Macrophage morphology in the tumor microenvironment predicts metachronous liver metastasis in gastric cancer: establishment and validation of a predictive model

Background Gastric cancer patients with metachronous liver metastasis (MLM) generally pose a significant clinical challenge, with 5-year survival below 10%. Macrophages are key components of a tumoral microenvironment, but the predictive value of their morphological features for MLM remains unexplored. This study aimed to construct a risk predictive model for MLM based on macrophage morphology in the tumor microenvironment. Methods To reduce major baseline imbalances between patients with and without MLM, propensity score matching (PSM) was performed. After matching, a retrospective analysis of 233 gastric cancer patients who underwent radical surgery between 2016 and 2020 was conducted. Macrophage morphological parameters in different tissues were quantified by Qupath. Patients were randomly divided into training (70%) and validation (30%) datasets. The optimal cutoff values for the continuous variables were determined using the Youden index. Univariate and multivariate logistic regression analyses in the training set were used to identify risk factors for MLM. A nomogram was constructed for clinical applications. The model’s value was validated through receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Results The incidence of MLM was similar in the training (38.7%, 63/163) and validation cohorts (32.9%, 23/70). The optimal cut off value of macrophage areas and perimeters in the tumor invasive front (IF) region, tumor region, peritumoral stroma (PS) region were 76.75 μm 2 , 35.77 μm, 71.35 μm 2 , 32.00 μm, 74.13 μm 2 , 31.32 μm. Multivariate analysis revealed that only the area and perimeter of macrophages in the IF region, the area of macrophages in the tumor region were independent predictive factors for MLM. These factors were incorporated into a predictive nomogram. The model demonstrated good discrimination, with an area under the curve (AUC) of 0.835 in the training set and 0.798 in the validation set. Calibration and decision curve analyses confirmed its clinical utility. Conclusion Macrophage morphology parameters, particularly in the IF and tumor regions, are significantly associated with MLM in gastric cancer. The morphology-based nomogram developed in this study provides an exploratory tool for risk stratification and may help inform individualized postoperative surveillance.