Biomarkers of Endothelial Dysfunction and Risk of Type 2 Diabetes Mellitus FREE

Type 2 diabetes mellitus is increasingly prevalent worldwide,¹ conferring major burdens on health and health care costs. Type 2 diabetes may be largely preventable,^2- 3 but a comprehensive understanding of its etiology is still needed. Development of atherosclerotic cardiovascular disease (CVD) is the principal complication in type 2 diabetes, but clinical CVD can also precede development of diabetes,⁴ lending support to the hypothesis that diabetes and CVD share common antecedents. A syndrome of insulin resistance may constitute this common antecedent,⁵ but mechanisms unifying diverse effects of insulin resistance are not well defined. Insulin resistance is an established factor in the pathogenesis of type 2 diabetes but has an uncertain association with CVD.⁶

Subclinical inflammation could be a unifying factor because it is a precursor of CVD, is associated with insulin resistance, and precedes development of type 2 diabetes.^7- 8 Inflammatory mediators may be pathogenic by inducing systemic endothelial dysfunction. This hypothesis localizes insulin resistance and atherosclerosis to a unifying tissue type: in large arteries, endothelial dysfunction leads to clinical CVD, whereas in the capillary and arteriolar endothelium, with a vast surface area in intimate contact with metabolically active, insulin-sensitive tissues, endothelial dysfunction may lead to type 2 diabetes.^9- 10 Identification of endothelial dysfunction as a type 2 diabetes precursor might expand options for diabetes prevention and treatment.

Endothelial dysfunction can be detected by measurement of elevated plasma levels of cellular adhesion molecules (CAMs), including E-selectin, intercellular adhesion molecule 1 (ICAM-1), and vascular cell adhesion molecule 1 (VCAM-1). Elevated levels of CAMs have been a consistent finding in cross-sectional studies of patients with type 2 diabetes^11- 12 and people at increased risk for diabetes.^12- 14 Retinal arteriolar narrowing, a marker of microvascular dysfunction, has been shown to predict incident type 2 diabetes,¹⁵ but prospective data directly linking endothelial dysfunction to development of diabetes remain sparse and inconsistent.^16- 17 In this study, we tested the hypothesis that elevated levels of E-selectin, ICAM-1, and VCAM-1 predict incident type 2 diabetes in women, independent of known diabetes risk factors, including obesity and subclinical inflammation.

The Nurses' Health Study began in 1976, when 121 700 female nurses aged 30 to 55 years and from 11 US states responded to a questionnaire of health-related information. Questionnaires have been administered biennially to update health information and identify new cases of disease.

During 1989-1990, 32 826 women free of diagnosed diabetes, coronary heart disease, stroke, or cancer provided blood samples. By 2000, 737 of these women had developed type 2 diabetes. Controls providing baseline blood samples were matched to diabetes cases by year of birth, date of blood draw, race, and fasting status (at least 8 hours overnight) at blood draw. In addition, another control was matched on these characteristics and also on body mass index (BMI) to each of the most obese cases (cases in the top 10% of the BMI distribution). We matched the most obese cases on BMI to improve statistical control for obesity, seeking to ensure that differences in risk of diabetes were not a function of incomplete control of confounding by higher levels of obesity in diabetes cases. Thus, we analyzed a sample of 785 controls.

Women who provided blood samples had a higher prevalence of obesity, a higher prevalence of a family history of diabetes, and a lower prevalence of current smoking but were otherwise similar to women not providing blood. Subjects provided written informed consent, and the study was approved by the institutional review board of Partners HealthCare System, Boston, Mass.

Baseline and incident type 2 diabetes were identified by self-report and confirmed by a validated supplementary questionnaire detailing symptoms, diagnostic laboratory test results, and diabetes treatment. Women were confirmed to have type 2 diabetes if they reported at least 1 of the following on the supplementary questionnaire: treatment with either insulin or an oral hypoglycemic agent, at least 1 classic symptom of diabetes (for instance, polyuria, polydipsia, weight loss) plus an elevated plasma glucose level, or an elevated plasma glucose level on 2 occasions. Elevated plasma glucose was defined as at least 140 mg/dL (≥7.8 mmol/L) fasting, at least 200 mg/dL (≥11.1 mmol/L) nonfasting, or at least 200 mg/dL (≥11.1 mmol/L) 2 hours after an oral glucose tolerance test for cases diagnosed before 1998; for cases diagnosed in 1998 and later, the fasting plasma glucose threshold was at least 126 mg/dL (≥7.0 mmol/L).¹⁸ Women were classified with incident diabetes if they met these criteria and were diagnosed at least 1 year after the date of blood collection. The validity of self-reported diabetes in this cohort has been confirmed with medical record review.¹⁹

Every 2 years, exposure status has been updated by questionnaire, including smoking, menopausal status and use or nonuse of postmenopausal hormone therapy, and body weight. We calculated BMI (measured in 1988) as weight in kilograms divided by the square of height in meters. In 1986, we assessed self-reported waist girth. The presence or absence of a family history of diabetes in first-degree relatives was assessed in 1982 and 1988. Information about physical activity was obtained in 1980, 1982, 1986, 1988, and 1992. Diet and alcohol consumption were assessed in 1980, 1984, 1986, and 1990 by using semiquantitative food frequency questionnaires. Physical activity and diet exposures were calculated as updated cumulative average levels. The reproducibility and validity of the food-frequency questionnaires have been described.²⁰ We summarized intake of cereal fiber, glycemic load, trans-fats, and the ratio of polyunsaturated to saturated fats (scored 1-5 for each, with 5 being the most healthy intake) to create a diet score (scored 4-20), with a high diet score associated with a reduced risk of type 2 diabetes.²¹ Reported weights have been shown to correlate with measured weights (r = 0.96); reported waist girth, with measured waist circumference (r = 0.89).^22- 23 Physical activity assessment has also been validated.²⁴

Women providing blood samples were sent a phlebotomy kit, returning the sample by overnight mail in a frozen water bottle. On arrival, samples were processed and frozen in liquid nitrogen until analysis; 97% arrived within 26 hours of phlebotomy. Quality control samples were routinely frozen with study samples; the long-term stability of plasma samples collected and stored under this protocol has been documented.²⁵ Study samples were analyzed in randomly ordered case-control pairs to further reduce systematic bias and interassay variation.

Levels of E-selectin, ICAM-1, and VCAM-1 were measured in 2002 by commercial enzyme-linked immunosorbent assay (R & D Systems, Minneapolis, Minn). These biomarkers are reliable markers of early atherosclerosis²⁶ and have modest correlations (r = 0.04-0.58) with endothelial dysfunction assessed directly by brachial artery flow-mediated vasodilatation or microcirculation iontophoresis methods in a variety of populations.^12,27- 29 C-reactive protein (CRP) levels were measured via a high-sensitivity latex-enhanced immunonephelometric assay (Dade Behring, Newark, Del). Insulin levels were measured by using a double antibody system with less than 0.2% cross-reactivity between insulin and its precursors (Linco Research, St Louis, Mo). Hemoglobin A_1c was measured by immunoassay (Roche Diagnostics, Indianapolis, Ind). The coefficients of variation for analytes were E-selectin, 4.50% to 6.22%; ICAM-1, 3.56%; VCAM-1, 8.48% to 9.77%; CRP, 2.07% to 4.47%; fasting insulin, 3.52% to 11.7%; and hemoglobin A_1c, 1.9% to 3.0%.

We conducted a prospective, nested case-control analysis of endothelial dysfunction biomarkers as predictors of incident type 2 diabetes. We compared baseline characteristics of study women by using t tests, χ² tests, or Wilcoxon rank-sum tests. We used conditional logistic regression to account for correlations introduced by matching to estimate the relative risk and 95% confidence intervals (CIs) for levels of biomarkers predicting type 2 diabetes. We divided the distributions of E-selectin, ICAM-1, and VCAM-1 into quintiles according to the distribution in controls and used regression models to estimate the significance of trend in relative risk across increasing quintiles and to estimate risk of diabetes in each quintile relative to the lowest quintile.

Using nested regression models conditioned on matching for age, race, and fasting status, we estimated crude relative risks of diabetes; then adjusted for BMI, family history of diabetes, smoking, alcohol intake, diet score, physical activity, and hormone use; further adjusted for these covariates and levels of CRP; and finally simultaneously adjusted for all diabetes risk factors and biomarker levels. In these analyses, we had at least 85% power to detect a significant linear trend (2-sided P<.05) across quintiles in which the relative risk in the top quintile relative to the bottom was at least 1.5.³⁰ In a subsidiary analysis among women providing fasting blood samples, we also adjusted nested regression models for levels of fasting insulin and hemoglobin A_1c. We used SAS (version 6.12; SAS Institute Inc, Cary, NC) for all analyses and defined statistical significance as P<.05.

Women who developed type 2 diabetes during 10 years of follow-up were more obese, had a greater prevalence of a family history of diabetes, less alcohol use, a less favorable diet score, less physical activity, and a lower prevalence of hormone therapy use than did controls (Table 1). Levels of fasting insulin, hemoglobin A_1c, E-selectin, ICAM-1, and VCAM-1 were all significantly higher at baseline in women who developed diabetes compared with those who remained nondiabetic at follow-up.

Elevated levels of endothelial dysfunction biomarkers measured at baseline significantly increased the relative risk of incident diabetes. In analyses conditioned on matching for age, race, and fasting status, the relative risk of diabetes was 7.50 (95% CI, 5.05-11.14) in the highest quintile of E-selectin relative to the lowest quintile; for ICAM-1, the relative risk was 4.29 (95% CI, 2.95-6.23); and for VCAM-1, the relative risk was 1.54 (95% CI, 1.10-2.15) (Table 2, model 1). After adjustment for BMI, family history of diabetes, smoking, diet score, alcohol intake, activity index, and postmenopausal hormone use, elevated levels of E-selectin and ICAM-1 but not VCAM-1 remained significant predictors of incident diabetes (Table 2, model 2).

These results remained similar when waist circumference was used instead of BMI (as in model 2, Table 2) as an index of obesity: for E-selectin, the relative risk of diabetes in the highest quintile was 5.08 (95% CI, 3.05-8.47); for ICAM-1, 2.46 (95% CI, 1.50-4.03); and for VCAM-1, 1.05 (95% CI, 0.66-1.70). Inflammation may be involved in the association between endothelial dysfunction and risk of diabetes, but further adjustment of model 2 for levels of CRP did not eliminate associations of E-selectin and ICAM-1 with incident diabetes (Table 2, model 3). Cellular adhesion molecules may act independently or in a causal pathway leading to diabetes, but simultaneous adjustment for all 3 biomarkers and other diabetes risk factors (Table 2, model 4) did not eliminate associations of E-selectin and ICAM-1 with diabetes.

We conducted subsidiary analyses (Table 3) to ensure that the baseline sample was free of diabetes (by excluding women with baseline hemoglobin A_1c levels >6.5%); that associations were not due to an effect of incipient diabetes on endothelial function (by excluding cases diagnosed during the first 4 years of follow-up); and that associations were not due to an effect of subclinical CVD at baseline (by excluding women with incident CVD during follow-up). Although blood pressure and plasma lipid levels have not been measured in this cohort, we attempted to account for their possible confounding effects by adjusting models for reported treatment for hypertension or hyperlipidemia. In all subsidiary analyses, levels of E-selectin and ICAM-1, but not VCAM-1, remained independent predictors of incident type 2 diabetes.

Hyperinsulinemia and hyperglycemia are potent risk factors for type 2 diabetes and may mediate effects of endothelial dysfunction on diabetes risk. We assessed these effects in the subset of women providing fasting blood samples in models adjusted for diabetes risk factors and levels of hemoglobin A_1c and fasting insulin (Table 4). In this analysis, elevated levels of E-selectin, but not ICAM-1 or VCAM-1, were still a significant independent predictor of incident type 2 diabetes.

Obesity and the mediating effects of increased adipocyte signaling could amplify the effect of endothelial dysfunction, increasing risk for type 2 diabetes. We assessed the joint effect of BMI and levels of E-selectin, ICAM-1, and VCAM-1 on risk of diabetes (Figure 1). BMI and endothelial dysfunction had only additive effects on diabetes risk: at every level of increasing BMI or E-selectin, there was a stepwise increase in the risk of diabetes. The most obese women with the highest levels of E-selectin had a 13.6-fold (P<.001) increased risk of diabetes relative to the leanest women with the lowest E-selectin levels. Patterns were similar for increasing levels of ICAM-1 and BMI (P = .02 for ICAM-1 at each level of BMI). Elevated levels of VCAM-1 did not substantially increase risk of diabetes at any level of BMI. In each of these analyses, P values for first-order biomarker × BMI interaction terms were not statistically significant.

A fundamental concept underlying the insulin resistance syndrome is that there is some shared precursor to type 2 diabetes and CVD. Insulin resistance has been a plausible candidate for this common precursor, but specific mechanisms whereby insulin resistance leads to both diabetes and CVD remain ill defined. Systemic inflammation is associated with insulin resistance and incident CVD and diabetes.^{7- 8,31- 32} A specific mechanism whereby inflammation may contribute to these disease processes is induction of endothelial dysfunction,³³ placing the vascular endothelium in the key unifying position in the shared pathogenesis of CVD and diabetes. Although there is prospective evidence linking endothelial dysfunction with incident CVD,³⁴ previous prospective data linking endothelial dysfunction to risk of diabetes have been indirect and inconsistent.^15- 17 In this study, we demonstrated that elevated plasma levels of molecular biomarkers of endothelial dysfunction were predictors of incident diabetes in a large cohort of initially nondiabetic women. Elevated levels of E-selectin, ICAM-1, and VCAM-1 raised the relative risk of diabetes by 1.5- to 7.5-fold. Associations of E-selectin and ICAM-1 with diabetes were independent of the confounding effects of usual risk factors for diabetes, and E-selectin remained a powerful predictor even after the potentially mediating effects of obesity, inflammation, and levels of insulin and hemoglobin A_1c were accounted for.

Subclinical tissue injury and increased fat mass elevate blood levels of inflammatory cytokines, especially tumor necrosis factor α and interleukin 6, which stimulate an acute-phase response marked by elevated levels of CRP.³⁵ Localization of this inflammatory cascade by vascular endothelial cells is mediated by CAMs. Their surface expression is a common endothelial response to a variety of toxic stimuli, and their shedding into the systemic circulation provides evidence of endothelial cell activation and endothelial dysfunction.²⁶ E-selectin, expressed exclusively by endothelial cells, is absent in inactive cells but is rapidly induced by inflammatory cytokines. ICAM-1 and VCAM-1 are expressed by endothelial cells and leukocytes in response to inflammatory cytokines, elevated levels of free fatty acids, oxidized low-density lipoprotein cholesterol, and advanced glycosylation end products occurring in diabetes.³⁶ These CAMs facilitate leukocyte rolling, adhesion, and transmigration into the subendothelial space, key steps in the early formation of atherosclerotic plaque. Thus, elevated levels of CAMs represent an early marker of several pathogenic processes associated with endothelial dysfunction and the potential development of CVD and type 2 diabetes.

Impaired endothelial function in large arterial beds is a key step in the pathogenesis of CVD, and elevated levels of CAMs predict the incident development of CVD events independent of standard CVD risk factors.³⁷ Endothelial dysfunction in the capillary and arteriolar microcirculation is a plausible key step in the evolution of tissue insulin resistance and may act by at least 2 mechanisms. Impaired endothelium-dependent vasodilatation and vasomotion may limit insulin-mediated capillary recruitment and microvascular redistribution of skeletal muscle blood flow from nonnutritive to nutritive flow routes, diminishing insulin delivery to metabolically active, insulin-sensitive muscle tissue.^10,38- 40 Also, altered endothelial permeability impairs insulin delivery to the interstitium. In the dynamic state, interstitial insulin levels appear to be a rate-limiting step determining insulin effectiveness.⁴¹ In addition to vascular insulin resistance, endothelial dysfunction is associated with impaired fibrinolysis, microalbuminuria, impaired function of insulin-sensitive lipoprotein lipase (contributing to a dyslipidemia with high triglyceride and low high-density cholesterol levels), impaired fibrinolysis, and impaired endothelium-dependent nitric-oxide–mediated vasodilatation (contributing to hypertension), all features of the insulin resistance syndrome.^9- 10 Recent experimental evidence in humans has placed endothelial dysfunction in a direct link between blood pressure and insulin sensitivity.⁴² The results of our analysis extend observations linking endothelial dysfunction to CVD and its risk factors, strengthening support for the hypothesis that endothelial dysfunction is also an important precursor of type 2 diabetes.

Results of our study extend other evidence linking endothelial dysfunction to insulin resistance and risk of diabetes. Mice with endothelial dysfunction by virtue of targeted knockout mutations in the endothelium-dependent nitric oxide synthase gene are insulin resistant and display features of the insulin resistance syndrome.⁴³ Endothelial dysfunction is a consistent finding in cross-sectional studies of established diabetes.^{11- 12,44- 45} Endothelial dysfunction and elevated levels of E-selectin, ICAM-1, and VCAM-1 have also been described in cross-sectional studies of nondiabetic subjects at increased risk of diabetes, including subjects with impaired glucose tolerance, and in nondiabetic first-degree relatives of patients with type 2 diabetes.^12- 14,46 Three previous epidemiologic studies have examined risk of incident diabetes. In the Atherosclerosis Risk in Communities (ARIC) Study, microcirculatory dysfunction, represented by retinal arteriolar narrowing, increased risk for diabetes by 71%, comparing subjects in the highest quartile of narrowing with those in the lowest quartile.¹⁶ Also in ARIC, levels of von Willebrand factor, another endothelial biomarker, was not an independent predictor of incident diabetes after BMI was accounted for, although coagulation factor VIII was an obesity-independent predictor of diabetes in women.¹⁶ In a study of 71 Pima Indians, elevated levels of several biomarkers, including E-selectin and ICAM-1, were not independently associated with incident diabetes.¹⁷ Treatment with drugs having beneficial effects on endothelial function also supports an association with insulin resistance and risk of diabetes. Insulin sensitization with troglitazone or metformin in type 2 diabetes lowers E-selectin levels, improves endothelium-dependent vasodilatation, and, when used in patients without diabetes, reduces risk of developing diabetes.^3,47- 50 Therapy with statins⁵¹ or angiotensin-converting enzyme (ACE) inhibitors⁵² improves insulin sensitivity and endothelial dysfunction and, in secondary analyses of CVD prevention studies, has been associated with an approximately 30% reduced risk of incident diabetes.^53- 54 However, whether amelioration of endothelial dysfunction by using statins or ACE-based therapies will actually prevent type 2 diabetes remains to be confirmed in clinical trials.

Our study has several limitations. We assessed endothelial dysfunction solely by means of plasma biomarkers and do not have other measures of vascular physiology. However, although expression of CAMs represents a specific, early stage in the evolution of endothelial dysfunction, plasma levels have modest correlations with directly assessed endothelial function.^12,27- 29 Our comprehensive assessment of diabetes risk factors allowed statistical control for important confounding factors in the pathogenesis of diabetes, but residual confounding could remain in the analysis. In particular, we did not measure levels of free fatty acids, atherogenic lipoproteins, or adipokines (for instance, adiponectin), known mediators of endothelial dysfunction and insulin resistance. Also, we measured only plasma insulin levels and did not have direct measurements of insulin resistance. Although plasma insulin levels in the physiologic range can induce insulin resistance in humans,⁵⁵ statistical control for insulin levels attenuated but did not eliminate associations of E-selectin with incident diabetes. In addition, we were able to control for obesity only by using BMI or waist circumference because we did not have more precise measures of body fat percentage or fat distribution. However, for levels of E-selectin at least, the association with diabetes remained strong even after extensive covariate adjustment. Although residual confounding by incompletely measured or unmeasured physiologic covariates may exist in our models, it seems unlikely that more complete statistical adjustment would completely eliminate the observed associations.

In addition, although hyperglycemia may lead to endothelial dysfunction,⁵⁶ our prospective study design and similar effects after subsidiary analyses excluding cases diagnosed during the first 4 years of follow-up or excluding women with hemoglobin A_1c levels higher than 6.5% support the conclusion that endothelial dysfunction preceded development of diabetic hyperglycemia. Further, although undetected subclinical atherosclerosis (which occurs in prediabetes and is associated with elevated levels of CAMs) might explain the observed associations, removal of data for women with incident CVD during follow-up did not alter the results. These findings indicate that baseline endothelial dysfunction was not a consequence of undetected, coincident diabetes or atherosclerosis but was an early abnormality clearly preceding the subsequent development of type 2 diabetes. Finally, the analysis was conducted in relatively healthy white women aged 30 to 55 years (range, 41-67 years at baseline). Results may not be generalizable to women of other ages or ethnic backgrounds or to men.

In conclusion, we found that elevated plasma levels of biomarkers reflecting endothelial dysfunction were powerful independent predictors of type 2 diabetes in initially healthy women. Our findings support the hypothesis that endothelial dysfunction may be a common pathogenic precursor to CVD and type 2 diabetes, placing endothelial dysfunction as a fundamental abnormality in the insulin resistance syndrome. These findings may have important implications for the prevention and treatment of type 2 diabetes. Therapies that improve endothelial dysfunction may prove important in the treatment of insulin resistance and in strategies to slow the accelerating worldwide epidemic of type 2 diabetes and its costly, morbid complications.