This web only provides the extract of this article. If you want to read the figures and tables, please reference the PDF full text on Blackwell Synergy. Thank you.
- Original Article -
Higher testosterone levels are associated with increased
high-density lipoprotein cholesterol in men with cardiovascular
disease: results from the Massachusetts Male Aging Study
Stephanie T. Page1, Beth A.
Mohr2, Carol L. Link2, Amy B.
O'Donnell2, William J.
Bremner1, John B. McKinlay2
1Department of Medicine, University of Washington, Seattle 98195, USA
2New England Research Institutes, Watertown 02472, USA
Aim: To study the relationship between circulating androgens (total testosterone [TT], free testosterone [fT] and
dihydrotestosterone [DHT]) and high-density lipoprotein cholesterol
(HDL-C) in men with and without cardiovascular disease (CVD).
Methods: Cross-sectional analyses included 1 661 baseline samples from the Massachusetts Male
Aging Study (MMAS), a population-based cohort of men ages 40_70 years. Serum hormones were measured by
radioimmunoassay and HDL-C was determined following precipitation of the lower density lipoproteins. CVD was
determined by self-report. Analyses were performed using multiple linear regression.
Results: TT and HDL-C were positively correlated in the entire sample
(r = 0.11, P = 0.0001). After adjusting for confounders, we found this
relationship was mostly limited to the 209 men with CVD. Among
men with CVD, TT (P = 0.0004), fT
(P = 0.0172) and DHT
(P = 0.0128) were all positively correlated with
HDL-C, whereas in men without CVD only TT correlated
with HDL-C (P = 0.0099).
Conclusion: Our results suggest that if androgens contribute to CVD in middle-aged
men, the effect is not related to a suppressive effect of endogenous T on HDL-C.
(Asian J Androl 2008 Mar; 10: 193_200)
Keywords: testosterone; high-density lipoprotein cholesterol; androgens; epidemiology
Correspondence to: Dr Stephanie T. Page, 1959 NE Pacific, Box 356426, Seattle, WA 98195, USA.
Tel: +1-206-616-0483 Fax: +1-206-616-0499
Received 2007-01-29 Accepted 2007-07-23
The incidence of cardiovascular disease (CVD) is greater in men than age-matched women . Major
cardiovascular risk factors also differ by gender, including lipid abnormalities, which might help to explain some of the
increased risk of CVD in men . In particular, age-adjusted levels of high-density lipoprotein cholesterol
(HDL-C) are lower in men than in women . Interestingly,
this difference is only manifest after puberty, supporting the concept that
androgens contribute to a reduction in HDL-C in men .
It has been proposed that sex hormones contribute to the
discrepancy in incident of CVD in men compared to pre-menopausal women . In men, low levels of serum
androgens are associated with increased risk of developing type 2 diabetes  and the metabolic syndrome , which
are both strongly linked to the development of CVD. In addition, androgens could affect CVD through their impact of
androgen on lipoprotein metabolism and HDL-C. Exogenous testosterone (T) lowers HDL-C when given to eugonadal
middle-aged men both in physiologic doses associated with male hormonal contraceptive regimens and in
supraphysiologic doses to athletes . Likewise,
androgen deprivation, either experimental or for the treatment
of prostate disease, increases HDL-C . These data
support the hypothesis that androgens inhibit
HDL-C production, or, perhaps, increase HDL-C catabolism.
In contrast to data from intervention trials,
epidemiologic analyses have found a positive relationship
between androgens and HDL-C [8, 9]. Moreover,
substitution of T in hypogonadal men with age-associated
hypogonadism results in either no change or only minor
decreases in HDL-C , whereas correction of
hypogonadism secondary to Klinefelters or Kallmanns syndrome
can increase HDL-C . These apparently conflicting
results suggest that the relationship between HDL-C and
androgens is complex and might be host dependent.
Using data from the Massachusetts Male Aging Study
(MMAS), a population-based cohort of 1 709 middle-aged
and older men, we investigated whether T and HDL-C are
associated in men with and without CVD. We hypothesized that if T is involved in the causal pathway between
HDL-C and CVD, higher androgen levels would be
associated with lower levels of HDL-C in men with CVD.
2 Materials and methods
The MMAS is a prospective, community-based, observational study of aging in middle-aged and older men.
The current report uses only baseline data
(T1: collected from 1987_1989). The design has been described
previously . At baseline, men aged 40_70 years from
11 cities and towns in the Boston, Massachusetts
metropolitan area, USA were randomly selected from annual
state census listings. To obtain a sample with
approximately equal percentages in each age decade (40_49,
50_59, 60_69 years), age-stratified cluster sampling was
used. Of those eligible, 1 709 (52%) agreed to
participate at T1. This response rate likely reflects, in part, the
early-morning phlebotomy, extensive in-home interview,
and absence of financial incentive involved in this study.
Trained interviewer/phlebotomists visited the men in their
homes, administered a standardized interview, and
obtained physical measures and blood samples. The New
England Research Institutes' Institutional Review Board
approved all protocols, including informed consent
2.2 Hormones and lipids
Two non-fasting blood samples were collected within
4 h of the subject's awakening to control for diurnal
variations in hormone levels. Samples were drawn 30 min
apart, pooled to help smooth episodic secretion,
transported in ice-cooled containers, and centrifuged within
6 h. The samples were stored at _20ºC until shipment
on dry ice to the central laboratory and then stored
frozen at _70ºC until being assayed.
All assays were performed at The Endocrine Laboratory, University of Massachusetts Medical School
(Worcester, MA, USA). Total T (TT) was determined
by radioimmunoassay (RIA) kit (Diagnostic Products
Corporation, Los Angeles, CA, USA). Hormones and lipids were assayed in 1994 on sera stored since
collection in 1987_1989. A structural equation model,
equivalent to a Deming regression, showed negligible change
as a result of assay drift or storage. The assay
cross-reactivity with dihydrotestosterone (DHT) was 2.8%.
The intra-assay and inter-assay coefficients of variation
(CV) were 5.4% and 8.0%, respectively. As noted
pre-viously, the distribution of MMAS serum TT levels is
similar to that reported by other large epidemiologic
studies that have used RIA techniques .
Sex hormone-binding globulin (SHBG) was measured
by RIA using kits (Farmos Diagnostica, Farmos Group,
Oulunsalo, Finland). The intra-assay and inter-assay CVs
were 8% and 10.9%, respectively. DHT was measured
by RIA column chromatography . The intra-assay
and inter-assay CV were 10.9% and 12.2%, respectively.
The Södergard equation was used to calculate free T
(fT), assuming a fixed albumin-bound concentration .
The Södergard equation produces estimates for fT, which
closely approximate those obtained from equilibrium
dialysis . Serum lipids were measured at the Lipid
Research Laboratory at Miriam Hospital, Brown
University (Providence, RI, USA). This lab participates in the
national survey for clinical laboratories sponsored by the
College of American Pathologists. HDL-C was determined on non-fasting serum samples following
precipitation of the lower density lipoproteins using Heparin
Manganese reagent .
2.3 Confounders and cardiovascular disease
Well-validated instruments were used: alcohol ,
physical activity (Stanford Five-City Physical Activity
Questionnaire ). Height and weight, waist and hip
circumference were measured using standard techniques
. Smoking and chronic disease (diabetes,
hypertension and CVD) were ascertained by self-report. Self-
report of heart disease was assessed at every timepoint
by asking "Have you ever been told by a health
professional that you have heart disease?". As reported
previously, using longitudinal data from MMAS, we have
found the concordance of self-report data compared to
medical report and National Death Index data combined
was approximately 80%. This is comparable to the
concordance rate between self-report and medical records
data reported in the published literature for ischemic heart
disease and cardiovascular conditions in general .
To determine prescription and non-prescription
medication use, the interviewer copied the medication name
from the label and queried the reason for use.
Medications were coded using a system based on the American
Hospital Formulary Service, as described previously .
2.4 Statistical analysis
Of the 1 709 in the original cohort, men who were
missing HDL-C data (n = 44) or all hormone data
(n = 4) were excluded from the analyses, resulting in a sample
size of 1 661.
Because of their skewed distributions, HDL-C, SHBG,
DHT, body mass index (BMI), waist circumference (WC), and waist to hip ratio (WHR) were log transformed
prior to modeling (Table 4 and Figure 1). However,
results are presented on the original scale. To test whether
HDL-C differed by categories, a two sample unpaired
t-test was used. Tests of whether a variable differed by CVD
status were done with a χ2-test or Fisher exact test for
categorical variables and a two sample unpaired
t-test for continuous variables. Correlations between HDL-C and
continuous predictors were assessed by the Pearson
Multiple linear regression analysis was used to model
HDL-C level as a function of hormone level and potential
confounders. The impact of TT, fT, SHBG and DHT was examined. Separate models were fit for each variable.
The confounding effects of the following variables
were examined: age, chronic disease (diabetes,
hypertension and CVD), medication (lipid-lowering,
prescription medications known to impact hormone levels),
smoking, alcohol intake and adiposity (BMI, WC, WHR).
Because of their correlation, the three adiposity measures
were modeled separately.
For all models, we tested for possible two-way
interactions between each hormone and each confounder.
An interaction is present if the impact of the hormone on
HDL-C varies by the level of a third variable. For many
of the models, we found a statistically significant
interaction between CVD and the hormone. This indicates
that the relationship between the hormone and HDL-C
differed depending on whether or not the man had CVD
(e.g. positively correlated in one group, negatively in
other; or association strong in one group and weak or
nonexistent in another). When an interaction was present,
we performed tests among the men with and without CVD to determine if the hormone_HDL-C association
existed within each group.
SAS software (SAS System for Windows 9.1; Cary,
NC, USA) was used to perform statistical analysis. The
level of significance was set at P < 0.05.
3.1 Baseline characteristics
Baseline demographic characteristics of the study
cohort (n = 1 709) have been presented previously 
and demographics of the analysis sample
(n = 1 661) were similar. The sample was predominantly white
(95.5%), employed (78%), with at least a high school education
(88%). By design (e.g. random sampling), these
demographics closely match those of the population of
Massachusetts in 1990 according to census data.
Descriptive statistics for baseline confounding variables,
hormone levels, and HDL-C are shown in Table 1. Subjects
ranged from 40_70 years old with a mean of
55.2 ± 8.7. The prevalence of CVD was 13%
(n = 209), similar to other community-based studies of older men [2, 21].
Lipid-lowering medication usage was rare (1%) while
9% of subjects reported a prescription medication that
could lower hormone levels. Mean (standard deviation)
TT was 17.9 (6.1) nmol/L, SHBG was 32.3 (16.3)
nmol/L, and average HDL-C was 1.10 (0.36) mmol/L (equivalent
to 42.5 mg/dL). Characteristics of the subjects with and
without CVD are shown in Table 2. As expected, men
with CVD were older, more likely to take lipid-lowering
medication, have hypertension and/or diabetes mellitus,
and had a greater degree of central adiposity, as assessed
by waist circumference and WHR, than men without CVD. Men with CVD also had higher total cholesterol,
lower HDL-C, and lower total and free testosterone than
those without known CVD.
3.2 Relationship between HDL-C and CVD and
established cardiac disease risk factors
The association between mean HDL-C and established coronary risk factors in the study cohort is shown
in Table 3. As expected, mean HDL-C was significantly
lower in men with CVD, diabetes and hypertension. Men
with higher alcohol intake had higher HDL-C.
Lipid-lowering medication had little impact on HDL-C,
possibly because of the low prevalence of use. Mean HDL-C
did not differ by hormone medication or smoking.
3.3 Androgens and HDL-C
Overall, in unadjusted analyses, TT was positively
correlated with HDL-C (r = 0.11,
P = 0.0001; Table 4) although the magnitude of the correlation coefficient was
small. SHBG and DHT were also positively correlated
with HDL-C (Table 4; r = 0.16 and
r = 0.06, respectively), whereas fT was not
(r = 0.01).
All three measures of adiposity, BMI, WC and WHR,
were inversely correlated with HDL-C (Table 4).
3.4 Androgens and HDL-C in men with and without
We went on to examine the relationship between
androgens and HDL-C in men with and without CVD after
controlling for confounders, including age and central
adiposity (Figure 1). The regression lines were adjusted
for WHR (as a measure of central adiposity), smoking,
alcohol consumption, age and the use of medications that
might affect hormone measures. The association between TT and HDL-C differed depending on whether
the man had CVD (P-value for interaction term 0.0130).
Among men with CVD, TT was significantly and positively associated with HDL-C
(P = 0.0004 for association between T and HDL-C among CVD cases).
Figure 1A illustrates the magnitude of this relationship: if two men
with CVD differed by 5 nmol/L in TT and all other
characteristics in the model were equal, the man with the
higher T would be expected to have a 6% higher HDL-C.
The association between TT and HDL-C among men without CVD
(P = 0.0099) was much weaker; a
5 nmol/L difference in TT would result in a 2% higher HDL-C.
Moreover, when WC or BMI were controlled for instead
of WHR, the hormone_HDL-C association also varied
by CVD status (interaction P = 0.0144 and 0.0289,
respectively). Among the men with CVD, a high TT was still associated with high HDL-C
(P = 0.0028 WC model;
P = 0.0053 BMI model). However, there was
no longer any association between TT and HDL-C among
the men without CVD when we adjusted for WC or BMI
instead of WHR (P = 0.2875 and
P = 0.1925, respectively).
The results for fT and DHT were similar to those for
TT (Figure 1B and 1D). When WHR was controlled
for, the association between fT and HDL-C varied by
CVD status and was present only among men with CVD
(interaction P = 0.0226). High fT was associated with high
HDL-C in this group (P = 0.0172). Similar results were
found when controlling for WC (interaction
P = 0.0389), although the interaction term did not reach significance
when adjusting for BMI (P = 0.0610). Analogous to the
results for TT, a man with 0.2 nmol/L higher fT would
be expected to have 6% higher HDL-C than a man with
similar WHR, cardiac risk factors and lower serum fT.
Moreover, in men with CVD there was a positive
association between DHT and HDL-C when WHR was controlled for (interaction
P = 0.0470; P = 0.0128 for the
correlation between DHT and HDL-C in this model); this
relationship, however, did not reach statistical
significance if WC or BMI replaced WC in the model (interaction
P = 0.1010 and
P = 0.1560, respectively).
Unlike TT, fT and DHT, the association between SHBG and HDL-C in the adjusted model did not differ by
CVD status (Figure 1C, P = 0.9854 for the interaction
term). Controlling for WC or BMI did not alter these
results (data not shown). However, SHBG was very highly positively associated with HDL-C regardless of
CVD status or which adiposity measure was controlled
for (P = 0.0001 for SHBG in all three adiposity adjusted
models without interaction terms; data not shown).
Similar to other cross-sectional analyses of middle
aged men [8, 9], in the present study of 1 661 men
enrolled in the MMAS, we found a positive relationship
between HDL-C and TT. HDL-C was inversely related
to chronic disease, including CVD, diabetes and
hypertension. After adjustment for confounders (age, WHR,
smoking, alcohol consumption and medications), we found
that the relationship between androgens and HDL-C was
mostly limited to the 209 men with CVD within the cohort.
Importantly, the positive relationship between HDL-C and
androgens was consistent, whether the assessment of
androgen level used TT, fT or DHT measurement. Moreover, although SHBG, which binds 60% of
circulating T, was strongly and positively correlated with
HDL-C, the strength of this relationship did not vary between
men with and without CVD. These results suggest that
the dichotomy between androgens and HDL-C in men with and without CVD that we observed is androgen
specific, and not a function of the carrier protein.
The hypothesis that T adversely impacts CVD risk is
largely based upon interventional trials that indicate that
exogenous androgens suppress HDL-C in men, even when given in physiologic doses . T has been
demonstrated to decrease HDL-C by increasing both hepatic
lipase (HL) activity  and scavenger receptor B1
expression , resulting in increased HDL-C uptake by
hepatocytes [23, 24]. However, it is difficult to
extrapolate these data to increases in CVD risk [23, 24],
because atherosclerosis is inhibited in transgenic animal
models with enhanced HL activity despite decreased
HDL-C . Furthermore, epidemiologic data have failed to
demonstrate an association between T levels and CVD
[24, 25] and, in fact, lower androgen levels are
associated with the development of type 2 diabetes, which would
increase the risk of CVD . Moreover, high
circulating T levels are associated with high, not low, HDL-C
levels in cross-sectional studies . If high levels of
circulating androgens contribute to CVD by lowering
HDL-C, one might expect that in men with CVD,
androgens would be associated with lower HDL-C levels; in
fact, our results showed just the opposite. Our data
support the hypothesis that if T increases CVD risk, this
effect is unlikely to be mediated through a negative
impact of T on HDL-C.
Because there is a very strong, positive association
between SHBG and HDL-C , analyses of free hormone is critical to discerning the effects of androgens.
In addition, although serum concentrations of DHT (the
product of 5α-reduction of T) are significantly lower
than T in men, DHT is a significantly more potent
androgen than T, at least in vitro. A strength of our data is the
consistency in the relationships between T, fT and DHT
and HDL-C in men with CVD. Although most studies find a positive relationship between TT and HDL-C, some
have suggested that this likely reflects the contribution
of SHBG (bound to T) , and that increased fT
results in a more atherogenic lipid profile, including lower
HDL . Our data argues against this, because both
higher fT and DHT levels were associated with higher
HDL-C in men with CVD. In addition, T levels might be
impacted by the presence of acute or chronic illness,
body composition, age and smoking . Others have
argued that the association between T and HDL-C is
mostly, if not completely, mediated by body composition,
and central adiposity in particular, because obesity is
associated with low T and low HDL-C . However,
when we controlled for BMI, WHR or WC in our analyses, androgens still were clearly related to HDL-C
in men with CVD. Such factors, or a selection bias,
might have influenced previous, small case control
studies in men undergoing coronary angiography, which failed
to find a relationship, or found a negative relationship,
between androgens and HDL-C in men with or without
CVD [8, 28].
In conclusion, using cross-sectional analyses of a
large cohort of community dwelling, middle-aged men,
we demonstrate a strong, positive relationship between
androgens and HDL-C in men with CVD. This is in contrast to men without CVD, where only TT, and not
fT or DHT, weakly correlated with increased HDL-C.
Our data suggest that any androgenic effect on CVD
risk is not mediated by an inhibitory effect of endogenous
T on HDL-C levels.
This work was supported by grants from the Endocrine Society (S. T. P.), VA Special Fellowship Program
in Advanced Geriatrics (S. T. P.), National Institute on
Aging (AG04763), and the National Institute of Diabetes
and Digestive and Kidney Diseases (DK51345 and DK44995).
1 Lloyd-Jones DM, Larson MG, Beiser A, Levy D. Lifetime
risk of developing coronary heart disease. Lancet 1999; 353:
2 Jousilahti P, Vartiainen E, Tuomilehto J, Puska P. Sex, age,
cardiovascular risk factors, and coronary heart disease: a
prospective follow-up study of 14 786 middle-aged men and women
in Finland. Circulation 1999; 99: 1165_72.
3 Kirkland RT, Keenan BS, Probstfield JL, Patsch W, Lin TL,
Clayton GW, et al. Decrease in plasma high-density
lipoprotein cholesterol levels at puberty in boys with delayed
adolescence. Correlation with plasma testosterone levels.
JAMA 1987; 257: 502_07.
4 Liu PY, Death AK, Handelsman DJ. Androgens and
cardiovascular disease. Endocr Rev 2003; 24: 313_40.
5 Ding EL, Song Y, Malik VS, Liu S. Sex differences of
endogenous sex hormones and risk of type 2 diabetes: a systematic
review and meta-analysis. JAMA 2006; 295: 1288_99.
6 Kupelian V, Page ST, Araujo AB, Travison TG, Bremner WJ,
McKinlay JB. Low sex hormone-binding globulin, total
testosterone, and symptomatic androgen deficiency are
associated with development of the metabolic syndrome in
nonobese men. J Clin Endocrinol Metab 2006; 91: 843_50.
7 Hurley BF, Seals DR, Hagberg JM, Goldberg AC, Ostrove
SM, Holloszy JO, et al. High-density-lipoprotein cholesterol
in bodybuilders v powerlifters. Negative effects of androgen
use. JAMA 1984; 252: 507_13.
8 Kiel DP, Baron JA, Plymate SR, Chute CG. Sex hormones
and lipoproteins in men. Am J Med 1989; 87: 35_9.
9 Barrett-Connor EL. Testosterone and risk factors for
cardiovascular disease in men. Diabete Metab 1995; 21: 156_61.
10 Whitsel EA, Boyko EJ, Matsumoto AM, Anawalt BD, Siscovick
DS. Intramuscular testosterone esters and plasma lipids in
hypogonadal men: a meta-analysis. Am J Med 2001; 111: 261_9.
11 Ozata M, Yildirimkaya M, Bulur M, Yilmaz K, Bolu E, Corakci
A. et al. Effects of gonadotropin and testosterone treatments
on Lipoprotein(a), high density lipoprotein particles, and other
lipoprotein levels in male hypogonadism. J Clin Endocrinol
Metab 1996; 81: 3372_8.
12 O'Donnell AB, Araujo AB, McKinlay JB. The health of
normally aging men: The Massachusetts Male Aging Study
(1987_2004). Exp Gerontol 2004; 39: 975_84.
13 Longcope C, Franz C, Morello C, Baker R, Johnston CC Jr.
Steroid and gonadotropin levels in women during the
peri-menopausal years. Maturitas 1986; 8: 189_96.
14 Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation
of simple methods for the estimation of free testosterone in
serum. J Clin Endocrinol Metab 1999; 84: 3666_72.
15 National Heart and Lung Institute LRCPLMC. Manual of
Laboratory Operations: Lipid and Lipoprotein Analysis. Vol
DHEW publication No. 75_628. 2nd edn. Washington, DC:
US Government Printing Office; 1974.
16 Khavari KA, Farber PD. A profile instrument for the
quantification and assessment of alcohol consumption. The Khavari
Alcohol Test. J Stud Alcohol 1978; 39: 1525_39.
17 Sallis JF, Haskell WL, Wood PD, Fortmann SP, Rogers T,
Blair SN, et al. Physical activity assessment methodology in
the Five-City Project. Am J Epidemiol 1985; 121: 91_106.
18 McKinlay SM, Kipp DM, Johnson P, Downey K, Carelton
RA. A field approach for obtaining physiological measures in
surveys of general populations: Response rates, reliability and
costs. Paper presented at: In: Proceedings of the Fourth
Conference on Health Survey Research Methods, USDHHS-PHS
Publication 84-3346, Washington DC, 1984.
19 St Sauver JL, Hagen PT, Cha SS, Bagniewski SM, Mandrekar
JN, Curoe AM, et al. Agreement between patient reports of
cardiovascular disease and patient medical records. Mayo
Clin Proc 2005; 80: 203_10.
20 Mohr BA, Guay AT, O'Donnell AB, McKinlay JB. Normal,
bound and nonbound testosterone levels in normally ageing
men: results from the Massachusetts Male Ageing Study. Clin
Endocrinol (Oxf) 2005; 62: 64_73.
21 Barrett-Connor E, Khaw KT. Endogenous sex hormones and
cardiovascular disease in men. A prospective population-based
study. Circulation 1988; 78: 539_45.
22 Herbst KL, Amory JK, Brunzell JD, Chansky HA, Bremner
WJ. Testosterone administration to men increases hepatic
lipase activity and decreases HDL and LDL size in 3 wk. Am
J Physiol Endocrinol Metab 2003; 284: E1112_8.
23 Langer C, Gansz B, Goepfert C, Engel T, Uehara Y, von Dehn
G, et al. Testosterone up-regulates scavenger receptor BI and
stimulates cholesterol efflux from macrophages. Biochem
Biophys Res Commun 2002; 296: 1051_7.
24 Wu FC, von Eckardstein A. Androgens and coronary artery
disease. Endocr Rev 2003; 24:183_217.
25 Yarnell JW, Beswick AD, Sweetnam PM, Riad-Fahmy D.
Endogenous sex hormones and ischemic heart disease in men.
The Caerphilly prospective study. Arterioscler Thromb 1993;
26 Bataille V, Perret B, Evans A, Amouyel P, Arveiler D,
Ducimetière P, et al. Sex hormone-binding globulin is a major
determinant of the lipid profile: the PRIME study.
Atherosclerosis 2005; 179: 369_73.
27 Gyllenborg J, Rasmussen SL, Borch-Johnsen K, Heitmann
BL, Skakkebaek NE, Juul A. Cardiovascular risk factors in
men: The role of gonadal steroids and sex hormone-binding
globulin. Metabolism 2001; 50: 882_8.
28 Kabakci G, Yildirir A, Can I, Unsal I, Erbas B. Relationship
between endogenous sex hormone levels, lipoproteins and
coronary atherosclerosis in men undergoing coronary angiography.
Cardiology 1999; 92: 221_5.