European Journal of Epidemiology

, Volume 24, Issue 6, pp 289–295

Endogenous testosterone and the prospective association with carotid atherosclerosis in men: the Tromsø study

Authors

    • Institute of Clinical MedicineUniversity of Tromsø
    • Department of MedicineUniversity Hospital of North Norway
  • S. H. Johnsen
    • Institute of Clinical MedicineUniversity of Tromsø
    • Department of NeurologyUniversity Hospital of North Norway
  • H. Schirmer
    • Institute of Community MedicineUniversity of Tromsø
    • Department of CardiologyUniversity Hospital of North Norway
  • I. Njølstad
    • Institute of Community MedicineUniversity of Tromsø
  • J. Svartberg
    • Institute of Clinical MedicineUniversity of Tromsø
    • Department of MedicineUniversity Hospital of North Norway
Cardiovascular Disease

DOI: 10.1007/s10654-009-9322-2

Cite this article as:
Vikan, T., Johnsen, S.H., Schirmer, H. et al. Eur J Epidemiol (2009) 24: 289. doi:10.1007/s10654-009-9322-2
  • 104 Views

Abstract

The role of testosterone in the development of cardiovascular disease is controversial. Recent observational studies, however, suggest a protective role of normal endogenous testosterone levels in the development of atherosclerosis. In a cohort from the Tromsø study, 1,101 men had both hormone-levels measured and the right carotid artery examined by ultrasound in 1994 and 2001. We studied the prospective association between sex hormone-levels and progression of carotid intima-media thickness (IMT) and plaque area from 1994 to 2001. We also performed a cross-sectional study of 2,290 men from the population in 2001. The data were analysed by univariate correlations, analyses of covariance and multiple linear regression analyses. In the cross-sectional study, we found an inverse association between testosterone levels and total carotid plaque area (P < 0.05), after adjusting for age, systolic blood pressure, smoking and use of lipid-lowering drugs. We found no prospective associations between sex hormone-levels and change in plaque area or IMT from 1994 to 2001. The lack of prospective associations in our study may be due to increased use of anti-hypertensive and lipid-lowering drugs from 1994 baseline to follow-up.

Keywords

ProspectiveCross-sectionalIntima-media thicknessPlaque areaAtherosclerosisTestosterone

Abbreviations

BMI

Body mass index

CCA

Common carotid artery

HDL

High density lipid cholesterol

IMT

Intima media thickness

SHBG

Sex hormone-binding globulin

Introduction

Men have a consistently earlier start of atherosclerosis than women. The role of testosterone in the development of atherosclerosis is still controversial, but it has become increasingly clear that higher physiological levels of testosterone is not the cause of a head-start of atherosclerosis in men [1, 2]. On the contrary, testosterone in men tends to be inversely related to both classical risk-factors of atherosclerosis and subclinical measures of atherosclerosis in cross-sectional analyses, suggesting a protective role of higher normal endogenous testosterone levels on the atherotrombotic process. Inverse associations have been found between testosterone levels and blood pressure [3, 4], left ventricular mass [3], serum triglyceride levels [5], waist circumference [6], HbA1c [7]. Positive associations have been shown between testosterone levels and HDL cholesterol [8, 9].

Carotid ultrasound measurement of intima-media thickness (IMT) and plaque area is considered a reliable way of assessing subclinical atherosclerosis. However, IMT and plaque are different ultrasound phenotypes and reflects different biological aspects of atherosclerosis, the former being more closely related to cerebrovascular events [10, 11], the latter being more closely associated with coronary atherosclerosis [12] and coronary events [13].

Negative associations between testosterone levels and carotid IMT have been reported in several cross-sectional studies [1417]. Although consistent, findings in cross-sectional analyses cannot distinguish between associations and causality, or disclose the direction of causality (either of which is plausible). One prospective study of 195 community-dwelling men aged 73–91 years showed a higher rate of IMT-progression over 4 years in men with serum testosterone levels in the lowest tertile [18]. However, this was a small study of the oldest old, and the finding needs to be replicated also in a general population.

We aimed to study the relationship between testosterone levels and carotid IMT and plaque area, both cross-sectionally and prospectively in a cohort of community-dwelling men. To the best of our knowledge, this is the largest population sample in which this issue has been addressed, and the first study to examine the relationship between testosterone and plaque area. We hypothesise that endogenous testosterone levels are inversely associated with IMT and plaque area, and that lower testosterone levels are associated with progression in IMT and plaque area.

Materials and methods

Subjects

The Tromsø study is a population-based prospective study with repeated health surveys, primarily focusing on cardiovascular and other chronic diseases.

Included in these analysis were men who participated in the fourth survey in 1994–1995 (for simplicity 1994 or baseline in the following) and in the fifth survey in 2001. The study populations have been described in detail previously [19]. For the cross-sectional analysis, we included 2,290 men with available data on both testosterone levels and carotid ultrasound from the fifth survey in 2001. For the prospective analysis, we included all men (n = 1,101) who had the right carotid artery examined by ultrasound and testosterone levels measured both in the fourth (1994) and the fifth survey (2001).

Cardiovascular risk factors

At both surveys, height and weight were measured in standing subjects wearing light clothing without shoes. Waist circumference was measured at the umbilical line according to written protocol. BMI (kg/m²) was calculated. Blood pressure was recorded with an automatic device (Dinamap Vital Signs Monitor, Critikon Inc., Tampa, FL, USA) by specially trained personnel. Non-fasting blood samples were drawn between 08:00 and 16:00 h in 1994, between 08:00 and 18:30 h in 2001. On both occasions, serum total cholesterol and triglycerides were analysed by enzymatic colorimetric methods with commercial kits (CHOD-PAP for cholesterol and GPO-PAP for triglycerides; Boehringer-Mannheim, Germany). Serum HDL cholesterol was measured after precipitation of lower-density lipoprotein with heparin and manganese chloride.

Self-administered questionnaires that included information about smoking habits, physical activity, medical history, use of anti-hypertensive and lipid-lowering drugs were completed and checked by trained nurses at both surveys.

Sex hormones

Serum samples from 1994 were analysed for sex hormones in the autumn of 2001. Serum samples from 2001 were analysed within 12 months. All samples were stored frozen at −70°C. On both occasions, the determination of total testosterone and SHBG was performed on Immulite 2000 (Diagnostic Product Corp., Los Angeles, CA, USA). The intra- and interassay coefficients of variation for the analyses were between 5 and 10%. Free testosterone values were calculated from total testosterone and SHBG using a fixed albumin concentration according to Vermeulen et al. [20].

Ultrasonography and measures of atherosclerosis

High-resolution B-mode and colour Doppler/pulsed-wave Doppler ultrasonography of the right carotid artery were performed as described previously [21, 22]. The same ultrasound imaging system and transducer (Acuson Xp10 128, ART upgraded, with a 7.5 MHz linear-array transducer, aperture size 38 mm) was used in 1994 and 2001. IMT was measured in 10 mm segments in three locations of the carotid artery: The near and far wall of the CCA and the far wall of the bifurcation. Frozen images from each segment were stored on high-resolution videotapes. The loss of parallel configuration of the near and far walls of the CCA served as a reference point for the start of the carotid bifurcation. Plaques were included in the measurements of IMT if they were located in areas predefined for IMT registrations. The ultrasonic images were analysed off-line with a computerised technique for automated ultrasonic image analysis. The mean IMT from the three preselected images was calculated for each location, and the average of the mean IMT in the three locations was used in the analyses. For measurement of IMT, the mean arithmetic difference between observers was −0.01 mm, and the mean absolute difference between observers was 0.11 mm. The mean arithmetic difference within observers was −0.01 mm and the mean absolute difference within observers was 0.10 mm. A plaque was defined as a localised thickening of the vessel wall of more than 50% compared to the adjacent (normal) IMT. In each subject, a maximum of six plaques were registered in the near and far walls of the common carotid, bifurcation and internal carotid, respectively. If more than one plaque was present, the sum of plaque areas was taken as the total plaque area. For measurement of plaque area, the mean arithmetic difference between observers was −1.0 mm2 (7.2% of mean plaque area, P < 0.05) and the limits of agreement ranging from −9.6 to +7.6 mm². The mean absolute difference was 2.9 mm² (21% of mean plaque area). The reproducibility study was performed on a sample of 111 participants in the Tromsø study in 1994 [21].

Statistical analyses

Distributions of population characteristics in 1994 and 2001 were expressed as mean and SD for continuous variables, and as percentages for categorical variables. Paired samples t-test was used for comparing means for continuous variables, and McNemar’s test for categorical variables.

Normal distribution was evaluated with determination of skewness and histograms. Total and free testosterone were considered normally distributed and total plaque area and IMT assumed normal distribution after log transformation. Change in plaque area, 2001 minus 1994 (Δplaque area) and change in IMT, 2001 minus 1994 (ΔIMT) were normally distributed.

In the cross-sectional analysis, ln plaque area and ln IMT were defined as dependent variables. Multiple linear regression analysis was performed to evaluate the independent contributions of sex hormones, adjusting for age and risk factors.

In the prospective analysis, baseline hormone levels were investigated in relation to progression of plaque area (Δplaque area) and IMT (ΔIMT), respectively. We also investigated the relation between change in hormone levels, 2001 minus 1994 (Δtestosterone and ΔSHBG) and Δplaque area and ΔIMT. Finally, we analysed for associations between baseline hormone levels and plaque area and IMT in 2001. Plaque area and IMT were analysed as dependent variables in multiple linear regression models, allowing for adjustment for age and cardiovascular risk factors. The same algorithm was applied with Δplaque area and ΔIMT as dependent variables. To assess whether a threshold was present, multivariate models using ANCOVA analyses were used to estimate mean Δplaque area and mean ΔIMT across quintiles of baseline levels of testosterone and change in levels of testosterone, respectively.

In both the cross-sectional and the prospective analysis, we only adjusted for risk factors that were significantly correlated (in univariate analysis using Pearson correlation) with plaque area and IMT, respectively. All statistical tests were two-tailed with statistical significance defined as P < 0.05. Model assumptions were assessed with residual analyses. Co-linearity diagnostics was applied for all regression analyses. The SPSS statistical software version 15.0 for windows was used for all analyses.

Ethics

The study protocol was approved by the Regional Committee for Medical and Health Research Ethics, North Norway and informed consent was obtained from all the participants.

Results

Cross-sectional analysis

Characteristics of the study sample in 2001 are presented in Table 1. Mean age was 66 years and the mean BMI was 26.9 kg m−2. The average mean IMT was 0.890 mm. One or more plaque was present in 67.7% and the mean total plaque area among these was 27.5 mm2. Lipid-lowering drugs were used by 18.3 and 26.1% used anti-hypertensive medication.
Table 1

Characteristics of 2,290 participating men: the Tromsø study 2001

 

n

Mean (SD) or %

Age (years)

2,290

66.0 (9.3)

BMI (kg/m2)

2,276

26.9 (3.5)

Total cholesterol (mmol/l)

2,290

6.1 (1.1)

HDL cholesterol (mmol/l)

2,290

1.4 (0.4)

Triglycerides (mmol/l)

2,290

1.6 (1.0)

HbA1C (%)

2,261

5.6 (0.8)

Testosterone (nmol/l)

2,290

13.9 (5.5)

SHBG (nmol/l)

2,290

43.5 (17.3)

Free testosterone (pmol/l)

2,290

244.9 (92.7)

Systolic blood pressure (mmHg)

2,289

143 (20)

Diastolic blood pressure (mmHg)

2,290

81 (11)

Self reported CHD (%)

2,243

19.3

Self reported CVD (%)

2,221

22.0

Current smoking (%)

2,128

25.6

Diabetes (%)

2,259

4.9

Current use of antihypertensive drugs (%)

2,233

26.1

Current use of lipid-lowering drugs (%)

2,205

18.3

Physical activity (vigorous) ≥ 3 h/week (%)

1,874

12.9

Intima media thickness (mm)

2,290

0.890 (0.188)

Carotid plaque present (%)

2,290

67.7

Total carotid plaque area (mm²)

1,550

27.5 (21.8)

BMI body-mass index, SHBG sex hormone binding globulin, CHD coronary heart disease, CVD cardiovascular disease

In Pearson’s correlation analysis, ln plaque area was correlated with age (r = 0.266, P < 0.001), testosterone (r = −0.076, P = 0.003), free testosterone (r = 0.109, P < 0.001), use of lipid-lowering drugs (r = 0.133, P < 0.001), current smoking (r = 0.058, P = 0.022) and systolic blood pressure (r = 0.177, P < 0.001).

Multiple linear regression models are presented in Table 2 with ln plaque area as dependent variable. Total testosterone was significantly and inversely associated with plaque area, also after adjusting for age, use of lipid-lowering drugs, systolic blood pressure and current smoking (P = 0.033). SHBG was borderline significant of being negatively associated with plaque area after adjustments (P = 0.054). Free testosterone was not associated with plaque area after adjustments (data not shown).
Table 2

Multiple linear regression models with ln plaque area as the dependent variable (2,290 men): the Tromsø study 2001

Independent variable

β

P

Independent variable

β

P

Testosterone

−0.053

0.033

SHBG

−0.049

0.054

Age

0.266

<0.001

Age

0.284

<0.001

Use of statin

0.157

<0.001

Use of statin

0.160

<0.001

Systolic blood pressure

0.134

<0.001

Systolic blood pressure

0.131

<0.001

Current smoking

0.133

<0.001

Current smoking

0.133

<0.001

R2

0.128

R2

0.127

SHBG sex hormone-binding globulin, β standardised regression coefficient, R2 the coefficient of determination

Ln IMT was correlated with age (r = 0.493, P < 0.001), total testosterone (r = −0.076, P < 0.001), free testosterone(r = −0.160, P < 0.001), SHBG (r = 0.082, P < 0.001) use of lipid-lowering drugs (r = 0.121, P < 0.001), systolic blood pressure (r = 0.303, P < 0.001), BMI (r = 0.085, P < 0.001) and waist circumference (r = 0.183, P < 0.001). In multiple linear regression models (data not shown), total testosterone was significantly associated with Ln IMT in age-adjusted analyses (β = −0.041, P = 0.023). But when adding the other variables, in any order, the independent contribution of total testosterone was lost and the effect estimates were greatly reduced. The association between free testosterone and Ln IMT was lost already when adjusting for age.

Prospective analysis

For these analyses, 1,101 men were available. General characteristics of the study sample in 1994 (baseline) and 2001 are shown in Table 3. All variables but triglycerides and physical activity changed significantly from 1994 to 2001. During follow up, there was a significant decrease in mean cholesterol-levels from 6.5 to 6.0 mmol/l, and a significant increase in mean total and free testosterone, from 13.2 to 14.0 nmol/l and from 206.1 to 246.1 pmol/l, respectively. There was a significant increase in the proportion of statin-users and users of antihypertensives, from 3.3 to 16.8% and from 14.5 to 25.6%, respectively, and a decreasing number of smokers, from 30.6 to 25.5%. Among the baseline variables, only age (r = 0.115, P < 0.001), use of lipid-lowering drugs (r = 0.072, P = 0.034) and systolic blood pressure were significantly correlated with Δplaque area. Use of lipid-lowering drugs was the only baseline variable that was significantly correlated with ΔIMT (r = −0.108, P = 0.001). There were no significant correlations between baseline sex hormone levels and ln plaque area or ln IMT in 2001.
Table 3

Characteristics of the men who had both their hormone levels measured and carotids scanned in 1994 and 2001

 

1994

2001

 
 

n

Mean (SD) or %

n

Mean (SD) or %

Pa

Age (years)

1,101

59.0 (9.3)

1,101

66.0 (9.3)

<0.01

BMI (kg/m2)

1,101

26.2 (3.2)

1,095

26.9 (3.6)

<0.01

Total cholesterol (mmol/l)

1,101

6.5 (1.2)

1,099

6.0 (1.1)

<0.01

HDL cholesterol (mmol/l)

1,097

1.4 (0.4)

1,099

1.4 (0.4)

<0.01

Triglycerides (mmol/l)

1,101

1.6 (1.0)

1,099

1.6 (1.0)

0.81

HbA1C (%)

1,022

5.4 (0.6)

1,078

5.5 (0.8)

<0.01

Testosterone (nmol/l)

1,101

13.2 (5.2)

1,101

14.0 (5.4)

<0.01

SHBG (nmol/l)

1,081

51.2 (22.4)

1,080

43.7 (17.4)

<0.01

Free testosterone (pmol/l)

1,081

206.1 (75.2)

1,080

246.1 (91.4)

<0.01

Systolic blood pressure (mmHg)

1,101

140 (19)

1,101

143 (20)

<0.01

Diastolic blood pressure (mmHg)

1,101

81 (11)

1,101

81 (12)

0.11

Self reported CHD (%)

1,095

12.5

1,080

18.4

<0.01

Self reported CVD (%)

1,093

13.6

1,072

20.9

<0.01

Current smoking (%)

1,101

30.6

1,100

25.5

<0.01

Diabetes (%)

1,097

2.5

1,087

4.6

<0.01

Current use of antihypertensive drugs (%)

1,095

14.5

1,069

25.6

<0.01

Current use of lipid-lowering drugs (%)

869

3.3

1,058

16.8

<0.01

Physical activity (vigorous) ≥ 3 h/week (%)

1,091

11.4

891

13.1

0.44

Intima media thickness (mm)

1,101

0.878 (0.190)

1,101

0.899 (0.197)

<0.01

Carotid plaque present (%)

1,097

51.3

1,101

64.4

<0.01

Total carotid plaque area (mm²)

563

23.1 (21.3)

709

27.0 (21.5)

<0.01

aPaired samples t-test for continuous variables, McNemar’s test for categorical variables

BMI body-mass index, SHBG sex hormone-binding globulin, CHD coronary heart disease, CVD cardiovascular disease

As shown in Table 4, there were no significant differences in plaque growth in men who had increased, unchanged or decreased testosterone levels between 1994 and 2001.
Table 4

Mean change in plaque area by quintiles of changes in testosterone and SHBG levels between 1994 and 2001

 

ΔPlaque areaa

ΔPlaque areab

Δtestosterone quintile 1 (<−2.88)

5.296

5.005

Δtestosterone quintile 2 (−2.88 to −0.50)

5.502

5.536

Δtestosterone quintile 3 (−0.50 to 1.90)

6.962

5.564

Δtestosterone quintile 4 (1.90 to 4.58)

5.358

4.873

Δtestosterone quintile 5 (>4.58)

5.023

4.798

ΔSHBG quintile 1 (<−16.0)

6.096

5.683

ΔSHBG quintile 2 (−16.0 to −8.0)

4.973

3.802

ΔSHBG quintile 3 (−8.0 to −4.0)

5.573

5.219

ΔSHBG quintile 4 (−4.0 to 2.0)

5.020

5.302

ΔSHBG quintile 5 (>2.0)

7.406

6.685

aAdjusted for age

bAdjusted for baseline (1994) age, systolic blood pressure, use of lipid-lowering drugs, BMI, smoking and HDL-cholesterol

BMI body-mass index, SHBG sex hormone-binding globulin

Discussion

We found an inverse linear association between total testosterone levels and total plaque area in a cross-sectional study of 2,290 community-dwelling men, after adjustment for other classical risk factors for atherosclerosis. No association was found between total and free testosterone levels and IMT in the cross-sectional analysis. To the best of our knowledge, this is the first study to show a relationship between total testosterone and carotid plaque area. Evidence is emerging for a link between testosterone and the atherosclerotic process. In a study of male rabbits, a beneficial effect of testosterone on plaque development was reported, probably mediated by the androgen receptor [23]. Inflammatory cytokines were inversely related to testosterone in one study [24], and testosterone replacement in hypogonadal men was recently reported to reduce inflammatory cytokines [25]. In male castrated rabbits, testosterone replacement reduced plaque area and levels of the inflammatory markers TNFα, IL6, sICAM and MMP2. The effects were reversed by addition of flutamide, again suggesting the effects to be mediated by the androgen receptor [26].

A few studies have investigated the relationship between IMT as a measure of carotid atherosclerosis and testosterone levels (including our previous cross-sectional study from the fourth Tromsø study [14]), where a negative association between both total testosterone [1416] and free testosterone [27, 28] with carotid IMT were reported, quite consistently, which is in contrast with our present results. In addition, we did not find any predictive effect of testosterone levels on progression of IMT or plaque area. In terms of IMT, our results are not in accordance with Muller et al. [18], who found an increased progression in IMT over 4 years in 195 men with testosterone levels in the lowest tertile, an association that was independent of other classical risk factors for CVD. They studied IMT and testosterone among the oldest old, while our population was relatively younger. However, repeating our analyses in a subgroup of men >60 years old, did not change our results. Furthermore, we did not find any threshold effect.

To appreciate the results in the present study, some issues need to be addressed. In our previous cross-sectional study from the 1994 Tromsø study [14] a handful of participants used lipid-lowering drugs, while in the present study a large part of the participants (18.3%) used lipid-lowering medication. The mean cholesterol level in the population was lower in 2001, probably as a reflection of the increase in statin use. Knowing that use of statins reduce the rate of carotid atherosclerosis progression, and even may cause regression of carotid IMT [29, 30], it could be a likely explanation for the discrepancy between our present and previous results [1416, 18]. This could also mask a potential predictive effect of testosterone on progression in IMT and plaque area in the present study. We therefore repeated both the cross-sectional and the prospective analyses after excluding all the statin-users, but the results were unchanged. However, excluding statin users may render a relatively healthier population left for analysis, and in addition, excluding almost 20% of the cohort means a fall in statistical power to detect changes. Other changes in the population making it less likely to detect progression of atherosclerosis, is a considerable increase in use of anti-hypertensive medication from 14.5% in 1994 to 25.6% in 2001, and a 5% reduction in smoking habits over the same period.

Our findings in the cross-sectional analysis are still contradictive, with an association between testosterone and plaque area, but not with IMT. Compared to IMT, carotid plaques are believed to represent a later stage of atherosclerosis related to endothelial dysfunction, oxidation, inflammation and cell proliferation [31], and are more related to coronary atherosclerosis and events [12, 13]. IMT, especially in its early development, reflects a hypertensive hypertrophic response of the medial cells and is stronger related to cerebrovascular disease. As a continuous ultrasound trait, IMT is a relatively insensitive measure of plaque evolution since plaque grows along the vessel wall faster than it thickens [32]. Plaque area may therefore be a more sensitive and representative measure of the atherosclerotic burden than both plaque thickness and IMT.

Regarding the significant increase in total and free testosterone between 1994 and 2001, which conflicts with earlier longitudinal reports [33, 34], there may be several explanations. The 1994 samples had been frozen for an average of 6.5 years before being analysed, while the 2001 samples were analysed within 12 months, and degradation in testosterone by time may have occurred. However, as the testosterone molecule is rather stable, an increase in testosterone level in the frozen sample may also have occurred due to the possible evaporation and thereby a concentration effect during long-time storage [35, 36]. Furthermore, we cannot rule out a systematic bias in the assay itself between 2001 and 2005, although all samples were analysed on the same instrument. On the other hand, differences in time of the day when the samples were drawn in 1994 and 2001 could not be an explanation, as the mean time of sampling was similar.

Hormone levels were based on a single serum sample drawn between 08:00 and 16:00 h, and not as morning samples as recommended due to the diurnal variation of testosterone. This may lead to non-differential misclassification of the exposure variable, and to weakening of associations in our study. Finally, in studies on progression, the measurement errors of at least two measurements are accumulated, giving substantially lower statistical power compared to baseline measurements only. Studies on progression require a higher level of reproducibility, more demanding techniques and highly standardised protocols that include the definition of anatomical landmarks, control of probe insonation angle, and careful circumferential scanning of segments to identify the maximum wall thickness. This may lead to underestimation of possible associations; especially in the prospective analysis were repeated measurement errors add to one other.

In conclusion, total testosterone was found to be negatively and independently associated with carotid plaque area in the cross-sectional analyses. We found no prospective associations between testosterone and carotid plaque area and IMT. The lack of prospective associations may be due to the increased use of lipid-lowering and anti-hypertensive drugs from baseline to follow-up, and a predictive effect of testosterone levels on the progression in carotid plaques should not be completely ruled out on the basis of our findings. Further research on the cause–effect relationship between testosterone and the atherosclerotic process is still warranted.

Acknowledgments

This study was supported by a grant from The Northern Norway Region Health Authority.

Copyright information

© Springer Science+Business Media B.V. 2009