Because of the common practice of measuring androgen when evaluating erectile dysfunction, androgen replacement therapy has proliferated. To our knowledge a causal relationship between altered levels of androgens and erectile dysfunction is not firmly established. In fact, there are few studies of androgen replacement and its effect on erectile dysfunction. In the Massachusetts male aging study testosterone was not although serum dehydroepiandrosterone was associated with patient reported erectile dysfunction. 26 It is believed that erection is achieved by synergistic reflexogenic and psychogenic mechanisms. The reflexogenic component involves sensory stimulation of the genitalia, which subsequently results in a spinal cord reflex erectile response and is largely androgen independent. The psychogenic component involves higher brain functions and depends on free testosterone. 25 Erectile dysfunction may be a consequence of a derangement in any vascular, neurological, psychogenic or endocrinological aspect of the erectile mechanism. What is relatively clear from the literature is that testosterone replacement is important in only a small subset of patients, comprising less than 5% of those with erectile dysfunction. 7
We compiled data from the literature in an attempt to create a basis for stimulating future investigation and provide insight into the rational use of androgen replacement for erectile dysfunction. From this meta-analysis of the usefulness of androgen replacement therapy for erectile dysfunction we noted that the response rate for primary etiologies was significantly higher than that for secondary etiologies, transdermal testosterone therapy was more effective than intramuscular or oral treatment, and intramuscular and oral methods of delivery were equivalent. In addition, there was a statistically significant difference in favor of testosterone over placebo, supporting a role for supplementation in select groups.
Why the response rate differs significantly in primary and secondary etiologies must be answered. A possible explanation is that men with primary testicular failure have a clearer causality of erectile dysfunction, including fewer confounding co-morbidities that would hamper a response to androgen supplementation. Those with secondary testicular failure may have associated problems, and so vascular, neurological, psychogenic and endocrinological aspects of erectile response may make it difficult ro establish a clear association with androgen deficiency. Also, better controlled studies with well developed inclusion and exclusion criteria are possible in men with primary testicular failure, providing less potential for selection bias.
One must also wonder why transdermal replacement induces a significantly better response than intramuscular or oral supplementation. The more physiological delivery pattern of the daily transdermal methods is an obvious consideration, and yet to our knowledge a mechanistic explanation is not forthcoming. It is possible that the improved response is due more to the nature of current investigation with better inclusion and/or exclusion criteria rather than to some intrinsic factor in the androgen level or rate of systemic decay. Transdermal studies are smaller than intramuscular studies (42 versus 238 cases), making direct comparison in this setting prone to error. The high response rate in the topical groups is questionable when considering the complex nature of erectile dysfunction, implying a potential bias in patient selection. A clinical trial comparing the response rates of transdermal and intramuscular routes is needed to address this issue.
A surprising finding was the similar response rate for intramuscular versus oral androgen therapy. This finding may reflect poor study design or selection bias in the treatment groups as well as poor overall response potential in these patients. The trend in differences in the scrotal versus nonscrotal application of transdermal delivery is a marginal finding that warrants crossover or comparison studies.
When considering placebo controlled crossover studies, several factors were noted. There is a surprising deficiency of well controlled studies. Because of the small number of these series, we did not use the more common Mantel-Haenszel analysis. In addition, we made no attempt to calculate a potential publication bias because of the few controlled investigations available for review. However, in our analysis we weighed each study by sample size. Ideally we would have preferred to know the placebo and testosterone response in each patient but this information was not provided. Thus, one must treat the samples as independent comparisons rather than as matched samples. Despite these difficulties we noted a statistically significant difference in favor of testosterone treatment over placebo, supporting a role for testosterone supplementation in select groups.
Controlled trials have been considered the gold standard for evaluating medical intervention. For most medical questions it has been possible to perform relatively small trials. 27 Meta-analysis has become the cornerstone of evaluating multiple, disparate clinical series in an attempt to provide information to assist physicians in clinical decision making. To avoid inherent flaws in performing this type of analysis it is important to choose a balanced group of trials for analysis. Experimental design must be similar, and inclusion criterion and outcomes comparable. In our meta-analysis we made every attempt to exclude articles that did not meet our aforementioned inclusion criteria. The systematic approach to identifying and abstracting research reports was the core of our project. Our identification of more than 17,000 reports of erectile dysfunction and/or androgen replacement with selection down to only 73 testifies to our rigorous inclusion criteria. A disappointing finding was that only 16 of these 73 reports were detailed enough for reliable data extraction, while only 11 of 16 provided details on etiology.
A potential critique of our analysis involves the inclusion criteria with regard to patient response. We included studies with a clear and quantifiable measure of erectile response with androgen supplementation, such as clinical improvement in potency, increased scores on sexual function surveys and diary entries of the number of erections. Each of these 3 measures is parallel and similar but not as good as those in more recent investigations of erectile dysfunction, such as combined penile tumescence monitoring, validated questionnaires and patient-partner diaries. 28 and 29 Given the diverse measures of response in the evaluated articles some variability was expected in results. Another potential critique of our efforts is the exclusion of nonpublished trials or the so-called publication bias. Because zero effect studies have a smaller chance of publication than statistically significant studies, there may be an increased risk of type I error. To estimate the significance of this phenomenon the theoretical number of unpublished studies with zero effect was included to change a significant into a nonsignificant result. Again, because of the small number of published series in this category, we omitted this step.
Androgen is necessary but not essential for normal libido and to our knowledge its exact role in erectile function remains unclear. When treated with visual or sexual stimuli, individuals with a castrate level of testosterone may achieve erections thought to be comparable to those in men with normal testosterone. 8 The clinical manifestations of androgen deficiency depend on the stage of sexual development in which they occur. In older males decreased virility and a catabolic effect are more predominant. In adult and aging males testosterone is needed to maintain virilization, libido, potency, secondary sexual characteristics, and muscle and bone mass. Although testosterone replacement seems reasonable in testosterone deficient men, to our knowledge supplementation has not yet been properly evaluated. Ideally androgen replacement therapy should reestablish a normal endocrine milieu that restores all androgen dependent physiological functions.
The methods of androgen replacement currently available for human use include parenteral preparations of the long acting esters of testosterone, enanthate and cypionate, oral preparations of 17α-alkylated derivatives and recently introduced transdermal preparations. The long acting 17β-hydroxy esters of testosterone, enanthate and cypionate are least costly but the efficacy and efficient use of health care expenditure in regard to these preparations are not universally accepted. Administering a 200 to 300 mg. dose every 2 to 3 weeks provides a practical option but it is far from an ideal physiological replacement. We made no attempt to standardize the frequency of administration or the dose since these factors varied among reports. In addition, the exact formulation of replacement therapy also varied. These variations may be viewed as a weakness of our review.
Oral androgen preparations have the potential for serious hepatotoxicity. However, in the rare patient who cannot receive androgen parenterally 25 or 50 mg. methyl testosterone, or 5 to 10 mg. fluoxymesterone may be given orally daily if hepatic function is closely monitored. The transdermal delivery of testosterone is convenient and reliable, and it better mimics physiological testosterone secretion. With such patches normal serum testosterone may be achieved in most men with hypogonadism. After applying the patch on the scrotal skin serum the testosterone level rapidly increases to the normal range, peaking between 2 and 6 hours after application. Because there is a gradual decrease into the hypogonadal range within 24 hours, the patch must be applied to the skin daily. When the patch is applied in the early morning, the diurnal testosterone rhythm may be closely mimicked. To our knowledge what has not been demonstrated is whether replicating this diurnal rhythm provides any physiological advantage.
An intriguing question raised by our study is the issue of appropriate first versus second line treatment of androgen related erectile dysfunction. We attempted to estimate the cost difference to address this issue. If testosterone cypionate were used as first line therapy, the estimated cost would be $68,057.75 for treating 100 patients for 6 months and then changing those in whom intramuscular therapy failed to transdermal treatment for the next 6 months (table 7). This cost assumes a 50% response rate for the initial 6 months on intramuscular therapy and an 80% rate for the next 6 months on transdermal therapy. Thus, theoretically 90 of 100 patients would respond during 1 year of treatment. This cost compares favorably to that of transdermal therapy as first line treatment at an initial yearly cost of $130,852.50 for a response in 80 of 100 patients in 1 year. The comparison does not include the cost of administering the injection, that is an office visit for some patients, home nurse visit or self-injection. Also, this comparison assumes that men refractory to intramuscular treatment would respond to transdermal therapy and there would be no inflation during treatment phase 2 when changing from intramuscular to transdermal delivery. One wonders whether the lower cost associated with parenteral testosterone as first line therapy for replacement would be offset by the marginal cost of its expected increased failure rate, increased physician time and expense as well as patient frustration with treatment failure. 30 Marginal costs are the key to understanding the role and influence of subsequent failure and necessary additional maneuvers in erectile dysfunction treatment. A marginal cost analysis of the various treatment sequence scenarios is an obvious need resulting from our report.