Background: The effects of testosterone on cognitive function in older men are incompletely understood. We aimed to establish the effects of long-term testosterone administration on multiple domains of cognitive function in older men with low or low-to-normal testosterone concentrations.

Methods: We did the randomised, double-blind, placebo-controlled, parallel-group TEAAM trial at three medical centres in Boston, Phoenix, and Los Angeles, USA. Men aged 60 years and older with low or low-to-normal testosterone concentrations (3·47-13·9 nmol/L, or free testosterone <173 pmol/L) were randomly assigned (1:1), via computer-generated randomisation, to receive either 7·5 g of 1% testosterone gel or placebo gel daily for 3 years. Randomisation was stratified by age (60-75 years vs >75 years) and study site. The testosterone dose was adjusted to achieve concentrations of 17·3-31·2 nmol/L. Participants and all study personnel were masked to treatment allocation. Multiple domains of cognitive function were assessed as prespecified secondary outcomes by use of standardised tests at baseline and months 6, 18, and 36. We did analyses by intention to treat (in men who had baseline assessments of cognitive function) and per protocol (restricted to participants who completed the study drug and had both baseline and 36 month assessments of cognitive function). The TEAAM trial is registered with ClinicalTrials.gov, number NCT00287586.

Findings: Between Sept 1, 2004, and Feb 12, 2009, we randomly assigned 308 participants to receive either testosterone (n=156) or placebo (n=152). 280 men had baseline cognitive assessments (n=140 per group). Mean follow-up time was 29·0 months (SD 11·5) in the testosterone group and 31·1 months (9·5) in the placebo group. The last participant completed the study on May 11, 2012. In the testosterone group, mean concentrations of serum total testosterone increased from 10·6 nmol/L (SD 2·2) to 19·7 nmol/L (9·2) and free testosterone concentrations increased from 222 pmol/L (62) to 364 pmol/L (222). In the placebo group, mean concentrations of serum total testosterone were 10·7 nmol/L (SD 2·3) at baseline and 11·1 nmol/L (3·2) post-intervention and free testosterone concentrations were 210 pmol/L (61) and 172 pmol/L (49), respectively. We recorded no between-group differences in changes in visuospatial ability (mean difference: Complex Figure Test -0·51, 95% CI -2·0 to 1·0), phonemic or category verbal fluency (phonemic fluency test 0·90, -1·3 to 3·1; categorical fluency test 1·1, -0·3 to 2·6), verbal memory (paragraph recall test 0·29, -1·2 to 1·8), manual dexterity (Grooved Pegboard Test 4·2, -1·3 to 9·7), and attention or executive function (Stroop Interference Test -2·6, -7·4 to 2·3) after adjustment for age, education, and baseline cognitive function. In both the intention-to-treat and per-protocol (n=86 per group) populations, changes in cognitive function scores were not related significantly to changes in total or free testosterone, or oestradiol concentrations.

Interpretation: Testosterone administration for 36 months in older men with low or low-to-normal testosterone concentrations did not improve cognitive function. Future long-term trials are needed to investigate the efficacy of testosterone replacement in patients with impaired cognition, such as people with Alzheimer's disease.

Funding: AbbVie Pharmaceuticals, Aurora Foundation, Boston Claude D Pepper Older Americans Independence Center, and Boston University's Clinical and Translational Science Institute.