Beyond insulin resistance in NASH: TNF-α or adiponectin?

Authors


Abstract

Adiponectin has antilipogenic and anti-inflammatory effects, while tumor necrosis factor α (TNF-α) reduces insulin sensitivity and has proinflammatory effects. We examined (1) the extent to which hypoadiponectinemia and TNF-α activation are features of nonalcoholic steatohepatitis (NASH) and (2) whether serum levels of these markers correlate with the severity of histological changes in 109 subjects with nonalcoholic fatty liver disease (NAFLD), including 80 with NASH and 29 with simple steatosis. By multivariate analysis, subjects with NASH had reduced adiponectin level and increased TNF-α and soluble TNF receptor 2 (sTNFR2)—but not leptin levels, compared with controls matched by age, sex, and body mass index; these differences were independent of the increased insulin resistance (by homeostasis model [HOMA-IR]) in NASH. When compared with simple steatosis, NASH was associated with lower adiponectin levels and higher HOMA-IR, but there were no significant differences in the levels of TNF-α and sTNFR2. The majority of subjects with steatohepatitis (77%) had adiponectin levels less than 10 μg/mL and HOMA-IR greater than 3 units, but only 33% of those with pure steatosis had these findings. HOMA-IR and low serum adiponectin were also independently associated with increased grades of hepatic necroinflammation. In conclusion, hypoadiponectinemia is a feature of NASH independent of insulin resistance. Reduced adiponectin level is associated with more extensive necroinflammation and may contribute to the development of necroinflammatory forms of NAFLD. (HEPATOLOGY 2004;40:46–54.)

Nonalcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver injury in many countries.1, 2 The spectrum encompasses simple fatty liver (steatosis), which is generally nonprogressive,3 and nonalcoholic steatohepatitis (NASH), which can lead to cirrhosis and liver failure.4, 5 Insulin resistance is central to the pathogenesis of NAFLD,6, 7 and recent data indicate that NASH should be considered the hepatic manifestation of the insulin resistance syndrome.6, 8–10 However, while steatosis appears very common, NASH is not universal among those with the insulin resistance syndrome.11, 12 It is therefore important to discern what factors in the host metabolic milieu modulate the development of NASH, and, in particular, the transition from steatosis to steatohepatitis.

Adipose tissue secretes several bioactive proteins, or adipokines, that regulate hepatic and peripheral glucose and lipid metabolism. These adipokines include leptin, tumor necrosis factor α (TNF-α), resistin, and adiponectin. Although resistin inhibits insulin action in animal models, its expression is not increased in human subjects with insulin resistance or type 2 diabetes.13 We recently suggested that raised serum leptin levels in NASH may be a reflection of the failure of leptin to stimulate hepatic lipid turnover, that is, hepatic leptin resistance.14 Likewise, raised serum TNF-α levels have been demonstrated in several studies of fatty liver disease.15, 16 TNF-α interferes with insulin signaling, thereby favoring steatosis, and may play a proinflammatory role in the pathogenesis of NASH.17 However, these earlier reports of TNF-α levels in NAFLD failed to match control subjects for adiposity, a factor which is likely to influence the serum level of TNF-α.18 Furthermore, the other marker of activation of the TNF-α system, namely, soluble TNF receptor 2 (sTNFR2) level was not determined.19, 20 This is important because it has been shown in several reports that the level of sTNFR2, but not sTNFR1, correlates well with the magnitude of insulin resistance.19, 21, 22

In contrast to TNF-α, adiponectin has antilipogenic effects that may protect nonadipocyte tissues, such as the liver, against lipid accumulation.23 In addition, the conditions most associated with the development of NAFLD, namely, obesity,24 insulin resistance,25 type 2 diabetes,26, 27 and dyslipidemia,28 all have reduced adiponectin levels. Adiponectin also has anti-inflammatory effects that could protect the fatty liver from the development of inflammation and cell injury (necroinflammatory change).29–31 Hence, adiponectin and TNF-α have opposing effects on insulin sensitivity and inflammation,32 and the balance between these two adipokine systems may be important to the pathogenesis of NASH.

We conducted the present study to test the hypothesis that hypoadiponectinemia and TNF-α activation occur in patients with NASH and that measures of these markers together with insulin resistance may help distinguish patients with NASH from those with simple steatosis. NASH subjects were compared with (1) controls matched for the factors known to affect adipokine levels (age, sex, and body mass index [BMI]) and (2) patients with fatty liver without inflammation (steatosis). We then determined whether the serum level of adiponectin and markers of the TNF-α system (TNF-α, sTNFR2) correlated with the severity of hepatic steatosis, necroinflammation, and fibrosis in NAFLD in a multivariate model.

Abbreviations

NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; TNF-α, tumor necrosis factor α; sTNFR, soluble TNF receptor; BMI, body mass index; LFT, liver function test; ALT, alanine aminotransferase; WHR, waist to hip ratio; HOMA-IR, homeostasis model; ROC, receiver operating characteristic.

Patients and Methods

We studied 109 adults (68 men and 41 women) with biopsy-proven NAFLD recruited at Westmead Hospital; 80 had NASH and 29 had steatosis without inflammation. Patients were referred for the assessment of abnormal liver function tests (LFT) or hepatic steatosis detected by ultrasonography; several have been the subject of earlier reports.6, 14, 33 In all patients, current and past daily alcohol intake was less than 40 g per week; details regarding alcohol consumption were obtained independently by at least 2 physicians and confirmed by close family members. All subjects had a normal serum albumin level, prothrombin time, and renal function. The diagnosis of type 2 diabetes mellitus was based on World Health Organization criteria.34 None of the patients were using thiazolidinediones.

The following disorders were excluded, as previously reported6: secondary causes of steatohepatitis and drug-induced liver disease, alcoholic liver disease, viral hepatitis, autoimmune hepatitis, primary biliary cirrhosis, sclerosing cholangitis, α1-antitrypsin deficiency, hemochromatosis, Wilson's disease, and biliary obstruction. The study protocol was approved by the Human Ethics Committee of the Western Sydney Area Health Service and written informed consent was obtained.

Pathology.

Liver tissues were stained with hematoxylin-eosin, reticulin, and Gomori trichrome stains and scored by an experienced hepatopathologist (J.G.K.). All cases showed macrovesicular steatosis affecting at least 5% of hepatocytes and were classified as either steatosis or steatohepatitis. In addition to steatosis, the minimum criteria for the diagnosis of steatohepatitis included the presence of lobular inflammation and either ballooning cells or perisinusoidal/pericellular fibrosis in zone 3 of the hepatic acinus.3, 5 Subjects with cirrhosis were either “definite” or “probable” cases of NASH-associated cirrhosis according to a recently proposed clinicopathological classification.4 All cases were scored using the method of Brunt.35 Steatosis was graded as follows: grade 1 (≥5% and <33% of hepatocytes affected); grade 2 (33%-66% of hepatocytes affected); or grade 3 (>66% of hepatocytes affected). As there was only 1 subject with a necroinflammatory grade of 3 (Table 1), those with grades 2 and 3 were combined for statistical analyses. The stages of fibrosis were grouped into 2 categories: mild (stages 0-2) and advanced (stages 3-4) fibrosis.

Table 1. Histological Findings for Subjects With Steatohepatitis and Steatosis
 Steatohepatitis (n = 80)Steatosis (n = 29)
  1. Abbreviation: NA, not applicable.

Steatosis grade  
 133 (41%)21 (72%)
 233 (41%)7 (24%)
 314 (18%)1 (3%)
Necroinflammatory grade  
 155 (69%)NA
 224 (30%) 
 31 (1%) 
Fibrosis stage  
 06 (8%)NA
 124 (30%) 
 219 (24%) 
 315 (19%) 
 416 (20%) 

Controls.

Eighty-two matched volunteers were enrolled as controls for the subjects with steatohepatitis. Fifty-two were recruited by advertisement in the local newspapers, and the rest were staff of Westmead Hospital contacted directly by the authors. All had normal physical examinations and LFTs, negative serology for viral hepatitis, and no history of liver disease. At our institute, the upper limits of alanine aminotransferase (ALT) were 47 U/L for males and 33 U/L for females, and the upper limit of aspartate aminotransferase were 45 U/L for both males and females. Only 3 male subjects had ALT levels greater than 40 U/L. The 53 male controls were frequency matched with the 48 male subjects with NASH by age and BMI; likewise, the 29 female controls were frequency matched with 32 female patients.36

Clinical and Laboratory Evaluations.

A complete physical examination was performed on each subject. Anthropometric evaluation included measures of BMI and central obesity (waist and hip circumferencesand waist-hip ratio [WHR]). BMI was calculated as weight (kg) divided by height (m) squared, and WHR was defined as the waist circumference (cm) divided by the hip circumference (cm).

On the morning of liver biopsy, venous blood samples were drawn after an overnight 12-hour fast to determine the levels of serum ALT, bilirubin, albumin, total cholesterol, triglycerides, glucose, insulin, C-peptide, adiponectin, leptin, TNF-α, and sTNFR2. Serum insulin was determined by a radioimmunoassay technique (Phadeseph Insulin RIA; Pharmacia and Upjohn Diagnostics AB, Uppsala, Sweden). Serum C-peptide was estimated by a competitive immunoassay (IMMULITE; Diagnostic Products Corporation, CA). Serum adiponectin levels were measured in duplicate by radioimmunoassay (Linco Research, St Charles, MO). Leptin (Diagnostic Systems Laboratories, Webster, TX) and TNF-α and sTNFR2 (Quantikine ELISA; R&D Systems, Minneapolis, MN) were measured using enzyme-linked immunosorbent assays. All other biochemical tests were performed using a conventional automated analyzer within the Department of Clinical Chemistry at Westmead Hospital.

Insulin resistance was calculated by the homeostasis model (HOMA-IR) using the following formula37: HOMA-IR = fasting insulin (mU/L) × plasma glucose (mmol/L)/22.5. Type 2 diabetes mellitus is part of the spectrum of disorders of insulin sensitivity, and this index has been shown to correlate well with the results of the euglycemic-hyperinsulinemic clamp in type 2 diabetic patients, including those treated with metformin and other oral hypoglycemic agents.38–40 However, HOMA-IR was not estimated in the 3 diabetic patients undergoing treatment with insulin.

Statistical Analyses.

The baseline characteristics of subjects with steatohepatitis were compared with those with steatosis or matched controls by 1-way ANOVA; P values were adjusted using Dunnett's method for multiple pairwise comparisons (for continuous variables). Chi-square test or Fisher exact test was used for nominal categorical variables, and Spearman rank correlation was used to assess association between continuous variables. Multiple logistic regression analyses with stepwise variable selection were then performed to compare the steatohepatitis subjects with the controls and to compare the subjects with steatohepatitis and steatosis; the covariates considered in these analyses were age, BMI, WHR, diabetes, HOMA-IR, adiponectin, TNF-α, sTNFR2, and leptin, together with their interaction with sex.

Among all subjects with NAFLD (simple steatosis or NASH), multiple ordinal regression analyses with forward stepwise variable selection were performed to identify independent predictors for the steatosis and necroinflammatory grade. Multiple logistic regression analysis with stepwise variable selection was used to identify the independent predictors for advanced fibrosis (stage 3 or 4). The covariates for these analyses were age, BMI, WHR, diabetes, HOMA-IR or C-peptide (see Results), adiponectin, TNF-α, sTNFR2, and leptin, together with their interaction with sex.

Area under the receiver operating characteristic (ROC) curve was used to illustrate the diagnostic ability of HOMA-IR and adiponectin in distinguishing between steatosis and steatohepatitis.

Results are expressed as mean ± SE unless stated otherwise. All analyses were performed with SPSS software for Windows, version 10 (SPSS Inc., Chicago, IL). A significance level of 5% was used throughout.

Results

Characteristics of Patients and Controls.

The histological findings of the subjects with steatohepatitis (NASH) and steatosis are shown in Table 1. The clinical and biochemical characteristics of the male and female cohorts with steatohepatitis are compared with their respective matched controls (with no known liver diseases) in Table 2; matching was obtained for age and BMI. When subjects with steatohepatitis were compared to those with steatosis, the male subjects were of similar age and BMI, but the females with steatosis were younger than their steatohepatitis counterparts. Despite having similar BMI, measures of central obesity (waist circumference and WHR) were higher for subjects with steatohepatitis compared to the matched controls; those with steatosis had intermediate values for WHR. The prevalence of diabetes and markers of insulin resistance (insulin, C-peptide, and HOMA-IR levels) showed trends similar to that of WHR, with NASH subjects having the highest values. By contrast, WHR was inversely associated with adiponectin levels (r = −0.4, P < .001). The difference in diabetes prevalence between subgroups was significant for women but not men (Table 2).

Table 2. Characteristics of Men and Women: Controls, Subjects With Steatosis, and Subjects With Steatohepatitis
CharacteristicMenWomen
Controls (n = 53)P Value*Steatosis (n = 20)P ValueSteatohepatitis (n = 48)Controls (n = 29)P Value*Steatosis (n = 9)P ValueSteatohepatitis (n = 32)
  • NOTE. Results expressed as mean ± SE or frequency (percentage).

  • Abbreviation: NS, not significant.

  • *

    Comparison between NASH subjects and controls, P values were adjusted using Dunnett's method as appropriate (see Patients and Methods).

  • Comparison between NASH subjects and fatty liver, P values were adjusted using Dunnett's method as appropriate (see Patients and Methods).

Age (y)45.9 ± 1.6NS41.2 ± 2.7NS45.2 ± 1.853.7 ± 2.0NS42.8 ± 4.4.00456.1 ± 1.9
BMI (kg/m2)29.4 ± 0.5NS29.1 ± 0.8NS30.7 ± 0.730.5 ± 0.8NS31.5 ± 1.5NS31.9 ± 0.8
Waist (cm)100 ± 2NS99 ± 2NS104 ± 292 ± 2<.001102 ± 5NS106 ± 3
WHR0.96 ± 0.01.0030.98 ± 0.01NS1.01 ± 0.010.83 ± 0.01<.0010.92 ± 0.04.021.01 ± 0.02
ALT (U/L)21 ± 2<.00168 ± 6.0391 ± 719 ± 8<.00172 ± 17NS81 ± 8
Diabetes mellitus4 (8%)NS2 (10%)NS9 (19%)3 (10%)<.0011 (11%).0217 (53%)
Glucose (mmol)5.4 ± 0.1.0055.6 ± 0.3NS6.0 ± 0.15.7 ± 0.4.0055.1 ± 0.2.017.9 ± 0.6
Insulin (μU/mL)10.2 ± 0.8<.00114.1 ± 1.3.00220.6 ± 1.49.2 ± 0.8<.00113.4 ± 2.5.0719.3 ± 2.0
C-peptide (nmol/L)0.70 ± 0.04<.0010.85 ± 0.06<.0011.25 ± 0.030.65 ± 0.06<.0010.76 ± 0.10.021.30 ± 0.14
HOMA-IR2.5 ± 0.2<.0013.5 ± 0.4.0015.5 ± 0.42.4 ± 0.3<.0013.2 ± 0.7.015.8 ± 0.6

Univariate Comparison of Adiponectin, Leptin, TNF-α, and sTNFR2 Between Subjects With Steatohepatitis, Matched Controls, and Subjects With Steatosis.

Adiponectin levels were significantly different between the 3 subgroups (P = .008 for females and P < .001 for males by ANOVA): the highest levels were in controls and the lowest in patients with NASH (Fig. 1). Mean adiponectin levels in the controls were about twice as high as the levels in NASH subjects, and the mean differences were more than 5 μg/mL for both male and female subjects (Fig. 1). The opposite trend was evident for leptin levels for males (P = .005 by ANOVA) but not females (P = .2 by ANOVA): for men, the highest levels were among steatohepatitis subjects (18.9 ± 2.1 ng/mL), intermediate levels were for steatosis (14.3 ± 2.2 ng/mL), and the lowest levels were in controls (11.5 ± 0.9 ng/mL). TNF-α level was higher for steatohepatitis subjects compared with controls (P < .001) but was similar between subjects with NASH and those with steatosis (P = .9) (Fig. 2). The sTNFR2 level was higher for steatohepatitis compared with both controls (P < .001) and subjects with steatosis (P = .03) (Fig. 3).

Figure 1.

Mean (± SEM) adiponectin levels in controls, subjects with steatosis, and subjects with steatohepatitis. P = .009 for female subjects and P < .001 for male subjects by ANOVA.

Figure 2.

Mean (± SEM) TNF-α levels in controls, subjects with steatosis, and subjects with steatohepatitis. P < .001 for steatohepatitis versus controls. P = .9 for steatohepatitis versus steatosis.

Figure 3.

Mean (± SEM) sTNFR2 levels in controls, subjects with steatosis, and subjects with steatohepatitis. P < .001 for steatohepatitis versus controls. P = .03 for steatohepatitis versus steatosis.

Multivariate Comparison Between Subjects With steatohepatitis, Matched Controls, and Subjects With Steatosis.

To assess whether the observed differences in adiponectin, leptin, TNF-α, and sTNFR2 levels between the 3 study groups were independent of the effect of insulin resistance, and to quantify the extent of these differences (odds ratio [OR]), multiple logistic regression analyses were performed comparing NASH subjects with the matched controls and with subjects having steatosis. Compared with matched controls, reduced adiponectin level and increased TNF-α and sTNFR2—but not leptin—levels occurred in NASH, independent of the increased HOMA-IR and WHR for these subjects (Table 3). Each 5-μg/mL decrease in adiponectin was associated with an OR of 6.1 for developing NASH (P = .001), and a 1-pg/mL increase in TNF-α resulted in a 2.0-fold increased risk of NASH (P = .004).

Table 3. Multiple Logistic Regression Analysis of Factors Associated With NASH Compared to Matched Controls
FactorOdds Ratio95% CIP Value
  1. NOTE. Intercept = −13.368 (SE = 4.884). R2 (Nagelkerke method) = 74%.

Adiponectin (per 5-μg/mL decrease)6.12.0–18.9.001
TNF-α (per pg/mL increase)2.01.2–3.3.004
sTNFR2 (per ng/mL increase)5.61.5–21.4.01
HOMA-IR (per unit increase)1.41.0–1.9.03
WHR (per 0.1 unit increase)2.71.1–6.9.04

When subjects with steatohepatitis were compared with those having steatosis in the multivariate model, HOMA-IR was significantly higher (OR 1.7; CI, 1.3-2.3; P = .001). There was a trend for adiponectin levels to be lower in steatohepatitis than in steatosis (OR 1.7; CI, 1.0-3.0; P = .06). There was no significant difference in the levels of the other adipokines and sTNFR2 between those with steatohepatitis and those with steatosis. Using adiponectin and HOMA-IR levels, the area under the ROC curve for distinguishing between steatohepatitis and steatosis was 0.79 (Fig. 4). Most subjects with steatohepatitis (77%) had an adiponectin level less than 10 μg/mL and a HOMA-IR level greater than 3 units, but only 33% of those with steatosis were in this category.

Figure 4.

ROC curve for differentiating steatosis and steatohepatitis using adiponectin and HOMA-IR (area under curve = 0.79).

Predictors for the Grade of Hepatic Steatosis in NAFLD.

The relationship between the severity of histological changes in NAFLD (steatosis or NASH) and the levels of adipokines and sTNFR2 were analyzed using multivariate models, adjusting for the effect of insulin resistance. Using multiple ordinal regression analysis, the independent predictors for increased steatosis grade were adiponectin level (negative association) and C-peptide levels (Table 4). HOMA-IR was not an independent predictor in the model after controlling for the effect of C-peptide, which is a surrogate marker of hyperinsulinemia. C-peptide level strongly correlated with HOMA-IR (r = 0.8, P < .001), and provides an indication of insulin resistance by reflecting the level of pancreatic insulin secretion.41

Table 4. Multiple Ordinal Regression Analysis for Factors Associated With the Steatosis Grade in Subjects With NAFLD (Simple Steatosis or NASH)
FactorOdds Ratio95% CIP Value
Adiponectin (per 5-μg/mL decrease)2.11.2–3.8.01
C-peptide (per nmol/L increase)2.51.2–5.6.02

To further explore the relationship between adiponectin levels and the extent of hepatic steatosis, we selected cutoff points for adiponectin levels to categorize the cohort into 3 groups (group 1, 0-5 μg/mL; group 2, 5.1-10 μg/mL; and group 3, >10 μg/mL; Table 5). Most subjects (73%) with adiponectin levels greater than 10 μg/mL had mild (grade 1) hepatic steatosis, while the majority of those (64%) with low adiponectin levels (≤5 μg/mL) had moderate-to-severe (grade 2 or 3) steatosis. The odds of moderate-to-severe steatosis increased by a factor of 2.13 (P = .006 by logistic regression analysis) per decrease in adiponectin level group, i.e. group 3 versus group 2, or group 2 versus group 1.

Table 5. Relationship Between Adiponectin Levels and Grade of Hepatic Steatosis
 Adiponectin Levels (μg/mL)
0–55.1–10>10
  1. NOTE. Results expressed as frequency (percentage).

Steatosis grade 116 (36%)20 (50%)16 (73%)
Steatosis grade 2 or 329 (64%)20 (50%)6 (27%)
Total45 (100%)40 (100%)22 (100%)

Predictors for the Necroinflammatory Grade and Fibrosis Stage in NAFLD.

The relationships between serum adipokines and hepatic injury in NAFLD were further examined by multivariate analyses to identify independent predictors of hepatic necroinflammation and fibrosis stage in subjects with steatosis or steatohepatitis. Even after controlling for the effect of HOMA-IR, decreased serum adiponectin levels were associated with higher hepatic necroinflammatory grades (Table 6). Thus, among the subjects with grade 2-to-3 inflammation (by definition, all had steatohepatitis), 21 of 25 (84%) had adiponectin levels less than 10 μg/mL, whereas 11 of 23 (48%) of those with simple steatosis had adiponectin levels greater than or equal to 10 μg/mL.

Table 6. Multiple Ordinal Regression Analysis for Factors Associated With the Necroinflammatory Grade in Subjects with NAFLD (Steatosis or Steatohepatitis)
FactorOdds Ratio95% CIP Value
Adiponectin (per 5-μg/mL decrease)2.41.4–4.0.001
HOMA-IR (per unit increase)1.21.0–1.4.02
Age (per 10-yr increase)1.41.0–2.0.05

Serum adipokine (adiponectin, TNF-α, leptin) and sTNFR2 levels were not associated with the extent of hepatic fibrosis in the multiple logistic regression model. HOMA-IR was the only independent predictor of severe fibrosis (OR 1.4; CI, 1.1-1.7; P = .001). Although subjects with severe fibrosis were older than those with lower stages of fibrosis (mean age, 53.9 ± 2.4 years and 44.9 ± 1.4 years, respectively), age was not a significant factor in the multivariate model (P = .08).

Discussion

This study confirms that hypoadiponectinemia and activation of the TNF-α system occur in subjects with NASH when compared to controls matched for age, BMI, and sex. Furthermore, adiponectin levels were lower for individuals with NASH than for those with the more benign form of NAFLD, simple steatosis. Lower adiponectin levels were also associated with higher grades of hepatic steatosis and necroinflammatory activity in these insulin resistance-related fatty liver disorders.

Animal models and preliminary uncontrolled human data have indicated that adiponectin could confer protective effects against alcoholic and nonalcoholic fatty liver diseases.23 The present data provide compelling evidence to support the hypothesis that adiponectin has anti-inflammatory hepatoprotective effects in humans with fatty liver attributable only to metabolic factors. Adiponectin is secreted by adipocytes and circulates at relatively high systemic concentrations to influence metabolic function.24 Reduced systemic levels of adiponectin have clinically significant effects and increase the risk of cardiovascular disease in both diabetic and nondiabetic patients.26, 42, 43 In the present study, decreased serum levels of adiponectin occurred in subjects with NASH when compared to matched controls and those with steatosis, independent of the effect of central obesity (WHR), insulin resistance, and other adipokines, including TNF-α and leptin. Type 2 diabetes and visceral obesity are major risk factors for NASH and are associated with reduced adiponectin levels.25–27 Our data support the concept that hypoadiponectinemia may be an independent mediator for the strong association between NASH and these clinical risk factors.

A potential confounding factor in the present study is that some of the controls could have underlying fatty liver disease, despite normal LFTs.44 In particular, we considered the possibility that higher adiponectin levels may be a feature of NAFLD subjects with normal ALT compared to those with raised ALT. If this is the case, the higher adiponectin levels in the control group may partly be a result of the inclusion of NAFLD subjects with normal ALT. However, this is an unlikely explanation because there was no significant association between ALT and adiponectin serum levels (Spearman r = −0.1, P = .3). Furthermore, it has been shown that individuals with NAFLD and normal or raised ALT have similar clinical and histological features that in turn predict serum adiponectin levels (BMI, WHR, gender distribution, prevalence of features of the metabolic syndrome, histological severity of hepatic steatosis and necroinflammation).44 Accordingly, we believe that the undiagnosed NAFLD subjects in the control group would likely have similar levels of adiponectin as the subjects in the NASH group. The unwitting inclusion of these subjects would actually decrease the power of this study to detect differences in adipokine levels between subjects with NASH and controls, rather than bias toward such a difference. Therefore, the fact that significant differences were noted is a strong indication that adiponectin could play an important hepatoprotective role against inflammatory recruitment in NAFLD.

Adiponectin and the Severity of Histological Changes of NAFLD.

The independent associations between lower serum adiponectin levels and measures of insulin resistance with higher grades of hepatic steatosis and necroinflammation support the hypothesis that both hypoadiponectinemia and insulin resistance may be independently involved in the pathogenesis of NAFLD, particularly in the progression from steatosis to steatohepatitis. Thus, NASH subjects had lower adiponectin levels and higher HOMA-IR than those with simple steatosis. In vitro studies have shown that adiponectin exerts several anti-inflammatory effects, including inhibition of nuclear factor κB activation45 and suppression of macrophage function.31 Furthermore, in animal models of alcoholic steatohepatitis, exogenous adiponectin suppresses hepatic inflammation and decreases the hepatic expression and plasma concentration of TNF-α.23 Peroxisomal proliferator activated receptor γ agonists have been found to increase adiponectin levels and to be associated with a reduction in hepatic steatosis.46 Two recent studies found that these agents also reduce hepatic necroinflammation in subjects with NASH,47, 48 consistent with our proposal of the anti-inflammatory effects of adiponectin.

In contrast to adiponectin, TNF-α and sTNFR2 levels did not correlate with the severity of necroinflammation, although they were increased in steatohepatitis compared to matched controls. In addition, serum levels of TNF-α were similar between subjects with steatosis and those with steatohepatitis. This indicates that while TNF-α activation occurs in NAFLD, the present evidence is not consistent with a role for TNF-α in potentiating the progression from steatosis to steatohepatitis. A larger study incorporating tissue measures of TNF-α signaling intermediates is required to clarify the role of TNF-α in the development of necroinflammation in NAFLD.

There was no association between the severity of hepatic fibrosis and the level of the adipokines measured. Unlike hepatic steatosis and inflammation, which can become altered over a short time in relation to changes such as lifestyle modification,49, 50 fibrosis progression may occur over decades in NAFLD.51 Adipokine levels fluctuate over time depending on the metabolic milieu. It is therefore not surprising that the cytokine levels observed in this cross-sectional study did not correlate with the extent of fibrosis. In the absence of a significant association between adiponectin levels and fibrosis stage, it is unlikely that the low circulating adiponectin in NASH was a result of the impairment in hepatic function. However, we cannot exclude the possibility that reduced adiponectin level in NASH may be secondary to hepatic necroinflammation.

Diagnostic Utility of Serum Adiponectin Levels.

An important issue in the clinical management of NAFLD is the differentiation between simple fatty liver and NASH. Natural history studies have shown that patients with simple fatty liver are unlikely to develop liver-related morbidity and mortality,3, 5 but patients with NASH may progress to cirrhosis, liver failure, and hepatocellular carcinoma.4, 52 At present, only liver biopsy can reliably distinguish between the two conditions. In this respect, measurements of adiponectin and HOMA-IR were reasonably accurate for differentiating between steatosis and steatohepatitis, as shown by the area under the ROC curve of 79% (CI, 68%-89%). These findings, however, need to be validated in external cohorts with a different mix of clinical features (e.g., including type 2 diabetes, obesity), which can affect HOMA-IR and adiponectin levels in the steatosis and steatohepatitis groups. A scoring system combining such noninvasive markers with age and other clinical features may permit the reliable differentiation of NASH from steatosis and hence the selection of patients with a higher probability of significant pathology for liver biopsy.

In conclusion, hypoadiponectinemia and elevated HOMA-IR appear to be more consistent and profound in NASH than in bland steatosis and may be one of the pathogenic links between central obesity and the development of necroinflammatory forms of NAFLD. Measurement of adiponectin levels and HOMA-IR hold promise for identifying steatohepatitis in subjects with NAFLD. Although subjects with NASH have increased TNF-α and sTNFR2 levels compared with overtly healthy controls, the levels of the markers of TNF-α activation do not correlate with the severity of necroinflammation. Future studies should explore the prognostic value of serum adipokines and insulin resistance for the long-term outcome of patients with NAFLD.

Acknowledgements

The authors thank Dr. Karen Byth for assistance in statistical analyses and Dr Isabelle Leclercq and the Endocrinology Laboratory, Westmead Hospital for advice on the biochemical assays. Drs. Rita Lin and Dev Samarasinghe and Ms. Lee Russell assisted with data collection.

Ancillary