Exercise for bone density: what kind of load actually counts

How hard you load a bone seems to matter. Heavy training raised bone density at the lower spine about three times as much as light training.¹ Walking on its own showed no significant effect on the spine.² The catch is that only four of the programs studied were genuinely heavy.¹

What these trials actually measure

Most studies here have the same shape. Postmenopausal women are assigned to an exercise program or to nothing, the program runs at least six months, and a scan measures bone density at the lower spine, the neck of the femur or the whole hip. Almost nothing in that design counts a broken bone.

The two measured sites are not arbitrary. The lower spine and the hip are where serious osteoporotic fractures tend to happen, so density there is the closest stand-in anyone has.

That is the central limitation. Density stands in for the thing anyone actually cares about. Across 43 trials in 4,320 postmenopausal women, exercise made no measurable difference to the number of fractures.³ The range of possible effects ran from a large benefit to a modest harm, so that result cannot tell either way. Density is what this evidence can speak to.

Walking is close to the control condition

The most common thing people do for their bones does the least to them. Across ten walking trials, walking on its own did not shift density at the lower spine, the wrist or the body as a whole. At the neck of the femur it did help, but only when the program ran longer than six months.²

A plausible reason is mechanical. Each footfall drives force up through the hip. Very little of it reaches the spine. The results follow the load.

Walking still earns its place, for the heart and the legs and the head. Changing the density of your spine is not one of the things it does.

Intensity is where the gradient shows

One analysis sorted the trials by how hard the exercise really was, because human studies had mostly tested light to moderate work.¹ It found 53 trials running 63 programs: 19 light, 40 moderate, and four heavy.¹

Programs in the intensity analysis that were heavy heavy intensity light or moderate

At the lower spine the gradient was clean. Heavy work added 0.031 g/cm² of density. Moderate added 0.012, light 0.010.¹ That is roughly three times the effect from the heavy end.

Lumbar spine bone density gain by intensity, g/cm²0 = no effect

Heavy intensity0.031 (0.012 to 0.049)

Moderate intensity0.012 (0.008 to 0.017)

Light intensity0.010 (0.005 to 0.015)

At the femoral neck it inverted. Light and moderate work did equally well, and heavy work showed nothing at all.¹ Four programs asked a question they were never big enough to answer.

Heavy has a definition worth carrying. In the resistance trials it means at least 70% of the most weight you could lift once.⁵ That is well past anything you could keep lifting while holding a conversation.

The same work points at frequency and duration too. Three sessions a week improved density at every site measured. Programs of 48 weeks or longer were the ones that moved the hip and femoral neck, and 40-minute sessions were what moved the spine.⁵ The trials behind those splits disagreed with each other badly, so treat them as leads rather than instructions.⁵

Where the load story falls apart

The evidence will not support a clean ranking of exercise types. A comparison of 84 exercise groups set weight-bearing exercise against resistance training against both together. All three moved density at all three sites, and their effects overlapped. At the lower spine: 0.40 for resistance training, 0.26 for weight bearing, 0.42 for the two combined.⁶

Those are standardized effects, read on a scale where 0.2 counts as small, 0.5 as moderate and 0.8 as large. All three land between small and moderate, and none stands apart from the others. The authors' own conclusion is that the benefit barely depends on which exercise you pick.⁶

The resistance-training work undercuts itself the same way. Across 17 studies the standardized effects came to 0.54 at the lower spine, 0.22 at the femoral neck and 0.48 at the whole hip.⁴ Then the breakdowns turned strange. Free weights beat machines at the hip, and training fewer than two days a week produced larger spine and hip changes than training more often.⁴ That flatly contradicts the three-sessions-a-week result.⁵

Two things survive the mess. Heavier beats lighter at the spine, and almost anything beats nothing. The exact prescription is not in the data.

What the density change is worth

Small, once you translate it. One analysis of exercise that drives force through the skeleton found modest gains at the femoral neck and lower spine, and estimated they would cut the 20-year risk of an osteoporotic fracture by about 11% and 10%.⁷

In the Cochrane review, the most effective type at the femoral neck was progressive resistance training for the legs, a 1.03% change in density from baseline. At the spine it was combined programs, at 3.22%.³

Those are percentages of a percentage. Nobody trains for a year and rebuilds a skeleton. What the trials describe is a modest gain over women who did not exercise.

What remains unknown

The heavy-training case rests on four programs.¹ Nothing here says whether heavy lifting is safe or useful for someone who already has osteoporosis, because the intensity analysis studied healthy postmenopausal women.¹ Whether impact work such as jumping and hopping adds anything is left open: the intensity analysis says only that resistance training, potentially combined with impact training, looked most effective.¹

The dose questions, how often and how heavy, get opposite answers from different analyses. One set of authors says plainly that evidence of this kind cannot settle them.^4,6 Reporting in the older trials was poor, particularly on how women were assigned to groups.³

And no trial here was built to count fractures. The outcome that actually matters is still an extrapolation from a density scan.⁷