Testosterone and male ageing: spinning the wheels

Two important research articles, published in a recent issue of the New England Journal of Medicine (NEJM), bracket the topic of prescribing testosterone for older men.1,2 Testosterone treatment for older men is based on considering male ageing as analogous to either menopause or pathologically based hypogonadism. The former is a false analogy — menopause has a unique natural history featuring complete failure of female reproductive hormones in mid-adult life, contrary to all other human hormonal systems which decline gradually and modestly with ageing. The latter is based on the superficial resemblance of non-specific symptoms in ageing men with those of most hormonal deficiencies or chronic diseases. This “andropause hypothesis” is not well substantiated, with the 2004 United States Institute of Medicine’s authoritative review3 concluding that available evidence did not justify a major study of testosterone replacement in older men comparable to the Women’s Health Initiative (WHI) study of oestrogen replacement in menopause. Nevertheless, the past two decades have seen an approximately 20-fold increase in testosterone prescribing despite no proven new indications. This is largely confined to the US, with minimal changes in Australia4 and other regional markets; however, that bandwagon could certainly be viewed as having left the station, fuelled by heavy direct-to-public drug advertising in the US.

One of the NEJM articles reports the European Male Ageing Study (EMAS) — a large observational study of male ageing involving more than 3300 men aged 40 years and over from population-based sources in eight European cities, and the European counterpart to the seminal Massachusetts Male Ageing Study (MMAS).1 The EMAS evaluated the relationship of non-specific physical and mental symptoms to serum testosterone levels, an approach originating from clinic-based studies5,6 and used in the analysis of the population-based data of the MMAS,7 to which the EMAS adds a large sample size.

Crucially, the statistical power of this study can neither overcome its logical flaws nor the inconsistency between its findings and conclusions. After noting statistically insignificant associations of low testosterone levels with clinically relevant physical or psychological features, the researchers focus on three sexual symptoms — erectile dysfunction, frequency of morning erections and sexual desire. Each symptom shows a weak association with serum testosterone levels, featuring a shallow breakpoint (at levels of 8–11 nmol/L) together with high false-positive and negative rates. Although the only consistent significant association is between all three sexual symptoms and a serum testosterone level threshold of 8 nmol/L, the researchers inexplicably propose a “definition” of “late-onset hypogonadism” requiring the presence of all three sexual symptoms plus a serum testosterone level of less than 11 nmol/L. The proposal is further undermined by the study’s findings that all the associations of the three sexual symptoms with low serum testosterone levels are nullified by adjustment for age, obesity and co-existing illnesses, indicating that they are attributable to confounding rather than any authentic correlation.

The study’s authors, in effect, overinterpret cross-sectional data to imply causality. Such quasi-longitudinal interpretation is especially unsafe when blood test results are not steady for the population. This is the case for serum testosterone, with evidence of downward temporal trends in America and Europe,8 probably due to progressive population increases in obesity. The impact of implied causality should not be underestimated — despite its ritual caveats against testosterone prescribing, this article is likely to encourage more overuse of testosterone.

The article resonates with past mistakes, notably the decades of excessive oestrogen prescribing, encouraged by overinterpreted observational studies and curbed only by the first placebo-controlled randomised trial of hormonal replacement therapy. Lest we sleepwalk down that same path, let us hope that it will not take decades before the “andropause hypothesis” undergoes rigorous testing.

The second NEJM article reports the early termination of a randomised, placebo-controlled clinical trial for excess adverse cardiovascular effects associated with testosterone supplementation.2 The study evaluated the somatic benefits of 6 months of daily use of testosterone gel in frail men over 65 years with low serum testosterone levels. As expected, this population had a high prevalence of cardiovascular disease, which would explain their low serum testosterone levels as a non-specific effect of chronic illness. The progressive excess of adverse cardiovascular effects in testosterone-treated men (23 men receiving testosterone v 5 receiving placebo in the trial’s total of 209 men) was unexpected but persisted despite the use of various definitions of adverse cardiovascular events (the original definition was broad and included unexplained syncope and peripheral oedema). As the study design was conventional in regard to testosterone dosage, titration and monitoring, these findings show a low cardiovascular safety margin for testosterone supplementation in frail older men. They differ from the findings of previous comparable placebo-controlled studies of testosterone use in older men, which reported no excess cardiovascular events,9 even in 12-month studies of men with cardiac failure.10 Nevertheless, these adverse findings cannot be considered surprising given the earlier onset and greater severity of cardiovascular disease in men, together with the refutation of the long-dominant hypothesis that oestrogen provides women with a degree of cardiovascular protection.11

As this second article2 highlights, the effects of a treatment that excessively increases risk for the most common cause of death — cardiovascular disease — overwhelm even substantial improvements in less common or non-fatal disorders. Hence, the study’s finding of a benefit (improved limb muscular strength) was overshadowed by adverse cardiovascular profiles, as has happened with other treatments, such as the WHI’s reductions in bone fractures and colorectal cancer, and the highly effective, gastric-sparing cyclooxygenase-2 (COX-2) inhibiting analgesics and the oral antidiabetic glitazones. A corollary is that overinterpreting the regular association of low serum testosterone in men with cardiovascular disease as a risk factor that might be ameliorated — rather than as a consequence — led to a misplaced focus on prostate cancer as the major risk of testosterone treatment in older men.

Taken together, these two studies can be construed as pressing the accelerator and the brakes at the same time on testosterone prescribing for older men . . . with probably the usual effect of spinning wheels. However, several reflections arise for Australian clinical practice.

First, Australia’s national guidelines for testosterone use,12 developed in 2000 and adopted by the Pharmaceutical Benefits Scheme as the criteria for subsidised testosterone prescriptions, are vindicated. They aim to restrict testosterone prescribing for age-related androgen deficiency without hindering it for pathologically based androgen deficiency. European and US guidelines (produced in 2005 and 2006, respectively) were republished recently, essentially unchanged.13,14 The northern hemisphere guidelines, in contrast to Australia’s, blur the distinction between pathologically based and age-related androgen deficiency, loosen the diagnostic criteria and lack regulatory force. Thus they minimise the diagnostic hurdle, leading to the perverse outcome of potentially encouraging rather than deterring unproven overuse of testosterone.

Second, the EMAS definition of “late-onset hypogonadism” is not suitable for implementation in practice. Findings of the MMAS confirm clinical experience and are reflected in all clinical guidelines:12-14 serial serum testosterone levels in older men are sufficiently variable7 to require the results of at least two blood samples taken weeks apart to establish a sustained low level of serum testosterone. Hence, a definition reliant on a sample of single testosterone measurements is likely to be highly error-prone, due to regression to the mean and other sources of variability. It also relies on testosterone measurements by mass spectrometry, a superior technology not yet available in Australian pathology laboratories despite recognised limitations of testosterone immunoassays.15 Similarly, the so-called “free” testosterone variable, also used in the study but not recommended, is calculated by an inaccurate formula unsuitable for individual diagnosis.16

Third, reflection on these two NEJM articles makes evident the need to reinforce bans on direct-to-public advertising of testosterone supplementation. This critical protection depends on industry self-regulation. Without this protection, an avalanche of misguided testosterone prescribing awaits us, analogous to the massive, unregulated marketing of non-proprietary impotence drugs advertised on billboards and in the media, apparently beyond regulatory reach.

Finally, testosterone prescribing for older men is best restricted to clinical trials where ethical oversight ensures appropriate design and warnings to participants. Age alone may not prove a valid indication, but this should not limit essential further research within the framework of placebo-controlled clinical trials aiming to define the benefits and risks of testosterone use in patients with the comorbidities of ageing, such as obesity, diabetes, metabolic syndrome and other chronic conditions.