Age‐related changes in testosterone and the role of replacement therapy in older men

Western populations represent an ageing society with continuing gains being made in life expectancy; a male born in Australia has a healthy life expectancy of 70 years with slightly lower expectations for those born in the UK and USA (WHO, 2001). The search for ways to maintain and enhance physical health and well-being is endless. Rather than aiming to simply live longer, people aspire to undertake 'active ageing' (WHO, 2002), with an emphasis on quality of life. Attention has been focussed on the possibility that testosterone replacement therapy may be beneficial to the quality of life of men in middle age and beyond, akin to that of the role of oestrogen replacement for symptomatic peri- and postmenopausal women (Hlatky et al., 2002). Although the clinical presentation of androgen deficiency in men with primary or secondary testicular failure is well recognized and the benefits of treatment well established (Allan Rossouw et al., 2002) serve to remind us of the need for properly designed and powered randomized, placebo-controlled data, and for caution in extrapolating cross-sectional or surrogate end-point data. We will discuss the extent of age-related decline in testosterone and the confounding effects of concomitant illness, the physiology of the changes in the hypothalamo–pituitary–testicular (HPT) axis, and the approaches to the laboratory assessment of hypoandrogenism in older men. The effects of androgens on key target tissues and the existing data from the limited number of placebo-controlled trials regarding the benefits and potential risks of its usage will be discussed. Unlike the predictable 90% fall in serum oestradiol across the menopause (Burger et al., 2002), testosterone levels in men begin to decline in the late third or early fourth decade and diminish at a constant rate thereafter (Baker et al., 1976b; Leifke et al., 2000; Harman et al., 2001). Many cross-sectional and several longitudinal studies have now documented significantly lower total testosterone levels in older men when compared to their younger counterparts (Feldman et al., 2002; Zumoff et al., 1982; Gray et al., 1991a, 1991b; Tenover Simon et al., 1992; Morley et al., 1997; Zmuda et al., 1997; Ferrini Harman et al., 2001). Those studies that have failed to identify this fall in testosterone have been criticized for the timing of sample collection (Harman Nieschlag et al., 1982). The absolute rate of decline in total testosterone levels in longitudinal studies varies threefold – from 0·11 nm (Harman et al., 2001) to 0·29 nm (Feldman et al., 2002) to 0·38 nm (Morley et al., 1997) per year. As a percentage of baseline values, cross-sectional studies estimate the rate of decline to be 0·5–0·8% per year (Davidson et al., 1983; Gray et al., 1991b; Simon et al., 1992), while longitudinal studies suggest a greater rate of decline, e.g. 1·6% per annum in the Massachusetts Male Ageing Study (MMAS; Feldman et al., 2002). Concomitant with the fall in total testosterone is a rise in SHBG levels with ageing (Field et al., 1994; Harman et al., 2001), estimated to be 1·3% per year in a cohort aged 40 years and over (Feldman et al., 2002). As a result of the combined effects of a rise in SHBG and a fall in total testosterone, calculated free testosterone levels decrease by approximately 2–3% per year (Feldman et al., 2002). The relationship between the changes in total testosterone and SHBG is unclear. Excluding obese men, in whom SHBG levels are reduced as a function of adiposity (Field et al., 1994), one group has described a fall in total testosterone from the age of 30 years with SHBG levels rising only from the age of 50 years (Leifke et al., 2000), whereas others have identified an earlier rise in SHBG with total testosterone levels falling only from the age of 55 years (Vermeulen et al., 1996). Ageing and obesity have opposing effects on SHBG levels. Given the increasing prevalence of obesity (Ball et al., 2002; 65% of Australian men are now considered to be overweight or obese) and the lack of understanding about the biological importance of indices of free testosterone, determining the prevalence of hypoandrogenism in ageing men is complex. Observational data on total testosterone levels across ages, reported in 1984 (Deslypere however, it is important to appreciate that falling testosterone levels in ageing men may not elicit a compensatory LH response (Kaufman Gupta et al., 2000; Diver et al., 2003). This diurnal pattern may be attenuated as a function of ageing (Bremner et al., 1983; Gupta et al., 2000) but as the loss is not universal (Diver et al., 2003), testosterone levels should be taken in the morning. Testosterone circulates 98% protein-bound (54% with low affinity to albumin and other proteins, and 44% with high affinity to SHBG; Sodergard et al., 1982) with only 2% as free testosterone. By analogy with other steroid hormones it has been proposed that free testosterone will relate to biological effect on tissue more than the total testosterone, and a number of methods for estimating this free fraction have been described (Morley et al., 2002). The clinical utility of this 'free hormone hypothesis' is unproven and no consensus exists on the optimum serum marker for testosterone (Anonymous, 2001). Although it is intrinsically appealing to suggest that free testosterone is a better index of androgen action (akin to the measurement of free rather than total thyroid hormone levels), direct evidence for this proposition is minimal. Limited observational data suggest a better correlation between muscle strength and either bioavailable testosterone (Perry et al., 2000) or a calculated free testosterone index (Roy et al., 2002) than with total testosterone. One prospective controlled study (Kenny et al., 2001) reported that bioavailable testosterone correlated better with the effect of testosterone supplementation on lean body mass (LBM) than did total testosterone levels. Nonetheless, the vast majority of data relate to total testosterone and it is this measure that primarily serves as the basis for making clinical decisions. The measurement of the free testosterone fraction by equilibrium dialysis is considered to be the 'gold standard' (Vermeulen et al., 1999) but it is time-consuming and expensive, and requires an estimation of total testosterone levels. Direct estimations by an analogue ligand radioimmunoassay (RIA) technique correlate poorly with equilibrium dialysis, significantly underestimating the actual free testosterone level, and their use cannot be recommended (Rosner, 2001). Bioavailable (also called 'non-SHBG bound' and comprising free and albumin-bound fractions) testosterone measurements are determined by RIA of the supernatant obtained after precipitation of SHBG-bound testosterone with 50% ammonium sulphate (O'Connor et al., 1973), and correlate well with estimates obtained by equilibrium dialysis (Vermeulen et al., 1999). These measures cannot be automated. Bioavailable testosterone values also show a circadian rhythmicity that is blunted with ageing (Plymate et al., 1989) and are subject to the same week-to-week variability seen with total testosterone levels (Morley et al., 2002). Calculations have been devised to determine the free component of testosterone based on the total levels of testosterone, SHBG and albumin, and make the reasonable assumption that the affinity constants for testosterone binding to both proteins are constant (Vermeulen et al., 1999). These values are usually described as the 'calculated free testosterone' levels and show a very good correlation with equilibrium dialysis measures. They are readily performed using standard assays and a computer algorithm. To date, however, there are no published data of population-based reference ranges. It should be noted that the free androgen index (FAI) is a derived unit-less calculation based on the formula FAI = (total testosterone/SHBG) × 100% and although validated for use in women is not applicable in men (Kapoor et al., 1993). The usefulness of serum LH as an 'internal' marker of androgen sufficiency is limited. Elevated levels support the diagnosis of testosterone deficiency in primary testicular failure, but many older men will have a serum LH within the young reference range along with a borderline low serum testosterone. Whether or not such older men are in fact eugonadal, as opposed to having a deficiency in hypothalamo–pituitary 'recognition' of actual deficiency (potentially implying benefit from testosterone therapy), is not clear. Testosterone treatment of the symptomatic 'low-normal testosterone/normal LH men' may seem attractive to some but lacks an evidence base and an understanding of its safety. Investigation of secondary testosterone deficiency (i.e. actual LH deficiency) must be considered as pituitary tumours and haemochromatosis (in particular) may occur in older men, and whilst rare, require specific evaluation (anterior pituitary hormone assessment, serum iron studies, and imaging) and treatment. The prevalence of androgen deficiency in the ageing male population is difficult to estimate due to the heterogeneity of the studied populations, the differing methods of estimating testosterone levels (total, free estimates) and the lack of consistency of nominal values for defining biochemical hypoandrogenism. Prevalence estimates are often based on the assumption that ageing men should be regarded as hypoandrogenic (synonymous with deserving of replacement therapy) when their testosterone measures fall below the lower limit of the healthy young adult male reference range. However, the latter is also subject to variability due to differences in assay platforms and reference ranges. Using differing and somewhat arbitrary cut-off points, one can obtain widely differing prevalence rates. The characteristics of the cohort of older men studied are critical in determining the reported prevalence of hypoandrogenism. In a study of institutionalized men aged 46–89 years almost 30% were classified as testosterone-deficient (defined as total testosterone levels below 10·4 nm; Swartz Kaufman Harman et al., 2001; Tenover, 2000). If the definition of hypoandrogenism is restricted to a total testosterone level < 8·7 nm, then only approximately 8% of these healthy cohorts would be classified as deficient (Tenover, 2000). If, however, free testosterone, as measured by equilibrium dialysis, is taken as the biological meaningful parameter, almost 1/3 of men classified as normal according to their total testosterone levels would be incorrectly labelled (Morley et al., 2002). The effects of ageing on testosterone levels may be confounded by other variables that must be considered in the assessment of androgen status. Older men recruited from medical clinics have lower testosterone levels than community-dwelling populations, suggesting that concurrent ill health depresses testosterone levels (McKinlay et al., 1989). Chronic illness, prescription medication, obesity or excessive alcohol were associated with a reduction in testosterone of 10–15% in all age cohorts in the MMAS men aged 40 years or over during 7–10 years of follow-up (Feldman et al., 2002). A direct effect on serum testosterone levels may be seen with medications such as opiates (change in LH pulsatility) or anticonvulsants (hepatic enzyme induction) (Zitzmann Dong et al., 1992; Handelsman, 1994). In a group of men aged 75 years or over admitted to hospital with acute illness, testosterone levels were 30% lower than matched controls and returned to normal upon recovery (Impallomeni et al., 1994). Both benign and malignant lung disease reduce testosterone levels irrespective of age, the effect being more marked in those with lung cancer (Blackman et al., 1988). Untreated diabetic men aged 50 years and over have also been found to have 15% lower testosterone levels than their body mass index (BMI)-matched, nondiabetic counterparts (Barrett-Connor, 1992). The effects of cardiovascular disease, depression and malignant prostate disease upon serum testosterone levels are dealt with below. The MMAS identified obesity as the most important determinant of total testosterone over time, with levels 25% lower in obese men when compared to their nonobese counterparts (Gray et al., 1991b). In this cohort, controlling for age, there was a 33% reduction in testosterone levels in the highest quintile BMI group when compared to the lowest (Field et al., 1994). Over 9 years of follow-up in these men obesity predicted a greater decline in total testosterone and SHBG with ageing (Derby et al., 2002). Whilst the lower SHBG levels associated with obesity contribute to low total testosterone levels, free testosterone levels may also be lower in obese as compared to nonobese men (Vermeulen et al., 1996). The effects of smoking on total testosterone levels are contradictory and the precise mechanisms of any effect are unclear. Elevated levels in current smokers in the order of 9–25% have been reported in middle-aged and older men (Deslypere Field et al., 1994; Vermeulen et al., 1996) but others have identified an inverse relationship between cigarette smoking and testosterone levels (Zmuda et al., 1997; Hsieh et al., 1998) while no effect was seen in current smokers comprising 18% of a cohort followed longitudinally in the Baltimore Ageing Study (Harman et al., 2001). The effects of alcohol upon testosterone are dependent upon the pattern and duration of usage. Acutely alcohol inhibits testosterone production (Gordon et al., 1976; Frias et al., 2002) whilst sustained stable alcohol intake in healthy older men does not influence total testosterone levels (Harman et al., 2001; Sparrow et al., 1980). Chronic alcoholic liver disease may effect androgen metabolism (Lester & Van Thiel, 1977) and elevate SHBG levels (thereby raising total testosterone levels) (Gluud, 1988) in addition to its well-known association with hypogonadism (Baker et al., 1976a). No correlation between exercise patterns and testosterone levels has been proven in population-based studies (Svartberg et al., 2003) perhaps related to the inaccuracies of men over a of years have baseline et al., 2001) and et al., 1998) testosterone levels when compared to their more but data regarding BMI was not in healthy older men has been to testosterone (Zmuda et al., but to a degree than seen in younger men et al., and this effect to be with et al., 2000). The is often marked by a group of et al., 2000) that women to seek from their men, the relationship between and is more complex. The age-related decline in serum testosterone to be associated with 1993). of these and physical and has been (Morley et al., 2000), and which both and variables et al., 2000). the and of these for low testosterone levels is low and has been properly There remains a lack of and validated for men most to be testosterone The must consider whether some or all of the are due to testosterone as opposed to other and consider these in with the testosterone levels. The diagnosis may be with low testosterone levels in an clinical However, the testosterone levels are often and the clinical diagnosis of androgen deficiency in the older then of the benefits and potential and risks from testosterone the increasing prescription of testosterone for a range of in ageing men very properly randomized, placebo-controlled studies have been performed & 2003) approximately men. These studies are in the of these trials is limited by the and of testosterone the duration of treatment only for and the of cohorts with a range of baseline testosterone levels. The effects of testosterone on key tissues of in men are below. The prescribing of testosterone in ageing men to effect for specific benefits increased muscle mass and has not been studied with to or quality of and cannot be However, one must in the with testosterone one can replacement physiology As demonstrated in data 2003), body can be in both young and older men by testosterone treatments serum levels across the the range. The relationship between the of cardiovascular disease and serum testosterone levels remains uncertain. It is that testosterone supplementation a for the of disease but data also support the that it may the cardiovascular of ageing men. The basis for the that androgens have a effect is derived from the association between male and of cardiovascular & and the use of with et al., studies have failed to identify a relationship between serum testosterone and the of in middle-aged and older men et al., 1996) but testosterone was only measured at baseline and the of other recognized cardiovascular factors on could not be Recent observational data et al., 2002) and studies of testosterone supplementation on surrogate of cardiovascular (Kenny et al., suggest that lower testosterone levels may greater of the between androgens and has been published et al., & 2003). These critical of the and studies of the effects of endogenous and androgens confirm the contradictory of current but suggest that the use of testosterone in older men should not be limited by of increasing cardiovascular & 2003). We therefore this on the clinical data to older men with age-associated declines in testosterone levels. studies of total testosterone in men with (as determined by clinical – have either failed to show a relationship or a negative association & 2003). a negative association was for both total and bioavailable testosterone in men with determined disease et al., 2000). prospective cohort studies of approximately middle-aged and older men & 2003) have no association between endogenous testosterone and after a follow-up However, an inverse association between serum testosterone and was found in a study of men (mean age after for age, BMI and multiple cardiovascular factors et al., 2002). The controlled trials of testosterone supplementation have not been of sufficient duration to determine effect on of cardiovascular disease and there is only limited about surrogate for a measure of function and a recognized marker for was not by either 12 months of testosterone (Kenny et al., or months of et al., 2001) therapy in older men with baseline total testosterone levels nm, although only small numbers were men total testosterone for months with testosterone or hCG did show a in function (Zitzmann et al., but the inverse relationship between of function and serum testosterone levels only in the and not in the range. The of these in older men with testosterone levels is unclear. Several groups have reported a beneficial effect of testosterone treatment on function in men with documented and et al., 2000). Although some et al., 1999), a serum testosterone levels from to reduced with the benefit seen in those subjects with the lowest baseline bioavailable testosterone levels et al., 2000). testosterone symptomatic and in older men with & 1993). These data are but are limited by small subject numbers and are the that testosterone a et al., a of cardiovascular was not by a (Zitzmann et al., and of were in both testosterone levels have also been to be correlated with et al., 1993). of men aged 60 years and over with baseline testosterone levels < nm with either or for months did not serum associated with et al., 2002). the effects of androgens on factors related to cardiovascular 1996) and may be by both the and the baseline of the The correlation of serum testosterone levels with established cardiovascular factors any of the association of testosterone with a negative association with et al., et al., 1992), (Simon et al., 1997) and adiposity et al., and an association with et al., A study of middle-aged men classified by total testosterone levels found that and total and levels were related to testosterone but after for measures of adiposity and only levels and remained significantly correlated with testosterone (Simon et al., 1997). In men matched for age and 12 weeks of testosterone and increased when compared to (Simon et al., 2001) and 9 months of a increased although levels were et al., 1993). the other hCG therapy for 12 weeks in an older cohort did not affect et al., 2002). The effects of testosterone therapy on