Dr. Weeks’ Comment: Dr. Shippen is a brilliant endocrinologist and he clearly discredits the sloppy science behind the hack job published earlier this year.
To the JAMA editor:
The article by Vigen et al has so many design flaws that the conclusions of increased risks of testosterone treatment should be questioned.(1)
1. The most important design flaw is the choice of angiography as the primary entry variable that had NO effect on outcomes. Previous VA studies and their own current analyses demonstrate no risk differences for treated or non-treated men for cardiovascular disease mortality(2). Therefore, this variable not only had no bearing on the results directly, but it resulted in many time-related statistical complexities regarding treatment timing and event relationships.
2. Additionally, this variable resulted in substantial exclusions that would have significantly increased the cohort numbers and important comparative results. For example, 2798 were excluded because testosterone treatment was started before angiography.
3. If the actual deaths in the treated (67/1223 – 5-5%) and untreated cohorts (681/7486 – 9.1%) are simply analyzed, there is an amazingly similar reduction in actual death rates of 39.6% (5.5/9.1%) in the treated cohort. This is dramatically similar to the 39% reduction in overall mortality in testosterone treated men at VA facilities reported by Shore et al (2). The complexities of statistical analysis due to time and treatment variables reverses this clear mathematical data. How can this dramatic reversal to a 30% INCREASE occur when the RAW actual data say the opposite occurred! That is a 70% difference between actual and analyzed (manipulated) statistics?
4. A serious flaw regards the time of testosterone treatment and relationship to event outcomes. For example, 1 in 6 of treated men filled only one prescription, yet these men were entered into the treatment cohort for the duration of the study. Duration of treatment or the time between treatments and endpoints is not clearly reported. It would be difficult, for example, to relate treatment for 1-3 months to an event 2-3 years later.
5. Testosterone testing post-treatment was available for only 60% of the treatment cohort and was reported as 332.2 ng/dl for the “first repeat testosterone measurement”. This low value suggests poor compliance with any of the treatments. Where are the testing data for any additional tests over the treatment periods? There is poor treatment data for 60% and NO reliable data to corroborate adequate treatment for 40% of the cohort. Who can rely on the published conclusions?
Eugene Shippen, M. D.
Medical Staff, St. Joseph Hospital,
2. Shores MM, et al. Testosterone treatment and mortality in men with low testosterone levels. J Clin Endocrinol Metab. 2012 June;97(6):2050-8.