Abstract
<title>Abstract</title> <p>Objectives To develop and validate a prognostic model based on preoperative clinical and computed tomography (CT) image findings to predict overall survival and recurrence in patients with colorectal liver metastases (CRLM) after undergoing hepatic resection. Methods We conducted a retrospective study using data from a cohort of 174 patients who underwent liver resection for CRLM at Memorial Sloan Kettering Cancer Center. Variable selection was based on clinical relevance, literature review, accessibility, and cost-effectiveness. Collinearity between predictors was inspected. Cox regression analysis was used for both overall survival and recurrence. Internal validation was conducted using bootstrap and Jack-Knife methods. Model performance was assessed through time-dependent receiver operating characteristic (ROC) curves, calibration plots, decision curve analysis, and dominance analysis. An intuitive web-based open-access calculator was developed for further clinical applicability. Results The study included 174 patients, with 87 (50.0%) deaths and 108 (62.1%) recurrences. The overall survival model had an area under curve (AUC) of 0.700 and a Brier score of 0.219. The recurrence model demonstrated an AUC of 0.724 and a Brier score of 0.203. Both models showed good calibration and yielded a highly positive net benefit in decision analysis. Clinical variables (e.g., presence of major comorbidity, tumor response (%), and receiving neoadjuvant chemotherapy) and radiological features (e.g., size, multiple metastases) were the leading and most informative variables. Conclusion The developed models, incorporating clinical and CT image findings, demonstrate good performance in predicting overall survival and recurrence in CRLM patients undergoing hepatic resection.</p>