Clinical Study Of Multi-indexes Models For Non-invasively Diagnosing Compensated Cirrhosis In Patients With Chronic Hepatitis B

Background:In patients with chronic hepatitis B(CHB),cirrhosis is the late stage in the long course of liver fibrosis,which is directly related to the chronic liver damage.The disease prognosis was significantly different between patients with compensated and decompensated cirrhosis.While patients with compensated cirrhosis owned a ten-year accumulated survival rate of 70%,with nearly 20%five-year mortality,patients' five-year survival rate in decompensated cirrhosis were only about one half,and even low to 35%.Anti-viral therapy of nucleot(s)ide analogues such as lamivudine, entecavir and telbivudine made it possible for us to reverse hepatitis B related fibrosis/cirrhosis.Being of great importance for improving patients' survival rate and quality of life,we should detect compensated cirrhosis as early as possible and take effective anti-virus therapy.Accurately determining the hepatic histological damages, liver biopsy is still considered as gold standard for assessing grade and stage of liver diseases.However,liver biopsy is an invasive method,and also yield false-negative results in nearly one-third of cases.Furthermore,liver biopsy is also limited by sampling error and inter-observer variability.As the role of reversing hepatic fibrosis of nucleot(s)ide analogue therapy becoming more and more obvious,it is necessary to explore non-invasive and reproducible alternative methods for detecting cirrhosis and evaluating anti-virus effect.Several models for significant fibrosis or cirrhosis had been introduced for hepatitis C,such as fibrosis marker of hyaluronic acid and other multiple variables models, including PGA,Forns,APRI,FibroTest and DS score.Among them,the application of Fibro Test had been reported in several validating study.However,these models were established mainly for hepatic fibrosis of hepatitis C and alcoholic liver diseases, but seldom for hepatitis B.As the pathogenesis and pathology of CHB were virtually different from hepatitis C,which study emphasis should be severe fibrosis and cirrhosis,model for hepatitis C may not be suitable for CHB.In recent years, FibroScan study was popular in Europe and Japan,which was another alternative method for detecting hepatic fibrosis in patients with chronic liver diseases.Relying on reported studies about non-invasive diagnosis for hepatic fibrosis or cirrhosis,and also large cohort of liver biopsied patients in our Hepatology Unit,our objective of present study was to construct a non-invasive model for detecting compensated cirrhosis in patients with CHB.With this model,most of the CHB patients would be determined as absence or presence of cirrhosis and then be free from liver biopsy.Of course,minority of patients still can not be accurately identified, which still need biopsy to determine the state of cirrhosis.Patients and Methods: Table 1 Total of 1051 patients with chronic hepatitis B included in studyTotal of 1051 patients with CHB of in Department of Infectious Disease of our hospital from AUG 1998 were analyzed(Table 1).All patients had been undergone percutaneous liver biopsies,and positive for serum HBsAg more than 6 months,these patients were compensated for liver function,characterized as Child-Turcotte-Pugh A grade.Exclusion criteria included autoimmune liver disease,evidence of alcoholic or fatty liver disease;drug induced hepatic injury,metabolic disease,combined positive for other hepatitis virus markers and other spleen,and gallbladder related diseases. Blood tests were performed within one week of liver biopsy,which including hematology,10 serum biochemical indexes,coagulations index and serum alpha fetoproteins(AFP).Patients' age were also recorded.Patients between AUG 1998 and DEC 2001 and patients after OCT 2007 were also detected fibrosis serum markers of hyaluronic acid(HA),laminin,typeⅣof collagen and typeⅢof precollagen(detected by radio-immunity assay).Ultrasound(US) examination was performed in patients after JAN 2003.The ultrasonic assessment of liver surface,liver parenchyma,gallbladder,hepatic vessel, splenomegaly and diameter of main portal vein were recorded.Patients after JUN 2007 were received FibroScan scanning in 191 patients.Blood tests and Imageological Examination analyzed in different periods were listed in Table 1 in detail.Liver biopsy specimens were obtained using a 16 gauge Menghini biopsy needle, and at least 1.0 centimeter long,each containing at least 5 portal tracts.The fibrosis score of portal tract was graded by METAVIR scoring system.Spearman's rho rank correlation and partial correlation analysis were used to evaluate correlation between liver inflammation and fibrosis.Continuous variables were expressed as mean±SD and compared using one-way ANOVA or student's t test.Ranked or symbol data were compared with Kruskal-Wallis' H test or Mann-Whitney U Test.For the formulation of predictive model,Logistic stepwise regression analysis was performed.The diagnostic value of formula was assessed by AUC.A two-sided P value of less than 0.05 was considered statistically significant. Statistical analysis was performed by SPSS software version 15.0.Results:1.Fibrosis Serum markers and routine blood test indexes.The rank correlation coefficient between hepatic fibrosis and serum HA,typeⅢprecollagen and laminin was 0.596,0.357 and 0.264,respectively(P=0.000).The correlation between fibrosis and typeⅣof collagen was not statistically significant. While controlling of liver inflammation and fibrosis,serum HA was correlated with hepatic fibrosis stages(r=0.298,P=0.000),and typeⅢprecollagen was correlated with inflammation grades(r=0.212,P=0.000).Among fibrosis markers in patients with cirrhosis,only serum level of HA was significantly higher than other fibrosis stages.AUC of HA for diagnosing cirrhosis was 0.846,which was higher than other fibrosis markers(0.646,0.513,0.705 for typeⅢofprecollagen,laminin and typeⅣof collagen in the same order).Diagnostic sensitivity and specificity of HA 120μg for determining cirrhosis was 83.3%and 71%.HA 94μg for excluding cirrhosis owned a negative predictive value(NPV) 93.7%and 220μg for determining cirrhosis with positive predictive value(PPV) 77.8%,and only 8.3%patients with mild fibrosis were found in cirrhosis determined patients.With excluding and determining cut-off value,serum HA would free 68%patients from biopsy.Patients' ages,platelet,ratio of AST/ALT,serum protein,serum direct bilirubin and prothrombin time were different between patients with and without cirrhosis.While incorporated with routine blood tests,hyaluronic acid and platelet were included to identify cirrhosis,with AUC of 0.888.With excluding model index of-1.76,NPV for cirrhosis was 98%,and PPV 74.4%for determining index -0.12,with mild fibrosis rate of 5.1%.With these two cut-off values,model consisting of hyaluronic acid and platelet would free 78%patients from biopsy.2.Routine blood tests for hepatitis and US score system.While inflammation controlled by partial correlation analysis,blood indexes of platelet,ratio of AST/ALT,albumin,bilirubin,bile acid,prothrombin time and white blood were correlated with fibrosis stages with descending order(r varied between 0.292 and 0.109).Apart from alkaline phosphatase,routine blood indexes in patients with cirrhosis were significantly different from other fibrosis stages.As more than 20%of data missed,indexes of bile acid,cholinesterase and AFP were excluded from logistic analysis.Model for cirrhosis diagnosis consists of platelet,bilirubin and protein ratio of albumin and globulin,with AUC 0.876.P value of Hosmer and Lemeshow test for model was 0.921.With model index-1.83,NPV for cirrhosis was 97.0%,and PPV 76%for determining index 0.4,with mild fibrosis rate of 2%.Used with these two cut-off values,model would free about 60%patient from biopsy. In US score system,there were significantly differences between fibrosis stages and ultrasonic indexes of liver parenchyma,gallbladder,hepatic vessel and splenomegaly.Ultrasonic scores were significantly different between F4 and other fibrosis,and significantly correlated with fibrosis stages(rho=0.461).Used with excluding and determining value,US score would free 50%patient from biopsy,with AUC 0.793.NPV of cirrhosis for normal ultrasound images of liver,spleen and gallbladder was 93.6%.Score of 9 determined cirrhosis with PPV 76%,with mild fibrosis rate 13.6%.If routine blood tests and US score combined used for identifying cirrhosis,model for cirrhosis consists of patients' age,blood platelet,US score,serum albumin and bilirubin,with AUC of 0.907.P value of Hosmer and Lemeshow test for model was 0.854.The cirrhosis index was correlated with fibrosis stages with coefficient of 0.637.Model would free 70%patients from biopsy,with NPV 97.4%for excuding index-2.18 and PPV 84.3%for diagnosing index 0.24,and mild fibrosis in cirrhosis diagnosed patients was only 2.9%.Model in validation set showed similar excluding but lower determining ability.3.FibroScan scanning and blood test,ultrasounds score system.The rank correlation and partial correlation coefficient between inflammation and liver stiffness were 0.689(P=0.000) and 0.127(P=0.080),respectively.The same coefficient of hepatic fibrosis and liver stiffness were 0.702(P=0.000) and 0.374 (P=0.000),respectively.These coefficients showed that liver stiffness was correlated mainly with hepatic fibrosis.Liver stiffness was significantly different between fibrosis stages,with AUC 0.911 for identifying compensated cirrhosis.Liver stiffness 13.5KPa owned a maximal cumulant of sensitivity and specificity,with NPV 97.1%. With this cut-off value,84.2%of non-cirrhosis patients were detected.While liver stiffness up to 21.8KPa,PPV for cirrhosis was 66.7%,with nearly 0 of mild fibrosis. With absence and presence of cirrhosis cut-off value,FibroScan would free 85% patients from biopsy,and 78%patients were identified correctly.Multi-variable Logistic regression analysis showed that compensated cirrhosis diagnosed model consisted of platelet and FibroScan liver stiffness,with AUC 0.912, and corresponding FibroScan AUC for cirrhosis was 0.913.With this model,the biopsy free rate was 80%and 72.8%patients would be correctly determined. Forthermore,mild fibrosis was 0 in the cirrhosis diagnosed patients.In 70 patients' FibroScan,routine serum tests and serum hyaluronic acid,FibroScan was the only variable for diagnosing cirrhosis.Discussion and Conclusion:In present study for patients with CHB,majority of routine blood tests for hepatitis were correlated with hepatic fibrosis stages.These variables included four type of indexes:demography of age,fibrosis markers of serum hyaluronic acid,typeⅢof precollagen and typeⅣof collagen,routine blood tests of white blood cell,platelet, AST,ALT,AST/ALT ratio,albumin,globulin,protein A/G ratio,bilirubin,direct bilirubin,bile acid and prothrombin time,and imageology of ultrasound and FibroScan liver stiffness.Patients' age,blood platelet,serum hyaluronic acid,albumin,protein A/G ratio, bilirubin,US indexes of liver,spleen and gallbladder and FibroScan liver stiffness were involved in cirrhosis diagnosis model construction.Furthermore,platelet was invoved in all multi-indexes models.Patients' age suggested the history of hepatitis activity,which would result in liver fibrosis progress.Thrombocytopenia in liver cirrhosis may be partially due to the accumulation and destruction of platelet in the portal hypertensive-enlarged spleen resulting from progressive liver fibrosis,and partially due to the impaired production of thrombopoietin in the cirrhotic liver. Serum albumin and bilirubin were the indexes in Child-Turcotte-Pugh score for liver function evaluation.Hypoalbuminemia and hyperbilirubinemia were both indication of impaired liver function resulted from continuous fibrosis.They were acceptable and diagnostic biochemical features of liver cirrhosis.Serum hyaluronic acid used as fibrosis marker was a custom practice in clinical work for years.US was also a common examination for detecting cirrhosis.Ultimately,FibroScan detected hepatic fibrosis by determining live stiffness,which was positively correlated with fibrosis stages. Regression analysis ultimately constructed several models for diagnosing compensated cirrhosis,including blood indexes(platelet,bilirubin and protein A/G ratio),combination of blood indexes and ultrasound(including patients' age, ultrasound score,platelet,bilirubin and albumin),platelet incorporated with hyaluronic acid and platelet plus FibroScan.The clinical significance of these diagnosing models was that most of patients with CHB would free from biopsies due to being identified as absence or presence of cirrhosis by excluding and determining cut-off value,while others were still needed biopsies to confirm the state of cirrhosis.Model of blood indexes,hyaluronic acid plus platelet,combination of blood indexes and US,FibroScan,FibroScan plus platelet were of more applicable value, with NPV more than 97.0%in cirrhosis excluded and mild fibrosis rate less than 5% in cirrhosis determined patients.By these model,the misdiagnosis patients were majority sever fibrosis(which would progressed to cirrhosis without anti-virus treatment),and thus can be ignored(table 2).Furthermore,there was no mild fibrosis in misdiagnosis patients identified by model consisted of platelet and FibroScan.The biopsy free rate of these model diagnosing compensated cirrhosis were 61.3%,73%, 71.5%,84.4%and 79.3%in the same order.Table 2 Evaluation of diagnostic value for compensated cirrhosis diagnosis by several modelPresent study indicated that model consisted by FibroScan single index or combined with platelet be of more applicable value in the non-invasive diagnosis of compensated hepatitis B related cirrhosis.Without FibroScan,hyaluronic acid plus platelet,blood indexes combined with US score would still play important role in non-invasive identification of compensated cirrhosis resulted from HBV infection.