Key takeaways
- TRAVERSE randomised 5,246 symptomatic men aged 45-80 with confirmed low testosterone and elevated cardiovascular risk to testosterone or placebo, with a mean follow-up of roughly 33 months.
- On the primary endpoint — cardiovascular death, non-fatal myocardial infarction and non-fatal stroke — testosterone was non-inferior to placebo.
- Testosterone raises hematocrit, can raise blood pressure modestly in some men, and shifts lipids in ways that vary by route and dose.
- The study that drove the alarm was a retrospective cohort of men in the Veterans Affairs system, which reported higher rates of death, myocardial infarction and stroke among testosterone users (Vigen et al., JAMA 2013).
- Testosterone met the non-inferiority criterion on the primary endpoint.
For about a decade, the sentence that ended most testosterone conversations was "but doesn't it cause heart attacks?" It was a reasonable question. It was based on data that turned out to be weak, it produced a regulatory warning, and it left a large number of men with genuine hypogonadism untreated while the field waited for a trial that could actually answer it. That trial has now reported. This is what it asked, what it found, and — the part that gets skipped — what it did not.
Why the question existed at all
Testosterone raises hematocrit, can raise blood pressure modestly in some men, and shifts lipids in ways that vary by route and dose. Those are real physiological effects with plausible cardiovascular implications, so the concern was never irrational. The problem was that the evidence used to act on it was observational, and observational data on testosterone is unusually treacherous: men who start treatment differ systematically from men who do not, and low testosterone is itself a marker of poor health rather than only a cause of it.
What the earlier data actually was
The study that drove the alarm was a retrospective cohort of men in the Veterans Affairs system, which reported higher rates of death, myocardial infarction and stroke among testosterone users (Vigen et al., JAMA 2013). It attracted immediate methodological criticism over how outcomes were classified and how the cohort was constructed, and it was subsequently corrected. It was never a randomised trial and could not establish causation.
Around it sat a set of older observational studies with the usual confounding problem — men starting therapy often carried multiple cardiovascular risk factors, which is frequently why their testosterone was low — and a small number of trials in frail elderly men using dosing that does not resemble modern practice. On that basis a regulatory label warning was added, prescribing fell, and the uncertainty hardened into received wisdom.
None of this meant testosterone was safe. It meant nobody knew, and a proper trial was overdue.
The trial that was designed to settle it
TRAVERSE — Testosterone Replacement Therapy for Assessment of Long-term Vascular Events and Efficacy Response in Hypogonadal Men — was built specifically to test cardiovascular safety, and its design choices are worth understanding because they determine what the result means:
- 5,246 men aged 45-80, with confirmed hypogonadism (testosterone below 300 ng/dL) and symptoms
- Deliberately high cardiovascular risk — existing cardiovascular disease or multiple risk factors. This was not a healthy population; it was the population in whom harm, if real, would show first
- Randomised to testosterone gel or placebo, which removes the confounding that ruined the observational literature
- Mean follow-up around 33 months
- A non-inferiority design with a primary composite endpoint of cardiovascular death, non-fatal myocardial infarction and non-fatal stroke
That last point matters more than any other. A non-inferiority trial is built to answer "is this worse?" — not "is this better?" (Lincoff et al., N Engl J Med 2023).
What it found
Testosterone met the non-inferiority criterion on the primary endpoint. Rates of cardiovascular death, heart attack and stroke were similar between groups. Two secondary signals were more frequent on testosterone: atrial fibrillation (3.5% versus 2.4%) and pulmonary embolism. Acute kidney injury did not differ.
In plain terms: in symptomatic men with confirmed low testosterone and elevated cardiovascular risk, replacement did not increase major adverse cardiovascular events over roughly three years. The decade-old question about heart attacks and strokes has an answer, and the answer is no.
What the trial did not establish
This is where the result is most often overstated, so it is worth being precise.
- It did not show cardiovascular benefit. Non-inferiority means "not worse." Anyone citing TRAVERSE as evidence that testosterone protects the heart is reading a safety trial as an efficacy trial.
- It did not licence treatment without a diagnosis. Every participant had confirmed biochemical hypogonadism plus symptoms. It says nothing about men with normal testosterone taking it anyway.
- It did not test high or supraphysiologic doses. It studied replacement to a physiologic range, titrated. Findings do not transfer to doses used outside medical supervision.
- It did not run indefinitely. Roughly three years of follow-up is substantial; it is not a statement about twenty.
- It did not resolve every endpoint. The same programme's fracture analysis produced an unexpected result in the other direction, which is a reminder that "cardiovascularly reassuring" is not "reassuring about everything."
The two signals that stayed
The atrial fibrillation and pulmonary embolism findings deserve attention rather than dismissal.
The absolute difference in atrial fibrillation was small, and the mechanism is not established. It is enough to change the conversation for a man with a history of AF, palpitations or significant structural heart disease — not enough, on its own, to change practice for everyone else.
Pulmonary embolism is usually discussed in the context of erythrocytosis, the rise in red cell mass that testosterone reliably produces. That effect is well described across the treatment literature (testosterone and erythrocytosis, meta-analysed), and it is the single most common reason a protocol needs adjusting. It is also the most manageable: it is visible on a routine complete blood count long before it causes anything, and it responds to dose reduction, a change of delivery route, or therapeutic phlebotomy.
The clinical pearl: TRAVERSE removed cardiovascular fear as the reason not to treat confirmed, symptomatic hypogonadism. It did not remove the reason to monitor. Hematocrit, blood pressure and lipids are checked on treatment because testosterone measurably moves them — and because the things that actually go wrong on a well-run protocol are the things a routine panel catches early.
What monitoring should look like
- Hematocrit at baseline and at intervals — the most important single number on treatment
- Blood pressure, measured properly rather than recalled
- A lipid panel including ApoB, which describes atherogenic particle burden better than LDL-C alone
- Total and free testosterone plus sensitive estradiol, to confirm the protocol is landing where intended
- An honest history around prior clots, atrial fibrillation and family history before starting
Bottom line
A large randomised trial in exactly the men who were most feared to be at risk found no increase in cardiovascular death, heart attack or stroke with testosterone replacement over roughly three years. That resolves the question that constrained treatment for a decade. It resolves it as a safety finding, not a benefit claim, in men with confirmed low testosterone and symptoms, at physiologic replacement doses, with monitoring. The residual atrial fibrillation and pulmonary embolism signals are small, real, and reasons for individual evaluation rather than blanket avoidance — and hematocrit remains the number that gets checked on every follow-up panel, because it is the one that moves.
Educational content, not medical advice. Laboratory interpretation and any treatment decision are made by a licensed physician after individual evaluation. Individual results vary.
