive margin; and the presence or absence of a residual adenomatous component and adverse outcome were examined by univariate and multivariate logistic regression analyses. Lymph node metastases were found in 2 patients, local recurrence in 4, and distant metastases in 2. Univariate logistic regression analysis showed that unfavorable histology at the invasive margin was significantly associated with lymph node metastasis or local recurrence (P = 0.0373), whereas the association of lymphatic invasion and vascular (lymphatic or venous) invasion with lymph node metastasis or local recurrence had marginal significance (P = 0.0785; P = 0.0990). Multivariate logistic regression analysis, with unfavorable histology at the invasive margin and lymphatic invasion as independent variables, showed that unfavorable histology alone had significance (P = 0.0373) in predicting adverse outcome. Widely accepted criteria such as massive submucosal invasion, positive vascular invasion, and poorly differentiated histology, were less useful in predicting adverse outcome. These results suggest that unfavorable histology at the invasive margin is a useful risk factor for predicting lymph node metastasis or local recurrence in patients with malignant colorectal polyps.