Subcutaneous Thigh Fat Area Is Unrelated to Risk of Type 2 Diabetes in a Prospective Study of Japanese Americans

Abstract

Aims/Hypothesis

Cross-sectional research has reported a negative association between subcutaneous thigh fat (STF) and type 2 diabetes prevalence but no prospective research on this association exists using direct measurements of STF obtained from imaging studies while adjusting for other fat depots. We studied the independent associations of intra-abdominal (IAF), subcutaneous abdominal (SAF), and STF with future risk of diabetes.

Methods

We prospectively followed 489 non-diabetic Japanese Americans (BMI 25.0-29.9 kg/m² 32.7%, ≥30.0 kg/m² 5.4%) over 10 years for the development of diabetes defined by use of hypoglycemic medication or a fasting ≥7.0 mmol/l or 2-hr ≥11.1 mmol/l plasma glucose during an OGTT. STF, SAF, and IAF area were measure by CT scan and mid-thigh circumference (TC) by tape measure at baseline.

Results

Over 10 years, 103 persons developed diabetes. STF area was not independently associated with odds of developing diabetes in univariate or a multiple logistic regression model (odds ratio, 1 SD increase 0.8, 95% CI 0.5, 1.2) adjusted for age, gender, BMI, IAF, and SAF. The only fat depot associated with diabetes odds in this model was IAF. TC was borderline significantly associated with a lower odds of developing diabetes (0.7, 0.5, 1.0, p=0.052)

Conclusions

Similar to other research, TC was negatively associated with diabetes risk, while STF was not, arguing that the negative association between TC and diabetes observed in other research is not due to thigh subcutaneous fat mass. IAF area emerged as the only measured fat depot that was independently associated with type 2 diabetes risk.

Cross-sectional research has shown a negative association between subcutaneous thigh fat and thigh circumference and glucose concentrations, dyslipidemia, and type 2 diabetes [1-3]. Investigators have shown that greater leg fat mass may be associated with lower prevalence of diabetes mellitus in obese participants [4, 5]. A potential causal role for thigh fat in the development of type 2 diabetes has also been suggested by a prospective population based study that demonstrated that the incidence of this outcome was negatively associated with thigh circumference and positively associated with waist circumference [6]. It is well recognized that pathological conditions that are accompanied by loss of gluteo-femoral fat such as glucocorticoid excess or partial lipodystrophy are associated with metabolic abnormalities [7]. Although these findings support a potential protective effect of thigh fat in the development of type 2 diabetes, other results do not support this association. A prospective investigation in Pima Indians showed that thigh circumference was less informative than waist circumference or generalized adiposity measures in predicting the risk of type 2 diabetes [8], and no significant association was seen between subcutaneous thigh fat depot size measured by computed tomography and type 2 diabetes in one cross-sectional study [9]. No prospective investigation exists to our knowledge on the association between directly measured thigh fat and risk of type 2 diabetes in normal, overweight, or obese individuals.

Other fat depots have also been associated with type 2 diabetes risk. A positive association between intra-abdominal fat and type 2 diabetes prevalence and incidence has been show in cross-sectional and prospective studies [9-12]. A less consistent association has been observed between subcutaneous abdominal fat and presence of metabolic risk factors or type 2 diabetes [9-12]. These data argue for differential associations between body fat and diabetes depending on fat location.

In this study we aimed to determine longitudinally the relationship between subcutaneous thigh fat, measured by computed tomography, and the development of type 2 diabetes while adjusting for potential confounding by other significant fat depots.

Research Design and Methods

Study participants

The study population comprised 519 non-diabetic Japanese-American men and women enrolled in the Japanese-American Community Diabetes Study. All individuals were of 100% Japanese ancestry. The details on the selection criteria and recruitment of individuals for this study have been previously published [13]. Briefly, participants were chosen as volunteers from a community-wide mailing list and telephone directory that included nearly 95% of the Japanese-American population in King County, Washington State. This population was followed prospectively over 10 years. Individual diabetes status was assessed at baseline, 2.5, 5, and 10-year follow-up visits. Thirty participants did not attend any of the follow-up visits due to death, inability to locate them or withdrawal from the study, leaving a total of 489 participants who were included in this analysis. This study received approval from the University of Washington Human Subjects Office and all participants provided written informed consent to participate.

Measurements

Evaluations were done at the General Clinical Research Center at the University of Washington. All research protocols used were reviewed and approved by the institution’s Human Subjects Review Committee and informed consent was obtained from all participants. BMI was computed as weight in kilograms divided by height in meters squared (kg/m²). Waist circumference was measured at the umbilicus in cm using a tape measure.

Normal glucose tolerance (NGT), impaired glucose tolerance (IGT), or type 2 diabetes classification was determined using a 75-g oral glucose tolerance test (OGTT) following a 10-hour fast. Diabetes status was defined as use of oral hypoglycemic medication or insulin, or fasting plasma glucose ≥126 mg/dl or 2-hr glucose ≥200 mg/dl. Participants were considered to have developed diabetes during follow up if this outcome occurred at the 2.5, 5, or 10 year follow-up time points. Plasma glucose was assayed by an automated glucose oxidase method. Family history of diabetes was considered positive if any first-degree relative had diabetes [13].

Single (1cm) CT scan slices were obtained at the abdomen (umbilicus level) and thigh (halfway between the greater trochanter and the superior margin of the patella) [14]. Cross-sectional measurements of fat area (cm²) were obtained from these on each participant as follows: subcutaneous abdominal fat, intra-abdominal fat (within the confines of the transversalis fascia), and left thigh subcutaneous fat [14]. The intra-observer variability for multiple measurements by a single observer of a single CT scan ranged from 0.2% to 1.4%. The point at which the mid-thigh CT was obtained was marked and the thigh circumference was assessed at this point using a tape measure.

Statistical analysis

Multiple logistic regression models were created to estimate the relationship between fat depots and diabetes occurrence while adjusting for other variables considered potential confounders. The variance inflation factor was used to identify a high likelihood of collinearity in multivariable models if it exceeded 4. Odds ratios for all continuous variables were calculated for a 1 standard deviation increase. The presence of interactions between gender and fat depots and gender and BMI were tested by insertion of first order interaction terms into the regression models. IBM SPSS Statistics 18 (SPSS) was used for statistical analyses.

Results

Over the follow-up period, 103 (19.8%) of the 489 participants developed diabetes. At baseline, study participants had a mean BMI of 24.1 kg/cm², a mean age of 52.2 years, 51.5% were male and 37.6% of individuals had a family history of diabetes (Table 1). Although the average BMI in this population was in the category usually defined as normal (BMI < 25.0 kg/m²), BMI ranged from 16.6 to 36.9 kg/m², with 61.9% of persons having a BMI < 25.0 kg/m², 32.7% between 25.0 and 29.9 kg/m², and 5.4% at 30.0 kg/m² or greater. Ranges of thigh measurements were as follows: mid-thigh circumference 36.2 to 63 cm; subcutaneous fat area 1.8 to 215.5 cm². Baseline characteristics by the development of diabetes during follow-up are presented in Table 1. All measures of overall and regional adiposity and circumference measures were positively and significantly correlated with each other except for a nonsignificant correlation between intra-abdominal fat and subcutaneous thigh fat (Table 2).

Table 1.

Characteristics of study participants at baseline by diabetes development during follow-up.

Characteristic	Study population	Did not progress to diabetes (N=386)	Progressed to diabetes (N=103)	p-value
Age (years)	52.2 (12.0)	51.2 (11.8)	57.9 (10.9)	<0.001
Male (%)	265 (51.5)	197 (51.0)	56 (54.4)	0.533
Fasting glucose (mmol/l)	5.2 (0.6)	5.1 (0.5)	5.6 (0.5)	<0.001
Positive family history of diabetes (%)	195 (37.6)	123 (31.9)	57 (55.3)	<0.001
BMI (kg/m2)	24.1 (3.2)	23.8 (3.1)	25.3 (3.6)	<0.001
Waist Circumference (cm)	86.0 (8.7)	85.1 (8.7)	89.3 (8.0)	<0.001
Subcutaneous Thigh fat (cm2)	65.0 (32.0)	65.2 (32.0)	63.0(30.5)	0.522
Intra-abdominal fat (cm2)	81.4 (49.6)	75.1 (46.1)	109.5 (51.9)	<0.001
Subcutaneous abdominal fat (cm2)	157.3 (76.4)	153.2 (78.1)	173.2 (70.2)	0.022
Mid-thigh circumference (cm)	49.3 (4.6)	49.3 (4.2)	49.0 (5.7)	0.382

Data are means (SD) or n (%).

Table 2.

Correlation matrix of measures of overall and regional adiposity displaying Pearson correlation coefficients. All associations were significant at p<0.001 except as noted.

Characteristic	BMI	Waist Circ.	Mid- Thigh Circ.	Subcutaneous Thigh Fat Area	Intra- Abdominal Fat Area	Subcutaneous Abdominal Fat Area
BMI	1
Waist Circ.	0.84	1
Mid-Thigh	0.67	0.45	1
Circ.
Subcutaneous	0.17	0.15	0.39	1
Thigh Fat
Area
Intra-	0.63	0.72	0.16	−0.07*	1
Abdominal
Fat Area
Subcutaneous	0.63	0.74	0.36	0.53	0.42	1
Abdominal
Fat Area

^*p = 0.118

Logistic regression analysis was performed to estimate the bivariate association between development of diabetes over the 10-year follow-up and each of the independent variables. Age, BMI, waist circumference, intra-abdominal fat area, subcutaneous abdominal fat area, fasting glucose, and family history of diabetes each showed a significant positive association with the development of diabetes in these analyses while subcutaneous thigh fat area and thigh circumference showed an insignificant negative association. Gender was not significantly associated with the development of diabetes (Table 3). Associations between thigh measurements and intra-abdominal fat and diabetes risk were estimated separately by gender. This analysis found a significant positive association between greater intra-abdominal fat area (1 SD increment) and diabetes risk [OR, 95% CI – men 2.1 (1.5, 2.8), women 2.0 (1.4, 2.7)]. The associations between thigh subcutaneous fat area and circumference and diabetes risk were nonsignificant for both men and women [thigh subcutaneous fat OR, 95% CI – men 1.2 (0.9, 1.6), women 0.8 (0.6, 1.2); thigh circumference – men 1.0 (0.8, 1.4), women 0.8 (0.5, 1.1).

Table 3.

Bivariate associations with incidence of diabetes over 10 years among Japanese Americans

Characteristic	Odds Ratio (95% Confidence Interval)	P-value
Age (years)	1.9 (1.5, 2.4)	<0.001
Male (%)	1.1 (0.7, 1.8)	0.548
BMI (kg/cm2)	1.5 (1.2, 1.9)	<0.001
Waist Circumference (cm)	1.6 (1.3, 2.0)	<0.001
Subcutaneous Thigh fat (cm2)	0.9 (0.7, 1.2)	0.532
Intra-abdominal fat (cm2)	2.0 (1.6, 2.5)	<0.001
Subcutaneous abdominal fat (cm2)	1.3 (1.0, 1.6)	0.022
Fasting glucose (mg/dl)	2.9 (2.2, 3.8)	<0.001
Family history of diabetes (%)	2.7 (1.7, 4.1)	<0.001
Mid-thigh circumference (cm)	0.9 (0.7, 1.1)	0.331

Odds ratios for continuous variables reflect a 1 SD magnitude increase.

Multiple logistic regression models were used to estimate the relationship between thigh fat and the development of type 2 diabetes while adjusting for other independent variables. All models adjusted for age, gender, and family history of diabetes. Subcutaneous thigh fat area was not independently associated with the odds of diabetes over 10 years in a model adjusted for age, gender, and family history of diabetes (Table 4, Model 1). After adjustment for BMI a negative but statistically nonsignificant association between subcutaneous thigh fat area and diabetes odds was observed (Table 4, Model 2). The magnitude of the association between subcutaneous thigh fat area and diabetes odds though diminished with the inclusion of intra-abdominal fat and abdominal subcutaneous fat areas in the model (Table 4, Model 3). No association was observed between subcutaneous thigh fat and type 2 diabetes odds in any of these models. There was no evidence of collinearity in the multivariable models shown in Table 4 as judged by the variance inflation factor.

Table 4.

Adjusted models of the incidence of diabetes over 10 years in relation to body composition among Japanese Americans.

Model	Variable	Odds Ratio (95% Confidence Interval)	P-value
1.	Age	1.9(1.5, 2.7)	<0.001
	Family History	2.5 (1.6, 4.0)	<0.001
	Male	1.4 (0.8, 2.6)	0.241
	Subcutaneous Thigh fat	1.1 (0.8, 1.5)	0.551
2.	Age	1.7 (1.2, 2.3)	<0.001
	Family History	2.3 (1.4, 3.7)	<0.001
	Male	0.6 (0.3, 1.2)	0.150
	BMI	1.8(1.4, 2.4)	<0.001
	Subcutaneous Thigh Fat	0.7 (0.5, 1.1)	0.108
3.	Age	1.6 (1.2, 2.2)	0.001
	Family History	2.3 (1.4, 3.6)	0.001
	Male	0.5 (0.2, 1.1)	0.087
	BMI	1.6(1.0, 2.6)	0.038
	Subcutaneous Thigh Fat	0.8 (0.5, 1.2)	0.279
	Intra-abdominal Fat	1.5 (1.1, 2.1)	0.033
	Subcutaneous Abdominal Fat	0.8 (0.5, 1.2)	0.286
4.	Age	2.1 (1.2, 3.7)	0.007
	Family History	2.2 (1.3, 3.4)	0.001
	Male	1.1 (0.7, 1.7)	0.831
	BMI	2.1 (1.5, 3.0)	<0.001
	Mid-thigh circumference	0.7 (0.5, 1.0)	0.052

Odds ratios for continuous variables reflect a 1 SD magnitude increase.

Intra-abdominal fat area was independently associated with the development of diabetes while subcutaneous abdominal fat area was not in a model that contained both of these measurements (Table 4, Model 3). Mid-thigh circumference was negatively associated with diabetes incidence (Table 4, Model 4) with higher circumference associated with lower odds of diabetes development, but this association was of borderline statistical significance (p=0.052). Insertion of intra-abdominal fat as an additional covariate into Table 4, Model 4 resulted in a diminution in the association between mid-thigh circumference and diabetes development (OR, 95% CI 0.75 (0.5, 1.1), p=0.183). Insertion of waist circumference into the regression models shown in Table 4 in place of BMI did not result in a change in the statistically significant association between diabetes development and intra-abdominal fat area, or the nonsignificant associations between this outcome and both thigh subcutaneous fat area and mid-thigh circumference (data not shown).

To assess whether the results of the models in Table 4 differed by gender, first order interactions between gender and fat depots and gender and BMI were tested in all models in Table 4. No significant interactions were observed (p-values for interaction terms > 0.05). In addition, odds ratios for the association between thigh subcutaneous fat and diabetes development were estimated by re-running Model 4, Table 4 for men and women separately. The results were similar for each gender [OR, 95% CI, men – 1.0 (0.7, 1.5), women 0.8 (0.5, 1.4)]. We also included a BMI × thigh subcutaneous fat interaction term in Models 2 & 3 in Table 4 to determine whether the association between thigh fat and diabetes risk might differ across the spectrum of overall adiposity. In neither model was the interaction term significant (Model 2, p=0.642, Model 3, p=0.414).

Discussion

In this study we failed to find a significant association between subcutaneous thigh fat area and cumulative incidence of type 2 diabetes. This finding was seen consistently in unadjusted analyses as well and multivariable models adjusted for age, gender, family history of diabetes, and overall and regional measures of adiposity. In adjusted models, thigh circumference was borderline significantly associated with diabetes incidence in the same direction as seen in other research. Greater thigh circumference was associated with a lower odds of incident diabetes in some models, suggesting that fat in this depot may exert a protective effect. Research to resolve the discordance between our findings and existing literature on thigh circumference and diabetes risk is needed. As a negative association between subcutaneous thigh fat area and metabolic syndrome prevalence among obese individuals has been reported in a cross-sectional study, we examined whether the association between this fat depot and diabetes risk varied by degree of overall adiposity as measured by BMI in our population [15]. We did not find this to be true, as our results support a negative association between diabetes risk and thigh subcutaneous fat area across the BMI spectrum observed among our participants.

Intra-abdominal fat area, on the other hand, was strongly associated with the risk of this outcome. Therefore, this study confirms the important role of visceral adiposity in the development of type 2 diabetes and provides evidence that subcutaneous thigh fat is not independently associated with type 2 diabetes risk. The higher risk associated with greater intra-abdominal fat area was independent of BMI and subcutaneous thigh and abdominal fat areas. In this paper we provide additional follow-up to our previous analysis of abdominal adipose areas in which most participants had only completed 5 years of follow-up [10]. The results from the previous analysis of these data showed a similar association between abdominal fat depots and type 2 diabetes risk but did not examine the role of thigh fat area.

Cross-sectional studies have reported an association between thigh circumference, fat area and type 2 diabetes [1, 2]. However, no report to our knowledge has prospectively examined the relationship between subcutaneous thigh fat measured by computed tomography and type 2 diabetes incidence while adjusting for direct measurements of other significant fat depots. Results from cross-sectional studies have shown inconsistent results in the prediction of diabetes frequency using varied methods of estimating thigh fat [2, 9]. One study reported greater lower body subcutaneous fat, measured by thigh skinfold thickness, to be associated with a lower likelihood of the presence of type 2 diabetes and insulin resistance, but no adjustment was performed for other regional body fat depots [2]. Using computed tomography, another cross-sectional study reported that individuals with type 2 diabetes and impaired glucose tolerance had similar amounts of subcutaneous thigh fat compared to those with normal glucose tolerance [9]. A major deficiency of such cross-sectional research is the inability to ascertain whether higher or lower measures of thigh fat preceded or followed the development of diabetes, insulin resistance, or other states of dysglycemia. Uncertainty of this temporal sequence prohibits definite conclusions to be made from cross-sectional research about whether the size of a fat depot may be causally related to diabetes or other outcomes.

A population-based prospective cohort study examined the relationship between thigh circumference and type 2 diabetes risk [6]. Logistic regression models adjusted for BMI, age, and waist circumference demonstrated an inverse relationship between type 2 diabetes risk and baseline thigh circumference in women only. This study had several limitations including lack of adjustment for regional body fat depots and measurement of thigh circumference as a proxy for thigh fat, which is prone to inter-observer variability, and furthermore cannot distinguish between muscle and fat components that contribute to overall circumference [14].

There are limitations to this prospective study. Thirty participants were lost to follow-up due to death, inability to locate or withdrawal from the study; however, we believe our follow-up rate of over 80% of participants over a period of 10 years makes our observations unlikely to be biased due to these missing data points. While we adjusted for multiple covariates, the possibility exists that confounding might still be present due to unmeasured covariates. The cohort we studied was restricted to Japanese Americans and therefore it is possible that our observations may not apply to other ethnic groups. While it may be that our failure to observe an association between thigh fat area and diabetes incidence could be due to insufficient power, we believe that our observations of the previously noted association between thigh circumference and diabetes incidence and between intra-abdominal fat area and risk of this outcome suggest that this is unlikely to be the case. CT scans of the thigh in this study focused on subcutaneous fat area only and did not define inter-muscular fat content, which has been shown to be related to metabolic abnormalities in post-menopausal women, although this depot is of much lesser importance than visceral fat [16]. We were also unable to examine thigh muscle area by CT scan as a possible explanation for the inverse association others have noted between thigh circumference and diabetes risk, as this measurement was not performed from the original CT images, which are no longer available for additional study. Higher muscle mass has been reported in association with lower glucose and insulin concentrations, and may help to explain the correlation between greater thigh circumference and lower diabetes risk [17]. To our knowledge no prospective investigation of the risk of type 2 diabetes in relation to thigh muscle mass or inter-muscular fat has been conducted. In addition, our population was on average of normal weight with a mean BMI of 24.1 kg/m². Whether these findings would hold for heavier populations is not known.

We did not directly measure body fat but instead thigh, subcutaneous abdominal, and intra-abdominal fat measurements were each determined by a single CT slice. A high correlation between a single CT-scan slice and direct measurement of visceral fat volume has been shown which limits the potential for bias [18, 19]. No studies to our knowledge have examined the correlation between a single CT-scan slice and direct measurement of thigh fat volume. It has been shown using dual-energy X-ray absorptiometry measures of thigh fat that these correlate well with adipose tissue mass measured by multi slice-CT techniques [20]. However, the correlation between a single CT-scan slice and subcutaneous thigh fat volume remains unexplored. Adjustment for total adiposity was performed using BMI. This represents an indirect measure of adiposity and may not have resulted in complete adjustment for body fat mass.

In conclusion, we found no evidence that directly measured subcutaneous thigh fat was related to diabetes incidence despite our finding of an inverse borderline significant association between thigh circumference and diabetes risk in our population. The concept of the ‘metabolically healthy’ thigh fat depot is not supported by our data. We have again demonstrated the importance of intra-abdominal fat area as the most important fat depot predictor of the development of type 2 diabetes. These observations provide evidence that intra-abdominal fat remains the key fat depot related to type 2 diabetes risk among Japanese Americans, while thigh fat area does not appear to have a role in the development of this outcome.

Abbreviations

95% CI

95% Confidence Interval

Circ.

Circumference

IAF

Intra-abdominal fat area

CT

Computed tomography

OR

Odds Ratio

OGTT

Oral glucose tolerance test

SAF

Subcutaneous fat area

SD

Standard deviation

STF

Subcutaneous thigh fat area