A Predictive Model for Postoperative Metastasis in Gastric Cancer Based on Preoperative Inflammatory Markers: A Retrospective Cohort Study from Western China.

Background & Objective: Postoperative metastasis is the predominant cause of poor long-term survival in gastric cancer (GC), where the systemic inflammatory response plays a pivotal role. Most existing predictive models are derived from general populations, and their applicability to specific ethnic and high-risk cohorts, such as the multi-ethnic population in Western China, remains unverified. This study aimed to develop a preoperative predictive model for postoperative metastasis specifically for this population by investigating the predictive value of novel preoperative inflammatory and metabolic markers: the Systemic Immune-Inflammation Index (SII), Uric Acid-to-Albumin Ratio (UAR), and Urea-to-Hemoglobin Ratio (UHR). Methods: Clinicopathological data from 656 GC patients who underwent surgery at Xinjiang Cancer Hospital between January 2018 and December 2023 were retrospectively collected. Based on the occurrence of postoperative metastasis, a 1:1 propensity score matching (matching factors: age, gender, BMI) was performed, resulting in 328 matched pairs (metastasis group vs non-metastasis group). Multivariable logistic regression analysis was used to identify independent predictors of postoperative metastasis, and a nomogram prediction model was constructed. The model's discrimination, calibration, and stability were internally validated using the area under the receiver operating characteristic curve (AUC), the Hosmer-Lemeshow test, and the Bootstrap method (1000 replicates). Results: Multivariable analysis identified SII (Q3 vs Q1: OR=1.80, P=0.009; Q4 vs Q1: OR=1.64, P=0.040), UHR (Q2 vs Q1: OR=1.84, P=0.025), T stage (T2 vs T1: OR=3.32, P<0.001), and N stage (N2 vs N1: OR=1.73, P=0.004) as independent risk factors for postoperative metastasis. Conversely, UAR (Q2 vs Q1: OR=0.44, P=0.015) was identified as a protective factor. The nomogram prediction model demonstrated a training set AUC of 0.684 and a bootstrap-corrected AUC of 0.669 upon internal validation. The model showed good calibration (Hosmer-Lemeshow test P=0.142). At the optimal cutoff value (0.417), the model's sensitivity was 88.1%, and the negative predictive value was 77.5%. Conclusion: This study successfully developed and validated a predictive model for postoperative metastasis in gastric cancer that integrates preoperative inflammatory markers (SII, UAR, UHR) and clinicopathological features (T stage, N stage) for the unique multi-ethnic population of Western China. The model exhibits good calibration and moderate discrimination, is particularly effective at identifying patients with low metastatic risk (high sensitivity), and serves as a useful auxiliary tool for individualized preoperative risk assessment and clinical decision support in this region.