A Retrospective Cohort Study to Predict Early Recurrence in Colorectal Liver Metastases Using a Nomogram Model.

Background: The incidence of colorectal liver metastases (CRLM) is rising, with only a subset of patients eligible for intent-to-cure treatment. Among these, up to 67% experience recurrence, with a worse prognosis for those with early recurrence. Reliable predictive models for early recurrence are needed.

Objective: To identify predictive factors for early recurrence in CRLM patients, construct a nomogram, and compare its predictive performance against a clinical risk score (CRS) model.

Methods: This study analyzed 240 CRLM patients who underwent intent-to-cure treatment at our center between January 2019 and August 2024. After applying inclusion and exclusion criteria, 198 patients were included. CRSs were calculated, and independent predictors of early recurrence were identified using univariate and multivariate Cox regression analyses. The nomogram model was evaluated using receiver operating characteristic (ROC) analysis, calibration, and decision curve analysis.

Results: Significant predictors of early recurrence included primary tumor location (p = 0.0014), primary tumor T stage (p = 0.0015), M stage (p = 0.0298), number of liver metastases (p = 0.003), metastatic tumor size (p = 0.0041), efficacy of neoadjuvant chemotherapy (p = 0.0043), and RAS mutation (p < 0.001). Independent predictors were primary tumor location, RAS mutation, number of metastases, and metastatic tumor size (p = 0.0047, p = 0.0116, p = 0.0423, and p < 0.0001, respectively). The nomogram model significantly outperformed the CRS model (AUC 0.790 vs. 0.604, p < 0.0001) and demonstrated superior clinical utility in decision curve analysis.

Conclusions: Primary tumor location, RAS mutation, and the extent of liver metastases are independent predictors of early recurrence in CRLM patients post-treatment. A nomogram integrating these factors demonstrated strong predictive performance, making it a practical tool for clinicians.