Analysis of Influencing Factors of Gastric Cancer Based on Lasso-Logistic Regression Model
Objective To explore the influencing factors of gastric cancer and construct the clinical prediction model.Methods From December 2020 to October 2023,the clinical data of 1000 patients with stomach neoplasm admitted to Putuo Hospital,Shanghai U-niversity of Traditional Chinese Medicine and Shuguang Hospital,Shanghai University of Traditional Chinese Medicine were collected.Af-ter data cleaning and eliminating abnormal values,the patients were divided into gastric polyps group(n=487)and gastric cancer group(n=479).Non-parametric test was used to screen out meaningful indicators,Lasso regression to screen out the characteristic factors re-lated to gastric cancer with non-zero coefficient,and stepwise Logistic regression analysis to screen out the factors with significant correla-tion,and Lasso-Logistic regression model was constructed.The receiver operator characteristic(ROC)curve was plotted to calculate the area under the curve(AUC)and the confusion matrix to evaluate the model efficiency.Results The results of multivariate Logistic re-gression analysis showed that age,white blood cell(WBC)count,monocyte(M)count,alanine amiontransferase(ALT),cancer anti-gen 724(CA724),cancer antigen 242(CA242),cancer antigen 50(CA50)and carcinoembryonic antigen(CEA)were independent factors affecting gastric cancer.Based on the results of multivariate Logistic regression analysis,the risk prediction nomogram model of gas-tric cancer was constructed.The AUC of test set was 0.91,the accuracy rate was 100％,and the recall rate was 100％;the AUC of valida-tion set was 0.93,the accuracy rate was 93.63％,and the recall rate was 74.1％.The model has good prediction efficiency.Conclusion In this study,8 common predictors of gastric cancer were constructed,and the Lasso-logistic regression prediction model had good differen-tiation,which could be used to complete the early screening of gastric cancer based on the physical examination reports of patients.
Gastric cancerLasso-Logistic regressionRisk factorsClinical prediction model
NETL
NSTL
万方数据
