Correlative Analysis Of Preoperative High Risk Factors And Construction Of Predictive Models For Splenic Hilar Lymph Node Meta-Stasis In Advanced Proximal Gastric Cancer

Objective:To investigate and analyze the clinical significance of preoperative serum inflammatory indexes,tumor markers and clinical tumor characteristics in advanced proximal gastric cancer with splenic hilar lymph node metastasis.A preoperative prediction model for splenic lymph node metastasis was established to provide a reference for individualized treatment of patients with advanced proximal gastric cancer.Methods:Clinical data of patients with progressive proximal gastric adenocarcinoma were collected from April 2016 to April 2020 in the Department of General Surgery of the940th Hospital of the Chinese People’s Liberation Army Joint Security Force.Whether the splenic hilar lymph nodes were metastatic or not was divided into splenic hilar lymph node metastasis group(No.10+group)and splenic hilar lymph node non-metastasis group(No.10-group)according to postoperative pathological tissue diagnosis.The preoperative general data,inflammatory markers,tumor markers and clinical tumor characteristics were compared between the two groups,respectively.To analyze the relationship between each factor and splenic hilar lymph node metastasis in advanced proximal gastric adenocarcinoma.Independent risk factors were derived from multi-factor binary logistic regression analysis,and the study samples were randomly divided into training and validation sets in proportion to their size.The training set data were used to establish a nomogram prediction model of splenic hilar lymph node metastasis in advanced proximal gastric cancer,and the validation set data were used to validate the prediction model.The efficacy,consistency and clinical utility of the prediction model were evaluated using the receiver operating characteristic curve(ROC)and area under the curve(AUC),calibration curve and decision curve analysis(DCA),respectively.Results:Clinical data of 236 patients with advanced proximal gastric adenocarcinoma were collected,including a total of 190 cases in the No.10-group and 46 cases in the No.10+group,with a splenic hilar lymph node metastasis rate of 19.5%(46/236).Univariate analysis showed that preoperative lymphocytes,albumin,NLR,PLR,LMR,CA19-9 positivity,tumor size,degree of hypofractionation,Borrmann type,Lauren type,depth of infiltration,lymph node enlargement and tumor site were strongly associated with advanced proximal splenic hilar lymph node metastasis,and the differences were statistically significant between the two groups(t=2.286,t=3.193,Z=-2.862,Z=-2.719,Z=-2.083,χ~2=8.749,χ~2=8.187,χ~2=20.428,χ~2=0.000,χ~2=0.000,χ~2=0.000,χ~2=0.000,χ~2=0.005,P<0.05).Multifactorial analysis showed that NLR(OR=1.473,95%CI=1.151～1.886),Borrmann type(OR=12.316,95%CI=2.932～51.744),degree of differentiation(OR=3.079,95%CI=1.051～9.016),lymph node enlargement(OR=4.918,95%CI=1.503～16.089)and tumor site(OR=2.710,95%CI=1.031～7.122)were independent risk factors for splenic hilar lymph node metastasis in advanced proximal gastric cancer.The results of multi-factor analysis were used to construct nomogram prediction model.The AUC=0.880(95%CI:0.822-0.938)for the prediction model in the training set and AUC=0.914(95%CI:0.836-0.991)in the validation set.The prediction curves of the training and validation sets in the calibration curves matched the actual observed curves reasonably well,and the goodness-of-fit tests were all P>0.05.A good net benefit was obtained with a threshold between approximately 0 and 0.65 for the training and validation sets in the decision curve analysis(DCA).Conclusion:Splenic hilar lymph node metastasis of advanced proximal gastric cancer was correlated with preoperative inflammatory markers,preoperative tumor markers and preoperative tumor clinical features.It was found that NLR,degree of differentiation,lymph node enlargement,tumor site(greater curvilinear side),Borrmann IV and III were independent risk factors for metastasis of splenic hilar lymph nodes in advanced proximal gastric cancer.The nomogram prediction model was constructed using the above variables,and internal validation showed that the model had good efficacy and was useful for assessing the preoperative status of splenic hilar lymph nodes in advanced proximal gastric cancer.