Article Text

Non-alcoholic fatty liver disease: non-invasive investigation and risk stratification
  1. J K Dyson1,
  2. S McPherson1,2,
  3. Q M Anstee1,2
  1. 1 Liver Unit, Newcastle Upon Tyne Hospitals NHS Trust, Freeman Hospital, Newcastle upon Tyne, UK
  2. 2 Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, UK
  1. Correspondence to Dr Q M Anstee, Institute of Cellular Medicine, The Medical School, Newcastle University, M4 090, 4th Floor, William Leech Building, Framlington Place, Newcastle-upon-Tyne NE2 4HH, UK; quentin.anstee{at}


Non-alcoholic fatty liver disease (NAFLD) encompasses a histological spectrum of liver disease, from simple steatosis through to cirrhosis. As the worldwide rates of obesity have increased, NAFLD has become the commonest cause of liver disease in many developed countries, affecting up to a third of the population. The majority of patients have simple steatosis that carries a relatively benign prognosis. However, a significant minority have non-alcoholic steatohepatitis, and have increased liver related and cardiovascular mortality. Identifying those at risk of progressive disease is crucial. Liver biopsy remains the gold standard investigation for assessing stage of disease but its invasive nature makes it impractical for widespread use as a prognostic tool. Non-invasive tools for diagnosis and disease staging are required, reserving liver biopsy for those patients where it offers clinically relevant additional information. This review discusses the non-invasive modalities available for assessing steatosis, steatohepatitis and fibrosis. We propose a pragmatic approach for the assessment of patients with NAFLD to identify those at high risk of progressive disease who require referral to specialist services.

  • fibrosis

Statistics from


Non-alcoholic fatty liver disease (NAFLD) is defined as fatty infiltration affecting greater than 5% of hepatocytes in the absence of excessive alcohol consumption (>20 g day for women and >30 g/day for men). NAFLD is strongly associated with insulin resistance and the metabolic syndrome.1 Against a background of increasing rates of obesity worldwide, NAFLD has become the commonest cause of liver disease in many developed countries, affecting up to a third of the population.2

The prevalence of NAFLD increases with age and varies between populations studied and with the modality of ascertainment.1 ,3 Increased waist circumference is probably the best single clinical predictor of underlying insulin resistance and the presence of NAFLD. The European DIONYSOS study identified NAFLD in 25% of normal weight subjects, in 67% of overweight subjects (body mass index (BMI) 25–29 kg/m2) and in 94% of obese subjects (BMI ≥30 kg/m2).2 ,4 In type 2 diabetes mellitus (T2DM), the prevalence of NAFLD is 40–70%.2 ,5–8 Over 90% of patients with NAFLD have one or more features of the metabolic syndrome and 30–40% have the full syndrome (three or more of: central obesity (waist circumference ≥94 cm or ≥80 cm for men and women, respectively), impaired fasting glucose (>5.6 mmol/L), hypertriglyceridaemia (>1.7 mmol/L), low serum high density lipoprotein cholesterol (<1.0 mmol/L or <1.3 mmol/L for men and women, respectively) and hypertension (>135/85 mm Hg)).9 The severity of NAFLD is also associated with the severity of the metabolic syndrome.10 ,11

Natural history and modifiers of disease progression

NAFLD encompasses a histological spectrum of liver disease from simple steatosis, through non-alcoholic steatohepatitis (NASH; characterised by inflammation and hepatocyte ballooning degeneration), to fibrosis and ultimately cirrhosis. The majority (70–90%) of patients with NAFLD have simple steatosis that carries a relatively benign prognosis.7 ,12 ,13 However, a significant minority have NASH and are at risk of progressive disease.12 ,13 To date, there are limited high quality prospective data on NAFLD progression, particularly in the primary care setting, as routine biochemical indices do not accurately reflect disease activity or fibrosis.14 ,15 Epidemiological studies indicate that, unlike steatosis, NASH is associated with a 10-fold increased risk of dying from liver disease (2.8% vs 0.2%, respectively) and a doubling of risk of a cardiovascular death compared with a reference population (15.5% vs 7.5%; p=0.04 for both).13 Up to 5.4% of patients with NASH develop complications of end-stage liver disease during long-term follow-up.2 ,13 ,16

The substantial interindividual variability in disease progression and outcome reflects the variety of environmental and genetic factors that influence pathogenesis (recently reviewed).17 Interest in the role of genetic factors has increased following the identification of polymorphisms in the patatin-like phosholipase domain containing 3 (PNPLA3) gene, associated with greater steatosis and more severe inflammation and fibrosis.18 ,19 Of the numerous genes that have been associated with NAFLD either in these studies or through candidate gene analysis, only a minority have been independently validated and so can be considered of proven significance. These include PNPLA3,18–23 mitochondrial Superoxide dismutase 2 (SOD2),24 ,25 phosphatidylethanolamine methyltransferase (PEMT)26 ,27 and Kruppel-like factor 6 (KLF6).28 ,29 A number of environmental factors have also been identified as significant modifiers of disease activity. These include diets high in fat and/or fructose (widely adopted as animal models for NAFLD research).30–32 There is also mounting evidence that the microbiome contributes to NAFLD pathogenesis throughout its progression from insulin resistance and abdominal obesity, through to fibrosing NASH.33 Small bowel bacterial overgrowth is commoner in patients with NASH than controls and is associated with more severe steatosis.34 ,35 Perhaps most intriguing are reports demonstrating the close interactions between the host innate immune system and the microbiome.36 Another identified factor is the presence of obstructive sleep apnoea.37–39 Finally, it should be remembered that NAFLD can coexist with other aetiologies of liver disease where it accelerates progression of liver disease due to hepatitis C,40 ,41 haemochromatosis42 and alcohol excess.43

Diagnosis and risk stratification

In light of the associated increased morbidity, there is a clinical imperative to distinguish those patients with NASH from those with steatosis, as the former are at greatest risk.16 Thus hepatological care and additional metabolic risk factor modification can be focused on those that will most benefit.15 ,16 Although population screening for NAFLD is not currently recommended, subjects with obesity, metabolic syndrome and T2DM are at very high risk of having NASH, so clinicians should have a high index of suspicion for the diagnosis in these subjects, even in the presence of normal liver function tests (LFTs).16 ,44

The majority of patients with NAFLD are asymptomatic and so are diagnosed following an incidental finding of abnormal LFTs or fatty liver on imaging.16 Where present, the typical biochemical abnormalities found in NAFLD are mildly elevated transaminases (alanine aminotransferase (ALT) >aspartate aminotransferase (AST)) and/or gamma-glutamyltransferase. However, routine biochemical tests are insensitive as ∼80% of patients have normal range ALT levels (men <40 IU/L and women <31 IU/L).45 Based on the assumption that traditional ‘normal ranges’ were derived from a cohort that contained individuals with undiagnosed liver disease, Prati et al proposed revised thresholds (men <30 IU/L and women <19 IU/L) derived from a cohort where overweight individuals were excluded.46 These values improve diagnostic sensitivity but not specificity. If elevated, ALT typically falls (and AST may rise) as fibrosis progresses to cirrhosis but absolute values have no relationship with histological disease activity and so are unhelpful in diagnosing NAFLD, distinguishing steatosis from steatohepatitis or quantifying fibrosis.14

Where abnormal clinical biochemistry is detected, alternative aetiologies of liver disease, including high alcohol consumption, chronic viral hepatitis, autoimmune liver disease and haemochromatosis, should be excluded.47 High ferritin levels are frequently seen in patients with NAFLD, raising diagnostic uncertainty about underlying genetic haemochromatosis. This phenomenon is well recognised and usually reflects underlying inflammatory activity. There is also evidence that it is indicative of steatohepatitis and more advanced fibrosis.48 ,49 If tested, a transferrin saturation <45% generally rules out haemachromatosis.50 Positive autoantibodies (antinuclear antibody ≥1:160 and/or antismooth muscle antibody ≥1:40), usually of relatively low titre, are present in 21–23% of patients with biopsy proven NAFLD.51 ,52 The clinical significance of this finding is uncertain. It generally does not indicate coexistent autoimmune hepatitis but may be associated with more advanced fibrosis and inflammation (although this has not been uniformly demonstrated).51 ,52 If there is uncertainty about the diagnosis of NAFLD in the context of positive autoantibodies, liver biopsy may be performed.

Liver biopsy allows direct assessment of hepatic triacyglycerol (TAG) accumulation, hepatocellular injury, inflammation and fibrosis. The widely adopted histological grading and staging system for NAFLD proposed by Kleiner et al records a ‘NAFLD activity score’ (NAS) based on the combined assessment of steatosis, inflammation and ballooning hepatocyte degeneration (table 1).53 The utility of this approach as a prognostic tool in routine clinical practice has proved controversial.12 ,54 More recently, the SAF score was proposed by Bedossa et al.55 This classifies biopsies into normal, steatosis or NASH based on the evaluation of steatosis (S), activity (A) and fibrosis (F) (table 1).

In contrast with NAS, the SAF score separates measurement of steatosis from markers of cellular injury (ballooning and lobular inflammation) and so may better discriminate NASH from more benign liver histology. However, its utility as a prognostic tool requires validation. The invasive nature of liver biopsy is a major limitation, making it impractical for widespread use as a prognostic tool. Therefore, a pragmatic approach to diagnosis and disease staging using non-invasive strategies is required, reserving liver biopsy for those patients where it offers clinically relevant additional information.15

When assessing subjects with suspected NAFLD there are three goals: (1) establishing a diagnosis of NAFLD; (2) discriminating simple steatosis from NASH; and (3) determining stage of fibrosis. This article will review current knowledge of non-invasive assessment of NAFLD and suggest a practical approach to diagnosis and risk stratification.

The ideal non-invasive test should provide similar information to a biopsy, detecting steatosis, measuring hepatocellular injury/inflammation and staging fibrosis. No one non-invasive test is able to do this. Numerous tests, each addressing specific histological characteristics, have been proposed, however, with some exceptions,56 ,57 few have been compared head to head. A further complication when assessing the fidelity of non-invasive markers is that hepatic parenchymal involvement may be patchy, causing biopsy sampling error.58 Therefore, the histological ‘gold standard’ may itself be flawed. Performance of markers with an area under the receiver operating characteristic curve (AUROC) >0.85 is generally considered acceptable.

Detection of steatosis

It is questionable whether knowledge of the degree of steatosis beyond the qualitative assessment required for diagnosing NAFLD is clinically relevant. There is little evidence to suggest that the amount of steatosis per se influences prognosis and it has not been shown that therapies specifically targeted at reducing hepatic TAG accumulation offer any prognostic benefit. Indeed, TAG accumulation is itself unlikely to be harmful, as it is driven by an adaptive response to hepatocyte stress through which potentially lipotoxic fatty acids are partitioned into relatively inert intracellular stores.1 ,59

Standard imaging modality based assessments of steatosis

Although, as will be discussed below, several blood based tests for steatosis have been proposed (table 2), these are of limited accuracy and not widely employed. The most clinically relevant non-invasive techniques are imaging based (summarised in table 3).

  • Ultrasound. Ultrasound is readily available and cost effective. It is widely adopted to detect steatosis, which causes increased hepatic echogenicity.72 ,73 However, ultrasound is subjective,66 not quantitative and sensitivity is limited when <33% of hepatocytes are steatotic.74 Hepatic fat content tends to diminish as cirrhosis develops and so may fall below the detection threshold in advanced NAFLD. Therefore, NASH is thought to cause 30–75% of cryptogenic cirrhosis.16 ,75 It is sometimes difficult to discriminate steatosis from mild/moderate fibrosis as these have similar sonographic appearances.72 ,73

  • CT. Relative differences in hepatic and splenic CT attenuation (Hounsfield units) may be used to quantify hepatic TAG accumulation (table 3). A number of different parameters have been proposed. The liver/spleen attenuation ratio correlates well with steatosis and has an acceptable sensitivity and specificity, although CT is most accurate when there is >30% steatosis.74

  • MRI and proton MR spectroscopy (MRI/ 1 H-MRS). These are the most accurate non-invasive measures of steatosis.66 1H-MRS exploits the differences in resonance frequency of protons in water and fat to generate a spectrum displaying signal peaks from water and fat. Due to its excellent reproducibility, MRS has been used to assess the prevalence of steatosis in population studies,45 ,76 and to monitor changes in TAG with weight loss.77 1H-MRS samples a greater portion of the liver than liver biopsy (27 g vs 75 mg), reducing sampling error.76 ,78 Hepatic TAG content can also be calculated using routine MRI sequences.79 Multi-echo MRI techniques measure fat content, comparing the signal phase and relaxation time from each component (fat and water) to calculate the relative abundance.80 Several different techniques have been proposed, discussion of the technical aspects of which falls outside the scope of this article. Studies have shown a close relationship between steatosis measured by these methods and histological grading81 but this has not been uniformly demonstrated.82

Table 2

Steatosis tests

Table 3

Radiological measures of hepatic triacylglycerol accumulation

Controlled attenuation parameter

As will be discussed below, Fibroscan is a proprietary ultrasound based modality that measures liver stiffness as a surrogate for hepatic fibrosis using transient elastography. Controlled attenuation parameter (CAP) is a technique developed to allow simultaneous measurement of steatosis (table 3). An initial study showed a significant relationship between the degree of steatosis and CAP (r=0.81).70 Subsequently, a study of 615 hepatitis C patients found CAP accurately detected steatosis independent of fibrosis (AUROC values were 0.80 for mild, 0.86 for moderate and 0.88 for severe steatosis).70 However, CAP has been studied in relatively few patients with biopsy proven NAFLD. Myers et al 69 examined 153 overweight patients with chronic liver disease (47% had NAFLD) reporting lower AUROC values (0.79, 0.76 and 0.70, respectively). Further histological correlation in a large scale study is needed to establish diagnostic thresholds in NAFLD.

Summary of efficacy

MRI and MRS are the most accurate non-invasive measures of steatosis but require specific expertise and are resource intensive, limiting their use to research studies. Ultrasound, although only qualitative, is acceptable to patients, cost effective and has reasonable sensitivity for >33% steatosis, making it a firstline investigation. CAP promises to be an attractive method if its efficacy is proven but currently remains experimental.

Discriminating steatosis from steatohepatitis

Obesity and insulin resistance promote increased hepatic free fatty acid flux that drives liver injury. A number of hepatocellular insults contribute to the transition from steatosis to NASH, including (1) lipotoxicity; (2) oxidative stress secondary to free radicals produced during β- and ω-free fatty acid oxidation; (3) endoplasmic reticulum stress; (4) endotoxin/TLR4 induced Kupffer cell activation; (5) cytokine release; and (6) immune system activation. These promote necrotic and apoptotic cell death,83–85 stellate cell activation, collagen deposition and hepatic fibrosis.86 Aspects of these processes have been assessed as markers of NASH. Table 4 summarises some of the more promising tests and the online supplementary table 1 lists a number of additional indices that have been studied.

Table 4

Steatohepatitis tests

Markers of insulin resistance and oxidative stress

HOMA-IR, a simple measure of insulin resistance, has been included in models to predict a diagnosis of NASH,90 but alone has no role in the diagnosis of NASH. Development of NASH is associated with increased hepatocyte oxidative stress.25 ,101 Markers of lipid perioxidation, including oxidised low density lipoprotein and thiobarbituric acid reacting substances can be detected in NASH102 but lack specificity for NASH in clinical practice.


NAFLD is associated with low grade systemic inflammation characterised by elevated circulating proinflammatory cytokines, including tumour necrosis factor α (TNFα), interleukin 6 (IL-6) and interleukin 8 (IL-8), as well as acute phase reactants such as C reactive protein and ferritin.103 ,104 Adiponectin (an anti-inflammatory adipokine) is secreted by adipose tissue and has been found to be negatively correlated with adiposity, insulin resistance and the metabolic syndrome.105

Jarrar et al 103 found that TNFα levels were higher, and levels of IL-8, visfatin and adiponectin were lower, in NASH than in simple steatosis. After multiple test adjustments, only differences in TNFα and adiponectin remained significant (p<0.003). Multivariate regression analysis showed that only TNFα was an independent predictor of histological fibrosis and disease progression in patients with NASH (p<0.0004) although its diagnostic ability is limited. The role of IL-6 in the pathogenesis of NASH is less clear with conflicting data in studies of patients with NAFLD.103 ,106

Low levels of adiponectin are seen in patients with NASH compared with subjects with simple steatosis,107 and hypoadiponectinaemia seems to have a role in the accumulation of hepatic fat, insulin resistance and development of NASH.108–110 Other studies have also noted a negative correlation between adiponectin levels and the severity of NASH.111 ,112

Leptin is also secreted by adipose tissue and is important in the regulation of food intake. The role of leptin in NAFLD remains unclear. Haukeland et al 113 found higher serum leptin levels in simple steatosis compared with NASH, whereas Lemoine et al 107 found higher levels in patients with NASH.

Further work is needed to establish the role of adipokines in the development and pathogenesis of NAFLD. Although cytokine and adipokine disturbance is seen in subjects with NASH, their changes lack specificity to be useful as clinical tests for NASH. It is possible that cytokines and adipokines could form part of a panel of biomarkers for NASH, but further study is needed.

Apoptosis and CK-18

Increased apoptosis, triggered through upregulation of Fas ligand and Fas receptor activation, is a prominent feature of NASH85 ,114–116 that has also been examined as a potential therapeutic target.84 ,117 Serum markers of apoptosis may help distinguish NASH from simple steatosis.115 Fragments of caspase cleaved cytokeratin-18 (CK-18), a major hepatocyte intermediate filament protein, are detectable in the bloodstream by ELISA. The M30 monoclonal antibody ELISA detects a neoepitope created by caspase cleavage and so is apoptosis specific118 while the M65 antibody detects both cleaved and intact CK-18 released during necrotic and apoptotic cell death.119 Most studies have used the M30 antibody, providing encouraging results.120 ,116 In the largest single study, Feldstein et al showed that raised plasma CK-18 fragments was an independent predictor of NASH.98 Both M30 and M65 predicted disease progression. This assay might be more accurate in differentiating NASH and simple steatosis, but further validation is needed.119 ,95

The combination of plasma CK-18 fragments with other serum markers of apoptosis might improve the diagnostic accuracy of tests for NASH. Combining plasma CK-18 fragments (M30) with soluble Fas improves the AUROC for diagnosing NASH.99 In addition, a model combining plasma CK-18 levels (cleaved and intact) with levels of adipokines (serum adiponectin and serum resistin) enabled the prediction of histological NASH with high sensitivity.95

Commercial panels

  • The NASH Test is a patented algorithm that predicts NASH with modest accuracy (table 4).92

  • The Palekar Index88 combines clinical characteristics with a panel of biomarkers of fibrosis. The presence of three or more risk factors gave a sensitivity of 73.7% and specificity of 65.7%.

  • Shimada et al 90 examined serum adiponectin levels, HOMA-IR and serum type IV collagen 7S levels to try and distinguish simple steatosis from NASH. Combining the three markers gave a sensitivity of 94% and specificity of 74%.

Imaging based modalities

No routine imaging based modality can reliably distinguish simple steatosis from NASH.74 The use of contrast enhanced ultrasound is an area of ongoing research. A small study demonstrated that the accumulation of Levovist microbubbles (between 5 and 20 min) was reduced in patients with NASH compared with NAFLD and healthy volunteers. Levovist is an ultrasound contrast agent which is injected intravenously in this context. Using this technique, the AUROC for a diagnosis of NASH was impressive at 1.0 when the cut-off value was set at 43.6 at 20 min.121 However, changes in signal intensity were not correlated with the histological degree of fibrosis and steatosis. This technique needs to be validated in a large cohort of patients.

Summary of efficacy

Overall, CK-18 fragments appear to be the most effective, although imperfect, single non-invasive test to distinguish NASH from simple steatosis. The role of changes in plasma CK-18 levels in the monitoring of disease following intervention has not been defined. Therefore, it is likely that a good non-invasive test for NASH will involve a panel of tests, probably including CK-18 levels.

Quantifying hepatic fibrosis

Repeated injury to the liver results in a wound healing response and ultimately leads to hepatic fibrosis. Due to persisting injurious factors, liver regeneration eventually fails and hepatocytes are replaced by extracellular matrix (ECM) composed of collagens (I, III, and IV), fibronectin, undulin, elastin, laminin, hyaluronan and proteoglycans.60 ,70 Serological markers for the evaluation of liver fibrosis can be divided into ‘indirect’ markers (that reflect alterations in hepatic function but not collagen turnover, eg, platelet levels) and ‘direct’ markers (associated with ECM deposition and turnover).122

Indirect markers: routine blood tests and ‘simple panels’

Significant hepatic fibrosis can lead to hepatocellular dysfunction and portal hypertension, which are reflected by changes in standard biochemical and haematological parameters. These tests, alone or combined as ‘simple panels’, are potentially attractive clinical tools as they are inexpensive and many indices are already routinely measured in patients with liver disease (table 5). Selected tests will be discussed further below, and the results of our recently reported head to head comparison in a large cohort with biopsy proven NAFLD are shown in table 6.57

  • With advancing fibrosis, serum ALT falls whereas AST remains stable or increases. Hence, the AST/ALT ratio (AAR) increases.133 Sheth et al found that an AAR >1 predicted hepatitis C related cirrhosis with 53.2% sensitivity and 100% specificity (positive predictive value (PPV) 100%, negative predictive value (NPV) 80.7%).123 However, a similar study reported a more modest PPV of 64% (95% CI 48 to 78%)134 although the latter did contain a proportion of high alcohol consuming patients which would be expected to raise AST levels. Our own study found that an AAR<0.8 had an excellent NPV (table 6), reliably excluding advanced fibrosis.57

  • The AST to Platelet Ratio Index was derived as a non-invasive test for advanced fibrosis/cirrhosis in hepatitis C135 ,136 but studies in NAFLD have proved disappointing.57 ,124 ,129

  • The BARD score, combining AAR, BMI and the presence of T2DM, was developed and validated in 827 patients with biopsy proven NAFLD fibrosis.127 While the NPV for advanced fibrosis with a score of <2 is impressive at 96%, the majority of NAFLD patients will score above this threshold due to obesity and diabetes, resulting in a poor PPV and limiting utility in practice.

  • The NAFLD Fibrosis Score, based on analysis of a 733 patient discovery set, is calculated using six routinely measured parameters found to be independently associated with advanced fibrosis on multivariate analysis (table 5). By applying a low cut-off (<−1.455), advanced fibrosis can be excluded with high accuracy (NPV 93%) while a high cut-off threshold (>0.676) offers accurate detection of advanced fibrosis (PPV 90%).131 Use of this score has been suggested to reduce the need for liver biopsy by ∼75%. This score has been independently validated in several other studies,129 ,137 ,138 and although the formula appears daunting, calculation can be performed using a simple online calculator (

  • The FIB-4 Score was originally derived in a cohort of patients with hepatitis C and HIV coinfection,139 but appears to be one of the best performing simple non-invasive tests for advanced fibrosis in NAFLD. In a study of 541 patients with biopsy proven NAFLD, the FIB-4 score performed better than several other simple scores for fibrosis in diagnosing/excluding stage 3–4 fibrosis with an AUROC of 0.80. A score of <1.3 had a 90% NPV for stage 3–4 fibrosis while a score of >2.67 had an 80% PPV with only a quarter of the cohort being unclassified 1.3 or above 2.67.128 Other studies have also found that the FIB-4 score narrowly out performs other simple non-invasive tests in predicting advanced fibrosis.57 ,120 ,140

Table 5

Fibrosis testsγ

Table 6

Head to head comparison of simple panels used to identify patients with fibrosis

In general, simple panels have relatively robust NPV and so can reliably exclude advanced fibrosis but have poor PPV (ranging from 27% to 79%).57 Therefore, these tests are unreliable at diagnosing advanced fibrosis but do have the potential to help with the triage of patients by reliably excluding advanced fibrosis. Using such tests may help to mitigate against the healthcare burden that such a large ‘at risk’ population places on resources by allowing those with ‘low risk’ scores to undergo further investigation. AAR, FIB-4 Score and NAFLD Fibrosis Score appear to be best suited to this (table 6)57 and are also effective in patients with normal range ALT levels.141

Direct markers of collagen turnover

  • Hyaluronic acid production is increased when collagen synthesis is accelerated and so is a marker of the increased ECM production seen in advanced NAFLD. Various small studies have found an AUROC of 0.75–0.97 for predicting advanced fibrosis.142–144 It is important to note that the studies used different cut-off points. Palekar et al 88 found that a cut-off for hyaluronic acid of 45.3 μg/L was a good predictor of advanced fibrosis. When hyaluronic acid and type IV collagen 7S domain were studied in a cohort of 72 patients with histologically confirmed NASH, significantly greater levels were observed with advanced fibrosis compared with those with lesser degrees. The AUROCs for hyaluronic acid and the type IV collagen 7S domain were 0.754 and 0.767, respectively, but only type IV collagen 7S domain remained significant after multivariate analysis adjusted for age, sex, platelet count, prothrombin time, AAR, BMI and T2DM.143

  • Liver fibrosis results in the deposition of collagen and release of propeptides, predominantly procollagen III. The terminal peptide of procollagen III correlates with NAS and its constituent components (p<0.001). The AUROC for discriminating between NASH and simple steatosis was 0.82–0.84 in patients with F0–3 fibrosis. A threshold of 6.6 ng/mL gave an NPV for advanced fibrosis of 95%–97% and 100% for cirrhosis.89

  • The Enhanced Liver Fibrosis test is a commercial panel of markers focusing on matrix turnover.132 It comprises tissue inhibitor of matrix metalloproteinase 1, hyaluronic acid and the aminoterminal peptide of procollagen III. When compared with the NAFLD Fibrosis Score, this test performed only marginally better for severe fibrosis (AUROC 0.93 vs 0.89) and moderate fibrosis (AUROC 0.90 vs 0.86), but combining the two enhanced efficacy (AUROC 0.98 for severe fibrosis and 0.93 for moderate fibrosis).56

  • Fibrotest is a commercial panel of biochemical markers of liver fibrosis which was initially validated in patients with hepatitis C. Ratziu et al have investigated its use in NAFLD.130 The AUROC for F2–F4 was 0.75–0.86 and for F3–F4 was 0.81–0.92. A meta-analysis showed the AUROC for advanced fibrosis was 0.84 without differences between the causes of liver disease.145

Imaging based modalities

Routine imaging modalities such as ultrasonography, CT and MRI are accurate in the diagnosis of advanced fibrosis and cirrhosis if features such as a nodular appearing hepatic parenchyma, enlarged caudate lobe or signs of portal hypertension are present. These features are usually absent in subjects with milder degrees of fibrosis and early cirrhosis, and so these modalities have no place in the staging of hepatic fibrosis. This is a rapidly advancing field, and a number of advanced elastography techniques measuring liver stiffness (hepatic parenchymal fibrosis reduces tissue elasticity) as a surrogate for hepatic fibrosis are more promising.

  • MR elastography shows potential for assessing liver fibrosis.146 One study has shown high accuracy (AUROC=0.93) for discriminating steatosis from NASH, with a sensitivity of 94% and a specificity 73% using a threshold of 2.74 kPa.147 Double enhanced liver MRI using two contrast agents had an accuracy of >90% for the diagnosis of advanced fibrosis (METAVIR ≥F3) in 101 patients with chronic liver disease.148 Another study reported that diffusion weighted MRI had similar diagnostic accuracy as ultrasound transient elastography, Fibrotest, AST to Platelet Ratio Index, Forn's Index and hyaluronic acid in the assessment of fibrosis in patients with chronic hepatitis C virus.149

  • Fibroscan employs ultrasound transient elastography (TE) to measure hepatic elasticity by quantifying the shear wave velocity with pulse echo ultrasound from low frequency vibrations that are transmitted into the liver.150 The resulting liver stiffness measurement correlates well with the degree of hepatic fibrosis in a range of liver diseases.150 ,151 In NAFLD, TE has an AUROC of 0.84 and 0.93 for the detection of ≥F2 and ≥F3 fibrosis, respectively, and performed better than a number of simple non-invasive scores for fibrosis.152 A cut-off of 7.9 kPa gives a high NPV of 0.96 for ≥F3 fibrosis but modest PPV (0.52). The optimum cut-offs for the use of TE in clinical practise is not known, but a low liver stiffness measurement appears to reliably exclude advanced fibrosis.

  • However, TE is operator dependent and may not give valid results in patients with central obesity, particularly if BMI >35 kg/m2 (OR 7.5, 95% CI 5.6 to 10.2; p=0.0001),153 age >52 years (OR 2.3, 95% CI 1.6 to 3.2; p=0.0001) and T2DM (OR 1.6, 95% CI 1.1 to 2.2; p=0.009). Wong et al found that the overall success rate of TE in patients with BMI ≥30 kg/m2 was 67.0% and in those with BMI ≥35 kg/m2 41% failed liver stiffness measurement acquisition.152 Clearly this is a significant limitation in a NAFLD population who are often older with central obesity, a high BMI and T2DM. The Fibroscan XL probe has been developed for use in obese patients which results in fewer test failures (1.1% vs 16%) and is more reliable (73% vs 50%; both p<0.00005). The AUROC of the XL and M probes were similar for ≥F2 fibrosis (0.83 vs 0.86; p=0.19) and cirrhosis (0.94 vs 0.91; p=0.28). Lower liver stiffness cut-offs are necessary with the XL probe.154 However, significant discordance between ≥2 stages in fibrosis on liver biopsy and TE still occurs in more than 10% of patients with BMI >28 kg/m2 with the XL probe.155

  • ARFI uses conventional B mode ultrasonography to generate an initial ultrasonic pulse to obtain a baseline signal and a subsequent high intensity ‘pushing pulse’. The response of the liver tissue (displacement) to the radiation force is quantified as shear wave velocity.156 ,157 There is an increase in the median velocity measured by ARFI with increasing severity of hepatic fibrosis. Yoneda et al 158 found that the optimal median ARFI velocity for the diagnosis of ≥F3 fibrosis was 1.77 m/s (AUROC 0.973, sensitivity 1.00, specificity 0.91). The median velocity in patients with simple steatosis was lower than that in healthy volunteers. This may make it difficult to distinguish between simple steatosis and NASH with mild fibrosis using ARFI. ARFI can be performed during standard ultrasound examination of the liver which, as discussed earlier, should form part of the routine evaluation of any patient with liver disease.

Summary of efficacy

The NAFLD Fibrosis Score and FIB-4 score are particularly suitable to use in clinical practice as a first stage triage of NAFLD patients. Some of the more advanced panels may offer a marginal improvement in NPV/PPV but they require additional assays and are more expensive to introduce. The Enhanced Liver Fibrosis panel is one of the few to be directly compared with the NAFLD Fibrosis Score, but performed only marginally better (AUROC 0.93 vs 0.89).56 TE offers a different approach but there remains some concern about inaccuracy, particularly in the older obese population that are at greatest risk of advanced NAFLD. Further studies are needed to clarify this.

A unified pragmatic approach

A large proportion of the population is at risk of NAFLD through being overweight and/or insulin resistant. Currently, there is very limited guidance for general practitioners regarding which patients with NAFLD to refer on for further evaluation. Reliance on elevations of ALT >1.5–2 times the upper limit of normal is grossly inaccurate and so other strategies are urgently needed to allow triage of patients.15

A number of different algorithms incorporating combinations of cell death and fibrosis markers with or without elastography have been proposed159 but prospective validation is lacking. Thus a pragmatic approach to risk stratify patients is all that can be proposed.

The first step is to identify those people who are at risk of NAFLD: patients with features of the metabolic syndrome (hypertension, central obesity, insulin resistance, dyslipidaemia), abnormal LFTs and fatty liver on imaging. All patients should undergo abdominal ultrasound (if not already performed) and have other liver diseases excluded.3

The next step is to identify those with advanced fibrosis or cirrhosis. This may be clear from liver synthetic dysfunction, routine imaging (nodular liver, portal hypertension) or high TE scores. ‘Simple panels’ (AAR, FIB-4, NAFLD Fibrosis Score) are inexpensive, use routine indices and have good NPV to reliably exclude advanced fibrosis.

The final step is to identify those with NASH who are at risk of progressive disease and require aggressive risk factor modification. CK-18, a marker of apoptosis, is currently the most useful non-invasive test to identify patients with steatohepatitis.

Musso et al have suggested three management algorithms for NAFLD. All patients should be assessed using the NAFLD Fibrosis Score and those at high risk should have a liver biopsy. In obese patients (algorithm 1), those at low or intermediate risk can be further risk stratified using CK-18. For non-obese patients, algorithm 2 uses CK-18 for low risk patients and Fibroscan with or without CK-18 to further delineate the intermediate risk group. Algorithm 3 uses Fibroscan for all low and intermediate risk patients with CK-18 to further risk stratify those with low Fibroscan scores. With these approaches, 85%, 88% and 90% of patients with NASH would be correctly identified, and 22%, 23% and 26% of patients with simple steatosis would be unnecessarily biopsied (for algorithms 1, 2 and 3, respectively).159

An example management algorithm is shown in figure 1. Low risk patients can be managed in primary care with lifestyle advice and risk factor modification. High risk patients require referral and liver biopsy to establish the stage of disease. CK-18 levels are not routinely available in all centres so patients at intermediate and high risk have to be managed according to the high risk arm of the algorithm. If the algorithm cannot be effectively implemented or if clinical suspicion remains high, onward referral is advised.

Figure 1

Example of a logarithm for clinical assessment of patients at risk of non-alcoholic fatty liver disease (NAFLD). The algorithm provides an example of a pragmatic approach to the diagnosis and assessment of NAFLD, accepting the many uncertainties that currently exist in this field.15 159 CK-18, cytokeratin-18; HCC, hepatocellular carcinoma; IR, insulin resistance; LFTs, liver function tests; liver screen, screening for viral, autoimmune and metabolic causes of abnormal LFTs; MS, metabolic syndrome; NASH, non-alcoholic steatohepatitis; USS, ultrasound.


NAFLD is now the commonest cause of liver disease, affecting up to a third of the population in developed countries. Many patients remain undiagnosed. The keys to managing NAFLD are risk stratification and risk factor modification for liver and cardiovascular disease. Identifying patients at high risk of disease progression allows care to be focussed on this group.

Interactive multiple choice questions

This JCP review article has an accompanying set of multiple choice questions (MCQs). To access the questions, click on BMJ Learning: Take this module on BMJ Learning from the content box at the top right and bottom left of the online article. For more information please go to: Please note: The MCQs are hosted on BMJ Learning—the best available learning website for medical professionals from the BMJ Group. If prompted, subscribers must sign into JCP with their journal's username and password. All users must also complete a one-time registration on BMJ Learning and subsequently log in (with a BMJ Learning username and password) on every visit.


QMA is the recipient of a Clinical Senior Lectureship Award from the Higher Education Funding Council for England (HEFCE).


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors All authors have contributed to the writing of this manuscript.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.