What Actually Builds Bone — And Why Most Exercise Doesn't

Your heart has a resting rate. Your nervous system has HRV. Your lungs and mitochondria have a VO₂max estimate. Your metabolism has a glucose curve. Your skeleton has nothing at all.

Of all the tissues we deliberately train, bone is the one no consumer device reads. It produces no daily signal and says nothing whatsoever when you neglect it. It is also one of the tissues that quietly decides how the last twenty years of a life go. Most people learn the first fact about their own skeleton in a clinic, from a scan, after something has already happened.

So the skeleton gets managed by inference instead. And the most comforting inference — I exercise regularly, so my bones are probably fine — turns out to be the one the evidence contradicts most bluntly.

Bone listens for force, not for time

Almost everything else we train responds to accumulated dose. More minutes in Zone 2, more total protein, more steps: the ledger adds up. Bone does not keep that kind of ledger.

Bone remodels around a set-point for mechanical strain — the framework Harold Frost called the mechanostat. What the tissue detects is not how long a load lasted but how hard and how fast it deformed the bone. Loads below the set-point are treated as background noise, no matter how many of them arrive.

The cleanest numbers on this come from a rat-tibia loading experiment by Cullen, Smith and Akhter. At high strain, 36 loading cycles were enough to drive a maximal bone-formation response. Drop the strain a little and it took 120 cycles for the same job. Drop it further and 400 cycles were needed. And at that lowest strain, 40 cycles produced no detectable response at all — not a small one, none.

The practical translation is uncomfortable for anyone whose training is built on volume. An hour of walking is thousands of loading cycles at a strain the skeleton has already decided to ignore. On bone's own accounting, most of them are rounding error.

Which is why some of the fittest people have the least dense hips

If cardiovascular fitness protected bone, elite endurance athletes would have the strongest skeletons in the population. A Norwegian cross-sectional study put that to the test, comparing 21 elite runners with 19 elite road cyclists.

The cyclists had lower bone mineral density at every site measured — total body, femoral neck, lumbar spine. Ten of the 19 cyclists met the American College of Sports Medicine threshold for low bone density, a Z-score of −1 or below. These were not sedentary people. They were among the most aerobically capable athletes in a wealthy country, and the paper notes they reported heavy lower-body resistance training. Sport type was the only factor significantly associated with low density.

Swimming tells the same story for the same reason: water carries the load so the skeleton doesn't have to. Cardiorespiratory fitness and skeletal loading are two separate accounts, and paying into one does not credit the other.

The trial that broke the "be careful" rule

For decades the standard advice to anyone with thinning bones was to avoid heavy loading. It sounded obviously sensible, which is usually a sign that nobody has tested it.

The LIFTMOR trial tested it. One hundred and one postmenopausal women with low bone mass, average age 65, randomised either to a home-based low-intensity programme — the thing most of them would have been told to do — or to twice-weekly, 30-minute supervised sessions of five sets of five reps above 85% of one-rep max: deadlift, back squat, overhead press, plus jumping chin-ups with a drop landing. Eight months.

Lumbar spine density rose 2.9% in the training group and fell 1.2% in the controls. Femoral neck density rose 0.3% against a 1.9% fall. Femoral neck cortical thickness went up 13.6% versus 6.3%. The training group gained 0.2 cm of height while the controls lost 0.2 cm. Every functional measure favoured the lifters.

Adverse events across eight months of heavy barbell work in women with osteopenia and osteoporosis: one. A minor lower-back spasm, two missed sessions. Compliance was 92%.

The old caution was not wrong about heavy loading carrying risk. It was wrong about light loading being the safe alternative — because in this trial the gentle arm lost bone, height and function while the heavy arm gained all three. Worth naming the fine print, though: every session was supervised and the loading was introduced gradually. The exposure that produced this result was a coached programme, not a heavy barbell on its own.

In the trial that tested this, the exposure was two supervised half-hour sessions a week for eight months — five sets of five, above 85% of one-rep max.

Men have skeletons too, and almost no data

Bone research has been so concentrated on postmenopausal women that the male evidence base is startlingly thin. A 2026 meta-analysis by Feng and colleagues pooled everything usable: 12 randomised trials, 1,061 middle-aged and older men, mean ages spanning 51 to 78.

Exercise raised lumbar spine density by 0.13 g/cm² (95% CI 0.01–0.26) and femoral neck density by 0.17 g/cm² (0.04–0.30), both at moderate certainty. Total hip did not reach significance (0.08; −0.05 to 0.20). Greater trochanter, femoral shaft and whole-body density showed nothing.

That pattern is the finding. Bone adapts where the force lands and nowhere else — there is no such thing as generalised bone training. The doses that worked were at least three sessions a week held for longer than six months, and multicomponent programmes beat single-modality ones. The authors also flag what is missing: modest sample sizes, methodological concerns in several trials, and no follow-up beyond 18 months anywhere in the literature.

The half of this we'd rather not print

A second 2026 meta-analysis, from Miao and colleagues, pooled 22 trials and 1,051 adults — mostly premenopausal women aged 18 to 45 — and found no significant overall effect on lumbar spine density (SMD 0.15; −0.09 to 0.38) or femoral neck density (0.06; −0.11 to 0.22). Only the women under 30 showed clear gains.

But the blood told a different story. Bone-formation markers rose clearly: osteocalcin (SMD 0.41) and bone-specific alkaline phosphatase (0.71), with P1NP up in the shorter trials. The resorption markers — CTX, TRACP-5b — did not move.

Read that honestly and it says something useful. The building signal switched on, the breakdown signal did not switch off, and density is such a slow readout that most trials end before it can show anything. In a healthy 30-year-old with little room to improve, "no change on the scan" and "nothing happened" are not the same statement.

The larger caveat is that none of these trials counted fractures. Bone density is a surrogate for the thing anyone actually cares about — a hip that survives a fall at 78 — and no exercise trial has run long enough to measure that properly. Which is also why not falling in the first place belongs in the same conversation as bone density.

What to do about it this month

Heavy or fast, and not much of it. Bone is the rare tissue where the honest prescription is fewer, harder repetitions rather than more — three or four working sets carry more skeletal signal than an hour of light movement, and the strength dose that works is smaller than most people assume.

Load the sites you care about. Hips need hip loading; the spine needs axial loading; wrists, in practice, need something in the hands. Impact is free and counts: hops, landings, stairs descended rather than avoided.

Technique comes before load, and progression comes in months. The connective tissue around the movement adapts on a slower clock than muscle, so strength will arrive before tolerance does. And judge the outcome in years — a DXA scan roughly 12 months apart is the only real readout, because nothing you wear reads bone.

If you already have a diagnosis of osteopenia or osteoporosis, the LIFTMOR result is a reason to start that conversation with your clinician, not a reason to skip it. What was tested was a supervised programme with graded loading — the supervision is part of the intervention, not an optional extra.

On the nutrition side, keep expectations modest and the wording precise. Calcium is needed for the maintenance of normal bones. Vitamin D contributes to the maintenance of normal bones and to the normal absorption and use of calcium. Vitamin K contributes to the maintenance of normal bones too, and protein does the same. That is the floor the building work stands on. It is not the building work. No nutrient has ever been shown to substitute for the mechanical signal.

The bottom line

The wellness industry — and we are part of it — has built a beautiful instrument panel for the tissues that answer back. Heart rate, HRV, sleep, glucose, temperature: all of them stream, all of them can be graphed by Friday. Bone answers nothing, on any timescale a dashboard understands, which is a large part of why it gets forgotten.

Our usual rule is to use numbers to correct fantasy rather than to replace experience. Here there are no numbers to use, so the discipline has to come from somewhere else — from deciding that something matters even when nothing is measuring it. Two or three genuinely heavy sets, twice a week, held for years, in a body that keeps showing up. It is the least trackable habit in the whole protocol, and one of the few where the evidence points in a single direction.