Abstract
Objective:
Although several epidemiological studies have investigated associations between TNF-α and insulin resistance, results have been inconsistent. We studied the relationship between TNF-α and glucose tolerance status as part of the Insulin Resistance Atherosclerosis Study.
Research Design and Methods:
Serum concentrations of TNF-α were measured in 1558 individuals in a triethnic population across a spectrum of glucose tolerance. Insulin sensitivity and insulin secretion were assessed by a frequently sampled iv glucose tolerance test (FSIGT).
Results:
Compared with those with normal glucose tolerance, circulating levels of TNF-α were elevated in individuals with impaired glucose tolerance (IGT) and type 2 diabetes mellitus (T2D) after adjusting for age, gender, ethnicity, clinic site, and body mass index (3.3, 3.5, and 3.7 pg/ml in subjects with normal glucose tolerance, IGT, and T2D, respectively; P < 0.05). Age-, sex-, and body mass index-adjusted levels of TNF-α differed by ethnicity, with Hispanics having the highest levels and African-Americans having the lowest (4.1, 3.6, and 3.0 pg/ml in Hispanics, non-Hispanic whites, and African-Americans, respectively; P < 0.05). TNF-α was correlated with waist circumference, high-density lipoprotein, triglycerides, plasminogen activator inhibitor-1 and insulin sensitivity index (SI) (r = 0.22, −0.30, 0.35, 0.31, and −0.25; P < 0.0001); however, correlations varied by ethnicity. After adjusting for demographics and adiposity, individuals characterized by increased insulin resistance (lower SI), had higher levels of TNF-α than subjects characterized by high insulin sensitivity (3.8 and 3.3 pg/ml in subjects with an SI below/above the median at baseline; P < 0.0001). No differences were found for acute insulin response.
Conclusions:
We confirm that TNF-α is associated with IGT and T2D in a large, multiethnic population, independent of measures of adiposity. Adjusted values of TNF-α, as well as relationships between TNF-α and variables related to T2D, varied by ethnicity. Increased TNF-α levels were predominantly associated with insulin resistance but not with primary defects in β-cell function.
Type 2 diabetes mellitus (T2D) is an international health problem of epidemic proportions. The prevalence of T2D has doubled among middle-aged adults in the United States since the 1970s (1), with 439 million people (∼7.7% of the population) estimated to be affected worldwide by 2030 (2). Onset of T2D frequently results in the development of vasculature complications, including atherosclerosis and retinopathy, and is characterized by hyperglycemia, insulin resistance, and β-cell dysfunction with obesity, sedentary lifestyle, and increasing age all acting as major contributory factors. Chronic, subclinical inflammation has emerged as a characteristic feature of T2D (3), with elevated circulating levels of C-reactive protein (CRP) (4), white cell count (5), plasminogen activator inhibitor-1 (PAI-1) (6), IL-6 (4, 7), IL-1β (7), and IL-18 (8) reported to be predictive of T2D onset.
TNF-α is a proinflammatory cytokine induced in response to injury and infection that plays immunoregulatory roles in a variety of inflammatory disorders, including cardiovascular disease (9). TNF-α is expressed predominantly by monocytes and macrophages as a 26-kDa transmembrane protein that is cleaved from the cell surface into a soluble, 17-kDa form by the matrix metalloproteinase, TNF-α converting enzyme (TACE/ADAM17). TNF-mediated responses are facilitated through interaction with either of two distinct transmembrane receptors, designated TNF receptor (TNFR)-1 (55 kDa) and TNFR-2 (75 kDa), that transduce signals through the activation of multiple cell signaling pathways.
Among its diverse functions, TNF-α is known to affect insulin signaling, lipid metabolism, and adipoctye function (10), and several studies provide evidence that TNF-α plays central roles in various components of the metabolic syndrome, including obesity-induced insulin resistance (11). TNF-α levels in adipose tissue of human subjects are positively correlated with body mass index (BMI) and hyperinsulinemia (12) and inversely associated with lipoprotein lipase activity (13), and prolonged administration of TNF-α is reported to result in hyperinsulinemia (14).
Several cross-sectional, population-based epidemiological studies have found positive associations between TNF-α (15, 16) and/or its receptors (16) with insulin resistance or other components of the metabolic syndrome. A smaller number of studies have examined levels of TNF-α/TNFR across a range of glucose tolerance status, with discrepant results reported. In these studies, TNF-α levels were found to be elevated in individuals with impaired glucose tolerance (IGT) (17, 18), in T2D subjects only (19) or not different (20–22).
Many of these investigations, however, relied on small numbers of participants with IGT and T2D and were composed of populations homogenous in age and ethnicity. In addition, only a small fraction of these studies used detailed measurements of insulin resistance. Therefore, in the present study, we measured serum concentrations of TNF-α in 1558 African-American, Hispanic, and non-Hispanic white men and woman, aged 40–69 yr, representing a distribution of glucose tolerance status, and assessed insulin sensitivity and secretion by a frequently sampled iv glucose tolerance test to determine the relationship between TNF-α and demographic and metabolic variables associated with insulin resistance and T2D.
Subjects and Methods
Study subjects
The Insulin Resistance Atheroclerosis Study (IRAS) is a population-based epidemiological study initiated in 1992 that aims to investigate the relationships between insulin resistance, cardiovascular risk factors, and clinical and subclinical cardiovascular disease in a large multiethnic population across varying glucose tolerance status. A full description of the design and methods for IRAS has been reported previously (23). In brief, IRAS is comprised of 1625 participants at baseline including African-Americans, Hispanics, and non-Hispanic whites from four clinical centers in San Antonio (TX), San Luis Valley (CO), Oakland (CA), and Los Angeles (CA). The recruitment strategy aimed to include individuals with an approximately equal distribution of glucose tolerance status [normal glucose tolerance (NGT), IGT, and T2D], gender, and age (40–69 yr). The IRAS protocol was approved by the local institutional review committees and required two approximately 4-h visits to complete (23). The oral glucose tolerance test (OGTT) and the frequently sampled iv glucose tolerance test (FSIGT) were conducted during the first and second visits, respectively, scheduled at an interval of approximately 1 wk apart. All participants gave informed consent.
The current report includes data on 1558 subjects in whom TNF-α levels were assessed at baseline. Height and weight were measured to the nearest 0.5 cm and 0.1 kg, respectively. Age and ethnicity were assessed by self-report.
Laboratory measurements
Plasma glucose measurements were obtained by a glucose oxidase technique on an automated autoanalyzer (Yellow Springs Equipment, Yellow Springs, OH). A fasting blood sample was taken and followed by a 75-g OGTT. A second blood sample was taken after 2 h, and glucose tolerance status was based on the World Health Organization criteria (24).
IGT and impaired fasting glucose (IFG) status were defined based on World Health Organization (24) (IGT) and American Diabetes Association (ADA) (25) (IFG) criteria, and nondiabetic individuals were stratified as follows: normal fasting glucose (NFG)/NGT (NFG and normal postchallenge glucose tolerance): fasting glucose less than 109.8 mg/dl and 2-h glucose less than 140.4 mg/dl (n = 640); NFG/IGT (isolated postchallenge hyperglycemia): fasting glucose less than 109.8 mg/dl and 2-h glucose between 140.4 and 199.8 mg/dl (n = 236); IFG/NGT (isolated fasting hyperglycemia): fasting glucose between 109.8 and 126.0 mg/dl and 2-h glucose less than 140.4 mg/dl (n = 59); IFG/IGT (combined fasting and postchallenge hyperglycemia): fasting glucose between 109.8 and 126.0 mg/dl and 2-h glucose between 140.4 and 199.8 mg/dl (n = 96). Subjects were also categorized using the 2003 ADA criteria of IFG (IFG = 100–125 mg/dl) (26).
Insulin sensitivity and insulin secretion were assessed by an FSIGT on a separate day from the OGTT, and serum insulin was measured by the dextran-charcoal RIA with a coefficient of variation of 19% (n = 163 split pairs). A modified protocol (27) that used an injection of insulin rather than tolbutamide (28) and a reduced-sampling protocol that used 12 plasma samples rather than 30 (29) was adopted. Insulin sensitivity is expressed as the insulin sensitivity index (SI), calculated by mathematical modeling methods (MINMOD, version 3.0, 1994; courtesy of Richard Bergman, Ph.D., Los Angeles, CA). Acute insulin response (AIR) was calculated as the mean increase in plasma insulin concentration above baseline, measured at 2 and 4 min after glucose administration.
TNF-α was measured at the Laboratory for Clinical Biochemistry Research (University of Vermont) in citrated plasma using a Quantikine HS Human TNF-α immunoassay (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions. All samples were analyzed in duplicate. The analytic coefficient of variation ranged from 8.4 to 11.8%. CRP, fibrinogen, and PAI-1 were measured as reported previously (30, 31).
Statistical analysis
All statistical analyses were conducted using the Statistical Analysis System (SAS, version 9.2; SAS Institute, Inc., Cary, NC). Descriptive data were stratified by quartile of TNF-α (percentage; mean values ± se) and tested for linear trends. Sensitivity of the findings to the use of quartiles was assessed by redoing the analyses using quintiles, and results were essentially identical.
The distribution of TNF-α was highly skewed, so a natural logarithm transformation was used to make it more approximately normal; back-transformed results (95% confidence interval) are presented in the tables. Differences in the concentrations of TNF-α for participants with NGT, IGT, and T2D were assessed at baseline in the overall population and stratified by ethnicity using an ANOVA, with a Tukey test used to maintain the overall type I error rate at 5% for post hoc pairwise comparisons. The models were then adjusted for the demographic and diabetes-related covariates age, gender, clinic site, smoking, BMI, or waist circumference and, in analyses of the overall population, ethnicity. Differences in TNF-α were also evaluated by gender, smoking status, and presence of hypertension in the overall population by Student's t test (for gender and hypertension) and ANOVA (for smoking status), respectively.
To determine whether TNF-α levels were more closely related to postglucose load or to fasting glycemia, nondiabetic individuals were stratified into the four different glucose tolerance categories described above. Differences in levels of TNF-α were compared among groups by ANOVA followed by Tukey for post hoc comparisons in the overall population. Additional analyses were performed, adjusting the models for the demographic covariates age, gender, ethnicity, and clinic site in one model and these variables plus BMI or waist circumference in a separate model.
Next, unadjusted partial Spearman rank correlations were estimated for TNF-α and diabetes-related variables in the overall population and stratified by glucose tolerance status. Partial Spearman correlations were then estimated after adjusting for age, gender, clinic site, ethnicity, and BMI or waist circumference in one analysis and these variables plus SI in a separate analysis. Similar analyses were also performed stratified by ethnicity. The z test for correlated correlation coefficients was used to compare correlations of different variables (32). Comparison of correlation coefficients by ethnicity was performed using the method described by Zar (33).
Subjects were stratified into groups based on low and high insulin resistance (defined by an SI below (n = 823) and above (n = 735) the median in the overall population at baseline; 1.1 × 104 minutes−1 × microunits−1 × milliliters−1) and by low (n = 760) and high (n = 764) first-phase insulin secretion (defined as an AIR above/below the overall population median at baseline; 37.5 μU/ml). After adjustment for covariates, an ANOVA was used to test for differences in lnTNF-α between groups overall and stratified by ethnicity.
Linear regression models were fit with lnTNF-α as the dependent variable. Age gender, ethnicity, smoking status, and clinic site were first entered into the model. Smoking status was not significant so was removed. Next, waist circumference, triglycerides, low-density lipoprotein (LDL), high-density lipoprotein (HDL), systolic and diastolic blood pressures were separately entered into the model that contained age, gender, ethnicity, and clinic site to obtain their partial R2 values. Only variables that had a P < 0.05 were retained. All significant variables were included in the final model to obtain the total R2. In subsequent analyses, SI, PAI-1 and 2-h glucose were included into the model as additional independent variables. Models were then analyzed and stratified by ethnicity.
Finally, similar linear regression analyses were used to model SI. Age, sex, ethnicity, and BMI or waist circumference were entered first. Next, TNF-α was entered separately followed by CRP, fibrinogen, and PAI-1. Subsequently, triglycerides, HDL, LDL, and systolic and diastolic blood pressures were entered separately.
Results
The median TNF-α concentration in the overall population was 3.6 pg/ml (interquartile range 2.75–4.48 pg/ml) and ranged from 1.1 to 18.8 pg/ml. Table 1 shows the distribution of diabetes-related variables in the IRAS cohort at baseline by quartiles of TNF-α. Participants with higher levels of circulating TNF-α were older, had higher body mass (BMI and waist), had increased prevalence of hypertension, higher levels of fasting, and 2-h glucose, had higher fasting and 2-h insulin, and had higher insulin resistance (lower SI), lower levels of HDL, higher levels of triglycerides, and higher circulating levels of the inflammatory and fibrinolytic proteins (CRP, fibrinogen, and PAI-1) (all P < 0.0001). No differences were observed in the levels of TNF-α by gender, smoking status, total cholesterol, LDL, or AIR.
Table 1.
Characteristics of the study population at baseline stratified by quartiles of TNF-α
| TNF-α quartiles | 1 | 2 | 3 | 4 | Pa |
|---|---|---|---|---|---|
| n | 386 | 401 | 385 | 386 | |
| TNF-α (pg/ml) | 2.2 (0.02) | 3.2 (0.01) | 4.0 (0.01) | 5.9 (0.09) | |
| Demographics | |||||
| Age (yr) | 53.6 (0.4) | 55.3 (0.4) | 55.6 (0.4) | 57.9 (0.4) | <0.0001 |
| Gender (male/female, %) | 41/59 | 47/53 | 45/55 | 45/55 | NS |
| Ethnicity, n (%) | <0.0001 | ||||
| Non-Hispanic whites | 130 (34) | 164 (41) | 149 (39) | 139 (36) | |
| African-American | 185 (48) | 116 (29) | 82 (21) | 58 (15) | |
| Hispanic | 71 (18) | 121 (30) | 154 (40) | 189 (49) | |
| Smoking (never/former/current, %) | 45/39/18 | 45/37/18 | 45/37/17 | 40/43/18 | NS |
| Obesity | |||||
| BMI (kg/m2) | 28.0 (0.3) | 29.0 (0.3) | 30.3 (0.3) | 30.6 (0.3) | <0.0001 |
| Waist circumference (cm) | 89.0 (0.6) | 92.2 (0.6) | 95.9 (0.7) | 96.8 (0.7) | <0.0001 |
| Hypertension (%, yes/no) | 31/69 | 37/63 | 41/59 | 46/54 | <0.0001 |
| Glucose/insulin | |||||
| Fasting glucose (mg/dl) | 113.8 (2.1) | 119.6 (2.5) | 131.1 (2.9) | 128.8 (2.5) | <0.0001 |
| Two-hour glucose | 160.2 (4.6) | 178.1 (5.1) | 204.9 (5.7) | 204.1 (5.6) | <0.0001 |
| Fasting insulin (μU/ml) | 15.9 (0.9) | 16.3 (0.8) | 19.9 (0.8) | 21.4 (0.6) | <0.0001 |
| Two-hour insulin | 82.4 (4.0) | 95.7 (4.8) | 101.7 (4.4) | 120.5 (5.3) | <0.0001 |
| Proinsulin (pmol/liter) | 7.1 (0.7) | 8.4 (0.5) | 11.1 (0.6) | 14.1 (0.9) | <0.0001 |
| Proinsulin/insulin | 0.50 (0.03) | 0.67 (0.08) | 0.69 (0.04) | 0.68 (0.03) | <0.05 |
| SI (×104/min−1 × μU−1 × ml−1) | 2.1 (0.1) | 1.9 (0.1) | 1.3 (0.09) | 1.2 (0.09) | <0.0001 |
| AIR (μU/ml) | 52.5 (2.8) | 52.8 (2.6) | 52.0 (2.4) | 56.1 (2.8) | NS |
| AIR (adjusted for SI) | 53.8 (2.7) | 54.5 (2.7) | 52.0 (2.7) | 55.2 (2.7) | NS |
| Lipids | |||||
| Total cholesterol (mg/dl) | 209.5 (1.9) | 212.9 (2.4) | 217.0 (2.2) | 211.3 (2.2) | NS |
| LDL (mg/dl) | 141.8 (1.8) | 141.6 (1.7) | 144.1 (1.9) | 137.7 (1.9 | NS |
| HDL (mg/dl) | 51.0 (0.7) | 45.9 (0.7) | 42.0 (0.7) | 40.5 (0.7) | <0.0001 |
| Triglycerides (mg/dl) | 108.8 (3.5) | 141.3 (4.6) | 172.3 (7.7) | 189.0 (7.0) | <0.0001 |
| Coagulation/inflammation | |||||
| PAI-1 (ng/ml) | 18.6 (1.0) | 22.8 (0.9) | 28.0 (1.0) | 31.7 (1.4) | <0.0001 |
| Fibrinogen (mg/dl) | 273.3 (2.9) | 277.7 (3.0) | 287.3 (2.8) | 291.8 (3.1) | <0.0001 |
| CRP (mg/liter) | 3.1 (0.31) | 3.4 (0.2) | 4.4 (0.3) | 5.5 (0.4) | <0.0001 |
Data are proportions (percentage), means (±se). NS, Nonsignificant.
P value for trend across TNF-α categories.
Circulating levels of TNF-α were significantly different by glucose tolerance status in the overall population (Table 2). These differences remained after adjusting for age, gender, ethnicity, clinic, and BMI or waist circumference. Subgroup analysis (adjusted for age and gender) showed that levels of TNF-α differed significantly by ethnicity, with Hispanics having the highest levels and African-Americans having the lowest (4.1, 3.6, and 3.0 pg/ml in Hispanics, non-Hispanic whites, and African-Americans, respectively; P < 0.05 for all comparisons). Differences in the levels of TNF-α by ethnicity were not explained by BMI, waist circumference, or SI. No interactions were found for ethnicity on the relationship between TNF-α and glucose tolerance status.
Table 2.
Concentrations of lnTNF-α overall and stratified by ethnicity in subjects with NGT, IGT, and T2D at baseline
| TNF-α (pg/ml) | NGT | IGT | T2D | P |
|---|---|---|---|---|
| Overall | ||||
| n | 696 | 354 | 507 | |
| TNF-α | 3.3 (3.2–3.4) | 3.6 (3.5–3.8)a | 3.9 (3.7–4.0)a,b | <0.0001 |
| Non-Hispanic whites | ||||
| n | 283 | 134 | 165 | |
| TNF-α | 3.3 (3.1–3.4) | 3.7 (3.5–3.9)a | 4.1 (3.9–4.3)a | <0.0001 |
| African-American | ||||
| n | 179 | 99 | 163 | |
| TNF-α | 2.7 (2.6–3.0) | 3.0 (2.9–3.3) | 3.3 (3.0–3.4)a | <0.05 |
| Hispanic | ||||
| n | 234 | 121 | 179 | |
| TNF-α | 3.7 (3.6–3.9) | 4.1 (3.8–4.3) | 4.5 (4.1–4.7)a,b | <0.0001 |
Data are backtransformed means (95% confidence interval).
P < 0.05 compared with NGT using Tukey's test.
P < 0.05 compared with IGT using Tukey's test.
Although levels of TNF-α (adjusted for age, sex, clinic site, and BMI or waist circumference) were higher in subjects with T2D compared with those with NGT in all three ethnic populations, significant differences between subjects with NGT and IGT were observed only in non-Hispanic whites, and differences between subjects with IGT and T2D were observed only in Hispanics (Table 2).
Compared with those with normal glucose homeostasis (NFG/NGT), nondiabetic subjects with isolated postchallenge hyperglycemia (NFG/IGT), or combined fasting and postchallenge hyperglycemia (IFG/IGT) had higher levels of TNF-α in the overall population (Table 3). Adjusting the model for age, gender, ethnicity, and clinic site did not alter the interpretation of the results. However, differences were no longer statistically significant after adjusting further for BMI, waist circumference, or SI (P = 0.27, 0.30, and 0.07, respectively). After stratifying by ethnicity, statistically significant differences were observed only between the NFG/NGT and NFG/IGT groups in non-Hispanic whites and the IFG/NGT and IFG/IGT groups in Hispanics; no differences were observed in African-Americans. Differences between NFG/IGT and the other subgroups were not observed when the analyses were performed using the 2003 ADA criteria of IFG.
Table 3.
Concentrations of lnTNF-α in nondiabetic subjects according to fasting and 2-h glucose tolerance status at baseline
| Model | TNF-α (pg/ml) | NFG/NGT | NFG/IGT | IFG/NGT | IFG/IGT | P |
|---|---|---|---|---|---|---|
| Overall | n | 640 | 236 | 59 | 96 | |
| TNF-α | 3.3 (3.2–3.4) | 3.6 (3.5–3.8)a | 3.5 (3.0–3.9) | 3.7 (3.4–4.0)a | 0.001 | |
| Non-Hispanic whites | n | 261 | 94 | 27 | 29 | |
| TNF-α | 3.3 (3.1–3.4) | 3.7 (3.4–4.0)a | 3.5 (2.9–4.3) | 3.7 (3.3–4.2) | 0.03 | |
| African-American | n | 160 | 50 | 15 | 40 | |
| TNF-α | 2.8 (2.7–3.0) | 3.0 (2.7–3.3) | 2.9 (2.1–3.5) | 3.4 (2.9–3.7) | 0.11 | |
| Hispanic | n | 219 | 92 | 17 | 27 | |
| TNF-α | 3.7 (3.6–3.9) | 3.9 (3.7–4.2) | 4.2 (3.3–5.2) | 4.4 (3.9–5.1)a | 0.04 |
Data are backtransformed means (95% confidence interval).
P < 0.05 compared with NFG/NGT using Tukey's test.
Correlations were found between TNF-α and waist circumference, fasting insulin, proinsulin, SI, HDL, triglycerides, PAI-1, and CRP in the overall population (Table 4). Adjusting the correlations for age, gender, clinic site, ethnicity, and BMI reduced the magnitude, but not the significance, of the correlation coefficients. For example, the adjusted correlation coefficients for fasting insulin, proinsulin, SI, HDL, triglycerides, PAI-1, and CRP were 0.17, 0.24, −0.17, −0.19, 0.21, 0.23, and 0.20, respectively (all P < 0.0001). Adjusting further for SI modestly reduced the magnitude of the results. For example, the partial correlation coefficients for HDL, triglycerides, PAI-1, and CRP were −0.17, 0.18, 0.19, and 0.17, respectively (all P < 0.0001).
Table 4.
Spearman correlation coefficients for TNF-α and metabolic syndrome-related variables overall and stratified by ethnicity
| Variable | Overall | White | AA | Hispanic |
|---|---|---|---|---|
| Demographic | ||||
| Age (yr) | 0.19a | 0.21a | 0.18b | 0.23a |
| BMI (kg/m2) | 0.16a | 0.24a | 0.12c | 0.23a |
| Waist (cm) | 0.22a | 0.28a | 0.21a | 0.24a |
| Blood pressure (mm Hg) | ||||
| Systolic | 0.07c | 0.07 | 0.13b | 0.19a |
| Diastolic | 0.02 | 0.04 | 0.06 | 0.04 |
| Insulin/glucose | ||||
| Fasting glucose (mg/dl) | 0.17a | 0.23a | 0.14b | 0.25a |
| Two-hour glucose | 0.19a | 0.24a | 0.15b | 0.24a |
| Fasting insulin (μU/ml) | 0.25a | 0.30a | 0.12c | 0.30a |
| Two-hour insulin | 0.17a | 0.18a | 0.12c | 0.18a |
| Proinsulin (pmol/liter) | 0.28a | 0.34a | 0.24a | 0.33a |
| Insulin secretion/resistance | ||||
| SI (×104/min−1 × μU−1 × ml−1) | −0.25a | −0.31a | −0.18a | −0.28a |
| AIR (μU/ml) | 0.02 | 0.007 | 0.11 | −0.05 |
| AIR (adjusted for SI) | 0.02 | 0.008 | 0.03 | −0.03 |
| Proinsulin/insulin | 0.14a | 0.12b | 0.15b | 0.21a |
| Lipids | ||||
| Total cholesterol (mg/dl) | 0.03 | 0.04 | 0.11c | 0.008 |
| HDL | −0.30a | −0.31a | −0.19a | −0.20a |
| LDL | −0.03 | −0.02 | 0.09 | −0.03 |
| Triglycerides (mg/dl) | 0.35a | 0.31a | 0.28a | 0.26a |
| Inflammation/coagulation | ||||
| Fibrinogen (mg/dl) | 0.14a | 0.15b | 0.11 | 0.20a |
| PAI-I (ng/dl) | 0.31a | 0.33a | 0.16b | 0.22a |
| CRP (mg/liter) | 0.27a | 0.26a | 0.25a | 0.33a |
AA, African-American.
P < 0.0001.
P < 0.005.
P < 0.05.
Correlations of TNF-α with waist circumference were larger than with BMI (r = 0.22 and 0.16 in the overall population, respectively; P < 0.001 for difference). Adjusted analyses using waist in place of BMI resulted in correlation coefficients of 0.14, 0.17, −0.13, −0.27, 0.29, 0.25, and 0.18 for fasting insulin, proinsulin, SI, HDL, triglycerides, PAI-1, and CRP, respectively (all P < 0.0001). Analyses stratified by ethnicity revealed significantly smaller correlations in African-Americans for TNF-α and fasting insulin (r = 0.12; P < 0.05 compared with whites and Hispanics) and PAI-1 (r = 0.16; P = 0.02 compared with whites) and a nonsignificant trend toward lower correlations with SI and HDL (r = −0.18 and −0.19; P < 0.005 for correlation coefficients).
Individuals who were characterized by a lower SI (i.e. were more insulin resistant) had higher levels of TNF-α than subjects with higher SI (3.8 and 3.3 pg/ml in subjects stratified by an SI below/above the median in the overall population at baseline, respectively; P < 0.0001). This difference was unaffected by adjusting for age, sex, ethnicity, clinic, and BMI or waist circumference. No differences, however, were observed in individuals stratified by AIR (a measure of first-phase insulin secretion) (3.5 and 3.5 pg/ml in subjects stratified by AIR below/above the overall population median at baseline, respectively; P = 0.93). Adjusting AIR for SI did not alter the results. Results were consistent across ethnicity.
Multivariate linear regression analyses used to model biological predictors of TNF-α showed that after forcing age, gender, race, and clinic site into the model, waist circumference, HDL (inversely), and triglycerides were all independently associated with lnTNF-α in the overall population (Table 5). In subsequent models that included SI, 2-h glucose, and PAI-1, SI, and 2-h glucose were excluded, and PAI-1 did not explain additional variance. Stratification by ethnicity revealed that HDL explained significantly more variance in whites compared with the other groups (partial r2 = 8.0, 1.9, and 3.7% in whites, African-Americans, and Hispanics, respectively; P < 0.05 compared with whites). Waist circumference explained less variance in African-Americans compared with whites and Hispanics (partial r2 = 4.0, 7.7, and 7.1%, respectively), although this observation was not statistically significant.
Table 5.
Multiple linear regression analyses with lnTNF-α as the dependent variable
| Ethnicity | Independent variable | Β | se (Β) | P | Partial R2 (%) |
|---|---|---|---|---|---|
| Overall | Waist circumference | 0.007 | 0.0007 | <0.0001 | 5.8 |
| HDL | −0.006 | 0.0007 | <0.0001 | 4.3 | |
| Triglycerides | 0.0006 | 0.00007 | <0.0001 | 3.8 | |
| Non-Hispanic white | Waist circumference | 0.008 | 0.001 | <0.0001 | 7.7 |
| HDL | −0.008 | 0.001 | <0.0001 | 8.0 | |
| Triglycerides | 0.0006 | 0.0001 | <0.0001 | 5.4 | |
| African-American | Waist circumference | 0.006 | 0.001 | <0.0001 | 4.0 |
| HDL | −0.004 | 0.001 | 0.002 | 1.9 | |
| Triglycerides | 0.001 | 0.0003 | <0.0001 | 3.8 | |
| Hispanic | Waist circumference | 0.008 | 0.001 | <0.0001 | 7.1 |
| HDL | −0.005 | 0.0007 | <0.0001 | 3.7 | |
| Triglycerides | 0.0005 | 0.0001 | <0.0001 | 3.3 |
After forcing age, sex, clinic site, and ethnicity into the model, waist circumference, triglycerides, HDL, LDL, and systolic and diastolic blood pressures were analyzed as independent variables. Only variables that had a P < 0.05 were considered in the final fitted model. R2 = 23.6% in the overall population; 21.9% in non-Hispanic whites; 10.3% in African-Americans; and 14.6% in Hispanics.
In multivariate linear regression models of SI (after forcing age, sex, ethnicity, clinic site, and BMI into the model), TNF-α was independently associated with SI in the overall population [β = −0.44 (±0.12); P = 0.0002]. Adjusting for waist circumference in place of BMI did not alter the results. When CRP, fibrinogen, and PAI-1 were added, TNF-α, fibrinogen, and PAI-1 remained significant [β = −0.29 (±0.12), −0.01 (±0.002), and −0.002 (±0.0008) for TNF-α, PAI-1, and fibrinogen, respectively; P < 0.05], and CRP was not significant. Adjusting for waist circumference in place of BMI excluded TNF-α from this model. Addition of HDL and triglycerides, two variables associated with TNF-α (as shown in Table 5) into the model containing demographics, BMI, PAI-1, and fibrinogen, resulted in TNF-α no longer being statistically significant. In analyses stratified by ethnicity (adjusted for demographics and waist circumference) TNF-α was independently associated with SI in non-Hispanic whites only [β = −0.54 (±0.20), −0.32 (±0.21) and −0.19 (±0.21) in whites, African-Americans, and Hispanics, respectively]. Adjusting for demographics and BMI yielded similar results.
Discussion
In our study of a triethnic population representing a wide distribution of glucose tolerance status, we have found that circulating levels of TNF-α were higher in participants with IGT and T2D compared with those who were neither. The associations between TNF-α and IGT/T2D remained after adjustments for possible confounders, such as demographics and adiposity. Significant differences in TNF-α levels were observed by ethnicity (4.1, 3.6, and 3.0 pg/ml in Hispanics, non-Hispanic whites, and African-Americans, respectively). TNF-α was correlated with waist circumference, HDL, triglycerides, and PAI-1, and the results suggested that TNF-α was more closely associated with increased insulin resistance than with defects in β-cell function.
Although several cross-sectional, population-based studies have examined associations between TNF-α (15, 16) with insulin resistance, or other components of the metabolic syndrome, only a small number of studies have measured these levels across a range of glucose tolerance status, and these reports have resulted in mixed outcomes. Our results are consistent with previous studies reporting elevated levels of TNF-α in individuals with IGT and T2D (17, 18) and is unique in regard to its multiethnic population, wide age range, and direct assessment of insulin resistance and secretion using the FSIGT. Inconsistencies between our results and those of some other studies (20–22) are likely explained by the limited power to detect differences in cohorts with smaller numbers of participants or ethnic differences in the sample populations.
Circulating levels of TNF-α were found to differ by ethnicity, with Hispanics having the highest levels and African-Americans having the lowest. Adjusting for BMI, waist circumference, or SI did not alter the interpretation of the results, and these differences are likely explained by additional factors, such as racial genetics, shared environments, etc. Other studies have also reported differences in TNF-α levels by ethnicity (34–36), with nonobese, nondiabetic Hispanics reported to have higher levels than non-Hispanic whites (34) and nondiabetic, overweight white woman having higher levels than similarly matched African-Americans (35). In the latter study, higher levels of TNF-α were attributed to differences in intraabdominal adipose tissue. Greater intraabdominal adipose tissue in whites was also associated with a more substantial impact of weight loss on reducing TNF-α levels in nondiabetic white woman compared with nondiabetic African-Americans (36). Additional investigations in multiethnic populations will be required to confirm the present findings.
TNF-α has been demonstrated to contribute to insulin resistance by a variety of mechanisms in animal models and cell-based studies, including inhibition of insulin receptor signaling, inhibition of glucose transport, and regulating lipid metabolism (10, 11). Our results are consistent with other population-based studies that have found associations between TNF-α and insulin resistance (15, 16, 18) and indicate that TNF-α is more closely related to increased insulin resistance rather than decreased insulin secretion. TNF-α was significantly correlated with fasting insulin, proinsulin, and SI but not with markers of insulin secretion, including AIR (Table 4). TNF-α levels were also found to be higher in individuals who were more insulin resistant, defined by an SI below the study population median, whereas no differences were found in those with low vs. high first-phase insulin secretion. These results are consistent with earlier reports in the IRAS cohort demonstrating that prediabetic individuals with decreased insulin sensitivity had higher levels of the inflammatory and fibrinolytic proteins CRP and PAI-1, compared with subjects with primary defects in first-phase insulin secretion (37), and findings that PAI-1 levels were elevated in subjects with T2D who were more insulin resistant (38).
TNF-α provided the first piece of molecular evidence mechanistically linking obesity, inflammation, and insulin resistance (39). In obese human subjects, TNF-α mRNA levels have been reported to be elevated in adipose tissue (12, 13), and several studies have postulated that TNF-α plays causal roles in obesity-mediated insulin resistance (11). Of interest in the present study was the finding that correlations of TNF-α with waist circumference were larger than those with BMI (r = 0.22 vs. 0.16 for waist circumference and BMI in the overall population, respectively; P < 0.001 for difference). The correlations of TNF-α and waist circumference were consistent across ethnicities (r = 0.28, 0.21, and 0.24 for non-Hispanic whites, African-Americans, and Hispanics, respectively), whereas correlations with BMI showed nonsignificant trends for being lower in African-Americans (r = 0.24, 0.12, and 0.23, respectively). This finding is consistent with reports in the literature suggesting that activation of the TNF-α system is more closely related to visceral adiposity than to sc fat (40, 41), which may be important given that abdominal-specific obesity is closely associated with insulin resistance (42).
PAI-1 production is also increased in obesity, and it is well appreciated that TNF-α up-regulates the expression of PAI-1 in adipocytes and adipose tissue (43). Furthermore, elevated levels of PAI-1 are reported to predict the development of T2D (6, 31). TNF-α was significantly correlated with PAI-1 in the present study (r = 0.31 in the overall population), consistent with TNF-α as a potential inducer of PAI-1 during the onset of T2D. Correlations with PAI-1 were larger in whites than African-Americans (r = 0.33 and 0.16, respectively; P = 0.02 for difference), a finding that may be attributed to ethnic differences in fat distribution and/or in mechanisms contributing to PAI-1 up-regulation. Along these lines, PAI-1 did not explain variance in TNF-α levels beyond those accounted for by waist circumference, HDL, and triglycerides. Using multiple linear regression models, independent associations (adjusted for demographics and BMI or waist circumference) with TNF-α and SI were significant in non-Hispanic whites only, whereas associations between PAI-1 and SI were significant across ethnicity, suggesting ethnicity-specific relationships between obesity, TNF-α, PAI-1, and insulin resistance. Furthermore, these results suggest that PAI-1 may be a more global marker of insulin resistance compared with TNF-α. Future studies designed to address the temporal associations between TNF-α, acute-phase reactants, and insulin resistance during the development of insulin resistance and diabetes would help to elucidate causal mechanisms among these variables.
Dyslipidemia, including increased serum levels of triglycerides and decreased HDL, is a characteristic feature of the metabolic syndrome and T2D. TNF-α was correlated with triglycerides and HDL in the present study (r = 0.35 and −0.30 in the overall population, respectively), consistent with reports that TNF-α can effect lipid metabolism by inhibiting free fatty acid uptake and lipogenesis and/or by enhancing intracellular lipolysis (10). Of interest in the present study was the finding that independent associations of TNF-α and HDL were larger in whites compared with African-Americans or Hispanics (partial r2= 8.0, 1.9, and 3.7%, respectively; P < 0.05 compared with whites), which again may suggest ethnic-specific differences in the relationships between TNF-α and features of the metabolic syndrome, including lipid metabolism.
The current study has several strengths including a large number of subjects with IGT and T2D comprised of three ethnicities and a wide age range, drawn from geographically diverse regions in the United States as well as the direct assessment of insulin resistance and secretion using the FSIGT. Limitations of the study include the fact that visceral fat was not directly measured because this may have provided insight into the observed differences in TNF-α by ethnicity. Another limitation is that the data are cross-sectional, so the temporal relationship between TNF-α and insulin resistance cannot be determined. Finally, the sample sizes in some categories were small (e.g. Tables 2 and 3), limiting the statistical power to detect differences.
In summary, we have shown increased levels of TNF-α in individuals with IGT and T2D in a large, multiethnic population, independent of measures of adiposity, and have demonstrated ethnic differences in TNF-α levels and in the relationships between TNF-α and variables related to the metabolic syndrome. Increased TNF-α levels were predominantly associated with insulin resistance but not with primary defects in β-cell function and were correlated with triglycerides, HDL, and PAI-1. These results are consistent with a mechanistic model in which TNF-α participates in causing insulin resistance and associated abnormalities including dyslipidemia and decreased fibrinolysis.
Acknowledgments
This work was supported by National Heart, Lung, and Blood Institute Grants HL-47887, HL-47889, HL-47890, HL-47892, HL-47902, HL-55208, and R01-HL-58329 and the General Clinic Research Centers Program Grants NCRR GCRC, M01 RR431, and M01 RR01346.
Disclosure Summary: N.C.O., P.W.C., A.J.G.H., S.M.H., L.E.W., and R.P.T. have nothing to declare. A.F. is currently employed by Eli Lilly and Co.
Footnotes
- ADA
- American Diabetes Association
- AIR
- acute insulin response
- BMI
- body mass index
- CRP
- C-reactive protein
- FSIGT
- frequently sampled iv glucose tolerance test
- HDL
- high-density lipoprotein
- IFG
- impaired fasting glucose
- IGT
- impaired glucose tolerance
- IRAS
- Insulin Resistance Atheroclerosis Study
- LDL
- low-density lipoprotein
- NFG
- normal fasting glucose
- NGT
- normal glucose tolerance
- OGTT
- oral glucose tolerance test
- PAI-1
- plasminogen activator inhibitor-1
- SI
- insulin sensitivity index
- T2D
- type 2 diabetes mellitus
- TNFR
- TNF receptor.
References
- 1. Fox CS, Pencina MJ, Meigs JB, Vasan RS, Levitzky YS, D'Agostino RB., Sr 2006. Trends in the incidence of type 2 diabetes mellitus from the 1970s to the 1990s: the Framingham Heart Study. Circulation 113:2914–2918 [DOI] [PubMed] [Google Scholar]
- 2. Shaw JE, Sicree RA, Zimmet PZ. 2010. Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 87:4–14 [DOI] [PubMed] [Google Scholar]
- 3. Schmidt MI, Duncan BB, Sharrett AR, Lindberg G, Savage PJ, Offenbacher S, Azambuja MI, Tracy RP, Heiss G. 1999. Markers of inflammation and prediction of diabetes mellitus in adults (Atherosclerosis Risk in Communities study): a cohort study. Lancet 353:1649–1652 [DOI] [PubMed] [Google Scholar]
- 4. Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. 2001. C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus. JAMA 286:327–334 [DOI] [PubMed] [Google Scholar]
- 5. Vozarova B, Weyer C, Lindsay RS, Pratley RE, Bogardus C, Tataranni PA. 2002. High white blood cell count is associated with a worsening of insulin sensitivity and predicts the development of type 2 diabetes. Diabetes 51:455–461 [DOI] [PubMed] [Google Scholar]
- 6. Festa A, D'Agostino R, Jr, Tracy RP, Haffner SM. 2002. Elevated levels of acute-phase proteins and plasminogen activator inhibitor-1 predict the development of type 2 diabetes: the insulin resistance atherosclerosis study. Diabetes 51:1131–1137 [DOI] [PubMed] [Google Scholar]
- 7. Spranger J, Kroke A, Möhlig M, Hoffmann K, Bergmann MM, Ristow M, Boeing H, Pfeiffer AF. 2003. Inflammatory cytokines and the risk to develop type 2 diabetes: results of the prospective population-based European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Diabetes 52:812–817 [DOI] [PubMed] [Google Scholar]
- 8. Thorand B, Kolb H, Baumert J, Koenig W, Chambless L, Meisinger C, Illig T, Martin S, Herder C. 2005. Elevated levels of interleukin-18 predict the development of type 2 diabetes: results from the MONICA/KORA Augsburg Study, 1984–2002. Diabetes 54:2932–2938 [DOI] [PubMed] [Google Scholar]
- 9. Bradley JR. 2008. TNF-mediated inflammatory disease. J Pathol 214:149–160 [DOI] [PubMed] [Google Scholar]
- 10. Cawthorn WP, Sethi JK. 2008. TNF-α and adipocyte biology. FEBS Lett 582:117–131 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Moller DE. 2000. Potential role of TNF-α in the pathogenesis of insulin resistance and type 2 diabetes. Trends Endocrinol Metab 11:212–217 [DOI] [PubMed] [Google Scholar]
- 12. Hotamisligil GS, Arner P, Caro JF, Atkinson RL, Spiegelman BM. 1995. Increased adipose tissue expression of tumor necrosis factor-α in human obesity and insulin resistance. J Clin Invest 95:2409–2415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Kern PA, Saghizadeh M, Ong JM, Bosch RJ, Deem R, Simsolo RB. 1995. The expression of tumor necrosis factor in human adipose tissue. Regulation by obesity, weight loss, and relationship to lipoprotein lipase. J Clin Invest 95:2111–2119 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Van der Poll T, Romijn JA, Endert E, Borm JJ, Büller HR, Sauerwein HP. 1991. Tumor necrosis factor mimics the metabolic response to acute infection in healthy humans. Am J Physiol 261:E457–E465 [DOI] [PubMed] [Google Scholar]
- 15. Hivert MF, Sullivan LM, Fox CS, Nathan DM, D'Agostino RB, Sr, Wilson PW, Meigs JB. 2008. Associations of adiponectin, resistin, and tumor necrosis factor-α with insulin resistance. J Clin Endocrinol Metab 93:3165–3172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Plomgaard P, Nielsen AR, Fischer CP, Mortensen OH, Broholm C, Penkowa M, Krogh-Madsen R, Erikstrup C, Lindegaard B, Petersen AM, Taudorf S, Pedersen BK. 2007. Associations between insulin resistance and TNF-α in plasma, skeletal muscle and adipose tissue in humans with and without type 2 diabetes. Diabetologia 50:2562–2571 [DOI] [PubMed] [Google Scholar]
- 17. de Rekeneire N, Peila R, Ding J, Colbert LH, Visser M, Shorr RI, Kritchevsky SB, Kuller LH, Strotmeyer ES, Schwartz AV, Vellas B, Harris TB. 2006. Diabetes, hyperglycemia, and inflammation in older individuals: the health, aging and body composition study. Diabetes Care 29:1902–1908 [DOI] [PubMed] [Google Scholar]
- 18. Zinman B, Hanley AJ, Harris SB, Kwan J, Fantus IG. 1999. Circulating tumor necrosis factor-α concentrations in a native Canadian population with high rates of type 2 diabetes mellitus. J Clin Endocrinol Metab 84:272–278 [DOI] [PubMed] [Google Scholar]
- 19. Miyazaki Y, Pipek R, Mandarino LJ, DeFronzo RA. 2003. Tumor necrosis factor α and insulin resistance in obese type 2 diabetic patients. Int J Obes Relat Metab Disord 27:88–94 [DOI] [PubMed] [Google Scholar]
- 20. Müller S, Martin S, Koenig W, Hanifi-Moghaddam P, Rathmann W, Haastert B, Giani G, Illig T, Thorand B, Kolb H. 2002. Impaired glucose tolerance is associated with increased serum concentrations of interleukin 6 and co-regulated acute-phase proteins but not TNF-α or its receptors. Diabetologia 45:805–812 [DOI] [PubMed] [Google Scholar]
- 21. Choi KM, Lee J, Lee KW, Seo JA, Oh JH, Kim SG, Kim NH, Choi DS, Baik SH. 2004. Comparison of serum concentrations of C-reactive protein, TNF-α, and interleukin 6 between elderly Korean women with normal and impaired glucose tolerance. Diabetes Res Clin Pract 64:99–106 [DOI] [PubMed] [Google Scholar]
- 22. Piemonti L, Calori G, Mercalli A, Lattuada G, Monti P, Garancini MP, Costantino F, Ruotolo G, Luzi L, Perseghin G. 2003. Fasting plasma leptin, tumor necrosis factor-α receptor 2, and monocyte chemoattracting protein 1 concentration in a population of glucose-tolerant and glucose-intolerant women: impact on cardiovascular mortality. Diabetes Care 26:2883–2889 [DOI] [PubMed] [Google Scholar]
- 23. Wagenknecht LE, Mayer EJ, Rewers M, Haffner S, Selby J, Borok GM, Henkin L, Howard G, Savage PJ, Saad MF, et al. 1995. The insulin resistance atherosclerosis study (IRAS) objectives, design, and recruitment results. Ann Epidemiol 5:464–472 [DOI] [PubMed] [Google Scholar]
- 24. 1985. Diabetes mellitus. Report of a WHO Study Group. World Health Organ Tech Rep Ser 727:1–113 [PubMed] [Google Scholar]
- 25. 1997. Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 20:1183–1197 [DOI] [PubMed] [Google Scholar]
- 26. Genuth S, Alberti KG, Bennett P, Buse J, Defronzo R, Kahn R, Kitzmiller J, Knowler WC, Lebovitz H, Lernmark A, Nathan D, Palmer J, Rizza R, Saudek C, Shaw J, Steffes M, Stern M, Tuomilehto J, Zimmet P. 2003. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 26:3160–3167 [DOI] [PubMed] [Google Scholar]
- 27. Bergman RN, Finegood DT, Ader M. 1985. Assessment of insulin sensitivity in vivo. Endocr Rev 6:45–86 [DOI] [PubMed] [Google Scholar]
- 28. Welch S, Gebhart SS, Bergman RN, Phillips LS. 1990. Minimal model analysis of intravenous glucose tolerance test-derived insulin sensitivity in diabetic subjects. J Clin Endocrinol Metab 71:1508–1518 [DOI] [PubMed] [Google Scholar]
- 29. Steil GM, Volund A, Kahn SE, Bergman RN. 1993. Reduced sample number for calculation of insulin sensitivity and glucose effectiveness from the minimal model. Suitability for use in population studies. Diabetes 42:250–256 [DOI] [PubMed] [Google Scholar]
- 30. Festa A, D'Agostino R, Jr, Howard G, Mykkänen L, Tracy RP, Haffner SM. 2000. Chronic subclinical inflammation as part of the insulin resistance syndrome: the Insulin Resistance Atherosclerosis Study (IRAS). Circulation 102:42–47 [DOI] [PubMed] [Google Scholar]
- 31. Festa A, Williams K, Tracy RP, Wagenknecht LE, Haffner SM. 2006. Progression of plasminogen activator inhibitor-1 and fibrinogen levels in relation to incident type 2 diabetes. Circulation 113:1753–1759 [DOI] [PubMed] [Google Scholar]
- 32. Kleinbaum D, Kupper L, Muller K. 1988. Applied regression analysis and other multivariable methods. 2nd ed Boston: PWS-KENT Publishing; 92 [Google Scholar]
- 33. Zar J. 2010. Biostatistical analysis. 5th ed Upper Saddle River, NJ: Prentice-Hall; 393–398 [Google Scholar]
- 34. Ho RC, Davy KP, Hickey MS, Melby CL. 2005. Circulating tumor necrosis factor α is higher in non-obese, non-diabetic Mexican Americans compared to non-Hispanic white adults. Cytokine 30:14–21 [DOI] [PubMed] [Google Scholar]
- 35. Hyatt TC, Phadke RP, Hunter GR, Bush NC, Muñoz AJ, Gower BA. 2009. Insulin sensitivity in African-American and white women: association with inflammation. Obesity (Silver Spring) 17:276–282 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Fisher G, Hyatt TC, Hunter GR, Oster RA, Desmond RA, Gower BA. 28 April 2011. Markers of inflammation and fat distribution following weight loss in African-American and white women. Obesity (Silver Spring) 10.1038/oby.2011.85 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Festa A, Hanley AJ, Tracy RP, D'Agostino R, Jr, Haffner SM. 2003. Inflammation in the prediabetic state is related to increased insulin resistance rather than decreased insulin secretion. Circulation 108:1822–1830 [DOI] [PubMed] [Google Scholar]
- 38. Haffner SM, D'Agostino R, Jr, Mykkänen L, Tracy R, Howard B, Rewers M, Selby J, Savage PJ, Saad MF. 1999. Insulin sensitivity in subjects with type 2 diabetes. Relationship to cardiovascular risk factors: the Insulin Resistance Atherosclerosis Study. Diabetes Care 22:562–568 [DOI] [PubMed] [Google Scholar]
- 39. Hotamisligil GS, Shargill NS, Spiegelman BM. 1993. Adipose expression of tumor necrosis factor-α: direct role in obesity-linked insulin resistance. Science 259:87–91 [DOI] [PubMed] [Google Scholar]
- 40. Bertin E, Nguyen P, Guenounou M, Durlach V, Potron G, Leutenegger M. 2000. Plasma levels of tumor necrosis factor-α (TNF-α) are essentially dependent on visceral fat amount in type 2 diabetic patients. Diabetes Metab 26:178–182 [PubMed] [Google Scholar]
- 41. Cartier A, Cote M, Bergeron J, Almeras N, Tremblay A, Lemieux I, Despres JP. 2010. Plasma soluble tumour necrosis factor-α receptor 2 is elevated in obesity: specific contribution of visceral adiposity. Clin Endocrinol (Oxf) 72:349–357 [DOI] [PubMed] [Google Scholar]
- 42. Kissebah AH, Vydelingum N, Murray R, Evans DJ, Hartz AJ, Kalkhoff RK, Adams PW. 1982. Relation of body fat distribution to metabolic complications of obesity. J Clin Endocrinol Metab 54:254–260 [DOI] [PubMed] [Google Scholar]
- 43. Alessi MC, Juhan-Vague I. 2006. PAI-1 and the metabolic syndrome: links, causes, and consequences. Arterioscler Thromb Vasc Biol 26:2200–2207 [DOI] [PubMed] [Google Scholar]