In 2013, Vigen and colleagues reported that among male veterans with low testosterone who had undergone coronary angiography, those who started testosterone therapy had a higher rate of death, myocardial infarction and stroke: a Kaplan-Meier cumulative event estimate of 25.7% at three years versus 19.9% in untreated men.[1] The finding, together with a meta-analysis published shortly after, drove the FDA's cardiovascular warning onto every testosterone label — where it stayed until 2025.
It is the most consequential testosterone study of the last two decades, and understanding what it did and did not show is how the rest of the evidence base makes sense.
What it did
- Design: retrospective national cohort study.
- Population: men with total testosterone below 300 ng/dL who underwent coronary angiography in the Veterans Affairs system between 2005 and 2011. 8,709 men met criteria.
- Exposure: 1,223 patients started testosterone therapy, at a median of 531 days after angiography. 7,486 did not.
- Primary outcome: composite of all-cause mortality, myocardial infarction and ischemic stroke.
What it found
Across 1,710 outcome events: in the no-therapy group, 681 died, 420 had MIs and 486 had strokes. In the therapy group, 67 died, 23 had MIs and 33 had strokes.[1]
Kaplan-Meier estimated cumulative percentages with events:[1]
| Time after angiography | No testosterone | Testosterone |
|---|---|---|
| 1 year | 10.1% | 11.3% |
| 2 years | 15.4% | 18.5% |
| 3 years | 19.9% | 25.7% |
Absolute risk difference at 3 years: 5.8% (95% CI, −1.4% to 13.1%).[1]
There was no significant difference in effect size between men with and without coronary artery disease (test for interaction, P = .41).
The interval that rarely gets quoted
The headline difference is 5.8 percentage points. Its 95% confidence interval runs from −1.4% to 13.1% — and it includes zero.
An interval spanning zero is consistent with testosterone therapy being associated with slightly fewer events as well as with substantially more. On this measure, the study's own data did not exclude no difference.
That is not a technicality. A warning applied to an entire drug class for ten years rested substantially on a retrospective cohort whose absolute risk estimate could not distinguish harm from no effect.
The other structural problems
It is retrospective and observational. Men were not randomized. The clinicians who prescribed testosterone chose whom to prescribe it to, after an angiogram — a decision informed by clinical information the model cannot fully recover.
The exposure timing is unusual. Therapy began a median of 531 days — nearly eighteen months — after the angiography that defined the cohort. A great deal happens in eighteen months, and the cohort's clock starts at a procedure rather than at treatment.
The event counts in the treated arm are small. Twenty-three myocardial infarctions and thirty-three strokes among 1,223 men. Estimates built on counts that size move considerably on small reclassifications.
The published record carries corrections. The abstract as indexed contains explicit [corrected] annotations on the results text, indicating that the reported figures were amended after initial publication.
What replaced it
The proper answer to a retrospective cohort is a randomized trial, and eventually the field ran one. TRAVERSE enrolled 5,246 men — a population deliberately selected for existing cardiovascular disease or elevated risk, much like Vigen's — randomized them, and followed them a mean 33 months. Major adverse cardiac events occurred in 7.0% on testosterone versus 7.3% on placebo, hazard ratio 0.96.
The Hudson individual-patient-data meta-analysis had already reached a compatible answer from pooled trial data, and the 41-trial meta-analysis has since confirmed it. In February 2025 the FDA removed the cardiovascular warning.[2]
What it is fair to take from it
Not that the researchers did poor work — a retrospective cohort was a reasonable thing to publish when no adequately powered trial existed, and the paper reported its own wide confidence interval.
The lesson is about what happened next. A finding whose absolute risk interval crossed zero became a class-wide label warning, and the warning shaped prescribing, insurance coverage and patient fear for a decade before a trial was run to test it.
The full narrative arc — this study, the meta-analysis that followed, and the warning — is told in how three years of contested evidence became a warning on every label.
Bottom line
Vigen 2013 reported higher rates of death, heart attack and stroke among treated men in a retrospective VA cohort, with a three-year absolute risk difference of 5.8% whose confidence interval included zero. It triggered a decade-long regulatory warning that randomized evidence has since not supported. Cite it as the origin of the controversy, not as evidence of harm.
References
2 sources
- Vigen R, O'Donnell CI, Barón AE, et al. Association of testosterone therapy with mortality, myocardial infarction, and stroke in men with low testosterone levels. JAMA. 2013;310(17):1829–1836. doi:10.1001/jama.2013.280386 · PMID 24193080
- U.S. Food and Drug Administration. FDA issues class-wide labeling changes for testosterone products. February 28, 2025. fda.gov
Practical questions about membership, labs and using the spray are answered in the Learn Hub FAQ.