Establishment And Evaluation Of A Risk Prediction Model For Malignant Hepatic Nodules Based On Logistic Regression Analysis

Objectives:A risk prediction model for malignant liver nodules was constructed and validated to provide screening methods for high-risk groups of liver malignant nodules.Methods:In this study,patients(n=1144 cases)who were initially diagnosed with liver nodular lesions by the Affiliated Hospital of North Sichuan Medical College from 2016 to 2020 were selected as the research objects.The results of needle biopsy or surgical pathological diagnosis were used as the grouping criteria for the malignant nodule group and the benign nodule group in this study.The patients were divided into training group(n=824)and validation group(n=320)according to the time of treatment and hospitalization.Referring to domestic and foreign literatures and guidelines,the influencing factors and routine examination indicators that may affect the risk of malignant liver nodules were selected,the hospital electronic medical record system was used for data collection,and SPSS26.0 and R3.5 software were used for statistical analysis.Univariate and multivariate Logistic regression analysis determined the risk factors of malignant liver nodules(P<0.05)and constructed the prediction model formula.Based on the data of the training group,the nomogram prediction model was constructed based on the established risk factors and verified in the validation group.The diagnostic performance and fit of the constructed model were evaluated by calibration,discrimination and Hosmer-Lemeshow test.Results:(1)The incidence of malignant liver nodules in the training group and the validation group was basically the same(60.56%vs 57.81%,P>0.05);(2)Univariate logistic regression analysis showed that the independent risk factors for malignant liver nodules were age,sex,nodule size,alpha-Fetoprotein(AFP),protein induced by Vitamin K Absence or Antagonist-Ⅱ(PIVKA-Ⅱ),total bilirubin(TBIL),y-glutamyltransferase isoenzyme(GGT),alanine aminotransferase(ALT),total bile acid(TBA),prothrombin time(PT),respectively;(3)The formula of the risk prediction model of malignant hepatic nodules constructed by multivariate Logistic regression analysis was:ln(P/1-P)=-15.595+0.037×age(years)+0.663×gender(male=1,female=2)+1.620×ln(nodule-size)+0.883×ln(PIVKA-Ⅱ)+0.681×ln(AFP)-1.131×ln(TBIL)+0.550×ln(GGT)-0.374×ln(ALT)-0.302×In(TBA)+3.624xln(PT).(4)The area under ROC(AUC),sensitivity,specificity,accuracy,positive prediction value(PPV)and negative prediction value(NPV)of the training group and the validation group were 0.969,93.38%,90.75%,92.35%,93.95%,89.94%,and 0.986,90.81%,94.26%,92.18%,95.45%,88.19%,respectively.(5)The H-L test results showed that the prediction model had good goodness of fit(χ2=6.955,P=0.542);the calibration curve showed that the predicted risk of malignant liver nodules in the line chart model had a good correlation with the actual incidence,and the predicted incidence was consistent with the actual incidence;(6)The AUC,sensitivity,specificity and accuracy of the model for the diagnosis of early hepatocellular carcinoma(Early-HCC)in the training group and the validation group were 0.942,88.64%,87.35%,87.82%and 0.956,87.04%,91.85%,90.47%,respectively.Conclusions:(1)Age,gender,nodule size,AFP,PIVKA-Ⅱ,TBIL,GGT,ALT,TBA and PT were independent risk factors for malignant liver nodules;(2)The risk prediction model of liver malignant nodules constructed in this study has good diagnostic efficiency and fitting degree,which can be used as a screening method for high-risk groups of liver malignant nodules,and it also shows good diagnostic accuracy in predicting Early-HCC.