点击显示 收起
1 From the Obesity Research Center, Department of Medicine, St Luke's–Roosevelt Hospital Center, Columbia University, New York, NY (JBA, AJK, LA, MW, EB, NR-K, IJ, SH, and DG), and Novartis Pharma Corporation, Cambridge, MA (BB)
See corresponding editorial on page 1153.
ABSTRACT
Background: African Americans (AAs) have a higher prevalence of obesity and type 2 diabetes than do whites. Higher insulin resistance and hyperinsulinemia have been reported in adult AAs than in whites. Differences in adipose tissue and its distribution may account for these findings.
Objective: The objective was to ascertain whether differences between AA and white women in adipose tissue (AT) and skeletal muscle (SM) volumes account for ethnic differences in insulin resistance.
Design: We used whole-body magnetic resonance imaging to measure AT and SM volumes and used the intravenous-glucose-tolerance test to measure insulin resistance.
Results: AAs (n = 32) were 29–42% more insulin resistant than were whites (n = 28) after adjustment for weight and height or any AT volumes (P < 0.05). After adjustment for SM volume, the difference decreased to 19% and became nonsignificant. AAs had a 163% greater acute insulin response to glucose than did whites; this difference was significant even after adjustment for insulin sensivitity index, weight, height, and any magnetic resonance imaging measures. With respect to regional AT volumes, an association independent of race, weight, height, and SM volume was found only between increased intermuscular AT and lower insulin sensitivity index.
Conclusions: Premenopausal AA women had significantly higher insulin resistance and acute insulin response to glucose than did their white counterparts. Whereas the difference in insulin resistance was partially accounted for by a greater SM volume in the AAs than in the whites, the difference in the acute insulin response to glucose was independent of any AT and SM measures and was disproportionately larger than expected according to the difference in insulin resistance. In addition, whole-body intermuscular AT was an important independent correlate of insulin resistance.
Key Words: Insulin resistance • race • intermuscular adipose tissue • IMAT • acute insulin response to glucose • AIRg • African American women • white women
INTRODUCTION
In the United States, obesity (1, 2) and type 2 diabetes (3-6) are more prevalent in African American adults than in non-Hispanic white adults. In addition, fasting insulin concentrations and insulin resistance—measured by the intravenous-glucose-tolerance test (IGTT), minimal model method, or the euglycemic hyperinsulinemic clamp method—have been reported to be higher in African American adults of all ages, both diabetic and nondiabetic (7-14), and in prepubertal children and adolescents (15-20) than in their white counterparts. These differences were significant even after adjustment for the degree of overweight, fatness, or behavioral factors (or all 3) in the groups of subjects (8, 14, 21-23). A larger insulin response to either oral or intravenous glucose in African American than in white adults and children has also been reported (7-8, 15-16, 20, 22). The underlying cause of these differences is unknown.
Larger amounts of total body fat and greater distribution of fat in the upper body, the visceral adipose tissue (VAT), or the abdominal subcutaneous adipose tissue (abdominal SAT) are associated with increased insulin resistance in populations of adult whites (24-27). Fat accumulation in the intermuscular adipose tissue (IMAT) depot of the thigh—ie, the adipose tissue (AT) under the muscle fascia and between muscle fibers—and lipid accumulation within the muscle cells—ie, intramyocellular lipid (IMCL)—was also associated with increased insulin resistance (28-34). Because measurements of fat distribution in the VAT or abdominal SAT depots at midwaist previously failed to explain the difference in insulin resistance between African American and white women (10, 13, 14), it was suggested that the accumulation of lipid elsewhere, such as between muscle fibers (in IMAT) or within muscle cells (IMCL), may account for this difference (10-14). We used whole-body magnetic resonance imaging (MRI) to investigate this question. This method allows the quantitative measurement of all body AT depots—ie, VAT, SAT (with its subdivisions of upper and lower body), whole-body IMAT, and whole-body skeletal muscle (SM) volume. Whole-body IMAT has been measured in women (35, 36), and its size is similar to that of VAT (35-37). However, both the role of the whole-body IMAT depot (in comparison with that of the VAT and SAT depots) and the relation of whole-body SM volume with respect to insulin resistance in African American and white women have not previously been characterized (28, 29, 36). Therefore, the goal of this study was to ascertain whether AT distribution—VAT, SAT, or IMAT (or all 3) and whole-body SM volumes, measured by whole-body MRI—can account for the differences in insulin resistance and in acute insulin response to glucose (AIRg) between African American and white women.
SUBJECTS AND METHODS
Subjects
Subjects were 28 white and 32 African American premenopausal women, lean and obese, aged >22 y. Women were recruited in an unselected fashion (except for exclusion criteria listed below) with a wide range of body mass index (BMI; in kg/m2): from 18 to 44. Subjects were included in the study if all 4 grandparents were reported to be of the same ancestry (ie, either white or African American). Inclusion criteria required that subjects had no chronic illnesses, did not take any medication on a regular basis, were not taking oral contraceptives, were weight stable (±2 kg) for > 6 mo before screening, and reported regular menstrual cycles at the time of the study. Subjects with diabetes were excluded from participation in this study after undergoing a screening oral-glucose-tolerance test; however, 3 African American women and 3 white women with impaired glucose tolerance were retained in the study (38). The women in the current study were sedentary, and the measurement of physical activity in a subset by using a questionnaire (data not shown) did not identify any differences between African Americans and whites. All studies were carried out at the General Clinical Research Center of St Luke's Hospital.
The subjects provided written informed consent before participating in the study. The Institutional Review Board of the Health Sciences Institute at St Luke's–Roosevelt Hospital Center approved the study.
Anthropometric measurements
Fasting weight and height were measured while the subjects were wearing undergarments. Minimum waist circumference (minimum circumference between the lower rib margin and the iliac crest, usually at the midpoint or midwaist) and maximum hip circumference (below the iliac crest, with the subject viewed from the front) were measured while the subjects were standing upright, with the heels together.
Skeletal muscle and regional adipose tissue volumes by whole-body magnetic resonance imaging
SM, VAT, SAT, and IMAT volumes were measured by using a whole-body MRI protocol as described previously (36, 37, 39). The entire body was visualized on a scout coronal image (6X Horizon; General Electric, Milwaukee, WI) and the axial level of L4–L5 was identified. The scans were acquired by using contiguous 10-mm axial slices, taken at 40-mm intervals, from below L4–L5 to the toes and from above L4–L5 to the fingertips (40–50 images for women of average height). SLICE-OMATIC image analysis software (version 4.2; Tomovision, Montreal, Canada) was used to analyze images on a PC workstation (Gateway, Madison, WI). All scans were read by the same technician (IJ). The technical error for 3 repeated readings by the same observer (IJ) of the same scan of MRI-derived SM, SAT, VAT, and IMAT volumes in our laboratory was 1.9%, 0.96%, 1.97% and 0.65%, respectively.
We defined IMAT as the AT visible between the muscle groups and beneath the muscle fascia (36). The gray-level intensity (threshold value) of the AT in the SAT region was measured and used as reference. This threshold value was reduced by 20% to identify the IMAT threshold. SAT and VAT were divided into upper- and lower-body regions at the L4–L5 vertebral space, which was identified for each scan on the scout coronal image. The procedure for calculating regional adipose tissue (AT) volume is to measure the relevant tissue in each slice by using threshold methods and manual delineation to draw boundaries between different tissues. The volume between slices is extrapolated from the area measurements.
Given the possibility that the tissue designated as SM by MRI measurements could have different densities in African American than in white women (see Discussion), we chose in the current study to express the SM measurement as volume rather than as weight. We then extended this to the other MRI measurements—specifically, AT—and thus all MRI measures are expressed in this study in liters (L).
Insulin sensitivity index and acute insulin response to glucose during the intravenous-glucose-tolerance test
The Bergman minimal model (MINMOD software, version 2.0; copyright RN Bergman, 1986) was used to quantify the insulin sensitivity index (SI). This measurement was made in all subjects during the follicular phase of the menstrual cycle. Glucose (0.3 g/kg body wt; 50% dextrose injection: Abbott, North Chicago, IL) was administered intravenously at time zero (T0). This was followed by an injection of tolbutamide (Orinase Diagnostic; Upjohn, Kalamazoo, MI) at T20 min. In 8 African Americans and 7 whites, insulin (0.03 units/kg; Humulin R; Lilly Inc, Indianapolis, IN) was used because of the unavailability of tolbutamide. Blood sampling, performed through a catheter placed in the contralateral arm, occurred at 30 time points during the 3 h after glucose administration. Plasma glucose and insulin were measured on all samples, and SI was calculated from these values by using the nonlinear mathematical (minimal) model of glucose disappearance. Studies in the literature show that the SI measured by using IGTT with intravenous insulin at T20 min is lower than that measured in the same subject by using IGTT with intravenous tolbutamide at T20 min (39). However, the difference between methods (14%) appears to be constant throughout the range of SI in nondiabetic subjects (39). Therefore, we used a value 14% higher than that obtained by the computer program for the subjects who underwent IGTT with insulin at T20 min, and we pooled the data. AIRg was calculated as the mean incremental value over baseline for the first 8 min after the glucose injection (20, 27). In addition, a categorical factor denoting the insulin subgroup (IGTT method factor) was entered in all relevant analyses of SI (see Statistical analysis).
Statistical analysis
Comparisons between groups were made by using analysis of variance (ANOVA) and analysis of covariance (ANCOVA) after assessment of the homogeneity of variance and of slopes. A general linear model (GLM) was used to model dependent variables (SI and AIRg) from categorical (ethnicity) and continuous (age, height, weight, or regional AT and SM volumes or all) independent variables. Both the significance of interactions between the independent variables and the significance of the regression coefficients were calculated for each analysis. Residuals after GLM were checked for normality, and naturally log-transformed dependent variables were used when a significant deviation from normality was detected. The GLM was used to determine which regional AT variable (ie, IMAT, SAT, or VAT) was associated with SI or AIRg independent of weight and height, SM volume, or ethnicity.
Two outliers were found for the AIRg variable (values >2 SD over the group mean for AIRg in the African American group). Because a review of data collection and assays for these data points confirmed the results, because none of the other variables measured in these 2 subjects were outliers compared with the means for their respective groups, and because analyses to measure AIRg gave substantially the same results whether or not these data points were included, we report the results with these data points included.
Because SI was measured by IGTT with the use of insulin instead of tolbutamide in a subgroup of women (7 African American and 8 white), a categorical factor denoting this subgroup (IGTT method factor) was entered in all the ANCOVA and GLM analyses of SI. Possible interactions were computed. No significant interactions were found in any of the analyses; the P value for this effect consistently was > 0.2. Therefore, all results of the IGTT method are reported for the combined groups. Analyses were conducted with STATISTICA software (version 6.0; Statsoft Inc, Tulsa, OK). P < 0.05 was considered significant.
RESULTS
Anthropometric and whole-body magnetic resonance imaging measurements
Weight was significantly higher in the African American than in the white women, whereas BMI, waist circumference, and waist-to-hip ratio were higher, but not significantly so, in the African American women (P > 0.1) (Table 1). Whole-body MRI measurements are shown in Table 2. Total SAT volume and its subdivisions at the L4–L5 vertebral space (upper- and lower-body SAT) were greater in the African Americans than in the whites but did not differ significantly between the groups (P > 0.15 for all). Total VAT and its subdivisions (upper- and lower-body VAT) did not differ significantly by race (P > 0.8, 0.6, and 0.2, respectively). Similar results were obtained for the abdominal SAT and VAT areas at L2–L3, which were calculated to allow comparison with previous studies (26). IMAT was larger in the African American than in the white women, but this difference was not statistically significant (P = 0.054). Mean SM volume was significantly larger in the African Americans than in the whites (P = 0.000031). In part, this was because the slope of SM (L) versus body weight (kg) was significantly steeper in the African American than in the white women (0.17 and 0.11 L · kg–1, respectively; P = 0.033) (Figure 1).
View this table:
TABLE 1. Characteristics and anthropometric measurements of the African American and white women1
View this table:
TABLE 2. Whole-body magnetic resonance imaging measurements of the study participants by race1
FIGURE 1.. Relation between body weight and skeletal muscle volume, measured with whole-body magnetic resonance imaging (MRI), in African American (•, straight line) and white (, dashed line) women. Skeletal muscle was measured by whole-body MRI. The slopes of the regression lines were significantly different (P < 0.05, general linear model).
Insulin resistance
The African American women were significantly more insulin resistant (ie, had lower SI) than were the white women (Table 3). We used ANCOVA to investigate whether AT and SM volume differences accounted for the difference in insulin resistance. Slopes for the relation between insulin resistance and covariates did not differ significantly by race (P > 0.2). The African American women remained more insulin resistant (ie, had lower mean SI) than did the white women after adjustment for weight and height (3.4 and 4.9 µU/mL–1 · 10–4 · min–1, respectively; P < 0.01), VAT (3.1 and 5.3 µU/mL–1 · 10–4 · min–1, respectively; P < 0.001) (Figure 2), SAT (3.3 and 5.0 µU/mL–1 · 10–4 · min–1, respectively; P < 0.01) (Figure 3), and IMAT (3.4 and 4.8 µU/mL–1 · 10–4 · min–1, respectively; P < 0.05) (Figure 4). The substitution of SAT and VAT subdivisions or areas at L2–L3 for the total VAT and SAT volumes did not change the results. However, when SM volume was used as a covariate, the difference in insulin resistance, albeit still present, was smaller and no longer statistically significant (3.7 and 4.6 µU/mL–1 · 10–4 · min–1, respectively; P = 0.17) (Figure 5).
View this table:
TABLE 3. Metabolic measurements of the study participants by race1
FIGURE 2.. Relation between total visceral adipose tissue (VAT) volume, measured with whole-body magnetic resonance imaging, and the insulin sensitivity index (SI), calculated by using the minimal model, from data obtained during the intravenous-glucose-tolerance test in African American (•, straight line) and white (, dashed line) women. The slopes of the regression lines were not significantly different (P = 0.42). SI was significantly lower in the African American than in the white women (P < 0.001, analysis of covariance) after adjustment for VAT. A general linear model was used to test homogeneity of slopes.
FIGURE 3.. Relation between total subcutaneous adipose tissue (SAT) volume, measured with whole-body magnetic resonance imaging, and the insulin sensitivity index (SI), calculated by using the minimal model, from data obtained during the intravenous-glucose-tolerance test in African American (•, straight line) and white (, dashed line) women. The slopes of the regression lines were not significantly different (P = 0.47). SI was significantly lower in the African American than in the white women (P < 0.001, analysis of covariance) after adjustment for SAT. A general linear model was used to test homogeneity of slopes.
FIGURE 4.. Relation between total intermuscular adipose tissue (IMAT) volume, measured with whole-body magnetic resonance imaging, and the insulin sensitivity index (SI), calculated by using the minimal model, from data obtained during the intravenous-glucose-tolerance test in African American (•, straight line) and white (, dashed line) women. The slopes of the regression lines were not significantly different (P = 0.32). SI was significantly lower in the African American than in the white women (P < 0.05, analysis of covarianc) after adjustment for IMAT. A general linear model was used to test homogeneity of slopes.
FIGURE 5.. Relation between total skeletal muscle (SM) volume, measured with whole-body magnetic resonance imaging, and the insulin sensitivity index (SI), calculated by using the minimal model, from data obtained during the intravenous-glucose-tolerance test in African American (•, straight line) and white (, dashed line) women. The slopes of the regression line were not significantly different (P = 0.86). SI did not differ significantly between the African American and the white women (P = 0.17, analysis of covariance) after adjustment for SM. A general linear model was used to test homogeneity of slopes.
GLMs were then used, in addition to ethnicity, to measure the contribution of larger SM or regional AT compartments (ie, VAT, SAT, or IMAT) to a lower SI. Ethnicity, weight, height, and SM, VAT, SAT, and IMAT (each of the latter considered separately) were added in the model. Because the residual values of SI obtained after adjustment in the multivariate analysis were not always normally distributed, log SI was used for these analyses. High IMAT but not high VAT or SAT was a predictor of low SI, independent of ethnicity and weight and height, and of SM (P < 0.05). In addition, high IMAT predicted a low SI, independent of VAT and SAT (entered in the model in addition to weight, height, and IMAT). The substitution of SAT and VAT subdivisions or areas at L2–L3 for the total VAT and SAT volumes did not change the results. SM volume was not an independent predictor of low SI in any of these models (P > 0.3 for all). Ethnicity remained an independent predictor of insulin sensitivity (significantly lower in the African Americans than in the whites: P < 0.05) in all models except those including both SM and IMAT volumes. The addition of age to any of these models did not change the results.
Acute insulin response to intravenous glucose
The values for the AIRg, measured during the first 8 min of the frequently sampled IGTT, were not normally distributed, and log AIRg was used for all analyses. The African American women had significantly (P < 0.001) higher unadjusted mean log AIRg than did the white women: 1.8 compared with 1.4 pmol · L–1 · min–1 (Table 3). This difference was significant even after adjustment for the degree of insulin resistance (SI) (Figure 6): adjusted mean log AIRg was 1.7 and 1.5 pmol · L–1 · min–1, respectively (P < 0.001). The slope of the log AIRg versus body weight was significantly steeper for the African American than for the white women (P < 0.05, not shown), but interactions by ethnicity observed for the relations of log AIRg to body-composition variables were not statistically significant [the ethnicity x SAT interaction, P = 0.054; the ethnicity x all other variables (ie, SM, VAT, and IMAT) interaction, P > 0.2 for all]. AIRg remained significantly higher in the African American than in the white women after adjustment for SM, VAT, SAT, or IMAT (P < 0.05 for all; data not shown). Because the slope of log AIRg versus body weight differed significantly between the 2 groups, the contribution of increased SM or regional AT compartments (ie, VAT, SAT, or IMAT) to a higher AIRg, independent of weight, was ascertained separately in the 2 groups. We used a set of GLMs in which, along with weight and height, the SM, VAT, SAT, and IMAT volumes were added separately as predicting variables in the model. In each group, neither SM nor any of the AT compartments predicted a higher AIRg, independent of weight and height. The addition of age to any of these models did not change the results.
FIGURE 6.. Relation between the insulin sensitivity index (SI) and the acute insulin response to intravenous glucose (AIRg) in African American (•, straight line) and white (, dashed line) women. SI was calculated by using the minimal model, and AIRg was calculated as the mean incremental value over baseline for the first 8 min after the glucose injection, with the use of data obtained during the intravenous-glucose-tolerance test. The slopes of the regression lines did not differ significantly (P = 0.15). AIRg was significantly higher in the African American than in the white women (P < 0.005, analysis of covariance) after adjustment for SI. A general linear model was used to test homogeneity of slopes.
DISCUSSION
The hypothesis that African Americans are more insulin resistant than are other racial groups (5) has been supported in several studies (7-15, 21-23). In those studies, differences in body weight or body composition (fat mass or fat-free mass) did not account for this higher degree of insulin resistance (7-15, 23). In the current study, we found significantly higher insulin resistance in the African American than in the white premenopausal women, and we examined whether this difference could be explained by AT distribution differences measured by whole-body MRI. Specific and unique aspects of our study were the measurement of whole-body SM volume and of its associated AT depot, IMAT. We found a significantly steeper increase in SM volume with an increase in weight and a borderline significantly higher IMAT volume (P = 0.054) in the African American than in the white women. The other AT compartments—ie,VAT, SAT, and SAT distribution (upper- and lower-body SAT, as well as the ratio of upper- to lower-body SAT; data not shown)—did not differ significantly between the groups (P > 0.15), whereas some previous reports suggested differences in upper-body fat distribution between African American and white women (26).
In the current study, none of the differences in the AT compartments statistically accounted for the difference in insulin resistance between the African American and the white women. Although we previously reported differences in whole-body IMAT between these 2 groups, albeit at higher levels of adiposity (37), adjustment for IMAT in the current study only partially decreased the difference in insulin resistance between groups (P = 0.05). However, adjustment for the SM volume, in ANCOVA or in statistical models that included weight, height, and IMAT, rendered the difference in insulin resistance statistically nonsignificant (P > 0.07). Therefore, we hypothesize that some aspect of the increased SM volume or an associated variable could partly explain the greater insulin resistance in the African American than in the white women. A greater amount of low-attenuation or low-density thigh muscle measured by computed tomographic scan was previously found in African American than in white women (12, 14). Low-density muscle signifies the deposit of a greater amount of lipid, free of visible AT, in and around the muscle (30). Because we measured IMAT separately from SM in the current study, the only amounts of muscle lipid unaccounted for would be lipid in nonvisible adipocytes (extramyocellular, EMCL) or inside the myocytes (IMCL). These difference in the size of these lipid depots would be too small to account entirely for the large difference in SM volume between the African American and the white women, but, if larger amounts of IMCL were present in the African Americans, that could account for their greater insulin resistance compared with that of the whites. However, this hypothesis will have to be confirmed by direct measurements of IMCL and EMCL.
Greater SM mass, measured as appendicular lean mass with the use of dual-energy X-ray absorptiometry (DXA) measurements, was reported in African Americans than in whites (41, 42). The fact that DXA separates all lipid from the lean tissue is further evidence that the greater SM (relative to weight) in the African Americans than in the whites cannot solely be explained by high IMCL and EMCL. Longer appendicular bone length has been proposed as a cause for the greater appendicular muscle mass in the African Americans than in the whites (41, 42). High androgens also could potentially be associated with both a larger SM mass and greater insulin resistance (43). The women in our study had regular menstrual cycles, and comparisons of testosterone and free testosterone concentrations in a subset did not identify any significant differences between the groups. However, our current data set is too limited to allow such hormonal differences to be entirely ruled out.
Some previous studies reported similar insulin resistance in the 2 groups (44-48), but others reported greater insulin resistance in the African Americans than in the whites (7-14). Differences in the amount of VAT may have masked ethnic differences in one of these studies (44), whereas, in others (45, 46), ethnic differences in insulin resistance may have been diminished by similarly large amounts of VAT in both cohorts. This possibility would be consistent with the reports of weaker relations (shallower slopes) of insulin resistance in association with greater VAT in African Americans than in whites (25, 47, 48). The range of VAT used and the lack of differences in VAT volume in the current study (P = 0.9) may have allowed us to find the ethnic differences in insulin resistance.
We also found a significantly higher AIRg in the African American than in the white women, and the difference persisted after adjustment for the AT and SM volumes and for SI. Although higher AIRg was reported in African American than in white adults (7), this difference was attributed to the greater insulin resistance in the former. A higher AIRg, disproportionate to the degree of insulin resistance, has been reported in African American than in white children (20, 21). We confirm here that, just as in children, nondiabetic premenopausal African American women have higher AIRg, at a rate out of proportion to their degree of insulin resistance, than do their white counterparts.
Among regional AT depots, a higher insulin resistance was independently associated only with a higher IMAT. Two previous studies related thigh (28) and whole-body (35) IMAT to insulin resistance, but neither showed an independent relation. It is not clear whether this discrepancy is due to differences in subject characteristics or to differences in the methods used, because neither of these studies specifically reported on the ethnicity of the populations studied. Independent associations of insulin resistance with IMAT and the disproportionate increase in AIRg may have important implications for the African American population. Increased IMAT and insulin resistance, with perhaps an associated increase in IMCL, may occur early in African Americans as they gain weight. Because insulin resistance and AIRg are the main determinants of glucose tolerance (49), any additional sources of insulin resistance or the inability to adequately increase AIRg (or both) will increase the risk of type 2 diabetes. Indeed, a lack of increase in AIRg at higher VAT was recently reported in African Americans with impaired glucose tolerance (25), African American adolescents (48), and African American low-birth-weight children (50) than in their white counterparts.
Our study's cross-sectional nature hinders the drawing of cause-and-effect inferences between associated variables, and the size of our population sample precludes a definitive conclusion with regard to interactions in the relations of SI and AIRg with some of the measured variables. The current results must be confirmed in larger samples that include men and women of all ages. Our study's strengths are the use of whole-body MRI measurements, which delineate more detailed AT depots, including IMAT, and the direct measurement of SI by IGTT.
In conclusion, African American premenopausal women have a higher degree of insulin resistance than do white premenopausal women, and this difference is partly accounted for by the greater SM volume in the former. In the 2 groups combined, insulin resistance is independently predicted by high IMAT. African American women also have a higher AIRg than do white women; the difference is disproportionately greater than expected according to the degree of insulin resistance, and it persists after adjustment for SM or any AT distribution measures. These findings underscore the complex relation between fat distribution and the factors that ultimately determine glucose tolerance, ie, insulin resistance and the insulin response to glucose.
ACKNOWLEDGMENTS
We thank the members of the Hormone and Metabolite and Body Composition Core Laboratories of the New York Obesity Research Center and especially Xavier Pi-Sunyer for support of the collection of data for this project.
JBA was responsible for the design of the study; JBA, AJK, LA, MW, EB, NR-K, and BB were responsible for subject recruitment and data collection; DG and IJ were responsible for magnetic resonance imaging analyses; JBA, AJK, and SH were responsible for data analysis; SH provided statistical expertise; JBA was responsible for data interpretation and manuscript writing; and JBA, AJK, LA, MW, EB, BB, SH, and DG were responsible for critical review of the manuscript for intellectual content. BB is an employee of Novartis Pharma Corporation; JBA has received research support from Novartis Pharma Corporation to study insulin resistance and its relation with body composition before and after weight loss. None of the other authors had any personal or financial conflicts of interest.
REFERENCES
Related articles in AJCN: