Black South African women are more insulin resistant than BMI-matched white women. The objective of the study was to characterize the determinants of insulin sensitivity in black and white South African women matched for BMI. A total of 57 normal-weight (BMI 18–25 kg/m2) and obese (BMI > 30 kg/m2) black and white premenopausal South African women underwent the following measurements: body composition (dual-energy X-ray absorptiometry), body fat distribution (computerized tomography (CT)), insulin sensitivity (SI, frequently sampled intravenous glucose tolerance test), dietary intake (food frequency questionnaire), physical activity (Global Physical Activity Questionnaire), and socioeconomic status (SES, demographic questionnaire). Black women were less insulin sensitive (4.4 ± 0.8 vs. 9.5 ± 0.8 and 3.0 ± 0.8 vs. 6.0 ± 0.8 × 10−5/min/(pmol/l), for normal-weight and obese women, respectively, P < 0.001), but had less visceral adipose tissue (VAT) (P = 0.051), more abdominal superficial subcutaneous adipose tissue (SAT) (P = 0.003), lower SES (P < 0.001), and higher dietary fat intake (P = 0.001) than white women matched for BMI. SI correlated with deep and superficial SAT in both black (R = −0.594, P = 0.002 and R = 0.495, P = 0.012) and white women (R = −0.554, P = 0.005 and R = −0.546, P = 0.004), but with VAT in white women only (R = −0.534, P = 0.005). In conclusion, body fat distribution is differentially associated with insulin sensitivity in black and white women. Therefore, the different abdominal fat depots may have varying metabolic consequences in women of different ethnic origins.
Obesity, particularly abdominal obesity, is associated with insulin resistance (1) and an increased risk of a number of conditions including type 2 diabetes (2). The prevalence of obesity in South African women is high, being 30.0% compared to 7.5% in South African men (3). Further, black South African women are more affected (31.8%) than white women (22.7%) (3). In 2000, 94% of type 2 diabetes in South African women could be attributed to excess body weight, surpassing global estimates (4).
Intriguingly, for the same level of adiposity black South African women, like African-American women, are more insulin resistant than their white counterparts (5,6,7); however, studies in South Africa and the United States have demonstrated that black women have significantly less visceral adipose tissue (VAT) than white women matched for BMI (5,8,9). On the other hand, black women have more abdominal subcutaneous adipose tissue (SAT), which has been shown to be a significant determinant of insulin sensitivity in some studies (9,10,11,12). SAT can be divided into two functionally distinct compartments, the deep and superficial depots. Emerging evidence indicates that deep SAT is more closely associated with insulin sensitivity than superficial SAT (10,11). An understanding of the relative contribution of abdominal adipose tissue depots to ethnic-specific differences in insulin sensitivity may aid in our understanding of the African paradox, namely insulin resistance with low VAT.
Factors other than abdominal fat distribution, such as dietary fat intake and physical activity also contribute to the variation in insulin sensitivity (7,13). These are often governed by the socioeconomic status (SES) of the individual, which has been shown to impact adipose tissue distribution (14) and have a marked influence on health status (15). Indeed, we recently showed that insulin sensitive black South African women had a higher SES and did more leisure and vigorous physical activity than their insulin-resistant counterparts (16).
We hypothesized that insulin resistance in black South African women was associated with increased deep SAT, as opposed to VAT, as well as other modifiable lifestyle factors. Therefore, the aim of the present study was to characterize the determinants of insulin sensitivity in black and white premenopausal South African women matched for BMI, focusing specifically on the relative roles of the adipose tissue depots, dietary intake, physical activity, and SES. Comparisons among ethnic groups enabled us to examine factors that are shared, as well as explore the unique features of an ethnic group that contribute to the variability in insulin sensitivity.
Methods and Procedures
The study population consisted of 13 normal-weight (BMI 18–25 kg/m2) and 16 obese (BMI >30 kg/m2) black women, and 14 normal-weight and 14 obese white South African women, who were recruited by advertisement in local newspapers and from local church groups, community centers and universities. Inclusion criteria were: (i) age 18–45 years, (ii) no known diseases or taking medication for any metabolic disorders, and (iii) not currently pregnant, lactating or postmenopausal. The study was approved by the Research Ethics Committee of the Faculty of Health Sciences of the University of Cape Town. Prior to participating in the study, procedures and risks were explained to the subjects, all of whom gave written informed consent.
Assessment of demographics, dietary intake, and physical activity. A demographic questionnaire was administered that included measures of SES, family history of type 2 diabetes (first degree relatives only), reproductive history, and personal health. SES was assessed on the basis of asset index, housing density, education, and employment (3). Asset index was based on 14 items reflecting individual and household wealth. Education was categorized by grades passed. Housing density was defined as the number of persons per room living in the household. Subjects were categorized as unemployed, students, informal employment, or employed. Based on a rank sum of these scores, a SES score was devised, which showed good internal reliability (standardized Cronbach's α = 0.829).
Dietary intake was estimated using a food frequency questionnaire previously validated in South African women (17). The questionnaire comprises 100 food items with food photographs to determine portion size. The questionnaire was administered by a registered dietician. Nutrient intake was calculated by means of the software program Food-Finder III, supplied by the Nutrition Intervention Research Unit (South African Medical Research Council, Parow, South Africa).
Physical activity energy expenditure was characterized using the Global Physical Activity Questionnaire. Total physical activity energy expenditure, in metabolic equivalents, as well as the domain and intensity of activity were calculated (18).
Body composition. Weight, height, waist (level of umbilicus), and hip (largest gluteal area) circumferences were taken. Fat and fat-free soft tissue mass were measured using dual-energy X-ray absorptiometry (Discovery-W, Software version 4.40; Hologic, Bedford, MA). Body composition of subjects that exceeded the scanning region was calculated using the arm-replaced method (19). Visceral, deep, and superficial SAT areas were measured using computerized tomography (CT, Toshiba X-press Helical Scanner; Toshiba, Tokyo, Japan) at the level of L4–L5 lumbar vertebrae, as described previously (10,11).
Insulin sensitivity. After an overnight fast, subjects underwent an insulin-modified frequently sampled intravenous glucose tolerance test to quantify insulin sensitivity. Baseline samples were drawn at −15, −5, and −1 min prior to the infusion of glucose (50% dextrose; 11.4 g/m2 body surface area) over 60 s commencing at time 0. At 20 min, human insulin (0.02 U/kg, Actrapid; Novo Nordisk, Bagsvaerd, Denmark) was infused over 5 min at a constant rate. Plasma glucose and serum insulin concentrations were measured in the three baseline samples and the 32 samples, drawn over 240 min following commencement of the glucose infusion. Glucose and insulin data from the frequently sampled intravenous glucose tolerance test were used to calculate the insulin sensitivity index (SI) using Bergman's minimal model of glucose kinetics (20).
Plasma glucose concentrations were determined using the glucose oxidase method (YSI 2300 STAT PLUS; YSI Life Sciences, Yellow Springs, OH). Serum insulin concentrations were determined by immunochemiluminometric assays using the ADVIA Centaur (Bayer Diagnostics, Tarrytown, NJ). The intra- and interassay coefficients of variation for plasma glucose and serum insulin concentrations were 0.6 and 2.5%, and 4.5 and 12.2%, respectively.
We undertook a power calculation based on data from Lovejoy et al. (9), which showed that we had 80% power to detect a difference of 2 × 10−5/min/pmol/l in SI at P < 0.05 with 25 subjects per ethnic group. Results are presented as means ± s.e., apart from physical activity data that are presented as medians and interquartile ranges. Two-way analysis of covariance adjusting for age, was used to compare anthropometric, dietary, and metabolic measures between normal-weight and obese, black and white women, with Bonferroni post hoc analyses. Ethnic differences in abdominal adipose tissue depots were also adjusted for total fat mass. Data were normalized by log transformation, if required. Item analysis was used to construct a SES score from the ranked sum scores of four measures of SES (housing density, asset index, education, and employment). Mann–Whitney U nonparametric statistics were used to compare physical activity and its components between ethnic and BMI groups. Data were analyzed using STATISTICA version 7 (Statsoft, Tulsa, OK).
The obese women were significantly older than the normal-weight women (Table 1). Consequently, all subsequent analyses were adjusted for age. Independent of BMI status, black women were of a significantly lower SES, had less formal education, more were unemployed, and a greater proportion had at least one pregnancy compared to their white counterparts (Table 1). There were no significant group differences in the proportion of subjects with a first degree relative with type 2 diabetes mellitus, nor in the proportion of subjects who currently smoked.
Table 1. Demographic profile and body composition of the participants according to ethnicity and BMI
By design, all measures of body composition and regional fat deposition were significantly greater in the obese than the normal-weight women (Table 1). The black women were shorter than the white women and weighed less. Fat-free soft tissue mass was significantly lower in black than white women, but not after adjusting for height (P = 0.17). Despite differences in stature, black and white groups were well-matched for BMI and total adiposity (dual-energy X-ray absorptiometry-derived fat mass and % body fat).
In normal-weight subjects, there were no differences by ethnicity in any measure of body size or the mass of any fat depot. However, there were significant interactions between ethnicity and BMI status for measures of centralization of body fat. Obesity was associated with a greater waist circumference, waist-to-hip ratio, and abdominal adipose tissue area in both ethnic groups. However, centralization of fat in obese black women was attributable to a greater amount of SAT and not VAT. In fact, obese black women had significantly less VAT, but more superficial SAT than obese white women. No ethnic differences in deep SAT were observed. Hence, VAT/SAT ratio of the obese black women was significantly lower than that in obese white women.
Dietary intake and physical activity energy expenditure
We next addressed the possible mechanisms of ethnic differences in fat depot expansion with obesity. Energy intake was greater in obese black compared to obese white women (Table 2), largely due to higher dietary fat intake. As a percentage of total energy, there were no ethnic differences in carbohydrate intake, but protein intake was lower and fat intake was higher in obese black than obese white women. There were no ethnic differences in the relative intakes of saturated or monounsaturated fat, but polyunsaturated fat intake (PUFA), and consequently, ω-6 fatty acid (n-6 FA) intake, were higher in black compared to white women. Alcohol intake was lower in black than white women. No ethnic or BMI differences were noted in relation to intakes of n-3 FA or added sugar.
Table 2. Dietary intake and physical activity of the participants according to ethnicity and BMI
Physical activity domain and intensity, but not total physical activity energy expenditure, were significantly different between ethnic groups (Table 2). White women engaged in vigorous intensity leisure activity, whereas black women did not engage in leisure physical activity, but expended more energy in travel, undertaken at a moderate intensity. Consequently, minutes of total and moderate vigorous activity per day did not differ by ethnicity. Black women slept more than white women. Physical activity energy expenditure was not influenced by BMI category.
Insulin sensitivity and associations with body composition and lifestyle factors
Fasting glucose levels did not differ between ethnic and BMI groups (Table 3) and all subjects had normal glucose tolerance according to American Diabetes Association Criteria (21). In keeping with SI being significantly lower in the black compared to white women, fasting serum insulin levels were significantly higher in the black women. For all levels of body fat and abdominal fat area, SI was lower in black women (Figure 1a–d). SI correlated negatively with total body fat and abdominal SAT (deep and superficial) in both black and white women, whereas SI correlated with VAT in white women only.
Table 3. Metabolic measures according to BMI and ethnicity
SI did not correlate with SES in black or white women (R = 0.25, P = 0.19 and R = 0.15, P = 0.16, respectively) nor with any measure of physical activity energy expenditure (R = 0.12, P = 0.57 and R = 0.07, P = 0.73, respectively). SI did not correlate with dietary intake in black women, but correlated with total fat and saturated fat intake (g) in white women (R = −0.41, P = 0.05 and R = −0.48, P = 0.02, respectively), independent of total body fat.
We have found that for the same level of adiposity, black women were more insulin resistant and had less VAT and more superficial SAT than white women. Insulin sensitivity did not correlate with VAT in black women. Rather, insulin sensitivity in black women was most closely associated with the subcutaneous depots, in particular deep SAT. In contrast, in white women the association with insulin sensitivity was equal between the VAT and SAT depots.
Previous studies in South Africa and the United States have shown that when matched for body fatness, black women have less VAT and more SAT, but are more insulin resistant than their white counterparts (5,6,7,8,9). However, to our knowledge, this is the first study to show that the higher SAT in obese black women is due to increased accumulation of superficial SAT rather than deep SAT. The relative contribution of abdominal adipose tissue depots to total body mass has implications for metabolic risk, particularly in terms of insulin resistance. Whereas SAT and insulin sensitivity have been shown to be associated (9,10,11,12), typically VAT has been shown to be the major determinant of insulin sensitivity (1,22). This stronger association of VAT with insulin sensitivity is thought to be due to its higher lipolytic activity and greater production of proinflammatory products that drain directly into the hepatic portal system (23). In this study, keeping with this, VAT was a significant determinant of insulin sensitivity in white women. In black women, the lack of association between VAT and insulin sensitivity and the observation that black women have less VAT than white women matched for BMI, despite being more insulin resistant, suggests that VAT is a less important determinant of insulin sensitivity in this population. Rather, the subcutaneous depots were more closely associated with insulin sensitivity in black women. These findings are in accordance with the literature about African Americans in whom the relationship between insulin sensitivity and SAT is stronger than that for VAT (12,24). The association between insulin sensitivity and SAT may be explained by the sheer volume of this depot, which has been shown in other studies to account for ∼80% of systemic circulating free fatty acids and the production of large amounts of adipokines that have effects at the liver, skeletal muscle, and pancreas (23).
Interestingly, deep and superficial SAT depots were equally associated with insulin sensitivity in white women, consistent with similar studies (10,24). In black women, the association between insulin sensitivity and deep SAT was stronger than for superficial SAT. The few studies in which deep SAT was shown to be a stronger correlate of insulin sensitivity included male subjects (10,11), who have greater proportions of deep SAT than superficial SAT (10). Deep SAT has been shown to be a metabolically distinct depot (25), with greater lipolytic activity (26), a lower intracellular pool of glucose transporter-4 and greater expression of resistin than superficial SAT (25).
Independent of total or abdominal fat distribution, there were significant ethnic differences in insulin sensitivity, suggesting that genetic and/or environment factors may also contribute to the variance in insulin sensitivity across ethnic groups. The discrepancies in SES between black and white women in this study do not allow us to distinguish between SES and ethnicity per se. Although we found no associations between SES and insulin sensitivity in either black or white women, a recent study reported that individual and neighborhood SES were independently associated with an increased prevalence of the metabolic syndrome in black and white women (27). Moreover, in a larger cohort of black South African women, we recently found that insulin-resistant women were of a lower SES, less educated, and less employed than their black insulin sensitive counterparts (16).
SES may also impact insulin sensitivity via its effects on lifestyle factors such as dietary intake and physical activity. We found a significant association between dietary fat intake, in particular saturated fat intake, and insulin sensitivity in white women, consistent with previous studies (28,29). Although black women consumed a higher fat diet than white women, we found no association between insulin sensitivity and dietary fat intake, which largely consisted of PUFA, high in n-6 PUFA. This contrasts with a randomized crossover study in which ingestion of a diet high in n-6 PUFA and low in saturated fat for only 5 weeks improved insulin sensitivity compared to a high saturated fat diet (29). However, the beneficial effects of PUFA on insulin sensitivity are only apparent in individuals with a lower total fat intake (<37% of total energy intake) (28), possibly explaining the discrepancy with our findings.
Although total activity was not different between groups, the marked ethnic differences in the domain and intensity of physical activity largely reflected differences in SES. Although leisure time physical activity has been associated with greater insulin sensitivity in black South African women (16), limited access to exercise facilities and issues of safety reduce leisure activity opportunities in underserved communities in South Africa (30). The failure to show an association between physical activity energy expenditure and insulin sensitivity in this study may relate to the small sample size and the variability of this indirect measure of physical activity (18). More direct measures of activity are required in future studies of this nature. Indeed, a recent study using whole room calorimetry showed that black South African women had lower total physical activity energy expenditure than white women matched for age and body fatness (31).
The strengths of this study lie in the high precision of the measurements of insulin sensitivity, body fatness, and body fat distribution. While we only performed single-slice CT scans to measure VAT and SAT areas as opposed to multiple-slice scans that yield volume measurements, we feel our data are still applicable as single-slice CT-derived VAT areas correlate strongly with VAT volume (10). Possible limitations of our study are the relatively small sample size and our inability to separate SES from ethnicity in the examination of the determinants of insulin sensitivity. This inability to differentiate SES and ethnicity could be due to bias in subject selection as the study cohort was not randomly drawn from the population, but rather comprised volunteers who responded to recruitment advertisements. However, our findings of ethnic-specific associations between abdominal adipose tissue depots and insulin sensitivity are in keeping with previous observations reported from the United States (9), suggesting that bias is unlikely to be a major factor. As our study was cross-sectional, future studies are required to examine the efficacy of interventions aimed at reducing insulin resistance in South African women. Dietary, exercise, and pharmaceutical interventions have been shown to result in preferential loss of deep SAT (32,33), the major determinant of insulin resistance in black women.
In conclusion, when matched for body fatness, black South African women were more insulin resistant and had less VAT and more superficial SAT than their white counterparts. Deep SAT was the most significant determinant of insulin sensitivity in the black women, whereas VAT and SAT correlated equally to insulin sensitivity in white women. These differences did not appear to be explained by major disparities in energy consumption or expenditure. Further studies are required to explore the biological basis of the ethnic-specific associations between subcutaneous adiposity and insulin sensitivity.
We thank the research volunteers for their participation in this study, Nandipha Sinyanya for her excellent field work, Judy Belonje for her expert technical assistance and Madelaine Carstens for performing the dietary analyses. Jack Bergman, Naomi Fenton of Symington Radiology, and Linda Bewerunge are thanked for performing the CT and dual-energy X-ray absorptiometry scans. This study was funded by the Medical Research Council of South Africa (Career Development Award to J.H.G.), the International Atomic Energy Agency, the National Research Foundation of South Africa, and Royal Society SA–UK Science Networks Programme, the University of Cape Town, the British Heart Foundation, the Wellcome Trust and the United States Department of Veterans Affairs.