—A diminished serum testosterone level in older men is associated with increased mortality. It is probably a marker of general health, not an independent risk factor.

By Sanjai Sinha, MD, FACP, Assistant Professor, Department of Medicine, Weill Cornell Medical College, New York, NY
Reviewed By Clifton Jackness, MD, Attending Physician in Endocrinology, Lenox Hill Hospital and the Mount Sinai Medical Center, New York, NY

Serum testosterone concentration peaks during adolescence and early adulthood. Males tend to have a 1% to 2% reduction per year in testosterone beginning in their 30s.1 A more precipitous decline may indicate a higher probability of mortality. Observational studies have associated low serum testosterone and late-onset hypogonadism with cardiovascular risk factors, diabetes, and hyperlipidemia.2-4 While these intermediary outcomes are clearly related to premature death, the effect of testosterone decline on longevity is less clear. Several studies have shown conflicting results.

A recent study of more than 4000 men concluded that low testosterone predicts cardiovascular mortality but is not associated with death from other causes.5 In another study, low testosterone levels were associated with increased mortality over 8 years of follow-up in a retrospective analysis of male veterans who attended a hospital clinic.6 Two population-based studies of comparatively healthy men with longer follow-up (15 to 16 years) did not find an association of testosterone with survival.7,8 The men in these studies, however, were considerably younger (mean age, early 50s), so the survival effects of testosterone insufficiency were less impactful. Finally, a prospective trial of 794 men concluded that men in the lowest quartile of testosterone levels had a 40% increased relative risk of death compared to those with higher levels, independent of age, adiposity, and lifestyle.9 This finding held up even after adjusting for the presence of diabetes, metabolic syndrome, and cardiovascular disease (CVD).

A recent meta-analysis examined studies like these to determine if pooled results might clarify our understanding of the relationship between testosterone levels and survival. Searching data sources from 1966 to 2010, Araujo and colleagues evaluated English-language observational studies that reported the association between endogenous testosterone and all-cause or CVD mortality.10 Of 45 studies selected for full-text review, 21 were chosen for systematic review, with 12 eligible for meta-analysis. Eleven of these reported on all-cause mortality (16,184 patients) and 7 reported on cardiovascular mortality (11,831 patients). A secondary objective was to assess whether clinical, demographic, or study-related factors influenced or modified study results. Examples of these factors included baseline testosterone level, smoking status, body mass index, age, length of follow-up, and type of testosterone assay.

The mean follow-up period of the studies was 9.7 years. Average patient age was 61 years, and average testosterone level was 467 ng/dL. The authors found that a 2.18 standard deviation decrease in total testosterone was associated with a 35% increased risk of all-cause mortality and a 25% increased risk of CVD mortality.

The validity of these summary estimates is limited because of significant between-study heterogeneity, however. Factors contributing to the relative risk heterogeneity included mean age of the population (P = .02), baseline total testosterone level (P = .018), number of years of follow-up (P = .010), and whether blood samples were collected in the morning (P = .03). Larger relative risks were observed in studies that included older (RR = 1.54; 95% CI, 1.28-1.85) versus younger (RR = 1.12; 95% CI, 0.92-1.36) men, patients with lower (RR = 1.55; 95% CI, 1.28-1.88) versus higher (RR 1.09; 95% CI, 0.88-1.36) testosterone levels, patients who were followed for no more than 9.6 years (RR = 1.63; 95% CI, 1.32-2.02) versus more than 9.6 years (RR = 1.13; 95% CI, 0.95-1.36) and in those that did not obtain morning blood samples (RR = 1.61, 95% CI, 1.28-2.03) versus those that did (RR = 1.15; 95% CI, 0.94-1.41).

The studies included generally healthy men, but there was still significant heterogeneity in the association between endogenous testosterone and all-cause mortality across the studies. Based on these findings, it appears that low total testosterone is a marker for all-cause mortality rather than a risk factor itself. The summary estimate may reflect the contribution of underlying health conditions to mortality, rather than the random effects of testosterone. And while low testosterone levels may indicate a higher probability of medical problems, a low level also may be the result of these medical problems. This distinction is very difficult to elucidate. Another limitation of this meta-analysis is that all eligible studies were observational, which makes selection bias an issue. This meta-analysis suggests that testosterone levels in middle-aged and elderly men are likely markers of general health. Whether testosterone treatment affects mortality is currently under investigation