Skip navigation

Erectile dysfunction treatments - relative efficacy


When there are a number of different treatments available for the same condition, this begs the question of which is "best". Best, of course, is one of those difficult questions, but usually involves:

  1. Efficacy: whatever we choose as the most relevant outcome, which treatment produces most of it?
  2. Harm: most treatments have some propensity to cause harm, so which causes least. Or, can we predict the type and extent and severity of any harm that may be caused.
  3. Cost: what does it cost to produce some benefit minus the cost of any harm. Often called cost-effectiveness, and measuring "bangs per buck", this argument often fails because budget-based medicine is more often practiced than evidence-based medicine.

In this section of Bandolier we have sought to bring information on erectile dysfunction treatments together, with brief reviews on alprostadil , sildenafil , yohimbine , phentolamine , apomorphine and VIP . We thought it relevant to include this short essay on relative efficacy, and how to think about it. [Relative efficacy rather than relative harm, because work on harm is a bit more tricky]

Thinking background

So let's go through a thinking process in stages.

First - we probably don't want information at all when there's not much of it, or it is of poor quality. Here that helps, because for phentolamine oral we have just one small trial, and for phentolamine with papaverine we have two small trials with fewer than 100 men in total. For oral apomorphine we have one randomised trial, but many different doses and regimens, so probably not enough information. For intracavernosal VIP we have two small trials. So at a stroke we can eliminate three of the treatments from consideration.

For alprostadil we have three trials and about 1100 men, for sildenafil four trials and about 1400 men, and for yohimbine we have 10 trials and about 660 men.

Second - we want to consider whether all the treatments are available. Yohimbine may or may not be, depending where we live, so three treatments are left for consideration.

Third - are the patients the same? If we exclude from consideration trials with men with particular problems, like spinal cord lesions or diabetes, as best we can judge the trials for all three remaining treatments are for erectile dysfunction of mixed aetiology, and to that extent are broadly comparable.

Fourth - are the outcomes used comparable? For the review we chose the outcome of patient/partner judgement of satisfactory erections suitable for intercourse, or actual intercourse, at home. Essentially, therefore, this was a threefold outcome of erection, sufficiently rigid for penetration, and of sufficient duration for satisfactory sexual activity. Some trials measured only rigidity in a laboratory setting, and these were not included. Ideally outcomes were reported on a patient basis, rather than on an event basis, which for Bandolier was a secondary outcome.

Most of the yohimbine trials had this outcome, as did the alprostadil trials. Those for sildenafil told us about erections suitable for intercourse, but also told us that a high percentage of them usually resulted in intercourse. Between these three treatments we therefore had rough comparability.

Fifth - are there any other problems we might need to consider? Well, yes, actually. The alprostadil trials all had an enriched enrollment, because only responders could enter the randomised trial. We have calculated that enriched enrollment would result in a better NNT and higher percentage of success.

Thinking result

On balance, then, the information we have for alprostadil, sildenafil and yohimbine is adequate and just about comparable. We think there may be some over-estimation of efficacy for alprostadil and for sildenafil, but can make a judgement about that if needed.

Relative efficacy result

The relative efficacy results can be displayed in two ways. Figure 1 has the result in the form of an NNT league figure, and Figure 2 displays the results in terms of percentages of men with a successful outcome (erection or intercourse) for treatment and placebo.

Figure 1: NNT league figure for erectile dysfunction

Figure 2: League figure of percentages


So with a few quibbles, we have some ideas about how the effectiveness of these three treatments relates one to the others.

It is likely that we have overestimated the efficacy of transurethral alprostadil for a real world situation. But the reality is that if prescribing were initiated in primary care it would be done on the basis of trying it out, as much as anything because some folk will hate the delivery method while others will consider it a bonus.

It is possible that we have overestimated the efficacy of sildenafil. But that depends on the interpretation of outcome, and whether intercourse with an erection is always the aim. For men with and without erectile dysfunction ejaculation frequency was higher than intercourse frequency (Dinsmore et al, 1998). The point is that without asking patients, we won't know.

Finally, check out a Cochrane Review protocol to see how they want to do it:

H Fink, T Wilt, R MacDonald, I Rutks, D Schow. Sildenafil for erectile dysfunction [protocol]. In The Cochrane Library Issue 2, 2001.