Training · Bone
Every link in the chain is real. The chain still does not hold.
The argument runs: training builds bone, bone means fewer breaks when you fall, and not breaking a hip in old age means living longer. It is a good argument. Every link in it has genuine evidence behind it, and we tested them one at a time rather than as a whole, because that is the only way to find out where an argument actually fails.
It fails on multiplication. A small effect times a partial effect times a time-limited effect is a very small effect. What survives is a different route to the same advice, and it is better evidenced than the one usually argued.
The chain, link by link
- 01
Training increases bone density
True, and smaller than almost anyone assumes
Spine density rises a few percent. Femoral neck density barely moves: in the best trial it rose 0.3% while controls lost 1.9%, so what loading buys at the hip is the avoided loss rather than a gain. Osteoporosis drugs move the same site by around 9%, roughly thirty times as much. Loading also has to be heavy or impactful to do anything at all, which rules out walking and swimming.
- 02
Higher bone density means fewer fractures
Strong for predicting, weaker for changing
Low density predicts fracture powerfully, with a relative risk of 2.6 per standard deviation at the hip, a steeper gradient than blood pressure carries for stroke. But the evidence that raising density lowers fracture risk comes almost entirely from drug trials, and there is a validated threshold of change that has to be cleared before a density gain counts as a surrogate for fewer fractures. Exercise does not reach that threshold.
- 03
Breaking a hip in old age kills you sooner
True, and concentrated in six months
With prospectively collected pre-fracture health data, the hazard ratio in the first six months is 6.28 and survives full adjustment, which is about as causal as observational evidence gets. After six months the excess vanishes: 1.04, not significant. The danger is real, acute, and shorter-lived than the decade usually quoted.
- 04
Therefore training means fewer fractures and longer life
The links multiply, and the product is small
Take the most favourable defensible number at each step and the arithmetic gives a mortality effect of roughly one to three percent of post-hip-fracture deaths. Each link is individually defensible. Multiplied together they do not carry the weight the argument wants them to. The most direct test of the underlying logic, a hip protector that does nothing but cushion the impact, shows a risk ratio of 1.15 in people living in the community.
What survives, and why it is a better argument
Three things come through the analysis intact, and none of them is about bone density.
The first is that exercise reduces the rate of falls by 23%, and Cochrane grades that high certainty across 108 trials. It is the best-evidenced finding in this entire area. Not falling beats surviving a fall, and the programmes that do it are built around balance and functional work rather than lifting alone.
The second is that leg power predicts fracture independently of bone density in men around 84, while raw force does not. That is a route from muscle to unbroken bones that bypasses density entirely, and it points at training for speed and power rather than only for load.
The third is that bone strength is not the same as bone density, and the standard scan measures the wrong one. Loading changes the geometry of a bone, its cortical thickness and how it resists twisting, and those change far more than density does. In the trial above, cortical thickness at the femoral neck rose 13.6% while density at the same site rose 0.3%. A DXA scan systematically understates what loading did.
So the advice survives and the reasoning changes. Train heavy, train for power, and train balance, because the case rests on not falling, on landing better, and on a bone that is structurally stronger than its density score suggests. Not on adding a few percent to a number on a scan.
The evidence behind each link
observational
Leg power predicted fracture even after adjusting for bone density. Raw strength did not.
Measured in
1,841 older men, median age 84, jump power and velocity measured on a force plate, 136 major osteoporotic fractures over 5 years, all self-reported fractures confirmed against medical records
Men with greater leg power and greater movement velocity broke fewer hips, spines, wrists and shoulders. The association held after adjustment for established fracture risk factors including bone density. Force at peak power, which is what most people mean by strength, did not carry the association. Power and velocity did.
HumanLeg power and velocity predict the risk of major osteoporotic fractures (MOF): the Osteoporotic Fractures in Men (MrOS) study (2026)What this does not show
That training to build power prevents fractures. Nobody was assigned to anything, and men who can still jump at 84 differ from men who cannot. What it does establish is that there is a route from muscle to fracture risk that does not run through bone density at all, which is the opposite of how this is usually argued.
randomised
Heavy lifting built bone in osteoporotic women, with one adverse event in the entire trial
Measured in
101 postmenopausal women with low bone mass, mean age 65, randomised to 8 months of twice-weekly 30 minute supervised high-intensity resistance and impact training at five sets of five above 85% of one-rep max, against a low-intensity home programme
Lumbar spine density rose 2.9% against a 1.2% loss in controls. Femoral neck cortical thickness rose 13.6% against 6.3%. Height was preserved, rising 0.2 cm while controls lost 0.2 cm. Every functional measure improved. One adverse event was recorded across the whole trial, a minor back spasm.
HumanHigh-intensity resistance and impact training improves bone mineral density and physical function in postmenopausal women with osteopenia and osteoporosis: the LIFTMOR randomized controlled trial (2018)What this does not show
A large density gain at the hip. Femoral neck density rose 0.3%, so the significant between-group difference came almost entirely from the control group losing 1.9% rather than from training adding much. The honest reading at the hip is that loading stopped the loss. What it does overturn is the standard advice that women with osteoporosis should avoid heavy loading, which this trial contradicts directly and under supervision.
observational
Low bone density predicts fracture better than blood pressure predicts stroke
Measured in
Eleven study populations, roughly 90,000 person-years of observation and over 2,000 fractures in women
Each one standard deviation decrease in hip bone density carried a relative risk of 2.6 for hip fracture (95% CI 2.0 to 3.5), and 2.3 at the spine for vertebral fracture. The authors note that this predictive ability is better than a one standard deviation rise in blood pressure is for stroke, and better than cholesterol is for cardiovascular disease.
ReviewMeta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures (1996)What this does not show
That raising density by a given amount lowers fracture risk by the matching amount. Predicting and modifying are different claims. The authors' own conclusion is that density predicts risk but cannot identify who will fracture, and they explicitly do not recommend screening healthy women by bone density on the strength of it.
observational
After a hip fracture the danger is concentrated in six months, not spread over a decade
Measured in
25,178 US Medicare beneficiaries followed a median 3.8 years, of whom 730 fractured a hip, with pre-fracture health and function collected prospectively rather than reconstructed afterwards
Adjusted only for age, sex and race, mortality was raised both early and late. Adjusted additionally for pre-fracture health, functional impairment, comorbidity and socioeconomic status, the excess was confined to the first six months, where the hazard ratio was 6.28 (4.82 to 8.19). Beyond six months it was 1.04 (0.88 to 1.23).
HumanExcess mortality following hip fracture: the role of underlying health status (2007)What this does not show
That the fracture is harmless after six months, or that it is merely a marker. A hazard ratio above 6 that survives full adjustment for prior health is about as close to causal as observational data gets. What it does correct is the widely repeated claim that a hip fracture shortens life for ten years. The authors conclude that all of the long-term excess was explained by the greater frailty of the people who fractured.
randomised
The device that does nothing but cushion the hip does not work outside care homes
Measured in
Pooled randomised trials of hip protectors: 14 studies and 11,808 participants in nursing or residential care, and five studies with 5,614 participants living in the community
In care homes, a small reduction in hip fracture, risk ratio 0.82 (0.67 to 1.00), which works out at 11 fewer fractures per 1,000 people. In the community, risk ratio 1.15 (0.84 to 1.58), which is no effect.
ReviewHip protectors for preventing hip fractures in older people (2014)What this does not show
That padding a hip cannot prevent a fracture in principle. Acceptance and adherence are poor, and people do not wear them at the moment they fall. It is on this page because it is the cleanest available test of the idea that the way to survive a fall is to make the impact survivable, and that idea performs badly the moment it has to survive contact with how people behave.
Why nobody has simply tested this
Fractures are rare events, which is a blessing for people and a problem for trials. Detecting a plausible reduction in hip fracture from exercise in a general older population needs tens of thousands of participants followed for years, and the estimate climbs further once you account for how few people in a real trial actually do the exercise they were assigned.
That is why the honest sentence is not that exercise fails to prevent fractures. It is that the trial capable of answering the question has never been run, and given what it would cost and who would own the result, it may never be. The same economics that empties the trial record across the rest of this site applies here too.
Questions people ask about this
- Does lifting weights prevent hip fractures?
- No randomised trial has ever shown a reduction in hip fracture from exercise, and the largest trial using fractures as its endpoint pointed slightly the wrong way. That is not the same as exercise being useless here. It means the trial that could answer it has never been run, and the sample size it would need runs to tens of thousands of people followed for years.
- So is training pointless for bone?
- Not at all, but the strongest route is not the one usually advertised. Exercise cuts the rate of falls by 23% with high-certainty evidence, which is the best-graded finding in this whole area, and leg power predicts fracture independently of bone density. Both of those run through not falling and through landing better, rather than through a denser femur.
- How much does training actually raise bone density?
- Spine density moves a few percent and femoral neck density barely moves at all. In the best trial the hip gain was 0.3% while untrained controls lost 1.9%, so the honest description at the hip is that loading stops the loss. For comparison, osteoporosis drugs move the same site by around 9%.
- Should someone with osteoporosis lift heavy?
- The standard advice says no. The trial that tested it directly put postmenopausal women with low bone mass under five sets of five above 85% of their one-rep max for eight months and recorded one adverse event in the entire study, a minor back spasm. That was under supervision, which is the part of the finding that should not be dropped.
- Does a broken hip really shorten your life?
- In the first six months, yes, and the effect survives adjustment for how healthy the person was beforehand, with a hazard ratio above six. Beyond six months the excess disappears once pre-fracture frailty is accounted for. The commonly repeated claim that a hip fracture shortens life for a decade does not survive that adjustment.