Diagnostic challenges of nonalcoholic fatty liver disease/nonalcoholic steatohepatitis

Watch a video presentation of this article

Watch the interview with the author

Abbreviations

AASLD

American Association for the Study of Liver Diseases

ALT

alanine aminotransferase

APRI

AST‐to‐platelet ratio index

ARFI

acoustic radiation force impulse

AST

aspartate aminotransferase

BARD

BMI, AST:ALT ratio, and diabetes status score

BMI

body mass index

CAP

controlled attenuation parameter

CK‐18

cytokeratin‐18

CT

computed tomography

EASL

European Association for the Study of the Liver

EASO

European Association for the Study of Obesity

ELF

enhanced liver fibrosis

FIB‐4

Fibrosis‐4

GGT

gamma‐glutamyltransferase

HAIR

hypertension, age, insulin resistance

MR

magnetic resonance

MRE

magnetic resonance elastography

MRI‐PDFF

magnetic resonance‐proton density fat fraction

NAFL

nonalcoholic fatty liver

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

NPV

negative predictive value

PPV

positive predictive value

TE

transient elastography

VCTE

vibration‐controlled transient elastography

Nonalcoholic fatty liver disease (NAFLD) is a global epidemic that ranges from isolated hepatic steatosis (nonalcoholic fatty liver [NAFL]) to steatosis plus inflammation (nonalcoholic steatohepatitis [NASH]) with or without fibrosis (Fig. 1).1 Whereas NAFL generally follows a benign course, NASH carries a significant risk for progression to fibrosis.2 The key diagnostic challenges in NAFLD are to accurately detect NASH and to quantify the degree of fibrosis to identify those at highest risk for liver‐related morbidity and mortality. Thus, when seeing a patient with possible NAFLD, the primary questions to answer are: (1) Does this patient have NAFLD? (2) Does this patient have underlying NASH? (3) Does this patient have any fibrosis? and (4) Does this patient have advanced fibrosis (stage 3 or 4)?

Figure 1.

Serum biomarkers and imaging modalities across the NAFLD spectrum.

Abbreviations: APRI, AST‐to‐platelet ratio index; ARFI, acoustic radiation force impulse; BARD, BMI, AST:ALT ratio, and diabetes status score; CK‐18, cytokeratin‐18; HAIR, hypertension, age, insulin resistance.

The Role of Liver Biopsy

Liver biopsy remains the gold standard for diagnosing NAFLD; however, its widespread use is limited by the risk associated with an invasive procedure, cost, and sampling error.3 Thus, noninvasive diagnostic modalities allow for risk stratification of patients with NAFLD to select those who would benefit most from liver biopsy, while potentially avoiding this invasive procedure in others.

Does This Patient Have Nonalcoholic Fatty Liver Disease?

Not all hepatic steatosis, defined as fat >5% to 10% of the liver parenchyma, is NAFLD.4 It is important to rule out other causes of hepatic steatosis, particularly alcohol (Table 1). NAFLD is typically associated with the features of the metabolic syndrome, which includes central adiposity, hypertension, dyslipidemia, and insulin resistance.4, 5 Thus, presence of one or more metabolic risk factors should raise clinical suspicion for NAFLD.

Table 1.

Causes of Hepatic Steatosis

Macrovesicular	Microvesicular
NAFLD Alcoholic liver disease Hepatitis C, genotype 3 Medications  Amiodarone  Corticosteroids  Methotrexate  Tamoxifen Wilson's disease Hemochromatosis Starvation Parenteral nutrition Lipodystrophy Abetalipoproteinemia	Reye's syndrome Acute fatty liver of pregnancy HELLP syndrome Medications  Antiretroviral medications  Valproate Inherited metabolic disorders  Lysosomal acid lipase deficiency  Lecithin‐cholesterol acyltransferase deficiency

Adapted with permission from Hepatology.4 Copyright 2018, American Association for the Study of Liver Diseases.

Hepatic steatosis is commonly detected incidentally on imaging such as ultrasound or computed tomography (CT). Notably, these modalities have poor sensitivity, detect fat only when 20% to 33% of the liver parenchyma is involved, and cannot accurately quantify the amount of hepatic fat present. Newer modalities such as magnetic resonance (MR) imaging–based spectroscopy, MR‐proton density fat fraction (MR‐PDFF), and transient elastography (TE)‐based controlled attenuation parameter (CAP) are more sensitive and allow for relatively accurate quantification of hepatic steatosis (Table 2).6 However, each of these imaging modalities has strengths and limitations that must be considered before implementation in the general population (Table 3).4 There are also several panels that have been proposed to diagnose hepatic steatosis, many of which have been used in population‐based studies aimed at estimating the epidemiology and natural history of NAFLD (Table 2).6

Table 2.

Noninvasive Methods for Detecting Hepatic Steatosis and Associated Test Characteristics

Test	Components	Low Cutoff	High Cutoff	Sensitivity/Specificity, %	PPV/NPV, %
SteatoTest	Age, sex, BMI, fasting glucose, cholesterol, triglycerides, ALT, bilirubin, GGT, haptoglobin, α2‐macroglobulin, apolipoprotein A1	<0.30	>0.72	90/90	63/93
NAFLD liver fat score	Metabolic syndrome, diabetes, fasting insulin, AST, AST/ALT	<−1.413	>1.257	95/95
Fatty liver index	BMI, waist circumference, triglycerides, GGT	<30	>60	87/86a
Ultrasonography				66‐94/66‐97
Ultrasound Fatty Liver Indicator (US‐FLI)	Liver brighter than kidney; liver brightness graded as mild/moderate (2 points) or severe (3 points). One extra point for each of the following: 1) Posterior attenuation of ultrasound beam, 2) Vessel blurring; 3) difficult visualization of gallbladder wall, 4) difficult visualization of diaphragm, 5) areas of focal sparing		Score ≥ 2	46/unknown	unknown/94
Unenhanced CT scan	Three available measures: liver parenchyma attenuation, liver to spleen attenuation difference, and liver to spleen attenuation ratio			85/100
TE‐CAP			>261	72/86	98/23
MRI‐PDFF			>3.71	96/100	100/70

Abbreviations: GGT, gamma‐glutamyltransferase; NPV, negative predictive value; PPV, positive predictive value.

predictor of absence of significant hepatic steatosis.

⁺predictor of presence of significant hepatic steatosis.

^aComparator group: liver ultrasound.

Data are from Machado and Cortez‐Pinto,17 Hernaez et al.,18 and Chen et al.12

Table 3.

Strengths and Limitations of Commonly Used Imaging Modalities in Nonalcoholic Fatty Liver Disease

Imaging Modality	Used to Assess	Strengths	Limitations
Ultrasound	Steatosis	Good for detection of moderate‐to‐severe steatosis Widely available Low cost Safe	Poor sensitivity and negative predictive value Unable to detect mild steatosis Not quantitative Fibrosis and steatosis have similar appearance Operator dependent Accuracy influenced by BMI
CT	Steatosis	Good for detection of moderate‐to‐severe steatosis Better specificity than ultrasound Provides additional anatomic information	Poor sensitivity Unable to detect mild steatosis Ionizing radiation exposure Limited by variable amounts of iron
MR imaging	Steatosis	Better sensitivity and specificity than ultrasound	Limited by high iron burden
CAP	Steatosis	Quantitative More sensitive than conventional ultrasound	Limited clinical experience
MR Spectroscopy	Steatosis	Quantitative Sensitive	Limited availability High cost Less accurate with nonhomogeneous fat distribution
TE	Fibrosis	Correlates with stage of fibrosis Point‐of‐care test	Accuracy reduced in obesity Severe steatosis may lead to false positives Operator dependent Accuracy influenced by BMI
Acoustic radiation force impulse	Fibrosis	Similar sensitivity/specificity as TE	Higher failure rates than TE Operator dependent Accuracy influenced by BMI
MR elastography	Fibrosis	Most accurate test for determining fibrosis stage Accuracy not affected by BMI, degree of steatosis	Limited availability High cost

Adapted with permission from Journal of Hepatology.5 Copyright 2016, European Association for the Study of the Liver.

Does This Patient Have Underlying Nonalcoholic Steatohepatitis?

Reliable noninvasive methods to detect NASH remain limited. Commonly investigated methods can be grouped into two broad categories: serum biomarkers and predictive models. Serum aminotransferases, which are often used in clinical practice as a surrogate for inflammation, have poor predictive value for NASH.4, 5 Serum alanine aminotransferase (ALT) greater than two times the upper limit of normal (>70 U/L) has only 50% sensitivity and 61% specificity for NASH.7 In addition, patients with NAFLD can have normal ALT levels, particularly as the disease progresses.4 Therefore, although elevated aminotransferases should raise suspicion for NASH, normal levels should not be used to exclude NASH.4, 5

Although serum biomarkers are not currently available for clinical use, many are under investigation (Table 4). These biomarkers broadly reflect the pathways involved in NASH development, including hepatocyte apoptosis, oxidative stress, and inflammation. Several diagnostic panels, such as the NashTest,8 use a combination of biomarkers and clinical factors to predict NASH (Table 4). Given the lack of reliable noninvasive tests, physicians must use clinical factors to risk‐stratify patients prior to liver biopsy. The presence of one or more features of the metabolic syndrome in a patient with NAFLD warrants referral to a specialist for consideration of liver biopsy.4, 5

Table 4.

Biomarkers and Complex Scores for Detecting Steatohepatitis

Test	Components	Cutoff	Sensitivity, %	Specificity, %	PPV, %	NPV, %
ALT7		>70 53‐71	50 72.2‐50	60.7 50.6‐60.7
CK‐1814		>216 >287	77 65	65 92
NashTest	Age, sex, height, weight, cholesterol, triglycerides, AST, ALT, bilirubin, haptoglobin α2‐macroglobulin, apolipoprotein A1	Undisclosed	33	94	66	81
NASH diagnostics	CK‐18, adiponectin, resistin	>0.2772 >0.3499	95 77	70 87	60 74	97 89
HAIR	Hypertension, insulin resistance, elevated ALT (>40)	≥2 parameters	80	89
The Nice Model	Metabolic syndrome, ALT, CK‐18	>0.14 >0.83	84 16	86 99	44 90	98 91
oxNASH15  Ballooning  Inflammation	Age, BMI, AST, 13‐hydroxyl‐octadecadenoic acids, linoleic acid	>55.2 >54.6	79 78	65 67	67 72	77 74
HepQuant STAT16	Serum concentration of tetra‐deuterated cholic acid 60 minutes after oral administration	>0.50 μMol	94	76	71	95

Abbreviations: HAIR, hypertension, increased ALT and insulin resistance; NPV, negative predictive value; PPV, positive predictive value.

Multiple cutoffs have been studied for several of these tests. These are shown in separate rows with their corresponding test characteristics.

Data are from Machado and Cortez‐Pinto.17

Does This Patient Have Any Fibrosis?

Hepatic fibrosis is the primary predictor of liver‐related mortality in NAFLD. Furthermore, this relationship is stage dependent, and higher fibrosis stage is associated with higher liver‐related mortality.9 Identifying patients with early‐stage fibrosis is key to implementing risk‐reduction strategies to prevent disease progression. Unfortunately, most diagnostic tests currently in use are best suited to detect advanced fibrosis. Even the most accurate imaging studies available are relatively insensitive for stage 1 fibrosis (Table 5).

Table 5.

Transient Elastography versus Magnetic Resonance Elastography for the Diagnosis of Fibrosis in NAFLD/NASH

	TE					MRE
Fibrosis Stage	Cutoff (kPa)	Sensitivity, %	Specificity, %	PPV, %	NPV, %	Cutoff (kPa)	Sensitivity, %	Specificity, %	PPV, %	NPV, %
≥1 versus 0	6.10	67	65	69	62	2.65	77	79	81	74
≥2 versus 0‐1	6.90	79	85	70	90	2.86	79	82	66	90
≥3 versus 0‐2	7.30	78	78	45	94	2.99	78	80	48	94
≥4 versus 0‐3	6.90	63	66	15	95	3.35	75	81	27	97

Abbreviations: NPV, negative predictive value; PPV, positive predictive value.

Data are from Park et al.13

Does This Patient Have Advanced Fibrosis (Stage 3 or 4)?

Several clinical prediction rules, serum biomarkers, and imaging techniques are available to detect advanced hepatic fibrosis (Table 6). Clinical prediction rules, including the NAFLD Fibrosis Score, the Fibrosis‐4 (FIB‐4) index, and the aspartate aminotransferase (AST):ALT ratio, have the advantage of using readily available, cost‐effective laboratory tests and have recently been shown to correlate with mortality in NAFLD.10 These scoring systems are best suited to rule out the presence of advanced fibrosis with negative predictive values >90%.6 Comparatively, the positive predictive value is modest, ranging from 55% to 79%.6 Thus, values greater than the upper cutoff require liver biopsy for confirmation of fibrosis, whereas a score less than the cutoff is likely sufficient to rule out advanced fibrosis and may reduce the need for liver biopsy by ∼75%.6

Table 6.

Complex Scores for Detecting Advanced Fibrosis (Stage ≥ 2) in Nonalcoholic Fatty Liver Disease/Nonalcoholic Steatohepatitis

Test	Components	Low Cutoff	High Cutoff	Sensitivity, %	Specificity, %	PPV, %	NPV, %	AUROC (CI)
NAFLD fibrosis score	Age, diabetes, BMI, AST, ALT, platelets, albumin	<−1.455	>0.676	78 33	58 98	30 79	92 86	0.81 (0.71‐0.91)
FIB‐4	Age, AST, platelets	<1.30	>3.25	85 26	65 98	36 75	95 85	0.86 (0.78‐0.94)
AST:ALT	AST:ALT	<0.8 1.0	>1.0	74 52	78 90	44 55	93 89	0.83 (0.74‐0.91)
BARD score	BMI, AST:ALT, diabetes	2	2	89	44	27	95	0.77 (0.68‐0.87)
APRI	AST:platelets	1	1	27	89	37	84	0.67 (0.54‐0.8)
ELF score	TIMP‐1, PIIINP, HA	0.375	>0.462	89 78	96 98	80 87	98 96	0.87 (0.67‐1.0)
FibroSurea (FibroTest)	Age, sex, bilirubin, GGT, haptoglobin, α2‐macroglobulin, apolipoprotein A1	30	70	77 15	77 98	54 73	90 76	0.81 (0.74‐0.86)

Abbreviations: APRI, AST‐to‐platelet ratio index; AUROC, area under the receiver operating characteristic; BARD, BMI, AST:ALT ratio, and diabetes status score; CI, confidence interval; GGT, gamma‐glutamyltransferase; HA, hyaluronic acid; NPV, negative predictive value; PIIINP, N‐terminal propeptide of type III procollagen; PPV, positive predictive value; TIMP‐1, tissue inhibitor of metalloproteinases 1.

^aTo provide a more comprehensive evaluation of liver injury in patients with NAFLD, NASH FibroSURE (LabCorp) combines FibroTest (for assessment of fibrosis), SteatoTest (BioPredictive; for assessment of steatosis), and NashTest (for assessment of NASH).

predictor of absence of significant hepatic fibrosis.

⁺predictor of presence of significant hepatic fibrosis.

Data are from McPherson et al.19 and Machado and Cortez‐Pinto.17

Several serum biomarkers and panels directly measure by‐products of fibrosis formation as a surrogate for hepatic fibrosis. One clinically available complex predictive model is the enhanced liver fibrosis (ELF) panel, which is composed of several individual biomarkers and has been shown to predict mortality in chronic liver disease.11 Similar to the clinical prediction rules, serum biomarkers/panels have good sensitivity to rule out advanced fibrosis, but they are less accurate at detecting early fibrosis with large ranges of indeterminate scores.

Several imaging modalities have been developed that allow for quantification of hepatic fibrosis with a higher degree of accuracy than serological tests. Both vibration‐controlled TE (VCTE) and MR elastography (MRE) use liver stiffness as a surrogate marker for fibrosis. VCTE, known commercially as FibroScan, is a point‐of‐care test that can be used in a clinic setting to predict advanced fibrosis with fairly high accuracy.6 VCTE is an appealing screening tool given its ease of use; however, its accuracy is significantly reduced in obese patients.6 MRE, whose accuracy is not as dependent on body mass index (BMI), has been shown in some studies to have better accuracy than VCTE in both obese12 and nonobese13 patients with NAFLD; however, its use is limited by high cost and limited availability.

Clinical Application

Both the American Association for the Study of Liver Diseases (AASLD) and the European Association for the Study of the Liver (EASL) have published practice guidelines that can assist clinicians in integrating noninvasive methods with clinical factors to make decisions on the utility of liver biopsy (Fig. 2).4, 5 Regardless of the noninvasive method used for risk stratification, it is important to remember that NAFLD is a dynamic disease, and thus ongoing risk assessment for liver disease progression over time is of paramount importance.

Figure 2.

Diagnostic flow chart to assess and monitor disease severity in the presence of suspected NAFLD and metabolic risk factors based on the most recent AASLD and EASL‐EASD‐European Association for the Study of Obesity (EASO) Clinical Practice Guidelines for the diagnosis and management of NAFLD.

¹See Table 1.^{4 2}See Table 6.⁵ *Risk factors for NASH include the metabolic syndrome, obesity, hypertension, dyslipidemia, and insulin resistance.

In summary, the current imperfect gold standard for the diagnosis of NAFLD/NASH is liver biopsy. A number of serum markers, imaging modalities, and clinical prediction rules are available as noninvasive alternatives to liver biopsy, but most have substantial limitations in clinical practice. To date, MRI‐PDFF and MRE seem to be the most accurate modalities for detecting hepatic steatosis and fibrosis, respectively.6 However, widespread use of these modalities is limited by cost and availability in clinical practice. TE is a more widely available tool for fibrosis assessment and offers accuracy close to that of MRE. Noninvasive detection of NASH and accurate determination of fibrosis stage remain key diagnostic challenges in need of further investigation.