In summary, patients with type 2 diabetes exhibited impaired endothelium-dependent vasodilation in vivo, elevated serum triglycerides, decreased LDL size, and normal antioxidant capacity. Of these parameters, LDL size was significantly correlated with endothelial function.

Abstract-A total of 75 in vivo endothelial function tests (intrabrachial artery infusions of endothelium-dependent[acetylcholine] and -independent [sodium nitroprusside] vasoactive agents) were performed in 18 type 2 diabetic patients (aged 58Ϯ2 years, body mass index 28.5Ϯ0.6 kg/m 2 , and fasting plasma glucose 229Ϯ11 mg/dL) and 27 matched normal subjects. These tests were performed before and 6 months after combination therapy with insulin and metformin and before and 6 months after metformin therapy only. Before insulin therapy, blood flow responses to acetylcholine (15 g/min) were significantly blunted in type 2 diabetic patients (7.5Ϯ0.7 mL ⅐ dL Ϫ1 ⅐ min Ϫ1) compared with normal subjects (11.6Ϯ0.9 mL ⅐ dL Ϫ1 ⅐ min Ϫ1 , PϽ0.01). During insulin therapy, the acetylcholine response increased by 44% to 10.8Ϯ1.6 mL ⅐ dL Ϫ1 ⅐ min Ϫ1 (PϽ0.05). Insulin therapy also significantly increased the blood flow responses to both low and high doses of sodium nitroprusside. We conclude that insulin therapy improves endothelium-dependent and -independent vasodilatation. These data support the idea that insulin therapy has beneficial rather than harmful effects on vascular function.

. S.M. and A.S. contributed equally to this article. Abbreviations: DCCT, Diabetes Control and Complications Trial; E/I ratio, ratio of length of R-R intervals during expiration to inspiration; ETDRS, Early Treatment of Diabetic Retinopathy Study; HF, high frequency; HFnorm, normalized high frequency component; HRV, heart rate variability; LF, low frequency; LFnorm, normalized low frequency component; RMSSD, square root of the mean square of R-R interval differences; TP, total power; UAER, urinary albumin excretion rate; VLF, very low frequency.A OBJECTIVE -Frequency domain analysis of heart rate variability (HRV) is used to assess cardiovascular autonomic function. There are no prospective data on the sensitivity of its various components to glycemia or other diabetes-related risk factors compared with conventional tests and with other complications of diabetes. RESEARCH DESIGN AND METHODS -In 1985, possible risk factors of future complications were determined in 115 children with type 1 diabetes. In 1996, the presence of complications (HRV analysis, conventional tests of autonomic function, urinary albumin excretion rate [UAER], and retinopathy) were assessed in 83 of these patients (age 32 ± 1 years, duration of diabetes 22 ± 1 years). RESULTS -Poor glycemic control (measured as lifetime glycemic exposure or HbA 1c in 1985)was the most important independent predictor of decreases in all measures of absolute power of HRV (total power [TP] and very low frequency, low frequency [LF], and high frequency [HF] power) and square root of the mean square of R-R interval differences but not of changes of normalized measures or ratios (normalized HF and LF, LF/HF). Other significant independent predictors of autonomic dysfunction were late age of onset of diabetes, female sex, and high BMI. To examine the sensitivity of the various tests to glycemia, the patients were divided into tertiles based on lifetime glycemic exposure (A 1c months). Glycemic exposure in the tertiles averaged 194 ± 25 A 1c months (20 years of HbA 1c 0.8% above normal), 556 ± 19 A 1c months(20 years of HbA 1c 2.3% above normal), and 963 ± 30 A 1c months (20 years of HbA 1c 4% above normal). Tests of complications that were significantly abnormal in patients already in the lowest tertile and were correlated with glycemia were TP and severity of retinopathy. Of conventional tests, only the ratio of length of R-R intervals during expiration to inspiration (E/I ratio) was significantly related to glycemic exposure, but it required high glycemic exposure (20 years of HbA 1c 4% above normal) to be abnormal. UAER was significantly increased only in the highest tertile of glycemic exposure.CONCLUSIONS -TP and retinopathy score were much more sensitive to antecedent glycemia than conventional tests of autonomic function or UAER and were significantly abnor-

Abstract-Normal insulin action in vivo involves a decrease in stiffness of large arteries (a decrease in aortic pressure augmentation). We determined whether the ability of insulin to decrease arterial stiffness is altered in uncomplicated type 1 diabetes. Nine type 1 diabetic men (age 28Ϯ2 years, body mass index 24Ϯ1 kg/m 2 ) and 9 matched normal men were studied under normoglycemic hyperinsulinemic (sequential 2-hour insulin infusions of 1 [step 1] and 2 [step 2] mU ⅐ kg Ϫ1 ⅐ min Ϫ1) conditions. Central aortic pressure waveforms were synthesized from those recorded in periphery with applanation tonometry on the radial artery and a validated reverse transfer function to construct the central aortic pressure wave every 30 minutes. This allowed the determination of aortic augmentation (the pressure difference between the first and the second systolic peaks) and the augmentation index (augmentation divided by pulse pressure), as the measure of stiffness of large arteries. Whole-body glucose uptake was 44% (step 1) and 37% (step 2) lower (PϽ0.001) in the diabetic patients than in the normal subjects. At baseline, before the insulin infusion, augmentation averaged 0Ϯ1 and 2Ϯ1 mm Hg (NS) and the augmentation index was Ϫ1.5Ϯ4.5% and 4.0Ϯ3.7% (NS) in the normal and diabetic subjects, respectively. After 1 hour of hyperinsulinemia, the augmentation index had decreased significantly (PϽ0.01) to Ϫ9.5Ϯ4.8% in the normal subjects but remained at 4.4Ϯ4.2% in the diabetic patients. A significant decrease was not observed in the diabetic patients until 150 minutes (Ϫ1.2Ϯ4.1%, PϽ0.05 versus baseline). Whole-body glucose uptake was significantly inversely correlated with the change in the augmentation index during step 1 (rϭϪ0.61, PϽ0.01). Insulin resistance in type 1 diabetes involves a defect in the ability of insulin to decrease central aortic pressure. This defect could predispose these patients to premature stiffening of large arteries.

These data are consistent with idea that altered neurotransmission is an important determinant of vascular reactivity of diabetic blood vessels to nitrovasodilators in vivo.