The Bone Paradox: When Better Density Means More Fractures

In 2017, the Testosterone Trials showed that testosterone therapy increased spinal bone density by 7.5% in older hypogonadal men — an effect comparable to bisphosphonates in postmenopausal women. For a field that had spent years studying whether testosterone helped bones, this was decisive confirmation.

In 2024, the TRAVERSE fracture substudy enrolled five thousand of those same kinds of men and found that testosterone therapy increased clinical fractures by 43%.

Both findings are methodologically sound. Both are real. And the apparent contradiction between them has generated more confusion in male endocrinology than almost any other result in the past decade.

The paradox is not a contradiction. It is a four-layer problem that, once you understand each layer, makes the fracture data not just explicable but predictable.

Layer One: Bone Is Not One Thing

A DXA scan gives you a number. That number, bone mineral density, is what most clinicians use to assess skeletal health. But bone has two structurally distinct compartments that respond to different hormones through different receptors — and a DXA scan captures them unequally.

Testosterone and estradiol — the hormone testosterone partially converts to via aromatase — have different effects on these two compartments. But the literature disagrees on which steroid protects which compartment, and this disagreement is itself informative.

Finkelstein and colleagues (JCI, 2016) ran one of the most elegant studies in bone endocrinology: 400 men received a GnRH agonist to suppress endogenous production, then graded doses of testosterone with or without an aromatase inhibitor. When aromatization was blocked — when estradiol couldn't be made from testosterone — spinal BMD fell regardless of testosterone dose. Their model: estradiol protects trabecular bone; testosterone, acting through the androgen receptor, protects cortical bone.

But Tenuta and Isidori (JCEM, 2025) present the opposite assignment in their comprehensive review, placing testosterone's pubertal role in cortical growth and periosteal expansion, while estradiol drives trabecular formation and resorption inhibition. Both are citing sound evidence. The compartment assignment depends on the context — pubertal vs. adult, direct receptor activation vs. paracrine signaling, mouse knockout vs. human interventional.

What everyone agrees on: DXA at the spine primarily reflects trabecular changes. If testosterone therapy improves trabecular bone, your DXA number goes up — but the cortical shell that gives bone its structural strength against falls may not have changed, or may have changed in the other direction.

This is the first layer: the DXA improvement is real but incomplete. It captures the trabecular response — the part testosterone genuinely helps. It may miss cortical effects that go the other way. A "better" DXA does not guarantee a stronger bone.

Layer Two: The Estradiol Threshold

Cortical bone protection depends not on testosterone directly but on estradiol — the product of aromatization. And there is a threshold below which cortical bone loses its protection.

The literature converges on a total estradiol level of approximately 90–110 pmol/L (25–30 pg/mL) as the minimum for adequate cortical bone maintenance in men. Below this, cortical resorption accelerates.

In the TRAVERSE trial, the testosterone-treated arm had estradiol levels of 31–32 pg/mL. The placebo arm: 19–20 pg/mL.

The treatment arm was right at the threshold — not safely above it. Individual variation in aromatase activity means some of those men were above 30 pg/mL and some were below. Up to 50% of middle-aged men have estradiol levels below this cortical protection threshold even with normal testosterone. The TRAVERSE population was not uniformly protected.

This matters because of a finding that reshapes how we think about aromatase and bone. Imai and colleagues (JBMR, 2025) showed in a knockout mouse model that the aromatase activity protecting bone comes from adipose tissue, not from osteoblasts. Osteoblast-specific aromatase knockout had no bone phenotype. Adipose-specific knockout produced significant trabecular loss and insufficient calcification via disrupted phosphate regulation. The fat makes the estradiol that saves the bones.

This connects to the metabolic picture. Lean men, post-weight-loss patients, and men with low visceral fat may have less adipose aromatase — and less local estradiol production — leaving their cortical bone less protected even if testosterone levels are adequate.

Layer Three: Where You Break Tells You Why You Broke

The TRAVERSE fracture substudy reported an overall hazard ratio of 1.43 (95% CI 1.04–1.97). But "fractures" is not one category. The specifics tell a different story.

Over 80% of the fractures in TRAVERSE were traumatic — from falls, not from osteoporotic collapse. The most common sites were ribs, wrists, and ankles. Classic osteoporotic fractures (hip, vertebral compression) were roughly equal between groups. And the fractures appeared early after starting treatment — too fast for bone remodeling, which takes months to manifest.

A research group at Brown University has now completed three PearlDiver database analyses on TRT and site-specific fractures. The results fracture the fracture data itself:

The same treatment, the same database, the same methodology — opposite results at different skeletal sites. Vertebral fractures went up nearly eight-fold. Hip fractures went down by nearly half.

The resolution maps to mechanism, not to bone quality:

This is supported by the broader physical activity literature. Van Gameren and colleagues (BMC Geriatrics, 2022) showed that increased physical activity initially increases fall risk — before the muscle and balance adaptations take hold and reduce it. The TRAVERSE fracture signal appeared early, consistent with more activity before protective adaptation.

Critically, TRAVERSE did not track falls, physical activity levels, or bone density. The fracture hypothesis is unfalsifiable within this dataset. Anawalt, speaking at the Androgen Society meeting in 2025, put it plainly: "These men did not have classic hypogonadism, didn't have severe hypogonadism and we don't know their BMD."

Layer Four: Which Men, Which Problem

The three major bone trials enrolled fundamentally different populations and got fundamentally different results. This is not noise — it is the population telling you which mechanism dominates.

Trial	N	Mean Age	Mean T	Duration	Key Bone Finding
TTrials Bone	211	~72	<275 ng/dL	1 year	Trabecular spine vBMD +7.5%
T4Bone	177	~60	≤403 ng/dL	2 years	Cortical vBMD tibia +3.1%, radius +2.9%
TRAVERSE Fracture	5,204	~63	~250 ng/dL	3.19 years	Clinical fracture HR 1.43 (1.04–1.97)

TTrials Bone (Snyder et al., JAMA Intern Med, 2017): the oldest, most hypogonadal men. The biggest trabecular benefit. These men had profound deficiency and their bones responded to repletion the way you'd expect — like a drought-stressed tree getting water.

T4Bone (Ng Tang Fui/Grossmann, JCEM, 2021): younger, mildly hypogonadal, obese men with prediabetes — and the benefit appeared in cortical bone, not trabecular. A different population, a different compartment benefited. 601 men had DXA, 177 had high-resolution peripheral QCT for microarchitecture. The T4Bone population had more adipose tissue — more aromatase — potentially explaining why their cortical bone responded while TTrials' trabecular bone did.

TRAVERSE (Snyder, Bauer et al., NEJM, 2024): functional hypogonadism, not classic pituitary disease. Mean testosterone around 250 ng/dL — low, but not profoundly so. High cardiovascular risk. This was not a population selected for bone disease, and the trial was not designed to study fractures. BMD was never measured. Falls were never tracked.

The NEJM correspondence (April 2024) following the publication crystallized the problem: Gruntmanis noted the population mismatch; Kovac, Weiss, and Arap questioned whether the result applied to truly hypogonadal men. Snyder and Bauer's reply acknowledged the limitations while defending the finding's statistical validity.

De Silva and Jayasena (JCEM, 2026) provided an important counterpoint from the other end of the severity spectrum: in congenital hypogonadotropic hypogonadism — the most severe form of secondary hypogonadism — treatment improves bone density but does not normalize it. Pooled lumbar Z-scores remained at −0.87, femoral neck at −0.70, despite treatment. Even where testosterone clearly helps, it isn't enough on its own.

What This Means — and What It Doesn't

The bone paradox is not an argument against testosterone therapy. It is an argument against using testosterone therapy as a bone drug.

The Endocrine Society already says this: for hypogonadal men at high fracture risk, use an antiresorptive — a bisphosphonate or denosumab — on top of testosterone, not instead of it. Tenuta and Isidori (JCEM, 2025) make the same recommendation explicitly, identifying the combination of TRT plus antiresorptive therapy as a research priority.

The evidence now explains why this layered approach is necessary:

What remains unknown is substantial. No trial has measured estradiol thresholds specifically in the fracture context. No study has tracked falls and fractures together in men starting TRT. The Arcand proximal humerus study from this PearlDiver series is now published (JSES RRT, 2026) but the full results are not yet freely available — the third fracture site in the divergence pattern.

The ADT Mirror

The opposite experiment — removing testosterone entirely with androgen deprivation therapy for prostate cancer — provides a brutal confirmation. ADT produces 2–8% BMD loss in the first year alone, osteoporosis in 9–53% of patients, and a 1.5–1.8× increase in fracture risk. If removing testosterone destroys bone that rapidly, replacing it should help. And it does — just not in the simple way DXA suggests.

The connection to the broader evidence base matters. My previous article on prostate cancer safety covered the metabolic harm of ADT. The measurement problem article documented the ±20% imprecision in testosterone assays — which means the estradiol measurements that determine bone protection are even less reliable (estradiol at low male levels is notoriously difficult to measure accurately). And the metabolic trap article documented the adipose-aromatase connection that Imai's 2025 study now confirms is the actual bone-protection pathway.

For Clinicians Reading This

If you are prescribing TRT to a hypogonadal man with osteopenia or osteoporosis:

The bone paradox is the clearest example of a theme running through every article in this series: testosterone therapy works, but it works within biological constraints that most clinical protocols ignore. The DXA number is not the bone. The compartment you measure is not the compartment that breaks. And the population you study determines which answer you get.

Bone, like everything in the HPG axis, is a system — not a number.