Development of a predictive model for metachronous liver metastasis in gastric cancer

Patients with metachronous liver metastasis (MLM) in gastric cancer generally have a poor prognosis. Early detection and accurate prediction of MLM are crucial for improving clinical outcomes. This study aims to identify the risk factors for MLM through clinical pathological parameters and develop a predictive model for MLM in gastric cancer.

A retrospective analysis of 1248 gastric cancer patients who underwent radical surgery between December 2016 and December 2020 was conducted. Patients were randomly divided into training (70%, n=873) and validation (30%, n=375) datasets. The optimal cutoff values for the continuous variables were determined using the Youden index. Univariate and multivariate logistic regression analyses were used to identify risk factors for MLM. A nomogram was developed based on the results of multivariate analysis. The model's value was validated through receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA).

The incidence of MLM was comparable between the training (10.3%, 90/873) and validation set (9.9%, 37/375). The optimal cutoff value was 3.315ng/ml for preoperative alpha-fetoprotein (AFP) level, 16.275U/ml for preoperative cancer antigen 125 (CA125) level, 0.280×10⁹/L for monocyte count and 1.430×10⁹/L for lymphocyte count, respectively. Univariate analysis showed that age, tumor size, pathological type, surgical method, T stage, N stage, TNM stage, neural invasion, lymphatic vascular invasion, number of lymph nodes harvested (LNH), preoperative total protein (TP), hemoglobin (HB), albumin (ALB), preoperative carcinoembryonic antigen (CEA), preoperative cancer antigen 19-9 (CA19-9), CA125, AFP levels, monocyte count, lymphocyte count, red blood cell (RBC) count and platelet count were considered as potential variables. Multivariate logistic regression analysis indicated that T stage, N stage, monocyte count, lymphocyte count, preoperative AFP and CA125 levels were independent predictive factors for MLM. The identified risk factors were further used to develop a predictive nomogram for MLM. The nomogram exhibited robust discriminatory performance, with an area under the curve (AUC) of 0.859 in the training set and 0.803 in the validation set. Moreover, the nomogram demonstrated excellent calibration and significant clinical utility.

This study successfully developed a predictive nomogram for MLM in gastric cancer. Besides conventional parameters, we identified and incorporated peripheral blood monocyte and lymphocyte counts as novel predictors, demonstrating their independent predictive value. Integrating these factors into nomogram could enhance predictive accuracy of MLM.