INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

CURRENT STATISTICS HAVE REVEALED that more than one out of every three adult women is overweight, with the largest percent, nearly 49%, being African-American (AA) women (27). These women also suffer more serious medical consequences, particularly with regard to insulin (Ins) resistance, hyperinsulinemia, diabetes, and coronary artery disease (CAD) mortality (3, 6, 22). However, not all overweight AA or Caucasian (CC) women fall victim to the metabolic perturbations associated with their excess weight, which may be more closely related to the central distribution of their weight. Excessive intra-abdominal fat or visceral adipose tissue (VAT) is significantly associated with glucose (Glu) intolerance, Ins resistance, and dyslipidemia (10), including elevated serum triacylglycerol (TG), and with reduced cardioprotective high-density lipoprotein (HDL) cholesterol. It is also associated with elevated systemic blood pressure (24).

With the advent of the more sophisticated computer technology, magnetic resonance imaging (MRI) techniques have been used to scan the intra-abdominal region and quantify the amount of the more hazardous VAT. Research supports measuring VAT area at one location to indicate the total amount of VAT; however, there is controversy regarding the optimal landmark to use when VAT area is measured in different racial populations (8, 36, 40). Although VAT volume and area at L₄-L₅ are similarly correlated with health-related variables (HRV) in obese men (37), we know of no studies examining this relationship in a mixed racial population of overweight women. Because there is no time penalty for acquiring multiple images and no X-ray radiation, it is just as easy to acquire several abdominal images as it is to acquire a single image. Thus it seems reasonable to examine whether VAT area at L₄-L₅ is an adequate surrogate measure for VAT volume in relation to HRV in overweight CC and AA women.

Conversely, anthropometric surrogates, such as waist circumference (WC), sagittal diameter (SD), and or waist-to-hip ratio (WHR), have provided a less expensive and more expedient alternative to the measurement of VAT. Again, there is controversy regarding the optimal anthropometric surrogates for VAT (7, 12, 20, 25), although WC and possibly SD have emerged as somewhat better surrogate measures (2, 20, 36). A more critical issue, however, is which anthropometric surrogate is the best predictor of HRV, which has been reported to be race dependent, differing between CC and AA women (7, 12, 23). Furthermore, research examining this question should also measure physical activity, stress, and nutrient intake in the same study to demonstrate that differences between groups are not related to these potential confounders.

We know of no studies evaluating the volume of VAT and its anthropometric surrogates to explain HRV among women and minorities while at the same time examining multiple confounders. Therefore, the purpose of this study was to determine whether the volume of VAT or its anthropometric surrogates could predict serum lipoproteins, blood pressure, and Ins resistance in a sample of overweight CC and AA women.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects. To participate in the study, subjects were required to be apparently healthy and premenopausal, displaying a body mass index (BMI) above 27.3 kg/m² (50). Subjects who were pregnant or lactating; amenorrheic; or taking medications that would affect blood pressure, carbohydrate, or lipid metabolism were excluded from the study. Both parents and both grandparents had to be of the same racial descent for the subject to be considered CC or AA. A total of 66 subjects (36 CC; 30 AA) met all eligibility requirements and fulfilled the criteria necessary for completion of the study. Subjects completed all testing procedures in accordance with the guidelines set forth by the Human Subjects Institutional Review Board at the University of Miami.

Anthropometric measurements. Body weight was measured to the nearest 0.1 kg. Body height was measured to the nearest 0.5 cm, and BMI was calculated as weight (kg) divided by height (m²). Circumferences were measured with a Gulick II plastic spring-tension measuring tape (Gays Mill, WI). WC was measured midway between the lower rib margin and iliac crest, whereas hip circumference was measured at the outermost points of the greater trochanters (50). The WHR was the ratio calculated between these two circumferences. The SD of subjects was measured in a reclining position with an anthropometer placed in the same position as that used to determine WC. This procedure is reported to maintain the depth of the abdomen, thereby providing a more accurate assessment of VAT (44). All anthropometric measurements were performed by the same investigator, who recorded the mean of the two measurements to the nearest 1.0 mm.

Blood pressure. After an overnight fast, systolic blood pressure (SBP) and diastolic blood pressure (DBP) were taken after subjects were in a seated position for a 5-min interval. Duplicate blood pressure measurements were taken from the left upper arm, averaged, and recorded to the nearest 2.0 mmHg with a 5-min interval separating measurements.

Serum measurements. Blood withdrawal from the antecubital vein occurred after a 12-h fast and was analyzed within a 1-wk period. Levels of total cholesterol (32) and HDL (48), HDL subfractions (16), and TG (29) were measured by a lipid laboratory within the Diabetes Research Institute, University of Miami. The very-low-density-lipoprotein (VLDL) cholesterol was estimated by TG divided by five, and low-density-lipoprotein (LDL) cholesterol was indirectly calculated by subtracting HDL and VLDL from total cholesterol (15). Serum standards used for calibration were developed by the Diabetes Research Institute and calibrated against serum samples from the Centers for Disease Control and Prevention Laboratory.

MRI. The abdominal region was examined by using MRI with a 1.5-T instrument (Siemens Medical Systems, Iselin, NJ). Spin-echo imaging was performed by using a T1 weighted sequence with a 147-ms repetition time and 4.8 ms echo time. Slice thickness was 10 mm with a 2.5-mm interslice gap. A total of seven slices was obtained in each subject with the central slice of the acquisition centered at the L₄-L₅ disc level. Acquisition time was 26 s by using this imaging sequence, enabling imaging to be performed during a single breath hold.

Diet record. Subjects were asked to keep a 3-day food log typical of their eating habits for 1 weekend day and 2 weekdays during the week preceding testing. Nutrient and alcohol intakes were analyzed by using a computerized analysis system (Dine Health Systems 1994). According to Basiotis et al. (4), a 3-day food log is sufficient to accurately assess nutrient intake. Subjects were also required to record the average number of cigarettes smoked daily.

Physical activity record. Subjects were administered the College Alumnus Questionnaire developed by Paffenbarger et al. (33) for assessing their participation in sports, recreational activities, and everyday physical activities. This questionnaire has been previously reported to be a reliable tool for the estimation of physical activity level in a similar group of subjects (1).

Education. This was measured by using a categorical scale that was modified from Rosmond et al. (39) On a 1-5 scale, all subjects had their education status evaluated accordingly: elementary school education received 1, vocational school education 2, senior high school education 3, undergraduate college education 4, and graduate school education 5.

Stress. All subjects completed the Daily Stress Inventory developed by Brantley et al. (5). This survey has been designed to evaluate minor psychological stress, and the survey has been shown to be a significant correlate of endocrine measures of stress.

Statistical methods. Means ± SD of all variables were calculated for CC and AA women. Differences between racial groups were assessed by using a Student's t-test for unpaired samples. Tests for normality were done on all physiological variables by using the Shapiro-Wilk W test. Because TG was not normally distributed, a log₁₀ transformation was performed, which adequately normalized the distribution of this variable. To obtain a normalized distribution for serum Glu levels, three outlying values (1 CC and 2 AA) were removed. Students' t-tests were conducted with and without the three outliers. Multiple-regression models were conducted to determine whether VAT or its anthropometric surrogates and race and/or a VAT × race interaction could predict HRV. The same regression model was repeated by substituting either percent body fat or SAT for VAT volume. Additionally, several potential confounders (e.g., alcohol, stress) were added to the VAT regression model to determine their contribution to the variance in HRV. In all regression analyses, age was treated as a nuisance variable and was regressed out of each HRV with the residuals serving as the new criterion variables in the subsequent multiple-regression analyses. To better understand significant interactions, Pearson's product-moment correlation matrix was performed by race after controlling for the effects of age. Partial correlations between HRV and VAT volume and percent body fat were performed after controlling for percent body fat and VAT volume, respectively, in CC and AA women. A significance level of P  0.05 was selected.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Results of the Student's t-test on physical characteristics (Table 1) indicated that there were no significant differences between CC and AA women for any of these measures.

Dietary information is presented in Table 2. It can be seen that there were no significant differences between CC and AA women for any reported nutrients. Substitution of absolute values in grams for the percentage of total fat, saturated fat, and carbohydrates also resulted in no significant differences between CC and AA women.

Table 3 presents the mean body composition and fat distribution values for CC and AA women. As can be seen, CC women had significantly higher VAT volume and area (P < 0.0001 for both). There was also a trend toward a higher age (P = 0.0631) and WC (P = 0.0584) in CC women.

Table 4 shows the similarity between CC and AA women for HRV. The only significant difference between groups was the higher TG level observed in CC women (P = 0.009). Serum Glu levels are presented after removal of three outliers; however, there were no significant differences in serum Glu levels either before or after their removal.

The multiple-regression analysis presented in Table 5 shows those HRV that could be significantly predicted by VAT volume, race, and a VAT volume × race interaction. HRV that could not be significantly predicted are not presented in this table. VAT volume was a significant and independent contributor explaining the variance in TG, HDL cholesterol, Ins, and Ins/Glu. It can be seen that VAT volume but not race contributed significantly to the variance in log TGs, and because VAT volume was greater in CC women, race differences in log TGs can be attributed specifically to differences in VAT volume. To support this finding, we reran a Student's t-test on log TGs in 13 CC and 13 AA women matched for VAT area and found no significant differences in their log TGs (4.90 ± 0.48 and 4.75 ± 0.54 for CC and AA women, respectively; P = 0.44).

Of particular interest in Table 5 was the fact that the VAT volume × race interaction contributed significantly to the prediction of Ins and Ins/Glu. Further evaluation showed that, after controlling for age, CC women showed a strong and positive relationship between VAT volume and Ins (r = 0.784; P < 0.0001) and Ins/Glu (r = 0.793; P < 0.0001). Conversely, AA women showed an essentially absent relationship [r = 0.061; P = not significant (NS) and r = 0.13; P = NS for Ins and Ins/Glu, respectively]. Addition of potential confounders to the regression model showed that alcohol had a significant impact on HDL (r² = 0.090; P = 0.016) and that self-reported stress significantly impacted Ins and Ins/Glu (r² = 0.068; P = 0.040 and r² = 0.129; P = 0.005 for Ins and Ins/Glu, respectively).

To determine the contribution of VAT volume apart from general adiposity, we reran the multiple regression in Table 5, substituting percent body fat for VAT volume. We found that percent body fat failed to result in a model that could significantly predict either log TG or HDL. In contrast, total adiposity contributed significantly and independently to the variance in Ins (r² = 0.1492; P = 0.0016) in a model that was significant (F = 4.2413; P = 0.007) and contributed significantly and independently to the variance in Ins/Glu (r² = 0.1649; P = 0.0013) in a model that was also significant (F = 4.744; P = 0.0050). In contrast to VAT volume, there was no race effect for percent body fat in predicting either Ins (r² = 0.0016; P = NS) or Ins-to-Glu ratio (r² = 0.015; P = NS). Substitution of SAT for VAT volume yielded results similar to those for percent body fat. SAT volume significantly and independently contributed to the variance in Ins (r² = 0.1264; P = 0.003) and Ins/Glu (r² = 0.1417; P = 0.003), with both models being significant (F = 3.653; P = 0.017 and F = 4.120; P = 0.012 for Ins and Ins/Glu, respectively). Similar to percent body fat, there was neither a race effect nor race × SAT volume interaction for Ins or Ins/Glu.

Shown in Table 6 are the correlations of HRV with VAT volume and percent body fat after controlling for percent body fat and VAT volume, respectively. In CC women, correlations between VAT volume and HRV were significant after controlling for percent body fat, whereas partial correlations between percent body fat and HRV were nonsignificant in these women. In contrast, partial correlations between VAT volume and HRV were minimal and nonsignificant in AA women; however, there was a trend toward a significant correlation and a stronger relationship between Ins/Glu and percent body fat in these women (r = 0.3381; P = 0.091).

To determine whether the relationship between VAT and Ins or Ins/Glu were due to differences in quantity of VAT, we performed the same multiple-regression analysis on a subset of 13 CC and 13 AA women matched for VAT area (within 2.0 cm²). When subjects were paired, the mean VAT area was 101.20 and 101.40 cm² for CC and AA women, respectively (P = 0.99). As shown in Table 7, Ins and Ins/Glu could be significantly predicted by VAT area, race, and VAT area × race interaction, with each predictor contributing significantly to the variance in the model. On further evaluation, the correlation between VAT area and Ins (r = 0.683) and Ins/Glu (r = 0.682) was positive and significant for CC women (P = 0.01 for both). Conversely, in AA women, the correlations between VAT area and Ins (r = 0.24) as well as Ins/Glu (r = 0.190) were negative and nonsignificant, thereby supporting the divergent relationship in CC and AA women observed for the entire sample. Furthermore, we found that 6 of the 13 matched CC and AA women had VAT areas below 110 cm, which is associated with minimal metabolic risk (49). The mean VAT area was 67.86 cm² for the six CC and AA women, respectively, with no significant differences observed between the two races in either age or VAT area. The correlation between VAT area and Ins and Ins/Glu was r² = 0.856 (P = 0.029) and r² = 0.824 (P = 0.043), respectively, for CC women, whereas in AA women it was r² = 0.362 (P = 0.479) and r² = 0.414 (P = 0.487), respectively.

Results obtained by using anthropometric surrogates for VAT volume and area are shown in Table 8. Because there were no significant interactions involving race and anthropometric surrogates, the interaction was excluded from the regression analysis. Results indicated the same HRV predicted by VAT and race could also be predicted by WC and race. Additionally, WC was a significant and independent predictor of the variance in HRV. The use of SD afforded the additional prediction of SBP but failed to predict the cardioprotective HDL. Interestingly, race was a significant and independent contributor to the variance in log TG by using either WC or SD in the model. Conversely, WHR did not result in the significant prediction of any HRV with the exception of log TG (F = 3.295; P = 0.043).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the present study, CC women showed significantly higher VAT volume and area than AA women having similar BMI values (P < 0.0001 for both). Given the lack of significant differences between CC and AA groups in potential confounders, higher VAT levels in CC women take on added clinical significance. The mean VAT area of 128.20 cm² found in CC women exceeded the 110-cm² critical threshold for increased CAD risk (49) and was more than 1.5 times higher than that observed for AA women. Conversely, significantly lower VAT levels in AA women may, in part, explain the blunted rise in CAD observed with increasing BMI levels in AA women (45).

Recently, there has been some speculation regarding the direct involvement of VAT in assessing one's CAD risk (14, 42). Although VAT drains into the portal vein, exposing the liver to increased concentrations of free fatty acids, investigators have questioned whether the amount of VAT is sufficient to impair the metabolic clearance of Ins leading to a state of hyperinsulinemia and perturbations in lipid metabolism (14). Our results indicated that TG, HDL cholesterol, Ins, and Ins/Glu could all be significantly predicted by VAT and race, with VAT contributing significantly to the variance in each model (Table 5). Therefore, our data corroborate the finding that VAT volume is predictive of HRV and that VAT area at L₄-L₅ is a viable alternative to VAT volume in both racial populations.

Total adiposity could also predict Ins and Ins/Glu similarly in CC and AA women; however, it was not as strong a predictor as VAT volume, which accounted for an extra 11.67% of the variance in Ins and an extra 15.28% of the variance in Ins/Glu. Because total adiposity could not predict log TG or HDL, VAT volume clearly represents the better predictor of HRV in a group of overweight women. Examination of our correlations, however, indicated that the ability of VAT to predict HRV was primarily mediated by the strength of that relationship observed in CC women.

One could speculate that the lesser relationship observed between VAT and Ins or Ins/Glu in AA women was due to the fact that they had lower VAT levels. However, in 6 of the 13 matched CC and AA women displaying low VAT areas, we found a significant relationship with Ins and Ins/Glu in CC women that was not observed in AA women. In fact, using the entire sample, we found 13 CC and 21 AA women displaying low VAT areas (<110 cm²) and showed the same trend indicated above. Thus it appears that a critical amount of VAT is not necessary to have a significant relationship with Ins or Ins/Glu.

It could be argued that, because Ins and Ins/Glu are strongly correlated (r = 0.975), the results of both variables are redundant. Ins, however, has been significantly and independently related to CAD and hypertension (9, 11, 38), and AA women show greater mortality from CAD (3, 6) and greater rates of hypertension than CC women (13). On the other hand, Ins/Glu, using the HOMA index, indicates risk for diabetes, and AA women have 100% greater risk for diabetes than CC men, whereas CC women have a 30% greater risk for diabetes than CC men (21). Thus the critical issue is that VAT volume was not able to predict either Ins or Ins/Glu (both of which represent different and relevant HRV) in AA women.

It should be pointed out that Albu and colleagues (2) did not find racial differences in the relationship between VAT area and Ins sensitivity. However, CC and AA subjects in that study had higher BMI and VAT levels than did subjects in the present study, with no racial differences observed for VAT area. They also used a dynamic model for examining Ins-Glu kinetics, which indicated peripheral Ins sensitivity. On the other hand, Dowling and Pi-Sunyer (12), using the identical measures of Ins sensitivity alluded to above, obtained results similar to ours. In their study, they used WHR cut points to indicate central and peripheral obesity and found significant differences between upper and lower body obesity and Ins sensitivity in CC women, whereas they found no differences between upper and lower body obesity and Ins sensitivity in AA women.

On the basis of the findings of our study, we can only speculate as to why we found no relationship between VAT and Ins and Ins/Glu in AA women. Research has shown a pituitary hyperresponsiveness occurring alone or in response to mental challenge that may play a role in the subsequent development of VAT and the related metabolic disorders of hyperinsulinemia and Ins resistance (31, 34, 35). Unfortunately, this has not been examined in a mixed racial population and should be examined further in both overweight CC and AA women. Interestingly, self-reported stress was the only confounder that significantly impacted both Ins (P = 0.040) and Ins/Glu (P = 0.005) in the present study. Thus we feel a more in-depth analysis of the hypothalamic-pituitary-adrenal axis function in response to stress may be relevant. There may also be a role for sex steroids in the development of VAT, hyperinsulinemia, and Ins resistance (18, 46). If there are significant differences between CC and AA women in sex steroids or in the impact of sex steroids on VAT or basal insulin levels, this, too, may account for differences in the relationship between VAT and Ins or Ins/Glu in CC and AA women. Certainly more research is needed to clarify this issue.

Finally, WC and SD were suitable anthropometric surrogates for predicting HRV. In fact, WC and race could predict the identical HRV as VAT and race (i.e., log TG, HDL, Ins, and Ins/Glu) and should be considered the best surrogate measure. Sjöström et al. (43) hypothesized that WC may be superior to VAT or other anthropometric surrogates because WC encompasses both SAT and VAT obtained in a single abdominal measurement and is closely associated with BMI. Our study showed that SAT area was >3.5 times the amount of VAT volume or area in both CC and AA women, and thus any WC measurement would reflect the dominance of SAT over VAT. Because the ability of SAT volume to significantly impact Ins and Ins/Glu occurs similarly for CC and AA women, WC is expected to be a better predictor of these variables in research using both CC and AA women. It should be recognized, however, that, in our study, VAT volume was clearly the stronger predictor of HRV, accounting for a larger percent of the variance (R²) in each model presented with less degrees of freedom (Table 5).

In conclusion, our study supports the use of VAT volume or area for predicting several HRV in CC women; however, its use in predicting either Ins or Ins/Glu appears significantly limited in AA women. Given their higher rates of diabetes and CAD mortality (3, 6, 22) as well as the fact that VAT levels were significantly lower in AA than CC women, the search for better predictors of HRV in AA women seems warranted. Anthropometric surrogates, specifically WC, can provide a suitable alternative to VAT for predicting several HRV in both CC and AA overweight women; however, it should be recognized that VAT was clearly the more powerful predictor of HRV in our study.