Fasting insulin reflects heterogeneous physiological processes: role of insulin clearance

Abstract

Several processes contribute to variation in fasting insulin concentration, including fasting glucose, insulin resistance, insulin secretion, and insulin clearance. Our goal was to determine the relative contribution of each of these insulin-related traits, plus anthropometric parameters, to fasting insulin among 470 Mexican Americans. The euglycemic hyperinsulinemic clamp yielded insulin sensitivity (M value) and metabolic clearance rate of insulin (MCRI). Acute insulin secretion was estimated by the insulinogenic index (IGI30) from the oral glucose tolerance test. Regression (univariate) and generalized estimating equations (multivariate) were used to describe the relationship of insulin-related traits to fasting insulin. Univarate analyses were used to select which traits to include in the multivariate model. In multivariate analysis, MCRI, M, BMI, waist circumference, and fasting glucose were independently associated with fasting insulin. Decreasing M and MCRI were associated with increasing fasting insulin, whereas increasing BMI, waist circumference, and fasting glucose were associated with increasing fasting insulin. Standardized coefficients allowed determination of the relative strength of each trait's association with fasting insulin in the entire cohort (strongest to weakest): MCRI (−0.35, P < 0.0001), M (−0.24, P < 0.0001), BMI (0.20, P = 0.0011), waist circumference (0.16, P = 0.021), and fasting glucose (0.11, P = 0.014). Fasting insulin is a complex phenotype influenced by several independent processes, each of which might have its own environmental and genetic determinants. One of the most associated traits was insulin clearance, which has implications for studies that have used fasting insulin as a surrogate for insulin resistance.

abnormalities of insulin action and/or hyperinsulinemia are central to several conditions such as diabetes mellitus, the metabolic syndrome, and polycystic ovary syndrome. Insulin may also affect the development of atherosclerosis, as elevated fasting insulin is a risk factor for subsequent atherosclerosis in epidemiologic studies (4, 17). Ambient insulin concentrations, which vary widely between individuals, are influenced by several distinct physiological processes, including insulin sensitivity, insulin secretion, and insulin clearance.

Because of ease of measurement, fasting insulin [or the closely related homeostatic model assessment of insulin resistance (HOMA-IR)] has been frequently used as a surrogate marker of insulin resistance in large studies. The correlation of fasting insulin-derived measures with gold-standard measures of insulin resistance (e.g., euglycemic clamp or frequently sampled intravenous glucose tolerance test) is imperfect (r values ranging from 0.2 to 0.7) (1, 23), suggesting the influence of other physiological factors. The relative contribution of insulin sensitivity, secretion, and clearance to the variability in fasting insulin concentrations is not known, particularly on a population scale.

Mexican Americans have a high prevalence of hyperinsulinemia/insulin resistance and the metabolic syndrome (15, 27, 28). In the Mexican-American Coronary Artery Disease (MACAD) study, insulin phenotypes were evaluated using the hyperinsulinemic-euglycemic clamp study (11), regarded as a gold-standard technique for direct physiological measurement of insulin sensitivity (3, 32). The clamp also allows calculation of the metabolic clearance rate of insulin (MCRI). In the MACAD study, insulin secretion was calculated from the oral glucose tolerance test (OGTT). Given that all these aspects of insulin action and dynamics were measured in the same subjects, we assessed the relationship of these three parameters, as well as fasting glucose and anthropometric markers, with fasting insulin concentration and identified MCRI, insulin sensitivity (M value), and BMI as quantitatively the most important correlates in this population.

EXPERIMENTAL PROCEDURES

Study subjects.

Associations with fasting insulin were assessed in participants of the Cedars-Sinai/UCLA Mexican-American Coronary Artery Disease (MACAD) study, a study of Mexican American families from Los Angeles (11, 12). In the present report, 128 families were included, comprising 470 subjects from the offspring generation (adult offspring of probands with CAD and the spouses of those offspring) who underwent phenotyping. By design, the offspring were free of diabetes and clinically manifested cardiovascular disease, thus avoiding secondary changes in phenotype caused by overt disease. All studies were approved by Human Subjects Protection Institutional Review Boards at UCLA and Cedars-Sinai Medical Center. All subjects gave informed consent prior to participation.

Phenotyping procedures.

Four hundred seventy adult offspring and their spouses underwent a phenotyping protocol that included an OGTT on one day, and a hyperinsulinemic-euglycemic clamp study on a separate day.

The fasting insulin measurements reported in this study were obtained after a 12-h fast, immediately prior to the hyperinsulinemic-euglycemic clamp. Fasting insulin was measured four times within 30 min prior to the clamp; the average of these four measurements was used as our fasting insulin value. Insulin was measured using the Human Insulin Specific RIA Kit (LINCO Research, St. Charles, MO; <0.2% cross-reactivity with proinsulin; manufacturer's inter- and intra-assay coefficients are 2.9–6.0 and 2.2–4.4%, respectively).

During the hyperinsulinemic-euglycemic clamp (3), a priming dose of human insulin (Novolin, Clayton, NC) was given and followed by infusion for 120 min at a constant rate (60 mU·m⁻²·min⁻¹) with the goal of achieving a plasma insulin concentration of 100 μIU/ml or greater. Blood was sampled every 5 min, and the rate of 20% dextrose coinfused was adjusted to maintain plasma glucose concentrations at 95 to 100 mg/dl. The glucose infusion rate (M value, mg·m⁻²·min⁻¹) over the last 30 min of steady-state insulin and glucose concentrations reflects glucose uptake by all tissues of the body (primarily insulin-mediated glucose uptake in muscle) and is therefore directly correlated with tissue insulin sensitivity (3). Often, an insulin sensitivity index (SI, mg·m⁻²·min⁻¹·μIU⁻¹·ml) is calculated as M/I, where I is the steady-state insulin level. In this study, to clearly distinguish between insulin sensitivity and insulin clearance in multivariate analyses, we relied on M as an approximation for insulin sensitivity in our primary analyses, because the calculations of SI and insulin clearance both use steady-state insulin in the denominator. The metabolic clearance rate of insulin (MCRI, ml·m⁻²·min⁻¹) was calculated as the insulin infusion rate divided by the steady-state plasma insulin level of the euglycemic clamp, as previously described (3, 13). To test the effect of the definitions of insulin sensitivity and clearance, we conducted secondary analyses that used SI as the insulin sensitivity measure.

The OGTT consisted of baseline glucose and insulin measurements followed by administration of 75 g of oral glucose with blood draws at 30, 60, 90, 120, and 180 min. Insulin secretion was obtained from the OGTT fasting and 30-min time point glucose and insulin measurements to estimate early insulin secretion, as the insulinogenic index [(insulin at 30 min − fasting insulin)/(glucose at 30 min − fasting glucose), μIU·dl·mg⁻¹·ml⁻¹] (18). The IGI30 is a measure of β-cell insulin response to oral glucose and therefore represents a dynamic measurement rather than a steady-state or basal assessment.

Data analysis.

Log-transformed (BMI, IGI30, fasting insulin) or square-root-transformed (M, SI, MCRI) trait values were used to normalize the distribution for all statistical analyses. Unpaired, two-sided t-tests were used to compare trait values between men and women. Simple regression was utilized for univariate analyses of fasting insulin vs. the traits of interest. Multivariate analyses were generated using generalized estimating equations (GEE) to assess the joint effects of BMI, waist circumference, fasting glucose, M (or SI), MCRI, and IGI30 on fasting insulin, adjusting for familial relationships. The weighted GEE1 (34) was computed assuming an exchangeable correlation structure and using the sandwich estimator of the variance to account for familial correlation present in family data. GEE was used to derive standardized regression coefficients, which in any one regression equation are measured on the same scale, with a mean of 0 and a standard deviation of 1. They are then directly comparable to one another, with the largest coefficient indicating which independent variable has the greatest influence on the dependent variable (26). We also generated multivariate models for fasting insulin in men and women separately; these analyses were followed by a repeat analysis in the entire cohort, including sex-by-trait interaction terms to evaluate the significance of any observed sex differences.

By use of procedures similar to those described above, additional exploratory analyses were conducted examining correlates of insulin secretion and insulin clearance separately.

RESULTS

The clinical characteristics of the 470 subjects (197 men, 273 women) who had quantitative assessment of insulin resistance are shown in Table 1. Insulin secretion (IGI30) was lower and insulin sensitivity (M and SI) and waist circumference significantly higher in men than in women. Fasting glucose was slightly but statistically significantly higher in the men. The other insulin-related traits did not differ significantly by sex.

Table 1.

Clinical characteristics of the study cohort

	Men (n = 197)	Women (n = 273)	P Value
Age, yr	33.0 (14.0)	34.0 (12.0)	0.58
BMI, kg/m2	28.2 (5.2)	28.2 (7.0)	0.98
Waist circumference, cm	95.5 (13.0)	88.5 (17.1)	<0.0001
Fasting glucose, mg/dl	94.9 (12.1)	92.0 (11.6)	0.0001
Fasting insulin, μIU/ml	11.3 (8.1)	12.0 (7.4)	0.95
M, mg·m−2·min−1	254.3 (155.9)	205.7 (119.7)	0.008
SI, mg·m−2·min−1·μIU−1·ml	1.92 (1.62)	1.70 (1.18)	0.039
MCRI, ml·m−2·min−1	479.2 (157.5)	483.9 (141.8)	0.46
IGI30, μIU·dl·mg −1·ml−1	1.10 (1.05)	1.43 (1.53)	0.0004

Data are medians (interquartile range). M, glucose infusion rate; SI, insulin sensitivity index; MCRI, metabolic clearance rate of insulin; IGI30, insulinogenic index.

Univariate analyses (Table 2) found significant relationships of fasting insulin with BMI, waist circumference, fasting glucose, M, SI, MCRI, and IGI30. Regression plots are displayed in Fig. 1. Joint analysis of BMI, waist circumference, fasting glucose, M, MCRI, and IGI30 on fasting insulin revealed that all except IGI30 were independently significantly associated with fasting insulin (Table 3). Increasing MCRI and M were associated with decreasing fasting insulin. Increasing BMI, waist circumference, and fasting glucose were associated with increasing fasting insulin. Similar results were obtained when SI was used as the insulin sensitivity measure [Suppl. Table S1 (supplementary materials are found with the online version of this paper at the Journal website)].

Table 2.

Univariate correlations of insulin-related and anthropometric traits with fasting insulin

	Entire Cohort		Men		Women
	R value	P value	R value	P value	R value	P value
Age	–0.035	0.45	–0.005	0.94	–0.068	0.26
BMI	0.56	<0.0001	0.60	<0.0001	0.54	<0.0001
Waist circumference	0.54	<0.0001	0.60	<0.0001	0.54	<0.0001
Fasting glucose	0.28	<0.0001	0.23	<0.0001	0.33	<0.0001
M	–0.47	<0.0001	–0.49	<0.0001	–0.45	<0.0001
SI	–0.60	<0.0001	–0.58	<0.0001	–0.63	<0.0001
MCRI	–0.48	<0.0001	–0.43	<0.0001	–0.52	<0.0001
IGI30	0.21	<0.0001	0.23	0.0012	0.20	0.0009

Fig. 1.

Regression plots on fasting insulin of the traits significantly associated with it in univariate analyses. The r values and P values are listed in Table 2. A: BMI. B: waist circumference. C: M value. D: MCRI. E: IGI30. F: fasting glucose.

Table 3.

Multivariate correlations of insulin-related and anthropometric traits with fasting insulin

	Standardized Coefficient	Standard Error	95% Confidence Limits		P Value
Entire cohort (51.7% fasting insulin variance)
BMI	0.20	0.062	0.081	0.32	0.0011
Waist circumference	0.16	0.071	0.025	0.30	0.021
Fasting glucose	0.11	0.034	0.042	0.18	0.0014
M	−0.24	0.045	−0.33	−0.16	<0.0001
MCRI	−0.35	0.037	−0.42	−0.27	<0.0001
IGI30	0.034	0.030	−0.024	0.093	0.25
Men (51.2% of variance explained)
BMI	0.31	0.11	0.10	0.53	0.0037
Waist circumference	0.19	0.11	−0.022	0.40	0.079
Fasting glucose	0.063	0.054	−0.043	0.17	0.24
M	−0.22	0.067	−0.35	−0.086	0.0012
MCRI	−0.33	0.066	−0.46	−0.20	<0.0001
IGI30	0.036	0.042	−0.046	0.118	0.39
Women (53.5% of variance explained)
BMI	0.14	0.075	−0.007	0.29	0.063
Waist circumference	0.15	0.081	−0.014	0.31	0.073
Fasting glucose	0.13	0.044	0.043	0.22	0.0032
M	−0.26	0.047	−0.35	−0.17	<0.0001
MCRI	−0.36	0.038	−0.44	−0.29	<0.0001
IGI30	0.037	0.037	−0.035	0.11	0.31

Significant values are indicated in boldface. P values were derived using generalized estimating equations to account for familial relationships.

Comparison of standardized coefficients allowed determination of the relative strength of each trait's association with fasting insulin (listed strongest to weakest): MCRI (−0.35, P < 0.0001), M (−0.24, P < 0.0001), BMI (0.20, P = 0.0011), waist circumference (0.16, P = 0.021), and fasting glucose (0.11, P = 0.014). These factors explained 52% of the variance in fasting insulin concentrations. Similar standardized coefficients were obtained from the model utilizing SI rather than M for insulin sensitivity: SI (−0.28, P < 0.0001), MCRI (−0.22, P < 0.0001), BMI (0.22, P = 0.0004), waist circumference (0.15, P = 0.015), fasting glucose (0.10, P = 0.0022) (Suppl. Table S1).

To determine whether these relationships differed by sex, we conducted multivariate analyses separately in the men and women. While MCRI exhibited the highest standardized coefficient within each sex group, the contribution of BMI to fasting insulin was relatively more important in men. The statistical significance of the differential contribution of BMI was confirmed by running the GEE model with interaction terms of sex with the other independent variables; only the sex-by-BMI interaction term was significant (P = 0.016).

Because insulin secretion and insulin clearance are the physical determinants of circulating insulin levels, we also conducted exploratory analyses characterizing the traits associated with each of these factors. Univariate analyses (Suppl. Table S2) identified BMI, SI, and waist circumference as associated with IGI30; these traits, as well as sex (Table 1) were analyzed jointly for their relationship with IGI30, revealing standardized coefficients of 0.19 for BMI (P = 0.011), 0.16 for SI (P = 0.0026), and 0.13 for sex (P = 0.0045) (Suppl. Table S3). Univariate analyses (Suppl. Table S4) identified BMI, waist circumference, and M for inclusion in multivariate analyses for MCRI, in which only waist circumference was significant (standardized coefficient −0.16, P = 0.012) (Suppl. Table S5).

DISCUSSION

We found that insulin clearance and insulin sensitivity, as well as BMI, waist circumference, and fasting glucose, are independently associated with variation in fasting insulin concentrations, together accounting for 52% of the variation in fasting insulin. Insulin clearance, insulin sensitivity, and BMI were the most strongly associated traits on a population level. To our knowledge, this is the first study reporting the joint effect of these insulin-related and anthropometric phenotypes on fasting insulin. Most prior studies examining trait relationships with fasting insulin were focused on anthropometric, ethnic, dietary, and lifestyle effects on insulinemia (10, 22, 25, 30). A twin study found nongenetic variation in BMI to be associated with fasting insulin; however, that study did not quantify insulin resistance, clearance, or secretion (24). The European Group for the Study of Insulin Resistance (EGIR) study, in which a large number of subjects underwent the euglycemic clamp, determined that the main associations with fasting insulin were, in descending order, BMI, insulin clearance, insulin sensitivity, and plasma glucose; these four factors accounted for 30% of the variation in fasting insulin (9). In that work, insulin secretion and body fat distribution were not included in the model; the authors assumed that the remaining variation was explained by insulin secretion. Another study utilizing the clamp found that MCRI and C-peptide (index of insulin secretion), but not hepatic insulin sensitivity, were significantly associated with HOMA-IR (which tracks closely with fasting insulin); that study did not evaluate anthropometric traits (21).

In our study, insulin sensitivity was strongly associated with fasting insulin, highlighting the importance of compensatory hyperinsulinemia in response to insulin resistance. This close relationship between fasting insulin and insulin sensitivity has led to the use of fasting insulin as a surrogate measure of insulin sensitivity by many investigators; however, given the contributions of insulin clearance, BMI, waist circumference, and fasting glucose to fasting insulin, it is clear that fasting insulin also reflects other aspects of insulin dynamics. This may contribute to the paucity of associations found with fasting insulin in large genome-wide association studies (6).

Of the dynamic factors, secretion and clearance, that physically determine insulin concentration, insulin clearance exhibited a stronger association with population variation in fasting insulin than acute insulin secretion. In support of this result, a clamp study of 10 men with intentional weight gain (BMI 21.8 to 23.8 kg/m²) found that reduction in insulin clearance was the primary contributor to the resulting increased basal and stimulated insulin levels (7). A second study, characterizing increased postprandial insulin levels going from normal weight to overweight to obese, found that insulin secretion plateaued in the overweight group, whereas reductions in insulin clearance progressed in obese and morbidly obese groups (8).

Insulin binding to its receptor triggers internalization of the insulin-insulin receptor complex. Within the cell, insulin is cleaved by insulin-degrading enzyme and other enzymes (5). Clearance of insulin occurs mainly in the liver and kidneys; however, given the above mechanism, any insulin-responsive tissue cell also clears insulin (5). Insulin clearance is the least studied aspect of insulin metabolism. Our results point to the need for further study of insulin clearance in hyperinsulinemic disorders. Insulin clearance is a highly heritable trait (13), raising the possibility that genetic determinants of insulin clearance may affect risk for hyperinsulinemic disorders such as diabetes mellitus or polycystic ovary syndrome.

The relatively weak contribution of our measurement of insulin secretion to fasting insulin might indicate that insulin secretion does not contribute strongly to interindividual differences in fasting insulin, at least in this population. However, IGI30 is an index of acute insulin secretion in response to a glucose load. Although a relationship between stimulated and basal insulin secretion is likely, it is probable that basal insulin secretion is a more important contributor to variation in fasting insulin concentrations. Fasting C-peptide, a reflection of basal insulin secretion, was not obtained in the MACAD study herein. Acute insulin secretion may be influenced by incretin hormones. Also, our estimation of insulin secretion by 30-min OGTT values may represent a less precise measurement of this trait than values obtained by procedures such as the hyperglycemic clamp or frequently sampled intravenous glucose tolerance test (FSIGT). A study utilizing the FSIGT found that insulin hypersecretion was more important than reduced hepatic insulin extraction in the hyperinsulinemia present in obese adolescents (2).

Fasting glucose exhibited a significant and independent association with fasting insulin levels. Fasting glucose likely reflects both basal insulin secretion and hepatic glucose output (31). As found in previous studies (25, 30), BMI and waist circumference were also independently associated with fasting insulin, even when included in the same regression model, suggesting that total adiposity as well as central fat distribution influence fasting insulin. This may be due to adipokines or similar factors produced by adipose tissue that may directly or indirectly regulate insulin production, clearance, and/or resistance. When the sex groups were analyzed separately, the association of BMI with fasting insulin remained significant in the men only. We speculate that this might be related to sexual dimorphism in adipokine levels [e.g., higher adiponectin in women (29)] and/or differences in the genetic regulation of adiposity (20).

The factors that we considered together explained 52% of the overall variation in fasting insulin levels, suggesting the existence of other physiological regulators of fasting insulin, such as diet and physical activity (10, 22, 25, 30) and/or basal insulin secretion. Our study was carried out in Mexican Americans, a group known for its high incidence of insulin resistance (15, 27, 28). Whether our results on the greater importance of MCRI and M on fasting insulin can be generalized to other populations is unknown; however, prior studies in European cohorts yielded partially consistent findings (9, 21). Caution in generalizing these findings is warranted, as Mexican Americans have been found to have lower insulin sensitivity and insulin clearance than non-Hispanic whites (16), but higher insulin clearance than African Americans (14, 19, 33).

Blood levels of insulin are directly controlled by insulin secretion (input) and insulin clearance (output). In separate multivariate models of these two traits, we found that insulin sensitivity, BMI, and waist circumference were all associated with insulin secretion, whereas only waist circumference was associated with insulin clearance. That insulin sensitivity, BMI, and waist circumference remained significantly associated with fasting insulin when analyzed jointly with insulin secretion and clearance suggests that there may be aspects of insulin secretion and insulin clearance that are not fully captured by the IGI30 and MCRI variables available to us.

Our measurement of insulin clearance was not perfect. The euglycemic clamp induces a hyperinsulinemic state, during which the physiological machinery of insulin clearance is likely to be saturated; therefore, the MCRI we calculated may be lower than expected at physiological insulin levels. The clamps performed in the MACAD study did not utilize radiolabeled glucose or insulin, which may have affected the accuracy of our insulin sensitivity and insulin clearance measurements, respectively.

In conclusion, insulin clearance, insulin sensitivity, and body mass index were the most important correlates of variability in fasting insulin in the population under study. Fasting insulin is thus a complex phenotype influenced by several independent processes, each of which might have its own environmental and genetic determinants. The convenience of fasting insulin in epidemiological and genetic studies must therefore be tempered by its heterogeneous underpinnings.

GRANTS

This study was supported, in part, by National Institutes of Health Grants R01-HL-088457, R01-DK-079888, P30-DK-063491, and M01-RR-00425 (General Clinical Research Center Grant from the NCRR), the Cedars-Sinai Winnick Clinical Scholars Award (to M. O. Goodarzi), and the Cedars-Sinai Board of Governor's Chair in Medical Genetics (J. I. Rotter).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).