0021-972X/98/$03.00/0
Journal of Clinical Endocrinology and Metabolism
Copyright © 1998 by The Endocrine Society
Vol. 83, No. 6
Printed in U.S.A.
Specific Impairment of Endothelium-Dependent
Vasodilation in Subjects with Type 2 Diabetes
Independent of Obesity*
ROBERT V. HOGIKYAN, ANDRZEJ T. GALECKI, BERTRAM PITT,
JEFFREY B. HALTER, DOUGLAS A. GREENE, AND MARK A. SUPIANO
Divisions of Geriatric Medicine and Endocrinology and Metabolism, Department of Internal Medicine,
and Institute of Gerontology, University of Michigan, and Michigan Diabetes Research and Training
Center, and Geriatric Research, Education, and Clinical Center, Department of Veterans Affairs
Medical Center, Ann Arbor, Michigan 48105
ABSTRACT
In subjects with type 2 diabetes in whom an impaired response
to an endothelial-dependent vasodilator has been characterized,
the populations have also been at least moderately obese. Obesity
has been characterized as an independent predictor of endothelial
dysfunction in nondiabetic subjects. We hypothesized that in normotensive subjects with type 2 diabetes compared with agematched control subjects, 1) endothelium-dependent vasodilation,
as demonstrated by the forearm blood flow (FABF) response to
intraarterial acetylcholine, would be decreased; 2) endotheliumindependent vasodilation, as demonstrated by the FABF response
to intraarterial nitroprusside, would be similar; 3) the degree of
insulin resistance, as measured by the insulin sensitivity index
(SI), would predict greater impairment in the FABF response to
acetylcholine; and 4) these relationships would be independent of
obesity. We measured FABF by venous occlusion plethysmography
during brachial arterial infusions of the endothelium-dependent
vasodilator acetylcholine and the endothelium-independent vaso-
T
HE CONTROL of blood flow and regulation of vascular tone involves the coordinated integration of
many systems. One component of this complex system
involves endothelial cell production of and vascular
smooth muscle response to nitric oxide, a potent vasodilator. Endothelial function has been characterized as impaired in human models of hypertension (1), hypercholesterolemia (2), aging (3), and some, but not all, groups of
subjects with insulin-dependent diabetes mellitus (4 –9) or
type 2 diabetes (10 –15). However, interpretation of findings in subjects with type 2 diabetes is complicated by
coexisting conditions, such as obesity, hypertension, hyReceived December 17, 1997. Revision received March 16, 1998. Accepted March 19, 1998.
Address all correspondence and requests for reprints to: Robert V.
Hogikyan, M.D., Geriatric Research, Education, and Clinical Center
(11G), Department of Veterans Affairs Medical Center, Ann Arbor,
Michigan 48105. E-mail: [email protected].
* Portions of this work were presented at the National Meeting of the
American Diabetes Association in Atlanta, GA, June 1995. This work was
supported in part by NIH Grants RR-00042 to the University of Michigan
General Clinical Research Center, AG-08808 to the Claude D. Pepper
Older Americans Independence Center, and DK-20572 to the Michigan
Diabetes Research and Training Center at the University of Michigan;
the Medical Research Service of the Department of Veterans Affairs; and
a grant from the John A. Hartford Foundation.
dilator nitroprusside in 20 control and 17 subjects with type 2
diabetes. We measured SI using the frequently sampled iv glucose
tolerance test. Among the diabetic relative to the control subjects
we identified a decrease in the acetylcholine-mediated percent
increase in FABF (P 5 0.02). Using the absolute FABF response to
acetylcholine and including adjustments for body mass index and
other covariates, the overall group difference remained and was
noted to be greatest in those subjects who had lower baseline
FABFs. In contrast, no significant difference in the nitroprussidemediated increase in the percent change FABF was identified between groups (P 5 0.30). Finally, the degree of insulin resistance,
as measured by SI, did not independently predict greater impairment of the FABF response to acetylcholine. This study is the first
to identify specific endothelial cell dysfunction that remains significant after adjustment for obesity in a population of normotensive subjects with type 2 diabetes. (J Clin Endocrinol Metab 83:
1946 –1952, 1998)
perlipidemia, and insulin resistance. Indeed, Steinberg et
al. identified similar impairments in endothelial-dependent vasodilation among obese subjects with type 2 diabetes and nondiabetic subjects with obesity and also identified an independent association of insulin resistance with
impaired endothelium-dependent vasodilation (12). These
findings suggested the possibility that endothelial dysfunction in obese type 2 diabetic subjects may be more
related to coexisting obesity than to diabetes.
The current study characterizes a group of normotensive subjects with type 2 diabetes with regard to endothelial cell function and insulin sensitivity. We used the
forearm blood flow (FABF) response to acetylcholine as an
index of endothelium-mediated vasodilation and the
FABF response to nitroprusside as an endothelium-independent measure of vasodilation. We controlled for obesity both by limiting the range of ideal body weight to less
than 150% and by incorporating body mass index (BMI)
and percent body fat into the statistical model used to
analyze the FABF responses, as the diabetic subjects had
an increased BMI relative to the control subjects. We also
tested the hypothesis that those subjects with the lowest
tissue sensitivity to the metabolic effects of insulin would
have the greatest impairment in the FABF response to
intraarterial acetylcholine.
1946
ENDOTHELIAL DYSFUNCTION IN TYPE 2 DIABETIC SUBJECTS
Subjects and Methods
Subjects
Twenty control subjects and 17 subjects with type 2 diabetes in otherwise good general health were recruited through newspaper advertisement and through the Human Subjects Core of the Geriatrics Center
at the University of Michigan. Subjects were screened before study entry
with a medical history, physical examination, and laboratory tests, including a complete blood count and routine chemistries. Subjects were
classified as type 2 diabetes by their primary care providers. All had
adult onset of their diabetes and no history of ketoacidosis. Glycosylated
hemoglobin values of the subjects with diabetes ranged from 6.5–14.7%
(normal, 4 – 8%). Body composition was estimated by bioelectric impedance using an RJL instrument (model BIA-103 B, RJL Systems, Mt.
Clemens, MI).
Subjects were excluded from participation if they exceeded 150% of
ideal body weight (Metropolitan Life Insurance Tables, 1983), had a
history of hypertension or a resting seated blood pressure greater than
160 mm Hg systolic or greater than 90 mm Hg diastolic, or had evidence
from history, physical exam, or laboratory results of other significant
underlying illness. Subjects did not have clinical signs of peripheral
vascular occlusive disease. Subjects were excluded if they were taking
a calcium channel blocker, an angiotensin-converting enzyme inhibitor,
or a b-blocker. Three of the subjects with diabetes were receiving replacement therapy for hypothyroidism with a TSH level that was in the
normal range. Diabetes treatment regimens included diet alone (n 5 4),
a sulfonylurea (n 5 6), insulin (n 5 5), a sulfonylurea and insulin (n 5
1), and metformin (n 5 1). With regard to known duration of diabetes,
four subjects were newly diagnosed, five reported diabetes duration as
5 yr or less, three reported diabetes duration as 10 yr or less, and four
reported diabetes duration as more than 10 yr. Twelve of the 20 control
subjects had no personal or family history of diabetes. Eight control
subjects underwent standard 75-g oral glucose tolerance tests because of
a history of incidental hyperglycemia or a family history of diabetes
mellitus. Subjects were excluded from the control group if their oral
glucose tolerance test was positive for diabetes mellitus by National
Diabetes Data Group criteria. Four of the control subjects met National
Diabetes Data Group criteria for impaired glucose tolerance. All subjects
gave written informed consent that was approved by the University of
Michigan human use committee.
1947
After the FABF measurement at the 6.4-mg dose, the acetylcholine infusion was stopped.
After a washout period lasting a minimum of 10 min, repeat measurement to establish a stable baseline FABF was carried out as described
above. To determine the effect of an endothelium-independent vasodilator on FABF, nitroprusside (Nitropress, Abbott Laboratories, North
Chicago, IL) was diluted in 5% dextrose to achieve stepwise increasing
infusion doses of 0.01, 0.04, 0.16, and 0.64 mg/dL FAVzmin. The nitroprusside was protected from light at all times. After the FABF measurement at the 0.64-mg dose, the nitroprusside infusion was stopped.
Frequently sampled iv glucose tolerance test (FSIVGTT)
On the day after the FABF protocol, after another overnight fast, the
FSIVGTT was carried out in all subjects. A FSIVGTT was performed on
the second day of study as described by Bergman (18), with the addition
of tolbutamide in control subjects and insulin in subjects with diabetes
to enhance the precision of the estimates of insulin action (19, 20). Each
medication was given as a bolus over approximately 25 s. This method
has been found to yield estimates of insulin sensitivity that are reproducible (14% mean coefficient of intraindividual variation) (21) and are
comparable to those obtained using the glucose clamp method (22).
Subjects were studied in the supine position. An iv catheter was placed
in the antecubital vein of one arm for the injection of glucose and
tolbutamide or insulin. Another catheter was inserted in a retrograde
manner into a dorsal hand vein of the contralateral arm, which was then
placed into a warming box heated to 60 C to obtain arterialized venous
blood samples for glucose and insulin determinations (19). Catheters
were kept patent by a slow infusion of 0.45% saline (,50 mL/h). Beginning 20 min after the iv lines were inserted, three baseline blood
samples were obtained, and blood pressure and heart rate were measured at 5-min intervals. Fifty percent glucose (300 mg/kg) was then
given as an iv push over 30 s. Blood samples (3 mL) were collected 2,
3, 4, 5, 6, 8, 10, 12, 14, 16, 19, 22, 23, 24, 25, 27, 30, 40, 50, 60, 70, 80, 90,
100, 120, 140, 160, and 180 min after the glucose bolus. Tolbutamide (137
mg/m2 body surface area) was given to control subjects iv 20 min after
the glucose injection to further stimulate insulin secretion. Insulin (0.05
U/kg) was given to the subjects with diabetes iv 20 min after the glucose
injection.
Study protocol
Analytical methods
All subjects reported to the General Clinical Research Center of the
University of Michigan Hospitals at 0730 h on each of the 2 days of study.
They were instructed to fast from 2200 h the night before each of the 2
study days and to abstain from smoking for 12 h before the studies.
Subjects were studied in the supine position in a quiet room maintained
at a constant temperature of 23–25 C. In subjects with diabetes, oral
hypoglycemics were discontinued 3 days before the study, and insulin
was withheld beginning with the afternoon dose on the day before the
study.
Mean arterial pressure (MAP) was determined from the electronically
integrated area under the intraarterial blood pressure curve from the
Marquette telemetry system (Marquette Electronics Series 7700, Marquette Electronics, Milwaukee, WI) just before each FABF measurement.
Baseline values reported from the FSIVGTT represent the mean of
three measurements before glucose administration for each variable.
Blood samples for plasma glucose and insulin were collected into chilled
glass tubes containing sodium heparin, stored on ice, and separated
immediately after each study. Plasma was stored at 270 C until assay.
Plasma glucose was measured by the Autoanalyzer (Sigma, St. Louis,
MO) glucose oxidase method, and plasma insulin was determined by
RIA in the Core Laboratory of the Michigan Diabetes Research and
Training Center. The percentage of glycosylated hemoglobin was determined at the Core Laboratory of the University of Michigan Diabetes
Research and Training Center using the Isolab Glyc-Affin GHb test kit
(Isolab, Akron, OH). Serum lipid levels were determined by the Clinical
Laboratory of the University of Michigan Health System using the
VITROS 950IRC analyzer (Johnson & Johnson Clinical Diagnostics,
Rochester, NY).
The insulin sensitivity index (SI) and a measure of glucose effectiveness (SG) were calculated from a least squares fitting of the temporal
pattern of glucose and insulin throughout the FSIVGTT using the MINMOD program (18). The acute insulin response to iv glucose (AIRG) was
calculated as the mean rise in plasma insulin above baseline 3, 4, and 5
min after iv glucose administration. KG, a measure of glucose tolerance,
is the rate of plasma glucose disappearance calculated as the least square
slope of the natural logarithm of the absolute glucose concentration
between 5–20 min after the glucose bolus (a normal value for KG is
.1%/min).
FABF protocol
Forearm volume (FAV) was measured using water displacement (7,
16). A 20-gauge 1.25-in. Angiocath catheter (Becton Dickinson Vascular
Access, Sandy, UT) was placed into the brachial artery of the nondominant arm. The catheter was connected to a pressure transducer (model
1290A quartz transducer, Hewlett-Packard, Andover, MA). One of the
basic electrocardiograph limb leads was monitored. FABF was measured using venous occlusion plethysmography during an intraarterial
infusion protocol we have previously described (17). Studies were performed at least 110 min after arterial catheter placement. To establish a
stable baseline, FABF was measured periodically until three consecutive
readings over approximately 10 min were similar. To determine the
effect of intraarterial infusions of acetylcholine on FABF, acetylcholine
(acetylcholine HCl, Sigma Chemical Co., St. Louis, MO.) was diluted in
5% dextrose to achieve stepwise increasing infusion doses of 0.1, 0.4, 1.6,
and 6.4 mg/dL FAVzmin. Each acetylcholine dose was administered by
an infusion pump (model 970T, Harvard Apparatus, South Natick, MA)
for 4 min before recording FABF during the fifth minute of each infusion.
1948
JCE & M • 1998
Vol 83 • No 6
HOGIKYAN ET AL.
Data and statistical analysis
Values are presented as the mean 6 se. P , 0.05 is considered
statistically significant. Statistical analysis was performed using StatView 4.5 (Abacus Concepts, Berkeley, CA) and SAS/PROC MIXED (23).
Baseline characteristics and results of the FSIVGTT were compared
between study groups using Student’s t tests. The homogeneity of the
gender distribution in both groups was tested using x2 analysis. To
describe the FABF responses to different doses of acetylcholine and
nitroprusside, two distinct analyses were performed. The FABF data
were analyzed as the percent change from baseline FABF comparing
groups, not adjusting for covariates, by repeated measures ANOVA. To
describe the absolute FABF response data to different doses of acetylcholine and nitroprusside, separate linear mixed effects (LME) models
were developed. A LME model (24) is an extension of the classical
analysis of repeated measures methods. Compared to standard repeated
measures analysis, LME allows for data missing at random, time-dependent covariates, and modeling of a covariance structure.
To develop the best LME models for acetylcholine and nitroprusside
responses, several models were considered for each infusate. To accommodate a skewed distribution of FABF responses and to stabilize the
variance, logarithmic transformation was applied to all FABF measurements, including those at baseline and during responses to each dose of
acetylcholine and nitroprusside. Each of the models takes into account
that several measures were repeated on the same individual. To describe
between-subject random variation and correlation of individual measurements, the optimal covariance structure was selected using Akaike’s
information criterion (25). Several independent variables, listed in Tables 1 and 2, and their interactions with stimulus dose and with study
group were considered in each model selection process. In choosing the
optimal model, the likelihood ratio test was used. To test the significance
of specific terms (e.g. study group effect) in the model, likelihood statistics were compared for two nested models with and without the terms
being tested (e.g. with and without group effect). To obtain P values, the
likelihood ratio test statistic was referred to the x2 distribution. The
number of degrees of freedom for the x2 distribution was determined by
subtracting the number of parameters in the models that were being
compared. To check the goodness of fit of the models developed, residuals were inspected by plotting them against predicted values and
against available covariates (not presented).
Results
Subject characteristics
The characteristics of the control subjects and those with
type 2 diabetes are compared in Table 1. The groups were
similar with respect to age, gender mix, MAP, and cholesterol
level. Although BMI was significantly greater in the diabetic
subjects than in the control subjects, there was no significant
difference in percent body fat between groups. The subjects
with type 2 diabetes differed from control subjects in all
parameters of glucose metabolism measured during the
FSIVGTT (Table 2). Fasting serum glucose (pre-FABF, Table
1; during the FSIVGTT, Table 2) and insulin levels and glycosylated hemoglobin percentages were higher among the
subjects with diabetes. Plasma insulin levels from one subject
with diabetes could not be determined due to the presence
of insulin antibodies. SI, SG, AIRG, and KG were all significantly less in the subjects with diabetes compared to those in
the control subjects.
FABF during vasodilatory intraarterial infusions
Acetylcholine. Baseline FABF was significantly higher in the
diabetic subjects (Table 1). A plot representing the mean
percent increase in FABF from baseline (to correct for the
baseline group difference in FABF) at each of the four intraarterial infusion doses of acetylcholine for control and
TABLE 1. Characteristics of type 2 diabetic and control subjects
Control
(n 5 20)
Age (yr)
Gender (M/F)
Body mass index (kg/m2)
Body fat (%)
Mean arterial pressure
(mm Hg)
Glycosylated hemoglobin
(%)
Baseline FABF (mL/
100 mL FAVzmin.)
Total cholesterol (mg/dL)
HDL cholesterol (mg/dL)
Fasting glucose (mg/dL)
Type 2
(n 5 17)
P value
57 6 3
12/5
28.4 6 0.9
27 6 2
97 6 2
0.28
0.33
,0.01
0.48
0.34
5.6 6 0.09
10.3 6 0.7
,0.01
3.4 6 0.2
4.7 6 0.5
0.02
193 6 8
43 6 3
98 6 2
195 6 10
36 6 3
169 6 12
0.87
0.14
,0.01
53 6 2
11/9
25.3 6 0.6
26 6 2
94 6 2
TABLE 2. Frequently sampled iv glucose tolerance test results
for type 2 diabetic and control subjects
Control
(n 5 20)
Fasting glucose (mg/dL)
Fasting insulin (mU/mL)
Sensitivity to insulin SI
(31024 min21/mU/mL)
Glucose effectiveness
(min21)
Acute insulin response
(mU/mL)
KG (%/min)
Type 2
(n 5 16)
P value
95.7 6 1.4
11.6 6 0.8
3.41 6 0.46
200 6 19
26.9 6 4.7
0.63 6 0.14
,0.01
,0.01
,0.01
0.19 6 0.001
0.15 6 0.001
,0.01
81.8 6 8.9
6.15 6 2.4
,0.01
2.03 6 0.16
1.17 6 0.09
,0.01
diabetic subjects is presented in Fig. 1. By repeated measures
ANOVA, the diabetic subjects demonstrated less percent
increase in FABF in response to intraarterial acetylcholine
than control subjects (by ANOVA, P 5 0.02). A linear mixed
effects model was developed using absolute FABF to estimate the influence of parameters of interest and to test hypotheses regarding FABF responses. To adjust for each subject’s baseline FABF, the baseline blood flow value was
entered into the model. The model revealed that the overall
difference in FABF response to acetylcholine between study
groups remained significant after adjusting for covariates (x2
5 16.25; df 5 8; P 5 0.04). FABF responses to individual
acetylcholine doses were significantly less in the type 2 diabetes group at the 0.4 (P 5 0.04) and 6.4 (P 5 0.02) mg/dL
FAVzmin acetylcholine doses.
The interaction among acetylcholine dose, study group,
and baseline FABF was also included in the model to take
into account the different slopes (P 5 0.001) of the FABF
responses during infusion of each of the four doses of acetylcholine in relationship to baseline FABF values in the
control and diabetic subject groups. This analysis indicates
the importance of including baseline FABF in the best-fit
model. The analyses performed for different covariates revealed that in the best-fit model the FABF response to acetylcholine, in addition to depending on the study group,
depends on acetylcholine dose (P 5 0.04) and positively on
subject age (P 5 0.04), but not on BMI, percent body fat,
gender, MAP, fasting insulin, plasma glucose, glycosylated
hemoglobin, diabetes duration, total cholesterol, or high density lipoprotein cholesterol. Specific to our second hypothesis, the FABF response to acetylcholine did not depend on
ENDOTHELIAL DYSFUNCTION IN TYPE 2 DIABETIC SUBJECTS
FIG. 1. Group mean data (mean 6 SE) for percent change in FABF
from baseline in response to intraarterial infusions of acetylcholine in
control (C) and type 2 diabetes mellitus subjects (by ANOVA, P 5
0.02).
1949
flow value was entered into the model as a covariate. The
model revealed that there was no overall significant difference between study groups (x2 5 9.01; df 5 8; P 5 0.34).
The interaction between nitroprusside dose and baseline
FABF was also included in the model to take into account the
different slopes (P 5 0.001) of the FABF responses at all four
doses of nitroprusside in relationship to baseline values in
the control and diabetic subject groups, indicating the importance of including baseline FABF in the best-fit model.
The analysis performed for different covariates revealed that
in the best-fit model, the FABF response to nitroprusside
depends on the nitroprusside dose, varies negatively with
subject age (P 5 0.02), and varies negatively with basal insulin level (P 5 0.008). The effects of BMI, percent body fat,
gender, MAP, plasma glucose, glycosylated hemoglobin, diabetes duration, total cholesterol, high density lipoprotein
cholesterol, SI, KG, AIRG, and SG were not significant; consequently, corresponding variables were not included in the
model.
Subset analysis
FIG. 2. Group mean data (mean 6 SE) for percent change in FABF
from baseline in response to intraarterial infusions of nitroprusside
in control (C) and type 2 diabetes mellitus subjects (by ANOVA, P 5
0.30).
the parameters of glucose metabolism, SI, KG, AIRG, or SG.
We also tested the relationship between the maximum FABF
response to acetylcholine and SI across both groups analogous to the analysis carried out by Steinberg et al. (12). In this
analysis, univariate linear regression reveals a positive relationship between the maximum FABF response to acetylcholine and the SI across both groups (r 5 0.364; P 5 0.03).
Nitroprusside. FABF before nitroprusside infusion was similar in the subjects with type 2 diabetes and the control
subjects (control, 5.19 6 0.33; type 2, 5.33 6 0.35 mL/dL
FAVzmin; P 5 0.8). A plot representing the mean percent
increase in FABF from baseline (to control for any baseline
variation in FABF) at each of the four intraarterial infusion
doses of nitroprusside for control and diabetic subjects is
presented in Fig. 2. By repeated measures ANOVA, the diabetic subjects demonstrated similar percent increases in
FABF in response to intraarterial nitroprusside as control
subjects (by ANOVA, P 5 0.30). A second linear mixed effects
model was developed using absolute FABF to estimate the
influence of parameters of interest and to test hypotheses
pertinent to FABF responses to different nitroprusside doses.
To adjust for each subject’s baseline FABF, the baseline blood
A secondary analysis was also performed to assess
whether endothelial dysfunction was present in type 2 diabetes when subject groups were more closely matched for
BMI. Subjects were eliminated from analysis who had BMI
values either greater than 30.0 kg/m2 or 21.0 kg/m2 or less.
This eliminated three control subjects and five diabetic subjects. In this subset, there was no significant difference in BMI
between subject groups (control, 25.5 6 0.6; type 2, 26.6 6 0.8
kg/m2; P 5 0.24). The findings in this subgroup analysis of
percent change in FABF responses to acetylcholine and nitroprusside were not altered from those of the larger group
(acetylcholine, P 5 0.03; nitroprusside, P 5 0.90).
Discussion
In this study we found that in normotensive subjects with
type 2 diabetes compared with age-matched control subjects:
1) endothelium-dependent vasodilation, as demonstrated by
the FABF response to intraarterial acetylcholine, is decreased; 2) endothelium-independent vasodilation, as demonstrated by the FABF response to intraarterial nitroprusside, is similar; 3) the degree of insulin resistance, as
measured by SI, does not predict greater impairment in the
FABF response to acetylcholine; and 4) these relationships
appear to be independent of obesity. The specific nature of
the impairment in response to acetylcholine is supported by
the lack of impairment in the increase in FABF in response
to the direct acting exogenous nitric oxide donor nitroprusside in type 2 diabetic compared with control subjects. We
also found that baseline FABF is an important determinant
of the response to both vasodilators and, therefore, was controlled for in the analysis of these responses.
In this study, we used intraarterial infusions of acetylcholine and nitroprusside to investigate differences between
control subjects and subjects with type 2 diabetes in their
FABF responses. Because intraarterial acetylcholine and nitroprusside caused dose-dependent increases in FABF, we
were able to quantitate sensitivity to these vasodilators. One
advantage of the brachial arterial infusion methodology is
1950
HOGIKYAN ET AL.
that FABF is varied over a 4- to 5-fold range without producing major systemic effects in heart rate or blood pressure
that could confound interpretation of the results, as has been
demonstrated in other studies (10).
Our finding of vasomotor dysfunction in subjects with
diabetes is in general agreement with previous studies of
vascular endothelial function in diabetic animals (26, 27) and
in humans with insulin-dependent diabetes mellitus (4 –9)
and type 2 diabetes (10 –12, 14, 15). Although the results from
these studies overall suggest that there is endothelial dysfunction in animals and humans with diabetes, the findings
have not been unanimous. In animal models of diabetes,
there is good evidence for specific endothelial dysfunction
(26, 27). In humans with insulin-dependent diabetes mellitus,
there is evidence for specific endothelial dysfunction (8), but
also evidence of intact endothelium-dependent vasodilation in response to metacholine and endothelium-independent vasodilation in response to nitroprusside (4). Investigations in humans have also tried to identify factors
such as glycemic control, cholesterol level, body composition, and age that may account for some of the variability
found in responses to both endothelium-dependent and
-independent vasodilators.
In humans with type 2 diabetes, three previous studies
have used venous occlusion plethysmography to define endothelium-dependent vasodilation during brachial arterial
infusions of acetylcholine (10, 13) and metacholine (11) in
middle-aged (10) and younger (11, 13) subjects with diabetes.
Two of these studies identified impaired responses in the
diabetic subjects to both the endothelium-dependent and
-independent vasodilators (10, 11), and one identified intact
endothelial function (13). When the responses to NG-monomethyl-l-arginine (10) and verapamil (11) were taken into
account, the studies that reported impaired endotheliumdependent responses concluded that multiple mechanisms
are involved in the vascular impairment characterized, inclusive of endothelial dysfunction. In the study by Avogaro
et al. (13), in which no differences in either endotheliumdependent or -independent responses were identified, a narrower dose range of acetylcholine was used, achieving lower
peak vasodilatory responses, which may have contributed to
their negative finding.
In the three studies discussed above, there was a range of
obesity among the subjects with diabetes. Subjects in the
study by Avogaro et al. were matched for BMI and were only
moderately obese, with mean BMIs of 27 and 29 kg/m2 in the
control and diabetic groups, respectively. In the study by
McVeigh et al. (10), the diabetic subjects had a higher mean
BMI than the control subjects, but as this difference was not
statistically significant, the FABF analyses did not take into
account BMI. In the study by Williams et al. (11) to address
the possible role of obesity, a subgroup analysis of a nonobese subset of the diabetic subjects was performed that demonstrated an impairment in endothelium-dependent and
-independent vasodilation.
Steinberg et al. studied two very obese (mean BMI, 36
kg/m2) young subject groups, one with and one without type
2 diabetes, using thermal dilution blood flow measurement
in the lower extremity (12). Among these obese subjects with
and without diabetes, there was an attenuated increase in leg
JCE & M • 1998
Vol 83 • No 6
blood flow in response to metacholine and no difference in
the blood flow response to intrafemoral artery nitroprusside
compared to the response in nonobese control subjects. However, no significant difference was observed in the vasodilator response to metacholine between the obese subjects
with and without diabetes. Results from the current study are
in agreement with the findings of Steinberg et al. of specific
endothelial dysfunction in subjects with type 2 diabetes.
However, they identified obesity as an important predictor
of endothelial dysfunction in type 2 diabetic subjects. The
current study was designed to test endothelial function in
normotensive type 2 diabetic subjects with mild to moderate
obesity. Our primary model analysis demonstrated that body
composition, as measured in this study, did not significantly
influence the finding of specific endothelial dysfunction
among the diabetic subjects. We also carried out a secondary
analysis, similar to that used in the study by Williams et al.
(11), which demonstrated specific endothelial dysfunction
when the subject groups were matched for BMI. Possible
explanations for the difference in the effect of obesity between our study and that of Steinberg et al. include there
being a threshold of obesity that needs to be surpassed before
it becomes an independent predictor of endothelial dysfunction and possible differences due to the choice of extremity
where endothelial function was tested. Additionally, measures of the distribution of body fat, particularly central
adiposity rather than total body fat, may have shown a relationship with the response to acetylcholine.
To better understand modulators of endothelial function
in addition to obesity, given the heterogeneity of the population of patients with type 2 diabetes as a whole, a number
of characteristics of the subjects in this study that were not
controlled for in the study design were also examined as
covariates. As described in Results, although most of the
potential covariates tested did not predict responses to the
intraarterial vasodilators infused, subject age did, both
within each group and when evaluating the combined
groups. Other studies have identified impaired endothelium-dependent vasodilation with increasing age in healthy
subjects (3, 28), whereas in obese subjects with type 2 diabetes, no change was demonstrated in endothelium-independent vasodilation with subject age (12). In the present
study, after identifying the best-fit model, age correlated
positively with acetylcholine response and negatively with
nitroprusside response. Steinberg et al. suggested that the
negative correlation between acetylcholine response and aging identified in earlier studies may be the result of the
increase in body fat with age (12). Controlling for covariates,
including obesity, in identifying a best-fit model for the current study may account for the difference in relationship
between age and the FABF response to acetylcholine identified compared to the relationship observed in other studies.
The other covariate that entered into the model of FABF
responses to both vasodilators is the baseline FABF. In a
study of 30 healthy young subjects, Chowienczyk et al. found
that FABF responses to acetylcholine, but not to nitroprusside, were positively correlated with baseline blood flow (29).
They suggest that baseline FABF is an important factor that
needs to be taken into account when interpreting the FABF
response to infusions of acetylcholine and nitroprusside. In
ENDOTHELIAL DYSFUNCTION IN TYPE 2 DIABETIC SUBJECTS
the present study, we found that FABF responses to both
acetylcholine and nitroprusside were directly related to baseline blood flow in each subject group except at the highest
dose of acetylcholine, at which this relationship only held for
the subjects with diabetes. As discussed earlier, the group
difference in FABF response to acetylcholine was most apparent in subjects with the lowest baseline FABF.
It has been suggested by Steinberg et al. that the endothelial
cell may be a target organ of insulin action with respect to the
nitric oxide system (12), as there is now evidence to suggest
that insulin exerts its vasodilating effect at least in part
through the release of nitric oxide (28, 30). As subjects with
type 2 diabetes are an insulin-resistant group, the current
study also sought to test whether subjects with the lowest SI
had the poorest responses to acetylcholine. Steinberg et al.
found that the extent of insulin resistance was correlated
with increasing impairment of endothelium-dependent vasodilation. In the current study, a similar comparison between the maximum FABF response to acetylcholine and SI
across both groups, using univariate linear regression, shows
a borderline significant positive relationship. However, the
narrow range of SI values among our diabetic subjects limits
the usefulness of such an analysis within the diabetes group
alone. It is also possible that we did not identify a relationship
between the FABF response and the sensitivity to insulin due
to the methods we used to characterize the sensitivity to
insulin. As has been shown previously, SI may differ depending on whether tolbutamide or insulin is used, the dose
of insulin, and whether the insulin is given as an infusion or
a bolus (31–34). To the extent that insulin has been shown to
at least in part exert its vasodilating effect through nitric
oxide release, it is also possible that had we measured vasomotor insulin sensitivity (blood flow response to insulin
infusion) we might have identified a relationship between
blood flow response to insulin and SI (35–38). The best-fit
linear mixed effects models for the acetylcholine and the
nitroprusside dose-response data failed to identify a significant correlation between any of the parameters of glucose
metabolism and the FABF response to either acetylcholine or
nitroprusside. These results held both within each group and
when evaluating the combined groups. A similar analysis of
the FABF response to nitroprusside did identify a negative
correlation with the baseline insulin level. However, the significance of this finding is unclear.
We acknowledge that our study design presents some
limitations in the interpretation of the results. In the analysis
of FABF responses to vasoactive infusates, four subjects with
impaired glucose tolerance were included in the control
group. This is a conservative approach that might tend to
diminish differences between control and diabetic subject
groups. When these subjects with impaired glucose tolerance
are excluded from the analysis, the same differences between
groups are present. With two control and one diabetic subject
receiving estrogen supplementation, and the study groups
statistically not different with regard to gender, we believe
that the subject groups are adequately matched for the potential of an estrogen effect (39), although this could not be
incorporated into the LME model. Two of the control and
none of the diabetic subjects were cigarette smokers at the
time of study, which again would tend to make it more
1951
difficult to detect a decrement in the FABF response to acetylcholine among the diabetic subjects (40). Eight of the diabetic subjects were taking an oral hypoglycemic agent, and
it is possible that there may have been a residual effect even
after it had been withheld for 3 days. More specifically, the
one subject taking metformin may have experienced increased blood flow, as has been characterized previously
(41). Peripheral vascular occlusive disease was not clinically
present in these subjects. However, the presence of subclinical peripheral vascular occlusive disease in the diabetic subject group may have contributed to the difference in FABF
responses to acetylcholine between groups. Finally, the baseline FABF before nitroprusside infusion was higher than the
preacetylcholine FABF. Although this could indicate incomplete return to baseline, which might limit our ability to
interpret the FABF response to nitroprusside, chance alone
is another possible explanation.
In summary, we have identified evidence for a specific
impairment of endothelium-dependent vasodilation in a
population of normotensive subjects with type 2 diabetes.
Obesity did not have a statistically significant influence on
this result. Baseline FABF was identified as an important
predictor of FABF responses to acetylcholine- and nitroprusside-mediated vasodilation and so was controlled for in the
data analysis. Finally, in this population, after controlling for
other variables, a LME model failed to identify a correlation
between parameters of glucose metabolism including SI and
the FABF response to acetylcholine. In conclusion, this is the
first study to identify specific endothelial cell dysfunction
that remains significant after adjustment for obesity in a
population of normotensive subjects with type 2 diabetes.
Acknowledgments
We thank Marla Smith and Eric Leiendecker for their technical assistance, and the nursing staff of the University of Michigan General
Clinical Research Center for their care of our subjects during this study.
Tolbutamide was a gift from Upjohn (Kalamazoo, MI).
References
1. Panza JA, Quyyumi Jr AA, Brush JE, Epstein SE. 1990 Abnormal endothelium-dependent vascular relaxation in patients with essential hypertension.
N Engl J Med. 323:22–27.
2. Creager MA, Cooke JP, Mendelsohn ME, et al. 1990 Impaired vasodilation of
forearm resistance vessels in hypercholesterolemic humans. J Clin Invest.
86:228 –234.
3. Gerhard M, Roddy M-A, Creager SJ, Creager MA. 1996 Aging progressively
impairs endothelium-dependent vasodilation in forearm resistance vessels of
humans. Hypertension. 27:849 – 853.
4. Elliott TG, Cockcroft JR, Groop P-H, Viberti GC, Ritter JM. 1993 Inhibition
of nitric oxide synthesis in forearm vasculature of insulin-dependent diabetic
patients: blunted vasoconstriction in patients with microalbuminuria. Clin Sci.
85:687– 693.
5. Smits P, Kapma J-A, Jacobs M-C, Lutterman J, Thien T. 1993 Endotheliumdependent vascular relaxation in patients with type I diabetes. Diabetes.
42:148 –153.
6. Calver A, Collier J, Vallance P. 1992 Inhibition and stimulation of nitric oxide
synthesis in the human forearm arterial bed of patients with insulin-dependent
diabetes. J Clin Invest. 90:2548 –2554.
7. Halkin A, Benjamin N, Doktor HS, Todd SD, Viberti G, Ritter JM. 1991
Vascular responsiveness and cation exchange in insulin-dependent diabetes.
Clin Sci. 81:223–232.
8. Johnstone MT, Creager SJ, Scales KM, Cusco JA, Lee BK, Creager MA. 1993
Impaired endothelium-dependent vasodilation in patients with insulindependent diabetes mellitus. Circulation. 88:2510 –2516.
9. O’Driscoll G, Green D, Rankin J, Stanton K, Taylor R. 1997 Improvement in
endothelial function by angiotensin converting enzyme inhibition in insulindependent diabetes mellitus. J Clin Invest. 100:678 – 684.
1952
HOGIKYAN ET AL.
10. McVeigh GE, Brennan GM, Johnston GD, et al. 1992 Impaired endotheliumdependent and independent vasodilation in patients with type 2 (non-insulindependent) diabetes mellitus. Diabetologia. 35:771–776.
11. Williams SB, Cusco JA, Roddy M-A, Johnstone MT, Creager MA. 1996
Impaired nitric oxide-mediated vasodilation in patients with non-insulindependent diabetes mellitus. J Am Coll Cardiol. 27:567–574.
12. Steinberg HO, Chaker H, Leaming R, Johnson A, Brechtel G, Baron AD. 1996
Obesity/insulin resistance is associated with endothelial dysfunction. J Clin
Invest. 97:2601–2610.
13. Avogaro A, Piarulli F, Valerio A, et al. 1997 Forearm nitric oxide balance,
vascular relaxation, and glucose metabolism in NIDDM patients. Diabetes.
46:1040 –1046.
14. Morris SJ, Shore AC, Tooke JE. 1995 Responses of the skin microcirculation
to acetylcholine and sodium nitroprusside in patients with NIDDM. Diabetologia. 38:1337–1344.
15. Pitei DL, Watkins PJ, Edmonds ME. 1997 NO-dependent smooth muscle
vasodilatation is reduced in NIDDM patients with peripheral sensory neuropathy. Diabet Med. 14:284 –290.
16. Parving HH. 1976 Increased microvascular permeability to plasma proteins in
short- and long-term juvenile diabetics. Diabetes. 25:884 – 889.
17. Hogikyan RV, Supiano MA. 1993 Homologous up-regulation of arterial
a-adrenergic responsiveness compensates for decreased sympathetic nervous
system tone. J Clin Invest. 91:1429 –1435.
18. Bergman RN. 1989 Toward physiological understanding of glucose tolerance:
minimal model approach. Diabetes. 38:1512–1527.
19. Foster HV, Dempsey J, Thompson E, Virduk E, DoPico GA. 1972 Estimation
of arterial PO2, PCO2, pH, and lactate from arterialized venous samples. J Appl
Physiol. 41:565–573.
20. Finegood DT, Hramiak IM, Dupre J. 1990 A modified protocol for estimation
of insulin sensitivity with the minimal model of glucose kinetics in patients
with insulin-dependent diabetes. J Clin Endocrinol Metab. 70:1538 –1549.
21. Ferrari P, Alleman Y, Shaw S, Riesen W, Weidmann P. 1991 Reproducibility
of insulin sensitivity measured by the minimal model method. Diabetologia.
34:527–530.
22. Bergman RN, Prager R, Volund A, Olefsky JM. 1987 Equivalence of the
insulin sensitivity index in man derived by the minimal model method and the
euglycemic glucose clamp. J Clin Invest. 79:790 – 800.
23. SAS Institute. 1996 SAS/STAT software: changes and enhancements through
release 6.11. Cary: SAS Institute.
24. Lindstrom MJ, Bates DM. 1988 Newton-Raphson and EM algorithms for
linear mixed-effects models for repeated-measures data. J Am Stat Assoc.
83:1014 –1022.
25. Akaike H. 1973 Information theory and an extension of the maximum likelihood principle. In: Petrov EBN, Csaki F, eds. Second International Symposium on Information Theory and Control. Budapest: Akademia Kiado;
267–281.
JCE & M • 1998
Vol 83 • No 6
26. Oyama Y, Kawasaki H, Hattori Y, Kanno M. 1986 Attenuation of endothelium-dependent relaxation in aorta from diabetic rats. Eur J Pharmacol.
131:75–78.
27. Meraji S, Jayakody L, Senaratne MPJ, Thomson ABR, Kappagoda T. 1987
Endothelium-dependent relaxation in aorta of BB rat. Diabetes. 36:978 –981.
28. Taddei S, Virdis A, Mattei P, et al. 1995 Aging and endothelial function in
normotensive subjects and patients with essential hypertension. Circulation.
91:1981–1987.
29. Simonson DC. 1988 Insulin sensitivity and the effects of antihypertensive
agents: implications for the treatment of hypertension in the patient with
diabetes mellitus. Postgrad Med J. 64:39 – 47.
30. Steinberg HO, Brechtel G, Johnson A, Fineberg N, Baron AD. 1994 Insulinmediated skeletal muscle vasodilation is nitric oxide dependent. J Clin Invest.
94:1172–1179.
31. Saad MF, Steil GM, Kades WW, et al. 1997 Differences between the tolbutamide-boosted and the insulin-modified minimal model protocols. Diabetes.
46:1167–1171.
32. Prigeon RL, Roder ME, Porte Jr D, Kahn SE. 1996 The effects of insulin dose
on the measurement of insulin sensitivity by the minimal model technique.
J Clin Invest. 97:501–507.
33. Alzaid AA, Dinneen SF, Turk DJ, Caumo A, Cobelli C, Rizza RA. 1994
Assessment of insulin action and glucose effectiveness in diabetic and nondiabetic humans. J Clin Invest. 94:2341–2348.
34. Welch S, Gebhart SSP, Bergman RN, Phillips LS. 1990 Minimal model analysis of intravenous glucose tolerance test-derived insulin sensitivity in diabetic
subjects. J Clin Endocrinol Metab. 71:1508 –1518.
35. Tack CJJ, Smits P, Willemsen JJ, Lenders JWM, Thien T, Lutterman JA. 1996
Effects of insulin on vascular tone and sympathetic nervous system in NIDDM.
Diabetes. 45:15–22.
36. Nolan JJ, Ludvik B, Baloga J, Reichart D, Olefsky JM. 1997 Mechanisms of
the kinetic defect in insulin action in obesity and NIDDM. Diabetes.
46:994 –1000.
37. Dela F, Larsen JJ, Mikines KJ, Galbo H. 1995 Normal effect of insulin to
stimulate leg blood flow in NIDDM. Diabetes. 44:221–226.
38. Dela F, Larsen JJ, Mikines KJ, Ploug T, Petersen LN, Galbo H. 1995 Insulinstimulated muscle glucose clearance in patients with NIDDM. Diabetes.
44:1010 –1020.
39. Taddei S, Virdis A, Ghiadoni L, et al. 1996 Menopause is associated with
endothelial dysfunction in women. Hypertension. 28:576 –582.
40. Celermajer DS, Adams MR, Clarkson P, et al. 1996 Passive smoking and
impaired endothelium-dependent arterial dilatation in healthy young adults.
N Engl J Med. 334:150 –154.
41. Jansson PAE, Andersson OK, Gudbjornsdoftir HS, Lonnroth PN. 1996 The
effect of metformin on adipose tissue metabolism and peripheral blood flow
in subjects with NIDDM. Diabetes Care. 19:160 –164.
Antiprogestins in the Next Millennium
Jerusalem, Israel
September 1–3, 1999
This is a joint meeting of Shaare Zedek Medical Center and The Weizmann Institute of Science, Israel.
For further information, please contact: Irving M. Spitz, M.D., D.Sc., Antiprogestins in the Next Millennium,
Institute of Hormone Research, Shaare Zedek Medical Center, P.O. Box 3235, Jerusalem, 91031 Israel.
Telephone: 1972-2-655-5188; Fax: 1972-2-652-2018; E-mail: [email protected]; Internet: http://
www.weizmann.ac.il/Biological–Regulation/antiprog/