Construction Of A LASSO Regression-based Prediction Model For The Development Of Cirrhosis In Chronic Hepatitis B

BackgroundHepatitis B virus(HBV)infection is the most common etiology of liver cirrhosis in our country.Chronic hepatitis B(CHB)progresses gradually and advances to the stage of cirrhosis with the progression of fibrosis,hepatic vascular proliferation,and pseudolobular formation.The early stage of cirrhosis is often asymptomatic and can be easily overlooked by patients,followed by several years or months of post-diagnosis into a symptomatic decompensation phase,and patients frequently require hospitalization due to various complications,experience impaired quality of life,and have increased mortality.Currently,non-invasive prediction models that can effectively detect cirrhosis in its early stages are still lacking.ObjectiveTo construct a nomogram prediction model for the development of cirrhosis in CHB using commonly used clinical indicators.MethodsCHB patients who underwent liver biopsy at the Infection Department of the First Affiliated Hospital and the Second Affiliated Hospital of Anhui Medical University between 2010 and 2018 were selected and divided into two groups:CHB alone group and cirrhosis group.Their general data,blood routine,liver function,HBV DNA quantification,alpha fetoprotein and other laboratory test indicators were collected and compared between groups.The least absolute shrinkage and selection operate(LASSO)regression analysis was employed to identify the predictors with significant predictive value for cirrhosis,followed by multivariate logistic regression analysis to establish a robust prediction model.Bootstrap resampling was performed 500 times for internal validation of the model,and a receiver operating characteristic curve was drawn and the area under the curve(AUC)calculated to assess model discrimination.A decision curve(DCA)was plotted to assess the model’s degree of benefit,and a calibration curve(CA)was drawn to assess the model’s degree of calibration.ResultsA total of 1087 CHB cases,135 of which were complicated by cirrhosis,were included,and all measures except HBV DNA quantification,alanine transaminase(AST)were statistically different between the two groups(P<0.05).The final selected predictor variables after LASSO regression analysis were age,AFP,albumin(ALB),globulin(GLB),glutamyl transpeptidase(GGT),and platelet count(PLT).The prediction model was established by multifactorial logistic regression analysis(Logit P=1.26+0.02 × age+0.001 × AFP-0.10 × ALB+0.07 × GLB+0.004 × GGT-0.02 ×PLT).The AUC of this prediction model was 0.83 with a 95%confidence interval(CI)0.79-0.87.The DCA curve suggested an increase in net benefit for patients using the established prediction model,and the CA curve suggested good agreement between the predictive effect of this prediction model and the actual outcome.ConclusionPatient age,serum ALB,GLB,PLT,GGT,AFP had some correlation with the development of early cirrhosis due to CHB.The prediction model,which was finally constructed by multivariate logistic analysis,was evaluated by ROC curve,DCA curve and CA curve analysis,had a good discrimination,benefit and calibration,and had some clinical significance.