Construction of a tumor-associated myeloid cell infiltration signature model based on LASSO-Cox regression for predicting postoperative overall survival in patients with gastric cancer
10.12025/j.issn.1008-6358.2026.20260074
- VernacularTitle:基于LASSO-Cox回归构建肿瘤相关髓系细胞浸润特征模型预测胃癌患者术后总生存期
- Author:
Leihao WANG
1
;
Siyuan WANG
1
;
Mengxin TIAN
1
;
Qiangjun GAN
1
;
Zhaoqing TANG
1
;
Xuefei WANG
1
Author Information
1. Department of Gastrointestinal Surgery, Zhongshan Hospital, Fudan University, Shanghai 200032, China;Department of General Surgery, Zhongshan Hospital (Xiamen), Fudan University, Xiamen 361015, Fujian, China.
- Publication Type:Originalarticle
- Keywords:
gastric cancer;
myeloid cell;
clinical prediction model;
prognosis
- From:
Chinese Journal of Clinical Medicine
2026;33(3):396-405
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the impact of tumor-associated myeloid cell infiltration signatures on postoperative overall survival (OS) in patients with gastric cancer. Methods A retrospective analysis was performed on patients who underwent radical gastrectomy at Zhongshan Hospital, Fudan University from January 2005 to December 2019. Patients were randomly divided into a training set (n＝167) and an internal validation set (n＝168) at a 1:1 ratio. Meanwhile, 147 patients with gastric cancer treated at Zhongshan Hospital (Xiamen), Fudan University during the same period were enrolled as the external validation set (n＝147). Immunohistochemistry was used in all patients to evaluate the distribution characteristics of 18 immune markers in three regions of gastric cancer tissues, namely primary tumor (PT), invasive margin (IM), and normal tissue (NT). These markers include 11 myeloid cell markers (CD11b, CD33, CD11c, CD14, CD15, CD16, CD68, CD86, CD163, CD206, and CD66b) and 7 immune checkpoints (CD73, IDO, LAG3, PD-1, SIGLEC9, SIRPA, and TIM3). LASSO regression and Cox proportional hazards model were applied to screen immune markers associated with postoperative OS in gastric cancer patients, and a predictive model for 1-, 3-, and 5-year postoperative OS was established. Data from the internal and external validation set were used for internal and external verification of the predictive model. Receiver operating characteristic (ROC) curves were plotted and the area under the curve (AUC) was calculated to assess the discriminative ability of the model. Results In the training set, LASSO regression and Cox proportional hazards model identified 4 immune marker signatures (CD14_PT, CD15_PT, CD206_PT, and SIGLEC9_PT) to construct the predictive model. ROC curve analysis showed that the AUCs of the model for predicting 1-, 3-, and 5-year postoperative OS were 0.74, 0.75, and 0.84, respectively. When the model was applied to the internal and external validation sets, the AUCs for 1-, 3-, and 5-year postoperative OS were 0.74, 0.72, 0.72 (internal validation set) and 0.72, 0.71, 0.74 (external validation set), respectively. Cox multivariate regression analysis showed that the risk score was an independent risk factor for postoperative OS in patients with gastric cancer. Conclusion The model constructed based on LASSO-Cox regression exhibits good predictive performance and can effectively predict postoperative OS in gastric cancer patients, which may serve as a basis for intensive treatment of high-risk patients after surgery.
