Noninvasive Prediction Model For Histological Stages Of Primary Biliary Cholangitis

Aim:To develop a simple noninvasive serological model to predict the severity of hepatic histological alterations in patients with primary biliary cholangitis.Materials and methods:Patients with PBC who underwent liver biopsy at the First Hospital of Jilin University from January 2013 to October 2022 were collected.A total of 114 qualified patients were selected according to the inclusion and exclusion criteria.The demographic characteristics(gender and age),laboratory tests and liver histological examination results were collected within one week prior to liver biopsy.All patients were randomly divided into model group(n=78)and validation group(n=36)according to ratio of 7:3.The patients were histologically staged according to Scheuer’s staging method.All commonly used serological parameters and liver pathological results were analyzed.The independent predictors of histological stages were selected to establish a noninvasive serological model.The scores of 22 noninvasive models(AAR,ALBI,AP index,APRI,CDS,Doha score,FCI,FI,FIB-4,FibroQ,GPR,GP model,GUCI,HB-F,King’s score,Lok index,NIKEI,NLR,Poh1 score,PRP,MELD,S index)were calculated and compared with the established model.Results:1.A total of 114 patients with PBC were included in this study,including 99 females(86.8%)and 15 males(13.2%).The ratio of male to female was 1:6.6,and the median age was 53(48,58)years.There were 33 patients with liver histological stage 1(29.0%),34 with stage 2(29.8%),16 with stage 3(14.0%)and 31 with stage 4(27.2%).There was no statistical differences in clinical characteristics between the patients in the model and validation group.2.In the model group,all indexes except age,sex,AST,and NE differed between pathological stages(P < 0.05).All serological parameters with absolute values of correlation coefficients>0.5 with histological stage by Spearman correlation analysis were included in the multivariate ordered logistic regression analysis.It was suggested that only TBA and RDW were independent predictors of histological stage of PBC.Multiple stepwise regression analysis was performed between the above indicators and histological stage to establish the model: Model S=﹣0.668+0.005×TBA+0.176×RDW.To facilitate the calculation and application in the clinic,the model was simplified to TR score=TBA×RDW/10.3.Compared with Model S and 22 other models,TR score was the most strongly correlated with histological stages(r=0.770,P < 0.0001).When predicting early histological lesions(S1)or liver fibrosis and cirrhosis(S3-S4),the AUROCs of TR score were 0.887(95% CI,0.809～0.965)and 0.893(95% CI,0.816～0.969),higher than Model S and all of the other 22 models included in this study.When predicting cirrhosis(S4),its AUROC is still as high as 0.921(95% CI,0.837-1.000).4.The AUROC,sensitivity and specificity of TR score and Model S were close in predicting each histological stage.In terms of calculation,TR score was simpler than Model S.And the diagnostic criteria of Model S were very close,while it of TR model spanned more.Therefore,TR score was chosen as the final prediction model.5.In the validation group,the TR score>23.923 was taken as the diagnostic criteria for distinguishing histological manifestations of S2-S4,and the accuracy,sensitivity,specificity,PPV and NPV of the TR score were 72.2%,80%,54.5%,83.3% and 54.5%,respectively.After the consistency test,the kappa value of the pathological gold standard was 0.345(P<0.05).Using TR score>61.189 as the diagnostic criteria for distinguishing S3-S4,the consistency rate,sensitivity,specificity,specificity,PPV and NPV were 75.0%,83.3%,70.8%,58.8% and 89.5%,respectively.The kappa was 0.491(P<0.01).Using TR score>97.465 as the diagnostic criteria for distinguishing S4,the consistency rate,sensitivity,specificity,specificity,PPV and NPV were 80.6%,77.8%,81.5%,58.3% and91.7%,respectively.The kappa was 0.533(P<0.01).Conclusions:1.TBA and RDW are independent risk factors for advanced histological stage in patients with PBC.2.Compared with 22 existing noninvasive models,TR score showed higher accuracy in predicting histological changes in the early stage(stage 1)and in the stage of liver fibrosis and cirrhosis(stage 3～4).3.TR score is a simple,inexpensive and stable noninvasive model,without complex calculation formulas and tools,and is expected to be a new noninvasive serological scoring model in diagnosing the histological stages of PBC.