Response by DR. MORGENTALER
Results of a newly reported study have inflamed concerns regarding testosterone therapy and the risk of cardiovascular disease (CVD). Here we review the current evidence about testosterone and cardiovascular risk and analyze results of the study.
Vigen et al retrospectively evaluated the charts of all male veterans who underwent coronary angiography in the US Department of Veterans Affairs (VA) hospital system between 2005 and 2011 and who had a serum testosterone level <300 ng/dL.1 Excluded from the study were men who started testosterone therapy prior to coronary angiography or before having a testosterone level checked or did not conform to Endocrine Society guidelines.2 The authors reported increased rates of death and cardiovascular events in men who received testosterone therapy for hypogonadism compared with those who did not receive testosterone therapy.1 These results, published in the November 2013 issue of JAMA, are called into question, however, based on a large body of literature showing just the opposite as well as by analysis of the study results, which shows questionable statistical manipulation and methodological flaws.
The authors reported that, at 3 years after angiography, the group receiving testosterone therapy had a higher event rate than those not receiving testosterone therapy (25.7% vs 19.9%, respectively), with an "absolute risk difference" of 5.8%.1 This is an invalid and misleading conclusion, because an absolute risk difference must be based on raw data, whereas the 5.8% figure was derived from Kaplan-Meier estimates of events. In fact, based on the raw data, 67 deaths, 23 MIs, and 33 strokes occurred among 1223 men who received testosterone therapy, for an event rate of 10.1% (123 of 1223 men). Among 7486 men who did not receive testosterone therapy, 681 deaths, 420 MIs, and 486 strokes occurred, for an event rate of 21.2% (1587 of 7486 men). Thus, the actual absolute rate of events was approximately twice as high among men who did not receive testosterone therapy.
After publication, the authors were required to revise their article, replacing the term "absolute risk" with "Kaplan-Meier estimated cumulative percentages with events," which more accurately reflects the highly statistical methodology in which adjustments were made for more than 50 variables. Sophisticated statistical analysis most certainly plays a vital role in biomedical research, but both investigators and readers must understand that the further they are removed from the raw data, the less reliable results become. Of particular concern in this instance are the facts that the actual event rate is less than half of that reported after statistical manipulation and that the relationship of the estimated event rates is completely reversed.
Methodological problems with this study include a muddled dataset. Whereas most studies compare results from distinct groups, all men in this study began in the non-testosterone therapy group.1 Upon filling a prescription for testosterone, men were reassigned to the testosterone therapy group. Therefore, all men in the treated group also contributed data to the untreated group. An event was attributed to the testosterone therapy group if a man filled his testosterone prescription the same day. Creating two distinct groups required multiple statistical assumptions with questionable reliability.
An even more serious methodological error was the exclusion of 1132 men who received testosterone therapy after having an MI or a stroke. Once an event occurred, these men had completed their contribution to the study,1 making subsequent testosterone therapy irrelevant. All of these events should have been attributed to the non-testosterone therapy group, which would raise the event rate for that group by 71.8% and would almost certainly result in, even after statistical manipulation, a higher cardiovascular risk for the non-testosterone therapy compared with the testosterone therapy group.
Finally, and of no small importance, a considerable body of evidence accumulated over more than 20 years, including numerous individual studies and two meta-analyses,3,4 indicates that men who have normal serum testosterone levels have a lower risk of CVD and overall mortality compared with hypogonadal men5,6 and that testosterone therapy in middle-aged to elderly men does not increase the incidence of cardiovascular events. (For a more detailed analysis, see Carson CC III. Clinical Consult, December 14, 2013, at www.testosteroneupdate.org/20111213ClinicalConsult.email@example.com.)