Development And Validation Of Noninvasive Models To Predict Significant Fibrosis In Patients With Chronic Hepatitis B

Objective:Non-invasive models of liver fibrosis have recently been developed as an alternative to liver biopsy.The aim of this study was to compare the diagnostic performance of 7 models(APRI,FIB-4,Forns' index,Fibrometer,Hepascore,SLFG and Fibroindex) and develop new algorithm in patients with chronic hepatitis B.Methods:This study is a diagnosis test.A total of 374 chronic hepatitis B cases were selected,which were confirmed by liver biopsy between March 2001 and August 2008 from Zhong Shan Hospital,Fudan University.Liver fibrosis was staged according to the METAVIR scoring system(fibrosis stage S2,S3 or S4 was termed significant fibrosis).Blood tests including HB,WBC,PLT,bilirubin,albumin,PT, transaminase,hepatitis virus markers,direct markers of liver fibrosis,et al were measured at the time of liver biopsy.All analysis was performed using the SPSS 15.0. Diagnosis accuracy was assessed by sensitivity,specificity,PPV,NPV,LR and ROC curves.Logistic regression was used for multivariate analysis.Results:1.The area under ROC curves(AUROC) were 0.66,0.73,and 0.74 respectively for significant fibrosis in APRI,FIB-4 and Forns' index.And the diagnosis accuracy was 63%,69%and 76%respectivelyby using the cutoff scores.The AUROC of APRI,FIB-4 and Forns' index for advanced fibrosis(S3-4) were 0.65,0.77,0.79, respectively.2.The AUROC were 0.80,0.83,0.83 and 0.86 respectively for significant fibrosis in Hepascore,Fibroindex,SLFG,and Fibrometer.And the diagnosis accuracy was 88%,83%,86%and 88%respectively by using the cutoff scores.The AUROC of Hepascore,Fibroindex,SLFG,and Fibrometer for advanced fibrosis(S3-4) were 0.95,0.88,0.93 and 0.94,respectively.3.A model named APAG was constructed consisting of Age,PT,Alb,GGT which could discriminate between chronic hepatitis B(CHB) patients with and without significant fibrosis.The AUROC was 0.79(95%CI 0.74-0.85,AdAUROC=0.81) for the training group(n=250 ) and 0.78(AdAUROC=0.80 ) for the validation group(n=124 ).In the training group,using a cutoff score of＜0.27,the presence of significant fibrosis could be excluded with high accuracy(82%NPV ); similarly,applying a cutoff score of＞0.80,the presence of significant fibrosis could be correctly identified with high accuracy(87%PPV ),and with the same certainty in the validation group.Application of this model may decrease the need for liver biopsy in staging of 33%CHB.4.Another model named PGPH was constructed consisting of PLT,GGT,PT,HA to discriminate with and without significant fibrosis.The AUROC was 0.86 (95%CI 0.79-0.92,AdAUROC=0.89) for the training group(n=120) and 0.86 (AdAUROC=0.88) for the validation group(n=55).In the training group,using a cutoff score of＜0.20,the presence of significant fibrosis could be excluded with high accuracy(86%NPV);similarly,applying a cutoff score of＞0.80,the presence of significant fibrosis could be correctly identified with high accuracy (92%PPV),and with the same certainty in the validation group.Application of this model may decrease the need for liver biopsy in staging of 47%CHB.Conclusions:1.The diagnostic value of the simple models,which don't include direct serum markers,is poor.The FIB-4 and Forns' index are more accuracy than APRI for the diagnosis of significant fibrosis,they may be proposed for clinical screening of significant fibrosis after validation in lager cohorts.2.The diagnostic value of the models including direct serum markers is much better than simple models,and may decrease the need of liver biopsy in about half patients.Among these models,the Fibrometer has the best diagnostic value. Validation in larger cohorts of patients and standardizing the measurement of direct serum marker are needed before these models are proposed for extensive clinical use.3.APAG and PGPH,two noninvasive models constructed in this study,showed the better diagnostic value than other models.However,it should be validated in larger cohort study.