Systematic review: non-invasive methods of fibrosis analysis in chronic hepatitis C


Dr R. K. Sterling, Virginia Commonwealth University Health System, Section of Hepatology, 1200 E. Broad Street, West Hospital, Room 1492, Richmond, VA 23298-0341, USA.


Background  Accurate determination of the presence and degree of liver fibrosis is essential for prognosis and for planning treatment of patients with chronic hepatitis C virus (HCV). Non-invasive methods of assessing fibrosis have been developed to reduce the need for biopsy.

Aim  To perform a review of these non-invasive measures and their ability to replace biopsy for assessing hepatic fibrosis in patients with chronic HCV.

Methods  A systematic review of PUBMED and EMBASE was performed through 2008 using the following search terms: HCV, liver, elastography, hepatitis, Fibroscan, SPECT, noninvasive liver fibrosis, ultrasonography, Doppler, MRI, Fibrotest, Fibrosure, Actitest, APRI, Forns and breath tests, alone or in combination.

Results  We identified 151 studies: 87 using biochemical, 57 imaging and seven breath tests either alone or in combination.

Conclusions  Great strides are being made in the development of accurate non-invasive methods for determination of fibrosis. Although no single non-invasive test or model developed to date can match that information obtained from actual histology (i.e. inflammation, fibrosis, steatosis), combinations of two modalities of non-invasive methods can reliably differentiate between minimal and significant fibrosis, and thereby avoid liver biopsy in a significant percentage of patients.


Accurate determination of the presence and degree of liver fibrosis is essential for predicting prognosis and for planning treatment of patients with chronic hepatitis C virus (HCV) infection.1 Percutaneous liver biopsy is considered the gold standard for assessing hepatic fibrosis.2 This method, however, is invasive and carries a significant rate of complications, ranging from 1% to 5%, and risk of mortality ranging from 1 in 1000 to 1 in 10 000.3–5 In addition, biopsy is subject to sampling error and inter-observer variability.6, 7 In the last 10 years, non-invasive methods of assessing liver fibrosis utilizing laboratory methods and imaging (Table 1) have been developed to reduce the need for biopsy. Because of confusing literature, we performed a systematic review of these non-invasive measures and their ability to replace biopsy for assessing hepatic fibrosis and inflammation in patients with chronic HCV in an attempt to put these studies into perspective.

Table 1.   Categories of variables used to assess hepatic fibrosis
  1. CTP, Child Turcotte Pugh; MELD, Model for End Stage Liver Disease; APRI, AST to platelet ratio index.

DemographicsBiochemical methodsImagingModelsBreath tests
Spider Nevi
Duration of haemodialysis
Body Mass Index
Time since liver transplantation
CTP score
MELD score
Alanine aminotransferase (ALT)
Aspartate aminotransferase (AST)
Platelet count
Albumin/Total Protein ratio
Gamma glutamyl transferase (GGT)
Transferrin Saturation
Mean corpuscular volume (MCV)
Prothrombin time
Serum immunoglobulins
Serum Insulin-like Growth Factor-1
Alpha-2 Macroglobulin
Apolipoprotein A1
Procollagen III peptide (PIIIP)
Hyaluronic Acid (HA)
Collagen types 1-IV
7S domain of type IV collagen (7S-IV)
Artificial Neural Networks (ANNs)
Matrix metalloproteinases (MMP-1, 2, and 9)
Tissue inhibitors of metalloproteinases
(TIMP-1 and TIMP-2)
MEGX (Monoethylglycinexylidide)
Serum apoptotic caspase activity
YKL-40(human cartilage glycoprotein-39)
Glycosylation of AGP (alpha-1-acid glycoprotein)
Plasma amino acids
Highly sensitive C-reactive protein(hs-CRP)
Proteomic markers
Ultrasonography (US)
US elastography
Magnetic resonance (MR) elastography
MR spectroscopy
Single photon emission computed tomography (SPECT)
Tissue strain imaging (TSI)
Fibrospect II
Hospital Gregorio
Maranon (HGM-1) index
Methacetin Breath Test (MBT)
Aminopyrine Breath Test (ABT)


A systematic review of PUBMED and EMBASE was performed for all articles published through 1 December 2008. The following search terms were used: elastography AND liver, elastography AND hepatitis, elastometry AND liver, elastometry AND hepatitis, Fibroscan, SPECT AND hepatitis, noninvasive liver fibrosis AND ultrasonography, noninvasive liver fibrosis AND ultrasound, noninvasive liver fibrosis AND Doppler, noninvasive liver fibrosis AND MRI, Fibrotest, Fibrosure (US patented name for Fibrotest), Actitest, APRI AND liver, Forns AND fibrosis, noninvasive AND liver AND fibrosis, and breath tests AND liver AND hepatitis C. The search was limited to adults and those in the English language and included all primary studies involving chronic HCV patients, HCV and human immunodeficiency virus (HIV) coinfection, and all primary studies which included patients with HCV in comparison with other aetiologies of chronic liver disease. This resulted in 401 studies as detailed in Figure 1. We excluded case studies, review articles and primary studies involving patients with liver diseases other than HCV [e.g. hepatitis B, alcoholic hepatitis, non-alcoholic fatty liver disease (NAFLD)]. This resulted in the subsequent exclusion of 250 studies with the remaining studies being divided into those that used serum tests, breath tests, and those that used hepatic imaging, including elastography, as well as studies that compared these various modalities.

Figure 1.

 Selection process for study inclusion in the systematic review of non-invasive assessment of liver fibrosis in patients with chronic hepatitis C.


Serum tests

Our search identified 87 studies utilizing serum tests in models to predict hepatic fibrosis. These models include both direct and indirect markers of liver fibrosis and include serum chemistries, HCV RNA, measures of hepatic metabolic activity, partial liver functions such as mitochondrial, microsomal, or cytosolic function, as well as extracellular matrix remodelling proteins or glycosaminoglycans, products of collagen synthesis or degradation, or enzymes involved in matrix synthesis or degradation.8 While several non-invasive models utilize tests that are not routinely available and are at additional costs, others incorporate routine clinical and laboratory data.

Models with routine tests

In an effort to help clinicians predict the probability of cirrhosis in patients with chronic HCV, Kaul developed a model that utilized available clinical and laboratory information with an AUROC of 0.938.9 Logistic regression identified platelet count < or = 140 000/mm,3 spider nevi, aspartate aminotransferase (AST) > 40 IU/L and male gender as independent predictors of cirrhosis. Male and female patients with normal platelet count and AST and no spider nevi had low (<2%) probability of cirrhosis. Male patients with abnormal values on all three other predictors had a 99.8% probability of cirrhosis. Other simple models have included AST/alanine aminotransferase (ALT) ratio and platelets, which correlated significantly with disease stage, except for patients with a history of alcohol abuse.10, 11 AST/ALT ratio had a sensitivity of 81% and a specificity of 55% in identifying cirrhotic patients who died within 1 year.12 However, several studies with these markers found poor correlation between serum markers and histological activity.13, 14 Benlloch also evaluated an index of prothrombin time (PT), albumin/total protein, AST and time since liver transplantation in a group of HCV-infected liver transplant recipients to differentiate mild from significant fibrosis [sensitivity 87%, specificity 71%].15 Boyacioglu found no correlation with fibrosis stage when evaluating age, haemodialysis duration, BMI, HCV RNA, ferritin, and ALT in a cohort of haemodialysis patients with HCV.13 Fibroindex was developed by Koda et al. and utilizes platelet count, AST and serum IgG, with AUROC for evaluation of significant fibrosis (F2-F3) and severe fibrosis (F3-F4) [0.826 and 0.848, respectively], comparable with that of both Forns [0.864 and 0.831, respectively] and AST to Platelet Ratio index [0.778 and 0.810, respectively].16

The AST to platelet ratio index (APRI) is a formula that utilizes measurements of serum AST concentration and platelet count. Its value is determined by the formula AST/(upper limit of normal)/platelet count (109/L) × 100.17 APRI is simpler to use than most of the other indices with performance similar to that of the Fibrotest (FT) and the Forns index. APRI was accurate in estimating fibrosis in patients with HCV [AUROC 0.87–0.89, sensitivity 94–100%, specificity 95–100%]18–20 and with HCV/HIV coinfection,21–23 although some studies have reported that it cannot replace liver biopsy in the accurate staging of fibrosis in patients with hepatitis C, as one study noted its inability to classify correctly 40–65% of patients with chronic HCV or HBeAg negative chronic hepatitis B.24

The Forns index uses four common clinical measurements: patient age, serum concentrations of total cholesterol and gamma-glutamyl transpeptidase (GGT) and platelet count. This method can be used to differentiate patients with mild (F0-F1) fibrosis from those with severe (F2-F4) fibrosis [AUROC 0.81], but it is less accurate in distinguishing among patients with grades F2-F4. Similar to FT, they felt that half of HCV patients without significant liver fibrosis could be identified with high accuracy using this index. Forns should not be used in patients with genotype 3; however, due to varying cholesterol levels.25 The Forns index has been validated in other cohorts [PPV 94% for significant fibrosis and NPV 100% for cirrhosis]26 and as a predictive tool for response to anti-HCV therapy.27

Originally developed for use in HIV-HCV coinfection, FIB-4 also utilizes routine laboratory tests to predict liver fibrosis. Based on multivariate logistic regression analysis, a simple index was developed: age ([yr] × AST [U/L])/((PLT [109/L]) × (ALT [U/L] × 1/2). Use of this index would correctly classify 87% of patients with FIB-4 values outside 1.45–3.25 and avoid biopsy in 71% of the validation set with AUROC 0.765, sensitivity 70%, specificity 97% for differentiating Ishak 0–3 from 4–6.28 This model was subsequently validated by Vallet-Pichard in a large cohort of HCV mono-infected patients, with the finding that using these ranges, 78% of 847 biopsies were correctly classified [AUROC 0.85 for severe fibrosis and 0.91 for cirrhosis].29

Other models have also been developed to predict fibrosis in HCV-HIV co-infected patients, such as the Hospital Gregorio Maranon (HGM-1) index,30 and the diagnostic performance of these, as well as previously mentioned models, has been compared. FIB-4, APRI, Forns, and platelet count did not differ significantly for fibrosis and cirrhosis, but could save liver biopsies in up to 56–76% of cases.31 Fibrometer, Hepascore and Fibrotest have also been shown to outperform SHASTA [an index of Serum Hyaluronic acid (HA), albumin, and AST], APRI, Forns and FIB-4 in other studies.32

Models with non-routine tests

Several non-invasive models include nonroutine measurements of extracellular matrix remodelling markers, such as amino-terminal propeptide of type III collagen (PIIIP), matrix metalloproteinase (MMP), tissue inhibitor of matrix metalloproteinase (TIMP), Hyaluronic acid (HA), and Type IV collagen (CL-4) alone or in combination with serum chemistries and HCV RNA. PIIIP was analysed in several studies, both alone and in combination with other potential markers of fibrosis. Mean PIIIP level was reflective of histological severity33 as well as an accurate marker of disease evolution.34 Trocme also found that an index of PIIIP/MMP-1 was significantly correlated (R2 = correlation coefficient) with fibrosis score [R2: F1 0.62, F2 0.61, F3 0.79, F4 0.96] and was better than HA and TIMP-1 [AUROC 0.77 for F1 v F2-F4 and 0.81 for F1–2 v F3-F4].35 Different PIIIP assays, especially when used in combination, have been shown to be significantly related to fibrosis, periportal necrosis and hepatic activity index (HAI).36 TIMP-1 and TIMP-2 have also been compared with MMP-2, with TIMP-1 and TIMP-2 yielding an AUROC of 0.73 for fibrosis determination.37 TIMP-1 and HA have also been shown to predict fibrosis accurately in HIV-HCV co-infected patients.38 Collagen types have been used to develop a Fibrosis Discriminant Score (FDS), with measurements of types III and IV plus serum ALT and albumin.39 However, Verbaan evaluated IgG, PIIIP, and CL-IV and found that the three markers correlated weakly with histological grade and stage of liver fibrosis and demonstrated a great degree of overlap.40

Saitou found that an increase in serum YKL-40, also known as human cartilage glycoprotein-39, PIIIP, and especially HA, were correlated with fibrosis progression, but HA was best for differentiating F4 from F0–3 [AUROC 0.854] and YKL-40 was best for differentiating F2–4 from F0–1 [AUROC 0.809].41 A model based on AST, platelet count, and HA has been shown to have better accuracy than YKL-40 and HA when used solely for the prediction of fibrosis in end stage renal disease patients with HCV.42 HA has also been shown to have higher diagnostic accuracy than PIIIP as a marker of fibrosis [AUROC 0.864 v. 0.691, respectively, for differentiation of extensive from mild fibrosis].43 HA concentrations have been shown to be higher in cirrhotic and fibrotic patients [sensitivity 88%, specificity 78% for cirrhosis]44 and moderately correlated with fibrosis scores,8, 45 but has also been shown to have a low correlation coefficient (0.45) when evaluated pre-treatment in a small cohort of patients.46 Yilmaz evaluated HA and high-sensitive-C-reactive protein (hsCRP) as a marker of necroinflammation and found that HA was accurate in predicting fibrosis score ≥1 and in predicting cirrhosis [AUROC 0.86], but cutoff values for hsCRP for predicting histological activity index (HAI) warranted further evaluation.47 HA, N-terminal procollagen III (PIIINP), laminin (LN), 7S subunit of Type IV collagen (7S-IV), TIMP, PGA (PT, GGT, Apolipoprotein A1), PGAA (PGA + delta2-Macroglobulin) and combinations of these have been evaluated with a high specificity for fibrosis (94%) and cirrhosis (90%) but with lower accuracy for fibrosis (43%) than cirrhosis (85%).48

In 2001, Imbert-Bismut reported on the Fibrotest, a composite of measurements of serum proteins, alpha-2 globulin, alpha-2 macroglobulin (A2MG), gamma globulin, apolipoprotein A, GGT, and total bilirubin, for the diagnosis of cirrhosis or F3/F4 fibrosis, with AUROC 0.923, PPV 85% and NPV 90%.49 It was felt that the number of biopsies could be reduced by up to 46% in chronic HCV patients using this test. FT has been utilized in several other reports with varying success,50–53 but has been shown to have significantly reduced performance in patients with normal ALT [Accuracy 73.5%].54 Because the model includes several non-routine tests, it is not readily available and requires additional costs. Also, because haptoglobin and bilirubin can be abnormal in the absence of liver disease, its applicability is limited. In an effort to also assess necroinflammatory activity, Actitest, a modified FT, also includes ALT measurement. Poynard assessed the discordance between biopsy and markers with discordance observed in 29%, 16% for fibrosis staging and 17% for activity grading. This was attributable to failure of markers in 2.4% and biopsy failure in 18%.55 Fibrotest has also been shown to be a better predictor than biopsy staging for HCV complications.56 Myers developed a similar index using age, gender and five biochemical markers –A2M, apolipoprotein A1 (ApoA1), haptoglobin, total bilirubin, and GGT, which was evaluated in both HCV and HCV/HIV co-infected patients. This index reduced the necessity for biopsy by 55%.57–59 FIBROSpect II (FS) is another diagnostic panel of the extracellular matrix remodelling markers HA, TIMP1, and A2M. FS has been shown to correctly identify F0-F1 fibrosis in 90% of study cases and 80% of F2-F4, but was not as accurate for intermediate stages of fibrosis (F1-F3). It was concluded that the high NPV of this assay in low F2-F4 populations may help avoid liver biopsy.60 Other studies have also shown clinical utility for FS in ruling out advanced fibrosis in hepatitis C by identifying patients with mild disease in whom treatment could be deferred.61, 62 Patel et al. found that FS differentiated mild from moderate-severe fibrosis in patients with chronic HCV and in a post-transplant cohort, but noted that assessing the utility of non-invasive biomarkers was limited by method variability and poor inter-observer agreement for histological staging.62 Several studies have shown that a better approach may be stepwise or combined use of these tests and combining APRI and FT or Hepascore increases diagnostic accuracy.63, 64

Other serum measures including caspase activity, serum globulins, plasma amino acids and insulin-like growth factor (IGF-1) have been developed. Proapoptotic activation of caspases is considerably enhanced in histological sections from HCV-infected liver tissue, suggesting an important role of apoptosis in liver damage. Caspase activation has also been studied as a marker of tissue damage in sera from patients with chronic HCV infection.65, 66 Zhang analysed amino acids with promising results for the prediction of cirrhosis using the optimal index of (Phe)/(Val) + (Thr + Met + Orn)/(Pro + Gly).67 Maruyama found that serum gamma (y)-globulin and immunoglobulin (IgG) were well correlated with HAI score (both; P < 0.0001), grading score (both P < 0.01) and staging score (both P < 0.0001).68 Lorenzo-Zuniga found that mean IGF-I values were significantly lower in patients with advanced fibrosis (F4, 65.9 ± 17.9 ng/mL) vs. F0, 145.2 ± 47.1; F1-F2, 150.3 ± 89.6; and F3, 121.4 ± 35.2 ng/mL; P < .05).69

Artificial neural networks (ANNs), built on clinical variables and patient data sets, have been developed to predict significant fibrosis in patients with chronic hepatitis C. Haydon developed a model of fifteen routine clinical and virological factors, which were collated from 112 HCV patients, with sensitivity, specificity, NPV and PPV values all greater than 92% for Ward-type ANNs for prediction of cirrhosis.70 Piscaglia also analysed ANNs in post-liver transplant patients with 100% sensitivity and NPV71 and Cucchetti found that ANNs measured the mortality risk of patients with cirrhosis more accurately than the model for end-stage liver disease (MELD) score.72

More recently, Gangadharan evaluated various proteomic markers.73 He found that inter-alpha-trypsin inhibitor heavy chain H4 (ITIH4) fragments, alpha1 antichymotrypsin, apolipoprotein L1(ApoL1), prealbumin, albumin, paraoxonase/arylesterase 1, and zinc-alpha2-glycoprotein were decreased in cirrhotic serum, whereas CD5 antigen-like protein (CD5L) and beta2 glycoprotein I (beta2GPI) were increased. A2M and immunoglobulin components increased with hepatic fibrosis, whereas haptoglobin and complement components (C3, C4, and factor H-related protein 1) decreased. In the investigation of another novel biomarker, Mooney investigated the glycosylation of alpha-1-acid glycoprotein (AGP), which was higher in patients with significant fibrosis/cirrhosis.74 It has been theorized that the changes in glycosylation in the setting of liver disease may be reflective of disease severity, and thereby act as a non-invasive marker of fibrosis, and a glycoprotein, galactose-deficient anti-Gal immunoglobulin G, has recently been identified that is altered both in amount and in glycosylation as a function of liver fibrosis and cirrhosis.75 Suzman et al. have also developed models using DNA microarray analysis in patients with HIV/HCV coinfection with accurate prediction of fibrosis and elimination of need for biopsy in 83% of patients.76 Although all these models can predict significant fibrosis with reasonable accuracy, their inclusions of non-routine tests limit their utility in clinical practice.

Imaging tests

Our search identified 57 studies utilizing radiological tests in models to predict hepatic fibrosis. They can be divided into ultrasound, magnetic resonance imaging, elastography, single photon emission computed tomography and tissue strain imaging.

Ultrasonography.  Ultrasound has been used to non-invasively assess the degree of liver fibrosis in patients with chronic HCV. Results vary with some studies showing an association between ultrasonography score and detection of cirrhosis with sensitivities ranging from 87.5% to 100% and specificities ranging from 81.5% to 93.5%.77–79 Vyas found that portal venous blood flow (PVBF), portal flow velocity (PFV) and gastric mucosal blood flow (GMBF) were all significantly slower in cirrhotic patients and PVBF and PFV were lower in Child’s class B/C than class A.80 Weickert et al. found statistically significant differences in all sonographic parameters between patients with and without cirrhosis, but sensitivity and specificity were significantly increased when assessment of the transmission of heart pulsations on the liver surface was included as part of the ultrasonographic examination [sensitivity 85% vs. 55% and specificity 93% v 86% respectively].81 Schneider et al. found that Doppler ultrasound alone was unable to discriminate between degrees of fibrosis, but portal venous undulations could predict liver cirrhosis with increased sensitivity (76.5%).82 Yamada et al. also used ultrasound to assess liver fibrosis by calculating a fibrosis extraction ratio (FER) (fibre volume/total volume), which was able to distinguish F0/F1 from ≥ F2 fibrosis with sensitivity of 55% in the HCV cohort.83 Hirata et al. derived an arterio-portal (A/P) ratio by evaluating hepatic hemodynamics, which was higher in patients with cirrhosis compared with controls and demonstrated statistically significant differences in the A/P ratio when comparing severe-to-mild or moderate fibrosis [R2 0.734–0.816].84 Lim et al. investigated the hepatic vein transit times (HVTT) for grading liver disease using an ultrasound microbubble contrast agent as a tracer. This study also utilized Doppler sonography to determine several indices to calculate portal vein congestive index, but found that there were no significant differences with increasing severity of liver disease and that these indices were difficult to reproduce reliably.79, 85 Abbattista, however, found that HVTT was significantly shorter in cirrhotic patients than in non-cirrhotic patients (P < 0.001) and distinguished between these patients with high accuracy.86 These results show that unenhanced Doppler ultrasound is not reliable in the discrimination of varying degrees of fibrosis, but that results can be improved with additional measurements such as heart pulsation at the liver surface and portal venous flow measurements.

Elastography.  Elastography or elastometry is another non-invasive method of measuring the mean stiffness of hepatic tissue with hepatic rigidity considered a marker of progressive fibrosis. Sandrin et al. evaluated this method by obtaining in vivo liver elasticity measurements using the shear elasticity probe, a device based on one-dimensional (1-D) transient elastography, which presents several advantages in that the transmitted elastic wave can be temporally separated from reflected elastic waves. Thus, the technique is less sensitive to boundary conditions than other elastographic techniques. The acquisition time is also short (typically less than 100 ms), which enables measurements to be made on moving organs. This makes transient elastography (TE) well adapted to the study of the liver. A probe (Fibroscan) with an ultrasonic transducer is used to transmit low frequency (50 MHz), low amplitude vibrations into the liver; these vibrations produce elastic shear waves that propagate throughout the liver. The probe also emits a pulse-echo ultrasound wave, which is used to determine the velocity of the shear wave. The velocity of the shear wave is directly related to liver stiffness.87

Comparisons of TE with biopsy results have shown that cutoff values can be established to distinguish mild/moderate fibrosis from severe fibrosis/cirrhosis, with validation studies showing variable results and with greatest statistical significance being demonstrated in the differentiation of cirrhosis from mild fibrosis [AUROC F = 4 (0.94), sensitivity ≥ 2 (85%), specificity F ≥ 2 (91%)].88, 89 Studies have utilized varying optimal stiffness cutoff values, making comparison between studies challenging. Overall, advanced fibrosis is more likely with a higher cutoff.89–93 Ganne-Carrie found an optimal cutoff value of 14.6 kPa for detection of cirrhosis, but a cutoff of 10.0 kPa and 14.1 kPa was adequate to achieve 95% sensitivity and specificity in their HCV patients with cirrhosis.94 Vergara found that the performance of TE was low for discriminating mild from significant liver fibrosis95 and Friedrich-Rust found that Spearman’s correlation coefficient between the elasticity scores using real-time elastography and histological fibrosis stage was low at 0.48.96 Arena et al. also found that TE was more suitable for identification of advanced fibrosis and that necroinflammatory activity influences TE measurements in patients without cirrhosis97 and Wong also found that TE might overestimate liver fibrosis when ALT is elevated.98 However, Chang et al. found good correlation between TE and fibrosis, but in an Asian cohort with only 8% of patients having HCV.99 In addition to fibrosis stage, Ogawa found that Fibroscan values correlated with serum levels of hyaluronic acid (HA) and type IV collagen.92 TE has also been used in assessing fibrosis in patients co-infected with HIV and HCV100 and in liver transplant recipients with recurrent HCV.101 Lastly, Foucher also used TE to determine cutoff values for the presence of Grade II-III oesophageal varices (27.5 kPa), Child-Pugh class B or C cirrhosis (37.5 kPa), history of ascites (49.1 kPa), and HCC (53.7 kPa) with an NPV >90%.102 Liver stiffness has also been shown to correlate with steatosis, necroinflammatory activity and hepatic iron deposition, as well as fibrosis.103 TE is limited, however, by its inability to be performed on patients with ascites (even if clinically undetectable), as well as on patients with narrow intercostal spaces or morbid obesity.

Magnetic resonance imaging.  Magnetic resonance (MR) imaging has also been utilized to evaluate liver fibrosis in patients with chronic HCV. Several types of enhanced MR imaging have been developed to evaluate the degree of liver fibrosis. Magnetic resonance elastography (MRE) demonstrated that patients with hepatic fibrosis have higher liver stiffness measurements than healthy volunteers104 and that those with mild fibrosis were able to be differentiated from those with moderate or advanced fibrosis, with mean hepatic shear elasticity being 2.24 ± 0.23 kPa in patients with F0-F1 fibrosis, 2.56 ± 0.24 kPa with F2-F3 fibrosis and 4.68 ± 1.61 kPa in patients with F4 fibrosis.105 In recent studies, MRE was found to have a higher technical success rate than ultrasound elastography and a better diagnostic accuracy than ultrasound elastography and APRI for the staging of liver fibrosis.106, 107 Aguirre et al. evaluated the accuracy of spoiled gradient echo (SPGE) and superparamagnetic iron oxide (SPIO) -enhanced MR in depicting hepatic fibrosis. Although diagnostic performance depended on the sequence and scoring system, sensitivity and specificity values >90% could be obtained with SPGE double-enhanced MRI.108 Lucidarme et al. also evaluated the ability of SPIO-enhanced MR to detect diffuse liver fibrosis by histopathological comparison in a cohort including HCV patients. When only the reticulation pattern was considered, SPIO-enhanced MR could detect direct signs of macroscopic fibrosis with a sensitivity of 61%. Their study was limited, however, in that their cohort consisted mainly of cirrhotic patients. Hence, no attempt was made to compare patients with stages F2, F3 and F4 fibrosis.109

Lim et al. investigated whether phosphorus 31 (31-P) MR spectroscopy could be used to assess the severity of HCV liver disease by measuring phosphomonoester (PME) and phosphodiester (PDE) ratio. This ratio has been viewed as an index of cell membrane turnover, therefore providing an indirect measure of severity of liver disease. They found overlap of the moderate hepatitis group with the mild hepatitis and cirrhosis groups, but there was a clear difference seen between the mild hepatitis and cirrhosis groups when using a PME/PDE ratio ≤0.2 to denote mild hepatitis and a ratio of ≥0.3 to denote cirrhosis.110

Other imaging methods.  Several additional imaging methods have been used to evaluate non-invasively liver fibrosis in HCV-infected patients. Single photon emission computed tomography (SPECT) was tested in HIV-HCV co-infected patients in an effort to correlate histological severity of liver fibrosis with SPECT results. A number of SPECT parameters were associated with histological changes, fibrosis, and cirrhosis. The minimum pixel count for spleen region of interest and maximum pixel count for right hepatic lobe correctly correlated 39 of 46 SPECT scans with biopsy results. Larger studies are needed, however, to validate these results.111 Groshar et al. also utilized SPECT with 99mTc-phytate colloid in patients with cirrhosis, 31 of whom had chronic HCV. This type of scan shows the distribution of the perfused Kuppfer cell mass, which are as equally affected by the fibrotic process as hepatocytes, and is proportional to the perfused hepatocyte mass. Cirrhotic patients showed a significant decrease in total liver uptake and a significant increase in total spleen uptake. Spleen volume was best at detecting liver cirrhosis, but total liver uptake correlated better with chronic liver disease severity. Total hepatic uptake represented the perfused hepatic mass and correlated inversely with clinical parameters that reflect the severity of liver fibrosis.112

Finally, tissue strain imaging (TSI), a myocardial examination technique, was also evaluated in a cohort of normal adults, patients with chronic hepatitis and patients with cirrhosis to determine its potential for assessing hepatic fibrosis. Liver biopsy was only obtained, however, in 7 of the 47 patients in the cohort. Strain values were calculated by dividing the change in length of the examined organ before and after motion by the length of the organ before motion. They did find statistically significant differences in the strain values between the three groups. One advantage of this method is that it can also be performed on patients with ascites, but one disadvantage is that strain imaging can currently only be performed using a cardiac transducer.113

Breath tests

There are several 13C breath tests available for the non-invasive determination of hepatocellular function and we identified seven studies meeting our criteria for evaluation. 13C-methacetin breath test (MBT) is essentially a microsomal liver function test in that 13C-methacetin is rapidly metabolized by healthy liver cells into acetaminophen and 13CO2 by a single dealkylation, and the increase of 13CO2 in breath samples can be quantified by isotope ratio mass spectrometry or nondispersive isotope-selective infrared spectroscopy. MBT has been shown to have high sensitivity (93.5–95%) and specificity (95–96.7%) in identifying cirrhotic patients, but patients with early fibrosis did not differ in change over baseline values from patients with advanced fibrosis as determined by liver biopsy.114, 115 Goetze et al. also evaluated MBT, but they evaluated the agreement of MBT results obtained by online molecular correlation spectroscopy as well as by mass spectrometry in patients with HCV. They found that in different degrees of liver fibrosis, there was an excellent linear relationship between breath test and isotope ratio mass spectrometry (IRMS) without evidence of the measurements diverting from the line of equality.116 MBT has also been shown to determine accurately fibrosis is patients with normal ALT117 and was more reliable than APRI or AST/ALT ratio.118 A shortened MBT (15 min) has also been shown to have a high sensitivity (93%) and specificity (94%).119

Another breath test is the 13C-aminopyrine breath test (ABT), where the aminopyrine is another substrate for labelling with the 13C isotope, much like the MBT. Schneider et al. found that the cumulative 13C recovery differed significantly between patients without relevant fibrosis (F0-F2) and cirrhosis (F5-F6), with sensitivity 73–82% and specificity 63–68%. The ABT did not differentiate patients with F3-F4 scores from the remaining two patient groups.120

Comparative studies.  Many studies have compared the various imaging and biochemical methods described above for the discrimination of degrees of liver fibrosis. Numerous methods have been compared with varying results as detailed in Table 2. We will summarize several of these here.

Table 2.   Comparative Studies
Author, yearNPatient populationTestComments
  1. ELFG algorithm, Age, hyaluronic acid, procollagen type III, and TIMP-1; GUCI, Goteborg University Cirrhosis Index - normalized AST x prothrombin-INR x 100/platelet count (× 10(9)/L); CDS – Cirrhosis Discriminant Score, which is derived from platelet count, ALT/AST ratio and PT; Pohl: includes AST, ALT, and platelet count; SAPI, splenic arterial pulsatility index; SELDI-TOF-MS, surface-enhanced laser desorption/ionization time-of-flight mass spectrometry; Bonacini’s discriminant score, derived from platelet count, ALT/AST ratio and INR; IR, Insulin resistance; Sydney, uses insulin resistance together with age, AST, cholesterol, and alcohol use; Fibrosure (US patented name for Fibrotest); FT-AT, Fibrotest-Actitest; FA by DIA, fibrotic area (FA) as calculated by a digital image analysis (DIA); LCAT, lecithin-cholesterol acyltransferase; Fibrometer, combines HA, PT, platelets, AST, α2 macroglobulin, urea and age; Hepascore, combines bilirubin, GGT, HA, α2 macroglobulin, age and gender.

Oberti, 1997130243Chronic liver disease63 clinical, biochemical, US, and endoscopic variablesHyaluronate and prothrombin index were the best predictive factors. Hyaluronate concentration of ≥60 μg/mL had a sensitivity of 97% and a specificity of 73% for the diagnosis of cirrhosis
Giannini, 2003140252HCVAST/ALT, MEGX (monoethylglycylxylidide), Platelet count, Prothrombin Time (PT)Combination of AST/ALT and/or PLT for prediction of cirrhosis had a PPV of 97% and an NPV 98%
Aube, 2004131106HBV, HCV, EtOH32 clinical, biochemical and US variablesDoppler US superior to clinical and biochemical methods for accurate diagnosis of cirrhosis, less accurate for severe fibrosis.
Saito, 200413375HCVElastometry, platelet countsBoth elastometry and platelet counts were correlated with fibrosis stage, but the deviation was smaller in the former
Castera, 2005134183HCVElastography, Fibrotest, APRIAUROC for FibroScan, FibroTest, and APRI were of the same order (.83, .85, and .78, respectively, for F ≥ 2; and .95, .87, and .83, respectively, for F = 4). The best performance was obtained by combining FibroScan and FibroTest, with AUROC .88 for F ≥ 2, and .95 for F = 4.
Colletta, 200514140HCVFibroscan v. FibrotestFor F2 or greater fibrosis: Fibroscan – Sens 100%, Spec 100%, NPV 100%, PPV 100% Fibrotest- Sens 64%, Spec 31%, PPV 33%, NPV 62%
Yan, 200514253 + 25 controlsPts with fibrosis (aetiology not specified)US, CTP, MELDUltrasonography compared to both CTP and MELD. Correlation coefficient 0.784, and 0.768, respectively.
Iacobellis, 20051431143HCVPlatelet counts <140 000/μL, US parameters including nodular liver surface, spleen and portal vein sizeAll indices had a specificity rate ≥90% in excluding bridging fibrosis/cirrhosis. None of the ultrasonographic parameters reached an acceptable sensitivity rate.
Cales, 2005144383Viral hepatitis51 blood markers, FT, Fibrospect, ELFG (European Liver Fibrosis Group Algorithm), APRI, FornsAUROC was 0.81 for FT, 0.883 for combination of platelets, prothrombin index, AST, A2M, hyaluronate, urea and age, 0.82 for Forns, 0.794 for APRI, and 0.892 for Fibrometer
Nunes, 200514540
INR, platelet, AST/ALT, APRI, Forns, procoll III, hyaluronic acid, YKL-40Diagnostic performance of non-invasive markers similar in mono and co-infected patients
Islam, 2005146179HCVVarious biochemical markers and indicesGUCI cutoff of 1.0 had sensitivity 80%, specificity 78%, NPV 97%, PPV 31% for diagnosis of cirrhosis.
Lackner, 2005121194HCVAST/ALT, CDS, AP, Pohl, APRI, platelet countCDS, AP, APRI and platelet count had similar AUROCs for significant fibrosis (0.71–0.80) and cirrhosis (0.89–0.91).
Kelleher, 2005122137HCV/HIVALT, AST, APRI, albumin, bilirubin, HA, YKL-40HA, albumin and AST accurately staged mild and advanced fibrosis
Borroni, 2006125228HCVAge-platelet index, cirrhosis discriminant score, APRI, Pohl, AST/ALTSpecificities of 5 measures 87–100% but sensitivities 17–67%.
Liu, 200614779HCV w/normal ALTUS, APRI, API, AST/ALTSAPI most discriminatory among Doppler indices and superior to APRI, API, AST/ALT for predicting significant fibrosis. SAPI set at 0.85 and 1.10 had sensitivity of 97% and 67%, specificity of 45% and 96%, PPV of 41% and 87% and NPV of 97% and 88%
Parise, 2006148206HCVHA, AST/ALT, APRI, GGTHA level showed best correlation with disease stage. HA and APRI showed best diagnostic accuracy
Gobel, 2006149117HCVSELDI-TOF-MS for multiple markersPanels correctly identified HCC and cirrhosis with high sensitivity and specificity
Bourliere, 2006126235HCVFT, APRI, FORNSThe AUROCs of these 3 methods were similar. 3 methods w/o biopsy allowed fibrosis to be well evaluated in 82% of pts.
Fabris, 200613340HCV with normal ALT (30 non-drinkers)AST/ALT, age-platelet, APRI, Forns, Bonacini’s discriminant scoreAmong non-drinkers, APRI >0.4 had a 100% sensitivity in identifying subjects with significant fibrosis and APRI ≤ 0.4 had a 100% NPV in excluding significant fibrosis.
Schneider, 200612083HCVABT, US, APRIFor ABT, sensitivity 73–83% and specificity 63–68% depending on sampling time. For APRI and US, sensitivity 77% and 88%; specificity 63% and 68%,
Romera, 2006124131HCVAST, ALT, GGT, platelet, cholesterol, IR, Sydney, Forns, APRIAUROC for absence of fibrosis 0.80 for Sydney, 0.71 for Forns, 0.71 for APRI; for advanced fibrosis, 0.88, 0.83, 0.82; NPV 74%, 02%, 67%.
Testa, 200613275HCVBiochemical markers, ABT, USVarious combinations effective in identifying or ruling out fibrosis in 75–80% of patients (BMI, APRI, PLT/Spleen Diameter ratio (SPD) and (APRI, ABT, PLT/SPD).
Wilson, 2006150210HCVFibrosure, APRI, ALTAll 3 had predictive value >95% for insignificant fibrosis and specificities of 89%, 73%, 73%.
Sebastiani, 2006127190CHCAPRI, Forns, FibrotestStepwise combination of these methods (3 algorithms) improves diagnostic performance and reduces need for liver biopsy 50–70%
Sene, 2006151138CHCFT-AT, Forns, APRI, age-platelet, platelet, hyaluronic acidAUC of FT-AT = 0.83. Discordance between FT-AT and Metavir in 29% of patients associated with haemolysis and male.
Maor, 200664132Haemophilia patients with HCV (27 co-infected with HIV)Fibrotest, APRI, Forns index, age-platelet index, hyaluronic acidConcordance of Fibrotest with APRI and/or Forns can be used to avoid biopsy. Fibrotest accurately identified advanced or minimal liver disease
Obrador, 2006152332HCVVarious biochemical and US methodsA score derived from 7 variables that predicted cirrhosis (age ≥ 60,platelet ≤ 100, AST:ALT ≥ 1, PT ≥ 1.1, caudate hypertrophy, right lobe atrophy and splenomegaly) ≥22 had a significant probability of cirrhosis (sensitivity 80%; specificity 96%; diagnostic accuracy 94%)
Masaki, 200610033
HCV/HIV haemophiliacs HCV mononfectedFibroscan, abdominal ultrasound (US), platelet counts, type IV collagen, procollagen type III, hyaluronic acidIn co-infected group, Fibroscan correlated with IV-coll, US, procoll type III, PLT and hyaluronic acid. In non-HIV group, Fibroscan correlated with AUS, platelet count, and IV-coll
Kawamoto, 200615330Normal, HBV, HCVFA by DIA, Fibroscan, platelet count, LCAT, pre-Albumin, APRI, hyaluronate, type IV collagenThe AUROC for FibroScan was higher than that for the other markers, but statistical significance was minimal.
De Ledinghen, 200613572HIV/HCV co-infected patientsElastography, platelet count, AST/ALT, AST/APRI, FIB-4Liver stiffness significantly correlated with fibrosis stage (Kendall tau-b = 0.48; P < 0.0001). AUROC curve of stiffness was 0.72 for F ≥ 2 and 0.97 for F = 4
Metwally, 2007154199
HCV (testing)
26 variablesScoring system developed using platelet count, AST, albumin. Cutoff of 4 had 99% specificity and 94% PPV; cutoff of 2 had 87% sensitivity and 95% NPV for severe fibrosis
Halfon, 200751356HCVFT, APRI, Fibrometer, hepascoreThe AUROCs of these 4 methods were similar
Leroy, 2007129180HCVMP3, Fibrotest, Fibrometer, Hepascore, Forns, APRIBest combinations selected 1/3 of patients for whom absence of significant fibrosis or presence of extensive fibrosis could be predicted with >90% certainty
Coco, 200788228Chronic viral hepatitisFibroscan, ALT, APRI, FORNS, Fibrotest, HAComparisons between APRI and FibroScan AUROCs showed that Fibro-Scan better identified both fibrosis ≥ F2 (P < 0.001) and cirrhosis (P < 0.001). The AUROC of FORNS for fibrosis ≥ F2 was 0.913, comparable to FibroScan.

Lackner compared AST/ALT ratio, cirrhosis discriminant score (CDS), which is derived from platelet count, ALT/AST ratio and PT, age to platelet index (AP), Pohl score (which includes AST, ALT, and platelet count), APRI, and platelet count alone and found that significant fibrosis was reliably predicted and cirrhosis reliably excluded by cutoff points for APRI and platelet count in 85% and 78% of patients respectively.121 Kelleher found that HA, albumin and AST accurately staged mild and advanced fibrosis,122 whereas Fabris found that APRI was better than other markers.123 However, Romera found that because of intermediate values, Sydney’s index, which uses insulin resistance together with age, AST, cholesterol, and alcohol use, Forns and APRI were not applicable in approximately 40% of patients.124 Combinations of markers have also been evaluated with variable results. Borroni evaluated combinations of five measures and found that this only slightly changed accuracy, but concluded that double cutoffs for cirrhosis discriminant score and APRI may reduce the number of patients requiring biopsy.125 Bouliere, however, evaluated FT, APRI and Forns and found that the three methods allowed fibrosis to be well evaluated in >80% of patients, but biopsy remained mandatory in approximately 20% of patients,126 whereas Sebastiani found that a stepwise combination of these three methods reduced need for liver biopsy in 50–70% of patients,127 which was similar to Maor’s results.128 Leroy evaluated several indices as well and found that the best combinations selected one-third of patients for whom absence of significant fibrosis or presence of extensive fibrosis could be predicted with >90% certainty.129

Radiological methods have also been compared with serum biomarkers. Oberti originally found that HA and prothrombin index [(now the international normalized ratio (INR)] were the best predictive factors when compared with other clinical, biochemical, US and endoscopic markers,130 whereas Aube found that Doppler US was superior to clinical and biochemical methods for accurate diagnosis of cirrhosis, but was less accurate for severe fibrosis.131 Testa also evaluated various combinations of tests, including ultrasound, and found these to be effective in identifying or ruling out fibrosis in 75–80% of patients.132

Analyses have also compared transient elastography with other non-invasive tests. Saito found that both elastometry and platelet counts were correlated with fibrosis stage, but the deviation was less with elastometry.133 Castera found that when FibroScan and FibroTest results agreed, liver biopsy confirmed them in 84% of cases for F ≥ 2, in 95% for F ≥ 3, and in 94% for F = 4.134 De Ledinghen also found that for the diagnosis of cirrhosis, AUROC curves of liver stiffness were significantly higher than those for platelet count, AST/ALT, AST/APRI and FIB-4.135 Lastly, Coco found that Fibroscan performed better than APRI, Forns, Fibrotest and HA and also that ALT was independently associated with stiffness.88


Liver biopsy is considered the gold standard for the histopathological assessment of liver tissue, as it is considered the most accurate for assessing severity and aetiology of liver disease as well as monitoring response to therapy. Biopsy can be utilized to diagnose, grade and stage multiple types of chronic liver disease; however, biopsy is not without inherent risks and there are patients in whom assessment of liver disease is needed, but biopsy is contraindicated.

Given these risks and potential sampling variability, there has been a substantial drive to develop non-invasive measures of hepatic fibrosis, especially for the grading of severity of chronic hepatitis C both prior to therapy and during therapy. There have been numerous methods including biochemical tests, hepatic imaging and breath tests, which have been developed to answer this need for non-invasive assessment. However, the literature is virtually overwhelming with numerous methods described above, many of which are not readily available outside the research institution and with significant costs. We have provided a concise literature review detailing the results of these various studies to help gain a better understanding of the various modalities currently being utilized to assess liver fibrosis non-invasively in patients with chronic hepatitis C infection.

The ideal assessment of fibrosis would involve a readily available test with the ability to discern accurately between degrees of fibrosis with high sensitivity and specificity. Several indices and models have been developed using routine laboratory tests, including APRI, Forns and FIB-4. A recent systematic review has detailed the results of 22 studies analysing the accuracy of APRI. They found that APRI could reduce the need for liver biopsy by 35% for significant fibrosis and concluded that the major strength of this non-invasive method was exclusion of significant fibrosis.136 The FIB-4 index has also been validated in HCV mono-infected patients, with 78% of the 847 biopsies correctly classified.29 With regard to HIV/HCV co-infected patients, systematic review and meta-analysis of diagnostic accuracy of non-invasive marker panels have shown acceptable performance of these models for assessment of fibrosis and cirrhosis.137

Several models of nonroutine measurements of extracellular matrix remodelling markers (PIIIP, MMP, TIMP, HA and CL-4) have also been developed. Fibrotest, a composite measurement of several serum proteins, as well as Actitest, a modified Fibrotest which includes ALT, have been developed as non-invasive fibrosis measurement tools. However, the components of these models are not readily available and have limited applicability because two components, haptoglobin and bilirubin, can give false positive results. Most trials reported that Actitest was highly accurate in patients with chronic hepatitis C with meta-analysis resulting in an AUROC range 0.73–0.87 for significant fibrosis and 0.75–0.86 for significant necrosis,138 although some have reported discrepancies.139 FIBROSpect II, a diagnostic panel of extracellular matrix remodelling markers has also been shown to differentiate mild from severe fibrosis, but was not as accurate for intermediate fibrosis. Other non-routine tests including proteomic markers, various serum protein and enzyme markers and the use of artificial neural networks are under investigation, but like the other aforementioned tests, their applicability is limited by availability, cost and the inability to differentiate intermediate fibrosis.

Various imaging tests including ultrasound, magnetic resonance imaging, elastography, single photon emission computed tomography and tissue strain imaging have been investigated for the assessment of liver fibrosis. Ultrasound is readily available and is used for HCC screening, but studies have shown that unenhanced Doppler ultrasound is not reliable in the discrimination of varying degrees of fibrosis. Results can be improved, however, with additional measurements such as heart pulsation at the liver surface and portal venous flow measurements. Several types of enhanced MR imaging have been developed to evaluate the degree of liver fibrosis, including MR elastography and 31-P MR spectoscopy, which have both been shown to differentiate mild from moderate or advanced fibrosis. The limitations of MR studies, however, include cost, availability, especially with enhanced MR imaging and the inability to differentiate degrees of intermediate fibrosis. The ability to differentiate intermediate fibrosis, thereby, does not make MR more appealing than the less expensive models using routine serum measurements. SPECT imaging is another non-invasive method that can separate normal from cirrhotic livers, but is not readily available and also does not show more diagnostic capability than other readily available tests. TE appears easy-to-perform and may be readily available in a few years. However, its use may be limited in patients with narrow intercostal spaces or morbid obesity. Although MRE may have the greatest accuracy, its cost, limited availability and procedure duration limit its widespread use.

The methacetin breath test, essentially a microsomal liver function test, has been shown to have high sensitivity and specificity in identifying cirrhotic patients. Utilizing the aminopyrine substrate in the ABT, however, did not differentiate patients with F3-F4 scores from patients with F1-F0 fibrosis and cirrhosis.

With the goal of using readily available modalities, as well as our findings that nonroutine methods did not demonstrate increased performance in the differentiation between minimal and significant fibrosis, we recommend using a readily available model such as APRI, FIB-4, Fibroindex or Forns for initial evaluation. As multiple studies comparing various methods of fibrosis have also demonstrated increased diagnostic accuracy with the combination of these tests,88–91, 94, 123 we recommend performing confirmatory testing with either a second of these tests or if available, transient elastography (Figure 2). This approach would effectively differentiate those patients with mild disease who could defer therapy from those with significant fibrosis, with liver biopsy only being performed on those patients with indeterminate results.

Figure 2.

 Algorithm for assessment of fibrosis in patients with chronic hepatitis C.

In conclusion, the accurate determination of the degree of liver fibrosis is essential for determining the need for treatment and for determining the prognosis in patients with chronic HCV infection. Liver biopsy is still considered the gold standard for making this determination, but great strides are being made in the development of accurate non-invasive methods for determination of severity of fibrosis. These methods include routine and nonroutine serum measurements, radiographic imaging and breath testing.


Declaration of personal interests: None. Declaration of funding interests: The initial literature review was funded by Roche.