Lymph node metastasis in T1 colorectal cancer: Risk factors and prediction model.
Shiki Fujino, Norikatsu Miyoshi, Masatoshi Kitakaze, Masayoshi Yasui, Masayuki Ohue, Hideki Osawa, Yoshihito Ide, Toshinori Sueda, Mitsuyoshi Tei, Takashi Takeda, Katsuki Danno, Yozo Suzuki, Shingo Noura, Kenji Ohshima, Eiichi Morii, Hidekazu Takahashi, Mamoru Uemura, Hirofumi Yamamoto, Kohei Murata, Yuichiro� Doki, Hidetoshi Eguchi
Abstract
Endoscopic resection is typically performed for early T1 stage colorectal cancer (T1 CRC). Additional surgery is subsequently recommended based on pathological findings; however, the current criteria may result in overtreatment. The present study aimed to re-examine the reported risk factors for lymph node (LN) metastasis in T1 CRC and develop a prediction model using a large multi-institutional dataset. In this retrospective study, the medical records of 1,185 patients with T1 CRC who underwent surgery between January 2008 and December 2020 were investigated. Slides pathologically re-assessable for additional risk factors were re-examined. A total of 251 patients with inadequate data were excluded, and 934 patients were randomly assigned at a ratio of 3:1 to the training and validation datasets. In the univariate analysis, left-sided CRC (P=0.003), deep submucosal invasion depth (P=0.005), poor histological grade (P=0.020), lymphatic invasion (P<0.001), venous invasion (P<0.001) and tumor budding grade 2/3 (P<0.001) were significant risk factors for LN metastasis. A nomogram predicting LN metastasis was developed using these variables, with an area under the received operating characteristic curve (AUC) of 0.786. The nomogram was validated using a validation set with an AUC of 0.721, indicating moderate accuracy. No LN metastases were observed in patients with <90 points using the nomogram; therefore, patients with a low nomogram score may avoid undergoing surgical resection. Prediction of LN metastasis using this developed nomogram may help identify patients who are at high-risk who require surgery.