Background
Previous cohort studies have associated lower testosterone concentrations with poorer cognitive function, and higher dementia risk, with clinical trials of testosterone providing inconclusive results.
Aims
To clarify whether endogenous concentrations of testosterone, and of other sex hormones, were associated with cognition and incident dementia in men.
Methods
A systematic review was conducted, identifying suitable studies from which individual participant data (IPD) were requested (PROSPERO registration: CRD42019139668). Eligible studies were prospective cohort studies measuring testosterone using mass spectrometry, with at least five years of follow-up data. IPD meta-analysis (IPDMA) models were fitted. Datasets that included both the outcome (baseline cognitive function or time to dementia diagnosis) and exposure (total testosterone [T] or sex hormone-binding globulin [SHBG] concentration) were analysed.
Results
The cross-sectional analysis included 9,741 men, median age 74 years, from 7 cohorts; the longitudinal analysis had IPD for 8,852 men from 5 cohorts (114,168 participant-years follow-up) plus aggregate statistics for an additional study (n=1,463). Harmonised z-scores of global cognition were not associated with T but were non-linearly with SHBG (estimated mean difference [MD] at medians of quintile 1, 26.5 [Q1] vs. quintile 5, 76.9 [Q5] nmol/L; MD, 0.088 [CI, 0.002 to 0.174]). Incident dementia was nonlinearly associated with T (adjusted hazard ratio, HR for Q1 vs Q3, 1.17, [CI 1.02 to 1.34] and with lower SHBG (Q2 vs Q5; adjusted HR, 0.82 [CI 0.70 to 0.96]).
Conclusions
Constraining study eligibility to those that assayed testosterone using mass spectrometry facilitated analyses of high-quality datasets, using multivariable non-linear models. Lower T is a potential biomarker for higher risk, and lower SHBG a potential biomarker for reduced risk of incident dementia in men, warranting further studies to explore potential mechanisms.