This large cohort study of middle-aged and older men demonstrates that serum testosterone and cFT are inversely associated with overall and cancer-related, but not CVD-related mortality. The study also demonstrates direct relationships between serum SHBG and overall, CVD-related and cancer-related mortality.
Previous studies of testosterone and mortality have been smaller, generally analysing between one to five thousand men, with several hundred to a thousand or more deaths.6-18 A nested case-control analyses drawn from a cohort of 11,606 men involved 825 men who died (369 CVD-related, 304 cancer-related) and 1,489 controls.11 The largest previous cohort analysis involved 5,350 men, with 1,533 deaths (428 CVD-related, 480 from cancer-related).19 Our analysis of 149,436 men from the UK Biobank, with 10,053 deaths (1,925 CVD-related, 4,927 cancer-related) provided unprecedented scope to clarify the associations of testosterone and SHBG with all-cause, CVD and cancer mortality, and to establish more precise estimates of effect sizes.
Additionally, previous multivariable models typically adjusted for between five to ten covariates,7,9,10,12-15,19,22 while more extensive models included 11 to 18 covariates.6,11,18,23 We adjusted for 26 covariates including age, and not only BMI but also waist circumference. Importantly, and unlike previous studies, we included adjustment for specific factors that raise SHBG, such as thyroid and liver disease and anticonvulsant medications.5
In unadjusted analyses, men with lower serum testosterone had higher risks of death from any cause, CVD and cancer. However, in multivariable analyses including SHBG, only men with the lowest testosterone concentrations had higher all-cause and cancer-related, but not CVD-related, mortality. Previous studies including SHBG in the model found either inverse associations,11,18 or no association9 of testosterone with all-cause mortality. Other studies adjusting for risk factors and comorbidities, not including SHBG, reported inverse12,13,15,16 or neutral6,7,19,22 associations of testosterone with all-cause mortality. In our fully-adjusted analysis, the inverse association of testosterone with all-cause mortality was independent of SHBG. The association was statistically robust, but largely limited to men with testosterone in the lowest quintile, who had a modest (14%) increase in risk
Several previous studies reported no associations of testosterone with CVD-related mortality,6,8,10,13,17,18 whereas others reported inverse associations of testosterone with CVD deaths.11,12,14,15,19 After adjusting for covariates including SHBG, we found no evidence that testosterone was associated with risk of CVD death. The HR was 1.01, with narrow confidence intervals (0.86-1.18), arguing against any substantive association of testosterone with this outcome.
Two previous studies suggested a relationship between lower testosterone concentrations and cancer-related mortality,11,16 whereas another study associated higher testosterone with deaths from lung cancer,22 and other studies have reported no association of testosterone with cancer deaths.12,19 We found a robust inverse association, which was limited to men with serum testosterone in the lowest quintile, who had a modest (20%) increase in risk of dying from cancer.
Associations of cFT with all-cause, CVD and cancer-related mortality were seen in univariable analyses. However, in multivariable analyses these were attenuated: men with the lowest cFT values had higher all-cause and cancer-related, but not CVD-related mortality. Results for cFT, adjusted for covariates, did not provide additional information beyond analysis of testosterone and SHBG for mortality outcomes.