Impact of Intra-Abdominal Fat and Age on Insulin Sensitivity and β-Cell Function

The study was a cross-sectional analysis of baseline data from 232 subjects (98 men and 134 women) who were recruited by advertisement to participate in a study on the effects of insulin sensitivity on the plasma lipid profile after egg consumption (12). A total of 12 subjects were excluded from the analysis because they met criteria for impaired fasting glucose (fasting glucose >6.1 mmol/l); thus, a total of 220 subjects (89 men and 131 women) form the basis of this report. All subjects were healthy and did not have a history of diabetes, dyslipidemia, or uncontrolled hypertension. The ethnic distribution of the study population consisted of 92.7% Caucasian, 3.2% African-American, 2.8% Asian-American, 0.5% Hispanic, and 0.9% Native-American subjects. All subjects signed informed consent to participate in the study, which was reviewed and approved by the human subjects review committee at the University of Washington.

Height and weight were measured and used to calculate BMI as weight/height² (kg/m²). Waist circumference was measured at the “natural waistline,” the smallest circumference. For those with no natural waistline, the measurement was made just above the iliac crest. Abdominal subcutaneous and intra-abdominal fat areas were quantified by a single-slice computed tomography scan at the level of the umbilicus (13). The variability of these measures made by a single observer is 1.5%, and the day-to-day variability is <1% (14).

A tolbutamide-modified frequently sampled intravenous glucose tolerance test was performed as previously described (15) to quantify insulin sensitivity, the acute insulin response to glucose (AIRg), and intravenous glucose tolerance. Briefly, after an overnight fast, three basal samples were drawn for insulin and glucose at 5-min intervals before glucose administration. Glucose (11.4 g/m² body surface area) was injected intravenously over 60 s beginning at time 0. Then, tolbutamide (125 mg/m² body surface area) was injected intravenously over 30 s, commencing 20 min after starting the glucose injection. Blood samples were drawn at 2, 3, 4, 5, 6, 8, 10, 12, 14, 16, 19, 22, 23, 24, 25, 27, 30, 35, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160, 180, 200, 220, and 240 min after glucose administration. Blood samples were separated and the plasma stored at −70°C before being assayed for glucose and immunoreactive insulin.

Plasma glucose concentrations were measured using the glucose oxidase method. Plasma immunoreactive insulin levels were measured using a modification of a double-antibody radioimmunoassay (16).

Fasting glucose and immunoreactive insulin values were calculated as the average of the three basal samples.

Using Bergman’s minimal model of glucose kinetics, parameters of glucose metabolism were quantified using the glucose and insulin data obtained from the frequently sampled intravenous glucose tolerance test (17). The insulin sensitivity index (S_i), which provides a measure of insulin’s ability to enhance glucose disposal, was calculated from the model. Glucose effectiveness at 0 insulin, which is a measure of the ability of glucose to promote its own disposal, was determined as S_g − (S_i × fasting insulin) (18). The administration of tolbutamide helps to improve parameter identifiability when the plasma glucose and insulin data are subject to analysis using this model (19).

The AIRg was calculated as the mean incremental insulin response above basal between 2 and 10 min after the intravenous glucose bolus. The glucose disappearance constant (K_g), a measure of intravenous glucose tolerance, was calculated as the slope of the natural log of glucose from 10 to 19 min, expressed as percent change per minute.

The magnitude of the insulin response was determined in part by the prevailing insulin sensitivity. Based on the known hyperbolic relationship between insulin sensitivity and the insulin response (11), we determined the disposition index, which was calculated as the product of S_i and AIRg. The calculation of this product provides a measure of β-cell function.

The cohort was divided into age quartiles for analysis. Because ages were listed as whole numbers, due to the presence of ties, quartiles are not equal in numbers. Subject characteristics and measures of glucose metabolism were compared by age quartiles using ANOVA with post hoc analysis. Continuous linear regression was performed to determine the relationships between age and variables of interest. Additionally, young (<50 years) and old (>60 years) subjects were pair matched for sex and intra-abdominal fat area and compared using Student’s paired t test. Stepwise multiple regression modeling was performed to examine the effects of age, sex, and subcutaneous and intra-abdominal fat on measures of S_i, AIRg, the disposition index, glucose effectiveness at 0 insulin, and K_g. Natural log transformation was performed if necessary to meet the assumption of normality of the statistical tests.

If the relationship between AIRg and S_i is hyperbolic (x × y = constant), then the relationship between lnAIRg and lnS_i is linear with a slope of −1. Thus, to assess whether the relationship between AIRg and S_i is hyperbolic in the entire cohort and specifically in the oldest age quartile, we performed linear regressions of lnAIRg on lnS_i. Because both the dependent (AIRg) and independent (S_i) variables include measurement with error, the properly weighted regression method, which incorporates error estimates of both the x and y variables, was used (20). Error estimates that were used for S_i and AIRg (coefficients of variation 16.9 and 20.6%, respectively) were based on studies performed in our lab (21). We also regressed lnAIRg on lnS_i and age quartile for all subjects, and we tested for significance using an age quartile × lnS_i interaction term. If this term was not significant, then the regression lines were parallel, i.e., the slopes were not significantly different. Data are presented as the means ± SE unless otherwise specified. A P value of ≤0.05 was considered significant.

The study cohort comprised 220 subjects with a broad range of age (26–75 years) and body size (BMI 18.7–40.4 kg/m²). Furthermore, they had broad ranges of S_i (0.7–30.0 ×10⁻⁵ min⁻¹/[pmol/l]), AIRg (32–2,638 pmol/l), K_g (0.60–3.90%/min), and disposition index (245–6461 ×10⁻⁵ min⁻¹) (Table 1).

The relationships between lnS_i and lnAIRg in the entire cohort of 220 was highly correlated (r = 0.412, P < 0.001), and the slope of the regression line was not significantly different from −1 (slope = −0.87 ± 0.13). To determine whether this relationship is also hyperbolic in older subjects, the same analysis was performed in the oldest age quartile (60–75 years). Again, lnS_i and lnAIRg were highly correlated (r = 0.477, P < 0.001), and the slope was not significantly different from −1 (slope = −1.1 ± 0.28), in keeping with the relationship between S_i and AIRg being hyperbolic. The hyperbolic relationship between S_i and AIRg for the youngest and oldest age quartiles are illustrated in Fig. 1. When the slopes of the regression lines describing the relationship between lnS_i and lnAIRg were determined for each age quartile, there was no statistically significant difference (age quartile × lnS_i interaction, P = 0.3), compatible with the hyperbolic relationship between S_i and the insulin response not being altered by differences in age.

Fasting plasma glucose and insulin did not differ based on age (Table 1). On the other hand, S_i was lower in the oldest quartile compared with the youngest quartile, whereas mean AIRg was lower in the oldest quartile, but this was not significant. Thus, based on the existence of the hyperbolic relationship between these two parameters in both younger and older subjects, when AIRg was adjusted for the prevailing insulin sensitivity, the disposition index was significantly lower in subjects in the oldest quartile. In keeping with this change, K_g was also decreased in the oldest age-group.

Simple linear regression analysis confirmed a significant negative correlation between age and S_i (r = −0.173, P = 0.01), age and the disposition index (r = −0.181, P < 0.01), and age and K_g (r = −0.173, P = 0.01). Neither fasting plasma glucose (r = 0.129, P = 0.06), fasting insulin (r = 0.001, P = 0.98), AIRg (r = 0.024, P = 0.73), nor glucose effectiveness at 0 insulin (r = 0.089, P = 0.19) were significantly correlated with age.

Intra-abdominal fat increased progressively with increasing age and was significantly greater in the oldest quartile (Table 2). There was no difference in BMI or subcutaneous fat between the age quartiles. The redistribution of abdominal fat toward more visceral fat was also reflected in the higher ratio of intra-abdominal fat to total abdominal fat areas, which increased with increasing age. The association between age and visceral adiposity was also demonstrable in linear regression models, which revealed positive correlations between age and intra-abdominal fat (r = 0.220, P < 0.001) and between age and the ratio of intra-abdominal fat to total abdominal fat areas (r = 0.277, P < 0.001). In contrast, neither BMI (r = 0.036, P = 0.60) nor abdominal subcutaneous fat area (r = 0.050, P = 0.46) were significantly correlated with age.

To determine whether insulin sensitivity, the acute insulin response, and β-cell function were associated with age independent of abdominal fat distribution, stepwise multiple linear regression analysis was performed using a model that included intra-abdominal fat, subcutaneous fat, age, and sex as the independent variables and the measure of glucose metabolism as the dependent variable (Table 3). Both intra-abdominal fat and subcutaneous fat were significantly associated with S_i, with intra-abdominal fat being more highly correlated. However, when adjusted for intra-abdominal fat and subcutaneous fat, age was no longer correlated with S_i. Intra-abdominal fat was significantly associated with AIRg, whereas age tended toward an inverse association with AIRg (r = −0.13, P = 0.06). Both intra-abdominal fat and age were significantly inversely associated with the disposition index. There was no significant effect of sex in any of the models.

To evaluate the effects of age and abdominal fat distribution on insulin-independent glucose disposal and intravenous glucose tolerance, the same stepwise multiple linear regression model was applied with glucose effectiveness at 0 insulin or K_g as the dependent variables. Neither age, sex, intra-abdominal fat, nor subcutaneous fat were associated with glucose effectiveness at 0 insulin. Age was associated with K_g, but intra-abdominal fat, subcutaneous fat, and sex were not. When glucose effectiveness at 0 insulin, AIRg, and S_i were added to the model as independent variables with K_g as the dependent variable, age was no longer significantly associated and glucose effectiveness at 0 insulin (r = 0.51, P < 0.001), AIRg (r = 0.551, P < 0.001), and S_i (r = 0.564, P < 0.001) were all significantly correlated with K_g.

Because intra-abdominal fat is so strongly associated with insulin sensitivity and β-cell function, to further investigate the effects of age on β-cell function, we pair matched 42 young (age 32–49 years) and older (age 60–74 years) subjects for sex and intra-abdominal fat (Table 4). Although intra-abdominal fat was matched, both BMI (P = 0.001) and subcutaneous fat area (P = 0.01) were significantly greater in the younger age group, whereas intra-abdominal fat/total abdominal fat area was significantly increased in the older group (P = 0.02). When matched for intra-abdominal fat and sex, insulin sensitivity was no longer significantly different between the younger and older subjects, but the disposition index tended to be lower in the older age-group (P = 0.06). K_g was also significantly lower in the older subjects (P < 0.05). Thus, whereas intra-abdominal fat may be the strongest predictor of S_i, other unidentified factors that occur with aging appear to contribute to the decline in glucose tolerance.

In this large cohort, we evaluated the effects of age and abdominal fat distribution on measures of insulin sensitivity and β-cell function in healthy adults and sought to validate use of the hyperbolic relationship between insulin sensitivity and the acute insulin response as a means to assess β-cell function in older populations. We found that this hyperbolic relationship does exist in both older and younger subjects; thus, these data support the use of the disposition index as a composite measure of β-cell function in all age-groups. Using this relationship between insulin sensitivity and the insulin response, we were then able to examine the impact of age and measures of abdominal fat distribution on β-cell function. Our findings demonstrate a strong association between increased intra-abdominal fat and decreased β-cell function, as well as demonstrating an independent association between increasing age and decreasing β-cell function.

The evidence for the hyperbolic relationship between insulin sensitivity and the insulin response to intravenous glucose and nonglucose secretagogues was first demonstrated in a cohort of healthy young subjects all <45 years of age (11). Calculation of the disposition index as determined by this relationship has been performed using data in subjects >45 years (4–6,22), but it has been suggested that using this relationship as a measure of β-cell function in older age-groups might not be appropriate (23). A key new finding in this study is demonstration of the existence of the hyperbolic relationship between S_i and AIRg in both younger and older subjects, thus providing support for the use of the disposition index as a composite measure of this interaction irrespective of age. With this in mind, the conclusions of studies that have recruited older subjects and have used this hyperbolic relationship to examine β-cell function (4–6,22) would be valid.

The effects of age on β-cell function independent of intra-abdominal fat in humans have not been clearly established. We found that although intra-abdominal fat was strongly associated with β-cell function, there was also a small independent association between age and β-cell function. Previous studies have demonstrated decreased insulin secretion in older subjects, but they did not specifically match for intra-abdominal fat (3,22). Others have demonstrated lack of an independent age effect on insulin sensitivity after adjusting for intra-abdominal fat, but they did not report similar analyses on the effect of intra-abdominal fat or age on measures of insulin secretion or β-cell function (6). The IRAS Family Study demonstrated independent effects of intra-abdominal fat and age on both insulin sensitivity and β-cell function (as measured by the disposition index) in Hispanic and African-American populations (10), but the validity of using the disposition index as a measure of β-cell function in older population groups had not been established. Our findings that intra-abdominal fat is most strongly correlated with β-cell function, but that age is also independently associated, support the findings from the IRAS Family Study. The mechanism(s) whereby increased visceral fat contributes to β-cell dysfunction is not known, but they could be mediated by free fatty acids as well as hormones secreted by adipocytes such as leptin, adiponectin, tumor necrosis factor-α, and interleukin 6, many of which have both central and peripheral effects.

Although increases in intra-abdominal fat may explain some of the age-related decline in β-cell function, additional mechanisms may contribute to the decrease in β-cell function with aging. Our group has previously shown that older subjects who undergo exercise training to normalize insulin sensitivity to that of young subjects continue to demonstrate defects in β-cell function (4), whereas others have shown that dietary carbohydrate content may impact β-cell function in older subjects (24). Animal data support the concept of a functional defect in insulin secretion that occurs with aging (25–27), with recent animal studies that controlled for body weight and visceral fat demonstrating a functional defect in insulin secretion independent of visceral fat and insulin action (27). The mechanisms whereby aging impairs β-cell function are not known but certainly warrant further investigation.

In addition to evaluating the effect of intra-abdominal fat on β-cell function, we have confirmed the important role of intra-abdominal fat in the insulin resistance of aging. Whereas the present study was cross-sectional and therefore cannot determine cause and effect, multiple studies have demonstrated that the decrease in insulin sensitivity with increased age is in large part associated with increases in intra-abdominal fat (6–9). Increased intra-abdominal fat is also recognized as a risk factor for the development of IGT and diabetes (28–32). That increased intra-abdominal fat accumulation may well be causal to the decline in S_i with aging is suggested by studies in aged rats demonstrating restoration of insulin sensitivity to that of young rats after the selective removal of visceral adipose tissue (33). In these studies, early removal of visceral fat prevented not only the progressive decline in insulin sensitivity over time but also delayed the progression to diabetes (33). Our findings that intra-abdominal fat is most strongly associated with insulin sensitivity and that when intra-abdominal fat is adjusted for, age is no longer associated with insulin sensitivity, add to this collective literature supporting the hypothesis that intra-abdominal fat has a role in determining insulin sensitivity and contributes to the age-related decline in insulin sensitivity.

There are a number of strengths of the current study, including the large number of subjects examined, the inclusion of similar numbers of men and women, and the broad range of ages and body habitus. However, because we selected only apparently healthy subjects, the older subjects in the cohort may represent healthy survivors, which is likely to have reduced our ability to detect differences between the age-groups and thus tends to strengthen our findings. Although we did not study subjects >75 years of age, we do not believe that this affected our findings because otherwise healthy adults >80 years of age have been shown to have glucose metabolism similar to those aged 61–79 years (34). In this study we quantified glucose tolerance as K_g from an intravenous glucose tolerance test. Although we did not perform oral glucose tolerance tests, which have demonstrated postchallenge hyperglycemia to be more common in older subjects (1), we do not believe this limits our findings because the older subjects had a lower K_g, which is consistent with the decreased glucose tolerance observed with the oral glucose tolerance test.

In summary, we have demonstrated that the hyperbolic relationship between insulin sensitivity and the acute insulin response is present in older populations, thus providing support for use of the disposition index as a composite measure of β-cell function irrespective of age. Furthermore, our analysis has demonstrated that the decline in β-cell function with age is associated with an increase in intra-abdominal fat, but that age is also independently associated with decreased β-cell function. This decline in β-cell function with age likely contributes to the increased prevalence of glucose intolerance and type 2 diabetes in the older population.

This work was supported in part by the Medical Research Service of the Department of Veteran Affairs; the American Egg Board; National Institutes of Health Grants DK-02456, DK-02654, DK-17047, DK-35747, DK-35816, DK-59417, HL-30086, HL-07028, and RR-37; the American Diabetes Association; the U.S. Department of Agriculture; and the McMillen Family Trust.

We thank the subjects who participated in the study and Diane Collins and the nursing staff of the General Clinical Research Center at the University of Washington for the care of the subjects. Brian Fish is thanked for his help with data management. We acknowledge the contribution of the staff of Immunoassay Core of the Diabetes Endocrinology Research Center, who performed the assays.