Diagnostic accuracy of biochemical markers of fibrosis in black African patients with chronic hepatitis B

Contradictory results of the accuracy of biochemical markers to predict the stage of fibrosis in black African patients with chronic hepatitis B (CHB) were previously published. We conducted a prospective cohort study to determine the diagnostic accuracy of aspartate aminotransferase to platelet ratio (APRI), aspartate aminotransferase to alanine aminotransferase ratio (AAR), platelet count, age-platelet (AP) index, and FIB-4 index for the prediction of significant fibrosis or cirrhosis in 117 black African patients (median age: 38 years, males: 73%) with CHB not previously treated. Among them, 45 had significant fibrosis and 18 had cirrhosis using the METAVIR score system. Factors associated either with significant fibrosis or cirrhosis were determined in logistic multivariate analysis. Areas under receiver operating curve were assessed and compared for APRI, AAR, AP index, FIB-4 index and platelet count. Sensitivity, specificity, positive and negative predictive values were determined for each biochemical markers. Multivariate analysis showed that aspartate aminotransferase (p < 0.0001) and platelets (p = 0.03) were the independent factors associated with significant fibrosis and only platelets (p = 0.01) were associated with cirrhosis. APRI (cut-off > 1.1) and FIB-4 index (cut-off > 2.1) ruled out significant fibrosis with high specificity of 84.7% and 86.1% respectively and negative predictive values of 78.2% and 72.9% respectively. More accurately, APRI (cut-off > 0.63) or FIB-4 index (cut-off > 1.26) ruled out cirrhosis with high sensitivity of 94.4% and 88.9% and high negative predictive values of 98.1% and 96.3% respectively. In conclusion, APRI and FIB-4 index are simple readily available markers to exclude significant fibrosis or more accurately cirrhosis in black African patients with CHB.


INTRODUCTION
Chronic hepatitis B (CHB) is the major cause of chronic liver disease that affects approximately 350 millions individuals worldwide leading to cirrhosis and hepatocellular carcinoma.Africa and Asia have the highest prevalence of hepatitis B virus infection worldwide [1,2].It is obvious that treatment of CHB is a challenge for clinicians in Africa.The assessment of liver disease in patients with CHB by the mean of liver biopsy is not a mandatory but needed at baseline to exclude others causes of liver disease or to evaluate the histologic damage in the liver before the initiation of treatment after long term follow up [2][3][4].However, liver biopsy is an invasive, costly and not completely safe mean that can lead to severe complications with a potential risk of mortality.In addition, it is limited by sampling error and poor concordance between two observers [5,6].
Viral hepatitis C is the more prevalent chronic liver disease in western countries [7].Non invasive means are constructed to determine the stage of fibrosis in patients with chronic hepatitis C with acceptable accuracies and therefore the need of liver biopsy can be obviated [8][9][10][11][12][13][14].
Several studies attempt to use some of these means in patients with CHB and yield unsatisfactorily results [15][16][17].WAI et al. show that aspartate aminotransferase (AST) to platelet ratio (APRI), aspartate aminotransferase to alanine aminotransferase (ALT) ratio (AAR) and platelet count were not accurate in predicting either significant fibrosis or cirrhosis among patients with CHB.These findings are confirmed by Kim et al. for the prediction of cirrhosis in patients with CHB.All these tests are validated in Asian patients [16,17].
The diagnostic accuracy of biochemical markers for the prediction of the stage of fibrosis or cirrhosis in black African patients was not evaluated enough.One study suggested that FIB-4 index was accurate to exclude significant fibrosis in black African patients with CHB [18].Recently discordant results were published by Bonnard et al. that demonstrated low accuracy of FIB-4 index and APRI for the prediction of significant fibrosis or cirrhosis in this ethnic group of population [19].
The aim of this study was to evaluate the AAR, APRI, platelet count, age-platelet (AP) index, and FIB-4 index for the prediction of significant fibrosis and cirrhosis in black African patients with CHB attending the hepatology unit of the Teaching hospital of Yopougon in Abidjan, Ivory Coast, West Africa.

Patients
All consecutive patients with positive serum hepatitis B surface antigen (HBsAg) and HIV-negative referred to the hepatology unit at Yopougon teaching hospital were eligible for the study.CHB infection was considered if HBsAg persists 6 months after the date of the onset of acute hepatitis B infection.For those with unknown date of the onset of acute hepatitis B infection, CHB was considered if HBsAg is present in a secondary blood test performed 6 months after the date of the previous positive blood test [1,2].The exclusion criteria were concomitant liver disease such hepatitis D superinfection, hepatitis C coinfection, autoimmune hepatitis, decompensated cirrhosis, hepatic schistosomiasis, if blood test were positive for antibodies against schistosomiasis or the presence in liver biopsy, granulomatous and fibrotic reaction related to hepatic schistosomiasis [20] concomitant treatment against HBV, and alcohol consumption over 50 g/l.All patients gave the consent for liver biopsy and data were used according to Helsinki declaration.

Clinical Data
Following clinical data collected: age, sex, occupation, ethnicity, source of contamination, presumably date of contamination, past history of vaccination, alcohol consumption.

Biological Data
Before liver biopsy, blood test were performed in the laboratory of the Yopougon teaching hospital and provided following variables, haemoglobin, hematocrit, mean corpuscular volume, platelet count (Coulter T-540 Coulter Corporation Hialeah, Florida USA), prothrombin index (Coagulometer Option 4 plus Biomerieux Germany), AST, ALT (Automate Analyzer, Hitachi Ldt, Tokyo, Japan).The upper limit of normal of AST and ALT were different according to sex and regarding the normal values assigned by the laboratory AST: 31 and 36 IU/L, ALT: 34 and 43 IU/L, for respectively male and female.

Virological Assay
HBsAg, hepatitis B e antigen (HBeAg), and antibody, antibody to hepatitis B core antigen (Anti-HBc) were performed using commercial assays (Cobas, Roche Diagnostics, Mannheim) if treatment was needed according to the result of liver histology, ALT level or viral load [3,4].All virological assays were assessed at the Centre Intégré de Recherche Bioclinique d'Abidjan (CIRBA) laboratory of Abidjan.

Histological Assessment
Liver biopsies if not contraindicated were performed after liver ultrasonography within 7 days after blood testing.Needles of 1.6 mm diameter (Hepafix, Braun, Melsungen) were used for all patients and liver biopsy was performed according the Menghini technique [5].Liver biopsy specimens were formalin-fixed, paraffinembedded and stained with hematoxilin-eosin-safran and perls coloration for iron load.Necroinfammatory activity and fibrosis stage were assessed by one senior pathologist (DMI), who was blinded to the clinical examination and blood tests and according to the METAVIR semi quantitative system [21].The fibrosis stage was as follow: F0: no fibrosis, F1: portal fibrosis, F2: fibrosis with few septa, F3: fibrosis with numerous septa, F4: cirrhosis.The grade of activity was as follow: A0: no histological activity, A1: mild activity, A2: moderate activity, A3: severe activity.The length of biopsy specimen was not recorded for all patients.

Diagnostic Target
Two diagnostic targets (significant fibrosis and cirrhosis) assessed by the histologic findings after liver biopsy were used to identify four groups of patients.Patients with significant fibrosis (F2, F3, and F4) were compared to those without significant fibrosis (F0 and F1) and patients with cirrhosis (F4) were compared with those without cirrhosis (F0, F1, F2, and F3).

Statistical Analysis
Continuous variables were expressed as median and interquartile range and qualitative variables were expressed as percentage.Mann-Withney U test and χ 2 or Fischer test (if appropriated) were used to compare quantitative and qualitative variables.The correlations between AST, ALT and platelet count with the stage of fibrosis were assessed by the non parametric Spearman's rho test (rho).The AAR, APRI, AP index and FIB-4 index were determined by formula previously published and resumed in Table 1 [8,11,13,14].Variables significantly associated with significant fibrosis or cirrhosis in univariate analysis were tested using a multivariate logistic regression.The diagnostic accuracy with the best cut-off point that maximizes the sensitivity and the specificity and express as sensitivity, specificity, predictive positive and negative values were determined by the receiver operating characteristic curve (ROC) for platelet count, AAR, APRI, AP index and FIB-4 index.The areas under ROC of platelet count, AAR, APRI, AP index and FIB-4 index of each group were compared according to the Henley and Mac Neil test [22].Value of area under ROC of 1.0 indicated an ideal test whereas value of 0.5 indicated that the test was not significant.All tests were two-tailed and performed by SPSS for Windows version 11.0 (SPSS Inc., Chicago, IL).P value under 0.05 was considered as significant.

Factors Associated with Significant Fibrosis or Cirrhosis
Patients with significant fibrosis or cirrhosis had   higher level of ALT (ULN) or AST (ULN) and lower platelet count than those without significant fibrosis or cirrhosis in univariate analysis (Table 3).AST (ULN) and ALT (ULN) levels correlated positively (both correlation coefficients rho = 0.4, p < 0.0001) whereas platelet count correlated negatively (rho = -0.28,p = 0.002) with the stage of fibrosis.Overall comparison demonstrated significant difference between AST or Platelets count and the stage of fibrosis (Figure 1).In logistic multivariate analysis, AST (p < 0.0001) and platelet count (p = 0.03) were independent predictors of significant fibrosis whereas platelet count was the only independent factor associated with cirrhosis (p = 0.01).The median values (with interquartile range) of biochemical markers are summarised in Table 4.Besides AAR for significant fibrosis, biochemical markers showed significant but modest area under ROC either for the prediction of significant fibrosis (fibrosis stage ≥ F2) or cirrhosis (F4) in our study (Table 5).

Comparison of Biochemical Parameters for the Prediction of Significant Fibrosis
As illustrated in Figure 2(a), APRI, platelet count, FIB-4 index and AP index had similar diagnostic accu-   racy for the prediction of significant fibrosis.Howeverbetter performances were observed with APRI and FIB-4 index (Table 5).With a cut-off > 1.1 of APRI, significant fibrosis could be correctly excluded in 61 (52%) of 117 patients with 78.2% of NPV.Similar results were obtained with FIB-4 index (cut-off > 2.1) that identified correctly 62 (53%) of 117 as patients with no significant fibrosis.However the number of false negative patients was lower with FIB-4 index than APRI (8.5% and 14.5 % respectively).

Comparison of Biochemical Parameters for the Prediction of Cirrhosis
All biochemical markers had better performances to exclude cirrhosis (Figure 2(b)) with high negative predictive values > 80%.APRI and FIB-4 index had maximal sensitivity respectively 94.4% and 88.9% (Table 5).With a cut-off > 0.63 of APRI, cirrhosis was correctly excluded in 52 (52.5%) patients among 117 with a NPV of 98.1%.Among 18 patients with histological proven cirrhosis, 17 (94.4%)were correctly classified and 1 (1.9 %) patient was false negative.Applying a cut-off > 1.26 of FIB-4 index, we found similar results, 51 (51.5%) patients of 117 correctly classified as patients with no cirrhosis.Thus 16 (88.9%)patients of 18 with cirrhosis were diagnosed and 2 (3.8%) were false negative.

DISCUSSION
We demonstrated in this study that biochemical mark-ers previously assessed in chronic hepatitis C such as APRI, AAR, FIB-4 index, platelet count and AP index had low accuracy regarding their respective areas under ROC, to predict the presence of significant fibrosis or cirrhosis among black Africans with CHB.However some of them could be used to exclude more accurately cirrhosis in low medicalized countries of Africa.Indeed we found that in patient with cirrhosis, APRI and FIB-4 index enabled to eliminate cirrhosis with high degree of certainty better than that published by Bonnard et al., in Burkina Faso [19].This is probably related to the size of our sample twice larger.Indeed our patients were slightly older than those from Mayotte and some of them had platelet count in normal range despite moderate to severe fibrosis (Figure 1(b)) leading to misclassification of FIB-4 index.The length of liver biopsy fragment was probably not large enough in our study to enhance the accuracy of FIB-4 index as previously demonstrated [18].In other hand the interaction of parasitosis or other blood transmitted diseases as malaria and CHB both prevalent in sub-Saharan African countries could reduce the global performance of these biochemical markers in African patients [24].Nevertheless APRI and FIB-4 index could be used to monitor patient with CHB when liver biopsy is difficult to perform.Both were accurate to exclude cirrhosis among black Africans with high degree of NPV.This probably means that APRI and FIB-4 index are more reliable when cirrhosis occurs in sub Saharan African patients with CHB than westerners or other black people.Recent studies conducted in Asia demonstrated discordant results of APRI and FIB-4 index for the prediction of stage of fibrosis.Wu et al., founded low accuracy of APRI and FIB-4 index to predict significant fibrosis comparable to our findings [25].However these two biochemical markers showed better performances to predict severe fibrosis.Kim et al. demonstrated that FIB-4 allowed detecting cirrhosis with high degree of certainty than APRI, and AAR [26].These recent findings emphasized the difficulty to provide definite and invariable performance of biochemical markers in patients with CHB whatever the geographical area.This study was conducted in West Africa where the endemicity of HBV is high and enrolled only black Africans.Most of them acquired HBV infection perinataly or during childhood which is the most common route of transmission of HBV in this area [1].Liver damage (fibrosis and cirrhosis) occurs mainly in adulthood with a high risk of onset of hepatocellular carcinoma.This explained that 45% of patients in our study had significant fibrosis or cirrhosis and were eligible for treatment [1,2].
We were not able to seek the date of the onset of acute hepatitis B infection, to determine viral load or HBeAg and antibody for most of patients because of the high cost of their determination and high number of missing values.These parameters were not included in the analysis.Despite these limitations, this study pointed out some specificities in a population of black African patients.Firstly, AST and platelet count were independent predictors of significant fibrosis and only platelet count were associated with cirrhosis in multivariate analysis.[23].However, we think that the procedure of calculation and interpretation is not easy to fulfil in clinical practice.Bonnard et al. found that Elastometry were more reliable than APRI and FIB-4 index to predict significant fibrosis in black African patients but the device for it determination is expensive and not available in most hospitals of sub Saharan African countries [19].Liver biopsy remains in this area the only means to quantify liver fibrosis in patients with CHB.Guidelines and consensus conferences recommend to monitor patients with cirrhosis by routine alpha-fetoprotein determination and liver ultrasonography every 6 months [2,3].Close surveillance is recommended to detect the onset of oesophageal varices if absent at the time of the diagnosis of cirrhosis by routine endoscopy [37,38].Lamivudine, a nucleoside analogue is widely used in Africa as part of highly antiretroviral therapy in HIV infected patients.It is a cost effective drug affordable to most Africans for the treatment of CHB.However, patients with cirrhosis treated by lamivudine need close surveillance because of the risk of hepatitis flares and liver decompensation during treatment [2,39].
Most of these recommendations are difficult to fulfil in developing countries of Africa.Biochemical markers such as APRI, and FIB-4 index easy to determine and cost effective could be used to identify African patients with CHB who do not have cirrhosis, and in whom close surveillance could be delayed during treatment by lamivudine.

CONCLUSION
This study demonstrated that biochemical markers currently used in chronic hepatitis C, had low accuracy regarding their areas under ROC.However APRI and FIB-4 index could be used to exclude cirrhosis in black African patients with CHB with high certainty.Further studies enrolling a large sample of black African patients with CHB are needed to establish the clinical relevance of the accuracy of biochemical markers in African patient.

Figure 1 .
Figure 1.Box plots of AST (a) or platelet count (b) according to the stage of fibrosis defined by META-VIR score system.The box represents the interquartile range; the top and the bottom of the box are respectively the 25 th and 75 th percentile.The line across the box is the median.The lower and upper values are indicated by the whiskers.The circles represent the outliers.

Figure 2 .
Figure 2. Receiver operating characteristic curve of the five biochemical markers for the prediction of significant fibrosis (a) or cirrhosis (b) according to the METAVIR stage of fibrosis in black African patients with chronic hepatitis B. APRI: aspartate aminotransferase to platelet count ratio, AAR: aspartate aminotransferase to alanine aminotransferase ratio, AP: age-platelet index and FIB-4 index.

Table 1 .
Formulas of biochemical markers for the prediction of necroinflammatory activity and fibrosis stage.

Table 2 .
Baseline characteristic of 117 patients included.

Table 3 .
Univariate analysis of parameters between patients with and without significant fibrosis, and between patients with and without cirrhosis.

Table 4 .
Distribution of biochemical markers values.

Table 5 .
Diagnostic accuracy of biochemical markers for the prediction of significant fibrosis and cirrhosis.
AAR: Aspartate aminotransferase to alanine aminotransferase ratio, APRI: Aspartate aminotransferase to platelet count ratio, AP index: age-platelet index.
[23]17]y, biochemical markers assessed in this study with significant values of area under ROC had high ability to exclude patients with cirrhosis.Each of them showed high NPV > 85%.Regarding the area under ROC, our study confirmed the findings made byWai et  al. and Kim et al.in Asian patients[16,17].Furthermore our study is consistent with that of Hongo et al., in which the areas under the ROC of APRI and AP index were also modest respectively 0.76 and 0.74[23].In contrast to the study ofWai et al.