Coronary artery disease (CAD) is one of the leading causes of mortality in men and women, being 5th in the rank order of disabilities in 1990. CAD is predicted to become the leading global cause of disease burden by 2020 (1). The age-adjusted morbidity and mortality rates from CAD are 2.5- to 4.5-fold higher in men than in women, the sex-specific gap narrowing after the menopause (2). The lifetime risk of CAD at the age of 40 years is 1 in 2 for men and 1 in 3 for women (3). This male preponderance is remarkably consistent across 52 countries with hugely divergent rates of CAD mortality and lifestyles (4). The universality of this disparity makes it likely that there is an intrinsic sexual dimorphism in susceptibility to CAD that may involve genetic, hormonal, lifestyle or ageing factors. Sex hormones can influence a multitude of factors implicated in the pathogenesis of atherosclerosis and coronary artery disease - the traditional view being that androgens are harmful and estrogens beneficial. However, the sex-specific disparity in CAD may involve diverse mechanisms ranging from in utero sex hormone imprinting, gender-specific behaviour, distribution of visceral body fat, to vascular and myocardial structural/functional adaptation to ageing, pressure overload and disease (5). As the therapeutic indications for male androgen therapy widen to 'non-classical' indications (6) including male contraception, physiological ageing, chronic debilitating conditions and hormone replacement in postmenopausal women (7), it becomes increasingly pertinent to consider whether natural or induced changes in levels of testosterone (T) or dehydroepiandrosterone (DHEA) will impact on the risks of coronary artery disease in men and women. This question is one of the major safety issues for androgen therapy. This chapter synthesises data from Medline publications on a variety of disciplines into a global assessment of the relationship between androgens and CAD in men and women

TESTOSTERONE (T) AND CORONAARY ARTERY DISEASE

Observational clinical studies

It is important to emphasize the limitations of observational studies on the associations between serum levels of endogenous androgens and CAD. The CAD endpoints were extremely variable (mortality, morbidity such as myocardial infarction (MI) and angina, angiography, ultrasound, arterial calcification, post-mortem findings and unspecified 'cardiac events'). Study groups were heterogeneous in terms of age, number of subjects and selection criteria. Most CAD patients will be on medications and modified their lifestyle. In some studies, selection of poorly-matched controls may have introduced biases. The time interval from MI to sampling varied from 3 months to many years and was not always standardised for diurnal variation of hormone levels. Only a minority of studies adjusted for confounding factors such as smoking, blood pressure, obesity, diabetes, and dyslipidemia. Finally, chronic illnesses including CAD can lower serum levels of T.

Endogenous T and CAD in men

Cross-sectional studies

Of the 33 cross-sectional studies (8-40) investigating the relationships between circulating testosterone (T) and CAD in men, seventeen showed no association while sixteen found lower levels of T in cases compared to controls (Table 1). Many of the studies also did not have sufficient statistical power due to the small number of subjects investigated. Two of the largest studies (16, 24) found a negative relationship between T and ischaemic heart disease or abdominal aortic calcification. As with all cross-sectional studies, the directionality between cause and consequence and significance of the observed negative relationship is unclear especially when CAD or generalized atherosclerosis can lower serum levels of T.

Prospective cohort or case-control studies

Seven (41-47) out of eight non-cross sectional studies showed no significant relationship or predictive value between T and CAD (Table 1). The four prospective cohort studies followed 1009 Californian men aged 40-79 over 12 years (42), 2512 men aged 45-59 in South Wales (Caerphilly) for 5 years (45), 890 Baltimore men aged 53.8±16 yr for up to 31 years (47) and 282 Dutch men for 6.5 years (24). In the first three, there was no correlation between baseline T levels and subsequent development of fatal or non-fatal CAD, stroke or heart failure after adjusting for relevant confounders. In the last (Rotterdam) study, low T was associated with progression of abdominal aortic atherosclerosis detected by radiological calcification. In the 4 case-control studies, baseline T levels in cases of CAD and matched controls from the Multiple Risk Factors Inverventional Trial (41), Honolulu Heart Programme (43), Baltimore Longitudinal Study of Ageing (44) and the Helsinki Heart Study (46) did not predict CAD events during observation periods of 6-8, 19-20, 9.5 and 5 years respectively.

In summary, none of the studies, cross-sectional or longitudinal, showed a positive relationship between endogenous T and CAD to suggest that high levels of this androgen may be a risk factor. On the contrary, CAD may be more common in men with low T.

Endogenous T and CAD in women

In two prospective studies of postmenopausal women (Table 2a), T, free T, bioavailable T, and androstenedione did not differ significantly between those with and without heart disease, or abdominal aortic calcification, at baseline and did not predict cardiovascular death or progression of aortic atherosclerosis during follow-up (24, 48). In contrast, the three cross-sectional studies (24, 49, 50) showed inconsistent relationships between T and CAD.

Polycystic ovarian syndrome (PCOS)

Indirect evidence for the atherogenicity of androgens in women stems from cross-sectional data, which consistently showed a cluster of cardiovascular risk factors including obesity, insulin resistance, dyslipidaemia, hypertension and impaired fibrinolysis in patients with PCOS. Based on calculated risk profiles, women with PCOS were predicted to have a relative risk for myocardial infarction of 7.4:1 (51).

In 102 consecutive women undergoing cardiac catheterisation, Wild et al 1990 (52) found a positive correlation between angiographic evidence of coronary artery disease and clinical evidence of hyperandrogenism (Table 2b). Similarly, in 143 women aged £60 years referred because of chest pain or valvular heart disease, ultrasound evidence of polycystic ovaries (in 42% of patients) was associated with an increased number of stenosed coronary arteries (53). The prevalence of self-reported CAD was found to be significantly higher in 28 women (45-59 yr) who had undergone ovarian wedge resection over 18 years ago compared to 752 aged - matched controls (56). Using B-mode ultrasound to detect premature atherosclerosis, significantly increased carotid artery intima-media thickness was found in small numbers of middle-age (>45 yr) patients with PCOS compared to age-matched controls irrespective of BMI, fat distribution and other risk factors (54, 55). Increased calcification of the coronary arteries (which correlates with atherosclerosis) was found in 32 premenopausal (30-45 yr) women with PCOS compared with 52 age-matched controls using electron beam computed tomography (57). However, in 18 healthy obese young women (32.7±1.9 yr) with PCOS, endothelium-dependent and endothelium-independent vascular responses were normal compared to age-matched controls (58).

Cardiovascular mortality (59) and morbidity (60) over a period of 30 years in 786 out of 1028 women (over 45 years of age) diagnosed to have PCOS on histopathological and hospital in-patient diagnostic records between 1930-1979, most of whom underwent ovarian wedge resection, were compared with 1060 age-matched controls. Despite significantly increased diabetes, hypertension, cholesterol and non-fatal cerebrovascular disease, standardised mortality ratio of 1.4 (95% CI 0.75-2.40) (59) and the odds ratio of 1.5 (95% CI 0.7-2.9) (60) for CAD were not significantly increased. In another cohort of 346 non-obese patients aged 30.3-55.7 yr diagnosed to have PCOS in a specialist clinic, the prevalence of cardiac complaints (including serious heart disease or cardiac arrest) ascertained by telephone questionnaire was not significantly different from that in 8950 age-matched females in the general population (61). However, both these studies suffer from methodological drawbacks such as under - ascertainment of PCOS (59, 60) and the relative young age of the smaller cohort (61). In a prospective cohort of 82,439 women from the Nurses Health Study (62) followed for 14 yr, prior menstrual irregularity at aged 20-35 yr was associated with an increased combined incidence of non-fatal and fatal myocardial infarction. The age-adjusted relative risks for women reporting usually irregular cycles or very irregular cycles were 1.25 (1.07-1.47) and 1.67 (1.35-2.06) respectively compared with those with very regular menses. There was also a non-significant trend for increased risks of ischaemic stroke. Though not confirmed by clinical or hormonal androgen excess, the diagnosis of PCOS was the most likely explanation for the history of irregular menstrual cycles in the study group.

In summary, endogenous T is unlikely to have a causal or protective role for CAD in postmenopausal women. On the other hand, younger women of reproductive age with PCOS and/or menstrual irregularity often have adverse cardiovascular risk profiles predisposing them to premature heart disease, although the evidence from prospective studies supporting an actual increased incidence of CAD is limited and indirect. Whether this association is causally related to chronic hyperandrogenemia per se, as opposed to the associated metabolic variables prevalent in PCOS, remain unclear. Nevertheless, given the high prevalence of PCOS in the general female population, this unresolved issue should have a high priority for further research. In the mean time, screening for CAD risk factors and lifestyle advices should be considered for women of reproductive age presenting with irregular menstrual cycles and PCOS.

Interventional clinical studies

The potential consequence of decreasing endogenous or increasing circulating levels of T on CAD is discussed under various clinical interventional scenarios.

Endogenous androgen deprivation

A frequently cited but misquoted study (63) compared the life expectancy of 297 castrated inmates with 735 intact inmates (white males) in a single state institution for the mentally retarded. Castrated males lived an average of 13.6 years longer than intact controls. However, the excess mortality in intact inmates was due to infections with no difference in cardiovascular disease mortality between the two groups. The authors correctly concluded that post-pubertal castration did not decrease the frequency of deaths due to cardiovascular disease.

Cross-gender anti-androgen treatment in 816 male-to-female orchidectomised transexuals aged 18-86 years (64) by administration of ethinylestradiol 100µg/day and cyproterone acetate 100mg/day for 7734 patient-years was not associated with any significant difference in cardiovascular mortality or morbidity compared to the general male population despite a 20-fold increase in venous thromboembolic complications.

Androgen excess from anabolic steroid abuse

It must first be emphasised that pathological data from men abusing exotic anabolic-androgenic steroids (AAS) in doses several orders of magnitude higher than those prescribed in clinical practice should not be extrapolated to the legitimate therapeutic use of approved testosterone preparations or indeed to androgen physiology.

Excessive androgen exposure in men is uncommon in clinical practice. However, AAS abuse, previously only prevalent in athletes and body builders (65), is said to have reached epidemic proportion in the general population in recent years especially amongst adolescents (66, 67, 68, 69). In three reviews of the literature between 1987 and 2000 (70, 71, 72), there were altogether 17 case reports of cardiovascular events in young male body builders abusing pharmacological doses of multiple anabolic agents. There were 11 documented cases of acute myocardial infarction, 4 of cardiomyopathy and 2 of strokes. Postmortem examinations in 2 young body builders using AAS who died of acute myocardial infarction did not reveal any lesions in the coronary arteries (73). It is not possible to draw firm scientific conclusions about the cardiac toxicity or atherogenicity of AAS from these sporadic case reports, when the baseline denominator information on prevalence and extent of exposure is shrouded in uncertainty and secrecy. Nevertheless, it has been conjectured that dose-dependent androgen - induced vasospasm, platelet aggregation, activation of coagulation cascade, atherogenic lipid profiles (decreased HDL-cholesterol and increased LDL-cholesterol) and abnormal left ventricular function and hypertrophy are relevant mechanisms precipitating sudden cardiac deaths in young power athletes and body builders (71).

Exogenous T treatment in men with CAD

Uncontrolled studies from the 1940s, largely of historical interests only, suggested that T may improve symptomatic CAD in men. More recent data are considered below.

Bolus intravenous or intracoronary injections of pharmacological doses of T acutely improved myocardial ischaemia or induced coronary artery dilatation in a small number of men with CAD. (74-76). Whether this acute pharmacological action of T can be translated into a therapeutic effect remains to be determined.

In a randomised placebo-controlled double-blind study T cypionate 200mg i.m. weekly for 8 weeks decreased ST segment depression during exercise testing in 25 men with positive tests (77). In a placebo-controlled crossover (but not double-blinded) study in 62 elderly men with CAD, oral T undecanoate for 4 weeks improved subjective symptom scores and resting ECG (78). Transdermal testosterone 5mg daily for 12 weeks increased the time to 1-mm ST segment depression in men with symptomatic CAD (79).

These preliminary data suggest ECG changes can improve after (maximum of 12 weeks) short-term T supplement in CAD patients with low T levels. It is unclear if the effects of T are based on specific cardiac actions or non-specific improvements in skeletal muscle performance during exercise testing or a central action in the brain. Whether there are real symptomatic or functional benefits or decreased cardiovascular mortality from T treatment in the long term also remain uncertain.

Exogenous T treatment in women

There is burgeoning interest in the use of testosterone as part of HRT in natural and surgically-induced menopause, adrenal deficiency and hypopituitarism (7, 80, 81, 82), even though the clinical and biochemical definition of female androgen deficiency has not been clearly established (83, 84, 85). Whether administration of testosterone to females in physiological or pharmacological doses will impact on the cardiovascular system is currently unknown but there is insufficient evidence to exclude potentially adverse effects. In a 20-year retrospective survey of 293 female-to-male transexuals aged 17-70 year (mean 34) treated for 2 months to 41 years (total exposure of 2418 patient-years) with oral testosterone undecanoate 160mg daily or mixed testosterone esters (Sustanon) 250mg i.m. every 2 weeks, there was no excess of cardiovascular (or all cause) mortality or morbidity compared with the general female population (64).

In summary, interventional studies to decrease endogenous testosterone or administration of exogenous testosterone do not suggest a causal relationship between the level of testosterone exposure and development of CAD. On the other hand, some preliminary information hints at possible short-term beneficial effects on myocardial ischaemia. However, prospective controlled data on cardiovascular disease endpoints (myocardial infarction, angina, mortality) from large-scale interventional studies in either men or women using physiological doses of androgens are currently lacking.

DEHYDROEPIANDROSTERONE (DHEA) AND CORONAY ARTERY DISEASE (CAD)

DHEA and its sulphate DHEAS are weak but highly abundant adrenal androgens which show a progressive age-related decline in both men and women from the third decade onwards (86, 87). There is a body of opinion suggesting that DHEA supplementation may be beneficial to the elderly in a variety of physiological functions including the prevention of cardiovascular disease (88, 90). It has been implied that, against an androgenic milieu in men, DHEA acts as a prohormone for conversion to metabolites with predominantly oestrogenic effects and potentially anti-atherogenic actions (89, 91).

Observational Clinical Studies

DHEA and CAD in Men

Many observational studies in men have attempted to demonstrate a correlation between serum DHEAS levels with different CAD endpoints including the extent of atherosclerosis assessed by autopsy, coronary angiography, carotid vessel thickness/pulse wave, aortic calcification and clinical disease states including angina, myocardial infarction and mortality (Table 3a). These have shown either an inverse (33, 36, 38, 92-97) (mostly cross-sectional), null (21, 35, 40, 94, 97-103 24) or positive (27, 46) relationship between DHEAS levels and CAD (Table 3a).

Of the nested case control or prospective cohort studies, all but four (46, 92, 94, 97) showed no association between DHEAS levels and incident CAD (Table 3a). In the Helsinki Heart Study of middle-aged dyslipidemic men, higher DHEAS levels were associated with an increased risk of CAD (46). In the Honolulu Heart Study of 6000 men of Japanese descent followed for 18 years (94), low DHEA was associated with fatal but not non-fatal CAD. In the Rancho Bernardo cohort study, a preliminary report of 242 men also showed a negative relationship between DHEAS and CAD mortality (92). However, in the full analysis of the same study on 942 men over 19 years (97), there was only a modest negative relationship between DHEAS and those that survived their cardiac events but none with CAD mortality. DHEAS levels appear to be associated with increased mortality from all causes of death in men over the age of 50 (94, 97, 99), giving rise to the notion that this is a non-specific marker of poor health and lack of adaptive capacity to acute illnesses or a secondary phenomenon consequent upon various diseases of ageing such as malignancies and heart failure (87, 101, 104).

DHEA and CAD in Women

A negative relationship between DHEAS and CAD and atherosclerosis has been documented in cross-sectional studies in younger women (93, 95, 105, 50). In contrast, a positive association between serum levels of DHEAS and cardiovascular and CAD mortality was shown in preliminary analyses on 30 CAD deaths during a 12-year follow-up of 289 postmenopausal women 60 to 79 years of age in the Rancho Bernardo study (106) (Table 3b). However, in the 19-year follow-up of the full cohort of 942 Rancho Bernardo cohort (107), cardiovascular and CAD mortality were not associated with serum DHEA levels at baseline (Table 3b). This lack of association was confirmed by six other shorter prospective studies (24, 99-103) and one cross-sectional study (49).

Taken together, data from observational studies in men and women do not support the hypothesis that DHEAS 'deficiency' is a risk factor for CAD fatalities or that DHEA may confer an anti-atherogenic action in men or women. Low DHEA may be a non-specific marker for ill health in general.

ANDROGENS AND CORONARY ARTERY DISEASE (CAD) - ANIMAL STUDIES

The influence of androgens on the development and progression of experimentally-induced atherosclerosis has been investigated in animal models of diet-induced atherosclerosis and atherosclerosis-susceptible transgenic mouse models (Table 4).

Animal studies with T

Diet-induced atherosclerosis

Larsen (108) investigated the effects of i.m. T enanthate in castrated male rabbits fed on cholesterol-rich diet. There was no difference in the cholesterol content of abdominal aorta lesion after 17 weeks. Bruck et al (109) demonstrated sex-specific effects of T and oestradiol in castrated male and female rabbits fed an atherogenic diet. After 12 weeks, aortic arch intimal thickness was reduced by i.m. oestradiol valerate 1mg/kg/week in females but not males and T enanthate 25mg/kg/week in males but not females and by combined oestradiol and T administration in both sexes. Interestingly, T treatment in female rabbits increased plaque sizes but oestradiol had no effect in male rabbits. These sex-specific anti-atherogenic effects of T and oestradiol were independent of changes in plasma lipoproteins. Alexandersen (110) showed that castration per se in male rabbits resulted in a doubling of aortic atherosclerosis compared to sham-operated controls. This can be reversed by oral T undecanoate 80mg daily or DHEA 500mg daily via lipid-dependent mechanisms. In addition, i.m. T enanthate 25mg twice weekly, which raised circulating T levels by 10-fold, decreased aortic atherosclerosis. This suggests that androgens in pharmacological doses may exert effects on the vasculature that protects against atherosclerosis.

Treatment of male chicks with T resulted in a dose-dependent increase in aortic atherosclerosis (111). Similarly, T treatment in female ovariectomised cynomolgus monkeys fed an atherogenic diet for 24 months increased the extent of coronary atherosclerosis by two-fold compared to untreated intact and ovariectomised controls (112). These effects were independent of changes in lipid levels. Surprisingly, the acetylcholine-induced atherosclerosis-related coronary artery vasoconstriction was reversed by T treatment suggesting an improvement in endothelial reactivity (112). Thus, in contrast to the neutral or beneficial effects of androgens observed in male animals, experimentally induced hyperandrogenism may enhance atherosclerosis in female animals. In view of the major sexually-dimorphic action of T, data from experimentally-induced atheroslerosis in female animals should clearly not be extrapolated to males.

Transgenic mouse models

Two studies investigating the effect of castration and exogenous T on atherosclerosis of apoE-deficient mice yielded conflicting results. In the study of Elhage et al. (113), castration at the age of 4 weeks had no effect on atherosclerosis in either male or female mice. In both sexes, application of subcutaneous T pellets for 8 weeks significantly decreased serum levels of cholesterol and inhibited the development of fatty streak lesions in the sinus aortae by about 30%. In the study by von Dehn et al (114), chemical castration by a GnRH antagonist to suppress endogenous T led to a decrease in atherosclerosis in both the sinus aortae and the ascending aorta despite increased cholesterol in male and decreased HDL-C in female mice. Raising T levels by silastic T implants in male mice led to small but significant increases of cholesterol levels and increased atherosclerotic lesions. Increasing T by silastic implants in female mice induced fewer atherosclerotic lesions and no changes in lipids. The discrepancy between the two studies may have resulted from the higher dosages of T in the second study. Another study performed in LDL-receptor knock-out mice also found an anti-atherogenic effect of T which was blunted by the parallel use of an aromatase inhibitor (115). The anti-atherogenic effect was therefore ascribed to oestradiol rather than T.

Animal studies with DHEA

In contrast to the conflicting data on the effects of exogenous T on diet-induced atheroclerosis, DHEA administration to rabbits seems to consistently decrease atherosclerosis. Thus all 4 studies (Table 4) (110, 116-118) in intact or castrated male and female rabbits treated by DHEA for between 5-30 weeks showed a significant reduction in the extent of aortic or cardiac transplant atherosclerotic lesions independently of changes in lipids. DHEA may therefore be considered favourable under these rather artificial experimental conditions. Although oestrogens were not measured in these studies, it is probable that DHEA administered in pharmacological doses to animals (rabbits) with little endogenous adrenal androgen production, would be converted to oestrogenic metabolites with potent actions on the vascular endothelium. This is supported by the findings of Hiyashi et al (118) who demonstrated that the anti-atherogenic effects of DHEA in ovariectomised female rabbits was partially (50%) blocked by the aromatase inhibitor fadrozole. Together with the fact that no specific receptors for DHEA has yet been identified, it is plausible that this steroid primarily acts as a prohormone for more potent metabolites. One should be circumspect in extrapolating the apparent beneficial actions of DHEA suggested by these animal studies to man since exposure to similarly high pharmacological doses has not been investigated.

EFFECTS OF ANDROGENS ON CARDIOVASCULAR RISK FACTORS

The net effect of androgens on cardiovascular risk is difficult to assess for at least six reasons. First, the risk factors for cardiovascular disease are numerous and ever increasing. T can influence several risk factors simultaneously, some of which at first sight appear beneficial, (e.g. lowering lipoprotein(a) (Lp(a)), insulin, fibrinogen, and plasminogen activator type 1 (PAI-1)), while others are considered adverse (e.g. suppressing HDL-C). Second, endogenous T appear to have opposite effects on cardiovascular risk factors to that of exogenously administered T. Third, the associations between serum concentrations of endogenous T and cardiovascular risk factors are confounded by the complex interactions between endogenous androgens, body fat distribution, and insulin sensitivity. Fourth, a causal relationship between some of the aforementioned changes in risk factors and atherosclerosis has not been proven. Of particular importance is the example where exogenous T-induced suppression of HDL-C may not necessarily be accompanied by changes in cardiovascular risk. Fifth, T can exert its metabolic effects directly or via its metabolites E2 and dihydrotestosterone. The effects of T and E2, in particular, can be either be additive (e.g. on Lp(a)) or counter-regulatory (for example on HDL-C). Sixth, polymorphisms in the genes encoding the androgen receptor, sex hormone binding globulin (SHBG) and 5a-reductase regulate genomic effects and bioavailability of T and dihydrotestosterone, respectively. Thus, at any given serum concentration of T, the metabolic effects at individual target tissue sites are pleiotrophic and complex.

Associations of endogenous T with cardiovascular risk factors

Cross-sectional population studies had found statistically significant correlations between plasma levels of T and various risk factors, which however were in opposite directions in men and women.

Men

In men T plasma levels were frequently found to have positive correlations with serum levels of HDL-C as well as inverse correlations with plasma levels of triglycerides, total cholesterol, LDL-C, fibrinogen and PAI-1 (119, 120). However, serum levels of T have even stronger inverse correlations with BMI, waist circumference, waist-hip-ratio (WHR), amount of visceral fat and serum levels of leptin, insulin and free fatty acids. After adjustment for these measures of obesity and insulin resistance, the correlations of the cardiovascular risk factors with T but not with visceral fat or insulin lost their statistical significance (121, 122). These findings indicate that a low serum level of T in eugonadal men is a component of the metabolic syndrome, characterized by obesity, glucose intolerance or overt type 2 diabetes mellitus, arterial hypertension, hypertriglyceridemia, low HDL-C, a pro-coagulatory and anti-fibrinolytic state for which insulin resistance is thought to be a critical component. Therefore, the frequently observed association of high T levels with a more favourable cardiovascular risk factor profile in men probably does not reflect direct regulatory effects of T on lipoprotein metabolism and the haemostatic system. Insulin suppresses the production of SHBG so that insulin resistance is commonly associated with low levels of SHBG and low levels of total T (123). Thus, in some populations, when serum levels of free T instead of total T were correlated with lipids and other cardiovascular risk factors, these associations disappeared.

The relationship between androgens, body fat distribution, and insulin sensitivity, of which the latter two are also involved in the regulation of HDL and triglyceride metabolism, is complex (124). It is not clear whether androgens regulate adipose tissue and insulin sensitivity or, vice versa, whether adipocytes and insulin regulate T levels. Probably a bi-directional relationship exists. Morbidly obese and insulin resistant men frequently have low serum levels of T which increase upon weight loss (125). E2 levels show the opposite changes to T with obesity and weight loss. It has therefore been suggested that obesity causes hypotestosteronemia by increased aromatisation of T to E2 in the adipose tissue. Supporting a role of insulin in the determination of T levels in men, infusion of insulin during euglycemic clamp increased T levels in obese men but not in lean men (126). On the other hand, hypogonadal men are frequently obese with increased levels of leptin and insulin (127). Body weight, leptin levels and insulin levels decrease upon substitution of T in hypogonadal men (128, 129). Even treatment of eugonadal obese men with T led to a decrease of visceral fat mass and, in parallel, improved insulin sensitivity and corrected dyslipidemia. In the opposite experiment, suppression of T by the GnRH-antagonist cetrorelix increased serum levels of leptin and insulin (130). Moreover, men with a lower number of CAG repeats in the androgen receptor gene who are more testosterone-responsive have less body fat than those with a high number of CAG repeats (131). These data indicate that in men, the dominant action in the bi-directional relationship is that T reduces fat mass especially in the abdomen and improves insulin action. In agreement with this androgens activate the expression of a-adrenergic receptors, adenylate cyclase, protein kinase A and hormone sensitive lipase in adipocytes (132). As a result, T stimulates lipolysis and thereby reduces fat storage in adipocytes.

Women

The relationships between endogenous androgens and obesity, insulin and cardiovascular risk factors in women are opposite to those in men. In cross-sectional studies, serum levels of T were found to have significant positive correlations with BMI and leptin levels (133). Low serum levels of SHBG, which is an indirect measure of female hyperandrogenism, were associated with high BMI and WHR as well as with high serum levels of leptin and insulin and low serum levels of HDL-C (121). Moreover, in a prospective study, 20% of women with SHBG-levels below the 5th percentile developed diabetes mellitus type 2 within the 12-year follow-up period (134). Thus, in women, hyperandrogenism is a component of the metabolic/insulin resistance syndrome. In agreement with this, women with PCOS frequently present with hypercholesterolemia, low HDL-C, hypertriglyceridemia, elevated fibrinogen and PAI-1, and a family history of diabetes mellitus. Because many women with PCOS are overweight and most if not all are insulin resistant, it is unclear whether hyperandrogenism is secondary to obesity and insulin resistance or hyperandrogenaemia itself contributes to obesity and insulin resistance (135, 136).

Insulin sensitivity contributes to the pathogenesis of hyperandrogenemia in PCOS. Insulin stimulates androgen synthesis in the ovaries via its cognate receptor and the inositolglycan pathway (137). Since the ovaries remain sensitive to insulin when other tissues such as fat and muscle are resistant, hyperinsulinemia can augment the LH- and ACTH-dependent androgen production in insulin resistant women with PCOS (138). In support of this, treatment of insulin resistant PCOS patients with metformin or the insulin sensitizer troglitazone significantly decreased serum levels of insulin as well as T, independently of BMI or gonadotropin levels (139, 140). Concomitantly, plasma levels of HDL-C increased and plasma levels of PAI-1 decreased. These data imply that hyperinsulinemia contributes to the functional ovarian hyperandrogenism in PCOS. On the other hand, lowering androgen levels with GnRH agonists and androgen receptor blockade in hyperandrogenic women improved insulin sensitivity and lipid profile (141, 142). The magnitude of these changes however is less than that usually encountered in PCOS. Furthermore, short-term lowering of ovarian androgen secretion by laparoscopic ovarian diathermy did not alter insulin or lipid levels (143). Thus, in adult women with PCOS, androgen excess is probably only a contributory factor for insulin resistance and its metabolic consequences rather than the root cause. This however does not exclude the possibility that androgens may have an aetiological role in PCOS. For example, experiments in marmoset monkeys recently showed that transient intrauterine or perinatal exposure to T predisposed female animals to central adiposity and insulin resistance in adult life (61). Supraphysiological doses of exogenous T or other androgens to women (144) or female rats (145) and cynomolgus monkeys (146) increased BMI, visceral fat and muscle mass and decreased insulin sensitivity. Hence there appears to be a vicious circle where androgen exposure in early life contributes to insulin resistance in adulthood. The resulting hyperinsulinaemia is critical to the pathogenesis of PCOS and exacerbates the androgen excess.

Children and adolescents

Prepubertal boys and girls do not differ significantly in their serum lipid and lipoprotein levels. In contrast to girls, in whom levels of HDL-C and LDL-C change little with puberty, sexually maturing boys experience a decrease in HDL-C and increases in LDL-C and triglycerides (147). However, these changes may not reflect effects of sex hormones only since they are confounded by other endocrine changes, for example in the growth hormone axis, which also regulates lipoprotein metabolism.

Effects of exogenous T and DHEA on cardiovascular risk factors

The effects of exogenous androgen on cardiovascular risk factors varied with the dose, route, duration and type of treatment, as well as the age, gender and conditions of the recipients (for review see 148). The most consistent findings were decreases in plasma levels of HDL cholesterol, lipoprotein(a) (Lp(a)) and fibrinogen, which are accompanied by much less prominent declines of LDL cholesterol and triglycerides .

HDL

In the majority of studies, substitution of T in hypogonadal men decreased HDL-C levels (148). In normal men, supraphysiological doses of T or androgen-like anabolic steroids decreased HDL-C by about 20% or more. Conversely, surgical castration or medical castration by GnRH analogues increased HDL-C (148). In hypogonadal patients or in elderly men, substitution of testosterone led to minor or no decrease in HDL-C. In a recent meta-analysis of 19 studies (149), there seems to be trends suggesting that suppression of HDL-C is directly correlated with the dose of T but inversely related to the age and duration of treatment. Transdermal application of testosterone or dihydrotestosterone also exerted less effect on HDL-C than oral and intramuscular adminstration.

Since low HDL-C is an important CAD risk factor and HDL exerts several potentially anti-atherogenic actions, lowering of HDL-C by T treatment may potentially increase cardiovascular risks. However, the epidemiological association of low HDL-C with CAD has not been proven to represent a causal relationship. Instead, low HDL-C often coincides with other components of the metabolic syndrome, and may therefore merely be a surrogate marker for other linked pro-atherogenic condition(s). Moreover, in transgenic animal models, only increases of HDL-C induced by apoA-I overproduction, but not by inhibition of HDL catabolism, were consistently found to prevent atherosclerosis (150). Therefore, the mechanism of HDL modification rather than changes in levels of HDL-C per se, appear to determine the (anti)-atherogenicity of HDL modification (Figure 1). Two genes involved in the catabolism of HDL are up-regulated by T, namely scavenger receptor B1 (SR-B1) and hepatic lipase (HL). SR-B1 mediates the selective uptake of HDL lipids into hepatocytes and steroidogenic cells, including Sertoli and Leydig cells of the testes, as well as cholesterol efflux from peripheral cells including macrophages. T up-regulates SR-B1 in the human hepatocyte cell line HepG2 and in macrophages and thereby stimulates hepatic selective cholesterol uptake and peripheral cholesterol efflux, respectively (151). HL hydrolyses phospholipids on the surface of HDL thereby facilitating the selective uptake of HDL lipids by SR-B1. The activity of HL in postheparin plasma is increased after administration of exogenous T (152) and slightly decreased by suppression of T after GnRH antagonist treatment (130). Increasing both SR-B1 and HL activities are consistent with the HDL lowering effect of T. Interestingly, in transgenic mice, overexpression of SR-BI or HL caused a dramatic fall in HDL-C but inhibited rather than enhanced atherosclerosis (150). This again demonstrates the fallacy of extrapolating the HDL-C lowering effect of T to increased cardiovascular risk.

Figure 1. Pathways of HDL metabolism and regulation by testosterone and oestradiol (described from top center and anti-clockwise) Mature HDL3 and HDL2 are generated from lipid-free apoA-I or poorly lipidated particles (preb1-HDL) as the precursors. These precursors are produced as nascent HDL by the liver or intestine, or are released by lipolysis of triglyceride rich lipoproteins (TGRL = VLDL and chlyomicrons), or by interconversion of HDL3 and HDL2. ABCA1-mediated lipid efflux from cells is important for initial lipidation; LCAT-mediated esterification of cholesterol generates spherical particles which continue to grow upon ongoing cholesterol esterification, and PLTP-mediated particle fusion and surface remnant transfer. These mature HDL particles also continue to accept cellular cholesterol by processes which are facilitated by the scavenger receptor BI (SR-BI) and LCAT. Larger HDL2 are converted into smaller HDL3 upon CETP-mediated export of cholesteryl esters from HDL onto apoB-containing lipoproteins, SR-B1-mediated selective uptake of cholesteryl esters into liver and steroidogenic organs, and HL- and EL-mediated hydrolysis of phospholipids. HDL lipids are catabolized either separately from HDL proteins, i.e. by selective uptake or via CETP-transfer, or together with HDL proteins, ie. via uptake through as yet unknown HDL receptors or apoE receptors. Both the conversion of HDL2 into HDL3 and the PLTP-mediated conversion of HDL3 into HDL2 liberate lipid-free or poorly lipidated apoA-I, which is either re-used for the formation of mature HDL or is filtrated into the kidney. Grey arrows represent lipid transfer processes, black arrows represent protein transfer processes. The hepatic expression and activity of both HL and SR-B1 was shown to be up-regulated by testosterone and down-regulated by oestradiol. In addition oestradiol up-regulates the hepatic expression and secretion of apoA-I. These actions of testosterone and oestradiol are in good agreement with their lowering and increasing effect on HDL cholesterol, respectively. In addition both testosterone and oestradiol stimulate SR-BI expression in macrophages and thereby cholesterol efflux from these cells onto lipidated HDL.

Any effects of exogenous DHEA on cardiovascular risk factors appear to be marginal. In men aged 60 - 84 yr, DHEA 100mg daily for 3 months decreased total and HDL-cholesterol (153) but this was not confirmed in a larger study (154). In postmenopausal women, application of DHEA results in a slight reduction of HDL cholesterol (155). Female patients with Addison's disease administered oral DHEA 50mg daily for 3-4 months showed either no change (156) or a decrease in total and HDL-cholesterol (157) when a relatively greater increase in testosterone was induced by DHEA bioconversion.

Lipoprotein(a)

Results of many case-control studies and most prospective population studies demonstrated that lipoprotein(a) (Lp(a)) levels higher than 30 mg/dl are an independent risk factor for coronary, cerebrovascular, and peripheral atherosclerotic vessel diseases, especially if it coincides with other cardiovascular risk factors (158). Although Lp(a) levels are predominantly determined genetically, administration of T to men consistently decreased serum levels of Lp(a) significantly by 25% to 59% (148, 159). Conversely Lp(a) levels were increased by 40% to 60% in men in whom endogenous T was suppressed by treatment with the GnRH analogues (148, 160, 161). The Lp(a) lowering effect of T is independent of E2, which also reduces Lp(a) levels. It is not known how T regulates Lp(a). It is also not known whether changes in Lp(a) induced by T will affect cardiovascular risk.

The haemostatic system

In agreement with an important role of thrombus formation in the pathogenesis of acute coronary events and stroke, prospective studies have identified various haemostatic variables as cardiovascular risk factors (162), among them fibrinogen and the fibrinolysis inhibitor, PAI-1 or tissue plasminogen activator antigen. Administration of supraphysiological dosages of T to 32 healthy men in a trial of male contraception, led to a sustained decrease of fibrinogen by 15 to 20% over 52 weeks of treatment (163). In this study the doubling of T levels initially also led to significant decreases of PAI-1, protein S, and protein C as well as to increases of antithrombin and ??thromboglobulin. Likewise PAI-1 was decreased in men who received the anabolic androgen stanozolol. However, suppression of T in patients with benign prostate hypertrophy, by GnRH analogue treatment, exerted no significant effects on plasma fibrinogen levels (164). In agreement with the lowering effects of T on PAI-1, T inhibited the secretion of PAI-1 from bovine aortic endothelial cells in vitro. Taken together the current data indicate that T lowers fibrinogen and PAI-1. However, these anti-coagulatory and pro-fibrinolytic effects may be opposed by pro-aggregatory effects on platelets since high dosages of androgens were found to decrease cyclooxygenase activity and thereby increase platelet aggregation (165).

EFFECTS OF ANDROGENS ON THE FUNCTION OF VASCULAR CELLS

Effects of androgens on vascular reactivity

An early hallmark of atherosclerosis is decreased vascular responsiveness to various hormonal stimuli either due to endothelial dysfunction or due to endothelium-independent disturbances in vascular smooth muscle cell physiology. As a result, decreased vasodilation and enhanced vasoconstriction can lead to vasospasm and angina pectoris. Moreover, endothelial dysfunction also contributes to coronary events by promoting plaque rupture and thrombosis (166). Testosterone can induce vasodilation or vasoconstriction via endothelium-dependent or endothelium-independent mechanisms, and by genomic or non-genomic modes of actions. The diversity of these findings appears to be due to differences in species, gender, concomitant disease and, most importantly, dosage of T.

Suggestive of an adverse effect of T, nitrate-induced and hence endothelium-independent dilation of the brachial arteries was significantly reduced in female-to-male transsexuals taking high-dose androgens (167). In another case-control study, castrated patients with prostate cancer had a greater flow-mediated (i.e. endothelium-dependent) dilation of brachial arteries than controls, whereas the nitrate-induced vasodilation did not differ between the groups (168). A positive association between the number of CAG repeats in exon 1 of the androgen receptor gene and endothelium-dependent as well as endothelial-independent vasodilatation of the brachial arteries in healthy men. Thus higher sensitivity to T correlated with lower vascular reactivity in eugonadal males (169).

In contrast to these observational studies, acute interventional studies with intravenous administration of T to male patients with CAD revealed apparently beneficial vasodilatory effects of T (see above section on Exogenous T treatment in men with CAD). Likewise, in vivo studies in monkeys and dogs of both sexes as well as most in vitro studies with animal vessels suggest that T exerts beneficial effects on vascular reactivity. After T treatment for two years in ovariectomized female cynomolgus monkeys, intracoronary injections of acetylcholine caused significant endothelium-dependent vasodilation in treated but not in untreated animals. In contrast, endothelium-independent vasodilation occurred normally in both groups (112). In dogs, T induced vasodilation of coronary arteries by endothelium-dependent and independent mechanisms (170, 171). In vitro studies with isolated rings of coronary arteries and/or aortas from rats, rabbits, and pigs also found that, in both sexes, T improved both endothelium-dependent and/or endothelium-independent vascular responsiveness (170-172). However, it must be emphasized that all these studies employed supraphysiological doses of T in the micromolar range. Teoh and colleagues (173) observed a direct vasodilatory effect of T on porcine coronary artery rings at micromolar but no effect at nanomolar dosages. In contrast, physiological doses of T inhibited the vasodilatory effects of bradykinin and calcium ionophores. Similarly, T inhibited the adenosine-mediated vasodilation of rat coronary arteries and impaired endothelium-dependent relaxation of aortic rings from rabbits which were either made hypercholesterolemic or exposed to tobacco smoke (174-176).

The cellular and molecular mechanisms by which T (and E2) regulate the vascular tone are little understood (Figure 2). Evidence for and against endothelium-dependent or endothelium-independent mechanisms have been found. Results of some studies suggest the involvement of endothelial nitric oxide (170, 171, 177). In dog coronary arteries, rat aorta, and rat cerebral arteries, the nitric oxide synthase inhibitor L-NMMA prevented T-induced vasodilation. However, in another in vitro study L-NMMA had no effect on T-induced vasodilation of rabbit aortas and coronary arteries (172). In agreement with the latter, in vitro expression of nitric oxide synthase in human aortic endothelial cells was stimulated by E2 but not by T (178). The involvement of prostaglandins is suggested by the observations that T increases the response of coronary arteries to prostaglandin F2? (175) and that dihydrotestosterone increases the density of thromboxane receptors in rats and guinea pigs cultured vascular smooth muscle cells (179). However, in some in vivo and in vitro animal studies, pretreatment with the prostaglandin synthesis inhibitor, indomethacin, had no effect on T-induced vasodilation, so that the role of eicosanoids mediating the actions of T on the arterial wall is still controversial.

It is unclear whether T regulates vasoreactivity by genomic or non-genomic effects or both (Figure 2). Androgen receptor expression has been found in rat aortic smooth muscle and endothelial cells. Steroid hormones can also regulate vascular tone by non-genomic mechanisms which involve plasma membrane steroid receptors as well as modulation of cell membrane channels, e.g. ATP-sensitive, voltage-dependent and calcium-activated potassium channels. Several observations suggest that T, especially in supraphysiological doses, modulates vascular tone via non-genomic modes of action and/or its secondary metabolites e.g. E2. First, the androgen receptor antagonists, flutamide or cyproterone acetate, did not inhibit the effects of T on rabbit or pig coronary arteries (172, 173). Second, the expression of aromatase in vascular endothelial cells (180) raises the possibility that T can be converted to E2 and exerts its vasoactive effects via estrogen receptors. However, neither the aromatase inhibitor aminogluthemide nor the estrogen-receptor antagonist ICI 182,780 prevented the T-induced vasodilation (170). Third, barium chloride attenuated the T-induced vasorelaxation of rabbit aortas and coronary arteries indicating that T modulates the opening of potassium channels in vascular smooth muscle cells (172).

Effects of T on macrophage functions

Circulating monocytes migrate into the vascular wall and differentiate into macrophages. They bind to modified lipoproteins which have permeated the vascular endothelium. The uptake of modified (e.g. by oxidation) lipoproteins by macrophages leads to the formation of large foam cells. These, together with T-lymphocytes, release inflammatory mediators which stimulate the proliferation and migration of smooth muscle cells. Human monocyte-derived macrophages express the androgen receptor in a gender-specific manner. Macrophages of male donors exhibit a fourfold higher expression of the androgen receptor than macrophages of female donors (181). There is also evidence that T regulates macrophage function by non-genomic effects via a G-protein-coupled, agonist-sequestrable plasma membrane receptor which initiates calcium- and 1,4,5-trisphosphate-signaling pathways (182).

Unregulated uptake of oxidatively modified lipoproteins via type A scavenger receptors leads to the intracellular accumulation of cholesteryl esters in macrophages and to foam cell formation (166, 183). E2 inhibits oxidation of LDL both in the presence and absence of cells including macrophages. By contrast, T increases the oxidation of LDL by placental macrophages in vitro (184). Moreover, dihydrotestosterone dose-dependently stimulates the uptake of acetylated LDL by scavenger receptor type A and, hence, the intracellular cholesteryl ester accumulation in macrophages. In addition to the higher expression of the androgen receptor in male donors, this effect was only seen in macrophages of male but not female donors. The stimulatory effect of dihydrotestosterone was blocked by the androgen receptor antagonist hydroxyflutamide (181).

Non-hepatic and non-steroidogenic cells cannot metabolize cholesterol and, therefore, can only dispose excess cholesterol by secretion. Cholesterol efflux from cells is hence central to the regulation of the cellular cholesterol homeostasis. Non-specific and passive (i.e. aqueous diffusion) as well as specific and active processes (i.e. receptor-mediated) are involved. To date, two plasma membrane proteins are known to facilitate cholesterol efflux. Interaction of the scavenger receptor B1 with mature lipid-containing HDL is thought to facilitate cholesterol efflux by re-organizing the distribution of cholesterol within bilayer plasma membrane (Figure 1). The ATP binding cassette transporter A1 mediates phospholipid and cholesterol efflux to extracellular lipid-free apolipoproteins by translocating these lipids from intracellular compartments to the plasma membrane and/or by forming a pore within the plasma membrane, through which the lipids are secreted (150). We have found that T up-regulates the expression of the scavenger receptor B1 in human monocyte-derived macrophages thereby stimulating HDL-induced cholesterol efflux. No effect of T was seen on the expression of the ATP binding cassette transporter A1 (151).

Effects of T on arterial smooth muscle function

Rat arterial smooth muscle cells express androgen receptors. In support of a physiological role, there is some evidence that T modulates endothelium-independent vasoreactivity (see section 6.1). Arterial smooth muscle cells also play an important role in atherosclerosis by proliferation, migration and matrix production (166). Whereas E2 can inhibit proliferation and migration of smooth muscle cells, T had no effect (185, 186). Moreover, the protection of female rabbits by E2, but not the protection of male rabbits by T, from atherosclerosis was associated with decreased incorporation of 5“-bromo-2“-deoxyuridine into DNA of neointimal cells, an in vivo marker of arterial smooth muscle cell proliferation (109).

OESTROGEN AND CAD IN MEN

There is compelling evidence indicating that an increasing number of physiological actions of testosterone in men are mediated by the oestrogen receptors (ERs) following conversion to oestradiol by site-specific aromatases in target tissues (187). The existence of two nuclear ER subtypes a and b as well as a membrane ER encoded by the same transcript as the nuclear receptor attest to the potential for many different biological estrogen effects. ERa, ERb and aromatase are detectable in coronary arteries of monkey and man (180, 188, 189). The extra-glandular production of estrogens (with circulating androgens as the immediate precursor substrate) may therefore play a role in male cardiovascular physiology and pathophysiology.

The importance of locally-produced oestrogens from aromatisation of testosterone in males for cardiovascular health has been highlighted by recent human and transgenic mouse models of aromatase deficiency and oestrogen resistance. In two men with undetectable circulating oestradiol and oestrone and high testosterone due to P450 aromatase deficiency (190, 191), dyslipidaemia with elevated total and LDL-cholesterol and triglyceride and decreased HDL-cholesterol was associated with insulin resistance (in the first patient only). These metabolic abnormalities were correctable by low dose oral or transdermal oestrogen replacement. Insulin resistance, acanthosis nigricans, low HDl-cholesterol, and impaired glucose tolerance were apparent in a 28 year old male with a null mutation in ER? gene causing oestrogen resistance (192). Ultrafast electron bean CT imaging showed calcium deposition in the proximal left anterior descending coronary artery indicating the presence of premature atherosclerosis (193). Flow-mediated brachial artery endothelial dependent NO-activated vasodilatation (membrane ER-mediated) in response to hyperaemia was absent showing marked endothelial dysfunction (194).

These clinical findings suggest that oestrogens are important in maintaining normal carbohydrate and lipid metabolism as well as normal endothelial-dependent NO-mediated vasodilatation in men. They are compatible with data from transgenic knockout models confirming that ERa is important in preventing adipocyt