Why TRAVERSE Settled It: The Design Decisions That Made the Trial Decisive

Between 2010 and 2023, the cardiovascular safety of testosterone therapy was argued over using a halted trial, two conflicting observational cohorts and a meta-analysis of trials designed for other endpoints. TRAVERSE ended the argument — not because it was larger, though it was, but because of four decisions made before enrolment opened.

This page is about those decisions. The results themselves are in the TRAVERSE main trial.

Decision one: make it a non-inferiority trial with a pre-set margin

Most trials ask whether a treatment is better. TRAVERSE asked whether testosterone was not meaningfully worse, and — crucially — defined in advance what "not meaningfully worse" meant:

Non-inferiority required an upper limit of less than 1.5 for the 95% confidence interval of the hazard ratio among patients receiving at least one dose.[1]

This is the design choice that does the real work. A conventional trial that finds no significant difference has failed to detect an effect, which is not the same as showing there isn't one. By committing to a margin beforehand, TRAVERSE made a null result interpretable: the observed hazard ratio was 0.96 (95% CI 0.78–1.17), and the upper bound of 1.17 sits well under the 1.5 threshold.[1]

"We didn't find a difference" and "we ruled out a difference larger than 50%" are different claims. Only the second settles anything, and only a pre-specified margin lets you make it.

Decision two: enrol the men most likely to have an event

TRAVERSE enrolled 5,246 men aged 45 to 80 who had preexisting cardiovascular disease or high risk of it, reported symptoms of hypogonadism, and had two fasting testosterone levels below 300 ng/dL.[1]

Deliberately enriching for cardiovascular risk is counterintuitive if you are hoping for a clean safety result. It is exactly right if you want the answer to mean something. Events are what a time-to-event trial runs on; a low-risk population produces too few of them to distinguish anything.

It also pre-empts the obvious objection. A safety signal that fails to appear in men already at high cardiovascular risk is more reassuring than one that fails to appear in healthy men — and it directly addresses the population the earlier scare was about. TOM had studied 209 frail men for six months; this studied 5,246 at-risk men for years.

Note the eligibility detail: two fasting testosterone measurements, not one. The trial applied the diagnostic standard that the prescribing boom of the 2000s often did not — see how prescribing tripled.

Decision three: treat to a target, not a fixed dose

Participants received daily transdermal 1.62% testosterone gel, dose-adjusted to maintain testosterone between 350 and 750 ng/dL, or placebo gel.[1]

Titrating to a range rather than fixing a dose means the trial tested testosterone therapy as it is actually practised. A fixed-dose trial would have left the question of whether the result applied at real-world doses permanently open.

Decision four: a hard composite endpoint, over a long enough period

The primary endpoint was the first occurrence of death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke, in a time-to-event analysis. A secondary endpoint added coronary revascularisation.[1]

These are events, not surrogates — no blood pressure readings standing in for strokes, no lipid panels standing in for infarcts. And the exposure was substantial: mean treatment duration 21.7 ± 14.1 months, mean follow-up 33.0 ± 12.1 months.[1]

The trial was also multicentre, randomised, double-blind and placebo-controlled — which removes, at a stroke, the confounding-by-indication problem that neither Vigen nor Sharma could escape, and which is why those two observational cohorts reached opposite conclusions from similar data.

What the design did not protect against

Being fair about a well-designed trial includes naming what it could not do.

1. Non-inferiority on MACE says nothing about other endpoints. The trial reported a higher incidence of atrial fibrillation, acute kidney injury and pulmonary embolism in the testosterone group.[1] Those are real findings inside a trial usually summarised as reassuring.
2. A safety trial is not an efficacy trial. TRAVERSE tested whether testosterone is safe for the heart, not whether it should be prescribed. Benefit is a separate question, answered — narrowly — by the T Trials.
3. The sub-studies do not all agree with the headline. Bone found more fractures on treatment.[2] The set is worth reading individually, starting with TRAVERSE and bone.
4. The population is specific. Men 45–80 with hypogonadism and cardiovascular disease or risk. Extrapolating to a healthy 35-year-old is not supported, in either direction.

Why it mattered beyond the label

TRAVERSE was an FDA-required postmarketing safety trial, arising from a September 2014 advisory committee recommendation for an industry-wide study.[3] Its results led directly to the February 2025 class-wide labeling changes, and to the further prescribing-information updates the FDA requested in June 2026 — both covered in what the 2025 label change did.

A regulator asked a question, specified the trial, and acted on the answer. That sequence is rarer than it should be.

Bottom line

TRAVERSE was decisive because it committed in advance to what would count as an answer, enrolled the population where the answer mattered most, dosed the way clinicians actually dose, and measured events rather than proxies.

The result — hazard ratio 0.96, upper confidence bound 1.17, against a pre-set margin of 1.5 — is a genuine ruling-out rather than a failure to find. That is the difference between this trial and everything that preceded it.

References

3 sources

1. Lincoff AM, Bhasin S, Flevaris P, et al. Cardiovascular Safety of Testosterone-Replacement Therapy. N Engl J Med. 2023;389(2):107–117. doi:10.1056/NEJMoa2215025 · PMID 37326322
2. Snyder PJ, Bauer DC, Ellenberg SS, et al. Testosterone Treatment and Fractures in Men with Hypogonadism. N Engl J Med. 2024;390(3):203–211. doi:10.1056/NEJMoa2308836 · PMID 38231621
3. U.S. Food and Drug Administration. FDA issues class-wide labeling changes for testosterone products. February 28, 2025. fda.gov