Context: Low testosterone levels have been associated with outcomes that reduce survival in men.
Objective: Our objective was to perform a systematic review and meta-analysis of published studies to evaluate the association between endogenous testosterone and mortality.
Data sources: Data sources included MEDLINE (1966 to December 2010), EMBASE (1988 to December 2010), and reference lists.
Study selection: Eligible studies were published English-language observational studies of men that reported the association between endogenous testosterone and all-cause or cardiovascular disease (CVD) mortality. A two-stage process was used for study selection. 1) Working independently and in duplicate, reviewers screened a subset (10%) of abstracts. Results indicated 96% agreement, and thereafter, abstract screening was conducted in singlicate. 2) All full-text publications were reviewed independently and in duplicate for eligibility.
Data extraction: Reviewers working independently and in duplicate determined methodological quality of studies and extracted descriptive, quality, and outcome data.
Data synthesis: Of 820 studies identified, 21 were included in the systematic review, and 12 were eligible for meta-analysis [n = 11 studies of all-cause mortality (16,184 subjects); n = 7 studies of CVD mortality (11,831 subjects)]. Subject mean age and testosterone level were 61 yr and 487 ng/dl, respectively, and mean follow-up time was 9.7 yr. Between-study heterogeneity was observed among studies of all-cause (P < .001) and CVD mortality (P = 0.06), limiting the ability to provide valid summary estimates. Heterogeneity in all-cause mortality (higher relative risks) was observed in studies that included older subjects (P = 0.020), reported lower testosterone levels (P = 0.018), followed subjects for a shorter time period (P = 0.010), and sampled blood throughout the day (P = 0.030).
Conclusion: Low endogenous testosterone levels are associated with increased risk of all-cause and CVD death in community-based studies of men, but considerable between-study heterogeneity, which was related to study and subject characteristics, suggests that effects are driven by differences between cohorts (e.g. in underlying health status).