Nonalcoholic fatty liver disease: From steatosis to cirrhosis


  • Geoffrey C. Farrell,

    Corresponding author
    1. The Storr Liver Unit, Westmead's Millennium Institute, University of Sydney at Westmead Hospital, Westmead, NSW, Australia
    Current affiliation:
    1. Department of Gastroenterology and Hepatology, The Canberra Hospital and the Australian National University, Canberra, ACT 2606, Australia
    • Director of Gastroenterology and Hepatology, The Canberra Hospital, PO Box 81, Woden, ACT 2606, Australia
    Search for more papers by this author
    • fax: (612) 62 81 51 79

  • Claire Z. Larter

    1. The Storr Liver Unit, Westmead's Millennium Institute, University of Sydney at Westmead Hospital, Westmead, NSW, Australia
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Nonalcoholic steatohepatitis (NASH), the lynchpin between steatosis and cirrhosis in the spectrum of nonalcoholic fatty liver disorders (NAFLD), was barely recognized in 1981. NAFLD is now present in 17% to 33% of Americans, has a worldwide distribution, and parallels the frequency of central adiposity, obesity, insulin resistance, metabolic syndrome and type 2 diabetes. NASH could be present in one third of NAFLD cases. Age, activity of steatohepatitis, and established fibrosis predispose to cirrhosis, which has a 7- to 10-year liver-related mortality of 12% to 25%. Many cases of cryptogenic cirrhosis are likely endstage NASH. While endstage NAFLD currently accounts for 4% to 10% of liver transplants, this may soon rise. Pathogenic concepts for NAFLD/NASH must account for the strong links with overnutrition and underactivity, insulin resistance, and genetic factors. Lipotoxicity, oxidative stress, cytokines, and other proinflammatory mediators may each play a role in transition of steatosis to NASH. The present “gold standard” management of NASH is modest weight reduction, particularly correction of central obesity achieved by combining dietary measures with increased physical activity. Whether achieved by “lifestyle adjustment” or anti-obesity surgery, this improves insulin resistance and reverses steatosis, hepatocellular injury, inflammation, and fibrosis. The same potential for “unwinding” fibrotic NASH is indicated by studies of the peroxisome proliferation activator receptor (PPAR)-γ agonist “glitazones,” but these agents may improve liver disease at the expense of worsening obesity. Future challenges are to approach NAFLD as a preventive public health initiative and to motivate affected persons to adopt a healthier lifestyle. (Hepatology 2006;43:S99–S112.)

Introduction: Early Insights and Contemporary Understanding

Associations between obesity, T2DM, and steatosis have long been recognized, as has the high prevalence of cirrhosis in diabetes. Cases of fatty liver disease with inflammation that resembled alcoholic steatohepatitis but occurring in nondrinkers were described 30 years ago, first in the Japanese literature and then in the United States.1, 2 Ludwig coined the term nonalcoholic steatohepatitis (NASH) in 1980. The more embracing term nonalcoholic fatty liver disease (NAFLD) has been adopted to cover the full spectrum of metabolic fatty liver disorders,3–7 particularly when histology is undefined. Other causes of steatohepatitis are sometimes referred to as “secondary NASH” but are better linked semantically to their known cause, for example, “alcoholic steatohepatitis” or “drug-induced steatohepatitis” (discussed in references 7 and 8). Until 10 years ago, no clues existed to the natural history or etiopathogenesis of NAFLD, and particularly how it could result in steatohepatitis or cirrhosis, but in the last decade there has been an explosion of interest in this disorder. Thus, Single Topic Conferences on NAFLD/NASH have been convened by the National Institutes of Health (1998), The Falk Foundation (The Hague, 2000), the American Association for the Study of Liver Diseases (2002), the Japanese Society of Hepatology and the European Association for Study of the Liver (both 2004), and the first book on NAFLD/NASH was published in 2005.7

Earlier classifications attempted to differentiate benign steatosis (with or without minor inflammation) from lesions that predicate adverse outcomes; the latter include ballooning degeneration of hepatocytes, Mallory bodies, and fibrosis.7, 9–11 In a recent move to reach consensus on pathological classification, the Pathology Committee of the NIH NASH Clinical Research Network have proposed a scoring system comprising 14 histological features.11 Although substantial sampling error can occur with needle biopsy of NAFLD,12 inter-rater variability between expert pathologists was high for fibrosis (kappa score, 0.84) and steatosis (0.79), lower for injury (0.56) and lobular inflammation (0.45). The unweighted sum of scores for steatosis, lobular inflammation, and hepatocellular ballooning was used to construct a new NAFLD activity score. This activity score can be used to classify cases into “NASH,” “borderline,” and “not NASH,” but agreement on diagnostic category is only moderate (0.61). Other aspects of NASH pathology,9–14 such as Mallory bodies and portal versus pericellular fibrosis, are more difficult to quantify or lack reproducibility. Nonetheless, Gramlich and colleagues (2004) emphasize the relationships between ballooning degeneration and Mallory bodies and perisinusoidal and perivenular fibrosis,14 confirming the concept that hepatocyte injury, an essential feature of NASH,7, 9–11 is the critical component of NAFLD that correlates with fibrogenesis.

Authors have differed on the level of alcohol consumption that can reliably distinguish between alcoholic steatohepatitis and NASH, from total abstinence (as in the original descriptions) to 20 to 40 g ethanol/day.2, 15–17 The latter approaches the threshold associated with an increased risk of cirrhosis in women.18 Small amounts of alcohol intake (1-2 standard drinks or 10-20 g ethanol/day) do not worsen hepatitis C.19 Although this finding is also likely to be true for NAFLD, it is harder to prove; lifetime total consumption may be more important.20 Lower levels of regular alcohol intake reduce the risk of cardiac events and improve insulin sensitivity in persons with T2DM, and probably in the metabolic syndrome.5 Because these are the most important risk factors for severity in NASH, it is possible that low-level alcohol intake has benefits rather than detriments. The NIH clinical research network on NAFLD/NASH has therefore agreed that the maximum allowable level of alcohol intake for definition of NAFLD is 2 standard drinks a day (140 g ethanol/week) for men, and one standard drink a day (70 g ethanol/week) for women; similar levels are adopted by the first book on NAFLD/NASH.7


NASH, nonalcoholic steatohepatitis; NAFLD, nonalcoholic fatty liver disease; T2DM, type 2 diabetes mellitus; PPAR-α, peroxisome proliferation- activater receptor alpha; IRS, insulin receptor substrate; IKK, inhibitor of kappaB kinase; TNFα, tumor necrosis factor alpha; SOCS, suppressors of cytokine signaling; IL-6, interleukin 6.

Current Scope and Future Dimensions


NAFLD/NASH has a very high prevalence in North and South America, much of Asia-Pacific (including Australia and New Zealand), the Middle East, and Europe. It is now the leading cause of referral to hepatology clinics in most regions, but accurate estimates of its incidence, prevalence, and natural history are lacking.15 Whereas NASH has been referred to as a disease of the “West,” altered socioeconomic circumstances and related changes in food intake, food composition, and physical activity (together referred to as “lifestyle”) may each play a role. Few studies document the relative importance of these factors, but a high intake of saturated fats was noted in one study,21 and another found a correlation between higher carbohydrate intake and liver inflammation.22 NAFLD/NASH is now regarded as a manifestation of the metabolic (or insulin resistance) syndrome, and the link between obesity, T2DM, cardiovascular disease and NAFLD is likely to reflect shared pathogenic factors.

Case definition and case ascertainment bedevil epidemiological studies of NAFLD/NASH and NASH-cirrhosis.15 Available data are based on tests that lack sensitivity and specificity, particularly unexplained aminotransferase elevations and abnormal (“bright”) hepatic ultrasonography. Some data bearing on prevalence of NAFLD/NASH are summarized in Table 1. Based on the Third National Health and Nutritional Examination Survey (NHANES III), which determined aminotransferase levels and excluded known causes of liver disease, the likely prevalence of NAFLD in North America and similar regions is 3% to 23%.4, 15 Population surveys using hepatic ultrasonography indicate similar prevalence (approximately 22%; 16% in lean, 76% in obese individuals).23–25 NHANES III was conducted a decade ago. A recent study using proton magnetic resonance spectrometry found approximately 30% of the U.S. population (including 45% Hispanics, 33% whites, 21% blacks; 42% white males, 24% white females) had increased hepatic triglyceride content.26 Another recent study based on ultrasonography found an apparent prevalence of NAFLD of 29% among healthy Japanese adults.27 Extrapolating data from liver biopsy or autopsy studies, 10% to 25% of such individuals, or 2% to 7% of the population, may have NASH and fibrosis or cirrhosis. In 20-year-old autopsy studies, the frequency of steatosis among those who died suddenly was 16% to 24%, and NASH was present in 2.1% to 2.4%.28, 29 Likewise, steatosis is found in approximately 20% of liver donors with normal aminotransferase levels.30, 31

Table 1. Some Data That Bear on the Prevalence of NAFLD and NASH*
Type of StudyPrevalence NAFLD (%)
  • Abbreviation: PMRS, proton magnetic resonance spectrometry

  • *

    Modified from McCullough.15

  • #

    Depending on ethnicity and gender, as follows: 45% Hispanic, 33% white, 24% black; 42% white males, 24% white females.

General population studies 
Ultrasound (references23–25, 27)∼22
 Bellentani et al.24 
 PMRS (Browning et al.26)24–45#
Liver enzymes (AT level) 
 Patt et al.3214–21
  NHANES III4, 153–23
Liver histology on selected groups 
Liver biopsy 
 7 studies—see McCullough1515–84
Autopsy-random deaths 
 Hilden et al.2824
 Ground2916 (NASH = 2.1)
Hospitalized deaths 
 Wanless and Lentz33 
Surgical patients 
 Adult living liver donors30, 3120
 Bariatric surgery15, 34–3956–86

More than 60% of men and 45% of women in Australia and the United States are overweight, at least one third of whom are obese; values range from 26% to 28% in some states. The prevalence of T2DM approximates 8%. It has been estimated that approximately 75% of those with obesity or T2DM have NAFLD,3, 15 whereas autopsy and biopsy studies show that approximately 20% of obese subjects have NASH.33–39 Thus, despite uncertainties about the precise prevalence of NAFLD and NASH, clearly obesity is now an important cause of cirrhosis.15 Because the prevalence of obesity and T2DM in North America and around the world has continued to rise since 1985, and because the frequency with which NASH is diagnosed in childhood is also increasing,7, 40, 41 the prevalence of NAFLD/NASH will likely continue to rise. This disorder will therefore contribute substantially to the burden of chronic liver disease in coming decades.

Risk Factors.

Obesity, hyperglycemia, T2DM, and hypertriglyceridemia are the best known.1–8, 24–27, 42–49 Earlier impressions that NAFLD/NASH was a female-predominant condition have been dispelled; it actually appears to be more prevalent in men.26, 42–49 The prevalence and severity of NAFLD/NASH appear to increase with age. Although cases with cirrhosis are now occasionally diagnosed in young adults, and particularly among those with hypothalamic and pituitary dysfunction,50 cases with liver complications are more often in the 6th through 8th decades of life.51, 52 This finding could be related to increasing rate of fibrotic progression with age53 or to mitochondrial dysfunction (which causes steatosis and hepatic insulin resistance) developing in the elderly.54

The strongest association of NASH is with central obesity (visceral adiposity), not overall obesity.43, 47–49, 55 Many individuals labeled as “non-obese” NAFLD on the basis of body mass index have central obesity; central obesity is a strong correlate with insulin resistance and a key feature of the metabolic syndrome.43, 44, 56 Thus, the prevalence of NAFLD (by ultrasonography) rose from 27% among those with normal fasting blood glucose, to 43% with fasting hyperglycemia and 62% among patients with T2DM.27 Genetic factors undoubtedly predispose to NAFLD/NASH.3, 6, 7, 57 Kindreds with more than 1 family member affected have been reported,58 whereas a family history of T2DM was found in two thirds of cases.43 The prevalence of steatosis is higher in Hispanics than among whites or blacks.26

A subset of factors has been linked to NAFLD severity (Table 2); the most powerful are hyperglycemia, T2DM, evidence of metabolic syndrome (e.g., arterial hypertension), obesity, and age.9, 42–49, 59–61 Employing these determinants and associated laboratory tests reflecting disease severity, investigators have constructed predictive indices (with euphonious acronyms) to identify individuals with the most or least severe forms of NAFLD (Table 2). The hope is that these scores can be used to select for or avoid liver biopsy.

Table 2. “Predictors” (Risk Factors and Laboratory Test Associations) for Fibrotic Severity in NASH
IndicesBARD59BARG60BAAT61HAIR34Angulo et al. 2005
  1. Abbreviations: AST, aspartate aminotransferase; ALT, alanine aminotransferase; BMI, body mass index; HgbA1c, hemoglobin A1c.

  2. *Unpublished data.

HgbA1c, hyperglycemiaYesYes
Insulin resistance indexYes

Natural History: Rate of Fibrotic Progression.

Although steatosis, even with minor inflammation, is nonprogressive,3–7, 15 the initial assessment of NASH as benign is not supported by current evidence.9, 15, 60–66 Progression of fibrosis has been demonstrated histologically in 32% to 37%,60, 65, 66 but apparent regression occurs in 18% to 29%.60, 66 Sampling error makes it difficult to be confident about rates of fibrotic progression,12 but they appear to be slow and variable, with regression also possible. Larger and longer studies are required to determine these. Obesity, diabetes, and initial fibrosis severity are the factors most conspicuously associated with fibrotic progression.60, 62, 65, 66 Estimated rates of cirrhosis development over 10 years have been 5% and 20% in 3 studies; the higher rate is for patients in a hospital setting,15, 62, 64 the lower comes from a community-based epidemiology project.52

Natural History: Clinical Outcome.

Patients with NASH/advanced stage fibrosis are at risk of developing liver complications within 7 years—the projected rate was approximately 30% in 1 study64 and up to 60% (13 of 21 cases) in another,52 which is similar to chronic hepatitis C with cirrhosis among untreated patients or those who experience nonresponse to antiviral therapy.64 Most patients with NASH/cirrhosis have T2DM or obesity and metabolic syndrome (Table 2). In the Verona diabetes study,67 the age-adjusted increased mortality for cirrhosis was second only to cancer and exceeded that for cardiovascular disease (which nonetheless remained the most common cause of death). The estimated rate of liver-related death over 10 years was 12% for patients monitored in a hospital setting.15, 64 In Olmsted County, Minnesota, survival for patients with NAFLD was less than for the age- and gender-matched general population, and liver disease was the third leading cause of death, compared with 13th in the general population.52 Among 420 cases of NAFLD (any severity), 7-year liver-related mortality was 1.7%; one patient required liver transplantation, and two developed hepatocellular carcinoma. Older age, impaired glucose tolerance, and cirrhosis were associated with higher risk of mortality.

It is therefore becoming apparent that age-adjusted mortality of patients with fibrotic NAFLD/NASH increases because of liver disease, albeit the absolute risk is relatively low.52 Rare cases of NASH appear to terminate in subacute liver failure,63 particularly after prolonged fasting or precipitant weight loss.68 The controversial issue of hepatocellular carcinoma will be discussed next.

Does NAFLD or Insulin Resistance Predispose to Hepatocellular Carcinoma?

Several cases of hepatocellular carcinoma have been described in patients with NASH, most often at diagnosis, but rarely during follow-up.52, 61, 62, 69–71 In one prospective study of defined cases of NASH-cirrhosis, no cases of hepatocellular carcinoma were found during 5-year follow-up,64 in striking contrast to 1% to 3% per annum with hepatitis C. In an earlier study of more than 20 years' follow-up of histologically defined NASH (varying severity), only one case of liver cancer was reported.62 In the larger Olmsted County community study,52 2 of 420 patients with NAFLD more broadly defined developed hepatocellular carcinoma during 7 years' follow-up, a rate that is much lower than would be expected for chronic hepatitis B or C. Diabetes is an “independent risk factor” for hepatocellular carcinoma irrespective of the presence of hepatitis B or C, or other liver diseases.72 However, the frequency of both diabetes and hepatocellular carcinoma increases as cirrhosis advances, so their association is not necessarily causal. A French study noted a relatively high incidence of hepatocellular carcinoma in obese patients with cirrhosis,61 but most cases were diagnosed at presentation or during the first year of follow-up, and other causative factors (previous hepatitis B virus infection, alcohol) were common.

Although it has been suggested that steatosis increases the risk of hepatocellular carcinoma with other causes,70, 73–75 the data derived from liver transplant series do not convincingly demonstrate that advanced NAFLD/NASH explains this association. Other causes of steatosis (hepatitis C, alcohol) need to be carefully appraised before these data can be accepted as evidence that NASH is an appreciable cause of hepatocellular carcinoma compared with other causes, such as chronic hepatitis B or C, hemochromatosis, and alcoholic cirrhosis. The available evidence supports the conclusion that patients transplanted for unambiguous NASH-cirrhosis rarely have coincident hepatocellular carcinoma.76–78 The statement that “cirrhotic-stage NASH… accounts for approximately 13% of all cases of hepatocellular carcinoma”75—a statement that references only the association between steatosis and primary liver cancer in explants70—is difficult to accept. Future studies to establish whether metabolic determinants, such as steatosis, insulin resistance, and glucose intolerance/T2DM, predispose to liver cancer must clearly distinguish whether fibrotic NASH is the cause, or whether the metabolic factors serve as epigenetic determinants of hepatocarcinogenesis initiated by other recognized causative factors. Meanwhile, the fact that patients with NASH-cirrhosis do not have the same high risk of hepatocellular carcinoma as more common causes has the management implication that liver cancer screening is unlikely to be cost-effective.

Steatosis, NASH or Cirrhosis?—The Diagnostic Challenge

NAFLD is most often diagnosed in asymptomatic (or tired) persons after the detection of raised aminotransferases during routine screening, or abnormal hepatic ultrasonography performed for another purpose, such as suspicion of gallstones. Alternatively, hepatomegaly may be detected during routine physical examination when liver tests are normal.26, 79 The presence of symptoms, usually nonspecific, does not appear related to disease severity; there can be improvement with modest weight reduction.80 The basis for metabolic liver disease can be revealed by a personal and lifestyle history documenting recent weight gain or expansion of abdominal girth. Patients with cirrhosis also may present with gastrointestinal bleeding from portal hypertension, or with muscle weakness, ascites, jaundice or hepatic encephalopathy.

Physical examination of patients with liver disease should now routinely include anthropometric measurements: body mass index and abdominal girth (waist circumference at umbilicus, or waist-to-hip ratio). Liver tests usually show minor nonspecific abnormalities. Alanine aminotransferase and gamma-glutamyl transpeptidase values are raised in most cases, but may be normal despite histologically significant disease.26, 79 Liver tests do not usually discriminate between steatosis, NASH, and cirrhosis, although a decrease in serum albumin, rise in aspartate aminotransferase:alanine aminotransferase ratio, and low platelet count often indicate cirrhosis (Table 2). It is important to appreciate when NAFLD is the most likely cause for aminotransferase elevation in a person with another possible liver disease,73, 81 including nonreplicative chronic hepatitis B virus infection, hepatitis C, and metabolic syndrome induced by corticosteroid therapy for autoimmune hepatitis or after liver transplantation. Hepatic imaging or liver biopsy may be required to clarify whether steatosis rather than the underlying liver disease is the basis for these changes.

The diagnostic workup of NAFLD should include fasting serum lipids, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and blood glucose. In addition to these metabolic tests, a test of insulin sensitivity should be performed. A dynamic test is preferred, such as a shortened 75 g glucose tolerance test with 1- and 2-hour serum insulin concentrations. At diagnosis, most cases of NASH have high serum insulin levels.43–45, 82 Serum C-peptide levels are also raised, reflecting increased pancreatic insulin secretion.43 Studies of cytokines and adipokines are of pathogenic interest in NASH,8, 83–86 which appears to be associated with more severe insulin resistance and lower adiponectin levels than uncomplicated steatosis,83, 85 but these assays do not yet have diagnostic utility.

Tests of Inflammation, and Iron Studies.

There is a need for biomarkers that reliably discriminate NASH from bland steatosis. An earlier suggestion that highly specific C-reactive protein may be valuable has not been confirmed.87, 88 Serum ferritin is elevated in approximately 60% of patients with NASH,3, 17 probably reflecting hepatocyte injury and liver inflammation rather than an increase in hepatic iron stores. Thus, although a possible role for iron in exacerbating NAFLD (particularly fibrogenesis) has been debated,8, 16, 17, 89–91 the weight of evidence is against this; some studies do show a weak association with C282Y.17, 91

Hepatic Imaging.

Increased echogenicity (“bright” scan) with ultrasonography or increased radiolucency with computerized tomography (compared with kidney) provide supportive evidence of steatosis. One study reported sensitivity of ultrasonography for steatosis was 89% with specificity of 93%; this compared with 77% and 89%, respectively, for fibrosis.92 Few studies compare ultrasound and computed tomography for diagnostic accuracy in NAFLD or NASH. In one such study, the sensitivity of both tests was 75% to 80% when appreciable (≥33% hepatocytes with stainable fat) steatosis was present.93 Hepatic magnetic resonance imaging is more sensitive than ultrasound for detecting minor degrees of steatosis,94 but both modalities readily detect heavy fat accumulation. By comparison, ultrasonography is operator-dependent, is affected considerably by body mass (obesity),95 and is relatively nonspecific; hepatic fibrosis or cirrhosis may give similar appearances. Magnetic resonance imaging also can be useful when fat distribution is heterogeneous on computed tomography so that the appearances cannot be distinguished from hepatic malignancy.96, 97

No imaging modality can distinguish NASH from simple steatosis.93, 98 Detection of cirrhosis by hepatic imaging has low sensitivity, although evidence should be sought for portal vein dilatation, retroperitoneal varices, or splenomegaly. Magnetic resonance imaging, proton magnetic resonance,26, 49, 55, 99 and dual-energy X-ray absorptiometry can quantify hepatic lipid stores and body fat distribution. These imaging modalities are currently of research interest26, 49, 55, 99 and in the future could be applied to monitor changes in hepatic steatosis as well as visceral adiposity. They are generally not yet available or are too expensive to recommend for routine clinical practice.

Liver Biopsy.

Liver biopsy is the only way to confirm the presence or absence of NASH in a person with features of NAFLD; it also remains the “gold standard” for fibrotic severity.7, 9–11 It remains unclear when liver biopsy is indicated because it is unlikely to alter management other than to reinforce the need for weight reduction.12, 100, 101

Noninvasive Assessment of Fibrotic Severity.

The strongest argument in favor of liver biopsy is the importance of detecting the presence of fibrosis in a patient with NASH. Several studies have shown that preexisting fibrosis is a strong predictor of subsequent progression to cirrhosis.9, 62, 66 Prospective studies that have determined the positive and negative diagnostic accuracy of various “predictive indices” are summarized in Table 2. Further research is needed to determine the confidence with which these (or new more accurate indices) can be applied in clinical practice to avoid the need for biopsy, and in what type of cases.

Other approaches to noninvasive assessment of fibrotic severity are currently being investigated. Serum hyaluronan appears to be a relatively accurate predictor of advanced stage fibrosis in NAFLD,102–105 but less so for distinguishing between minor degrees of fibrosis in NASH and uncomplicated steatosis.105 Thus, the negative predictive value was 96%, but the positive predictive value was only 51% when a cutoff for serum hyaluronan levels was set at 42 μg/L, assuming a prevalence of severe fibrosis in the population of NAFLD patients is 20%.102 Likewise, Sakugawa and colleagues103 found similar results for serum hyaluronic acid (cutoff 43 μg/L) with a negative predictive value of 78% and a positive predictive value of 92% for severe NASH fibrosis versus nonsevere NASH.103 Serum type VI collagen 7S domain gave very similar results.103 An International Multicenter Cohort study compared serum levels of 9 surrogate markers of liver fibrosis among over 1,000 patients with various chronic liver diseases subjected to liver biopsy. The study included age, hyaluronic acid, amino-terminal propeptide of type III collagen, and tissue inhibitor of matrix metalloproteinase A in a predictive algorithm, which was separately evaluated in a validation population. The algorithm was relatively sensitive (90%) for detecting fibrosis and accurately detected the absence of fibrosis (negative predictive value for significant fibrosis 92%).104 The performance of this algorithm was excellent for NAFLD.

Transient elastography (FibroScan) is another promising novel approach for determining fibrotic stage in hepatitis C and possibly other liver diseases.106 A limitation in NAFLD/NASH may be that technical difficulties can make it impossible to obtain reliable measurements in obese subjects. In summary, growing evidence indicates that noninvasive markers of liver fibrosis may be used with other clinical-pathological variables to reliably identify the vast majority of persons with NAFLD who do not have significant fibrosis, as well as those who probably do. Further definitive studies are required before these tests can be applied to replace liver biopsy in substantial subsets of patients with NAFLD.

Pathogenic Relationships

Insulin Resistance and Steatosis, Steatosis and Hepatic Insulin Resistance.

Metabolic studies in humans and animals show inextricable relationships between visceral adiposity (central obesity), steatosis, and insulin resistance.3, 5–8, 23, 24, 26, 34–39, 43–49, 55, 83, 84, 107–111 Insulin resistance is an impaired tissue response to insulin. In peripheral tissues (muscle, adipose), the consequence is decreased uptake of glucose; however, hepatic insulin resistance confers failure of insulin to stimulate glycogen synthesis and suppress gluconeogenesis.108, 109, 111 The result is failure of insulin to downregulate hepatic glucose production.

The etiopathogenic sequence of steatosis coming before hepatic insulin resistance has repeatedly been demonstrated with short-term high-fat feeding in rats108, 112 and in liver-specific lipoprotein lipase transgenic mice.113 Furthermore, correction of steatosis reverses hepatic insulin resistance in leptin-deficient ob/ob mice.111 In humans, attention is currently focused on mitochondrial defects as a primary cause of steatosis because of impaired β-oxidation of fatty acids.54, 107 Such mitochondrial defects could possibly have a genetic basis114 and are likely worsened by aging and environmental factors54, 107, 109 such as high saturated fats. Mitochondrial abnormalities have been described in the livers of patients with NASH.110, 115 Once created, steatosis could then cause hepatic insulin resistance.54, 107, 109, 113 Conversely, the near universal association of histologically defined NASH with peripheral insulin resistance, and usually with two or more features of the metabolic syndrome.33, 43, 45, 51 is consistent with an etiopathogenic role for peripheral insulin resistance in this clinically important form of NAFLD. Furthermore, the recurrence of NASH after liver transplantation is consistent with the disorder arising secondary to the metabolic milieu, rather than as a primary hepatic defect.

A key issue in NASH pathogenesis is whether hepatic insulin resistance could give rise to cellular injury and hepatic inflammation in the liver, as well as itself resulting from both inflammation and steatosis. Recent studies show the latter is possible by molecular mechanisms that block hepatic insulin receptor signaling.116 In both muscle and liver, a favored molecular mechanism of insulin resistance is intracellular accumulation of fatty acids and their metabolites that activate protein kinase C isoforms.107–109, 111, 113 Protein kinase C catalyzes serine/threonine phosphorylation of the insulin receptor and its key signaling intermediates, insulin receptor substrate [IRS]–1 and IRS-2, thereby impairing the tyrosine phosphorylation required for physiological signaling. Effects on IRS-2 are critical for insulin receptor signaling in the liver. In addition to protein kinase C, other serine/threonine kinases can abrogate insulin receptor signaling. These include inhibitor of kappaB kinase [IKK]-β and c-Jun N-terminal kinase, both of which can be activated directly by oxidative stress or indirectly by engagement of the tumor necrosis factor-alpha (TNFα) type 1 receptor. Cytokines also induce suppressors of cytokine signaling (SOCS), another family of proteins that can mediate hepatic insulin resistance.116

In both NASH and some forms of experimental steatohepatitis, enhanced hepatic expression of cytochrome P450 (CYP)2E1 occurs.117–121 This expression creates oxidative stress associated with impaired insulin receptor signaling.120, 121 TNFα, another candidate molecule in the transition of steatosis to steatohepatitis, is liberated from adipose tissues of obese persons.86, 109 It could worsen hepatic insulin resistance via activation of IKK-β or c-Jun N-terminal kinase, or by induction interleukin-6 (IL-6). In turn, IL-6 induces SOCS3, as reviewed elsewhere.116 The possible importance of SOCS3 receives support from studies of a transgenic mouse in which primary liver inflammation is induced by liver-specific expression of IKK-β kinase122; this process stimulates IL-6 release to cause both local (hepatic) and systemic insulin resistance.

Alternatively, steatosis and steatohepatitis could be attributable to the combined effects of severe peripheral insulin resistance and relative failure of humoral (adipokine) mediators that combat the effects of high insulin levels and fasting hyperglycemia on hepatic lipid turnover. The latter include increased fatty acid uptake and synthesis (lipogenesis), suppression of mitochondrial β-oxidation, and impaired triglyceride secretion as very-low-density lipoproteins.123 Some, but not all, studies have characterized low serum adiponectin levels as a factor that distinguishes NASH from simple steatosis.83–85 Adiponectin facilitates transport of fatty acids into mitochondria for β-oxidation and suppresses hepatic fatty acid synthesis, thereby countering the effects of high serum insulin levels. Interestingly, serum levels of adiponectin fall before onset of other metabolic complications, such as T2DM and cardiovascular disease. Furthermore, administered adiponectin can reverse experimental steatohepatitis by combating release of TNF-α.124, 125 The possible relevance of adiponectin and cytokines as potential therapeutic targets in NASH is discussed more fully elsewhere.126

Transition of Steatosis to Steatohepatitis.

Cell injury may occur when the capacity of hepatocytes to safely store fat is overwhelmed by continued uptake,127, 128 local synthesis, or impaired egress of fatty acids123, 126; these fatty acids then become toxic to the cell in a pathobiological process termed lipotoxicity. Lipotoxicity can cause cell death by the direct effects of lipid mediators on apoptosis. Alternatively, liberation of oxidized lipids and their peroxidation products (chemotactic aldehydes and organic acids) might be instrumental in recruiting and perpetuating the inflammatory response that characterizes NASH. The fatty liver is predisposed to forms of injury that involve oxidant stress. Steatosis could therefore provide the setting (or “first hit”) for NASH, but a “second hit” may be needed to cause cellular injury or recruit inflammation.129 The “two-hit” model of NASH pathogenesis was proposed in response to the experimental observations that endotoxin causes focal cytolysis and inflammation with exaggerated release of TNFα in rodent models of steatosis,86 and the livers of patients with NASH as well as animal models of metabolic steatohepatitis over-express CYP2E1,117–121 a microsomal fatty acid oxidase that generates reactive oxygen species. Oxidative stress, TNFα, other cytokines, and chemokines are all present in NASH,3, 5–8, 83, 86 but the ways in which they initiate or perpetuate steatohepatitis remain uncertain. Recent data indicate that TNFα may not be essential83, 130–132 and that oxidative stress may recruit inflammation via activation of nuclear factor-kappa B.122, 131

Fibrogenesis and Cirrhotigenesis.

Other factors might mediate or amplify the pathogenic processes that lead to hepatic fibrosis or cirrhosis. These include mitochondrial injury,109, 110, 115 stellate cell activation,133–135 and microvascular injury.82, 136 Oxidative stress could be directly involved in fibrogenesis,133 as shown by the protective effects of vitamin E in a nutritional model of steatohepatitis (N Phung, J George, unpublished data). Alternatively, mobilization of profibrogenic cytokines (connective tissue growth factor, transforming growth factor beta) could be involved, liberated in response to hyperglycemia or necroinflammatory change.133–135, 137 Many of these processes reflect interactions between host genes57 and environmental factors in complex, interactive pathogenic networks. For example, feeding obese (fa/fa) rats a high-fat diet was sufficient to cause oxidative stress associated with steatohepatitis and periportal fibrosis.127 Thus, emerging concepts of NASH pathogenesis might shed the notion about “two hits” struck sequentially, as multiple interactive pathogenic networks are more likely, central to which are the factors that predicate accumulation of fat within the liver in a form that causes tissue injury.

Reversal of Steatosis, Steatohepatitis, and NASH/Cirrhosis: What Is Achievable?

As for obesity, T2DM, cardiovascular disease, and several cancers, public health measures are indicated to prevent or reverse overnutrition and underactivity, the factors that lead to central obesity, insulin resistance, and NAFLD/NASH. The efficacy and cost-efficacy of such measures have not been clarified. More details are required to establish the specific roles of physical activity, energy intake, and food composition (content of carbohydrate, fat and dietary fiber; glycemic index, type of lipids and micronutrients)—these are important areas for future research. Meanwhile, applying similar principles of diabetes prevention/early intervention to NAFLD/NASH may be reasonable.138, 139 The logistics need to be worked out for professional advice and supportive programs of ‘lifestyle intervention’ (Table 3), how to pay for them, and the roles of primary care physicians, specialist physicians, dietitians, personal trainers, or case managers.

Table 3. Components of “Lifestyle Change” in Diabetes Intervention Studies*
• Attempt modest weight reduction by modulating energy intake and the other measures listed below
• Reduce total fat intake to <30% energy intake
• Replace saturated with unsaturated fats
• Replace fat with complex carbohydrates containing at least 15 g fiber/day
• Diet rich in fruit and vegetables
• Minimum of 140 minutes of exercise/week—fast walking as standard
• Use of case manager to advise, encourage, monitor progress

An extensive literature documents the beneficial effects of slow weight loss on serum aminotransferase levels in patients with NAFLD,139, 140–144 as reviewed elsewhere.144 “Crash dieting” or other approaches to weight reduction equivalent to fasting must be avoided because precipitant and profound weight loss [greater than 2 lbs (1 kg)/week] has been associated with worsening liver test abnormalities and accelerated hepatic fibrosis, or even liver failure.43, 62, 63, 68 Evidence shows that modest and sustained weight reduction, particularly in association with exercise, not only improves aminotransferase levels and reduces steatosis, but also resolves steatohepatitis and reverses hepatic fibrosis.39, 80, 141–143 Furthermore, correction of steatosis reverses hepatic insulin resistance.111 The concept that steatosis reversal can “unwind” NASH receives strong support from experimental studies in mice fed a methionine- and choline-deficient diet, in which the peroxisome proliferation-activater receptor-alpha (PPAR-α) agonist Wy-14,643 reversed all elements of fibrosing steatohepatitis within just 12 days (Fig. 1A).135 Equally impressively, in humans with morbid obesity subjected to bariatric surgery to effect gradual and sustained weight reduction, follow-up biopsies 1 to 4 years later showed resolution of NASH in most cases, including reduced severity of fibrosis, even in some cases with cirrhosis (Fig. 1B-C).39, 143

Figure 1.

Reversal of hepatic fibrosis in experimental steatohepatitis and in NAFLD/NASH. (A) Hepatic collagen deposition in livers of mice fed a methionine- and choline-deficient (MCD) diet for 8 weeks [Aa], or 9 weeks [Ab]. Other groups of animals were fed the MCD diet for 8 weeks and treated with Wy-14,643 (0.1% wt/wt) for the last 5 days [Ac], or 12 days [Ad]. Sirius red staining shows pericellular fibrosis (arrows), which is resolved by treatment with Wy-14,643. (From Ip et al.,135 used with permission.) (B) Improvement in hepatic fibrosis after bariatric surgery. Two cases from Dixon et al.143 are shown as cases 1 and 2, before [1Ba,c] and 40 and 28 months after surgery, respectively [1Bb,d]. Considerable reduction in fibrosis is evident (Sirius red stain). (Slides and clinical information provided by courtesy of Dr John Dixon and Professor Prithi Bhathal.) (C) Improvement in cirrhosis morphology after biliopancreatic diversion. Two cases from Kral et al.39 are shown (patients 9 and 10), before [1Ca,c] and after surgery (15 months for patient 9, 30 months for patient 10) surgery [1Cb,d]. In addition to clearance of steatosis, architectural remodeling and reduction in fibrotic severity is evident (Trichrome stain). Figure reprinted from Kral et al.,39 copyright 2004, with permission from Elsevier. (D) Histological changes in liver biopsies taken before and after treatment with rosiglitazone (4 mg twice a day for 48 weeks). 1Da-b, subject 2 pretreatment and posttreatment liver biopsies, respectively (hematoxylin-eosin, original magnification ×20) showing improvement in both necroinflammatory score (from grade 2 to grade 0) and fibrosis (from stage 2 to stage 1). 1Dc-d, subject 3 showing improvement in zone 3 pericellular fibrosis from predominant to delicate (trichrome stain, original magnification ×40). (From Neuschwander-Tetri et al.,147 used by permission and with the assistance of the authors.)

Pharmacological therapy of NASH has elicited considerable interest.144 Although the paradigm established with PPAR-α agonists in mice indicates possible utility of agents that alter hepatic lipid storage and turnover,135 little published experience exists of PPAR-α agonists used for NAFLD/NASH in humans,145, 146 and there are species differences concerning the importance of PPAR-α in the liver. Conversely, considerable interest has been shown in use of the PPAR-γ agonist thiazolidinediones (“glitazones”) against NASH. In preliminary studies, troglitazone (now withdrawn because of hepatotoxicity), rosiglitazone, and pioglitazone all improved serum aminotransferase levels.144 Furthermore, small controlled or open studies have demonstrated that the latter two agents substantially improve liver histology after 1 year of therapy.147, 148 In addition to beneficial effects on insulin resistance, steatosis, necroinflammatory change, and fibrosis all improved substantially (Fig. 1D).

PPAR-γ agonists act principally on adipocytes to facilitate their differentiation and enhance their storage capacity for triglycerides, thereby effectively “syphoning fat” out of the liver into subcutaneous deposits, as well as improving insulin sensitivity. As a result, weight gain is an unwanted effect; such an effect occurred in at least two thirds of patients in the rosiglitazone and pioglitazone NASH treatment trials.147, 148 Metformin is another agent that has modest effects in improving insulin sensitivity, albeit less than lifestyle modification.139 Metformin was more effective than vitamin E or prescriptive diet in improving steatosis, liver inflammation, and fibrosis in NAFLD.149 Among other agents studied for efficacy in NASH (Table 4), a randomized controlled trial has now shown that ursodeoxycholic acid is ineffective,150 but vitamin E has shown some promise, particularly in combination with other agents,151–154 as has another antioxidant, betaine.155, 156 A pilot study of pentoxifylline, an inhibitor of TNFα production, improved liver test abnormalities; however, 9 of 20 treated patients could not continue therapy because of nausea.157 Further trials of these agents are continuing, with longer-term histological outcomes and sustained benefits the benefit most sought.

Table 4. Drugs Used to Treat NASH
Class of drug; agentsType of StudyEffectsComments/Adverse Effect
 Metformin149Open-label, randomized trial (n = 55) vs. vitamin E (n = 28) or prescribed diet (n = 27) for 52 weeksImproved AT more often; reduced metabolic syndrome severitySelected (n = 14) metformin-treated subjects had improved steatosis, necroinflammatory change and fibrosis
 Pioglitazone148Open-labelImproved AT and liver histologyN = 18; loss of steatohepatitis; weight gain average 4%
 Pioglitazone + vitamin E152Randomized (open-label)Improved AT and liver histologyN = 21; both groups improved, but greater histological improvement in combination group
 Rosiglitazone147Open-labelImproved AT and liver histologyN = 30; 10% withdrew, weight gain in 67%, median 7.3% of body wt.; relapse posttreatment
 Vitamin E151Open-labelImproved ATChildren, n = 11, treated 4–10 months; no histology
 Vitamin E153Open-labelImproved ATN = 22; no histology
 Vitamin E + C154RCT—vs. placeboImproved histologyN = 45, but only 9 biopsied posttreatment; AT changes not mentioned
 Betaine155RCT—vs. placeboImproved ATN = 191, No histology
 Betaine156Open-labelImproved ATN = 8; improved histology
Other hepatoprotectant   
 Ursodeoxycholic acid150RCT—vs. placeboAT improved in both groups (no difference)N = 100, No difference between groups; no histology

Conclusions: Practical Recommendations

As yet no accepted drug treatment of NASH has been named. Application of pharmacological therapy should be regarded as experimental and best conducted within the constraints of a randomized controlled clinical trial. In the absence of such evidence, it seems reasonable to conclude that attempts to treat obesity constitute first-line therapy in patients with NAFLD/NASH. Thus, early intervention should attempt to increase physical activity while implementing dietary and other antiobesity measures. The emphasis needs to be on slow and modest reduction of body mass, not exceeding 2 pounds (1 kg)/week, coupled with increased physical activity.80, 138, 141 The latter may have beneficial effects on peripheral insulin sensitivity, even without weight loss. Whereas patients with NAFLD can have raised serum aminotransferase levels without NASH, or NASH without raised aminotransferases,79 it seems pragmatic during implementation of lifestyle measures to attempt normalization of serum aminotransferases as an indication of improvement in liver disease. Alternatively, ultrasonography could be used in a patient known to have normal aminotransferase levels. For the morbidly obese (body mass index >35 kg/m2) and those with refractory obesity, specific measures such as lipase inhibitors or safe forms of bariatric surgery should be considered. For all patients, the emphasis should be on reducing abdominal girth rather than dramatically lowering body weight. What is ultimately required is life-long maintenance of these beneficial changes.


The authors thank Dr. Narci Teoh for assistance in preparation of parts of this manuscript.