目的 探讨错配修复 （MMR）与胃癌患者术前炎性指标和临床病理特征的关联，构建以胃癌患者术前炎性指标和临床病理特征为基础的胃癌MMR预测模型，为胃癌MMR状态评估提供新思路。 方法 纳入2020年9月—2023年10月行手术治疗的254例胃癌患者，依据MMR蛋白表达情况将患者分为MMR表达正常［MMR稳定（pMMR）］组和MMR表达缺陷（dMMR）组，收集2组胃癌患者的术前炎性指标和临床病理特征资料。采用χ²检验分析2组胃癌患者炎性指标和临床病理特征与MMR的关联性；筛选dMMR的独立预测因子，构建列线图；采用受试者工作特征（ROC）曲线和校准曲线评价模型的性能，采用临床决策曲线评价预测模型的临床实用性。 结果 研究共纳入254例胃癌患者，其中pMMR组患者221例（87%），dMMR组患者33例（13%）。2组胃癌患者年龄、肿瘤发病部位、肿瘤分化程度、肿瘤最大径、血小板/淋巴细胞比值（PLR）、碱性磷酸酶（AKP）、AKP/白蛋白（AL）比值（AAR）、纤维蛋白原（FB）/淋巴细胞比值（FLR）、FB/AL比值（FAR）、D-二聚体（D-D）和FB比较差异有统计学意义（P<0.05）。单因素和多因素Logistic回归分析，肿瘤最大径［比值比（OR）=2.958，95%置信区间（CI）：1.196~7.314，P=0.019］、肿瘤发病部位（OR=4.013，95%CI：1.596~10.089，P=0.003）、肿瘤分化程度（OR=3.006，95%CI：1.250~7.230，P=0.014）、FAR（OR=2.793，95%CI：1.179~6.616，P=0.020）和糖类抗原199（CA199）（OR=0.279，95%CI：0.084~0.929，P=0.038）是dMMR的独立预测因子。基于炎性指标和临床病理特征构建的胃癌MMR预测模型ROC曲线下面积（AUC）值为0.800，灵敏度为0.851，特异度为0.606，P<0.01。验证列线图的校准曲线能够很好地拟合到理想曲线上，且Hosmer-Lemeshow拟合优度检验P=0.412；临床决策曲线显示模型具有良好的净收益。 结论 胃癌患者术前炎性指标和临床病理特征与胃癌MMR状态存在关联，肿瘤最大径、肿瘤发病部位、肿瘤分化程度、CA199和FAR是dMMR胃癌的独立预测因子，基于上述独立预测因子构建的胃癌患者MMR预测模型，可高效预测dMMR胃癌患者MMR状态。

Objective To discuss the associations of mismatch repair （MMR） in gastric cancer with preoperative inflammatory indicators and clinicopathological features in the gastric cancer patients， and to construct a gastric cancer MMR predictive model based on preoperative inflammatory indicators and clinicopathological features of the gastric cancer patients， and to provide new ideas for evaluation of MMR status in gastric cancer. Methods The data of 254 gastric cancer patients who underwent surgical treatment from September 2020 to October 2023 were included. According to the expression of MMR protein， the patients were divided into MMR normal （proficiout MMR， pMMR） group and MMR deficient （dMMR） group. The preoperative inflammatory indicators and clinicopathological features data of the gastric cancer patients in two groups were collected. The associations between inflammatory indicators， clinicopathological features， and MMR in dMMR group and pMMR group were analyzed usingChi-square test. The independent predictive factors for dMMR were selected to construct the nomogram. Receiver operating characteristic （ROC） curve and calibration curve were used to evaluate the predictive efficacy， and decision curve was used to evaluate the practicality of the predication model. Results A total of 254 gastric cancer patients were included in the study， with 221 patients （87%） in pMMR group and 33 patients （13%） in dMMR group. There were statistically significant differences （P<0.05） in age， tumor location， tumor differentiation degree， maximum tumor diameter， platelet-to-lymphocyte ratio （PLR）， alkaline phosphatase （AKP）， alkaline phosphatase-to-albumin ratio （AAR）， fibrinogen（FB）-to-lymphocyte （FLR）， FB-to-albumin（AL） （FAR）， D-dimer （D-D）， and FB of the gastric cancer patients between dMMR group and pMMR group. Univariate and multivariate Logistic regression analysis revealed maximum tumor diameter ［odd ratio（OR）=2.958， 95% confidence interval （CI）：1.196-7.314， P=0.019］， tumor location （OR=4.013，95%CI：1.596-10.089， P=0.003）， tumor differentiation （OR=3.006， 95%CI： 1.250-7.230， P=0.014）， FAR （OR=2.793， 95%CI：1.179-6.616， P=0.020）， and carbohydrate antigen 199（CA199） （OR=0.279， 95%CI：0.084-0.929， P=0.038） were the independent predictors of dMMR. The area under the ROC curve （AUC） value of the gastric cancer MMR prediction model constructed based on inflammatory indicators and clinical pathological characteristics was 0.800 with the sensitivity of 0.851 and the specificity of 0.606. The calibration curve of the nomogram was found to fit the ideal curve well，and in Hosmer-Lemeshow test P=0.412， the clinical decision curve showed a better net benefit. Conclusion The preoperative inflammatory indicators and clinicopathological features are associated with MMR in gastric cancer； maximum tumor diameter， tumor location， tumor differentiation， CA199， and FAR are the independent predictors of dMMR. The prediction model based on the above predictors could predict the MMR status of the dMMR gastric cancer patients.