Visceral Adipose Tissue Cutoffs Associated With Metabolic Risk Factors for Coronary Heart Disease in Women

The subjects for this study included peri- and postmenopausal women (at least 45 years of age) who were originally recruited for various weight loss and exercise training studies at the University of Maryland in Baltimore. Women were initially screened via telephone and excluded from participation if they were pregnant, were premenopausal (i.e., regular menstrual cycles), had smoked in the past 5 years, were regular exercisers (>15 min twice a week), had significant weight change (±5%) in the past year, or were taking antidepressants or medications affecting lipid or glucose metabolism. Perimenopausal women were those with irregular menstrual cycles not associated with pathology, such as polycystic ovarian disease, in the 6 months before the study. Postmenopausal women had no menses for at least 1 year before the study and follicle-stimulating hormone values >20 mIU/ml.

Written informed consent was obtained on the first clinic visit from each woman according to the guidelines of the University of Maryland Institutional Review Board for Human Research. Initial screening evaluations, including a medical history, physical examination, 12-lead resting electrocardiogram, fasting blood profile (complete blood count and Chem 18), and graded exercise stress test (22) were performed to exclude women with evidence of ischemia, hypertriglyceridemia (>400 mg/dl), untreated hypertension (blood pressure >160/90 mmHg), cancer, liver, renal, or hematological disease, or other medical disorders. A total of 233 women met these criteria, and their baseline data were used in the cross-sectional secondary analysis for this study.

Measurements of body composition, body fat distribution, fasting and 2-h postprandial (75-g) glucose concentrations (23), and fasting lipoprotein lipid and insulin concentrations were conducted over the course of 2 days. A fasting blood sample was drawn on both days to provide a duplicate measurement of lipoprotein lipids, and the values reported are the average of these 2 days. If the total cholesterol and triglyceride values differed by >5%, or the HDL cholesterol values differed by >10%, a third sample was drawn, and the values reported are the average of these measurements.

Height and weight were measured and BMI calculated as weight in kilograms divided by the square of height in meters. WHR was calculated from the duplicate measurement of the minimal waist circumference to the circumference of the maximal gluteal protuberance. Waist (minimal circumference) and hip (maximal gluteal protuberance) circumferences were measured in duplicate. Percent body fat, fat-free (bone and muscle) mass, and adipose tissue mass were measured using dual-energy X-ray absorptiometry (DPX-L; Lunar Radiation, Madison, WI). A single-slice CT scan taken midway between L4 and L5 was performed using a GE Hi-Light CT scanner to measure abdominal VAT and subcutaneous adipose tissue areas, as previously described (24).

Venous blood samples for the measurement of lipoprotein lipids, glucose, and insulin were collected in chilled tubes containing 1 mg EDTA per milliliter of blood. Plasma was separated by centrifugation at 4°C, and lipoprotein lipids were measured as previously described (25). In our laboratory, the inter- and intra-assay coefficients of variation were 6.2 and 1.5%, respectively, for the measurement of total cholesterol, 7.6 and 2.6% for triglycerides, 9.2 and 2.7% for HDL cholesterol, and 7.6 and 2.8% for LDL cholesterol. Glucose was measured using the glucose oxidase method (Beckman Glucose Analyzer; Beckman, Fullerton, CA), and insulin was measured by radioimmunoassay with an insulin-specific antibody (cross-reactivity with proinsulin <0.2%) (Linco, St. Louis, MO). The lower limit of detection of this assay in our laboratory is 12 pmol/l. Intra- and interassay coefficients of variation of pooled control sera average 5 and 9%, respectively.

Statistical analyses were performed using SPSS version 10.1. Data were first tested for a normal distribution using the Shapiro-Wilk test for normality. Triglyceride and HDL₂ cholesterol concentrations were not normally distributed. Log transformation normalized the distribution of both variables so the logarithm of these data were used for parametric statistical analyses. Partial correlation coefficients adjusted for age and race were computed to identify statistically significant correlations between VAT and disease risk factors. The level of VAT associated with elevated CHD risk factors was determined in two ways. First, women were divided into quintiles based on their level of VAT (n = 46–47 per group) and risk factors were compared between these groups using ANCOVA with calculation of simple contrasts and controlling for age and race. Next, logistic regression analyses, adjusted for age and race, was used to calculate the relative risks for having a low HDL cholesterol (<1.0 mmol/l or 40 mg/dl) (26), hypertriglyceridemia (≥1.7 mmol/l or 150 mg/dl) (26), high LDL/HDL cholesterol ratio (≥3.0), and impaired glucose tolerance (2-h glucose >7.8 mmol/l or 140 mg/dl [23]). The level of significance was set at P < 0.05, and all data are reported as mean ± SD.

The women in this study ranged in age from 45 to 73 years (59 ± 6 years). Of the 233 subjects, 22 were perimenopausal (9%) and 211 postmenopausal (91%) and 50 were African American (21%) and 183 Caucasian (79%). Although no women were on lipid-lowering medications, some did use hormone replacement therapy (HRT), nonsteroidal anti-inflammatory drugs, and antihypertensive medications (Table 1). A total of eight women were diagnosed with diabetes based on a 2-h postprandial glucose >200 mg/dl. As shown in Table 1, although the majority of women were overweight/obese, there was a wide range in body weight, body fatness, abdominal fat distribution, and metabolic risk factors among both African-American and Caucasian women.

VAT was significantly related to total fat mass (r = 0.43, P < 0.001) and age (r = 0.33, P < 0.001). Independent of age, there were no differences in VAT between peri- and postmenopausal women. Likewise, VAT did not differ between women on and off HRT. However, African-American women had less VAT than Caucasian women (P < 0.05) (Table 1) despite having a higher fat mass (P < 0.05) and no differences in age or HRT use.

Partial correlation coefficients were calculated to determine whether there was a linear relationship between metabolic CHD risk factors and VAT. This analysis was adjusted for age and race, but not for total fat mass, since we were interested in assessing the association between risk factors and the absolute amount of VAT. Concentrations of total and LDL cholesterol and systolic and diastolic blood pressure did not correlate with VAT in these women; however, all other risk factors were related to VAT (Table 2).

We also assessed whether there were racial differences in the relationships between metabolic CHD risk factors and VAT or waist by including a race times VAT or race times waist interaction term in the regression model. None of the interactions were significant for any of the dependent variables tested, indicating that in our sample of women, the relationships between these CHD risk factors and VAT or waist are similar in Caucasian and African-American women. Thus, data were combined for further analyses.

Figure 1 shows the average lipoprotein lipid concentrations by quintile of VAT. Women in the lowest VAT quintile (≤105 cm²) had significantly higher concentrations of HDL and HDL₂ cholesterol and lower LDL/HDL cholesterol ratios and triglyceride concentrations than women in the remaining four quintiles (P values <0.05–0.001). Women in the second lowest VAT quintile (106–139 cm²) had higher HDL and HDL₂ cholesterol concentrations and lower LDL/HDL cholesterol ratios than women with a VAT ≥163 cm² (P < 0.05), as well as had lower triglyceride concentrations than women with a VAT ≥193 cm² (P < 0.05).

The glucose and insulin risk factors are shown by VAT quintile in Fig. 2. Women in the lowest VAT quintile (≤105 cm²) had lower fasting glucose concentrations than women in the remaining four quintiles (P values <0.01–0.001) and had lower 2-h glucose concentrations than women in the two highest VAT quintiles (≥163 cm², P < 0.05). Women in the lowest VAT quintile had lower fasting insulin concentrations than women in the remaining quintiles (P values <0.05–0.001), and women in the three lowest VAT quintiles had lower insulin concentrations than women in the two highest VAT quintiles (≥163 cm², P < 0.05).

Table 3 shows the age- and race-adjusted relative risks for having a defined CHD risk factor in women with a VAT greater than the first quintile (i.e., ≥106 cm²). We did not include HDL₂ cholesterol or fasting glucose and insulin in this analysis because there are no established criteria for HDL₂ cholesterol and insulin, and there were only 19 women with impaired fasting glucose (≥110 mg/dl). Women with a VAT level of 106–162 cm² are ∼2.5 times more likely to have a low HDL cholesterol (P < 0.05), while women with a VAT ≥163 cm² are ∼5.5 times more likely to have a low HDL cholesterol (P < 0.01) and ∼4.0 times more likely to have a high LDL/HDL cholesterol ratio (P < 0.05) compared with women with a VAT ≤105 cm². In addition, women with a VAT level >193 cm² are 3.3 times more likely to be hypertriglyceridemic (P < 0.05). Women with a VAT level ≥163 cm² are at a higher risk of having impaired glucose tolerance (P < 0.01). Thus, a VAT ≥106 cm² is associated with a significantly elevated risk and a VAT ≥163 cm² with an even greater risk for these metabolic CHD risk factors.

Waist circumference correlated positively with VAT (r = 0.60, P < 0.001), although, as seen in Fig. 3, there is a great deal of variability in waist circumference for a given VAT level. All risk factors, except total and LDL cholesterol and the ratio of LDL/HDL cholesterol, were related to waist circumference (Table 2).

We were interested in determining whether waist circumference provided an accurate estimate of these high-risk (≥106 cm²) and very high-risk (≥163 cm²) VAT cut points. We used the equation of the regression line of the waist circumference–to–VAT relationship to determine the corresponding high-risk (89 cm) and very high-risk (96 cm) waist circumferences and then tested whether women with waist circumference >89 cm were at a heightened risk of having adverse CHD risk factors (Table 3). This analysis showed that a waist circumference of 89–95 cm was associated with low HDL cholesterol and high LDL/HDL cholesterol ratio (P < 0.05), although the relative risks for these conditions did not increase with increasing waist circumference. In addition, a high waist circumference (i.e., >89 cm) was not associated with a high triglyceride concentration. However, women with a waist circumference ≥96 cm did have a 3.2-fold increased risk for impaired glucose tolerance (P < 0.05).

The results of this study show that in middle-aged and older women, a VAT area ≥106 cm² is associated with elevated risk for having low HDL cholesterol concentrations, hypertriglyceridemia, high LDL/HDL cholesterol ratio, impaired glucose tolerance, and hyperinsulinemia compared with those with a VAT ≤105 cm². A VAT ≥163 cm² is predictive of an even greater risk for these metabolic risk factors for CHD, relative to lower VAT levels. These analyses were adjusted for age and race because of the effects of these factors on the selective accumulation of visceral fat. Although visceral fat mass is, in part, affected by total body fat mass, we chose not to adjust our analyses for total fat mass since we were more interested in determining an absolute level of VAT associated with CHD risk factors. This decision was based on evidence showing that there are large interindividual differences in the amount of visceral fat at any level of body fat content (27,28).

This critical VAT cutoff of 106 cm² is similar to the value of 110 cm², which was shown by Williams et al. (21) to be associated with elevated risk factors for CHD, including high total cholesterol and triglycerides, low HDL cholesterol, and elevated blood pressure in women. In addition, this value matches that of the critical value of 100 cm² shown by Despres and Lamarche (19) to be associated with metabolic risk factors for type 2 diabetes and CHD in younger women. These results are comparable despite the fact that we studied an overall older and more obese sample of women. Moreover, it is important to point out that in the few studies to date that have attempted to identify cut points for VAT, some have used receiver-operating characteristic curve analyses, while others used a similar approach (logistical regression analyses) to ours. We chose the latter approach because, from a clinical standpoint, the degree of relative risk may be more meaningful than a likelihood ratio of sensitivity or specificity. The fact that the 100- to 110-cm² value is consistent across studies suggests that the different analyses produce similar results. Thus, our data confirm the finding that a VAT of ∼100–110 cm² is predictive of adverse lipid (HDL cholesterol and triglyceride) and glucose risk factors in women. However, unlike these other studies (19,21), total and LDL cholesterol were not related to VAT in the sample of women we studied. Instead, age was the only independent predictor of total and LDL cholesterol in our women (data not shown). This finding may be due to the older age or much higher degree of total obesity in our sample, as evidenced by their average age (59 years) and percentage body fat (46%).

The association between an abdominal fat distribution and a greater incidence of metabolic health problems was reported as early as the 1950s (29). Subsequently, it was discovered that the associations observed between anthropometric measures of abdominal fat and higher risk for CHD likely resulted from the excess accumulation of VAT (8–11,12). Furthermore, while the amount of visceral fat correlates with the amount of total body fat, visceral obesity and its associated metabolic complications are not only evident in obese subjects. Studies show that visceral fat accumulation also contributes to metabolic risk factors in healthy, nonobese individuals (10,13–15). In addition, obese individuals who are “metabolically healthy” have very little visceral fat accumulation (30). Thus, the majority of evidence to date suggests that assessment of visceral fat may be more clinically relevant than assessment of total body fat for the prediction of subsequent CHD onset.

Despite the clinically significant information gained from assessment of VAT, its measurement is costly, it involves radiation exposure, and it is not a practical screening tool for the general population. Therefore, anthropometric or other surrogate measures of VAT need to be defined. For example, waist circumference, which has been shown to be the best predictor of VAT in several studies (31–34), has been proposed as a screening tool for assessment of disease risk (3). In the present study, there was only a modest correlation between waist circumference and VAT, as well as a large degree of variation in waist circumferences for a given VAT level. In addition, the waist circumference cut points that corresponded to the high-risk and very high-risk VAT cut points in these women were predictive of an increased risk for low HDL cholesterol and glucose intolerance, although the relative risks were not as high as those for the VAT cut points.

Recently, Despres and colleagues (4,35) have proposed a novel screening approach for identifying viscerally obese men at risk for CHD that utilizes waist circumference and fasting triglyceride concentrations. As noted by this group, the cut points they propose as useful in young and middle-aged men (waist circumference >90 cm and triglyceride concentration >2 mmol/l) may not be useful in individuals of a different sex, age, or race. For example, this group reports a much higher correlation between waist circumference and VAT (up to 0.87) in their study populations of young Caucasian men and women (32). In addition, they have shown that this relationship is age specific in that values of waist circumference that corresponded to “critical” levels of VAT were lower in subjects >40 than in those <40 years of age (33). Because of the preferential accumulation of VAT with age (16,36), a specific waist circumference may not correspond to a similar VAT level in younger and older individuals. This relationship may be even more complex in postmenopausal women, who are also predisposed to selective fat storage in the visceral region (37). Thus, further research is needed to define visceral obesity and identify simple markers of VAT that accurately predict risk factors and incidence of CHD in older women.

Some discussion is necessary regarding the limitations of this study. First, the women in this study were not randomly selected but were recruited for ongoing weight loss and exercise studies at the University of Maryland. In addition, we did not include women with frank CHD and/or diabetes and our sample was relatively healthy with respect to the risk factors studied. This could potentially limit the generalizability of our findings and/or may change the VAT cut points found by our analyses. Moreover, our sample included Caucasian and a small number of African-American women. Although there were no significant interactions between race and VAT for the risk factors tested, larger sample sizes are needed to definitely test whether there are ethnic differences in the degree of metabolic abnormalities for a given VAT level. In addition, adipose tissue or body weight cut points are a guide that can be used for defining associations with risk or incidence of disease. Like the BMI cut points, a critical level of VAT does not establish causality for an elevated disease risk. However, these values can be used as a guide for defining “visceral obesity” and may prove useful for identifying women most likely to benefit from preventative interventions. Further research is needed to confirm and/or refine these cut points for assessment of incidence of CHD (and to determine whether they differ by sex, ethnicity, or age-group), as well as to determine the percent of individuals in the population with this amount of VAT, i.e., prevalence of visceral obesity.

This study was supported by National Institutes of Health Grants RO1 NR03514, R29 AG14066–02, K01 AG00747, K01 AG00685, and T32 AG-00219 and by the Department of Veterans Affairs Baltimore GRECC.

We are grateful to the nursing and laboratory staff of the University of Maryland’s Division of Gerontology and GRECC for their assistance in the conduct of this research. We especially thank all of the women who volunteered to participate.