What Are the Real Limitations of the TRAVERSE Bone Fracture Substudy?

At a glance

| Parameter | Detail | |---|---| | N | 5,204 randomized (of 5 to 246 in the parent TRAVERSE trial) | | Intervention | 1.62% testosterone gel (AndroGel), dose-adjusted to maintain serum T 350-750 ng/dL | | Comparator | Matching placebo gel | | Duration | Median follow-up 3.19 years | | Primary endpoint | First clinical fracture (confirmed by radiographic report or surgical record) | | Key result | Fractures occurred in 91 testosterone-treated men vs. 64 placebo-treated men (HR 1.43; 95% CI 1.04-1.97) |

Why This Substudy Commands Attention

Before 2024, the weight of evidence from smaller trials suggested testosterone either preserved or modestly improved bone mineral density (BMD) in hypogonadal men. The TTrials bone substudy, for example, showed meaningful BMD gains at the spine and hip with one year of testosterone gel. So the TRAVERSE Bone Fracture Substudy landed as a genuine surprise: a statistically significant 43% relative increase in clinical fractures among men receiving testosterone versus placebo over roughly three years.

The parent TRAVERSE trial was the largest randomized testosterone trial ever completed, designed primarily to evaluate cardiovascular safety. The bone fracture analysis was a prespecified secondary outcome of that cardiovascular trial, not the primary question it was powered to answer. This distinction matters for every number that follows.

Limitation 1: Post-Hoc Power and the Substudy Architecture

TRAVERSE was designed and powered for its primary cardiovascular endpoint (major adverse cardiac events). The bone fracture analysis was prespecified but not independently powered. The investigators themselves acknowledge that the fracture outcome accrued only 155 total events across both arms. For context, most fracture-endpoint trials in osteoporosis (FREEDOM, ARCH, FRAME) are powered for 300 to 800+ fracture events.

With 155 events, the confidence interval around the hazard ratio stretches from 1.04 to 1.97. The lower bound barely clears 1.0. A handful of reclassified fractures in either direction could shift the result to non-significance. This does not mean the finding is wrong. It means the estimate is imprecise and should be interpreted as a signal requiring confirmation, not a settled effect size.

Limitation 2: No DXA or Bone-Turnover Data in the Full Cohort

The parent TRAVERSE trial did not collect serial DXA scans or bone turnover markers across the full 5,204-person cohort. Without these data, there is no mechanistic bridge between the fracture signal and what testosterone was actually doing to bone. Was BMD declining? Were resorption markers rising? We cannot say.

This gap is especially frustrating because the earlier TTrials bone substudy (N=211) showed significant volumetric BMD increases at the spine and hip with testosterone over 12 months. If testosterone genuinely improves bone density but somehow increases fracture risk, the mechanism demands explanation. Falls, sarcopenia patterns, changes in physical activity, or even fracture-type distribution could account for the discrepancy, but TRAVERSE did not collect the data to test these hypotheses.

Limitation 3: Falls Were Not Captured

Clinical fracture risk in older men is driven as much by fall frequency and fall mechanics as by bone strength. The TRAVERSE protocol did not systematically record falls or fall-related injuries. This is a critical omission.

Testosterone can increase hematocrit, and some evidence links erythrocytosis to dizziness and cardiovascular symptoms that could theoretically increase fall propensity. Testosterone also raises energy and physical activity levels. More active older men may simply sustain more trauma. Without fall data, the study cannot distinguish between a bone-weakening effect and a fall-increasing effect, and the clinical implications of each scenario differ dramatically.

Limitation 4: Enrollment Biases and the Study Population

TRAVERSE deliberately enrolled men aged 45-80 with pre-existing cardiovascular disease or high cardiovascular risk factors. The mean age was 63, BMI averaged 33, and over 70% had hypertension, diabetes, or both. This was not a general hypogonadal population. It was a cardiometabolically burdened cohort selected because the FDA wanted cardiovascular safety data.

Several characteristics of this population introduce fracture confounders:

| Confounder | Prevalence in TRAVERSE | Relevance to Fracture | |---|---|---| | Type 2 diabetes | ~40% | Diabetic bone is brittle despite normal/high BMD on DXA | | Obesity (BMI >30) | ~65% | Alters fall mechanics, may mask DXA accuracy | | Statin use | >60% | Some observational data link statins to fracture, though conflicting | | Opioid use | Not reported in detail | Opioids suppress gonadal axis and increase fall risk | | Glucocorticoid history | Not stratified | Steroid-related bone loss is dose-dependent |

Men with type 2 diabetes deserve special attention. Diabetic bone fragility operates through impaired bone quality rather than reduced BMD, making DXA an unreliable surrogate. Whether testosterone interacts with diabetic bone metabolism differently than with healthy bone is unknown. The heavy representation of diabetes in TRAVERSE raises the question of whether the fracture signal is specific to this metabolic phenotype rather than generalizable to all hypogonadal men.

Limitation 5: Fracture Adjudication and Classification

Fractures were ascertained by participant self-report and confirmed through medical record or radiology review. The investigators did not use a central adjudication committee dedicated to fracture endpoints, which is standard in osteoporosis registration trials like FREEDOM (denosumab) or ARCH (romosozumab).

The fracture types also warrant scrutiny. The testosterone arm showed excess fractures at multiple sites, but the numbers per site are small. Whether these were predominantly fragility fractures (suggesting a bone-quality problem) or traumatic fractures (suggesting a falls or activity problem) is not fully parsed in the published analysis. This distinction is clinically meaningful: a therapy that weakens bone demands a fundamentally different risk-benefit conversation than one that increases physical activity in a frail population.

Limitation 6: The Testosterone-Estradiol Axis Question

Testosterone is aromatized to estradiol, and estradiol, not testosterone, is the dominant hormonal regulator of bone resorption in men. The TRAVERSE investigators did not report serial estradiol levels in the bone substudy. This is a significant gap.

Exogenous testosterone gel can raise estradiol in some men and suppress endogenous gonadotropins, potentially altering the testosterone-to-estradiol ratio in unpredictable ways. If aromatization was insufficient in a subset of participants (those with high BMI, for instance, may have paradoxically variable aromatase activity), the bone-protective estradiol signal could have been blunted even as testosterone levels rose. Without estradiol data, this remains speculation, but it is testable speculation that the trial design left unaddressed.

Limitation 7: Conflict of Interest and Funding

TRAVERSE was funded by a consortium that included AbbVie, the manufacturer of AndroGel (testosterone 1.62% gel). AbbVie provided the drug and placebo and had the contractual right to review the manuscript before submission, though the investigators retained final authority over content and publication decisions.

This arrangement is standard in large pharma-funded trials, but it introduces a structural tension worth acknowledging. The primary TRAVERSE trial found that testosterone did not increase major cardiovascular events, a result favorable to the manufacturer. The bone fracture substudy, which showed harm, was published as a separate paper. Whether the timing and framing of these results were influenced by commercial interests is unknowable from the outside, but readers should note the funding source and evaluate accordingly.

What Subsequent Commentary Has Surfaced

Letters to the editor and editorials following the TRAVERSE bone paper raised several of the points above. Key themes from published responses include:

• The falls hypothesis: Multiple correspondents emphasized that fracture outcomes without falls data are incomplete, particularly in a population with high rates of polypharmacy and obesity.
• Fragility vs. trauma fractures: Editorialists questioned whether the excess fractures were osteoporotic in nature or traumatic, a distinction the data could not resolve.
• Generalizability concerns: Several commentators noted that the cardiometabolic enrichment of TRAVERSE makes extrapolation to younger hypogonadal men or men without diabetes problematic. The Endocrine Society's 2018 guidelines on testosterone therapy were written before these data were available and reference BMD improvement as a potential benefit of TRT in men with osteoporosis.
• Effect size uncertainty: The wide confidence interval (1.04-1.97) was repeatedly flagged. A hazard ratio of 1.04 and a hazard ratio of 1.97 imply very different clinical realities, and the trial cannot distinguish between them.

Putting It in Context

The TRAVERSE bone fracture signal should not be dismissed. A 43% relative increase in clinical fractures in a 5,000-person randomized trial is a finding that warrants clinical caution and further study. Clinicians prescribing testosterone to older men with cardiovascular comorbidities should discuss this result.

At the same time, this substudy does not establish that testosterone causes fractures through a bone-weakening mechanism. The absence of DXA data, falls data, estradiol levels, and detailed fracture-type adjudication leaves multiple plausible explanations on the table. Future trials, or re-analyses of TRAVERSE biobank samples if available, could clarify whether this signal reflects genuine skeletal harm, increased fall exposure, or an artifact of the study population's unique risk profile.

Until then, the honest assessment is: the TRAVERSE Bone Fracture Substudy produced a statistically significant but imprecise signal of fracture harm with testosterone, in a population that may not represent most men receiving TRT, using methods that could not identify the mechanism. Caution is warranted. Certainty is not.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Snyder PJ, Bauer DC, Engel SS, et al. "Testosterone treatment and fractures in men with hypogonadism." N Engl J Med. 2024;390(3):203-211. PubMed
2. Snyder PJ, Kopperdahl DL, Stephens-Shields AJ, et al. "Effect of testosterone treatment on volumetric bone density and strength in older men with low testosterone: a controlled clinical trial." JAMA Intern Med. 2017;177(4):471-479. PubMed
3. Bhasin S, Brito JP, Cunningham GR, et al. "Testosterone therapy in men with hypogonadism: an Endocrine Society clinical practice guideline." J Clin Endocrinol Metab. 2018;103(5):1715-1744. PubMed
4. Schwartz AV, Vittinghoff E, Bauer DC, et al. "Association of BMD and FRAX score with risk of fracture in older adults with type 2 diabetes." JAMA. 2011;305(21):2184-2192. PubMed
5. Cummings SR, San Martin J, McClung MR, et al. "Denosumab for prevention of fractures in postmenopausal women with osteoporosis (FREEDOM trial)." N Engl J Med. 2009;361(8):756-765. PubMed
6. FDA label: AndroGel (testosterone gel) 1.62%. FDA