Abstract
Colorectal cancer (CRC) is a leading cause of cancer-related mortality, with stage II and III CRC patients at significant risk of metachronous distant metastasis despite curative resection. While prior studies have developed models to predict specific types of distant metastasis, such as metachronous liver metastasis and peritoneal metastasis, there remains a need for a comprehensive model addressing any distant metastasis in stage II-III proficient mismatch repair (pMMR)/microsatellite stability (MSS) CRC. This study aims to develop a novel prediction model for metachronous distant metastasis in stage II-III pMMR/MSS CRC. We performed a retrospective study of data from 110 hospitalized stage II-III MSS CRC patients who underwent radical resection between June 2017 and February 2024. Univariate and stepwise multivariate logistic regression analyses were conducted to screen predictive factors. The variance inflation factor was employed to investigate multicollinearity among the final predictors. Model performance was evaluated through the receiver operating characteristic curves, calibration curve, and decision curve analysis. Internal validation was performed using 500 bootstrap iterations. The optimal model incorporated 3 key features: pT stage, pN stage, and vascular invasion. The model demonstrated good discriminatory ability with an area under curve of 0.830 (95% CI 0.749-0.911). Using a cutoff value of 0.226, sensitivity and specificity were 0.794 and 0.697, respectively. Internal validation confirmed the model' robustness, with an area under curve of 0.809 (95% CI 0.711-0.907). The calibration curve demonstrated a high level of consistency between the predicted and actual probabilities. Decision curve analysis demonstrated the model's high clinical utility. This nomogram including pT stage, pN stage, and vascular invasion provided a practical tool for predicting metachronous distant metastasis in patients with stage II-III pMMR/MSS CRC undergoing radical resection, exhibiting excellent discrimination, accuracy, and clinical applicability.