Research literature

TRAVERSE Explained: The Testosterone Cardiovascular Safety Trial

Published October 26, 2026 · 6 min read

Large safety trials are rare, expensive and slow, which is why the few that exist get cited for years. TRAVERSE is one of them. It was designed to answer a narrow regulatory question that hung over testosterone products for a decade: in middle-aged and older men with low testosterone, does replacement raise the risk of major heart events? This article explains the design, the main result, the secondary safety findings that got less attention, and what the trial cannot say. It is a good worked example for the framework in our guide to reading a clinical trial result.

Why the trial was done

Whether testosterone products raise cardiovascular risk had been debated for years, and TRAVERSE was designed to test it directly in a large randomized trial. It was sponsored by a consortium of testosterone manufacturers and overseen by the Cleveland Clinic Coordinating Center for Clinical Research, with Labcorp Drug Development maintaining the clinical database (source 3). That funding structure is worth knowing: it does not by itself make the result wrong, but readers should know who paid.

Design

TRAVERSE was a randomized, double-blind, placebo-controlled noninferiority trial run at 316 US sites. It randomized 5,246 men aged 45 to 80 who had symptoms of hypogonadism and two fasting testosterone measurements below 300 ng/dL, and who either already had cardiovascular disease or were at high risk of it (sources 1 and 2). In the full analysis set of 5,204 men, 2,847 had pre-existing cardiovascular disease and 2,357 had elevated risk (source 2). Participants used a daily 1.62% transdermal gel, with the dose adjusted to keep blood levels in a target range, or an identical placebo gel (source 1). The study was event-driven, meaning it continued until enough primary events had occurred. Mean treatment duration was about 21.7 months and mean follow-up about 33 months (source 1). Reports describe the duration slightly differently, so we use these figures and note the difference.

The primary endpoint, called MACE, was the time to the first occurrence of cardiovascular death, nonfatal myocardial infarction or nonfatal stroke (sources 1 and 3).

What "noninferiority" means here

Most trials ask whether a treatment is better than placebo. A noninferiority trial asks something different: is the treatment not worse than placebo by more than a pre-set margin? TRAVERSE defined noninferiority as the upper limit of the 95% confidence interval for the hazard ratio being below 1.5 (source 1). That was the line the result had to stay under. It is a deliberately tolerant bar for a safety question, because the aim was to rule out a large increase in risk rather than to detect a small one. A noninferiority result therefore does not mean the two groups were proven equal.

The primary result

Source: sources 1 and 2. A hazard ratio of 0.96 with a confidence interval that spans 1 means the trial found no detectable difference in the primary outcome; it does not show that testosterone reduced events. A sensitivity analysis that censored follow-up more than 365 days after the last dose gave a hazard ratio of 1.02 (95% CI 0.81 to 1.27), with 154 events (5.9%) versus 152 (5.8%) (source 2). The similarity of the two analyses matters because high discontinuation can blur a trial's result: about 61% of men in each group had stopped the study drug by the end (source 2). When many participants stop treatment, the trial is less able to detect a difference that exists only while the drug is being taken.

The safety findings that got less attention

The headline was reassuring on heart attack, stroke and cardiovascular death. The same trial also reported imbalances in several other events, listed here because a safety trial should be read in full:

• Nonfatal arrhythmias requiring intervention: 134 men (5.2%) on testosterone versus 87 (3.3%) on placebo (P = .001) (source 2).
• Acute kidney injury: 2.3% versus 1.5% (P = .04) (source 2).
• Atrial fibrillation and pulmonary embolism: both were reported more often with testosterone (source 1). A 2024 review article gives atrial fibrillation counts of 91 versus 63 (P = .02) and notes that fracture counts were also higher on testosterone, 91 (3.5%) versus 64 (2.4%), mostly fall-related and early (source 4). That review expresses its author's own interpretation, so we cite its counts but not its conclusions.
• Prostate cancer: 12 men (0.5%) versus 11 (0.4%), P = .87, with high-grade cancers also similar and low in number in the review's account (sources 2 and 4).

The investigators themselves described the arrhythmia and kidney findings as unexpected, according to the HCPLive report (source 2). Safety imbalances in secondary events can arise by chance when many outcomes are compared, but they can also be real, and trials are not usually powered to tell the difference. That is a reason to read them as signals for further study, not as findings of harm and not as noise.

What the trial cannot show

• Duration. Mean follow-up of under three years does not speak to what happens over many years of exposure.
• Population. Participants were men aged 45 to 80 with two low testosterone readings, symptoms, and cardiovascular disease or risk. The results do not transfer to younger men, women, or people with normal testosterone levels.
• Formulation. One daily gel was tested. Results for other forms and regimens cannot be assumed to match.
• Treatment exposure. With about 61% discontinuing in each arm, intention-to-treat results reflect the policy of assigning treatment, not the effect of taking it continuously.
• Arrhythmia capture. According to the 2024 review, the trial did not record scheduled electrocardiograms, so atrial fibrillation was captured only as reported medical events (source 4).
• Efficacy. TRAVERSE was a safety trial. Its secondary reports on symptoms, bone, anemia and diabetes are separate analyses and are not covered in this article.
• Funding. It was industry-sponsored, as noted above.

How to read it

TRAVERSE is a good example of two things at once. On its pre-specified primary question, the answer was that testosterone gel was not inferior to placebo for major cardiovascular events within the trial's window and population. On the broader safety picture, it also found more arrhythmias and acute kidney injury in the testosterone group, and its length, population and discontinuation rate limit how far the result reaches. Holding both facts together is what separates a careful reading from either a reassurance or a warning. For the general toolkit (estimands, confidence intervals, what a noninferiority margin does), see How to Read a Clinical Trial Result, and for why a single account of someone's experience cannot answer a question like this, see Personal Accounts vs. Trial Data.

Sources

1. "Cardiovascular Safety of Testosterone-Replacement Therapy," abstract of Lincoff AM et al., N Engl J Med 2023;389:107-117, doi:10.1056/NEJMoa2215025 (abstract text as reproduced by UroToday; the journal page itself was not opened): urotoday.com
2. HCPLive, "TRAVERSE trial: no increased cardiovascular risk with testosterone replacement therapy": hcplive.com
3. Cleveland Clinic newsroom, "Cleveland Clinic-led trial finds testosterone replacement therapy does not increase incidence of major adverse cardiac events for men with low testosterone," June 16, 2023: newsroom.clevelandclinic.org
4. Hackett G, review of the TRAVERSE trial, 2024 (PMC; author's own interpretation, used here only for counts): pmc.ncbi.nlm.nih.gov

Related reading

• How to Read a Clinical Trial Result
• Personal Accounts vs. Trial Data
• Glossary

← Back to all articles