Abstract

Background:

Studies on the relationship between regional fat and insulin resistance yield mixed results. Our objective was to determine whether regional fat distribution, independent of obesity, is associated with insulin resistance.

Design:

Subjects included 115 healthy, overweight/moderately obese adults with body mass index (BMI) 25–36.9 kg/m2 who met predetermined criteria for being insulin resistant (IR) or insulin sensitive (IS) based on the modified insulin suppression test. Computerized tomography was used to quantify visceral adipose tissue (VAT), sc adipose tissue (SAT), and thigh adipose tissue. Fat mass in each depot was compared according to IR/IS group, adjusting for BMI and sex.

Results:

Despite nearly identical mean BMI in the IR vs. IS groups, VAT and %VAT were significantly higher in the IR group, whereas SAT, %SAT, and thigh sc fat were significantly lower. In logistic regression analysis, each sd increase in VAT increased the odds of being IR by 80%, whereas each increase in SAT decreased the odds by 48%; each increase in thigh fat decreased the odds by 59% and retained significance after adjusting for other depots. When grouped by VAT tertile, IS vs. IR individuals had significantly more SAT. There was no statistically significant interaction between sex and these relationships.

Conclusion:

These data demonstrate that after adjustment for BMI and VAT mass, sc abdominal and thigh fat are protective for insulin resistance, whereas VAT, after adjustment for SAT and BMI, has the opposite effect. Whether causal in nature or a marker of underlying pathology, these results clarify that regional distribution of fat-favoring sc depots is associated with lower risk for insulin resistance.

It was first shown more than 25 yr ago that whole-body insulin-mediated glucose uptake (IMGU) was decreased in obese individuals (1). It is now clear that IMGU can vary widely in overweight/obese individuals and that a substantial number of these individuals are insulin sensitive, without the metabolic abnormalities that characterize insulin resistance (2). Although the reason for this variability is unclear, some have attributed it to differences in regional fat distribution (3, 4). Epidemiological studies have shown that upper body fat and waist circumference predict the development of type 2 diabetes and cardiovascular disease (57). Visceral adipose tissue (VAT), in particular, has been implicated because it correlates with waist circumference and demonstrates greater catecholamine-stimulated lipolysis and inflammation (8, 9), compared with sc adipose tissue (SAT). Because VAT accounts for only 7–15% of total body fat, however, it seems unlikely that it could play a more important role than sc fat in the development of insulin resistance.

Studies using “gold standard” radiological techniques to quantify sc vs. visceral fat mass, along with quantitative measures of IMGU, are contradictory as to which fat depot is most closely related to insulin resistance (1015). These studies are small, and most have included individuals over a wide body mass index (BMI) range from 19–40 kg/m2 and above. Most show both SAT and VAT to be associated with insulin resistance. Three studies limited to obese individuals, however, showed VAT alone to predict insulin resistance (1113). Alternatively, one study showed that in multiple regression analysis, SAT alone independently predicted insulin resistance (14). Because BMI may confound the independent effect of regional fat distribution, studies including a small number of subjects over a wide BMI range may be limited in ability to ascertain relative contribution of regional fat depots.

The goal of the current study was to isolate the contribution of differences in regional fat distribution from differences in total adiposity by comparing two groups of individuals who were similarly obese, but quite disparate with regard to IMGU.

Subjects and Methods

Healthy, overweight or moderately obese subjects were recruited via newspaper advertisement in the cities surrounding Stanford University. Eligibility requirements included age 35–65 yr, BMI 26–36.9 kg/m2, stable weight for 3 months, no major organ disease, fasting plasma glucose concentration below 7.0 mmol/liter, and absence of lipid-lowering, diabetogenic, or weight loss medications. Subjects with a history of eating disorder, bariatric surgery, or liposuction were excluded. The study was approved by the Stanford University Human Subjects Committee, and all subjects gave written, informed consent.

IMGU was quantified by a modification (16) of the insulin suppression test as originally described and validated (17). Briefly, after an overnight fast, subjects were infused for 180 min with octreotide (0.27 μg/m2 · min), insulin (25 mU/m2 · min), and glucose (240 mg/m2 · min). In steady state (150–180 min), plasma insulin concentrations are similar for all individuals, and steady-state plasma glucose (SSPG) concentration provides a direct measure of IMGU in response to an infused glucose load; the higher the SSPG concentration, the more insulin resistant the individual. Based on a prior study of the distribution of SSPG concentrations in 490 healthy nondiabetic adults (18), we classified volunteers as being insulin resistant (IR) or insulin sensitive (IS) if their SSPG concentration was in the top or bottom 40th percentile of the SSPG distribution as previously described. Individuals not falling into either category were excluded.

Clinical measurements obtained in the General Clinical Research Center after overnight fasting included lipid/lipoprotein concentrations, height and weight in light clothing, waist circumference, race/ethnicity, and resting blood pressure (average of six).

The volume of SAT and VAT was quantified with computerized tomography scans (California Advanced Imaging, Atherton, CA) done at L4–5. Using a Siemens Sensation 4 CT Scanner, with scan variables set to 120 kV, effective mass of 165, and a 40-cm field of view, 10 slices (1-cm thick area) of the abdomen through L4–5 were obtained. Subcutaneous thigh fat was quantified midway between the greater trochanter and patella of the right leg. %VAT and %SAT were calculated as VAT/(VAT + SAT) and SAT/(VAT + SAT), respectively.

Statistical tests comparing IR vs. IS subgroups used analysis of covariance or general linear regression, adjusted for sex and BMI. χ2 test was used for categorical variables. Logistic regression analysis with IR group as dependent variable was conducted to formally test for sex-fat interaction in predicting IR group and to determine the independent contribution of VAT, SAT, and thigh fat with adjustment for sex and BMI in the group as a whole. Standardized odds ratios (Z-scores) are presented. Because regional fat mass differs by sex, box and whisker plots depicting regional fat mass according to IR group (Fig. 1) are shown separately by sex, whereas statistical tests were performed on sexes combined due to negative sex interactions in the relationships of interest. Serum triglycerides and %VAT were log-transformed for statistical analyses; other variables were normally distributed. P values <0.05 were considered statistically significant.

Fig. 1

Boxplots showing distribution of BMI, %VAT, %SAT, and sc thigh fat in females and males grouped according to insulin resistance. P values reflect comparison of IR vs. IS subjects (men and women combined), via general linear regression with adjustment for sex and BMI (top left panel not adjusted for BMI). Substitution of VAT and SAT (cm3) in place of %VAT and %SAT in same model yielded P = 0.005 and P = 0.027, respectively. *, Data points outside the 1.5 × interquartile range.

Fig. 1

Boxplots showing distribution of BMI, %VAT, %SAT, and sc thigh fat in females and males grouped according to insulin resistance. P values reflect comparison of IR vs. IS subjects (men and women combined), via general linear regression with adjustment for sex and BMI (top left panel not adjusted for BMI). Substitution of VAT and SAT (cm3) in place of %VAT and %SAT in same model yielded P = 0.005 and P = 0.027, respectively. *, Data points outside the 1.5 × interquartile range.

Results

By selection, SSPG concentrations were significantly higher in the IR vs. the IS subgroups (224 ± 35 vs. 86 ± 22 mg/dl; P < 0.001). Also by design, the IR and IS subgroups were not statistically significantly different in terms of BMI (30.7 ± 2.7 vs. 30.1 ± 2.6 kg/m2; P = 0.33). Although there were no significant differences in age, race, waist circumference, blood pressure, or low-density lipoprotein-cholesterol concentrations, the IR subgroup had significantly higher fasting plasma glucose (102 ± 11 vs. 94 ± 7 mg/dl) and triglyceride (141 ± 69 vs. 89 ± 44 mg/dl; P < 0.001) concentrations, and lower high-density lipoprotein-cholesterol (44 ± 13 vs. 56 ± 14 mg/dl) concentrations in comparison with the IS subgroup.

Figure 1 depicts a comparison of BMI and fat in different regions. Although regional fat distribution is shown separately for each sex, formal testing for a sex-fat interaction in predicting IR status was negative for all fat depots, indicating that statistical comparisons should be done with sexes combined. For sexes combined, BMI did not differ significantly. Despite this, %VAT was significantly greater in the IR vs. IS group (after adjustment for potential confounding by sex distribution and BMI), whereas the percentage of SAT and thigh fat showed the opposite relationship. Absolute VAT and SAT mass also showed a bidirectional relationship with IR group, with IR characterized by higher VAT (P = 0.005) and lower SAT (P = 0.027). As expected, VAT and %VAT were significantly greater in males (P < 0.001), whereas SAT and thigh fat were significantly higher in females (P < 0.001). Between-sex differences persisted after adjustment for BMI and insulin resistance.

The results of logistic regression analysis (Table 1) show the odds of being IR vs. IS for each standardized unit of SAT, VAT, %VAT, or thigh fat, as well as the independent contribution of male sex and BMI. Each SAT increment was associated with a 42% decreased risk of being IR (Model 1, P = 0.04), whereas each VAT increment was associated with an 80% increased risk of being IR (Model 2, P = 0.04). %VAT and %SAT increments were associated with 88% increased and decreased risk of IR, respectively (Model 3), and each thigh fat increment was associated with a 59% decreased risk of IR (Model 4). Inclusion of both VAT and SAT in the multivariate model demonstrated independent associations, in opposite directions, for both VAT [odds ratio (OR), 1.77; 95% confidence interval (CI), 1.04–3.02] and SAT (OR, 0.56; 95% CI, 0.34–0.94) with IR status, whereas BMI and male sex were not significant predictors (Model 5). Inclusion of thigh fat with SAT, VAT, or both showed a persistent independent protective effect of thigh fat (P = 0.01), with VAT and SAT losing statistical significance (P = 0.10 and 0.89). BMI remained a significant predictor of IR status only in models not containing VAT.

Table 1

Multiple logistic regression analysis of 115 males and females for regional fat predictors of insulin resistance (IR group)

Model Standardized OR 95% CI P valuea 
Model 1    
    SAT (Z-score) 0.58 0.35–0.98 0.04 
    Male sex 1.56 0.62–3.89 0.34 
    BMI (kg/m21.19 1.00–1.42 0.05 
Model 2    
    VAT (Z-score) 1.80 1.03–3.15 0.04 
    Male sex 1.39 0.54–3.59 0.49 
    BMI (kg/m21.01 0.87–1.17 0.89 
Model 3    
    %VAT (Z-score) 1.88 1.10–3.23 0.02 
    Male sex 1.15 0.42–3.15 0.78 
    BMI (kg/m21.07 0.93–1.22 0.36 
Model 4    
    Thigh (Z-score) 0.41 0.22–0.78 0.006b 
    Male sex 0.59 0.16–2.09 0.41 
    BMI (kg/m21.34 1.08–1.65 0.007 
Model 5    
    VAT (Z-score) 1.77 1.04–3.02 0.04 
    SAT (Z-score) 0.56 0.34–0.94 0.03 
    Male sex 0.89 0.31–2.52 0.83 
    BMI (kg/m21.14 0.95–1.36 0.16 
Model Standardized OR 95% CI P valuea 
Model 1    
    SAT (Z-score) 0.58 0.35–0.98 0.04 
    Male sex 1.56 0.62–3.89 0.34 
    BMI (kg/m21.19 1.00–1.42 0.05 
Model 2    
    VAT (Z-score) 1.80 1.03–3.15 0.04 
    Male sex 1.39 0.54–3.59 0.49 
    BMI (kg/m21.01 0.87–1.17 0.89 
Model 3    
    %VAT (Z-score) 1.88 1.10–3.23 0.02 
    Male sex 1.15 0.42–3.15 0.78 
    BMI (kg/m21.07 0.93–1.22 0.36 
Model 4    
    Thigh (Z-score) 0.41 0.22–0.78 0.006b 
    Male sex 0.59 0.16–2.09 0.41 
    BMI (kg/m21.34 1.08–1.65 0.007 
Model 5    
    VAT (Z-score) 1.77 1.04–3.02 0.04 
    SAT (Z-score) 0.56 0.34–0.94 0.03 
    Male sex 0.89 0.31–2.52 0.83 
    BMI (kg/m21.14 0.95–1.36 0.16 
a

Interaction term for sex-regional fat mass term did not alter results, was not statistically significant, and was thus removed from final models.

b

P < 0.05 after adjustment for VAT, SAT, male sex, and BMI.

Table 1

Multiple logistic regression analysis of 115 males and females for regional fat predictors of insulin resistance (IR group)

Model Standardized OR 95% CI P valuea 
Model 1    
    SAT (Z-score) 0.58 0.35–0.98 0.04 
    Male sex 1.56 0.62–3.89 0.34 
    BMI (kg/m21.19 1.00–1.42 0.05 
Model 2    
    VAT (Z-score) 1.80 1.03–3.15 0.04 
    Male sex 1.39 0.54–3.59 0.49 
    BMI (kg/m21.01 0.87–1.17 0.89 
Model 3    
    %VAT (Z-score) 1.88 1.10–3.23 0.02 
    Male sex 1.15 0.42–3.15 0.78 
    BMI (kg/m21.07 0.93–1.22 0.36 
Model 4    
    Thigh (Z-score) 0.41 0.22–0.78 0.006b 
    Male sex 0.59 0.16–2.09 0.41 
    BMI (kg/m21.34 1.08–1.65 0.007 
Model 5    
    VAT (Z-score) 1.77 1.04–3.02 0.04 
    SAT (Z-score) 0.56 0.34–0.94 0.03 
    Male sex 0.89 0.31–2.52 0.83 
    BMI (kg/m21.14 0.95–1.36 0.16 
Model Standardized OR 95% CI P valuea 
Model 1    
    SAT (Z-score) 0.58 0.35–0.98 0.04 
    Male sex 1.56 0.62–3.89 0.34 
    BMI (kg/m21.19 1.00–1.42 0.05 
Model 2    
    VAT (Z-score) 1.80 1.03–3.15 0.04 
    Male sex 1.39 0.54–3.59 0.49 
    BMI (kg/m21.01 0.87–1.17 0.89 
Model 3    
    %VAT (Z-score) 1.88 1.10–3.23 0.02 
    Male sex 1.15 0.42–3.15 0.78 
    BMI (kg/m21.07 0.93–1.22 0.36 
Model 4    
    Thigh (Z-score) 0.41 0.22–0.78 0.006b 
    Male sex 0.59 0.16–2.09 0.41 
    BMI (kg/m21.34 1.08–1.65 0.007 
Model 5    
    VAT (Z-score) 1.77 1.04–3.02 0.04 
    SAT (Z-score) 0.56 0.34–0.94 0.03 
    Male sex 0.89 0.31–2.52 0.83 
    BMI (kg/m21.14 0.95–1.36 0.16 
a

Interaction term for sex-regional fat mass term did not alter results, was not statistically significant, and was thus removed from final models.

b

P < 0.05 after adjustment for VAT, SAT, male sex, and BMI.

Finally, when grouped according to VAT tertile, IS individuals compared with IR individuals had greater SAT mass (mean ± sd): VAT tertile 1, 146 ± 30 vs. 130 ± 40 cm3, P = 0.003; tertile 2, 181 ± 45 vs. 139 ± 41 cm3, P = 0.003; and tertile 3, 176 ± 55 vs. 139 ± 41 cm3, P = 0.065, via analysis of covariance adjusting for BMI and sex.

Discussion

The results of this study show that among overweight/moderately obese men and women, after adjusting for BMI, greater absolute or relative VAT increases the risk for insulin resistance, whereas increased absolute or relative SAT decreases the risk for insulin resistance. These findings stand in contrast to most prior studies on abdominal fat distribution, which show that insulin resistance is positively correlated with both VAT and SAT. Although some studies have shown through multivariate analysis that VAT (1113) or SAT (15) is a stronger predictor, none have previously shown that SAT is inversely associated with insulin resistance. This discordance is likely due to prior studies including a wide BMI range, from lean to very obese, whereas we included only overweight/moderately obese individuals. The resulting effects are likely 2-fold: 1) VAT and SAT measurements from the wider BMI range reflect to some degree the effect of obesity, per se, on insulin resistance, whereas a narrower BMI range, with additional statistical adjustment for BMI, highlights the effect of fat distribution independent of fat mass; and 2) it is possible that SAT is protective in obese individuals but associated with increased or neutral risk in normal-weight individuals, thus minimizing the ability to detect the protective effect in a group including normal-weight individuals.

Three small studies limited to overweight/obese individuals hint at a protective effect of SAT. These studies showed correlation coefficients between SAT and IMGU of 0.17, 0.06, and 0.02, in contrast to correlation coefficients between VAT and IMGU of −0.40, −0.40, and −0.42 (1113). That increased SAT may be beneficial for a given degree of obesity was also highlighted in an analysis of men in the Framingham study (19), which showed that among individuals with VAT in the highest tertile, those with SAT also in the highest tertile had a significantly lower risk of metabolic syndrome. This supports our results, which showed that when grouped according to VAT tertile, IS, compared with IR individuals, had significantly more SAT.

Perhaps even more dramatic in the current study, is the protective effect of thigh fat: in models containing all regional depots, thigh fat remained the only significant predictor of IR status. These findings extend those reported in several studies showing that increased thigh fat or leg/trunk ratio were independently associated with surrogate estimates of insulin sensitivity (20, 21), although the associations applied only to women with BMI greater than 27 kg/m2. The only prior study using a precise physiological measurement of insulin sensitivity (14) demonstrated the contrary in 54 subjects with wide-ranging BMI (19–41 kg/m2); thigh fat was positively associated with insulin resistance, although not statistically significant after adjustment for other depots. Thus, the current findings are the first using a quantitative measurement of IMGU to show that thigh sc fat is independently protective for insulin resistance.

In summary, the results of this study confirm prior suggestions that VAT is associated with insulin resistance. Furthermore, they demonstrate that SAT is associated with decreased risk for insulin resistance, independent of VAT and BMI, and that thigh sc fat is protective independent of all regional depots and BMI. These observations should not be misconstrued as demonstrating causality. Careful studies will be required to ascertain whether these patterns of regional fat distribution are the cause or consequence of insulin resistance, or perhaps a marker of a related process that determines both regional fat distribution and insulin resistance.

Abbreviations:

  • BMI

    Body mass index

  • IMGU

    insulin-mediated glucose uptake

  • IR

    insulin resistant

  • IS

    insulin sensitive

  • SAT

    sc adipose tissue

  • VAT

    visceral adipose tissue.

Acknowledgments

The authors thank Dr. Philip Lavori (Stanford University) for his statistical advice.

Funding for this study was provided by National Institutes of Health (NIH)/National Institute of Diabetes and Digestive and Kidney Diseases Grants 1 R01 DK071309-01, 5RO1DK071333, and 5K23 RR16071; by NIH Grant RR 000070; and by the NIDDK Intramural Research Program.

T.M., F.A., A.L., and C.L. recruited and performed all metabolic studies; A.L. oversaw Health Insurance Portability and Accountability Act compliance and Institutional Review Board requirements; T.M. collected and performed statistical analyses on the data; and T.M. prepared the manuscript, which was reviewed by all authors.

Clinical Trials Identifiers: NCT00285844, NCT01336777

Disclosure Summary: The authors have nothing to declare.

References

1.
Olefsky
J
,
Reaven
GM
,
Farquhar
JW
1974
Effects of weight reduction on obesity: studies of carbohydrate and lipid metabolism.
J Clin Invest
 
53
:
64
76
2.
McLaughlin
T
,
Abbasi
F
,
Lamendola
C
,
Reaven
G
2007
Heterogeneity in the prevalence of risk factors for cardiovascular disease and type 2 diabetes in obese individuals: effect of differences in insulin sensitivity.
Arch Intern Med
 
167
:
642
648
3.
Pouliot
MC
,
Després
JP
,
Nadeau
A
,
Moorjani
S
,
Prud'Homme
D
,
Lupien
PJ
,
Tremblay
A
,
Bouchard
C
1992
Visceral obesity in men. Association with glucose tolerance, plasma insulin, and lipoprotein levels.
Diabetes
 
41
:
826
834
4.
Mathieu
P
,
Poirier
P
,
Pibarot
P
,
Lemieux
I
,
Després
JP
2009
Visceral obesity. The link among inflammation, hypertension, and cardiovascular disease.
Hypertension
 
53
:
577
584
5.
Vague
J
1956
The degree of masculine differentiation of obesities: a factor determining predisposition to diabetes, atherosclerosis, gout, and uric calculous disease.
Am J Clin Nutr
 
4
:
20
34
6.
Ohlson
LO
,
Larsson
B
,
Svärdsudd
K
,
Welin
L
,
Eriksson
H
,
Wilhelmsen
L
,
Björntorp
P
,
Tibblin
G
1985
The influence of body fat distribution on the incidence of diabetes mellitus. 13.5 years of follow-up of the participants in the study of men born in 1913.
Diabetes
 
34
:
1055
1058
7.
Lapidus
L
,
Bengtsson
C
,
Larsson
B
,
Pennert
K
,
Rybo
E
,
Sjöström
L
1984
Distribution of adipose tissue and risk of cardiovascular disease and death: a 12 year follow up of participants in the population study of women in Gothenburg, Sweden.
Br Med J Clin Res Ed
 
289
:
1257
1261
8.
Liu
A
,
McLaughlin
T
,
Liu
T
,
Sherman
A
,
Yee
G
,
Abbasi
F
,
Lamendola
C
,
Morton
J
,
Cushman
SW
,
Reaven
GM
,
Tsao
PS
2009
Differential intra-abdominal adipose tissue profiling in obese, insulin-resistant women.
Obes Surg
 
19
:
1564
1573
9.
Bolinder
J
,
Kager
L
,
Ostman
J
,
Arner
P
1983
Differences at the receptor and postreceptor levels between human omental and subcutaneous adipose tissue in the action of insulin on lipolysis.
Diabetes
 
32
:
117
123
10.
Garg
A
2004
Regional adiposity and insulin resistance.
J Clin Endocrinol Metab
 
89
:
4206
4210
11.
Brochu
M
,
Starling
RD
,
Tchernof
A
,
Matthews
DE
,
Garcia-Rubi
E
,
Poehlman
ET
2000
Visceral adipose tissue is an independent correlate of glucose disposal in older obese postmenopausal women.
J Clin Endocrinol Metab
 
85
:
2378
2384
12.
Ross
R
,
Aru
J
,
Freeman
J
,
Hudson
R
,
Janssen
I
2002
Abdominal adiposity and insulin resistance in obese men.
Am J Physiol Endocrinol Metab
 
282
:
E657
E663
13.
Ross
R
,
Freeman
J
,
Hudson
R
,
Janssen
I
2002
Abdominal obesity, muscle composition, and insulin resistance in premenopausal women.
J Clin Endocrinol Metab
 
87
:
5044
5051
14.
Goodpaster
BH
,
Thaete
FL
,
Simoneau
JA
,
Kelley
DE
1997
Subcutaneous abdominal fat and thigh muscle composition predict insulin sensitivity independently of visceral fat.
Diabetes
 
46
:
1579
1585
15.
Abate
N
,
Garg
A
,
Peshock
RM
,
Stray-Gundersen
J
,
Grundy
SM
1995
Relationships of generalized and regional adiposity to insulin sensitivity in men.
J Clin Invest
 
96
:
88
98
16.
Pei
D
,
Jones
CN
,
Bhargava
R
,
Chen
YD
,
Reaven
GM
1994
Evaluation of octreotide to assess insulin-mediated glucose disposal by the insulin suppression test.
Diabetologia
 
37
:
843
845
17.
Greenfield
MS
,
Doberne
L
,
Kraemer
F
,
Tobey
T
,
Reaven
G
1981
Assessment of insulin resistance with the insulin suppression test and the euglycemic clamp.
Diabetes
 
30
:
387
392
18.
Yeni-Komshian
H
,
Carantoni
M
,
Abbasi
F
,
Reaven
GM
2000
Relationship between several surrogate estimates of insulin resistance and quantification of insulin-mediated glucose disposal in 490 healthy, nondiabetic volunteers.
Diabetes Care
 
23
:
171
175
19.
Porter
SA
,
Massaro
JM
,
Hoffmann
U
,
Vasan
RS
,
O'Donnel
CJ
,
Fox
CS
2009
Abdominal subcutaneous adipose tissue: a protective fat depot?
Diabetes Care
 
32
:
1068
1075
20.
Hunter
GR
,
Chandler-Laney
PC
,
Brock
DW
,
Lara-Castro
C
,
Fernandez
JR
,
Gower
BA
2010
Fat distribution, aerobic fitness, blood lipids, and insulin sensitivity in African-American and European-American women.
Obesity
 
18
:
274
281
21.
Aasen
G
,
Fagertun
H
,
Halse
J
2008
Regional fat mass by DXA: high leg fat mass attenuates the relative risk of insulin resistance and dyslipidaemia in obese but not in overweight postmenopausal women.
Scand J Clin Lab Invest
 
68
:
204
211