TRAVERSE Bone Fracture Substudy Subgroup Analyses: Who Responded Most and Least

At a glance

| Parameter | Detail | |---|---| | N | 5,204 (bone fracture analysis population) | | Intervention | 1.62% testosterone gel (AndroGel), dose-adjusted to target 350-750 ng/dL | | Comparator | Matching placebo gel | | Median follow-up | 3.19 years | | Primary endpoint | Incidence of first clinical fracture (confirmed by radiographic adjudication) | | Key result | HR 1.43 (95% CI 1.04-1.97); 91 fractures (testosterone) vs 64 fractures (placebo) |

Why Subgroup Data Matters Here

When a trial produces a result that contradicts prior evidence, the first clinical question is always: does the signal come from everyone, or from a specific subset? For decades, observational data and smaller RCTs suggested testosterone replacement therapy (TRT) either preserved or improved bone mineral density (BMD) in hypogonadal men. The TTrials Bone substudy showed volumetric BMD gains at the spine and hip with one year of testosterone gel. So when the TRAVERSE fracture substudy reported a 43% increase in clinical fractures, the mismatch demanded granular subgroup interrogation.

The parent TRAVERSE trial enrolled men aged 45 to 80 with serum testosterone <300 ng/dL and either established cardiovascular disease or elevated CV risk. The fracture substudy was pre-specified and adjudicated independently.

Pre-Specified Subgroup Analyses

The TRAVERSE investigators defined several subgroups before unblinding, designed to test whether the fracture signal concentrated in identifiable populations.

Age (<65 vs ≥65 years)

Men 65 and older carried the majority of fracture events in both arms, consistent with age-related bone loss. The hazard ratio point estimate was numerically higher in the ≥65 group, but confidence intervals overlapped substantially with the younger cohort. The interaction p-value did not reach significance (p-interaction > 0.20), meaning the trial could not confirm that age modified the testosterone-fracture relationship. This matters clinically because the Endocrine Society guideline already recommends against routine TRT initiation in men over 65 without clear hypogonadal symptoms, and these data offer no reassurance for that population.

Baseline Testosterone Level

Participants were stratified around the cohort median baseline testosterone (approximately 220 ng/dL). Men with the lowest baseline levels, those most classically "hypogonadal," did not show attenuation of fracture risk. In fact, the point estimate trended toward a larger hazard ratio in men with baseline testosterone below the median. This finding undermines the intuitive hypothesis that the men most deficient in testosterone would benefit most from replacement. The FDA's 2015 label revision for testosterone products already narrowed the indication to men with structural or genetic hypogonadism, and the TRAVERSE fracture data reinforce that caution.

BMI (<30 vs ≥30 kg/m²)

Obesity is common in this population. Roughly 60% of TRAVERSE participants had BMI ≥30. The pre-specified analysis showed that the fracture hazard ratio was numerically larger among men with higher BMI, though again the interaction test was non-significant. One mechanistic hypothesis is that obese men have higher aromatase activity, converting exogenous testosterone to estradiol at rates that may alter bone remodeling dynamics in unpredictable ways. The observation aligns with prior evidence that the relationship between adiposity, sex steroids, and fracture risk is nonlinear.

Prior Fracture History

Men with a documented prior fracture at baseline were expected to be at highest absolute risk. The subgroup analysis showed that these men had high event rates in both arms, but the relative increase with testosterone was actually smaller (point estimate closer to 1.0) than in men without prior fractures. This could reflect a ceiling effect, survivors who already adapted to fracture risk, or simply the small number of events in this subgroup limiting statistical power.

Post-Hoc Exploratory Analyses

Beyond the pre-specified subgroups, the investigators and subsequent commentators examined additional strata.

Race and Ethnicity

The TRAVERSE cohort was predominantly White (approximately 63%), with meaningful representation of Black (15%) and Hispanic (12%) participants. The published fracture data did not report formal race-stratified hazard ratios for the bone substudy, a limitation the authors acknowledged. Given that Black men generally have higher BMD and lower fracture rates at baseline, and that the parent TRAVERSE cardiovascular analysis showed no significant race-based interaction for the primary CV endpoint, the absence of race-specific fracture data represents a gap. Clinicians prescribing TRT to non-White patients cannot assume the overall 43% increase applies uniformly to their population.

Fracture Site

The increase in fractures was not driven by a single skeletal site. Upper extremity, lower extremity, and vertebral fractures all contributed events. The pattern suggests a systemic rather than site-specific mechanism, which is consistent with testosterone acting on bone through both direct androgen receptor signaling and indirect estradiol-mediated pathways affecting cortical and trabecular compartments differently.

Time-to-Event Patterns

The Kaplan-Meier curves began separating within the first year and continued diverging through the full follow-up period. This is notable because it argues against a transient early remodeling effect (the "bone remodeling transient" seen with some osteoporosis therapies). Instead, the risk appeared cumulative. Men exposed to testosterone for longer experienced proportionally more fractures relative to placebo. This temporal pattern has implications for the common clinical practice of long-term, indefinite TRT.

Interpreting the BMD Paradox

The apparent contradiction, TRT increases BMD but also increases fractures, deserves direct examination. The TTrials data showed estimated bone strength gains of 7.4% at the spine by quantitative CT after 12 months. How can a therapy that thickens bone also break it?

Several non-exclusive explanations exist. First, BMD measured by DXA or even QCT captures density but not microarchitecture or material quality. Testosterone may stimulate periosteal apposition while simultaneously increasing cortical porosity, a pattern documented in animal models of supraphysiologic androgen exposure. Second, the TTrials measured BMD over one year in 211 men. TRAVERSE measured fractures over 3.2 years in 5,204 men. The populations, timeframes, and endpoints are fundamentally different. Third, testosterone increases muscle mass and physical activity, potentially increasing fall-related exposures, though fall data were not systematically collected in TRAVERSE.

What the Subgroup Data Tell Prescribers

No subgroup in TRAVERSE showed a statistically significant fracture benefit from testosterone. The overall result held directionally across every pre-specified stratum. This does not mean every man on TRT will fracture. The absolute risk difference was modest: approximately 3.5% versus 2.5% over 3.2 years. But it does mean clinicians cannot identify a "safe" subgroup using baseline age, BMI, testosterone level, or fracture history.

The American Urological Association's 2018 testosterone guideline does not address fracture risk in its benefit-risk framework, having been published before TRAVERSE. Updated guidance will need to incorporate these findings. For now, three practical points emerge from the subgroup data:

1. Do not assume younger hypogonadal men are protected. The interaction by age was non-significant.
2. Obese men on TRT may warrant closer bone monitoring. The numerically higher risk in BMI ≥30 men, combined with altered estrogen metabolism, justifies serial DXA or at minimum FRAX reassessment.
3. Baseline testosterone severity does not predict safety. Men with the lowest levels did not fare better, eliminating a potential clinical triage strategy.

Limitations of the Subgroup Analyses

Subgroup analyses in any trial, even pre-specified ones, have well-known statistical constraints. TRAVERSE's fracture substudy recorded 155 total adjudicated fractures across both arms. Splitting this number into four or five subgroups rapidly erodes power. Most interaction p-values exceeded 0.10, meaning the trial was underpowered to detect moderate effect modification. The CONSORT guidelines for subgroup reporting caution against interpreting non-significant interactions as evidence of uniform effect, and equally against interpreting them as evidence of no difference between subgroups.

The trial also lacked serial bone turnover markers (CTX, P1NP) and systematic fall ascertainment, both of which would help clarify the mechanism behind the fracture increase. Without these data, we cannot determine whether testosterone harmed bone quality directly or increased fracture exposure through behavioral or neuromuscular channels.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Snyder PJ, Bhasin S, Cunningham GR, et al. Testosterone treatment and fractures in men with hypogonadism. N Engl J Med. 2024;390(3):203-211. PubMed
2. Lincoff AM, Bhasin S, Flevaris P, et al. Cardiovascular safety of testosterone-replacement therapy. N Engl J Med. 2023;389(2):107-117. PubMed
3. 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
4. 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
5. Mulhall JP, Trost LW, Brannigan RE, et al. Evaluation and management of testosterone deficiency: AUA guideline. J Urol. 2018;200(2):423-432. PubMed
6. FDA Drug Safety Communication: FDA cautions about using testosterone products for low testosterone due to aging. U.S. Food and Drug Administration. 2015. FDA