Endocrinology and Metabolism

Subdivisions of subcutaneous abdominal adipose tissue and insulin resistance

David E. Kelley, F. Leland Thaete, Fred Troost, Trina Huwe, Bret H. Goodpaster


Whereas truncal (central) adiposity is strongly associated with the insulin resistant metabolic syndrome, it is uncertain whether this is accounted for principally by visceral adiposity (VAT). Several recent studies find as strong or stronger association between subcutaneous abdominal adiposity (SAT) and insulin resistance. To reexamine the issue of truncal adipose tissue depots, we performed cross-sectional abdominal computed tomography, and we undertook the novel approach of partitioning SAT into the plane superficial to the fascia within subcutaneous adipose tissue (superficial SAT) and that below this fascia (deep SAT), as well as measurement of VAT. Among 47 lean and obese glucose-tolerant men and women, insulin-stimulated glucose utilization, measured by euglycemic clamp, was strongly correlated with both VAT and deep SAT (r = −0.61 and −0.64, respectively; both P < 0.001), but not with superficial SAT (r = −0.29, not significant). Also, VAT and deep SAT followed a highly congruent pattern of associations with glucose and insulin area under the curve (75-g oral glucose tolerance test), mean arterial blood pressure, apoprotein-B, high-density lipoprotein cholesterol, and triglyceride. Superficial SAT had markedly weaker association with all these parameters and instead followed the pattern observed for thigh subcutaneous adiposity. We conclude that there are two functionally distinct compartments of adipose tissue within abdominal subcutaneous fat and that the deep SAT has a strong relation to insulin resistance.

  • obesity
  • insulin resistance
  • visceral adiposity

in obesity, insulin resistance (IR) is affected not only by the total amount but also the distribution of adipose tissue (AT) and, in particular, an upper body or truncal fat distribution (10,36). During the past decade and more, the study of abdominal adipose tissue distribution by computed tomography (CT) and magnetic resonance imaging has identified the importance of visceral adiposity (11, 25,32). There is a strong association between visceral adiposity and IR patterns of glucose homeostasis (7, 9, 31), including in type 2 diabetes mellitus (4). As well, visceral AT (VAT) is related to essential hypertension, dyslipidemia, and other factors, such as impaired fibrinolysis, that contribute to a heightened risk of cardiovascular disease (5, 8, 33). Moreover, changes in VAT, whether as increases over time or decreases in relationship to weight loss interventions, predict associated changes in glucose and insulin homeostasis (14, 16, 26).

However, a debate has arisen regarding the singular importance of visceral compared with subcutaneous AT (SAT) in relation to IR. Abate and colleagues (1, 2), reported that subcutaneous rather than visceral adipose tissue volume was the stronger correlate with insulin-resistant glucose metabolism. Perhaps the difference is related to gender, as the studies of Abate and colleagues involved men, whereas several of the key studies that find greater importance of VAT were performed in women (7, 9, 31). However, in a recent cross-sectional study of lean and moderately obese men and women, we found that VAT and SAT had a similar association with IR (17). Furthermore, among women, Carey et al. (6) found that truncal adiposity, measured by dual-energy X-ray absorptiometry (DEXA), a technique that does not distinguish between visceral and subcutaneous depots, had a robust relation with IR. In addition, Misra et al. (29) reported that SAT of the abdomen located in the posterior as opposed the anterior half of the abdomen was the site of adipose tissue deposition that was more strongly associated with IR. To account for this curious observation, these investigators suggested that the SAT of the posterior half of the abdomen is mostly “deep” subcutaneous adipose tissue (deep SAT).

There is a well described fascial plane within the SAT of the abdomen (18, 28), with the superficial adipose layer possessing compact fascial septa (Camper's fascia), whereas the deeper layer of adipose tissue has more loosely organized fascial septa (Scarpa's fascia). Fat lobules of the two sites also differ. The superficial layer is characterized by small tightly packed lobules, whereas those of the deeper layer are larger and distributed in an irregular manner (28). The thickness of the deep layer appears more variable among individuals and especially in relation to obesity (3). The presence of these fascial planes and differences in histology are well recognized with respect to liposuction, which generally is targeted toward the deep layer (15, 23). Given the anatomical basis for considering the two layers of SAT in the abdomen different and the ability to delineate the fascial plane utilizing CT (22), the current study was undertaken to examine these adipose tissue depots from a metabolic perspective. The related purpose was to address current controversies regarding the importance of subcutaneous abdominal adipose tissue in relation to IR.



Volunteers were recruited by advertisement, and 47 men and premenopausal women who met inclusion and exclusion criteria were selected to participate. Fifteen volunteers (7 men and 8 women) were lean [body-mass index (BMI) <25 kg/m2], and 32 volunteers (15 men and 17 women) were obese (BMI >27 kg/m2). Inclusion criteria included normal glucose tolerance, being nonsmokers, not taking any chronic medications, having stable weight for ≥6 mo preceding the studies, and being sedentary (defined as not engaged in >2 exercise sessions weekly). Individuals with diabetes mellitus, hyperlipidemia, coronary heart disease, or vascular disease were excluded, as were those having hypertension or being treated with antihypertensive agents. As part of the screening examination, volunteers fasted overnight and in the morning had a 75-g oral glucose tolerance test with venous samples for glucose and insulin obtained at 30-min intervals for 2 h. The protocol had been reviewed and approved by the University of Pittsburgh Institutional Review Board, and all volunteers had given written informed consent.

Computed tomography.

Computed tomography (9800 CT scanner, General Electric, Milwaukee, WI) was used for determinations of VAT and SAT areas, as previously described (35). Briefly, a cross-sectional scan of 10 mm thickness centered at the L4-L5 vertebral disc space was obtained using 170 mA with a scanning time of 2 s and a 512 × 512 matrix. The boundary between visceral and subcutaneous AT was defined by use of the abdominal wall musculature in continuity with the deep fascia of the paraspinal muscles, as has been previously described (11). This method was then modified to determine the areas of deep and superficial SAT, according to the method of Johnson et al. (23). To do this, the image was purposely displayed at a more negative attenuation setting to clearly visualize the fascial plane within SAT, and a cursor was used to demarcate this boundary. Representative images using this approach are shown in Fig. 1.

Fig. 1.

Representative cross-sectional abdominal computed tomography (CT) scans of a lean (A) and an obese (B) research volunteer are shown with demarcations of visceral adipose tissue (AT; large arrowheads), deep subcutaneous (open arrows) and superficial subcutaneous (closed arrows) AT depots. The fascia (small arrowhead) within subcutaneous abdominal AT was used to distinguish superficial from deep depot. In the two CT scans shown, area of superficial subcutaneous AT was similar (144 vs. 141 cm2), whereas areas for deep subcutaneous (126 vs. 273 cm2) and visceral AT (VAT) (84 vs. 153 cm2) were quite different. Insulin-stimulated glucose metabolism was 6.1 and 4.0 mg ⋅ min 1 ⋅ kg FFM 1 in lean and obese volunteers, respectively. FFM, fat-free mass.

Within each of the three regions defined by the cursor (i.e., VAT, deep SAT, and superficial SAT), the cross-sectional area of AT was measured as the pixels (0.6 mm) in the fat density range [−190 to −30 Hounsfield Units (HU)], using commercially available software (GE Medical Systems, Milwaukee, WI). To determine the proportions of anterior and posterior SAT and with respect to anterior and posterior distribution of deep SAT and superficial SAT, the method of Misra et al. (29) was used. This entailed placing a line in the anterior-to-posterior direction through the middle of the vertebral body and then placing a line perpendicular to this at the midpoint to separate the anterior from the posterior compartment. To determine the area of subcutaneous adipose tissue in the thigh, cross-sectional CT of the midthigh was performed, as previously described (24).

Dual-energy X-ray absorptiometry.

To determine overall fat mass (FM) and fat free mass (FFM), dual-energy X-ray absorptiometry (DEXA) (Lunar model DPX-L, Madison, WI) was performed using software version 1.3Z. This computerized analysis was also employed to measure abdominal FM. Abdominal AT volume was assessed according to the method of Jensen et al. (21). Briefly, a region of interest on the trunk was defined by using the diaphragm as the cephalad limit and the top of the iliac spine as the caudal boundary. Care was taken to position each volunteer with arms lifted away from the trunk so as to include only the abdominal area in this analysis.

Insulin sensitivity.

For subjects to be prepared for determinations of insulin sensitivity, they were instructed to consume a weight-maintaining diet containing ≥200 g of carbohydrate for ≥3 days before measurements of insulin sensitivity and to avoid exercise or strenuous exertion for 2 days before these studies. On the day before measurement of insulin sensitivity, subjects were admitted to the University of Pittsburgh General Clinical Research Center. That evening, they received a standard dinner (10 kcal/kg: 50% carbohydrate, 30% fat, 20% protein) and then fasted until completion of the insulin sensitivity measurement. An overnight timed urine collection (∼12 h duration) was obtained for nitrogen measurements to estimate protein oxidation with systemic indirect calorimetry. At about 7:00 AM, a catheter was placed in a forearm vein for later infusion of glucose and insulin and to start a primed (20 μCi), continuous (0.20 μCi/min) infusion of HPLC-purified [3-3H]glucose (New England Nuclear, Boston, MA). Isotope was started 100 min before beginning the insulin infusion to allow 4 h for isotope equilibration before determination of glucose specific activity and rates of systemic glucose utilization during the final 40 min of the 3-h insulin infusion. Data on these procedures have been reported previously (16,17). Samples were collected for determination of serum insulin, cholesterol, triglyceride, and leptin. Supine blood pressure in the right arm was recorded. Continuous infusion of regular insulin (Humulin, Eli Lilly, Indianapolis, IN) was given at a rate of 40 mU ⋅ m 2 ⋅ min 1for 3 h. Euglycemia was maintained with an adjustable infusion of 20% dextrose (500 ml) to which 80 μCi of [3-3H]glucose was added to maintain stable plasma glucose specific activity (12). Plasma glucose determined at 5-min intervals during the clamp remained stable during the last 40 min [5.04 ± 0.11 (SD) mM; coefficient of variation = 3.5%]. Blood samples for measurement of [3-3H]glucose specific activity were collected every 10 min during the final 40 min of insulin infusion. Systemic indirect calorimetry was performed in the postabsorptive state and during the last 30 min of insulin infusion by use of an open circuit spirometry metabolic monitor system (DeltaTrac, Anaheim, CA), to estimate glucose and fat oxidation (13).


Plasma glucose was measured by an automated glucose oxidase reaction (YSI 2300 Glucose Analyzer, Yellow Springs, OH). Glucose specific activity was determined with liquid scintillation spectrometry after deproteinization of the plasma with barium sulfate and zinc hydroxide. Serum insulin and leptin were determined by commercially available RIA kits (Pharmacia, Uppsala, Sweden; Linco Research, St. Louis, MO). For the insulin RIA, cross-reactivity of insulin with proinsulin was 41%, and the inter- and intra-assay coefficients of variation were between 5.4 and 5.8%. Serum cholesterol and triglycerides were determined by a spectrophotometric assay; serum high-density lipoprotein (HDL) cholesterol, total cholesterol, triglycerides, and apoprotein-B (apoB) were measured by spectrophotometric assay.


Rate of plasma glucose appearance and utilization (Rd) were calculated with the Steele equations (34), as modified for variable rate glucose infusions that contain isotope (12). The area under the curve (AUC) for glucose and insulin after glucose ingestion was calculated by a trapezoid method.

Statistical analysis.

Data are presented as means ± SD, unless otherwise indicated. ANOVA was used to compare men and women and lean and obese groups for body composition parameters and clinical and metabolic characteristics. Regression analysis was used to determine the relationships between body composition and insulin sensitivity. All statistical measurements were made with SigmaStat (version 2.0, Jandel Scientific Software, San Rafael, CA).


Body composition.

The body composition characteristics (means ± SE) of the lean and obese men and women are shown in Table 1. BMI differed for lean and obese groups as by study design but without any effect of gender. However, systemic FM did differ by gender, being greater in women as well as being greater in obese subjects. This pattern also pertained to thigh AT and for superficial SAT of the abdomen, because women had greater AT in these depots. For deep SAT in the abdomen, there was not any effect of gender. However, mean values in obese subjects were approximately threefold greater than in lean subjects. VAT was approximately two- to threefold greater in obese subjects compared with lean subjects, and there was a tendency for men to have greater VAT, but this gender difference did not achieve statistical significance (P = 0.08). The volume of abdominal AT, determined by DEXA, was greater in obesity. Abdominal adipose tissue measured by DEXA was highly correlated to cross-sectional abdominal AT area (cm2) measured by CT (r = 0.97,P < 0.01).

View this table:
Table 1.

Body composition in men and women

Deep SAT in the abdomen of obese subjects comprised a significantly greater cross-sectional area than did VAT (P < 0.001). The majority of deep SAT was located in the posterior (dorsal) half of the abdomen; the mean value was 76 ± 1% located posteriorly with a relatively narrow range (67–87%). Obesity and gender did not influence this pattern. Superficial SAT was more evenly distributed around the circumference of the abdomen with a mean value of 55 ± 2 and 45 ± 2% anterior and posterior, respectively. There were differences in proportions of superficial SAT and deep SAT (expressed as a percentage of cross-sectional abdominal AT) when examined by gender and in relation to obesity. Superficial SAT accounted for 45 ± 3 and 41 ± 2% of abdominal AT in lean and obese women, which was significantly greater (P < 0.01) than the respective amounts in lean and obese men at 28 ± 3 and 28 ± 2%. For deep SAT, expressed as a percentage of cross-sectional abdominal AT, there was a significant effect of obesity (P < 0.01) but not of gender. The values for lean and obese women were 32 ± 3 and 37 ± 2% and for lean and obese men the percentages were 36 ± 3 and 44 ± 2%. For VAT, the percentages of abdominal AT were 23 ± 3, 36 ± 3, 20 ± 2, and 27 ± 2%, for lean and obese women and lean and obese men, respectively, and these differences were significant for obesity (P < 0.01) but not for gender. VAT was highly correlated with deep abdominal AT (r = 0.76, P < 0.01), but the association with superficial abdominal AT was more modest (r = 0.43, P < 0.01).

Metabolic variables.

Values for insulin-stimulated glucose metabolism (glucose Rd), 2-h glucose tolerance AUC, 2-h insulin AUC, fasting plasma insulin, leptin, and lipids are shown in Table2. Glucose Rd was reduced in obesity. However, consistent with the inclusion criteria of normal glucose tolerance, there was not a significant difference between lean and obese subjects for glucose AUC, whereas insulin AUC did differ; obese men and women had higher fasting insulin (P < 0.001). There were also highly significant group and gender differences for plasma leptin (both P < 0.001), with leptin higher in women and in the obese group. Despite the fact that none of the volunteers was hypertensive, mean arterial blood pressure (MAP) was significantly higher in obesity (P < 0.001) and in men (P < 0.01). Values for MAP were: 82 ± 3, 88 ± 3, 89 ± 2, and 94 ± 2 mmHg, respectively, in lean women and men and obese women and men (Fig.2). There were no significant differences by obesity or gender for HDL cholesterol, apoB, triglyceride, or LDL cholesterol, as shown in Table 2.

View this table:
Table 2.

Metabolic variables in men and women

Fig. 2.

A: rates of insulin-stimulated glucose metabolism (mg ⋅ min 1 ⋅ kg FFM 1) plotted against cross-sectional areas of VAT (cm2), in women (●) and men (○); overall correlation is significant: r = −0.61, P < 0.001. B: values of overall subcutaneous abdominal AT (SAT) plotted against insulin-stimulated glucose metabolism; overall correlation, r = −0.53, P < 0.001. C: values of deep SAT plotted against insulin-stimulated glucose metabolism; overall correlation, r = −0.64, P < 0.001. D: values of superficial SAT plotted against insulin-stimulated glucose metabolism; overall correlation, r = −0.29, P < 0.001.

Relation of abdominal AT depots to metabolic variables.

Shown in Table 3 are the correlation coefficients between glucose Rd, glucose AUC, and insulin AUC with regard to the distribution of abdominal AT. With respect to glucose Rd, there was a significant negative correlation with FM but not with thigh AT or superficial SAT; for these the correlation coefficients were very similar. The plots relating glucose Rd to subdivisions of abdominal adipose tissue are shown in Fig. 2. However, deep SAT was significantly and negatively correlated with glucose Rd, and the correlation coefficient was highly similar to that of VAT. Moreover, the sum of VAT and deep SAT had a slightly stronger correlation (r = 0.68; P < 0.001) than either component considered individually. In multiple regression analysis, total body fat and VAT accounted for 45% of the variance in insulin sensitivity, yet deep SAT remained independently associated with insulin sensitivity (r 2 = 0.51, including total body fat, VAT, and deep SAT in the model). The amount of superficial SAT was not associated with insulin sensitivity after adjustment for total body fat and either VAT or deep SAT. The association for truncal AT, measured by DEXA, and for total SAT was of intermediate strength, being weaker than VAT and deep SAT, yet stronger than superficial SAT, and it was similar to the association between FM and glucose Rd.

View this table:
Table 3.

Correlation coefficients between body composition and insulin sensitivity

The relation of adiposity to glucose AUC or insulin AUC was not as robust as that found for insulin sensitivity measured by the euglycemic clamp method, perhaps because all subjects had normal glucose tolerance. Nonetheless, the respective values for correlation coefficients were highly similar for VAT and deep SAT and quite different from the value of superficial SAT, which in turn followed a pattern quite similar to thigh AT. With respect to fasting insulin, the correlation coefficients for VAT and deep SAT were 0.57 and 0.58, respectively (both P < 0.001), whereas the coefficient for superficial SAT was 0.26 (P = 0.10), and again this was similar to that found for thigh AT (r = 0.27, NS).

Shown in Table 4 are the correlation coefficients between AT depots and plasma leptin, MAP, and plasma lipids. Plasma leptin was significantly correlated with all parameters of adiposity but was most strongly related to thigh AT and superficial SAT; the respective correlation coefficients were nearly identical. The correlation between leptin and deep SAT and VAT was also similar but of lesser strength than superficial SAT. With respect to MAP, VAT and deep SAT were the only depots to manifest significant correlation and were of highly similar value, whereas in contrast, the association with superficial AT was quite weak and again followed the pattern observed for thigh AT. As is also shown in Table4, these same patterns of similarity between VAT and deep SAT were found for associations with apoB, triglyceride, and HDL cholesterol, whereas the patterns for superficial SAT differed from those for deep SAT and instead followed the patterns found for thigh AT.

View this table:
Table 4.

Correlation coefficients between body composition and lipids, leptin, and blood pressure


The current study was undertaken to examine the novel hypothesis that superficial and deep depots of subcutaneous abdominal adiposity, defined anatomically by a fascial plane that divides the two depots and differing in histological characteristics (22), might also differ in regard to their association with insulin resistance. The findings clearly indicate that strong differences do exist. Superficial SAT manifests a powerful relation to plasma leptin but a weak association with insulin resistance, and in these and other respects, it follows a pattern observed for thigh subcutaneous adipose tissue, a depot generally regarded as a weak determinant of insulin resistance. In contrast, the deep subcutaneous adipose tissue of the abdomen manifests a robust relation to IR and other key aspects that define the insulin resistance syndrome (e.g., blood pressure, fasting insulin, and lipids); moreover, it does so in a pattern nearly identical to that observed for visceral adiposity. Therefore, from the perspective of understanding body composition and insulin resistance, these results indicate that it is not accurate to “lump” these two differing adipose tissue depots into a single category, but instead it may be useful to “split” the depots in accord with the anatomic demarcation of the fascial plane (18).

In the surgical literature of the past decade, more specifically that related to the procedure of liposuction, there have been a number of reports on the fascial plane within subcutaneous adipose tissue of the abdomen and at other sites (3, 15, 22, 28). This follows by many decades earlier reports on the presence of a fascial plane in subcutaneous abdominal adipose tissue that was recognized by dissection methods in humans and other species (20). Moreover, while the superficial adipose tissue has a well defined framework of vertically oriented closely spaced septa dividing the adipose tissue's relatively small lobules, the septal structure of the deeper adipose layer is less well defined, and the adipose tissue is more loosely organized (28). However, these distinctions have not been scrutinized in regard to their potential impact upon obesity-related IR, despite sustained interest during the past several decades in the role of truncal obesity.

Our interest in the potential metabolic significance of deep vs. superficial subcutaneous abdominal adipose tissue was stimulated by the report of Misra et al. (29). These investigators reported that subcutaneous fat located in the posterior half of the abdominal wall was more strongly related to IR than was subcutaneous abdominal fat located anteriorly. The explanation suggested for this curious observation was that posterior subcutaneous adipose tissue may largely represent the deep subcutaneous depot, although the investigators were not able in their study to measure the two depots directly. In the current study, the hypothesis that deep subcutaneous abdominal adipose tissue is associated with IR is confirmed; moreover, in also confirming that approximately three-fourths of the deep subcutaneous adipose tissue was found in the posterior half of the abdominal wall, our findings provide insight into the observations of Misra et al.

In the current study, visceral adipose tissue had a strong (negative) correlation with insulin resistance as measured by the euglycemic insulin infusion method. Moreover, visceral adipose tissue was associated with blood pressure, fasting insulin, and lipids in a pattern entirely consistent with numerous previous reports indicating the correlation of visceral adiposity with the insulin resistance syndrome (5, 8, 14, 36). The novel finding is that, in all these respects, deep subcutaneous abdominal adipose tissue mirrored those relationships found for visceral adiposity. Indeed, deep subcutaneous abdominal and visceral adipose tissue content was independently associated with insulin resistance. In recent years, despite the strong body of data linking insulin resistance to visceral adiposity, some debate has arisen regarding the relative importance of subcutaneous abdominal adipose tissue. Two studies by Abate and colleagues (1, 2), in which insulin sensitivity was determined by the “gold standard” glucose clamp method and abdominal adipose tissue volumes were carefully measured by MRI, yielded the provocative finding that subcutaneous abdominal adiposity was of greater importance than visceral adiposity. A recent study from our laboratory yielded similar results (17). These studies appeared to conflict with previous studies that had emphasized the importance of visceral adiposity (7, 8, 14,31). Based on the current findings of the starkly different associations found for deep vs. superficial subcutaneous adipose tissue, we believe that this controversy can now be substantially resolved.

Part of the difference in the data of Abate and colleagues (1, 2) compared with that of others (7, 9, 29) might be that the latter studies involved only women, whereas those of Abate and colleagues involved only men. Even when matched for body mass index, women generally have more adipose tissue than men do, and this gender difference is principally due to greater subcutaneous adipose tissue. These gender differences are reconfirmed in the present study with respect to superficial subcutaneous abdominal adipose tissue, as well as thigh subcutaneous adipose tissue. Among women, compared with men and regardless of obesity, superficial subcutaneous abdominal adipose tissue comprised a greater proportion of the overall amount of subcutaneous abdominal adipose tissue. In contrast, we did not observe a gender effect on the amount of deep subcutaneous abdominal adipose tissue; differences were related to obesity, and in both men and women this depot was strongly correlated to insulin resistance. Thus, if subcutaneous abdominal adipose tissue is considered only as a single depot, the adverse effect of the deep depot will be diluted by inclusion of the superficial depot, and this effect will be somewhat greater in women.

The anatomical differences between deep and superficial abdominal adipose tissue in association with insulin resistance may also have implications for the site-specific differences in adipocyte biology (30). For example, evidence from ex vivo studies on human adipocytes indicates that catecholamine-induced lipolysis is enhanced in visceral compared with subcutaneous abdominal tissue (19), primarily due to an increase in visceral adipocyte β3-adrenergic receptor function (19, 27). This results in greater fatty acid delivery to the portal circulation (27), which may adversely affect in vivo insulin action. Although these results cannot be interpreted with respect to deep vs. superficial subcutaneous adipose tissue, it is possible that analogous differences in metabolic activity exist between these depots. Indeed, studies in swine have found increased lipogenic enzyme levels in deep vs. superficial subcutaneous tissue (20), but these results have yet to be confirmed in humans. Future studies are required to elucidate physiological differences in deep vs. superficial abdominal adipocytes, which may provide insight into mechanisms of insulin resistance of obesity.

In summary, we have utilized computed tomography to clearly visualize the fascial plane of abdominal subcutaneous adipose tissue and to thereby measure superficial and deep depots in lean and obese men and women. The novel finding is that deep subcutaneous abdominal adipose tissue is strongly related to insulin resistance and in a manner nearly identical to that of visceral adiposity. In contrast, superficial subcutaneous abdominal adipose tissue had a weak association with insulin sensitivity but a strong association with plasma leptin and, overall, followed the pattern observed for thigh subcutaneous adipose tissue.


We would like to gratefully acknowledge the participation and cooperation of the research volunteers. We would also like to thank the research staffs at the General Clinical Research Center, the Obesity and Nutrition Research Center, and most particularly, those of the Department of Radiology. In addition, we would like to acknowledge the contributions of Sue Andreko, Nancy Mazzei, Amy Meirer, and Rena Wing.


  • Address for reprint requests and other correspondence: D. E. Kelley, Division of Endocrinology and Metabolism, East-1140 Biomedical Science Tower, University of Pittsburgh, Pittsburgh, PA 15261 (E-mail:kelley{at}msx.dept-med.pitt.edu).

  • This project was supported by funding from National Institutes of Health R01 DK-49200–04, 5M01 RR-00056 (General Clinical Research Center), and 1P30 DK-46204 (Obesity and Nutrition Research Center). National Institutes of Health National Research Service Award DK-07052–23 supported B. Goodpaster. During these studies, F. Troost was a visiting student at the University of Pittsburgh from the University of Maastricht, The Netherlands.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.


View Abstract