Malignant colon polyps: predicting lymph node metastasis following endoscopic excision

Aim The risk of lymph node metastasis (LNM) of malignant colon polyps (MCPs) is partly estimated by histologic features of the sampled polyp. However, the routinely available histologic data is limited to tumor grade and status of lymphovascular invasion (LVI). Methods The NCDB for colon cancer 2004-2018 was utilized. Patients with pT1Nx adenocarcinoma arising in a polyp and undergoing partial colectomy with >= 12 retrieved nodes were selected. NCDB 2004-2017 was used as a training cohort to develop two scoring systems based on a multivariable regression for predictors of LNM including clinical characteristics, grade, and LVI: a nomogram scoring system (NSS) and a simplified scoring system (SSS). These models were internally validated using NCDB 2018 to calculate precision metrics for each model. Results Six thousand sixty-nine patients were selected in the training cohort. 64.5% of MCPs were in the sigmoid, and LNM rate was 11.2%. Multivariable regression identified younger age, females, hindgut location, higher grade, and LVI as significant predictors of LNM. LNM risk was 1.2% when all unfavorable predictors were absent and exceeded 10% when NSS > 70 or SSS >= 3. In the 2018 validation cohort, 723 patients were scored per NSS and SSS, and the negative predictive value for both was 96%. Conclusion Estimating LNM risk in MCPs by applying clinical characteristics along with limited histologic data can help inform decision-making when considering formal oncologic resection. The NSS and SSS demonstrated comparable predictability of LNM among pT1Nx MCPs. The models require external validation and may be strengthened by incorporating additional endoscopic and pathologic characteristics.