Nonalcoholic steatohepatitis versus steatosis: Adipose tissue insulin resistance and dysfunctional response to fat ingestion predict liver injury and altered glucose and lipoprotein metabolism


  • Potential conflict of interest: Nothing to report.


Nonalcoholic fatty liver disease (NAFLD) ranges from simple steatosis (SS) to nonalcoholic steatohepatitis (NASH). Though liver-related risk seems confined to NASH, it is currently unclear whether NASH has a higher risk of cardiovascular disease (CVD) and diabetes than SS as a result of the coexistence of obesity and other cardiometabolic confounders. Adipose tissue is an emerging modulator of liver disease in NAFLD and of cardiometabolic disease in the general population. We evaluated in SS and NASH (1) glucose homeostasis and cardiovascular risk profile and (2) the effect of adipose tissue dysfunction, assessed in fasting conditions and postprandially, on liver injury, glucose and lipoprotein metabolism, and markers of early atherosclerosis. Forty nonobese, nondiabetic, normolipidemic biopsy-proven NAFLD patients (20 with SS and 20 with NASH) and 40 healthy subjects, matched for overall/abdominal adiposity and metabolic syndrome, underwent an oral fat load test, with measurement of plasma triglyceride-rich lipoproteins, oxidized low-density lipoproteins, adipokines, and cytokeratin-18 fragments, and an oral glucose tolerance test with minimal model analysis to yield glucose homeostasis parameters. Circulating endothelial adhesion molecules were measured, and adipose tissue insulin resistance (adipose IR) index and visceral adiposity index were calculated. Despite similar fasting values, compared to SS, NASH showed a more atherogenic postprandial lipoprotein profile, an altered adipokine response (i.e., higher resistin increase and an adiponectin fall), and hepatocyte apoptosis activation after fat ingestion. Adipose IR index, endothelial adhesion molecules, and hepatic insulin resistance progressively increased across NAFLD stages. NASH, but not SS, showed an impaired pancreatic β-cell function. On multiple regression analysis, adipose IR index and postprandial adiponectin independently predicted liver histology and altered cardiometabolic parameters. Conclusion: Adipose tissue dysfunction, including a maladaptive adipokine response to fat ingestion, modulates liver injury and cardiometabolic risk in NAFLD. (HEPATOLOGY 2012;56:933–942)

Nonalcoholic fatty liver disease (NAFLD) affects 30% of the general adult population and up to 80% of obese and diabetic subjects1 and is considered the hepatic manifestation of metabolic syndrome. NAFLD ranges from simple steatosis (SS) to steatosis plus necroinflammation (e.g., nonalcoholic steatohepatitis; NASH) with or without fibrosis.2 Emerging evidence suggests that NAFLD is associated with an increased risk of liver-related complications, of type 2 diabetes mellitus (T2DM), and of cardiovascular disease (CVD).3 Though liver-related risk appears to be confined to NASH, it is currently unclear whether NASH confers a higher risk of CVD and T2DM than SS as a result of the varying epidemiological association of NASH and SS with other cardiometabolic confounders, including obesity and metabolic syndrome, which prevents any inference on the independent effect of NASH/SS on cardiometabolic risk and on mechanisms underlying both liver disease progression and cardiometabolic risk in NAFLD.

Adipose tissue is emerging as a key mediator of cardiometabolic disorders in the general population and of liver disease in NAFLD, likely through the modulation of lipotoxic free fatty acid (FFA) metabolism and of pro- and anti-inflammatory cytokine secretion.4-5 Consistently, abdominal adiposity excess, as quantified by magnetic resonance, correlated with liver fat in healthy subjects and with severity of inflammation and fibrosis in NASH.5, 6

Among dietary factors, excessive fat ingestion has been consistently linked to the pathogenesis of obesity, CVD, and T2DM in the general population7 and to liver injury in animal models of NASH,8 whereas the evidence connecting dietary fat excess to NAFLD in humans is controversial.

We hypothesized that adipose tissue dysfunction may mediate both liver disease progression and cardiometabolic risk in NAFLD, and that a maladaptive adipocyte response to dietary fat modulates liver injury and cardiometabolic risk in NAFLD, even in the absence of overall/abdominal adiposity excess. We therefore assessed the association of adipose tissue dysfunction, evaluated both in fasting conditions through two recently proposed indexes (the adipose tissue insulin resistance [adipose IR] index5 and the visceral adiposity index [VAI]9, 10) and dynamically after an oral fat challenge, with liver histology, glucose/lipoprotein metabolism, and markers of early atherosclerosis in nondiabetic, nonobese patients with biopsy-proven SS or NASH matched for overall/abdominal adiposity and traditional cardiometabolic risk factors.


adipose IR, adipose tissue insulin resistance; AI, adaptation index; ALT, alanine aminotransferase; ANOVA, analysis of variance; ApoA1, apolipoprotein A1; ApoE, apolipoprotein E; AUC, area under the curve; BMI, body mass index; CGI, CP-genic index; CI, confidence interval; CK-18, cytokeratin-18; CRP, C-reactive protein; CVD, cardiovascular disease; ER, endoplasmic reticulum; FFAs, free fatty acids; HDL-C, high-density lipoprotein cholesterol; hs-CRP, high-sensitivity CRP; iAUC, incremental area under the curve; ICAM-1, intercellular adhesion molecule 1; IR, insulin resistance; IV, intravenous; LDL-C, low-density lipoprotein cholesterol; MTP, microsomal triglyceride transfer protein; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score; NASH, nonalcoholic steatohepatitis; NT, nitrotyrosine; OGIS, oral glucose insulin sensitivity; OGTT, oral glucose tolerance test; OR, odds ratio; oxLDLs, oxidized low-density lipoproteins; SEM, standard error of the mean; SS, simple steatosis; T2DM, type 2 diabetes mellitus; Tg, triglycerides; TRLPs, triglyceride-rich lipoproteins; VAI, visceral adiposity index; VLDL, very-low-density lipoprotein.

Patients and Methods


Among 85 patients referred by family physicians to our hepato-metabolic clinic for chronic liver enzyme elevations, we selected 40 nonobese, nondiabetic, normolipidemic biopsy-proven NAFLD patients (20 with SS and 20 with NASH) who were matched for age, sex, body mass index (BMI), waist circumference, and metabolic syndrome features without clinical evidence of CVD. NASH was defined according to Brunt's definition,11 and histological features were scored according to NASH Clinical Research Network criteria.12 Besides histological evidence, the diagnosis of NAFLD required all the following criteria: persistently (>6 months) elevated liver enzymes; ultrasonographic bright liver without any other liver or biliary tract disease; negative viral markers; and a history of alcohol consumption <20 g/day in men and <10 g/day in women, as assessed by a detailed interview extended to family members and by a validated questionnaire filled in daily for 1 week by patients.

Exclusion criteria were as follows: obesity (BMI ≥30 kg/m2); diabetes (fasting plasma glucose ≥126 mg/dL or plasma glucose ≥200 mg/dL at +2 hours on oral glucose tolerance test [OGTT] or antidiabetic drugs); overt dyslipidemia (fasting serum cholesterol ≥200 mg/dL or plasma triglyceride [Tg] ≥200 mg/dL); exposure to occupational hepatotoxins or drugs known to be steatogenic, hepatotoxic, or to affect lipid/glucose metabolism; positive autoimmune or celiac disease markers; and abnormal serum α1-antitripsin, ceruloplasmin, or thyroid hormones. Mutations in the hemochromatosis genes, HFE and transferrin receptor 2, were detected in patients and controls using multiplex amplification reaction (Nuclear Laser Medicine, Milan, Italy). Liver iron concentration and hepatic iron index were assessed from 2 mg of dry-weight tissue by atomic absorption spectroscopy.

Of 103 healthy subjects enrolled in a population-based cohort study, 40 controls matched for age, gender, BMI, waist circumference, and metabolic syndrome features were randomly identified. To further rule out subclinical liver disease in controls, besides a negligible alcohol intake (<20 g/day in men and <10 g/day in women) and healthy abdomen ultrasound, the upper healthy alanine aminotransferase (ALT) limit was set at 30 (males) and 20 U/L (females).13

Patients and controls gave their consent to the study, which was conducted according to the Declaration of Helsinki.


Percent body fat was estimated by the bioelectrical impedance analysis method (TBF-202; Tanita, Tokyo, Japan), which was previously validated against dual X-ray absorption.14 Abdominal visceral fat area (cm2) was calculated using the equations developed by Stanforth et al. and validated against computed tomography in Caucasians.15

Adipose Tissue (Dys)Function Indices.

The adipose IR index was calculated as the product of fasting FFAs × fasting plasma insulin5; the VAI, incorporating BMI, waist circumference, fasting Tg, and high-density lipoprotein cholesterol (HDL-C) was calculated as previously described.10

Genetic Analyses.

Patients and controls were genotyped for microsomal triglyceride transfer protein (MTP) −493 G/T and apolipoprotein E (ApoE) polymorphisms, which are well-known modulators of lipoprotein metabolism and liver disease in NAFLD16 (see Supporting Appendix).

Alimentary Record.

Subjects filled in a daily dietary record for 1 week, according to the European Prospective Investigation into Cancer and Nutrition protocol, which was analyzed using the WINFOOD database (Medimatica, Teramo, Italy), as previously described.17

Inflammatory Markers and Cytokines.

Serum high-sensitivity C-reactive protein (hs-CRP), adiponectin, tumor necrosis factor-alpha, resistin, and leptin were measured as described in the Supporting Appendix.

Endothelial Dysfunction.

Soluble adhesion molecules E-selectin and intercellular adhesion molecule 1 (ICAM-1), which are validated markers of endothelial dysfunction and subclinical atherosclerosis,18 were measured (see Supporting Appendix).

Nitrosative Stress.

Plasma nitrotyrosine (NT) was chosen as a marker of nitrosative stress, which is involved in the pathogenesis of cardiovascular complications in insulin-resistant conditions and liver injury in NASH.19, 20 Fasting plasma NT was determined by a commercial enzyme-linked immunosorbent assay kit product by HyCult Biotechnology b.v. (sold in Italy by Pantec, Turin, Italy).

Oral Fat Load.

In westernized countries, people spend a substantial part of their life in the postprandial phase and postprandial lipemia is an emerging CVD risk factor.21 Within 2 weeks of completion of the alimentary record, participants underwent a 10-hour oral fat tolerance test, as previously described17 (see Supporting Appendix). Blood samples were drawn and immediately stored at −20°C every 2 hours for 10 hours. The following parameters were measured: (1) plasma total cholesterol, Tg, HDL-C, apolipoprotein A1 (ApoA1), and FFAs were measured by automated enzymatic methods; (2) Tg-rich lipoproteins (TRLPs) were isolated through preparative ultracentrifugation and subfractionated (see supporting Appendix); and (3) circulating oxidized low-density lipoproteins (oxLDLs), adipokines (e.g., resistin and adiponectin), and cytokeratin-18 (CK-18) fragments (a marker of hepatocyte apoptosis and hepatic necroinflammation in NASH3) were also measured (see Supporting Appendix).

OGTT-Derived Indices of Glucose Homeostasis.

After completion of the alimentary record, participants underwent a standard 75-g OGTT: Circulating glucose, insulin, and C-peptide were measured every 30 minutes and analyzed through the minimal model method to yield parameters of glucose homeostasis (see Supporting Appendix). Insulin sensitivity was estimated from a model of glucose clearance, which provides oral glucose insulin sensitivity (OGIS), an index of whole body insulin sensitivity, and muscle and hepatic insulin resistance (IR) index were calculated from the OGTT, as previously validated against a clamp in nondiabetic subjects.22, 23 The following indices of β-cell function were also calculated: the CP-genic index (CGI) and the integrated index of β-cell function adaptation index (AI), which relates β-cell insulin secretion to insulin sensitivity and represents an integrated parameter of β-cell function. The OGTT-derived AI was previously validated against the frequently sampled intravenous (IV) glucose tolerance test minimal model in NAFLD24 and nondiabetic subjects.25, 26 The incretin effect (i.e., the effectiveness of ingested glucose in stimulating β-cell insulin secretion, compared to IV glucose) was also calculated, as described in the Supporting Appendix.

Statistical Analysis.

Data are expressed as mean ± standard error of the mean (SEM). Differences across groups were analyzed by analysis of variance (ANOVA) and then by Bonferroni's correction, when variables were normally distributed; otherwise, Kruskal-Wallis' test, followed by Dunn's post-hoc test, was used to compare nonparametric variables. Normality was evaluated by Shapiro-Wilk's test. Fisher's exact test or the chi-square test were used to compare categorical variables, as appropriate.

Area under the curve (AUC) and incremental AUC (iAUC) of parameters measured during the oral fat test and the OGTT were computed by the trapezoid method. Multivariate repeated-measures ANOVA was used to test the interaction between time and group during the oral fat load test. When a significant interaction was found between factors, differences across groups were analyzed by ANOVA followed by Bonferroni's correction, if variables were normally distributed; otherwise, Kruskal-Wallis' test was performed, followed by Dunn's post-hoc test, to compare nonparametric variables. Differences were considered statistically significant at P < 0.05.

Analysis of dietary, anthropometric, and metabolic parameters and of genetic polymorphisms was made using Spearman's correlation test. Genetic polymorphisms were modeled as an additive effect, with quantitative predictor variables reflecting the number of risk alleles (0, 1, or 2). When a relation was found on univariate analysis, multiple regression analyses were used to estimate the relationship between different variables after log transformation of skewed data.

A logistic regression model was used to identify independent predictors for NASH and for advanced (i.e., stage 3) fibrosis. Based on univariate analysis, the covariates were OGIS, homeostasis model assessment, C-reactive protein (CRP), endothelial adhesion molecules, fasting and postprandial adiponectin/resistin/CK-18/FFAs, adipose IR index, and VAI.

Statistical analyses were done using STATISTICA software (version 5.1; StatSoft Italia, Padua, Italy).


Baseline Characteristics.

Compared to controls, NAFLD patients had higher adipose IR index, plasma endothelial adhesion molecules (e.g., E-selectin and ICAM-1), CRP, and NT and lower adiponectin levels. Compared to patients with SS, NASH had higher adipose IR index, endothelial adhesion molecules, and NT, but did not differ in any other anthropometric or metabolic parameter (Table 1). The percentage of patients with impaired glucose tolerance was higher in NASH than in SS, although the difference was not statistically significant.

Table 1. Baseline Characteristics of Patients With NASH, SS, and Controls
CharacteristicsControls (n = 40)Steatosis (n = 20)NASH (n = 20)P Value
  • Data are presented as mean ± SEM. Differences were considered statistically significant at P < 0.05. In the last column, P value of NASH versus simple steatosis is shown.

  • Abbreviations: BP, blood pressure; AST, aspartate aminotransferase; total C: total cholesterol; MetS, metabolic syndrome (according to ATP III criteria); impaired fasting glycemia, fasting plasma glucose ≥110 mg/dL; impaired glucose tolerance, plasma glucose 140-199 mg/dL at +2 hours on OGTT.

  • *

    P < 0.05 versus controls;

  • P < 0.01 versus controls.

Age, years50 ± 347 ± 447 ± 40.697
Sex, % males6866670.702
Smokers, %3436330.679
Family history of T2DM, %1515200.968
BMI, kg/m225.1 ± 1.625.1 ± 1.525.2 ± 1.60.729
Fat mass, %21 ± 322 ± 323 ± 20.451
Systolic BP, mmHg129 ± 4131 ± 6130 ± 70.668
Diastolic BP, mmHg85 ± 385 ± 587 ± 60.129
Waist, cm90 ± 490 ± 591 ± 50.372
Visceral fat area, cm291 ± 392 ± 593 ± 70.692
MTP −493 G/T, %    
ApoE genotype, %    
Tg, mg/dL80 ± 14105 ± 1692 ± 190.203
LDL-C, mg/dL112 ± 14122 ± 19108 ± 210.106
HDL-C, mg/dL50 ± 252 ± 351 ± 20.210
Glucose, mg/dL94 ± 595 ± 696 ± 50.693
Insulin, μU/mL5.3 ± 2.810.1 ± 4.812.0 ± 3.80.259
AST, U/L13 ± 242 ± 452 ± 40.101
ALT, U/L15 ± 497 ± 9117 ± 90.245
ICAM-1, mg/mL194.2 ± 8.3239.4 ± 8.2*279.1 ± 9.30.029
E-selectin, mg/mL18.5 ± 2.325.3 ± 2.445.9 ± 2.80.004
CRP1.2 ± 0.51.9 ± 1.1*2.7 ± 1.20.029
NT (mmol/mL)5.1 ± 4.916.1 ± 9.227.8 ± 15.30.012
Adipose IR, mol/L/pmol/L17.1 ± 1.949.5 ± 4.3*82.4 ± 8.20.0003
VAI1.15 ± 0.331.16 ± 0.281.19 ± 0.340.348
MetS, %3131310.999
Abdominal obesity, %3450.714
Hypertension, %4955570.801
Impaired fasting glycemia, %815200.693
Impaired glucose tolerance, %610250.405
High Tg, %1210110.999
Low HDL-C, %1415200.348
Histological steatosis, % hepatocytes34 ± 641 ± 70.432
Lobular inflammation1.8 ± 0.4
Hepatocellular ballooning1.7 ± 0.3
NAS score 2.0 ± 0.35.7 ± 0.50.0001
Fibrosis stage1.9 ± 0.6

Histological features of patients with NASH and SS are described in the Supporting Appendix.

Alimentary Record.

There was no difference in daily total energy and macro- or micronutrient (including type of fat) and antioxidant vitamin intake among controls, SS patients, and patients with NASH. Daily alcohol intake was also similar across groups: 11 ± 4 versus 13 ± 6 versus 15 ± 6 g (P > 0.5 for all comparisons).

Oral Fat Load.

ApoE and MTP −493 G/T genotype distribution did not differ between NASH, SS, and controls (Table 1).

Despite similar fasting values across the three groups, postprandial Tg, very-low-density lipoprotein (VLDL)1 and VLDL2 of both intestinal and hepatic origin, oxLDLs, and resistin responses progressively increased across SS and NASH. HDL-C and ApoA1 levels fell more consistently in NASH than in SS and were deeper in SS than in controls. Plasma low-density lipoprotein cholesterol (LDL-C) and insulin values did not significantly change from baseline during the test (Table 2; Figs. 1 and 2).

Table 2. Data Derived From Oral Fat Load Test
 Controls (n = 40)Steatosis (n = 20)NASH (n = 20)P Value
  • Data are presented as mean ± SEM. In the last column, P value of NASH versus simple steatosis is shown.

  • Abbreviation: LDL-CD, LDL-conjugated dienes.

  • *

    P < 0.05 versus controls;

  • P < 0.01 versus controls;

  • P < 0.001 versus controls.

Fasting Tg, mg/dL80 ± 14105 ± 1692 ± 190.2030
iAUC Tg, mg/dL × hours118 ± 46358 ± 69*451 ± 820.0160
Fasting FFAs, mMol/L465 ± 68768 ± 81*1,012 ± 2130.0090
iAUC FFAs, mMol/L × hours321 ± 3571,506 ± 541*9,311 ± 9170.0001
Fasting VLDL1-apoB48, mg/dL3.6 ± 1.04.1 ± 1.24.3 ± 1.20.3780
iAUC VLDL1-apoB48, mg/dL × hours5.7 ± 1.29.5 ± 2.318.7 ± 4.40.0060
Fasting VLDL2-apoB48, mg/dL1.3 ± 0.51.7 ± 0.51.9 ± 0.60.8860
iAUC VLDL2-apoB48, mg/dL × hours1.4 ± 0.73.4 ± 0.9*6.0 ± 1.10.0120
Fasting VLDL1-apoB100, mg/dL5.5 ± 1.05.8 ± 1.16.8 ± 1.60.2350
iAUC VLDL1-apoB100, mg/dL × hours3.6 ± 1.310.2 ± 2.323.0 ± 3.60.0020
FastingVLDL2-apoB100, mg/dL2.1 ± 0.72.3 ± 0.62.4 ± 0.60.3470
iAUC VLDL2-apoB100, mg/dL × hours (mg/dL x hr)(mg/dL x hr)1.7 ± 0.54.9 ± 1.89.9 ± 2.00.0030
Fasting LDL CD, uA 234 nm/uA 200 nm × 1006.4 ± 1.56.2 ± 1.46.7 ± 1.50.7190
iAUC LDL CD, uA 234 nm/uA 200 nm × 100 × hours1.9 ± 0.55.1 ± 1.7*8.5 ± 2.00.0100
iAUC HDL-C, mg/dL−19 ± 4−40 ± 5*−86 ± 60.0090
Fasting ApoA1, mg/dL109 ± 4107 ± 5110 ± 50.8810
iAUC ApoA1, mg/dL) × hours−1 ± 8−30 ± 9*−221 ± 140.0001
Fasting resistin, ng/mL3.9 ± 0.34.1 ± 0.54.1 ± 0.60.5410
iAUC resistin, ng/mL × hours1.0 ± 0.23.1 ± 0.7*6.9 ± 1.10.0040
Fasting adiponectin, ng/mL11,086 ± 1,8536,334 ± 9085,483 ± 9980.3920
iAUC adiponectin, ng/mL × hours8,273 ± 6261,231 ± 921−9,240 ± 1,5790.0001
CK-18 fragments, IU/L69 ± 9108 ± 10251 ± 110.0090
iAUC CK-18 fragments, UI/L × hours79 ± 891 ± 11195 ± 100.0070
Figure 1.

Oral fat load: postprandial responses in plasma triglycerides (A), VLDL1-apoB100 (B), VLDL2-apoB100 (C), VLDL1-apoB48 (D), VLDL2-apoB48 (E), and LDL conjugated dienes (F). Data are presented as mean ± SEM.

Figure 2.

Oral fat load: postprandial responses in plasma HDL-cholesterol (HDL-C) (A), apolipoprotein A1 (apoA1) (B), resistin (C), adiponectin (D), free fatty acids (FFA) (E) and cytokeratin-18 (CK-18) fragments (F). Data are presented as mean ± SEM.

Adiponectin significantly increased in controls postprandially. Despite similar fasting levels in SS and NASH, adiponectin increased postprandially in SS, whereas it slowly decreased in NASH. Plasma CK-18 levels increased postprandially in NASH, but not in SS and controls, indicating a significant apoptosis activation after fat ingestion in the former (Table 2; Fig. 2).

OGTT-Derived Parameters of Glucose Homeostasis.

Time course of circulating glucose, insulin, and C-peptide is shown in Fig. 3, and OGTT-derived parameters of glucose homeostasis are reported in Table 3. OGIS progressively declined and hepatic insulin resistance progressively increased across SS and NASH. Pancreatic β-cell function was similar between controls and SS, whereas NASH patients showed an impaired pancreatic β-cell function and a reduced incretin effect, compared to SS and controls. After excluding patients with impaired glucose tolerance, the difference in β-cell function between NASH and SS remained significant.

Figure 3.

Oral glucose tolerance test (OGTT): time course of circulating glucose (A), insulin (B) and C-peptide (C). Data are presented as mean ± SEM.

Table 3. OGTT-Derived Parameters of Glucose Homeostasis
ParametersControls (n = 40)Steatosis (n = 20)NASH (n = 20)P Value
  • Data are presented as mean ± SEM. Differences were considered statistically significant at P < 0.05. In the last column, P value of NASH versus simple steatosis is shown. CGI = ΔC-peptide30′/Δglucose30′ during the OGTT. AI was computed by multiplying CGI × OGIS. Hepatic extraction is the percentage of secreted insulin extracted by the liver.

  • *

    P < 0.05 versus controls;

  • P < 0.01 versus controls;

  • P < 0.001 versus controls.

OGIS, mL × min−1 × m−2459.1 ± 9.5403.0 ± 12.8*371.6 ± 11.70.041
Hepatic insulin resistance, g/dLglucose × μU/mLIns × min−22,841 ± 2064,803 ± 339*6,271 ± 5120.005
Muscle insulin sensitivity0.04 ± 0.010.02 ± 0.01*0.02 ± 0.01*0.623
Hepatic insulin extraction, %79 ± 680 ± 778 ± 90.556
CGI, ngC-pep × ginline image619 ± 47584 ± 38378 ± 290.009
AI, ngC-pep × ginline image × mL−1 × m−2284,121 ± 18,105231,852 ± 10,980144,398 ± 7,1560.006
Incretin effect, %75 ± 368 ± 455 ± 40.021

Correlative Analysis.

The main univariate correlation coefficients between anthropometric and metabolic variables are shown in Supporting Appendix Table 1.

Adipose IR index independently predicted plasma E-selectin (β = 0.50; 95% confidence interval [CI]: 0.45, 0.56; P = 0.009), ICAM-1 (β = 0.49; 95% CI: 0.44, 0.55; P = 0.010), and NT (β = 0.46; 95% CI: 0.42, 0.51; P = 0.021).

Similarly, iAUC adiponectin predicted E-selectin (β = −0.47; 95% CI: −0.42, −0.53; P = 0.010), ICAM-1 (β = −0.51; 95% CI: −0.46, −0.57; P = 0.008), and NT (β = −0.45; 95% CI: −0.40, −0.51; P = 0.025).

The NAFLD activity score (NAS) was independently predicted by adipose IR index (β = 0.49; 95% CI: 0.44, 0.54; P = 0.018) and CK-18 fragments (β = 0.48; 95% CI: 0.43, 0.53; P = 0.021).

On logistic regression analysis, the presence of NASH was independently predicted by adipose IR index (odds ratio [OR]: 2.0; 95% CI: 1.5-2.6; P = 0.019), iAUC adiponectin (OR: 0.5; 95% CI: 0.2-0.9; P = 0.020), and fasting CK-18 fragments (OR: 1.9; 95% CI: 1.2-2.5; P = 0.021).

Advanced fibrosis was predicted by adipose IR index (OR: 2.3; 95% CI: 1.8-2.9; P = 0.010) and iAUC adiponectin (OR: 0.5; 95% CI: 0.2-0.8; P = 0.004).

The results of multiple regression analysis of the oral fat load and OGTT parameters are reported in Supporting Appendix Tables 2 and 4, respectively.


The main findings of our study were the following: (1) Compared to matched SS patients, NASH patients had a higher adipose IR index and a more proinflammatory resistin/adiponectin acute response to fat ingestion, suggesting that (2) compared to matched patients with SS, NASH patients showed higher levels of endothelial adhesion molecules, a more proatherogenic lipoprotein profile, and a significant hepatocyte apoptosis activation after fat ingestion. NASH showed also more severe hepatic insulin resistance as well as impaired pancreatic β-cell function and incretin response. Altogether, these findings suggest that NASH has a greater potential for developing T2DM and CVD than SS, regardless of its association with obesity and metabolic syndrome. (3) Liver histology and most cardiometabolic abnormalities were predicted by adipose IR index and postprandial adiponectin response to fat, thereby placing adipose tissue dysfunction at the core of liver and cardiometabolic disease in NAFLD.

After fat ingestion, NASH patients showed a highly atherogenic postprandial lipoprotein profile and an increase in circulating markers of hepatocyte apoptosis, compared to SS. The higher postprandial TRLP elevation occurred despite similar fasting lipoprotein profile and steatosis severity, with the latter being the main determinant of hepatic VLDL1 secretion rate27: Though we cannot tell whether it depends on higher hepatic VLDL output or impaired TRLP catabolism or both, the inverse correlation of adiponectin response with postprandial lipoproteins and CK-18 fragments in the whole study population (Supporting Appendix Table 2) links adiponectin response to fat ingestion to postprandial hyperlipemia and hepatocyte apoptosis. Consistent with our findings, adiponectin extensively modulates lipid metabolism by promoting hepatic and muscle glucose and FFA oxidation and systemic TRLP clearance through lipoprotein lipase activation.28

Strikingly, adiponectin decreased postprandially in NASH, whereas it increased in controls and, to a lesser extent, in SS. This finding is in agreement with part, but not all, of the literature, depending on meal composition and test duration. A decrease in serum adiponectin similar to that observed in our NASH patients has been observed with high-fat, but not high-carbohydrate, meals, generally in diabetic, but also in nondiabetic, insulin-resistant subjects, starting after 4-6 hours subsequent to meal ingestion.29, 30 The plasma adiponectin fall has been shown to be preceded in vivo by a decrease in adipocyte adiponectin messenger RNA expression: Given the high abundance of adiponectin in the blood, as compared with other adipokines, and its long half-life (3-6 hours), it is likely that acute inhibition of adiponectin synthesis after meal ingestion may require a longer observation time to yield a significant plasma adiponection reduction, thereby explaining why most oral fat load tests currently used, which last ≤4 hours, have missed these changes.

Our data suggest that postprandial lipemia evokes a physiological “compensatory” increase in adiponectin secretion aiming at restoring baseline plasma lipid levels by enhancing FFA oxidation and TRLP catabolism. This response is progressively lost across different NAFLD stages and may contribute to liver injury (as suggested by postprandial CK-18 fragment elevation) and cardiometabolic derangement of these patients.

Mechanisms underlying failing adiponectin response to fat ingestion cannot be elucidated by our study. The ability of insulin to acutely suppress adiponectin gene expression and reduce circulating hormone in lean and obese subjects is well known,31, 32 but fasting insulinemia was similar between NASH and SS and did not significantly change during the oral fat load test. The amount and type of dietary fat regulates adiponectin secretion without affecting insulin levels33 by acting on nuclear sterol regulatory element-binding protein 1c,34 but dietary fat intake of our NASH and SS patients were similar. We speculate that a strong genetic background may underlie both the severe adipose tissue resistance to insulin antilipolytic action, leading to very high circulating FFA levels, and in the maladaptive response to fat ingestion, leading to an altered adiponectin/resistin response. Among novel candidate genes conferring susceptibility to obesity and its complications, the transcription factor, activator protein 2 beta, has been recently shown to promote adipocyte hypertrophy and IR and to down-regulate adiponectin expression postprandially, shifting adipokine expression toward a proinflammatory profile.35, 36 Another potential culprit is that the endoplasmic reticulum (ER) resident thiol protein, ERp44, a key mediator of ER stress, has been shown to regulate post-translational modifications of adiponectin, reducing adipocyte content and secretion of this adipokine.37, 38

NASH showed also more severely impaired hepatic insulin action and pancreatic β-cell function than SS. The correlation of hepatic IR and β-cell function indices with adipose IR index and postprandial adiponectin response (Table 4 and Supporting Appendix Table 2) suggest the absence of the protective action of adiponectin on β-cells and the rise in lipotoxic FFAs may synergistically impair hepatic insulin sensitivity and pancreatic β-cell function, further increasing the risk of T2DM, compared to SS.39, 40

Table 4. Multiple Regression Analysis: OGTT-Derived Parameters of Glucose Homeostasis Independently Predicted by Adipose IR and iAUC Adiponectin in Study Subjects (n = 80)
Dependent variableβ (95% CI)P Value
Adipose IR index
 OGIS−0.45 (−0.35, −0.72)0.025
 Hepatic insulin resistance0.45 (0.39, 0.53)0.020
iAUC adiponectin
 Hepatic insulin resistance−0.49 (0.42, 0.56)0.012
 CGI0.45 (0.41, 0.50)0.021
 AI0.46 (0.41, 0.51)0.018
 Incretin effect0.47 (0.42, 0.53)0.020

Lomonaco et al. have recently connected adipose IR index to glucose metabolism, traditional cardiovascular risk factors, and histological fibrosis in obese NAFLD patients with metabolic syndrome.41 We expanded the concept of adipose tissue dysfunction by showing that insulin-resistant and lipolytic adipocytes are also characterized by a maladaptive adipokine response to fat ingestion, which may be central for liver injury and cardiometabolic complications of NASH.42 These two features of dysfunctional adipocytes are detectable even in the absence of adiposity excess, which may subsequently derive from these functional changes: In fact, besides their systemic actions, adiponectin and resistin act in an autocrine and paracrine manner by enhancing FFA oxidation and adipocyte lipoprotein lipase activity,43 respectively: Therefore, the imbalance between these two adipokines may promote adipocyte enlargement and obesity.

Our study has strengths and limitations: The strengths are the careful selection of nonobese otherwise healthy biopsy-proven NAFLD subjects and their thorough cardiometabolic characterization; the limitations are the small number of patients and the cross-sectional design, which prevents any causal inference from the present study.

Future research should elucidate molecular mechanisms underlying adipocyte dysfunction and verify the appealing concept that a maladaptive adipocyte response to a chronic, daily, repetitive stress such as fat ingestion links chronic overfeeding to obesity and its complications. In the meantime, emerging evidence suggests that the restoration of adipocyte insulin sensitivity with thiazolinideniones and incretin analogs may synergistically benefit liver disease and associated cardiometabolic abnormalities in NAFLD.44-45

Our data provide also a rationale for prospectively evaluating simple, fasting adipose tissue dysfunction indices, such as adipose IR index, as a screening tool for individuating NAFLD patients at increased overall health-related risk in larger independent cohorts of unselected NAFLD patients.46, 47