Bone density

Most age-related fractures aren't really about old bones — they're about bones that never got loaded heavily enough to stay strong. Walking barely moves the spine and swimming leaves the skeleton untouched, while heavy resistance training with impact loading moves both hip and spine — and it now has the endpoint that matters, cutting major fractures by about a quarter.

Bone has been quietly reclassified over the last decade. It is no longer treated as inert structural scaffolding; it's a highly vascularized endocrine organ that secretes osteocalcin into the brain, traffics calcium between the skeleton and the arterial wall, and tracks all-cause mortality independent of fractures. The dominant intervention to defend it is mechanical — heavy, progressive resistance training with impact loading. Nutrition (protein, calcium, vitamin K2), chronobiology, and gut/endocrine status are the supporting cast.

Why bone is on the longevity short list

Evidence rating: Moderate — the associations below are large and replicated across cohorts, but observational, so residual confounding cannot be fully excluded. Throughout this article, a hazard or risk ratio above 1.0 means higher risk and below 1.0 means lower; the bracketed 95% confidence interval (CI) is the range where the true value most plausibly lies, and when it crosses 1.0 the result is compatible with no effect.

Three independent signals converge:

All-cause mortality. Low bone mineral density (BMD) is associated with higher mortality across cohort after cohort, and the relationship persists after adjusting for age, sex, body mass, smoking, and metabolic comorbidity.^[1] The gradient runs the way you would expect, and one widely cited paper is internally inconsistent about it. Its Table 2 reports osteoporosis at about 40% higher all-cause mortality than normal bone mass (HR 1.40, 95% CI 1.13–1.75) with osteopenia indistinguishable from no effect (HR 1.05, 0.89–1.24); the running text of the same paper prints those two figures the other way round. The table, the discussion, and the age-stratified analysis (osteoporosis HR 2.80, 2.29–3.42 in adults over 50) all agree: it is osteoporosis that carries the mortality signal.^[2] What that excess mortality is made of has not been established. The same cohort tested each cause of death separately and found none of them significant once confounders were accounted for — cardiovascular death, cancer death, heart disease and stroke were all compatible with no association. Fracture-related mortality was never analysed at all, in either direction. The honest reading is that low bone mass tracks something broader about cardiometabolic health, not that it kills through a known route.
Osteosarcopenia is the lethal combination. When bone loss and muscle loss occur together, the pooled signal in 14,429 prospectively followed adults is about 53% higher all-cause mortality (RR 1.53, 1.28–1.78) vs. healthy controls.^[3] The one study that has compared the combination against each condition alone did so in 572 hospitalised older patients, where three-year mortality was 46% with osteosarcopenia against 9% in those with neither condition; the authors also report the combination adding 30% over sarcopenia alone and 8% over low bone mass alone, without making clear whether those are percentage points or relative increases.^[4] That is an acute-care population in their mid-seventies, so read it as evidence that the two conditions compound rather than as a number a healthy midlife reader can apply. Mechanical and endocrine cycles couple the two: atrophied muscle stops loading bone and secretes inflammatory factors that suppress osteoblasts. Treating one without the other is incomplete.
The bone-vascular axis is bidirectional. As bone demineralizes, calcium does not simply excrete — it preferentially deposits in arterial walls, accelerating atherosclerosis and vascular stiffness.^[5] Pro-inflammatory cytokines from active bone loss — tumour necrosis factor-α (TNF-α) and interleukin-6 (IL-6) — damage vascular endothelium directly, and calcified arteries reciprocate by secreting sclerostin, Dkk-1, and SFRP — all of which inhibit osteoblast bone formation, locking the system into a feedback loop. The two systems are coupled tightly enough that it is worth treating them as one problem — though the review that assembles this mechanism is explicit that "the causal relationship between these two systems… remains unclear."

A fourth, more recently appreciated link: bone is endocrine to the brain. In mice, osteocalcin secreted by osteoblasts crosses the blood-brain barrier and binds GPR158 in the cortex and hippocampus, where it drives brain-derived neurotrophic factor (BDNF) and appears necessary for normal spatial memory. Treat this as an interesting hypothesis rather than a reason to train: two independent laboratories in 2020 built osteocalcin-free mice and found none of the predicted metabolic or behavioural abnormalities, and the review that assembles the pathway concedes the mechanism "remains incompletely understood."^[6] Circulating osteocalcin declines in parallel with age-related cognitive decline; supplementing it (or stimulating its release through mechanical bone loading) reverses age-related deficits in animal models. This places resistance training squarely on the cognitive longevity short list — see Dementia prevention.

Beyond DXA: the Trabecular Bone Score

DXA (dual-energy X-ray absorptiometry) — the standard areal-BMD measurement — quantifies how much mineral is present in a 2D projection. It says nothing about the 3D architecture or material quality of the trabecular matrix. That gap is now filled by the Trabecular Bone Score (TBS), a non-invasive index extracted from the same lumbar DXA image that captures trabecular connectivity and spacing.

TBS category	Score	Interpretation
Normal microarchitecture	> 1.31	Robust, well-connected trabecular network
Partially degraded	1.23 – 1.31	Onset of structural decay; a reason to take loading and nutrition seriously, not a treatment threshold on its own
Degraded	< 1.23	High fragility-fracture risk and elevated mortality

The mortality signal from TBS is independent of BMD. Evidence rating: Moderate (observational, but adjusted for BMD). In a cohort of 2,641 US adults aged 60 and over followed for a median of nearly twelve years, those in the lowest tier of that analysis had about 47% higher all-cause mortality (HR 1.47, 1.10–1.96) and roughly double the cancer-specific mortality (HR 2.07, 1.17–3.67), with the association robust to BMD adjustment.^[7] TBS also tends to be lower in people carrying more central fat: in a NHANES sample of nearly 4,000 women averaging fifty, waist circumference was the strongest metabolic correlate of a degraded score, and the association held even in women with a body-mass index under 25, where soft-tissue artefact is negligible.^[8]

Practical takeaway, with a boundary: if you are having a lumbar DXA anyway, ask for the TBS read-out — the International Society for Clinical Densitometry considers it appropriate from age 40. What it will not do is stage you on its own. The 2023 positions state that TBS "should not be used alone to determine treatment recommendations", that it is most likely to change management in adults over 40 who are already close to the drug-treatment threshold on the Fracture Risk Assessment Tool (FRAX), and that tracking TBS change over time "is unlikely to be helpful with the current version of the TBS algorithm".^[9] A degraded score is a reason to take the loading and nutrition prescription seriously; it is not, on its own, a trigger for treatment.

Heavy lifting beats walking by an order of magnitude

Bone is a piezoelectric, mechanosensitive tissue. Under high mechanical strain, fluid shear inside the osteocyte lacunar-canalicular network triggers electrical and biochemical signals that recruit osteoblasts (Wolff's Law). The body is metabolically lazy: it builds and maintains only the bone that habitual loading actually demands.

Harold Frost's mechanostat formalised this: bone only adds mass when habitual strain exceeds a minimum effective strain threshold, and it removes mass when strain falls below a lower disuse threshold — so an activity has to be unusual in magnitude, not merely repeated often, to register.^[10]

This is why the well-meaning "walk for your bones" guidance has aged poorly. Evidence rating: Moderate that low-impact aerobic exercise does not build lumbar-spine bone. The hip is not settled — the same review found a small but significant gain at the femoral neck. A meta-analysis of eight walking trials in postmenopausal women found no meaningful effect on lumbar-spine BMD — a difference of just 0.007 g/cm² that could easily be chance (weighted mean difference 0.007 g/cm², p=0.09) — while at the femoral neck it found a small positive effect that just reached significance (0.014 g/cm², 95% CI 0.000–0.028); its stated conclusion is that walking preserves bone at the femoral neck while doing nothing for the spine, and total hip could not be pooled at all. The trials were small and of poor methodological quality, and the authors still recommend that other forms of exercise providing greater targeted skeletal loading be used.^[11] Swimming is the clearest case of a sport that spares the skeleton: the review of 64 studies found competitive swimmers with bone density similar to sedentary controls and lower than athletes in high-impact sports, and concluded that swimming "does not seem to negatively affect bone mass" — the water simply removes the ground-reaction force.^[12] Water-based exercise classes for older adults are a different question and do show small gains at the spine and hip.^[13] The American College of Sports Medicine (ACSM) position stand on physical activity and bone health draws the same line, recommending weight-bearing, high-impact and resistance activity rather than low-impact aerobic work as the osteogenic prescription for adults.^[14] These modalities are excellent for the cardiovascular system, and they are not the osteogenic stimulus — but "they do nothing for bone" is too strong. A 2026 network meta-analysis of 74 trials ranks mind-body exercise combined with resistance as the most effective modality at both the lumbar spine and the femoral neck, and reports that aerobic and mind-body work each show site-specific effects of their own.^[15] Whole-body vibration is the borderline case: 2025–2026 meta-analyses find a statistically significant but low-certainty effect on total-femur BMD only — not femoral neck or lumbar spine — confirming it is at best a modest adjunct for those who cannot perform impact loading, not a substitute.^[16] The intervention that does move spine and hip BMD is now well-defined.

The LIFTMOR protocol

The Australian LIFTMOR (Lifting Intervention For Training Muscle and Osteoporosis Rehabilitation) trial randomized postmenopausal women with diagnosed osteopenia or osteoporosis to high-intensity resistance and impact training (HiRIT) vs. a low-intensity home program, twice weekly for 8 months under supervision.^[17] Evidence rating: Strong — a supervised RCT in the exact population (diagnosed low bone mass), replicated. The protocol was deliberately aggressive against entrenched clinical fears that heavy loading would shatter fragile bone:

Compound barbell lifts: deadlift, overhead press, back squat, 5 sets × 5 reps above 85% of one-rep max (1RM).
Impact loading: jumping chin-ups with heavy drop landings.
Twice weekly, ~30 minutes per session, after a graded transition period.

Over 8 months the HiRIT group gained ~2.9% BMD at the lumbar spine and ~0.3% at the femoral neck with significant improvements in functional performance, vs. losses in the low-intensity arm. The follow-on trial from the same group, MEDEX-OP, reproduced the direction at about two-thirds the size: lumbar-spine BMD rose 1.9% on the same heavy protocol against 0.1% on a low-intensity comparator, and total hip — a co-primary outcome — did not favour the heavy arm.^[18] LIFTMOR reported a single adverse event, a minor back spasm, in participants screened for conditions and medications that affect bone;^[19] MEDEX-OP reported seven adverse events across its four groups and described both programmes as well tolerated. Neither trial reported a fragility fracture. Heavy progressive loading is both the most effective osteogenic stimulus available and safe when supervised, even in already-fragile bone.

Density is a surrogate; the endpoint that matters is a broken bone. Pooling eleven controlled exercise trials covering nearly 20,000 participant-years, exercise cut low-trauma major osteoporotic fractures by about a quarter (RR 0.75, 95% CI 0.54–0.94), with almost no variation between studies.^[20] Two details are worth carrying: the analysis found no significant advantage for programmes that progressed intensity, and none for programmes lasting longer than a year — so the fracture benefit does not obviously track the "heavy or nothing" framing that the density evidence supports.

How to dose osteogenic loading — the mechanostat rules

LIFTMOR tells you what to do; the mechanostat literature tells you how to dose it. Four rules follow from it, and they explain why ordinary activity fails. Evidence rating: Moderate — the dosing rules are well-replicated in animal loading models and the tennis-arm natural experiment, but the precise human force thresholds are extrapolations.

Force magnitude and rate, not volume, set the stimulus. Jump landings that reach roughly four to five times body weight, applied fast, are the loads that exceed the thresholds this literature works with — measured directly in premenopausal women, hopping and jumping produced 3.90–5.38 body weights at rates of 192–329 body weights per second.^[21] All three common jump types raise the bone-formation marker P1NP acutely by 8–11%, with no significant difference between them.^[22] Maximal countermovement and squat jumps may not exceed walking-level joint loading at the hip and ankle — the hard-force version of the "walking is not enough" thesis.
Mechanosensitivity saturates within a few cycles. Osteocytes desensitize rapidly, so piling on repetitions in one bout yields steeply diminishing returns. Dividing a daily dose into several short bouts markedly enhanced the osteogenic response in the foundational rodent loading work,^[23] and the follow-up that varied the interval directly found about 8 hours of recovery between them sufficient to restore full mechanosensitivity.^[24] The practical rule is "little and often": short osteogenic bouts spread across the day complement (don't replace) the twice-weekly heavy session.
Loading is site-specific. Bone responds only where it is loaded. The cleanest demonstration is the tennis "natural experiment": in 105 national-level female tennis and squash players, the playing arm carried between about 9% and 16% more bone mineral content than the non-playing arm depending on the site measured (vs 3–5% side-to-side difference in non-playing controls), rising to as much as 23% at the humerus in those who started before menarche.^[25] Bone builds where it is loaded and nowhere else, so choose movements that load the hip and spine specifically.
It reverses — bone training is maintenance, not a one-time deposit. In the ACTLIFE RCT of early-postmenopausal osteopenic women, the lumbar-spine BMD advantage over controls was lost after just 3 months of detraining, even though strength and power gains persisted longer.^[26] Programs must be continuous, not seasonal.

What the 2026 ACSM update endorses

The American College of Sports Medicine's 2026 Resistance Training Position Stand — synthesizing 137 systematic reviews and more than 30,000 participants — reinforces the general frame:^[27]

The largest gain comes from moving from no resistance training to any; train all major muscle groups at least twice weekly.
For strength, lift heavier — around 80% of one-rep max (1RM) — with progressive overload.

For bone specifically the picture is narrower than the strength literature. In the pooled resistance-training trials, the heavy subgroup — defined there as at or above 70% of one-rep max — was significantly better at the total hip and femoral neck but not at the lumbar spine, and the best-supported frequency was three sessions a week.^[28] A 2026 synthesis that compared heavy against light-to-moderate loading head-to-head in adults over 50 found no significant bone difference at either site — lumbar spine and femoral neck effects both compatible with chance — even though strength clearly favoured the heavy arm.^[29] Heavy loading is the best-evidenced route to the hip; the claim that lighter loading does nothing for bone is not supported by the trials that tested it directly.

Mechanical silence is independently toxic

Evidence rating: Weak. A narrative review from the International Osteoporosis Foundation's rehabilitation working group argues that prolonged sedentary time harms bone independently of how much you train, while describing that evidence as emerging.^[30] The independence is genuinely contested: in a Canadian cohort of more than 8,000 adults followed for a decade, with physical activity adjusted for, neither baseline nor sustained sedentary time predicted bone-density change or fragility fracture.^[31] What is not contested is the trade-off: replacing sitting time with walking tracks with lower osteoporosis risk, which is a reason to move rather than a reason to fear the chair. The implication parallels the active-couch-potato finding for cardiovascular risk — see Sitting. One LIFTMOR session is not a license for nine sedentary hours. Habitual movement and frequent posture change matter for bone as well as heart.

Protein, calcium, and the vitamin K2 traffic problem

Protein is anabolic to bone, not catabolic

Evidence rating: Moderate for protein; Moderate for the calcium-form and K2 detail below. The decades-old worry that high-protein diets "leach calcium from the bones" via metabolic acidosis has never been substantiated — but the umbrella review that examined it did not settle the question either way, citing both a cohort in which higher animal-versus-plant protein tracked faster femoral-neck loss and a trial in which shifting part of the protein from animal to plant sources raised bone turnover. That same review, behind the German nutrition guideline, rated the evidence insufficient across almost every bone outcome, with two exceptions: a possible reduction in hip-fracture risk of about 11% at higher versus lower intake, and possible evidence of no effect of extra protein on total bone density.^[32] What is not in doubt is that protein is a structural requirement — the matrix is built from it. High-quality protein activates the growth regulator mTORC1 (via leucine), upregulates insulin-like growth factor 1 (IGF-1, which stimulates renal calcitriol production and intestinal calcium absorption), and provides the amino acids the bone matrix is built from.

Consensus bodies set the floor lower than the bone literature's working targets: ESPEN and PROT-AGE both recommend at least 1.0–1.2 g/kg/day for healthy older people, rising to 1.2–1.5 with illness.^[33] ^[34] This site's working bands are 1.2–1.6 g/kg/day for an active midlife adult, and 1.6–2.0 g/kg/day for an older adult who trains hard, is in a caloric deficit, or has a chronic illness (including glucagon-like peptide-1 (GLP-1) receptor-agonist users — see GLP-1). The historical EFSA / FAO PRI of 0.83 g/kg is now considered inadequate for skeletal preservation in adulthood. See Protein for the full evidence base.

Calcium intake is falling and anti-nutrient intake is rising

A serial NHANES analysis running from 1999 to 2023 documented a population-level shift, though its four headline numbers span different windows. Phytate intake rose from 594 to 834 mg/day and oxalate from 242 to 281 mg/day between 1999–2000 and 2017–2020; mean dietary calcium peaked at 1,025 mg/day in 2009–2010 and fell to 900 mg/day by 2021–2023; and mean femoral-neck bone density fell from 0.849 to 0.775 g/cm² between 2009–2010 and 2017–2020.^[35] These are separate cross-sectional samples rather than the same people followed over time, and the authors conclude only that such diets "may contribute". The combination of less calcium, more anti-nutrient binding, and likely reductions in habitual mechanical loading is showing up in the population-level skeleton.

Anti-nutrients matter for plant-source calcium specifically:

Source	Calcium content	Anti-nutrients	Net bioavailability
Spinach	High	Very high oxalate	Poor — most calcium is bound and excreted
Kale, bok choy	Moderate	Low oxalate	High — counts as bioavailable plant calcium
Almonds, sesame	Moderate	Phytate	Moderate; soaking/sprouting helps
Whole grain millet, teff	High	Very high phytate	Poor without dephytinization (soaking, fermenting)
Dairy (yogurt, kefir, cheese)	High	None	High; the population reference for bioavailability

If supplementing to reach 1,000–1,200 mg/day total, the chemical form matters:^[36]

Calcium carbonate — 40% elemental calcium and the cheapest form, but it needs stomach acid to dissolve, so take it with food. In people with no gastric acid, a fasting dose of carbonate is absorbed about five times less well than in people with normal acid — yet the same study found absorption completely normal when the carbonate was given with breakfast.^[37] Drug-induced acid suppression does the same thing: 20 mg of omeprazole daily for a week significantly cut fasting absorption of calcium carbonate in women over 65.^[38] So if you take a proton-pump inhibitor (PPI), either use citrate or take carbonate with a meal. Frequently causes constipation and bloating.
Calcium citrate — 21% elemental, absorbed independent of stomach acid, can be taken on an empty stomach. The right choice for older adults, anyone on acid-reducing medication, or anyone with reflux.
Calcium hydroxyapatite — slower, more stable absorption; mimics bone matrix. Reasonable but usually unnecessary.

Vitamin K2 directs the traffic

Vitamin K2 is the cofactor that carboxylates osteocalcin, which binds calcium into the skeleton, and Matrix Gla Protein, which holds it out of soft tissue and vessel walls. Supplementation does move those markers: pooled across nine trials it raised osteocalcin and bone-specific alkaline phosphatase and lowered the uncarboxylated fraction, though the authors describe the resorption-marker change as of "uncertain clinical relevance" and note that no bone-density or fracture outcome has yet been demonstrated.^[39] The idea that undirected calcium ends up in arteries rather than bone is mechanistically coherent and clinically unproven. The two clinically relevant subtypes:

MK-4 — short half-life, fast tissue uptake, and the only form with fracture-endpoint trials behind it, at prescription doses of about 45 mg/day in Japan. Those trials do not bear weight. The meta-analysis that reported large fracture reductions pooled seven Japanese menaquinone trials, drawn from a literature in which fourteen menatetrenone papers — twelve of them by one author — have since been retracted for fabricated data; its own senior author later published a caution that the review had included problematic trials.^[40] ^[41] The one large clean trial — 4,378 women, 45 mg/day, three years — missed its primary endpoint in both strata and reported more adverse events on the drug,^[42] and modern pooled estimates are non-significant.^[43] Requires multiple daily doses to maintain status.
MK-7 — long half-life (~3 days), once-daily dosing supports continuous carboxylation status. Strong mechanistic and biomarker support; lacks the fracture-endpoint trials of high-dose MK-4.

For most healthy adults the practical choice is MK-7 at 90–180 µg/day, paired with vitamin D3 and adequate magnesium. Don't take K2 supplementally if you're on warfarin without coordinating with the prescriber. See Vitamin K2 for the full evidence picture and Vitamin D.

Where the nutrition evidence has shifted

Two corrections and two positive food trials are worth folding into the picture:

Vitamin D in already-replete adults does not cut fractures. The VITAL fracture ancillary randomized 25,871 generally replete adults to 2,000 IU/day vitamin D3 and found no reduction in total, non-vertebral, or hip fractures, including in those with low BMD or taking calcium.^[44] The nuance: correcting deficiency matters; supplementing the replete does not. See Vitamin D and Calcium for the dosing detail. Evidence rating: Strong.
Excess preformed vitamin A (retinol) raises hip-fracture risk. A meta-analysis of 283,930 participants found high retinol intake raised hip-fracture risk by about 40% (RR 1.40, 95% CI 1.03–1.91) and high total vitamin A by about 29% (RR 1.29, 1.07–1.57); beta-carotene did not.^[45] A "more is not better" caution against high-dose retinol supplements, large amounts of liver, and cod-liver-oil megadosing. Evidence rating: Moderate.
Prunes (50 g/day) preserve hip BMD. In a 12-month RCT of 235 postmenopausal women, the 50 g/day dose (~4–6 prunes) held total-hip BMD steady while controls lost ~1%; the 100 g/day arm showed no significant hip benefit and had high dropout — so the actionable dose is 50 g, not 100 g.^[46] Evidence rating: Moderate.
Collagen peptides (5 g/day) raised BMD in one RCT. 5 g/day specific collagen peptides for 12 months increased lumbar-spine and femoral-neck BMD and raised P1NP in 131 postmenopausal women with low BMD; a 2025 meta-analysis found benefit greatest when combined with calcium and vitamin D.^[47] Several such trials are industry-linked. Evidence rating: Weak / preliminary.

The older "alkaline diet / acid-ash leaches calcium from bone" narrative does not hold up: alkaline potassium salts reduce urinary calcium, but systematic reviews find insufficient evidence that dietary acid load causes BMD loss or fractures — mechanistically plausible, not established. This parallels the protein-acidosis myth above.

Cortisol, insulin resistance, and the gut

Evidence rating: Moderate across this section — the cortisol and insulin-resistance associations are consistent but observational; the Basigin mechanism and much of the gut-bone axis are still preclinical.

Chronic stress damages bone through a specific protein

Chronic glucocorticoid elevation — prescription steroids, but also persistent endogenous cortisol from psychological stress — degrades bone profoundly. Mild Autonomous Cortisol Secretion (MACS) is associated with worse bone quality even where the density reading looks normal: 75 patients compared against matched referents showed no difference in bone density at any site but lower TBS (1.389 vs 1.475), lower osteocalcin and lower tibial trabecular volume.^[48] The mechanism was clarified in 2025: in mice, glucocorticoids drive skeletal stem cells to release a protein called Basigin, which disrupts their normal lineage behaviour and distorts the bone's vascular network — vessel numbers rise while their structure becomes malformed.^[49] An antibody against Basigin given alongside the steroid prevented the bone loss; given to two-year-old mice on its own, it restarted remodelling and improved bone mass. Both results are in mice, with human cells tested only in culture — a plausible therapeutic class, not an available one.

A Mendelian-randomization analysis suggests the two problems are partly the same problem: genetically higher cortisol raised osteoporosis risk at the femoral neck, and sarcopenia-related traits accounted for a slice of the estimate at the lumbar spine.^[50] The study rests on three genetic variants in one ancestry and reports opposite directions at the two sites, so treat it as a hint that muscle is on the causal path rather than as a measured quantity — and note that a genetic mediation estimate says nothing directly about what training does. See Stress.

Insulin resistance uncouples bone density from bone strength

Insulin resistance is not straightforwardly bad for the density reading — the bone is loaded by extra body weight and exposed to anabolic insulin — but the effect is smaller than it looks: in a UK Biobank analysis the crude positive association between a surrogate insulin-resistance index and bone density lost significance once age, sex and body mass were accounted for.^[51] What is clearer is that bone quality can be worse even when the density reading is not. Measured directly in the tibia, women with type 2 diabetes had bone material strength about 10% below matched controls with no difference in bone density or microarchitecture — the mechanical deficit was invisible to a density scan, and the leading explanation is advanced glycation end-products (AGEs) cross-linking the collagen matrix.^[52] A "normal" DXA in a metabolically unhealthy adult should not be reassuring. Glycemic control and insulin sensitivity belong on the bone-health checklist, which makes the dietary pattern, exercise, and sleep prescriptions for Metabolic flexibility directly relevant.

The gut-bone axis

A diverse, fiber-fermenting microbiome produces short-chain fatty acids (butyrate, propionate, acetate) that suppress osteoclast differentiation and enhance paracellular calcium absorption. Conversely, dysbiosis raises systemic TNF-α and IL-6 and shifts the balance toward bone resorption.^[53] Probiotics are the part of this story that has not held up. In women aged 75 to 80, a year of Lactobacillus reuteri did not build bone — both groups lost it, and the probiotic only slowed the loss, by about 1% of tibial volumetric density with a confidence interval whose lower bound sat almost exactly at zero.^[54] A larger, longer trial of the same strain from the same group — 239 early-postmenopausal women over two years, two doses — then found no effect on that endpoint or on any other predefined bone outcome.^[55] Support for Lactobacillus rhamnosus GG is preclinical. The general dietary lever is the same one that protects everything else: high plant diversity, polyphenols, fermented foods, adequate fiber. See Fermented foods.

Chronobiology: the nocturnal resorption peak

Evidence rating: Moderate — the circadian resorption rhythm is well-demonstrated; the specific timing-of-dosing figures below come from small trials and a chronotherapy review, so treat them as directional.

Bone resorption is not constant. Continuous serum sampling shows that CTX (the C-terminal telopeptide of type I collagen, a resorption marker) rises sharply at night during the fasting/sleeping phase, while bone formation markers stay relatively flat across the 24 hours. A 2025 study demonstrated that this rhythm is intrinsically circadian — it persists under constant conditions rather than being driven purely by meal timing or the sleep-wake cycle.^[56] The circadian gene BMAL1 keeps bone turnover coordinated — in mice, deleting it from bone-forming cells raises the RANKL signal that recruits osteoclasts; circadian disruption (chronic shift work, sustained sleep deprivation, evening blue-light exposure) lowers BMAL1 and accelerates net bone loss.

This has practical chronotherapy implications:^[57]

Intervention	Optimal timing	Mechanism
Calcium supplementation	Evening / pre-sleep	1,000 mg taken at 23:00 for two weeks cut total daily bone-resorption markers by about 20%; the same dose at 08:00 did nothing measurable^[58]
Mechanical loading (training)	Any time you will actually do it	In rodents the anabolic response is larger in the early active phase, but the human meta-analysis found no difference between morning and evening training for strength, size or bone density; morning daylight is worth having for circadian phase regardless
Time-restricted eating	Daytime 8–12h window	Post-meal release of the gut hormone GLP-2 (glucagon-like peptide-2) suppresses daytime resorption; eating-clock alignment stabilizes total bone mineral content
Teriparatide (rare clinical)	Morning	12-month trial: 9.1% lumbar BMD gain morning vs. 4.8% evening
Salmon calcitonin (rare clinical)	Pre-dinner	75% suppression of resorption when taken pre-dinner or in the evening, against 40–50% in the morning — the driver is dosing before a meal, not the hour

For a healthy adult the actionable items are simple: train at whatever hour you will keep doing it, take calcium with the evening meal rather than at breakfast, eat in a daytime window, and protect sleep — see Sleep and Circadian rhythms. Shift workers carry an independent bone-loss risk that is hard to fully neutralize and worth flagging.

Sex steroids and the timing of bone loss

The perimenopausal rapid-loss window

The fastest bone loss of a woman's life is not gradual age-related decline — it is a roughly three-year window running from one year before the final period to two years after it, during which women lose about 7% of lumbar-spine bone mass, most of the 10.6% lost across the whole decade.^[59] This makes the timing of any intervention decisive. Hormone therapy prevents fractures. A 2001 meta-analysis of 22 small trials found about a quarter fewer non-vertebral fractures overall (RR 0.73), with a larger effect in trials whose participants averaged under 60 (RR 0.67).^[60] But the largest randomised dataset since then found the benefit the same at every age: 8.6% of women on oestrogen-plus-progestin fractured over 5.6 years against 11.1% on placebo, about 2.5 fewer fractures per hundred women, and "the effect did not differ in women stratified by age."^[61] The timing hypothesis is well supported for cardiovascular outcomes; for fractures it rests on subgroups of underpowered trials. Protection also lasts only as long as treatment does — fracture risk climbs back toward never-user levels over the decade after stopping.^[62] The 2022 position statement of The North American Menopause Society (renamed The Menopause Society in 2024) affirms hormone therapy for prevention of bone loss and fracture in appropriate candidates. Evidence rating: Strong for the loss window; HT decisions are individual and belong with a clinician.

Male osteoporosis — and why estradiol, not testosterone, runs the show

Men are badly under-served by bone care: excess mortality one year after a hip fracture at age 80 is about 8% in women and 18% in men — roughly double.^[63] The counterintuitive mechanism: estradiol — not testosterone — is the dominant sex steroid regulating bone resorption in men. Finkelstein's selective-suppression experiments — healthy men put into temporary gonadal shutdown, then given graded testosterone with or without a drug blocking its conversion to oestrogen — found bone resorption rose as testosterone fell, and rose considerably more when oestrogen production was blocked too. The authors' conclusion is that oestrogens are the primary regulator of bone in adult men, with testosterone contributing.^[64] In the Testosterone Trials bone substudy of older men, testosterone raised spine trabecular volumetric BMD by ~7.5% over a year, with the gain correlating more tightly with the rise in estradiol than testosterone — but the trial was underpowered for fractures,^[65] and the much larger TRAVERSE fracture subtrial found the opposite of a benefit: clinical fractures occurred in 3.5% of men on testosterone against 2.5% on placebo over about three years — roughly one extra fracture per hundred men treated.^[66] Those men had hypogonadism plus established or high cardiovascular risk, so the figure does not transfer directly to a healthy midlife man — but it is the reason a density gain is not an anti-fracture claim. Treat this as a BMD/strength signal, not validated anti-fracture therapy. The Endocrine Society advises screening all men ≥70, and men 50–69 with risk factors. Evidence rating: Moderate.

A practical corollary: several common chronic medications drive secondary bone loss and warrant DXA/FRAX awareness rather than reflexive discontinuation — glucocorticoids (the most potent), proton-pump inhibitors (about 30% higher hip-fracture risk in meta-analysis, RR 1.30, 1.19–1.43, and similar whether use was short or long),^[67] selective serotonin reuptake inhibitors (SSRIs), aromatase inhibitors, and androgen-deprivation therapy. Discuss lowest effective dose and deprescribing of unnecessary chronic PPIs with the prescriber.

A note on weight loss and GLP-1

Rapid weight loss — including with GLP-1 receptor agonists — lowers bone density, especially at the hip. The one randomised trial that separated the components is encouraging about the fix and unclear about the cause: over a year after a very-low-calorie diet, exercise combined with liraglutide kept hip and spine density unchanged versus placebo, while liraglutide alone lost more bone than exercise alone at both sites.^[68] Whether that reflects lost mechanical loading or something the drug does directly is not established — the arms did not lose identical amounts of weight. The mitigation is the same prescription this page already makes: resistance and impact training, ≥1.2–1.6 g/kg protein, and adequate calcium and vitamin D alongside any intentional weight loss. Evidence rating: Moderate.

What's coming: the regenerative pipeline

Three preclinical / early-clinical programs may shift the therapeutic landscape over the next 5–10 years, but none is currently a substitute for mechanical loading and nutrition in a healthy adult:

GPR133 / AP503 — A mechanosensitive adhesion G-protein-coupled receptor (ADGRD1) that responds to the physical force of loading, stimulates osteoblasts and inhibits osteoclasts. A screened small-molecule agonist, AP503, raised bone mass and strength in healthy mice and in a mouse model of postmenopausal bone loss.^[69] A potential anabolic class distinct from bisphosphonates — and, notably, a molecular receptor for exactly the mechanical loading this page prescribes.
CCN3 — A hormone released by neurons in the hypothalamus that maintains skeletal density during the calcium drain of breastfeeding. UCSF-led work with UC Davis collaborators showed it activates skeletal stem cells directly and accelerates fracture repair in young and old mice of both sexes, with hydrogel delivery localising the effect.^[70] Every result in a living animal is in mice; human cells have been tested only in culture, and two of the investigators have since co-founded a company to commercialise it.
ARPA-H NITRO program — A federally funded fast-track for regenerating cartilage and the bone beneath it. The name expands to Novel Innovations for Tissue Regeneration in Osteoarthritis: it is a joint-repair programme, not a bone-density one, and its relevance here is the injectable bone regeneratives it is developing along the way. Results so far are in osteoarthritic animal models; the IND-enabling preclinical work is underway and first-in-human trials are slated for late 2027.^[71]

The point is to understand that the field is moving from antiresorptive maintenance (bisphosphonates, denosumab) toward genuine structural rebuilding. The point is not to defer mechanical loading and nutrition while waiting — those interventions remain the floor against which any future drug will be added.

A practical bone-density protocol for healthy midlife adults

Know when a DXA is actually indicated. The 2025 US Preventive Services Task Force update looked specifically at adults aged 40 and over and still recommends screening only for women 65 and older, and for postmenopausal women under 65 whose fracture risk is elevated on a formal risk tool such as the Fracture Risk Assessment Tool (FRAX); for men at any age it concluded the evidence is insufficient to recommend either way, and it establishes no rescreening interval.^[72] Ask for an earlier scan if you have a specific reason — glucocorticoid exposure, an eating-disorder history, premature menopause, a parental hip fracture, chronic kidney disease, or a low-trauma fracture already behind you. A baseline in your late 40s is not wrong, but it is a personal choice made outside the guidelines, not a standard of care.
Lift heavy, twice a week. Compound barbell movements (squat, hinge, vertical/horizontal press, vertical/horizontal pull) at ≥70–85% 1RM, 5–8 reps, supervised initially. LIFTMOR itself used five sets of five above 85% of one-rep max; the pooled trial evidence supports anything at or above 70%. The protocol is safe even in low-bone-mass populations when coached, and is the single most osteogenic stimulus available. See Resistance training.
Add impact loading — and do it most days, not twice a week. The one trial that dose-ranged it randomised women to 0, 2, 4 or 7 days a week of 50 unilateral multidirectional hops. Two days a week — 100 landings — produced no femoral-neck change at all; only the daily arm did, at about 1.8%.^[73] A twelve-month programme of the same daily hopping raised femoral-neck density in older men,^[74] and a six-month unilateral hopping programme did the same in postmenopausal women, with no adverse signal on knee-cartilage imaging.^[75] In each case the comparison was against the participant's own untrained leg. Multidirectional hopping is the cleanest hip evidence there is; jump squats and low-box jumps are reasonable substitutes but have not been tested at the hip the same way.
Don't sit through the rest of the day. Aim for something like 7,000 steps and take hourly posture breaks — but treat that as a target rather than a floor, since everything below it counts too. The step evidence is about mortality, not bone; what it buys the skeleton is the habitual low-level loading a single gym session does not supply — see Sitting.
- Train balance to prevent the fall. Nearly every fragility fracture requires a fall, so fall prevention is half the equation. Tai Chi cuts the number of people who fall by about a quarter (RR 0.76) and is a low-impact complement for older or deconditioned readers, even though it does little for BMD itself.^[76] See Mobility and balance.
Protein 1.2–1.6 g/kg/day (1.6–2.0 if you are older and training hard, in a caloric deficit, or managing a chronic illness), spread across three or four meals at ~30–40 g each. See Protein.
Calcium 1,000–1,200 mg/day, with the evening meal. Prefer dairy / kale / bok choy over spinach for plant sources. Calcium citrate if you take a PPI or have low gastric acid — and supplement only the gap, never as a large isolated bolus. See Calcium.
Vitamin D3 to a 25(OH)D of ~30–50 ng/mL (75–125 nmol/L) plus vitamin K2 (MK-7) ~90–180 µg/day to direct the calcium into bone rather than artery. See Vitamin D.
Adequate magnesium (200–400 mg/day of elemental magnesium from food plus supplement if needed) — a cofactor for activating vitamin D. Higher magnesium intake tracks with higher bone density but has never been shown to reduce fractures, so this is a plausibility play. See Magnesium.
Treat metabolic and adrenal health as bone health. Visceral adiposity, insulin resistance, and chronic high cortisol all damage the matrix in ways DXA may miss. See Metabolic flexibility and Stress.
Sleep close to 8 hours. Bone resorption follows an intrinsic circadian rhythm, and the clock that governs it is set by regular sleep and daylight. Shift work is a flagged risk factor.
Don't smoke; cap alcohol. Both are directly toxic to bone (osteotoxic) — smoking suppresses osteoblasts and lowers estrogen, and heavy alcohol impairs bone formation and raises fall risk. See Alcohol.

What's overhyped or wrong

"Walking is enough for bones." The mechanical strain produced by ambulation is below the osteogenic threshold for the lumbar spine, where pooled walking trials find nothing at all. It buys a little at the femoral neck. Walking is excellent cardiovascular care; it is not the osteogenic prescription.
"Calcium and vitamin D will protect my skeleton." Necessary, not sufficient. Without mechanical loading the substrate has no signal to assemble around.
"Heavy lifting will break my fragile bones." LIFTMOR and follow-on trials specifically tested this in adults already diagnosed with osteopenia and osteoporosis under supervision and reported zero fragility fractures. The 20th-century habit of prescribing only low-impact exercise to high-risk patients is one of the most expensive pieces of received wisdom in geriatric medicine. There are two real exclusions, both from the UK expert consensus rather than from caution: avoid postures involving a high degree of spinal flexion — loaded rounding of the back, not the neutral-spine hinge of a deadlift — and if you have already had a vertebral fracture or multiple low-trauma fractures, keep impact to about the level of brisk walking.^[77] Roughly two-thirds of vertebral fractures are silent, so "my back doesn't hurt" is not evidence of an intact spine.
"Whole-body vibration plates build bone." Modest at best in meta-analyses; not a substitute for progressive resistance with impact.
"Spinach is a great calcium source." Oxalate binds the calcium; net bioavailability is poor. Eat spinach for other reasons.
"DXA looks normal, so my bones are fine." DXA misses microarchitectural decay (TBS catches more), and in insulin-resistant adults it can be falsely reassuring because AGEs compromise bone quality without a density change.
"Generic bone-health supplement stacks fix it." Most commercial blends are under-dosed on K2, over-dosed on undirected calcium, or built around evidence-light ingredients (e.g., strontium citrate). The targeted stack is plain: D3 + K2 (MK-7) + magnesium + adequate dietary protein + calcium adjusted to dietary intake.
"Stem cells / GPR133 / CCN3 will fix it soon." Promising, but preclinical or early-clinical. Mechanical loading is the only intervention with a 30-year evidence base and a today-actionable protocol.