Owing to the high prevalence of various chronic liver diseases, cirrhosis is one of the leading causes of morbidity and mortality worldwide. In recent years, the development of non-invasive tests of fibrosis allows accurate diagnosis of cirrhosis and reduces the need for liver biopsy. In this review, we discuss the application of these non-invasive tests beyond the diagnosis of cirrhosis. In particular, their role in the selection of patients for hepatocellular carcinoma surveillance and varices screening is highlighted.
Despite geographical differences, chronic liver diseases are highly prevalent worldwide. It is estimated that at least 350 million and 120 million people globally are chronically infected with hepatitis B virus (HBV) and hepatitis C virus, respectively.
The diagnosis of cirrhosis is not as simple as it seems. Evidently, the diagnosis is straightforward when a patient has already developed clinical manifestations of portal hypertension such as ascites, varices and hypersplenism. Nonetheless, these signs are absent in patients with early cirrhosis, and the radiological features of early cirrhosis are subtle and unreliable.
In recent years, the development and application of non-invasive tests of liver fibrosis have revolutionized hepatology practice. Numerous studies have confirmed the accuracy of these tests in fibrosis staging and the diagnosis of cirrhosis. In general, the tests have high negative predictive value in excluding advanced fibrosis and cirrhosis and have been recommended by the European Association for the Study of the Liver as initial assessment in patients with various liver diseases.
In addition, cirrhosis is not one single disease but encompasses a broad spectrum of clinical condition ranging from compensated disease to decompensated disease. As the disease progresses, various complications of portal hypertension may develop. The development of hepatocellular carcinoma (HCC) further drifts the clinical course and leads to major morbidity and mortality. Therefore, one important part of the management of cirrhosis is to identify and treat major complications early. In this review, we first provide an overview on non-invasive tests of liver fibrosis. Since the diagnosis of cirrhosis is only the first step in the management of cirrhosis, we further discuss the potential application of these tests in the risk stratification of cirrhosis and prediction of cirrhotic complications.
Non-invasive tests of liver fibrosis have been a hot research area in the past decade. At the beginning, the main focus was to reduce the burden of liver biopsy by confidently identifying patients who are very unlikely to have significant fibrosis on one hand and those who are very likely to have advanced fibrosis or cirrhosis on the other. Treatment decisions can then be made accordingly, and patients in the middle (gray zone cases) may undergo liver biopsy or be observed over time. In general, non-invasive tests of liver fibrosis can be divided into serum tests and physical measurements.
The advantages of serum tests include high applicability (successful measurements can be made in most cases) and relatively simple logistics. Doctors may obtain blood samples at their clinics and send them to designated laboratories even for more specific biomarkers. Serum tests can be divided into class I biomarkers and class II biomarkers. Class I biomarkers specifically measure the activity of fibrogenesis or fibrinolysis. In contrast, class II biomarkers do not measure fibrosis directly but represent parameters that correlate with fibrosis. For example, aspartate aminotransferase (AST) is a marker of hepatic necroinflammation and not fibrosis. However, patients with fibrosis and cirrhosis often have increased AST levels. Although class I biomarkers are expected to directly reflect fibrosis and be more accurate than class II biomarkers, this has not been consistently demonstrated in prospective studies.
In any case, at present there is no single marker that can adequately reflect fibrosis. Therefore, in most situations several biomarkers or a combination of biomarkers and other clinical features are used. Some of the combined panels such as FibroTest, FibroMeter and enhanced liver fibrosis (ELF) score have been commercialized. It should be noted that such combined tests are modeled against liver histology, which is an imperfect reference standard. In other words, even if the models can 100% faithfully reflect liver histology, the accuracy of liver histology to diagnose fibrosis and cirrhosis will be the ceiling of accuracy of the new models.
Some of the class II biomarkers are more generic. Examples include the AST-to-alanine aminotransferase (ALT) ratio and the AST-to-platelet ratio index (APRI). In other cases, owing to the pathophysiology of different liver diseases, the class II biomarkers are more disease-specific. For instance, metabolic factors are overrepresented in the NAFLD fibrosis score, which should only be applied in patients with NAFLD.
The other main class of non-invasive tests of liver fibrosis relies on physical measurement of liver stiffness and elasticity. Although the cutoffs of the measurements are determined with reference to histology, these tests are not modeled against histology and theoretically may achieve better prediction than histology. In fact, studies in patients with chronic viral hepatitis suggest that transient elastography may be better than histology in predicting overall mortality.
Transient elastography by FibroScan (Echosens, Paris, France) is currently the most commonly used method to measure liver stiffness.
Newer techniques such as acoustic radiation force impulse and shear-wave elastography allow simultaneous visualization of the liver parenchyma and measurement of liver elasticity.
For simplicity, doctors usually adopt one or more recommended cutoff values in the interpretation of any of the non-invasive tests above. In reality, however, the more extreme the values of the non-invasive tests are, the more we are confident in whether a patient has fibrosis/cirrhosis or not. For example, a probability-based interpretation of liver stiffness measurements (LSM) has been proposed.
HCC is one of the most important complications in patients with chronic liver diseases.
HCC surveillance improves the prognosis of patients by identifying tumors of smaller sizes, fewer numbers of tumors, and longer overall survival.
The key components of an optimal surveillance program include accurate risk stratification and reliable surveillance tools. Hence there is a need for accurate HCC risk prediction to assist prognostication as well as decision on the need for HCC surveillance.
Various non-invasive tests of liver fibrosis have been tested to predict the risk of HCC. Among them LSM with transient elastography is the most widely-studied. A dose-response relationship between LSM and HCC risk was demonstrated in patients with either chronic hepatitis B (CHB) or chronic hepatitis C (CHC).
FibroTest is one of most popular serum-based non-invasive tests for liver fibrosis.
Another serum-based non-invasive test, enhanced liver fibrosis (ELF) test has been increasingly used to assess liver fibrosis by its or in combination with LSM in CHB patients.
The risk factors of HCC in patients with chronic hepatitis B are well known, and various groups have derived HCC risk scores based on the factors.
LSM has also been integrated with age, gender, HBV DNA level into a regression formula to predict HCC with good accuracy.
Portal hypertension is one of the most lethal complications of chronic liver disease.
However, EGD is unpleasant. Some patients at risk would never agree to undergo EGD because of the perceived discomfort. On the other hand, many other patients go through the procedure with negative findings. This may cause unnecessary suffering and increase health costs. The use of thrombocytopenia and splenomegaly has been proposed to identify patients with portal hypertension and varices, but the accuracy remains limited.
Ascites, hepatic encepalopathy, and variceal bleeding are some of the complications of liver cirrhosis. The development of these events is due to portal hypertension. The gold standard for detecting portal hypertension is through the measurement of hepatic venous pressure gradient (HVPG). However, it is invasive and not readily available. Therefore, a number of studies have evaluated the use of non-invasive tests to predict portal hypertension (
Splenomegaly and hypersplenism are features of portal hypertension. Indeed, spleen stiffness (54 kPa) can predict survival free complications among HCV-related cirrhotic patients.
Furthermore, the use of duplex doppler ultrasonography has been an attractive non-invasive test to determine portal hypertension by assessing the vascular anatomy and its hemodynamics. Parameters such as portal blood flow and velocity, resistive and pulsatility indices have also been explored to predict HVPG although with conflicting results. In a study of cirrhotic patients with different etiologies, portal vein velocity was found to correlate with HVPG, but the findings have not been confirmed by all studies.
A Korean group further developed a simple risk score comprising bilirubin and platelets to predict HVPG.
As mentioned above, EGD should be performed for varices screening in cirrhotic patients, but the procedure is unpleasant and often omitted. Since non-invasive tests of fibrosis can reflect portal hypertension, it is logical to consider their application in selecting patients for EGD. LSM has the highest accuracy with a sensitivity of 76-91% and specificity varies of 28-88% to detect esophageal varices (
In the recent Baveno VI guideline, some of these non-invasive tests were already incorporated to stratify patients.
Non-invasive tests of liver fibrosis have revolutionized the management of chronic liver diseases. Compared with routine clinical assessments, the non-invasive tests allow more confident diagnosis of cirrhosis and can also reflect the severity of cirrhosis and/or portal hypertension. They can therefore be used to select patients for HCC surveillance and varices screening.
That said, there are still a number of questions regarding the use of the non-invasive tests. The interval of testing is currently undefined, and it is unclear to what extent the changes in the non-invasive tests reflect fibrosis progression. In addition, patients with treated viral hepatitis often have reduced liver stiffness and improved serum tests of fibrosis.
alanine aminotransferase
AST-to-platelet ratio index
aspartate aminotransferase
chronic hepatitis B
chronic hepatitis C
esophagogastroduodenoscopy
enhanced liver fibrosis
hepatitis B virus
hepatocellular carcinoma
hepatitis C virus
hazard ratio
hepatic vein pressure gradient
liver stiffness measurement
non-alcoholic fatty liver disease
Factors | CU-HCC score |
LSM-HCC score |
---|---|---|
Age | ||
> 50 years | +3 | +10 |
≤ 50 years | 0 | 0 |
Albumin | ||
≤ 35g/L | +20 | +1 |
> 35g/L | 0 | 0 |
Total bilirubin (μmol/L) | ||
> 18 | +1.5 | |
≤ 18 | 0 | |
HBV DNA | ||
> 200,000 IU/mL | +4 | +5 |
2,000–200,000 IU/mL | +1 | 0 |
≤ 2,000 IU/mL | 0 | 0 |
Cirrhosis | ||
Yes | +15 | |
No | 0 | |
Liver stiffness measurement | ||
≤ 8.0 kPa | 0 | |
8.1-12.0 kPa | +8 | |
> 12.0 kPa | +14 |
HBV, hepatitis B virus; LSM, liver stiffness measurement.
Total CU-HCC score ranges from 0 to 44.5. Scores of 0 to 4, 5 to 19 and 20 to 44.5 indicate low, intermediate and high risk respectively.
Total LSM-HCC score ranges from 0 to 30. Scores of 0 to 10, 11 to 20 and 21 to 30 indicate low, intermediate and high risk respectively.
Author | N | Non-invasive tests | Cutoff | Endpoint | AUROC | Sensitivity | Specificity |
---|---|---|---|---|---|---|---|
Tasu 2002 [ |
50 | Hepatic arterial acceleration index by duplex ultrasound | 1 ms-2 | HVPG 12 mmHg | 0.83 | 65% | 95% |
Portal vein velocity | 17cm/s | HVPG 12 mmHg | - | 69% | 67% | ||
Bolognesi 2001 [ |
40 | Estimated portal pressure by echo-color Doppler | HVPG 16 mmHg | - | 82% | 70% | |
Berzigotti 2006 [ |
31 | Renovascular impedance | 0.7 | HVPG 16 mmHg | - | 52% | 100% |
Kim 2007 [ |
76 | Damping index by Doppler hepatic vein waveform | 0.6 | HVPG 12 mmHg | 0.86 | 76% | 82% |
Choi 2003 [ |
138 | Portal vein velociity & flow | HVPG | No correlation | No correlation | ||
Splenic vein velocity & flow | HVPG | No correlation | No correlation | ||||
Pulsatility index | HVPG | No correlation | No correlation | ||||
Resistive index | HVPG | No correlation | No correlation | ||||
Portal vein velocity | Change in HVPG with terlipressin | No correlation | No correlation | ||||
Taurel 1998 [ |
40 | Portal vein time-average mean blood velocity | - | HVPG 12 mmHg | - | - | - |
Portal vein flow | - | HVPG 12 mmHg | - | - | - | ||
Hepatic artery resistance index | - | HVPG 12 mmHg | - | No correlation | No correlation | ||
Superior mesenteric artery resistance index | - | HVPG 12 mmHg | - | No correlation | No correlation | ||
Baik 2006 [ |
78 | Hepatic waveforms by Doppler ultrasound | Mono-phasic waveform | HVPG 15 mmHg | - | 74% | 95% |
Robic 2011 [ |
65 | LSM | 21.1 kPa | Portal hypertension-related complications | 0.734 | 100% | 41% |
Lemoine 2008 [ |
92 | LSM | 20.5 kPa | HVPG 10 mmHg | 0.76 | 63% | 70% |
LSM | 34.9kPa | HVPG 10 mmHg | 0.94 | 90% | 88% | ||
Procopet 2015 [ |
202 | LSM | 13.6kPa | HVPG 10 mmHg | Training cohort- 0.94 | Training cohort- 95.6% | Training cohort- 75.3% |
Validation cohort- not reported | Validation cohort- 86.7% | Validation cohort- 69.3% | |||||
LSM | 21.1 kPa | HVPG 10 mmHg | Training cohort- 0.94 | Training cohort- 88.9% | Training cohort- 87.6% | ||
Validation cohort- not reported | Validation cohort- 80% | Validation cohort- 96.6% | |||||
Lok score | 0.73 | HVPG 10 mmHg | Training cohort- 0.84 | Training cohort- 87% | Training cohort- 80.6% | ||
Validation cohort- not reported | Validation cohort- 66.7% | Validation cohort- 88.5% | |||||
Risk score | -1 | HVPG 10 mmHg | Training cohort- 0.80 | Training cohort- 76.4% | Training cohort- 78.9% | ||
Validation cohort- not reported | Validation cohort- 60% | Validation cohort- 79.3% | |||||
FIB-4 | 3.25 | HVPG 10 mmHg | Training cohort- 0.79 | Training cohort- 71.4% | Training cohort- 73.1% | ||
Validation cohort- not reported | Validation cohort- 53.3% | Validation cohort- 82.8% | |||||
Colecchia 2014 [ |
92 | Spleen stiffness | <54 kPa | Low risk for clinical decompensation | - | 97% | 63% |
Vizzutti 2007 [ |
61 | LSM | 13.6 kPa | HVPG 10 mmHg | 0.99 | 97% | 92% |
LSM | 17.6 kPa | HVPG 12 mmHg | 0.92 | 94% | 81% | ||
Carrion 2006 [ |
124 | LSM | 8.74 kPa | HVPG 6 mmHg | 0.93 | 90% | 81% |
Vizzutti 2007 [ |
66 | Splenic artery resistance index | 0.6 | HVPG 12 mmHg | No correlation | - | - |
Superior mesenteric artery pulsatility index | 2.7 | HVPG 12 mmHg | No correlation | - | - | ||
Right renal artery resistance index | 0.65 | HVPG 12 mmHg | No correlation | - | - | ||
Park 2009 [ |
61 | Risk score with platelets and bilirubin | -1 | HVPG 10 mmHg | 0.91 | 88% | 86% |
AUROC, area under the receiver-operating characteristics curve; HVPG, hepatic vein pressure gradient; LSM, liver stiffness measurement
Author | Etiology | N | Non-invasive tests | Cutoff for varices | Cutoff kr large varices | AUROC | Sensitivity | Specificity |
---|---|---|---|---|---|---|---|---|
Castera 2009 [ |
Chronic hepatitis C | 124 | LSM | 21.5 kPa | 0.84 | 76% | 78% | |
Platelet count | <140x109/L | 0.69 | 56% | 76% | ||||
Fibrotest | 0.78 | 0.72 | 72% | 69% | ||||
Prothrombin index | 80% | 0.68 | 44% | 84% | ||||
AST/ALT ratio | 1 | 0.83 | 68% | 89% | ||||
AST-to-platetlet ratio index | 1.3 | 0.62 | 68% | 64% | ||||
Lok index | 0.6 | 0.81 | 68% | 82% | ||||
Sebastiani 2010 [ |
Mixed | 620 | Lok index + Forns’ index | Lok index (0.9) + Forns’ index (8.5) | Retrospective -0.82 | Retrospective -65% | Retrospective -82% | |
Prospective -0.81 | Prospective -79% | Prospective -62% | ||||||
Patanwala 2013 [ |
Primary biliary cirrhosis | 529 | New Castle Varices in PBC score | 0.3 | 0.863 | 93% | 61% | |
Colecchia 2011 [ |
Biliary atresia | 31 | AST-to-platetlet ratio index | 0.96 | 0.88 | 86% | 81% | |
LSM | 10.6 kPa | 0.92 | 87% | 88% | ||||
Platelet/spleen | 1.06 | 0.90 | 73% | 93% | ||||
LSPS (LSM x spleen diameter/platelet) | 9.2 | 0.96 | 91% | 92% | ||||
Mangone 2012 [ |
Mixed | 87 | Platelet/spleen | 936.4 | 0.67 | 姑% | 64% | |
Galal 2011 [ |
Mixed | 154 | Hyaluronic acid | - | 207 ug/L | 0.92 | 94% | 78% |
Hong 2009 [ |
Chronic hepatitis B | 146 | Portal vein diameter | 12 mm | 0.74 | 84% | 57% | |
Spleen width | 46 mm | 0.74 | 72% | 76% | ||||
Regression Function | 0.3631 | 0.78 | 87% | 60% | ||||
Giannini 2003 [ |
Mixed | 266 | platelet/spleen diameter ratio | 909 | 0.92 | 100% | 71% | |
Barrera 2009 [ |
Not specified | 67 | platelet/spleen diameter ratio | 830.8 | 0.78 | 77% | 74% | |
Thabut et al. 2006 [ |
Not specified | 99 | FibroTest | 0.8 | 0.77 | 92% | 21% | |
Vermehren et al. 2011 [ |
Mixed | 166 | Acoustic radiation force impulse of the spleen | 3.04 m/s | 0.58 | 90% | 25% | |
LSM | 20.5 kPa | 0.53 | 91% | 28% | ||||
Morishita 2014 [ |
Chronic hepatitis C | 135 | Acoustic radiation force impulse | 2.05 m/s | 2.39 m/s | OV 0.89; High risk OV 0.87 | OV 83%; High risk OV 81% | OV 76%; High risk OV 82% |
Platelet count | 8.25x10^4/mm3 | 7.95 | OV 0.74; High risk OV 0.66 | OV 67%; High risk OV 64% | OV 67%; High risk OV 63% | |||
FIB-4 | 6.21 | 7.7 | OV 0.75; High risk OV 0.74 | OV 71%; High risk OV 67% | OV 69%; High risk OV 78% | |||
AST-to-platetlet ratio index | 1.5 | 1.62 | OV 0.68; High risk OV 0.67 | OV 59%; High risk OV 64% | OV 64%; High risk OV 68% | |||
Cherian 2011 [ |
Mixed | 229 | Child-Pugh score | class B/C | class B/C | not reported for OV/large OV | OV-not reported; large OV 95% | OV-not reported; large OV 25.7% |
Spleen diameter | 150 mm | 160 mm | OV-not reported; large OV 0.63 | OV-not reported; large OV 67% | OV-not reported; large OV 54.7% | |||
Platelet count | 100,000/uL | 90,000/uL | OV-not reported; large OV 0.70 | OV-not reported; large OV 59% | OV-not reported; large OV 64% | |||
Portal vein diameter | 13 mm | - | OV-not reported | OV-not reported | OV-not reported | |||
Eslam 2013 [ |
Mixed | 280 | Adiponectin + platelet count + homeostasis model assessment of insulin resistance (HOMA-IR) | Adepoectin 19.2 ug/L platelet 100 x103 HOMA IR 4 | -estimation cohort 0.88 | -estimation cohort 91% | -estimation cohort 87% | |
-validation cohort 0.80 | -validation cohort 87% | -validation cohort 82% | ||||||
Taourel et al. 1998 [ |
Alcoholic liver disease | 40 | Hepatic artery resistance index | 0.72 | - | - | - | |
Vizzutti et al. 2007 [ |
Chronic hepatitis C | 61 | LSM | 17.6 kPa | 0.76 | 90% | 43% | |
Vizzutti 2007 [ |
Chronic hepatitis C | 66 | Splenic artery resistance index | 0.6 | No correlation | - | - | |
Superior mesenteric artery pulsatility index | 2.7 | No correlation | - | - | ||||
Right renal artery resistance index | 0.65 | No correlation | - | - | ||||
Park 2009 [ |
various | 61 | Risk score with platelets and bilirubin | -1 | 0.82 | 82% | 76% | |
Procopet 2015 [ |
various | 202 | LSM | 13.6 kPa | Training cohort- 0.90 | Training cohort-88% | Training cohort- 71% | |
Validation cohort- not reported | Validation cohort- 93% | Validation cohort- 79% | ||||||
LSM | 21.1 kPa | Training cohort- 0.90 | Training cohort- 84% | Training cohort- 84% | ||||
Validation cohort- not reported | Validation cohort-71% | Validation cohort- 88% | ||||||
Lok score | 0.73 | Training cohort- 0.86 | Training cohort- 87% | Training cohort- 80% | ||||
Validation cohort- not reported | Validation cohort- 57% | Validation cohort- 83% | ||||||
Risk score with platelets and bilirubin | -1 | Training cohort- 0.84 | Training cohort- 77% | Training cohort-82% | ||||
Validation cohort- not reported | Validation cohort- 57% | Validation cohort- 75% | ||||||
FIB-4 | 3.25 | Training cohort- 0.80 | Training cohort- 74% | Training cohort- 72% | ||||
Validation cohort- not reported | Validation cohort- 57% | Validation cohort- 88% |
ALT, alanine aminotransferase; AST, aspartate aminotransferase; AUROC, area under the receiver-operating characteristics curve; LSM, liver stiffness measurement.