Among the complications of percutaneous liver biopsy are pain (10%-30%), bleeding (which may be severe and necessitate blood transfusion or emergency surgery), biliary peritonitis, and pneumothorax. In large series, mortality has been reported to range from 0.1%-0.01%[7,8]. Percutaneous liver biopsy is contraindicated in the presence of coagulopathy, thrombocytopenia, and ascites. These contraindications may be in part obviated by transjugular liver biopsy; however, this method is not widely available.

An average needle biopsy specimen represents only 1/50 000 of a human liver. Sampling error thus may play an important role in diseases that exhibit a patchy rather than homogenous distribution within the liver. In a study investigating simultaneous biopsies from the left and right liver lobes during laparoscopy, fibrosis scores obtained in both biopsy sites differed by at least one stage in 33% of the patients[9]. Furthermore, Bedossa et al[10] demonstrated that the length of the biopsy core is positively related to the precision of fibrosis scoring. Likewise, Colloredo et al[11] reported that in liver biopsy specimens of inadequate size stage is likely to be underscored in chronic viral hepatitis. Based on these studies, liver biopsy cylinders with a length of ≥ 20 mm (at a width of 1.4 mm) and/or at least 11 complete portal tracts have been identified as minimal requirements for optimal histological evaluation of fibrosis in chronic hepatitis C. However, with respect to the evaluation of noninvasive fibrosis tests, most studies did not accurately report whether these requirements for adequate liver biopsy specimens were met[12].

As scoring is subjective, observer error also plays an important role. Interobserver variation is largely dependent on the experience of the pathologist. In a study evaluating interobserver variation between 10 different experienced pathologists, substantial agreement was found for staging fibrosis (kappa 0.78) while variation was considerably higher for grading inflammatory activity[13,14].

The interpretation of studies assessing hepatic fibrosis is further hampered by the lack of standardization of hepatic fibrosis scores. Several scoring systems have been developed that classify the degree of hepatic fibrosis either on a 5-step scale (F0-F4) including the Knodell fibrosis score[15], the Scheuer fibrosis score[16], and the METAVIR fibrosis score[13], or on a 7-step scale (F0-F6) such as the Ishak fibrosis score[17]. Significant fibrosis has been defined as F2-F4 or F3-F6 and cirrhosis as F4 or F5-F6, respectively. To date, liver pathologists have not reached a universal consensus on the standardization of scoring systems. However, histopathological scoring of fibrosis by different systems appears to be quite robust as comparison of the Ishak and METAVIR fibrosis scores yielded excellent agreement (weighted kappa 0.998)[14].

Almost three decades ago, the ratio of aspartate aminotransferase to alanine aminotransferase (AST/ALT ratio, AAR) has been proposed as a surrogate marker of hepatic fibrosis, with values > 1 being suggestive of cirrhosis[18]. This finding is related to an increased release of mitochondrial AST, decreased AST clearance and/or impaired synthesis of ALT in advanced liver disease. However, discrepant results have been published on the diagnostic accuracy of the AAR. Giannini et al[19] reported high diagnostic accuracy of the AAR for prediction of cirrhosis and significant fibrosis[20]. In contrast, Lackner et al[21] found the diagnostic accuracy of AAR to be clearly inferior to that of other indirect fibrosis tests based on routine laboratory parameters.

Hepatic fibrosis may lead to thrombocytopenia as a consequence of impaired synthesis of thrombopoietin and/or sequestering of platelets in an enlarged spleen. Surprisingly, few data exist on the diagnostic value of platelet count per se although the platelet count has been included in several composite fibrosis scores. Ono et al[22] reported the use of platelet count could discriminate F4 from F1-F3 in 75%-80% of patients with chronic hepatitis C. In our own study, a platelet count of < 150 × 10⁹/L had a positive predictive value (PPV) > 90% for significant fibrosis, whereas at a cut-off of ≥ 150 × 10⁹/L it had a negative predictive value (NPV) > 90% for cirrhosis[21].

Platelet count has been combined with age in the age-platelet index[23] or with AAR and prothrombin time in the cirrhosis discriminant score (CDS)[24] but the diagnostic accuracy of these composite scores was not superior to platelet count per se[21]. In addition, platelet count is a component of AST to platelet ratio index (APRI), model 3, Forns index, Fibrometer, and FibroIndex.

The APRI was described by Wai et al[25] from Anna Lok’s group at Ann Arbor University. It is calculated as

APRI = [(AST/Upper limit of normal)/platelet count (10⁹/L)] × 100

This test is derived from readily available laboratory parameters and usually requires a pocket calculator. Its diagnostic accuracy for both significant fibrosis and cirrhosis has been confirmed by several external studies[21,26-30]. Using the cut-offs proposed by Wai et al, approximately 50% of the patients can be correctly classified without a liver biopsy.

Lok et al[31] proposed another prognostic model for prediction/exclusion of cirrhosis in patients with chronic hepatitis C which is based on platelet count, AST, and prothrombin time expressed as international normalized ratio (INR). This index may be derived from the regression formula:

log odds = -5.56 - 0.0089 × platelet (× 10⁹/L) + 1.26 ×AST/ALT ratio + 5.27 × INR

or by a calculator available at the web site of the HALT-C trial (http://www.haltctrial.org). Using model 3 at a cut-off of < 0.20, cirrhosis could be excluded with an NPV of 99%.

Forns et al[32] developed an index derived from age, gamma-glutamyl transferase (GGT), cholesterol, and platelet count in a study of 476 untreated HCV patients, which is calculated as follows:

7.811 - 3.131 × ln (platelet count) + 0.781 × ln (GGT) + 3.467 × ln (age) - 0.014 × (cholesterol [mg/dL])

In their study, the area under the receiver operating characteristic curve (AUROC) for prediction of significant fibrosis (F2-F4 according to the Scheuer classification) was 0.86 in the test set and 0.81 in the validation set. The diagnostic accuracy of this index has been confirmed in patients with HIV-HCV coinfection[33].

French investigators analyzed an extensive array of biochemical tests in 339 patients with chronic hepatitis C and identified a panel of 5 markers which could best predict the stage of fibrosis: α2-macroglobulin, haptoglobin, apolipoprotein A1, GGT, and total bilirubin[34]. This test has been marketed as Fibrotest^TM (Biopredictive, Paris, France) in Europe and as Fibrosure^TM (LabCorp, Burlington, NC) in the United States. In contrast to the above mentioned indirect fibrosis tests, calculation of the Fibrotest by a patented algorithm is subject to payment of a fee to the manufacturer. The Fibrotest has been validated internally and externally in several studies in patients with chronic hepatitis C[35-38]. Interestingly, a recent study suggested that Fibrotest was a better predictor than histologic staging for complications of chronic hepatitis C[39]. Fibrotest was also found to predict fibrosis in alcoholic[40] and non-alcoholic[41] fatty liver disease. The diagnostic accuracy of this test is limited by hemolysis (leading to a reduction in haptoglobin), Gilbert’s syndrome (increasing the bilirubin level), and recent or ongoing infection (leading to elevations of α2-macroglobulin and haptoglobin).

Recently, Japanese investigators proposed another index composed of platelet count, AST, and gamma globulin:

1.738 - 0.064 (platelets [× 10⁴/mm³]) + 0.005 (AST [IU/L]) + 0.463 (gamma globulin [g/dL])

The AUROC for prediction of significant fibrosis was 0.83 (0.82 in the validation set) which was slightly superior to APRI or the Forns index assessed in the same study[29].

Hyaluronic acid (HA) is a polysaccharide present in ECM and elevated in serum in patients with hepatic fibrosis. Commercial test kits are available from Corgenix (Westminster, Colorado). The diagnostic accuracy of HA was found to be superior to that of procollagen type III N-terminal peptide (PIIINP)[42] and its diagnostic accuracy was confirmed in a large study of 486 HCV patients[43]. HA was also found to be an accurate marker of severe hepatic fibrosis in nonalcoholic fatty liver disease[44].

Procollagen peptides: PIIINP is a product of cleavage of procollagen and has been proposed as a serum marker of hepatic fibrosis more than two decades ago[45]. However, in several studies in patients with chronic HCV infection, this biomarker showed only moderate diagnostic accuracy[42,46,47].

Matrix metalloproteinases and inhibitors: Excess ECM proteins are degraded by matrix metalloproteinases (MMPs) which are in turn inhibited by tissue inhibitors of metalloproteinases (TIMPs). Since both MMPs and TIMPs are related to matrix protein turnover, their serum levels have been used as fibrosis markers. However, studies on the correlation of MMP-1, MMP-2, MMP-9, TIMP-1, and TIMP-2 with hepatic fibrosis have produced conflicting results[48-50]. TIMP-1 is a component of several composite fibrosis panels (see below).

YKL-40: YKL-40 is a glycoprotein believed to play a role in ECM degradation. In a study of 109 patients with chronic HCV infection, this biomarker showed similar diagnostic accuracy to HA for significant fibrosis, while its ability to diagnose cirrhosis was inferior to that of HA[51].

The Fibrospect II assay (Prometheus Laboratories Inc., San Diego, CA) uses HA, TIMP-1 and α2-macroglobulin. This test was found to accurately predict significant fibrosis in a study of 294 HCV patients and validated in an external cohort of 402 HCV patients[52]. This test was further validated in a study comparing the diagnostic accuracies of APRI, Forns index, Fibrotest, and Fibrometer[26].

In a European multicenter study of 1021 patients with chronic liver disease of different etiologies, an algorithm consisting of age, HA, PIIINP, and TIMP-1 was developed[53]. Using the Scheuer fibrosis score as a reference test, its overall diagnostic accuracy was similar to that of other noninvasive fibrosis tests (AUROC 0.78 for significant fibrosis, 0.89 for cirrhosis). Performance of the algorithm was slightly lower in a subgroup of patients with chronic hepatitis C (n = 325, AUROC 0.77 for prediction of F3-F4). This test is being marketed as Enhanced Liver Fibrosis (ELF^TM) test by Siemens Medical Solutions Diagnostics (Tarrytown, NY).

This model was developed by Australian investigators and is derived from bilirubin, GGT, HA, α2-macroglobulin, age and sex. High diagnostic performance was reported for both significant fibrosis and cirrhosis[54], but external validation yielded somewhat lower diagnostic accuracies[30].

The Fibrometer test incorporates α2-macroglobulin, HA, AST, platelet count, prothrombin index, urea, and age. Cales et al[26] reported superior diagnostic accuracy of Fibrometer to Forns index, Fibrotest, and APRI. However, this finding was not confirmed in an external validation study[30]. Several Fibrometers are now commercially available at BioLiveScale (Angers, France) for assessment of fibrosis in chronic viral hepatitis (Fibrometer V), alcoholic liver disease (Fibrometer A), and metabolic steatopathy (Fibrometer S).

Transient elastography by Fibroscan^TM (Echosens, Paris, France) is a promising technique that estimates the degree of hepatic fibrosis by measuring liver stiffness[55]. The Fibroscan transmits a vibration of low frequency into the liver from a probe that includes an ultrasonic transducer. The vibration waves induce an elastic shear wave whose velocity is proportional to the stiffness of the tissue. Shear wave velocity is measured by pulse-echo ultrasound and results are expressed in kPa. In cirrhotic patients, liver stiffness measurements (LSM) show a wide range from approximately 12 to 75 kPa. The Fibroscan samples a volume of approximately 4 cm³ which is considered more representative of the entire liver than a needle biopsy specimen. Measurements can be quickly performed in 5 min and are highly reproducible. However, steatosis may confound its value to estimate fibrosis. Furthermore, measurement is heavily limited or impossible in obesity, ascites or in patients with narrow intercostal spaces.

The Fibroscan has been studied in various liver diseases including chronic hepatitis C[56] and primary biliary cirrhosis[57] demonstrating good diagnostic accuracy. A small study in patients with chronic HCV infection and persistently normal transaminases even reported values of 100% for sensitivity, specificity, PPV and NPV, respectively[58]. In a study of 711 patients with chronic liver disease of various etiologies, LSM was closely related to fibrosis stage (r = 0.73) and high diagnostic accuracy was found for the diagnosis of cirrhosis (AUROC 0.96)[59]. Ganne-Carrie et al[60] further elaborated LSM for diagnosing cirrhosis and confirmed high diagnostic accuracy (AUROC 0.95) at a cut-off value of 14.6 kPa with some false-negative results due to inactive or macronodular cirrhosis. Another large study evaluated the success rate and performance of LSM in 935 patients with chronic HCV infection and reported successful measurements in 97% of the patients, while fatty thoracic belt was the major limiting factor[61]. Interestingly, recent data indicate a close correlation between LSM and portal pressure as estimated by the hepatic venous pressure gradient (HVPG) in patients with HCV-related cirrhosis[62] as well as recurring hepatitis C following liver transplantation[63]. Vizzutti et al[62] reported a good correlation between LSM and HVPG (r = 0.82) at lower portal pressures (< 12 mmHg) and suggested that, at a cut-off of 13.6 kPa, LSM could reliably predict or exclude clinically significant portal hypertension (CSPH, i.e. a portal pressure of ≥ 10 mmHg) which is the prerequisite for the formation of esophageal varices. Thus, LSM may be the method of choice to characterize the severity of cirrhosis in patients without CSPH while at the same time it could be used to identify patients with CSPH that may benefit from portal pressure measurement by hepatic vein catheterization. This approach should be prospectively tested in future studies.

Another attractive approach is the development of real-time ultrasound systems equipped with special elastography modules that allow estimation of liver fibrosis with conventional ultrasound probes during a routine sonography examination. First results obtained with the Hitachi EUB-8500 system were recently presented[64]. In 79 patients with chronic viral hepatitis, the AUROC for diagnosis of significant fibrosis was 0.75 (elasticity score) or 0.93 (combined elasticity-laboratory score).

Aguirre et al[65] from the University of California at San Diego proposed noninvasive diagnosis of liver fibrosis using double contrast material-enhanced MR imaging. In a retrospective study in 101 patients, this method could detect advanced fibrosis (METAVIR fibrosis score ≥ 3) with a sensitivity and specificity of > 90%.

Castera et al[66] investigated APRI, Fibrotest, Fibroscan and combinations thereof in 183 patients with chronic hepatitis C. For significant fibrosis (but not cirrhosis), the combination of Fibroscan and Fibrotest had superior diagnostic accuracy (AUROC 0.88) to that of respective individual tests.

The sequential use of several markers may overcome some of the limitations of individual markers. Sebastiani et al[67] developed three different sequential algorithms (including APRI, Fibrotest and/or Forns index) for the diagnosis of significant fibrosis with elevated ALT, significant fibrosis with persistently normal ALT, and cirrhosis, respectively, which allowed to classify 100% of the patients for each entity.