The effect of statins on testosterone in men and women, a systematic review and meta-analysis of randomized controlled trials

1Statins are extensively used in the West for secondary prevention of cardiovascular diseases, and have contributed to a reduction in cardiovascular disease mortality rates. Statin use will become more common globally with the emerging epidemic of cardiovascular disease in developing countries and the increasing availability of statins off patent. Randomized controlled trials (RCTs) show statins reduce mortality rates more than other commonly used lipid-modulating therapies, such as niacin, fibrates and ezetimibe (where the evidence is limited), and are beneficial for some cardiovascular diseases poorly correlated with cholesterol, although statins increase the risk of diabetes, with a dose-dependent effect. In RCTs statins also reduce inflammation and improve immune function. Currently, there is no known, coherent explanation for this specific pattern of pleiotropic effects, and no identified 'active ingredient', despite relevance to drug discovery, existing therapies and modifiable risk factors.

2Statins inhibit the enzyme 3-hydroxy-3-methyl-glutaryl-CoA (HMG-CoA) reductase, which reduces cholesterol production. In contrast niacin blocks the breakdown of fats, fibrates activate peroxisome proliferator-activated receptor (PPAR)-α and ezetimibe inhibits cholesterol absorption. All modulate circulating lipids, but only statins also inhibit de novo synthesis of cholesterol throughout the body. Concern has always existed that statins might impair testosterone production either by reducing availability of its preferred substrate, that is, locally produced de novo cholesterol in the gonads and elsewhere, or by inhibiting steps in the steroidogenesis process, but this concern has been considered of little clinical significance. Lowering androgens potentially explains some of statins' effects. An RCT among men found testosterone therapy enhanced glucose metabolism, whilst impaired glucose metabolism may be a side effect of androgen deprivation therapy. Men with naturally low androgens due to Klinefelter's syndrome are more prone to diabetes. Extensive experimental evidence indicates that testosterone is immunomodulatory and impairs immune response. Given statins' physiological mechanism would be expected to reduce androgens as well as the similarity between the effects of statins and of reducing androgens, we hypothesized a priori that the pleiotropic effects of statins could be due to lowering of androgens, that is, that statins lower androgens and that lower androgens might mediate the pleiotropic effects of statins. Here, we carried out a meta-analysis of placebo-controlled randomized trials, to avoid bias by indication, in men and women to examine whether statins reduced total testosterone.

3We implemented this meta-analysis following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (Additional file 1). Two reviewers (CMS and SLAY) independently searched for and selected studies, resolving differences by consensus. Two statisticians (GF and BJC) extracted information from the selected studies.

4We systematically searched the PubMed, Medline and ISI Web of Science databases until the end of 2011 using search term '(Testosterone OR androgen) AND (CS-514 OR statin OR simvastatin OR atorvastatin OR fluvastatin OR lovastatin OR rosuvastatin OR pravastatin)' with the search limited to RCTs of studies in humans in English, because an initial search suggested the relevant literature was in English. From our search, we discarded any studies that were not relevant from the title or abstract and read the remaining to identify placebo-controlled randomized trials of statins with testosterone reported. We also used the references of the selected trials to identify additional relevant trials.

5We included any published placebo-controlled randomized trial in English of at least 2 weeks' duration examining the effect of statins on testosterone in adults, because statins act rapidly on cholesterol. We did not consider trials in children, because children have different levels of testosterone, and statins are rarely used in children. We did not otherwise select by participant characteristics, because statins are used for the same purpose in a wide variety of patients, and there is no reason to think that statins have different effects by patient subgroup.

6A statistician (GF) extracted information by trial arm on number of participants, testosterone assays used, and mean and standard deviation of the change in testosterone during the trial. Where this information was not provided in full, the statistician (GF) made conservative estimates based on the information available. A second statistician (BJC) then checked the information extracted and any estimates. The reviewers (CMS and SLAY) independently used an established tool to evaluate the quality of each trial, and a sensitivity analysis was done excluding the lower quality trials.

7We used funnel plots to assess publication bias, and I2 to assess heterogeneity between trials. To obtain an estimate of the difference in mean testosterone between statin and placebo groups, we combined the results of the selected trials using inverse variance weighting and a fixed or random effects model depending on the level of heterogeneity, using the 'metacont' function from the 'meta' package of R 2.14.1 (R Development Core Team, Vienna, Austria). We considered men and women together and separately because of the different hormone milieu by sex. We did no other subgroup analysis.

8This study is an analysis of published data, which does not require ethics committee approval.

9The initial searches yielded 27 studies. Two additional trials were found from the bibliographic search. We discarded 18 studies. Three were clearly not relevant based on title, four had no control group, three had the control group taking other drugs (namely neomycin, cholestyramine or other lipid-lowering drugs and atorvastatin plus ezetimibe), four were not randomized controlled trials of statins, two were largely duplicates of other included publications and two provided insufficient information. Details of these 18 excluded studies are given in Additional file 2. Of the 11 trials remaining, 1 small trial of 15 men and 7 women providing only graphs had an implausibly high post-intervention value of testosterone for women in the placebo group (mean approximately 5.2 nmol/l for 3 women, no standard deviation given), so these 7 women were also excluded. Figure 1 shows the search strategy resulting in these 11 placebo-controlled randomized trials.

10Table 1 shows there were 5 trials over 25 years of 501 men, mainly middle aged with hypercholesterolemia, taking typical doses of statins. There were 6 recent trials of 368 young women with polycystic ovary syndrome. Most of the trials were in western settings and were carried out by teams affiliated with or funded by pharmaceutical companies. Most of the trials were of average quality as shown in Additional file 3, with the larger, more recent ones generally of higher quality.

11Funnel plots gave little indication of publication bias among men (Figure 2), but among women, the trials were more diverse. For completeness Figure 2 also shows the funnel plot for men and women together.

12Figure 3 (top panel) shows a forest plot for men. The trials were homogeneous (I2 = 0%), with the overall estimate within the confidence intervals of all trials. Statins lowered testosterone by about 3.4% among men (-0.66 nmol/l, 95% confidence interval (CI) -0.14 to -1.18) in both fixed and random effects models, and the estimate was similar (-0.73 nmol/l, 95% CI -0.20 to -1.26) including only the higher quality trials. Figure 3 (middle panel) shows a forest plot for women. The trials were heterogeneous (I2 = 90.1%). Statins lowered testosterone by about 12.3% among women (-0.40 nmol/l, 95% CI -0.05 to -0.75) in a random effects model, with similar results when the most extreme study was omitted. The estimate for women was also similar, but included no effect (-0.50 nmol/l, 95% CI 0.06 to -1.06) when considering only the higher quality trials. For completeness, Figure 3 (bottom panel) shows a forest plot for men and women together. The trials were heterogeneous (I2 = 79%). Statins lowered testosterone overall (-0.44 nmol/l, 95% CI -0.13 to -0.75) in a random effects model; results were similar when each of the three most influential trials was removed, or when only considering the higher quality trials (-0.57 nmol/l, 95% CI -0.12 to -1.02).

13This meta-analysis of placebo-controlled randomized trials suggests statins reduce testosterone. Among men the evidence was homogeneous and largely related to typical doses for the target group using statins for the prevention of cardiovascular disease, where statins potentially lowering testosterone has always been a concern. Among women less concern has existed. Very few trials have examined testosterone among women using statins for the prevention of cardiovascular disease, and our findings were less robust for women than men. Nevertheless, statins have recently been discovered as an effective anti-androgen treatment for polycystic ovary syndrome. A recent meta-analysis of statin therapy for women with polycystic ovary syndrome concluded that statins reduced testosterone based on the same literature.

14To the best of our knowledge no previous meta-analysis of placebo-controlled randomized trials has assessed the effect of statins on testosterone among men. Two trials comparing simvastatin 80 mg/day with 40 mg/day among in total 640 men found median testosterone lower by 10.3% and 7.5% respectively after 48 weeks, consistent with the 3.4% reduction here among men mainly using simvastatin 20 mg/day, suggesting a possible dose response of statins on testosterone.

15The clinical significance of this reduction in testosterone with statins is difficult to gauge. The normal range of testosterone is wide and sexual function similar across the range. Erectile dysfunction is a rare side effect of statins, perhaps because statins' have beneficial effects on cardiovascular function that would counteract changes of this magnitude in testosterone. However, large changes in testosterone (and libido) can occasionally occur with statin use, which are reversible by statin withdrawal. The impact on population health may be more germane, where statins causing diabetes could be another side effect. We are not aware of any study examining whether lowering testosterone mediates the effect of statins on diabetes, as this possibility has not, to the best of our knowledge, been considered before, and remains speculative. Moreover, observational studies suggest serum testosterone has sex-specific physiological effects on diabetes, negative among men, but positive among women, when statins increase diabetes in both sexes. We could not identify any RCT confirming testosterone therapy increases diabetes incidence among women. In postmenopausal women, RCTs of testosterone therapy indicate little effect on glucose metabolism. In younger women, some RCTs indicate that low doses of testosterone may improve glucose metabolism; female to male transsexuals given high doses of testosterone have improved glucose metabolism. Nevertheless, distinct effects of testosterone on diabetes may occur by dose, sex and age with the reduction in testosterone with statin treatment insufficient to modulate diabetes risk consistently. This meta-analysis also raises the question as to whether lowering testosterone is a side effect of statin therapy or contributes to statins' mode of action, which could inform new treatments and prevention policies. Observationally testosterone is inversely associated with cardiovascular mortality; whether testosterone is causal or a marker of health is unknown. No RCT has shown testosterone therapy reduces cardiovascular events; two RCTs of testosterone therapy were halted because of adverse, mainly cardiovascular events, among men allocated to testosterone. Natural experiments suggest lower testosterone protects against specifically ischemic heart disease mortality, with a relatively lower risk in men legally castrated or with Klinefelter's syndrome. Physiologically lowering testosterone may reduce thromboxane and platelet activation, specifically relevant to reducing ischemic cardiovascular disease but not to diabetes. However, whether lowering testosterone with statin treatment modulates cardiovascular disease, via these or other pathways, has not been examined.

16Despite providing a meta-analysis of all known placebo-controlled randomized trials, limitations exist. First, given when they were conducted, not all the trials had high quality scores. However, they had the expected effects on lipids. Second, not all the trials were intended to assess the effects of statins on testosterone. However, the trial, largely designed for this purpose, is influential, and alone found statins reduced testosterone. Third, testosterone was not assayed in the same way in all trials (Table 1). Testosterone is difficult to measure, particularly among women. Heterogeneity for women could be due to differing, possibly suboptimal, testosterone assay techniques, for which we compensated by using a random effects model. Imprecision in the assays also reduces power but would not bias a comparison between statin and placebo groups unless statins interfered with the testosterone assay. Steroids can cause assay interference. We could find no evidence that statins cause assay interference. Fourth, too few trials existed to assess dose-response effects or compare hydrophilic (fluvastatin, rosuvastatin, and pravastatin) to lipophilic (atorvastatin, lovastatin, and simvastatin) statins. However, atorvastatin, pravastatin or simvastatin do not differ substantially in their effects on cardiovascular disease. Fifth, other RCTs may have assessed but not reported the effects of statins on testosterone. Given, the concern that statins lower testosterone, these might perhaps be trials where statins reduced testosterone among men. Sixth, RCTs are not always tagged as such and could be missed. To check we searched using 'trial' as a search term instead of restricting the search to RCTs, which gave the same selection (data not shown). Seventh, evidence concerning the effect of statins on testosterone in post-menopausal women is lacking. Eighth, considering the effects of statins on testosterone for men and women together may be invalid, hence results stratified and pooled by sex. Ninth, the effect of statins on testosterone in different ethnic groups is lacking. Effects might differ in settings, such as China, where peak testosterone, androgen related parameters and ischemic heart diseases mortality rates are all lower than in western populations. Finally, cross-sectional studies were not included, because these provide evidence from which it is difficult to assess causality. The larger cross sectional studies only considered men and generally observed lower testosterone among statin users than non-users.

17This meta-analysis shows that statins reduce testosterone. This finding does not demonstrate that androgens mediate any health effect of statins, but raises the question as to whether testosterone modulation plays a role in statins' effects on health, particularly among men where testosterone is an important hormone. Whether reducing testosterone enhances or impairs the protective effects of statins on cardiovascular mortality needs to be determined urgently, because it could enable the development of enhanced statin based treatments, the development of other drugs in the same class and the identification of potentially modifiable population-wide risk factors for several non-communicable chronic diseases.

18HMG-CoA: 3-hydroxy-3-methyl-glutaryl-CoA reductase; PPAR: peroxisome proliferator-activated receptor; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT: randomized controlled trial.

19BJC has received research funding from MedImmune Inc., and consults for Crucell MV. The authors report no other potential conflicts of interest.

20CMS originated the idea for the paper. CMS and SLAY carried out the systematic search. GF and BJC did the data extraction and analysis. CMS drafted the paper with assistance from SLAY, GF and BJC. All authors reviewed the paper critically and have read and approved the manuscript for publication. All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. CMS is the guarantor. All authors have read and approved the final manuscript.

21The pre-publication history for this paper can be accessed here:

22http://www.biomedcentral.com/1741-7015/11/57/prepub