Potential conflict of interest: Nothing to report.
Liver biopsy is recommended before antiviral treatment, particularly for patients with hepatitis C virus (HCV) genotype 1 infection, but it may cause complications and is limited by sampling error. Several non-invasive tests comprising routine laboratory parameters (simple fibrosis tests) have been proposed to predict fibrosis in chronic HCV. The aim of the current study was to validate and compare the diagnostic accuracies of the simple fibrosis tests, aspartate aminotransferase (AST)/alanine aminotransferase (ALT) ratio (AAR), cirrhosis discriminant score (CDS), age-platelet (AP) index, Pohl score, AST-to-platelet ratio index (APRI), and platelet count per se. Staging was performed in liver biopsy specimens of 194 treatment-naïve patients with chronic HCV according to Ishak et al. by two independent pathologists. Receiver operating characteristic curve analysis showed comparable diagnostic accuracies of CDS, AP index, APRI, and platelet count for prediction of significant fibrosis (F3-F6) (area under the ROC curve [AUROC], 0.71, 0.74, 0.80, and 0.71, respectively; pathologist A) and for prediction of cirrhosis (F5-F6) (AUROC, 0.91, 0.91, 0.90, and 0.89, respectively; pathologist A). Diagnostic accuracy of APRI for prediction of significant fibrosis was superior to that of AAR (P < .05). Significant fibrosis was reliably predicted by APRI ≥ 1.5 and platelet count <150 ×109/L in 24% and 22% of the patients, respectively, whereas cirrhosis was reliably excluded by APRI <2.0 and platelet count ≥ 150 ×109/L in 85% and 78% of the patients, respectively. In conclusion, simple fibrosis tests may render liver biopsy unnecessary only in a minority of patients with chronic HCV. Improved serum fibrosis markers with greater sensitivity for severe fibrosis or cirrhosis are needed. (HEPATOLOGY 2005;41:1376–1382.)
Chronic infection with the hepatitis C virus (HCV) is one of the most common causes of chronic liver disease in the western world and is complicated by cirrhosis or hepatocellular carcinoma in approximately 20% and 5% of patients, respectively.1, 2 Current antiviral treatment with pegylated interferon alfa plus ribavirin clears HCV infection in approximately 40% to 90% of patients, depending on HCV genotype.3
Because antiviral treatment is costly and may be associated with significant side effects, the indication for antiviral therapy has to be evaluated carefully. Antiviral treatment is indicated for most patients infected with the easier-to-treat HCV genotypes 2 or 3. In contrast, the decision for treatment in the more difficult-to-treat HCV genotype 1–infected patients should be based on additional prognostic factors, such as the degree of hepatic fibrosis at the time of liver biopsy. Accordingly, recent American Association for the Study of Liver Diseases (AASLD) practice guidelines recommend antiviral treatment for patients with HCV genotype 1 infection only in cases with significant liver fibrosis (METAVIR fibrosis score ≥ 2/Ishak fibrosis score ≥ 3).3 This concept is based on the finding that the risk of developing cirrhosis is closely related to the stage of fibrosis.4 Liver biopsy is thus recommended to aid in the decision to treat patients infected with HCV genotype 1, but its value is being questioned because of its potential risks and the concern of sampling error.5 Therefore, noninvasive tests for hepatic fibrosis are desired by clinicians as well as patients.6
Several laboratory tests, scores, and indices have been proposed for noninvasive prediction of hepatic fibrosis in patients with chronic hepatitis C.7–23 Among these, aspartate aminotransferase (AST)/alanine aminotransferase (ALT) ratio (AAR),7–12 cirrhosis discriminant score (CDS),13 age-platelet index (AP index),14 platelet count per se,15 Pohl score,16 and AST-to-platelet ratio index (APRI)17 are based on routine laboratory parameters and therefore readily available in clinical practice. All of these tests were reported to predict the presence of significant fibrosis, bridging fibrosis, or cirrhosis with considerable diagnostic accuracies. However, discrepant results were reported for some of these tests when evaluated in different study populations.24, 25
The aim of the current study was to evaluate and compare diagnostic accuracies of AAR, CDS, AP index, Pohl score, APRI, and platelet count for the prediction of significant fibrosis or cirrhosis.
HCV, hepatitis C virus; AST, aspartate aminotransferase; ALT, alanine aminotransferase; AAR, AST/ALT ratio; CDS, cirrhosis discriminant score; AP index, age-platelet index; APRI, AST-to-platelet ratio index; HAI, hepatic activity index; INR; international normalized ratio; ULN, upper limit of normal; ROC, receiver operating characteristic; PPV, positive predictive value; NPV, negative predictive value; AUROC, area under the ROC curve.
Patients and Methods
The study cohort included 211 consecutive patients with chronic hepatitis C who underwent percutaneous liver biopsy at the Medical University Graz or at the Landeskrankenhaus Hoergas, Austria, between 1994 and 2004. All patients tested positive for the presence of HCV RNA using a polymerase chain reaction assay and did not suffer from additional causes of chronic liver disease as confirmed by standard clinical, serological, biochemical, and radiological criteria. Additional exclusion criteria were antiviral treatment before liver biopsy, alcohol consumption in excess of 20 g/d, and previous liver transplantation.
Liver tissues were fixed in formalin immediately after biopsy and embedded in paraffin; 4-μm sections were cut with a microtome. One section of each biopsy was stained with hematoxylin-eosin for assessment of hepatic inflammatory activity and the other section with the chromotrope aniline blue for evaluation of hepatic fibrosis. Biopsy specimens with at least six portal fields were considered representative.5 Seventeen patients whose biopsy specimens showed fewer than six portal fields were excluded, leaving 194 patients for analysis. Histological grading (modified hepatic activity index [HAI]) and staging (fibrosis score) were performed independently by two pathologists (pathologist A, C.L.; pathologist B, S.L. or B.L.) as described by Ishak et al.26
AST, ALT, prothrombin time (international normalized ratio [INR]), and platelet count were routinely determined in the clinical laboratory. Because of the introduction of the International Federation of Clinical Chemistry reference method for the determination of aminotransferase activities at 37°C, the upper limits of normal (ULN) for AST and ALT changed in the course of the study (ULN before March 2003: AST, 18 U/L; ALT, 22 U/L; after March 2003: AST, 35 U/L ♂ or 30 U/L ♀, ALT, 45 U/L ♂ or 35 U/L ♀). Therefore, both AST and ALT were transformed into multiples of the ULN for further analysis except for the calculation of AAR. The reference range for platelet count was 140 to 440 ×109/L. Laboratory parameters were available on the day of liver biopsy (n = 96) or from within 1 month of the date of liver biopsy (n = 98). From these routine laboratory values, AAR, CDS, AP index, Pohl score, and APRI were calculated exactly as originally described13, 14, 16, 17 (Table 1).
Table 1. Simple Fibrosis Tests Composed of Routine Laboratory Parameters
Abbreviations: AAR, AST/ALT ratio; CDS, cirrhosis discriminant score; AP index, age-platelet index; APRI, AST-to-platelet ratio index; ULN, upper limit of normal.
AP index is the sum of the above (possible value 0–10).
Positive: AAR ≥1 and platelet count <150 ×109/L
([AST/ULN]/platelet count [×109/L]) × 100
Statistical analysis was performed using the SPSS software version 11.0 (SPSS Inc., Chicago, IL). Patient characteristics are given as mean ± SD or median and interquartile range as appropriate. Bivariate Spearman's rank correlation coefficient (R) was calculated to measure the relationship between variables. Kappa statistics was used to measure agreement between the two pathologists in determining the degree of fibrosis. Receiver operating characteristic (ROC) curves were constructed for each test. To evaluate diagnostic accuracies of simple fibrosis tests, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the area under the receiver operating characteristic curve (AUROC) were calculated. Individual AUROCs of continuous variables, such as AAR, APRI, and platelet count, were compared according to the procedure proposed by Hanley and McNeil.27
Clinical characteristics of the patients and standard laboratory tests at the time of liver biopsy are summarized in Table 2. HCV genotype was known in 141 patients. Genotype 1 was present in 119 patients, genotype 2 in 3 patients, genotype 3 in 10 patients, and genotype 4 in 9 patients. Ishak's modified HAI ranged from 1 to 11. Significant fibrosis (F3-F6) was found in 50% or 46% and cirrhosis (F5-F6) in 16% or 17% of patients (pathologist A or B, respectively). Good agreement was found between pathologists A and B for the Ishak fibrosis score in predicting significant fibrosis (Kappa = 0.79) and cirrhosis (Kappa = 0.93).
Table 2. Characteristics of the Study Population (n = 194)
48 ± 12
Male sex, n (%)
Genotype 1, n (%)
25.0 ± 4.0
2.2 ± 1.9
2.8 ± 2.0
Prothrombin time (INR)
1.01 ± 0.11
Platelet count (×109/L)
204 ± 66
Length of biopsy core (mm)
19 ± 8
Number of portal tracts, median (Q1, Q3)
11 (9, 16)
NOTE. Results are given as means ± SD, unless indicated otherwise.
Abbreviations: BMI, body mass index; ULN, upper limit of normal; INR, international normalized ratio; Q1, first quartile; Q3, third quartile.
Modified HAI score, median (Q1, Q3)
6 (4, 7)
5 (4, 6)
Mild (0–6), n
Moderate (7–12), n
Severe (13–18), n
Ishak fibrosis score, median (Q1, Q3)
2.5 (1, 4)
2 (1, 3)
Weak to moderate correlations were found between HAI versus AST, ALT, and INR, respectively (Table 3). Hepatic fibrosis weakly correlated with ALT, INR, AAR, and CDS, whereas moderate correlations were found with AST, platelet count, AP index, and APRI (Table 4). ROC curves evaluating the diagnostic accuracies of AAR, CDS, AP index, APRI, and platelet count were constructed to predict significant fibrosis (Fig. 1) or cirrhosis (Fig. 2). Comparison of AUROCs for continuous variables (AAR, APRI, and platelet count) by the procedure proposed by Hanley and McNeil27 showed superior diagnostic accuracy of APRI over AAR for prediction of significant fibrosis (Table 5). Similar results were found for all correlations and ROC analyses when hepatic fibrosis was assessed by either pathologist A or B (Tables 3–5).
Table 3. Correlation of Grade (Modified Hepatic Activity Index) With Standard Laboratory Parameters
Modified HAI vs.
Bivariate Spearman's Rank Correlation Coefficient
Abbreviations: ULN, upper limit of normal; INR, international normalized ratio.
Table 5. Performance of Simple Fibrosis Tests for Prediction of Significant Fibrosis (F3–F6) or Cirrhosis (F5–F6)
Ishak Fibrosis Score vs.
Abbreviations: AUROC, area under the receiver operating characteristic curve (confidence intervals are given in parentheses); AAR, AST/ALT ratio; CDS, cirrhosis discriminant score; AP index, age-platelet index; APRI, AST-to-platelet ratio index.
P < .05 vs. AUROC 0.5.
P < .05 vs. AAR (comparison of AUROCs according to Hanley and McNeil27).
To compare our results with those of the previous publications, sensitivity, specificity, PPV, and NPV of simple fibrosis tests were calculated for each pathologist's staging using categories and cutoff values exactly as originally described (Table 6). For platelet count, an optimal cutoff of 150 ×109/L was determined by ROC analysis. High PPVs for significant fibrosis and high NPVs for cirrhosis were found with platelet count as well as APRI. Platelet count per se demonstrated high predictive values for both diagnosis of significant fibrosis (PPV 93%, pathologists A and B, respectively) and exclusion of cirrhosis (NPV 95%, pathologists A and B, respectively) at the same cutoff level of 150 ×109/L. Similarly, an APRI ≥ 1.5 had a PPV of 91% or 83% (pathologist A or B, respectively) for significant fibrosis, whereas an APRI < 2.0 had an NPV of 91% (pathologists A and B, respectively) for cirrhosis. A platelet count of less than 150 ×109/L or an APRI of 1.5 or greater allowed reliable diagnosis of significant fibrosis in 22% or 24% of our patients, respectively (i.e., in about half of the patients with ‘true’ significant fibrosis) whereas a CDS of 8 or higher was diagnostic for bridging fibrosis (F4-F6) in only 2% of the patients. Conversely, a platelet count of 150 ×109/L or greater or an APRI of less than 2.0 reliably excluded cirrhosis in 78% or 85% of the patients, respectively.
Table 6. Diagnostic Accuracies of Simple Fibrosis Tests for Prediction of Significant Fibrosis (F3–F6), Bridging Fibrosis (F4–F6), and Cirrhosis (F5–F6)
Patients Classified (%)
Diagnostic Accuracy (%)
Ishak Fibrosis Score
Pathologist A/B (literature)
NOTE. Data are shown for Ishak fibrosis score as assessed by pathologist A or B (values reported in the original studies are given in parentheses). The cutoffs described in the original reports were used for all calculations. For AAR and platelet count, Ishak stages F5-F6 in our study were compared with presence of cirrhosis in the Giannini study12; for CDS, Ishak stages F4-F6 in our study were compared with modified Knodell stages F3-F4 in the Bonacini study13; for AP index, Ishak stages F3-F6 in our study were compared with METAVIR stages F2-F4 in the Poynard study14; and for Pohl score, Ishak stages F4-F6 in our study were compared with METAVIR stages F3-F4 in the Pohl study.16
Overall the performance of AP index, Pohl score, and APRI in our study was in good agreement with the data reported by Poynard et al.,14 Pohl et al.,16 and Wai et al.,17 respectively. In contrast, PPV of AAR and platelet count (cutoff 130 ×109/L) for cirrhosis was considerably lower in our study than in the report by Giannini et al.12 (Table 6).
Several routine laboratory parameters, used either alone or in combination, have been evaluated as noninvasive predictors of hepatic fibrosis.7–18 In addition, special fibrosis markers such as hyaluronic acid19 and several multicomponent tests such as a five-marker index composed of α2-macroglobulin, apolipoprotein A1, haptoglobin, total bilirubin, and gamma-glutamyl-transpeptidase (Fibrotest, Biopredictive, France)21, 22 or, more recently, an algorithm combining age, hyaluronic acid, amino-terminal propeptide of type III collagen, and tissue inhibitor of matrix metalloproteinase 123 were proposed for non-invasive diagnosis of hepatic fibrosis. Assessment of these special markers is, however, costly and not routinely available in most clinical laboratories. All noninvasive fibrosis tests evaluated in the current study (AAR, CDS, AP index, Pohl score, and APRI) are derived from four readily available laboratory parameters, that is, AST, ALT, platelet count, and prothrombin time (INR).
Simple fibrosis tests are not yet widely applied in clinical practice and recent AASLD guidelines on diagnosis and treatment of chronic hepatitis C do not include recommendations for their use.3 This may be related to the fact that their diagnostic accuracies have not yet been sufficiently validated by other investigators in different study populations. In the current study, we confirm diagnostic accuracies at the reported cutoffs for CDS, AP index, Pohl score, and APRI but not AAR for significant fibrosis or cirrhosis. It should be pointed out that platelet count per se had similar diagnostic accuracies as the other simple fibrosis tests comprising platelet count among other laboratory parameters.
Consistent with our findings, Wai et al.17 also reported low diagnostic accuracy of AAR in predicting significant fibrosis (AUROC 0.60) and cirrhosis (AUROC 0.76) (Giannini and Testa,24 author reply). Differences in the study populations and the histopathological assessment of fibrosis may be responsible for these discrepancies. For example, prevalence of cirrhosis was 36% in the population studied by Giannini et al.12 but only 15% in the Wai study17 and 16% to 17% in our study. In the Giannini study, cirrhosis was mainly diagnosed on the basis of clinical signs of portal hypertension (and not on histology), indicating more advanced cirrhosis in their patients. Furthermore, it should be pointed out that four different staging systems were used in the respective studies: Giannini et al. used the Scheuer system,28 Bonacini et al. applied a modified Knodell system,29 Poynard et al. and Pohl et al. used the METAVIR system,30 and the Ishak system26 was used in the Wai study and in our study. Comparison between these studies is further hampered by the fact that the different staging systems use either a 5-step (F0-F4) or a 7-step (F0-F6) scale, even though good agreement of staging by the METAVIR and Ishak systems was recently reported.31
In contrast to the other tests, APRI uses 2 different cutoff values for exclusion or prediction of significant fibrosis (<0.5/≤1.5) and thus allowed reliable diagnosis or exclusion of significant fibrosis in approximately 50% of patients in the study by Wai et al.17 We could confirm a high PPV for significant fibrosis of an APRI ≥1.5 in 24% of our patients; however, the NPV of an APRI <0.5 for exclusion of significant fibrosis was found to be slightly lower (80% vs. 86% in the Wai et al. study). This discrepancy may be attributable to differences in the study populations. APRI was recently evaluated in a large cohort of treatment-naive patients with chronic hepatitis C undergoing various clinical trials.25 The diagnostic accuracy of APRI in predicting hepatic fibrosis or cirrhosis was somewhat lower than in the Wai study and in our study. This may be related to the use of different staging systems (the Scheuer system vs. the Ishak system) and to the fact that histology was assessed by several pathologists in different trial centers as opposed to one single pathologist in the Wai study or two independent pathologists in our study.
The inverse correlation of platelet count and the degree of hepatic fibrosis in chronic hepatitis C has been noted by several investigators.13–17 Thrombocytopenia in patients with advanced hepatic fibrosis may be explained by portal hypertension leading to pooling of platelets in an enlarged spleen, a myelosuppressive action of HCV, or reduced hepatic production of thrombopoietin.32 However, only a few studies have assessed the diagnostic value of platelet count per se for prediction of hepatic fibrosis in chronic hepatitis C.12, 15, 20, 22 Comparison of these studies with our findings is again hampered by differences in the study design and staging systems. Murawaki et al.20 evaluated 165 patients with chronic hepatitis C and fibrosis stages F0-F3 (Scheuer system) and reported that the platelet count at a cutoff of 160 ×109/L was 68% sensitive and 71% specific for prediction of F2-F3 fibrosis.20 Giannini et al.12 reported high diagnostic accuracy of platelet count for prediction of cirrhosis at a cutoff level of 130 ×109/L. However, at this cutoff we found a lower sensitivity and PPV of platelet count for prediction of cirrhosis (Table 6), which may be explained by differences in the composition of the study cohort. Myers et al.22 compared diagnostic accuracies of Fibrotest, prothrombin time, platelet count, and AP index in 323 patients with chronic hepatitis C and reported an AUROC of 0.67 for platelet count and of 0.72 for AP index for prediction of F2-F4 fibrosis (METAVIR system), which correspond well to our findings.22
Based on a comprehensive systematic review, Gebo et al.33 concluded that biochemical blood tests and serologic tests have only modest value in predicting fibrosis on liver biopsy. We agree that the moderate diagnostic accuracy of simple fibrosis tests for intermediate stages of fibrosis limits their application. Overall, the simple fibrosis tests evaluated in this study were less accurate for prediction of significant fibrosis than for prediction of cirrhosis, and none of them were found to reliably exclude significant fibrosis. Nevertheless, both APRI ≥ 1.5 and platelet count <150 ×109/L can identify significant fibrosis with a PPV >90% in subsets of patients with chronic hepatitis C. In summary, by using specific cutoff points, clinically important stages of chronic hepatitis C can be reliably identified without a liver biopsy: (1) significant fibrosis can be predicted in approximately 25%, and (2) cirrhosis can be excluded in approximately 80% of patients typically seen in a referral center.
The authors thank Profs. Helmut Denk and Gerald Hoefler for fruitful discussion of the manuscript.