Should Nonalcoholic Fatty Liver Disease Be Included in the Definition of Metabolic Syndrome?: A cross-sectional comparison with Adult Treatment Panel III criteria in nonobese nondiabetic subjects

A total of 197 Caucasians (87 women, age range 24–63 years, BMI range 19.9–29.9 kg/m²) were selected from a population-based cohort participating in previous institutional studies over the past 5 years (14). All were nondiabetic and nonobese, in good general health, and with normal findings on medical history, physical examination, blood count, and chemical screening battery. Insulin resistance was defined by a homeostasis model assessment of insulin resistance (HOMA-IR) index >2. The HOMA-IR index closely correlated with clamp measures in nondiabetic Northern Italian subjects (15) and with oral glucose tolerance test (OGTT)-derived indexes of insulin sensitivity in our subjects with NAFLD (G.M., unpublished data); furthermore, a cutoff value >2 predicted insulin resistance and increased cardiovascular risk in the general population independent of traditional risk factors (2–4,14,15).

NAFLD was diagnosed by persistently (>6 months) elevated aminotransferases (defined by alanine aminotransferase [ALT] ≥30 units/l in men and ≥20 units/l in women, based on recently proposed cutoff values, which increase the sensitivity for detection of NAFLD) (16,17) and ultrasonographic bright liver with no other liver disease; histological confirmation was available for 66% of the subjects. Exclusion criteria were a history of alcohol consumption >70 g/week (assessed by a detailed inquiry of patients and relatives and a validated questionnaire filled in daily for 1 week by the patients); diabetes (fasting plasma glucose ≥126 mg/dl or ≥200 mg/dl at +2 h with a standard oral glucose load on an OGTT); obesity (BMI ≥30 kg/m²); positive serum markers of viral disease; and exposure to occupational hepatotoxins or to drugs known to be steatogenic or to affect glucose metabolism.

Modified ATP III criteria for the diagnosis of metabolic syndrome were the following: hypertension (systolic/diastolic blood pressure ≥130/85 mmHg or receiving antihypertensive therapy); hypertriglyceridemia (fasting plasma triglycerides ≥150 mg/dl [1.7 mmol/l] or receiving lipid-lowering therapy; low plasma HDL cholesterol (<40 mg/dl [1.03 mmol/l] in men and <50 mg/dl [1.29 mmol/l]) in women]); impaired glucose regulation (impaired fasting glycemia, i.e., fasting plasma glucose ≥100 but <126 mg/dl [5.6–7.0 mmol/l] or impaired glucose tolerance, i.e., plasma glucose ≥140 mg/dl [7.8 mmol/l] at +2 h on an OGTT]; or abdominal obesity, modified according to ethnic-specific cutoff values to increase sensitivity for “metabolically obese” lean subjects (for Europeans, cutoff ranges were a waist circumference >94 cm [37 inches] in men and >80 cm [31 inches] in women) (12). A diagnosis of metabolic syndrome required fulfillment of at least three criteria.

Nitrosative stress is believed to be involved in diabetic endothelial dysfunction and cardiovascular complications (1) and in liver oxidative injury in NAFLD (18,19). Fasting plasma nitrotyrosine, as a marker of nitrosative stress, was determined by a commercial ELISA kit product by HyCult Biotechnology (Pantec, Turin, Italy).

Soluble adhesion molecules E-selectin, vascular cell adhesion molecule-1 (VCAM-1), and intercellular adhesion molecule 1 (ICAM-1) have been associated with endothelial dysfunction and early cardiovascular disease (20,21). Serum E-selectin, VCAM-1, and ICAM-1 levels were measured by a solid-phase ELISA (R&D Systems, Minneapolis, MN). Minimal detectable doses and intra- and interassay coefficients of variation (CVs) were, respectively, <0.1 ng/ml, 4.7–5.0%, and 7.4–8.8%, 0.17–1.26 pg/ml, 2.3–3.6%, and 5.5–7.8%, and <0.1 ng/ml, 4.7–5.0%, and 7.4–8.8%.

Serum tumor necrosis factor-α (TNF-α), leptin, and adiponectin were measured by a sandwich ELISA (R&D Systems Europe, Abingdon, UK). For TNF-α the kit has a sensitivity of 0.12 pg/ml in a 200-μl sample size and a range of 0.5 to 32 pg/ml. Intra- and interassay CVs were 5.9 and 12.6%, respectively. For adiponectin, the kit has a sensitivity of 0.25 pg/ml in a 50-μl sample size and a range of 3.9 to 250 ng/ml. The intra- and interassay CVs were 3.4 and 5.8%, respectively. Resistin was measured by a biotin-labeled antibody-based sandwich enzyme immunoassay (BioVendor Laboratory Medicine, Brno, Czech Republic).

Data are expressed as means ± SD. Differences between groups were analyzed by ANOVA when variables were normally distributed; otherwise (for triglycerides, insulin, HOMA-IR, adipokines, nitrotyrosine, and adhesion molecules), the Mann-Whitney test was used. A χ² test was used to compare categorical variables. Normality was evaluated by the Shapiro-Wilk test. For multiple comparisons, ANOVA and the Kruskal-Wallis test, followed by the Bonferroni correction or a Dunn test, were used, as appropriate. Spearman correlation coefficients were used to estimate the relationship between variables. Logistic regression analysis was applied when multiple associations were detected on univariate analysis. Differences were considered statistically significant if P < 0.05.

Sensitivity, specificity, positive and negative predictive value, and likelihood ratios of different criteria for the presence of insulin resistance were calculated. The positive likelihood ratio is the true-positive rate divided by the false-positive rate; the negative likelihood ratio is the inverse of the true-negative rate divided by the false-negative rate. For each parameter 95% CIs were provided.

Of the subjects, 50% were overweight (39% of insulin-sensitive and 71% of insulin-resistant subjects) (Table 1). The prevalence of NAFLD was 30% (8% in insulin-sensitive and 73% in insulin-resistant subjects). Insulin-resistant subjects also had higher circulating markers of oxidative stress and endothelial dysfunction and lower adiponectin levels.

The prevalence of different features of the metabolic syndrome and of NAFLD were higher in insulin-resistant subjects. Seventy-three percent of insulin-resistant versus 43% of insulin-sensitive subjects were hypertensive (P = 0.001), 54% of insulin-resistant versus 32% of insulin-sensitive subjects had abdominal obesity (P = 0.017), 33% of insulin-resistant versus 18% of insulin-sensitive subjects were hypertriglyceridemic (P = 0.054), 21% of insulin-resistant versus 2% of insulin-sensitive subjects had low HDL cholesterol, and 54% of insulin-resistant versus 31% of insulin-sensitive control subjects had impaired glucose regulation (P = 0.012). The prevalence of metabolic syndrome was 20% according to ATP III criteria (12).

Table 2 shows the relationship of different ATP III criteria and of NAFLD to insulin resistance. Hypertension and NAFLD were the most sensitive (73% for both) and had the highest negative predictive value (81 and 87%, respectively) for insulin resistance. The negative likelihood ratio (i.e., the extent to which the odds of insulin resistance decrease if the test result is negative) was greatest with hypertension (0.23) and NAFLD (0.14). HDL cholesterol and NAFLD were most specific for insulin resistance (98 and 92%, respectively) and had the highest positive predictive value (83 and 81%, respectively) and positive likelihood ratio (i.e., the odds of insulin resistance if the test result is positive: 5.00 and 4.39, respectively).

ATP III criteria had a positive predictive value of 62% for the presence of insulin resistance; the addition of NAFLD to ATP criteria yielded a prevalence of metabolic syndrome of 32%, with a positive predictive value for insulin resistance of 72%. The sensitivity of ATP III criteria for insulin resistance was 38%, but rose to 69% with the inclusion of NAFLD as a criterion (P = 0.002). The specificity of ATP III criteria for insulin resistance was 89% without NAFLD and 87% with NAFLD (P = 0.002). The positive likelihood ratio for insulin resistance increased from 1.64 to 2.53 when NAFLD was included (P = 0.010), whereas the negative likelihood ratio, i.e., the odds of insulin resistance in the absence of a certain condition, decreased from 35 to 18% with the inclusion of NAFLD in the diagnostic criteria of metabolic syndrome (P = 0.0002).

With subgrouping of insulin-resistant subjects according to the presence of NAFLD, patients with NAFLD had higher levels of nitrotyrosine and of soluble adhesion molecules than subjects without fatty liver (Table 3).

The main Spearman correlation coefficients are presented in Table 1 of the online appendix (available at http://dx.doi.org/10.2337/dc07-1526). HOMA-IR correlated with age, ALT (Fig. 1 of the online appendix), HDL cholesterol, triglyceride-to-HDL cholesterol ratio, C-reactive protein, adiponectin, nitrotyrosine, waist, adhesion molecules, and the number of ATP criteria.

On logistic regression analysis, HOMA-IR >2 was independently predicted by NAFLD, adiponectin, and nitrotyrosine (Table 4). Nitrotyrosine was independently predicted by NAFLD and C-reactive protein, NAFLD (β = 0.36; P = 0.008) and HOMA-IR independently predicted E-selectin, ICAM-1 was independently predicted by ALT and adiponectin, and adiponectin predicted VCAM-1 (Table 4).

By directly comparing NAFLD with ATP III criteria for metabolic syndrome in nonobese nondiabetic subjects, we found that 1) fatty liver, as diagnosed by ALT levels and ultrasonography, is more closely associated to insulin resistance than to ATP III criteria and 2) the presence of NAFLD in insulin-resistant subjects implies more severe systemic oxidative stress and endothelial dysfunction, independently of metabolic syndrome, adiposity, and adipokines.

The aim of the ATP III criteria is to identify individuals at increased risk for cardiovascular disease and diabetes to allow early treatment. Current ATP III criteria were selected because they tend to cluster together, share insulin resistance as their common denominator, and have individually been associated with an increased cardiovascular risk. However, they correlate weakly with the presence of insulin resistance, having a sensitivity and positive predictive value of 46 and 76%, respectively, in the general population (10). Because of their low sensitivity, many cases of insulin resistance remain undiagnosed, particularly in nonobese nondiabetic subjects, in whom the diagnosis of metabolic syndrome is less assisted by obesity and glucose criteria (10–13).

ATP III criteria consistently identified only 39% of insulin-resistant subjects in our population, with a positive predictive value of 62% and a positive likelihood ratio of 1.64. The presence of NAFLD alone doubled the sensitivity and significantly elevated positive predictive value and positive likelihood ratio, keeping the same specificity for insulin resistance (92% vs. 89%). The addition of NAFLD to the ATP III criteria improved their sensitivity by 72%, from 39 to 69%, with only a slight decrease in specificity, from 89 to 87%. Pathogenetically, these findings suggest that in nonobese nondiabetic subjects hepatic fat accumulation is more tightly related to insulin resistance than visceral adiposity, as estimated by waist circumference or any other feature of the metabolic syndrome, as defined by ATP III criteria.

The mechanism(s) underlying the association between NAFLD and insulin resistance are under investigation, but impaired hepatic lipid and lipoprotein handling and increased oxidative stress may enhance liver fat accumulation and lead to insulin resistance by nuclear factor-κB pathway activation (22–26). Increased nitrosative stress, in particular, seems to be operating in NAFLD even in the absence of insulin resistance (27).

The second intriguing finding relates to the additive information that NAFLD carries in the setting of insulin resistance. Patients with NAFLD displayed more severe oxidative stress and endothelial dysfunction than nonsteatotic insulin-resistant subjects, despite similar HOMA-IR and adiposity, and NAFLD independently entailed more severe oxidative stress and endothelial dysfunction.

Nitrotyrosine is an index of peroxynitrite formation, which is believed to play a key role in diabetic endothelial dysfunction and in liver oxidative injury in NAFLD (18,19). Mechanisms underlying increased oxidative stress in NAFLD cannot be elucidated in our study, but impaired mitochondrial β-oxidation, dietary saturated fat excess, and reduced antioxidant intake have been proposed (22,23,26). Increased oxidative stress can induce not only steatosis but also necroinflammation and fibrosis in NAFLD: high oxidative stress experimentally impaired VLDL secretion, leading to hepatocyte triglyceride accumulation (24,28); oxidation end products trigger the inflammatory cascade and extracellular matrix deposition, paralleling the severity of liver fibrosis in NAFLD (23).

Soluble adhesion molecule levels were higher in NAFLD than in insulin resistance without fatty liver, and fatty liver predicted increased E-selectin and ICAM-1 levels independently of HOMA-IR, adipokines, and visceral fat accumulation. Plasma concentrations of these molecules were consistently related to incident cardiovascular disease in apparently healthy individuals in large prospective studies (20,21). Our findings therefore suggest that NAFLD may be an early marker of endothelial dysfunction, independently of insulin resistance and traditional risk factors (7). Mechanism(s) linking NALFD to endothelial dysfunction are unclear, but impaired lipoprotein metabolism and oxidized LDL accumulation are potential candidates (22,23,26,27).

In nonobese, nondiabetic insulin-resistant subjects, the presence of NAFLD may therefore indicate a host of unsuspected derangements in oxidative balance and endothelial function, not routinely assessed, that contribute to increased cardiovascular disease risk: in day-to-day clinical practice, a diagnosis of NAFLD would simply require chronically elevated liver enzyme levels, an ultrasonographic bright liver, and exclusion of viral infection and of exposure to hepatotoxins (including alcohol) by interview of patient and relatives. Early identification of such patients at higher cardiometabolic risk may trigger earlier lifestyle and pharmacological interventions. Consistently, therapeutic measures in NAFLD also ameliorate insulin sensitivity and cardiovascular risk profile, the ultimate goal of a diagnosis of metabolic syndrome (29–31).

Further studies are required to test the validity of this proposal and to overcome the limitations of our study: its cross-sectional nature, which prevents any causal inference, and the absence of obese and diabetic subjects. However, the prevalence of NAFLD and insulin resistance is even higher in obesity and diabetes, so the association between NAFLD and insulin resistance could be further strengthened. Consistent with previous findings, lower sex-specific ALT cutoff levels identified insulin-resistant subjects more accurately (16,17). However, because liver fat content was not directly measured, some control subjects might have fatty liver despite normal ultrasound and liver enzyme levels; even so, this definition of NAFLD proved to be useful in clinical practice. Furthermore, misclassification of NAFLD would attenuate the magnitude of the difference in oxidative stress and endothelial dysfunction observed toward the null hypothesis, making our results a conservative estimate of the relationship between NAFLD, oxidative stress, and endothelial dysfunction.