Skip navigation

Testosterone for erectile dysfunction


Clinical bottom line

There is limited evidence that testosterone can help men with erectile dysfunction. In five randomised comparisons with placebo with 109 men, testosterone was better than placebo in four, and the overall NNT was 2.1 (1.5 to 3.0). There is good evidence that testosterone supplementation works best in men with low initial serum testosterone concentrations below 12 nmol/L, and is ineffective in men with serum testosterone concentrations above this.

Information on trial quality is limited, and this result has insufficient weight to drive clinical practice.


P Jain et al. Testosterone supplementation for erectile dysfunction: results of a meta-analysis. Journal of Urology 2000 164: 371-375.

AM Isidori et al. Effects of testosterone on sexual function in men: results of a meta-analysis. Clinical Endocrinology 2005 63: 381-394.

Review 1

The first review sought papers in which testosterone was given as the only methods of therapy for erectile dysfunction. Included studies had to have a clear and quantifiable measure of erectile response, rather than desire, and which included a number of measures like surveys and patient diaries. Electronic databases were searched, with the date of the last search in 1998.

Formal evaluation of trial quality was not performed, and randomised and case control studies were included.


Overall, 16 studies were included, but we are only told that five trials (from 1979 to 1994) with 109 patients were randomised, and had comparison to placebo. The results of those trials are shown in Figure 1:

Figure 1: L'Abbé plot of erectile responders to testosterone and placebo

With placebo, 9/54 men had a response (eight in one trial). The percentage response was 17% (95% CI 7-27). For testosterone 36/55 men had a response (65%, 95% CI 53-78). The relative benefit was about 4 (because of zero or 100% values in many cells, accurate computation is impossible). the NNT for one man to have a response with testosterone compared with placebo was 2.1 (1.5 to 3.0).

Adverse events were not described.

Review 2

The second review sought randomised trials only, and only six of the 16 studies in the first review were included in the second review. Information was abstracted from trials for a range of outcomes, including serum testosterone levels at baseline and after treatment. The analysis used standardised mean differences, so that simple explanations of the results were not easy.


Searching found 17 papers, 14 of which were both randomised and blind. Five studies were on hypogonadal men (testosterone below 7 nmol/L, seven on eugonadal men (testosterone above 12 nmol/L), and five on mixed types. There were 656 men, with 284 randomised to testosterone, 274 to placebo, and 88 in crossover design studies. Overall the mean age was 58 years. Most studies had fewer than 20 men in each trial arm. There were large differences in testosterone preparation used, dosing, delivery method, and prevalence of sexual dysfunction at baseline.

In studies with baseline testosterone below 10 nmol/L, , with between three and six studies reporting outcomes, testosterone significantly improved:

In studies with baseline testosterone above 10 nmol/L, results were much less impressive, with less difference between testosterone and placebo, and with statistical significance for:

There was some, limited, evidence that testosterone that no effect was found in men with baseline testosterone above 12 nmol/L, and that for both libido and erectile function the magnitude of effect was inversely related to the baseline testosterone concentration in the study population: men with lower testosterone did better.


These would make useful teaching papers. The first review tries to do many things, and probably fails to meet many of the standards we look for in meta-analyses while remaining enthusiastic, and interesting, and even possible relevant. The second review is much better in that it was of mainly randomised and blind studies, but they were small, and some of the extrapolations are limited in their utility.

There is also the issue of searching efficiency. Some studies in the second, later, review were not found by the first, earlier, review, even though they were published at that time. At least two small trials were available for the second review, but not included, when they perhaps might have been (though they were too small to have changed any results). We should not be censorious, because searching has improved, and labelling of trials has improved, considerably, in recent years.

The main problem is one of numbers. There just weren't enough to make any sense, despite attempts to analyse, for instance, the relative efficacy of different modes of testosterone delivery. For what it's worth, transdermal seemed better than others, and not scrotal administration or intramuscular injection.

These reviews give a reasonable idea of where we are now, and what the literature is. Useful if we were considering testosterone development, or if all other acceptable treatments had failed.