The review included 15 randomised controlled trials (RCTs; n=862), comprising 8 parallel trials (n=781) and 7 crossover trials (n=81). The 15 RCTs involved 17 comparison groups.
In general, the quality of the studies was limited. Only one study reported using allocation concealment. Blinding was fully reported in 4 studies, although all but one clearly or probably blinded the participants and treatment providers. Loss to follow-up exceeded 15% in 5 studies. The authors noted the possibility of reporting bias and publication bias.
Among men with low testosterone levels, testosterone had a large but not statistically significant effect on satisfaction with erection (effect size 0.80, 95% CI: -0.10, 1.60; 4 studies); there was a high level of inconsistency between the studies (I-squared 86%). The subgroup analysis showed a larger treatment effect in younger men (effect size 1.80, 95% CI: 1.00, 2.70; 2 studies) than among older men (effect size 0.10, 95% CI: -0.60, 0.80; 2 studies).. Among men with low-normal or normal testosterone levels there was a small but statistically significant effect favouring testosterone over placebo for satisfaction with erection (effect size 0.34, 95% CI: 0.03, 0.65; 7 datasets from 6 studies). Inconsistency was very low (I-squared 5%).
Among men with low testosterone levels, testosterone had a large statistically significant effect on libido (effect size 1.31, 95% CI: 0.40, 2.35; 5 studies); there was a high level of inconsistency between the studies (I-squared >90%). Among men with low-normal or normal testosterone levels, there was no significant difference between the groups (5 datasets from 4 studies) and no inconsistency (I-squared 0%).
The meta-analysis found no significant effect of testosterone on overall sexual satisfaction, regardless of the testosterone level at baseline (10 studies).
Systematic subgroup and sensitivity analyses failed to explain inconsistencies between the groups. The outcomes did not differ significantly between the groups with low and low/normal testosterone levels at baseline.