Insulin resistance and intima–media thickness in the carotid and femoral arteries of clinically healthy 58-year-old men. The Atherosclerosis and Insulin Resistance Study (AIR)


Björn Fagerberg MD PhD, Department of Medicine, Sahlgrenska University Hospital, 413 45 Gothenburg, Sweden (fax: + 46 31 88 57 47; e-mail:


Abstract. Bokemark L, Wikstrand J, Attvall S, Hulthe J, Wedel H, Fagerberg B (Sahlgrenska University Hospital, and the Nordic School of Public Health, Göteborg, Sweden). Insulin resistance and intima–media thickness in the carotid and femoral arteries of clinically healthy 58-year-old men. The Atherosclerosis and Insulin Resistance Study (AIR). J Intern Med 2001; 249: 59–67.

Objective. To examine whether insulin resistance was associated with ultrasound-assessed measures of atherosclerosis in men with varying degrees of obesity.

Design. A random selection of subjects from the general population were divided into quintiles of a body mass index/blood glucose score that was shown to be a valid and reproducible index of the degree of insulin sensitivity as assessed by the clamp technique. Every fourth man in quintiles 1 and 5 and every 20th man in quintiles 2–4 (in total, 104 men) were selected for an ultrasound examination of the carotid and femoral arteries and a euglycaemic hyperinsulinaemic clamp examination, adjusted for fat-free mass.

Setting. A university hospital.

Subjects. A total of 104 clinically healthy 58-year-old men of Swedish ancestry.

Results. The mean common carotid artery intima–media thickness (IMT), but not the common femoral IMT, correlated significantly with glucose infusion rate (GIR) (r = − 0.20, P < 0.05), systolic blood pressure (r = 0.20, P < 0.05), pulse pressure (r = 0.23 P < 0.01), heart rate (r = 0.20, P < 0.05), HDL cholesterol (r = − 0.18, P < 0.05), log triglycerides (r = 0.28, P < 0.01), apoA1 (r = − 0.20, P < 0.05), apoB (r = 0.21, P < 0.05), LDL particle size (r = − 0.22, P < 0.05) and plasma insulin (r = 0.20, P < 0.05). In a multiple regression, common carotid IMT was independently associated with log triglycerides (β = 0.25, P = 0.012) and pulse pressure (β = 0.21, P = 0.031) (R2 = 8.7%, P = 0.005)

Conclusions. Insulin sensitivity, measured with the gold standard euglycaemic hyperinsulinaemic clamp method, showed similar associations with ultrasound-assessed measures of atherosclerosis in the carotid arteries as established cardiovascular risk factors, but only triglycerides and pulse pressure contributed independently to the variability in the common carotid intima–media thickness.


Insulin resistance is characterized by an abnormally low response of target cells to insulin and is associated with a clustering of risk factors for cardiovascular disease [1]. Obesity is a rapidly increasing health problem in modern society and is also the strongest determinant of the degree of insulin sensitivity [2]. However, conclusive evidence that insulin resistance is associated with atherosclerotic disease is still lacking.

The association between the degree of insulin resistance and atherosclerotic disease has so far been examined in a number of cross-sectional studies in different populations, using varying techniques. A few small studies have used the euglycaemic hyperinsulinaemic clamp technique. Laakso et al.[3] found that insulin resistance was more common amongst subjects with ultrasound-assessed plaques in the carotid or femoral arteries. We performed a pilot study in hypertensive and normal men and were the first to report that insulin-mediated glucose uptake was inversely related to intima–media thickness (IMT) in the common carotid artery [4]. Bressler et al.[5] found that subjects with angiographically documented coronary artery disease were more insulin-resistant than a disease-free control group. The IRAS study, a large, tri-ethnic multicentre study in the USA, used the minimal model technique to assess insulin sensitivity and measured the intima–media thickness in the carotid artery with ultrasound in a mixed population of healthy subjects and patients treated for hypertension or diabetes [6]. The results showed that insulin resistance was associated with atherosclerosis amongst Hispanics and non-Hispanic whites, but not blacks. Three other smaller studies have used the minimal model or other methods to measure insulin sensitivity and ultrasound techniques to assess vascular morphology in normal subjects, diabetics or hypertensives [7–9]. The results were inconsistent, as there was either a varying degree of association between insulin resistance and carotid artery IMT [7,8], or no association at all with plaques in the femoral artery [9]. Most of these studies have also included subjects taking different types of medicine, including antihypertensive agents which are known to affect insulin sensitivity, thereby confounding the relationship to atherosclerosis [10].

So far, no study has examined the relationship between insulin sensitivity measured using the gold standard technique, i.e. the euglycaemic hyperinsulinaemic clamp method, and assessed IMT in both the carotid and femoral arteries in a population-representative sample. Accordingly, the present study was initiated with the objective of testing the hypothesis that insulin resistance is associated with subclinical atherosclerosis amongst clinically healthy 58-year-old men in the general population with varying degrees of obesity.

Materials and methods


The inclusion criteria were age 58 years, male sex and Swedish ancestry. Exclusion criteria were cardiovascular or other clinically overt disease, treatment with cardiovascular drugs that might disturb the measurements performed in the study or unwillingness to participate. The subjects were randomly selected amongst men in the County Council register and were invited to a screening examination. Invitation letters were sent to 1728 men, of whom 1432 replied and 852 underwent the screening examination.

The study outline is presented in Fig. 1. The goal was to screen about 800 subjects, who, in connection with the screening examination, were divided into quintiles of a body mass index (BMI)/blood glucose score, which allowed immediate stratification and selection for further studies, including the present one, which is a substudy of a larger study of all subjects. The following equation was used:

Figure 1.

Study outline.

inline image

This algorithm was based on a previous study of clinically healthy men of similar age who had undergone a euglycaemic hyperinsulinaemic clamp examination [4]. The correlation coefficient between the BMI/blood glucose score and observed insulin sensitivity was 0.81. In the present population sample, this score correlated significantly with insulin sensitivity measured with the euglycaemic hyperinsulinaemic clamp method, when expressed as insulin-mediated glucose uptake adjusted both for body weight and for fat-free mass (FFM) (r = 0.69, P < 0.001, and r = 0.59, P < 0.001, respectively, n = 104). In a reproducibility study of 60-year-old men obtained from the background population (n = 32) examined at an interval of 2 weeks, this score showed a coefficient of variation of 3.4% and a correlation coefficient of r = 0.99 between the two examinations.

A random sample of every fourth man in quintile 1 (indicating low sensitivity) and quintile 5 (indicating high insulin sensitivity) and every 20th man in quintiles 2–4 was invited to an ultrasound examination and a euglycaemic hyperinsulinaemic clamp examination. The result of the latter examination showed that the insulin-mediated glucose uptake was 4.9 ± 2.7, 9.7 ± 2.4 and 9.9 ± 2.8 mg kg−1 FFM min−1, respectively, for those in the score-based quintiles 1, 2–4 and 5 (P < 0.001 for trend).

The sample size calculation was based on the results of a previous pilot study of 23 clinically healthy men selected from the general population [4]. The correlation coefficient between common carotid IMT and insulin-mediated glucose uptake as assessed by the euglycaemic hyperinsulinaemic clamp technique was −0.59 (P < 0.01). A more cautious approach showed that, by studying 100 subjects, it would be possible to establish a significant correlation at the 5% level if the correlation coefficient was at least 0.20.

The subjects received both written and oral information before they gave their consent to participate. The study was approved by the Ethics Committee at Sahlgrenska University Hospital.


Established questionnaires were used to evaluate history of previous disease, current disease and smoking habits [11]. The total number of smoking years was multiplied by the number of cigarettes smoked daily and the product was termed ‘cigarette-years’. Body weight was measured on a balance scale with the subject dressed in underwear. Blood pressure was measured twice when the subject had been resting in the supine position for 5 min using the appropriate cuff size in relation to arm size [12]. Diastolic blood pressure was determined as Korotkoff phase V. A 12-lead standard electrocardiogram was recorded. Heart rate was recorded from the electrocardiogram. Whole blood glucose was measured with the glucose oxidase technique. Blood samples were drawn, and serum and plasma were frozen in aliquots at − 70°C within 4 h.

Euglycaemic hyperinsulinaemic clamp   The euglycaemic hyperinsulinaemic clamp examination was performed ad modum DeFronzo et al., slightly modified according to a previous description [13]. After a priming dose, the insulin infusion rate was 1 mU min−1 kg−1 body weight, continuing for 120 min until the end of the examination. During the clamp, the target whole blood glucose concentration was 5 mmol L−1 and the glucose infusion rate was adjusted in connection with each determination of whole blood glucose if necessary. After the clamp examination, FFM was measured using the dual-energy X-ray absorptiometry body composition model (Lunar DPX-L) [14]. Insulin sensitivity was calculated as the glucose infusion rate per minute adjusted for FFM (GIRFFM) during the final hour of the examination [13]. A reproducibility study of this variable has shown a coefficient of variation of 14.7% and a correlation coefficient of r = 0.70 between repeated examinations [13].

Ultrasound examination   The visit for the ultrasound examinations always took place in the morning. The men were asked to fast overnight and to refrain from smoking on the morning of the visit. After supine rest, blood was drawn and plasma frozen for later assessment of insulin peptides. The ultrasound examination was performed with an ultrasound scanner (Acuson 128) with a 7-MHz linear transducer with an aperture of 38 mm. The electrocardiographic signal (lead II) was simultaneously recorded to synchronize the image capture to the top of the R wave to minimize variability during the cardiac cycle. Both the left and right carotid arteries were scanned at the level of the bifurcation, and images for IMT measurements were recorded from the far wall in the common carotid artery and the carotid artery bulb. The right femoral artery was examined distal to the inguinal ligament at the site where the artery divides into the superficial femoral artery and the profound femoral artery. It was not possible to obtain readable images in a few subjects, presented below in more detail.

Intima–media thickness   IMT was defined as the distance from the leading edge of the lumen–intima interface to the leading edge of the media–adventitia interface of the far wall (10 mm segment in the carotid artery, 15 mm in the femoral artery).

The images were measured in an automated analysing system, based on automatic detection of the echo structures in the ultrasound image, but with the option of making manual corrections by the operator [15,16].

Biochemical methods   At screening, plasma insulin was determined in all subjects with a radioimmunoassay (Pharmacia Insulin RIA, Pharmacia Diagnostics, Uppsala, Sweden). Serum concentrations of total cholesterol, HDL cholesterol and triglycerides were assessed as previously described [17–19].

LDL particle size was analysed with gradient gel electrophoresis and measured with a computerized method developed at the Wallenberg Laboratory for Cardiovascular Research in cooperation with Chalmers University of Technology [20]. The coefficient of variation is 0.3%, with a correlation coefficient of r = 0.99 with the same serum run on different gels on different days [20].

Statistical methods

All statistics were analysed using SPSS for Windows 7.5 (Chicago, IL). The results are presented as mean values, standard deviations and numbers (%). Skewed variables were log transformed. ANOVA, Dunnet's test (post-hoc analyses) and t-test were used for comparison of continuous variables and the χ2 test for categorical data. A non-parametric Spearman's rank correlations test was used in the analysis, with the relationships illustrated using Pearson's correlation coefficient. Partial correlation coefficients and forward, stepwise multiple regression were used in the analyses examining the association between the studied variables. Two-sided P < 0.05 was considered statistically significant.


The eligible subjects (n = 818) had the following characteristics: BMI, 26.2 ± 3.3 kg m−2; waist:hip ratio, 0.94 ± 0.06; systolic and diastolic blood pressures, 136 ± 19 and 82 ± 10 mmHg, respectively; plasma insulin, 9.38 ± 5.21 mU L−1; serum cholesterol, 6.04 ± 1.09 mmol L−1; HDL cholesterol, 1.27 ± 0.35 mmol L−1; LDL cholesterol, 4.07 ± 0.97 mmol L−1; and serum triglycerides, 1.57 ± 0.89 mmol L−1. The characteristics of the group of subjects who were selected for the ultrasound and euglycaemic hyperinsulinaemic clamp examinations (n = 104, Table 1) were very similar to the entire group of subjects who underwent the screening examination (n = 818).

Table 1.  Characteristics of the subjects as a group and divided into tertiles of common carotid intima–media thickness (IMT)
Tertiles of common carotid artery IMT
 n   =  104Lowest
n = 34
n = 35
n = 34
  • a

    P-value refers to difference between tertiles (ANOVA). Only P-values < 0.05 are given.

  • *

    P  < 0.05 tertile 1–2;

  • P  < 0.05 tertile 1–2, 1–3;

  • P  < 0.05 tertile 2–3;

  • §

    P  < 0.01 tertile 1–3;

  • P  < 0.01 tertile 2–3;

  • * *

    P  < 0.05 tertile 1–3, 2–3;

  • † †

    P  < 0.05 tertile 1–3.

Weight (kg)82.8 ± 14.679.1 ± 13.687.0 ± 15.382.2 ± 14.1 ≤ 0.20
Height (cm)178 ± 6178 ± 6179 ± 6177 ± 4 ≤ 0.20
Body mass index (kg m−2)26.4 ± 4.424.9 ± 4.027.0 ± 4.426.4 ± 4.6 ≤ 0.20
Total body fat (kg)21.6 ± 8.420.0 ± 9.422.9 ± 7.721.7 ± 7.8> 0.20
Waist circumference (cm)95.6 ± 12.291.5 ± 11.998.8 ± 12.196.3 ± 11.70.040*
Waist:hip ratio0.94 ± 0.070.91 ± 0.070.96 ± 0.060.96 ± 0.080.007
Blood pressure
 Systolic (mmHg)125 ± 19119 ± 19130 ± 20127 ± 180.048
 Diastolic (mmHg)78 ± 1076 ± 979 ± 1177 ± 9> 0.20
Pulse pressure (mmHg)48 ± 1543 ± 1451 ± 1349 ± 17 ≤ 0.20
Heart rate (beats min−1)62 ± 860 ± 963 ± 863 ± 7 ≤ 0.20
Serum cholesterol
 Total (mmol L−1)6.02 ± 1.105.74 ± 0.945.97 ± 1.136.35 ± 1.16 ≤ 0.20
 HDL (mmol L−1)1.26 ± 0.361.29 ± 0.351.34 ± 0.401.12 ± 0.280.029
 LDL (mmol L−1)4.09 ± 1.033.85 ± 0.984.01 ± 0.954.43 ± 1.10 ≤ 0.20
Serum triglycerides (mmol L−1)1.57 ± 1.101.32 ± 0.741.38 ± 0.592.01 ± 1.600.003§
ApoA1 (g L−1)1.42 ± 0.241.44 ± 0.241.49 ± 0.261.32 ± 0.180.008
ApoB (g L−1)1.21 ± 0.271.14 ± 0.271.17 ± 0.261.32 ± 0.250.011**
LDL peak particle size (nm)26.3 ± 0.726.4 ± 0.626.5 ± 0.726.0 ± 0.80.015‡
Blood glucose (mmol L−1)4.9 ± 0.54.8 ± 0.55.0 ± 0.45.0 ± 0.7> 0.20
Plasma insulin (uU L−1)9.32 ± 5.068.08 ± 3.899.41 ± 5.4510.48 ± 5.51 ≤ 0.20
GIRFFM (mg kg−1 min−1)8.30 ± 3.29.3 ± 3.28.2 ± 3.17.4 ± 3.10.044††
Smoking status
 Cigarette-years299 ± 379214 ± 289328 ± 351351 ± 476> 0.20
 Never (n, [%])38 [37]15 [44]10 [29]13 [37]> 0.20
 Previous (n, [%])42 [41]11 [32]15 [43]16 [47]> 0.20
 Current (n, [%])23 [22]8 [24]10 [29]5 [15]> 0.20

Ultrasound measurements and cardiovascular risk factors

Table 2 gives the results of the ultrasound examinations in all patients and in each tertile of common carotid IMT. The subjects in the highest tertile of common carotid IMT had thicker mean IMT in the carotid bulb and thicker IMT in the femoral artery in comparison with the subjects in the remaining tertiles.

Table 2.  Results of the ultrasound measurements in the entire group and in those divided into tertiles of common carotid intima–media thickness (IMT)
Tertiles of common carotid IMT
 n   =  104Lowest
n = 34
n = 35
n = 34
  1. LD, lumen diameter. aP-value refers to difference between tertiles (ANOVA). Only P-values < 0.05 are given. bNot done, as this variable is a selection criterion or closely related. *P < 0.01 tertile 1,2–3; **P < 0.05 tertile 1–3.

Carotid artery
 IMT, mean communis (mm) (n = 103)0.80 ± 0.120.67 ± 0.050.80 ± 0.030.94 ± 0.06Not doneb
 IMT, mean bulb (mm) (n = 101)1.03 ± 0.320.87 ± 0.121.04 ± 0.261.18 ± 0.450.000*
 LD, mean communis (mm) (n = 103)6.25 ± 0.696.08 ± 0.586.46 ± 0.726.20 ± 0.73 ≤ 0.20
Common femoral artery
 IMT, mean femoralis (mm) (n = 101)1.07 ± 0.460.92 ± 0.391.08 ± 0.431.20 ± 0.540.047**
 LD, mean femoralis (mm) (n = 95)8.90 ± 1.199.11 ± 1.209.03 ± 1.218.55 ± 1.10 ≤ 0.20

As shown in Table 1, the highest tertile of common carotid IMT was associated with low serum concentrations of HDL cholesterol and apoA1, high levels of circulating apoB and triglycerides, and a low glucose infusion rate.

In corresponding analyses, the highest tertile of carotid bulb IMT was associated with high serum concentrations of cholesterol (P = 0.028), LDL cholesterol (P = 0.035), and apoB (P = 0.032), and the highest tertile of femoral artery IMT was related to high serum triglyceride levels (P = 0.049) and cigarette-years (P = 0.001) (data not shown).

Glucose infusion rate and ultrasound measures

Table 3 and Fig. 2 demonstrate that, of all ultrasound-assessed measures of the carotid and femoral arteries, only the common carotid IMT correlated significantly with insulin-mediated glucose uptake. There was no relationship between the femoral artery IMT and glucose infusion rate (data not shown).

Table 3.  Correlations between glucose infusion rate adjusted for fat-free mass and ultrasound variables
  1. IMT, intima–media thickness; LD, lumen diameter. *P < 0.05.

Carotid artery
 IMT, mean communis (mm) (n = 103) − 0.20*
 IMT, mean bulb (mm) (n = 101) − 0.01
 LD, mean communis (mm) (n = 103) − 0.14
Common femoral artery
 IMT, mean (mm) (n = 101)0.02
 LD, mean (mm) (n = 95)0.06
Figure 2.

Scattergram of the relationship between common carotid intima–media thickness (IMT) and glucose infusion rate.

Table 4 demonstrates that glucose infusion rate and a number of variables, including serum triglycerides (Fig. 3), were associated with common carotid IMT. In a stepwise multiple regression equation with common carotid IMT as dependent variable and the statistically significant variables in Table 4 as independent variables, only log serum triglycerides (β= 0.25, P = 0.012) and pulse pressure (β = 0.21, P = 0.031) contributed to the variability in IMT (R2 = 8.7%, P = 0.005).

Table 4.  Correlations between common carotid intima–media thickness and variables associated with the metabolic syndrome
  • *

    P  < 0.05,

  • * *

    P  < 0.01.

Waist circumference0.12
Systolic blood pressure0.20*
Pulse pressure0.23**
Heart rate0.20*
HDL cholesterol − 0.18*
ApoA1 − 0.20*
LDL peak particle size − 0.22*
Plasma insulin0.20*
Glucose infusion rate − 0.20*
Figure 3.

Scattergram of the relationship between common carotid intima–media thickness (IMT) and log serum triglycerides.

Triglycerides correlated significantly with BMI (r = 0.44, P < 0.001), waist:hip ratio (r = 0.58, P < 0.001), serum insulin (r = 0.50, P < 0.001), glucose infusion rate (r = − 0.43, P < 0.001), LDL particle size (r = − 0.71, P < 0.001), HDL cholesterol (r = − 0.52, P < 0.001), and ApoA1 (r = − 0.33, P < 0.01).


The results of the present study showed that insulin sensitivity was inversely associated with the common carotid IMT. This association was of the same magnitude as that observed for established cardiovascular risk factors such as blood pressure [21] or apoB [21], although it was not an independent relationship.

To our knowledge, the present study is the first one to examine the relationship between IMT in several vascular beds prone to atherosclerosis and insulin sensitivity in a population-based sample, utilizing the gold standard method for assessing insulin sensitivity, i.e. the euglycaemic hyperinsulinaemic clamp technique. In comparison with previous studies, we have also kept a number of potentially confounding factors constant. It is known that race, age, cultural factors, concurrent medication and current disease may affect insulin sensitivity [10,22–24]. Hence, we studied only white men all of the same age (58 years) and cultural background, and excluded all subjects with ongoing cardiovascular medication or current disease. Although we excluded subjects with treated hypertension and clinical diabetes mellitus, 20% of the subjects had a resting systolic blood pressure above 140 mmHg and 34% of the subjects were insulin-resistant according to a suggestion based on 18 previous studies [25]. This is a feasible way to study the atherosclerotic disease mechanisms, although one must keep in mind that the results can be inferred only to the background population.

The euglycaemic hyperinsulinaemic clamp method was used to assess insulin sensitivity measured as the glucose infusion rate adjusted for FFM. It is important to adjust for FFM, as the often-used procedure of adjusting only for body weight underestimates insulin sensitivity amongst obese subjects [13,26]. The euglycaemic hyperinsulinaemic clamp method is associated with a variability of 12–15% in subjects randomly selected from the general population [13,27].

The ultrasound method was based on a strictly documented protocol with automated reading and low variability [15]. We found that common carotid IMT was the only ultrasound-assessed measure that was associated with glucose infusion rate. It may be argued that IMT is a composite measure and that a thick intima–media complex may not only be caused by atherosclerosis. However, we observed that a large common carotid IMT was associated with other characteristics of atherosclerotic disease, i.e. intima–media thickening in the carotid bulb and the femoral artery, as well as with well-known risk factors of atherosclerosis such as high blood pressure, low HDL cholesterol, high serum levels of apoB and triglyceride, and a small LDL particle size [16,21,28]. In addition, there is also evidence from other studies that common carotid IMT is related to atherosclerosis [16,28–35]. Five studies have shown that common carotid IMT predicts cardiovascular diseases such as myocardial infarction and stroke [28,36–39].

The interpretation of our data is that a large common carotid IMT, as a valid indicator of established atherosclerotic disease, was associated with a low insulin sensitivity, as assessed by the best available method. The observation that this association was not independent of serum triglycerides and pulse pressure is probably explained by the close relation between insulin resistance and many other risk factors. An observational, cross-sectional study is a blunt tool to clarify the true nature of the associations between highly interrelated factors, and information on causality can only be obtained with other study designs. One may speculate that serum triglyceride levels might be a composite marker of all pro-atherogenic mechanisms related to the metabolic syndrome. Thus, serum triglycerides showed statistically significant associations with most components in the metabolic syndrome and is closely related to the accompanying atherogenic lipoprotein pattern, including small LDL particles.

Our conclusion is that we have found an association between insulin resistance and common carotid IMT, but that there was no relation to IMT in the bulb or femoral artery. Serum triglycerides and pulse pressure were independently associated with common carotid IMT.


This work was supported by grants from the Swedish Heart-Lung Foundation, the Swedish Medical Research Council (12270,10880), King Gustav V and Queen Viktoria Foundation and AstraZeneca, Mölndal, Sweden. We thank the staff at the Wallenberg Laboratory for Cardiovascular Research; Eva-Lena Alenhag, Lisbet Birke, Anna Frödén, Maria Gustavsson, Caroline Schmidt, Inger Wendelhag, Marianne Åmand and Aira Lidell for excellent technical assistance.

Received 9 February 2000; revision received 5 May 2000; accepted 28 June 2000.