[Development and validation of a prediction model for colorectal cancer liver metastasis]

Objective: To develop a prediction model for colorectal cancer liver metastasis (CRLM) using machine learning algorithms based on routine clinical laboratory data. Methods: A retrospective case-control study was performed on 5 026 patients with colorectal cancer admitted to Qilu Hospital of Shandong University from January 2019 to August 2023. Clinical information and 70 routine laboratory indicators at admission were collected. Patients were randomly divided into training and validation sets in a 7∶3 ratio. Feature selection was performed stepwise using LASSO regression followed by multivariate logistic regression. A random forest (RF) prediction model was constructed in the training set. The model performance was comprehensively evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, decision curve analysis (DCA), calibration curve, and Brier score, and further validated in the validation set. SHapley Additive exPlanations (SHAP) analysis was used to interpret the feature importance of the final model. Results: 11 laboratory indicators were screened to develop an RF predictive model, including AFP, CEA, CA19-9, ADA, HDL-C, LDH, Cys C, GLDH, Hcy, PT, and lymphocyte ratio. The AUC of the RF model, CEA, CA19-9, and combined detection of the two markers was 0.867, 0.737, 0.738, and 0.786, respectively. Delong test revealed that the predictive performance of the RF model was superior to that of the other detection methods (all P<0.05). Model interpretation using the SHAP algorithm revealed that carcinoembryonic antigen, carbohydrate antigen 19-9, and other indicators were key predictors of CRLM. Conclusion: An interpretable machine learning model is developed for the early prediction of liver metastasis and personalized treatment regimens in colorectal cancer patients through the effective integration of clinical laboratory data, providing valuable clinical decision support to clinicians and improving patient prognosis.