Design: Hierarchical Bayesian network meta-analysis with dose-response modelling
Population: 124 randomised trials, 18,429 adults aged 40 and over
Primary outcome: Change in bone mineral density at lumbar spine, femoral neck and total hip
Effect: Brisk walking or jogging: lumbar spine +0.013 g/cm² (95% CrI 0.006–0.021), total hip +0.021 (0.005–0.038); meaningful benefit at about 600 METs-min/week

A Bayesian network meta-analysis pooled 124 randomised trials in 18,429 adults aged 40 and over to ask which exercise, at what dose, protects bone. At the lumbar spine, brisk walking or jogging (mean difference 0.013 g/cm², 95% credible interval 0.006 to 0.021) and combined aerobic-resistance training (0.013, 0.010 to 0.016) came out ahead. At the femoral neck it was brisk walking or jogging (0.009, 0.001 to 0.018) and mind-body exercise (0.007, 0.002 to 0.013). At the total hip, only brisk walking or jogging reached benefit (0.021, 0.005 to 0.038).

The dose findings are the practical part. The relationship was non-linear and broadly an inverted U: a clinically meaningful difference appeared at around 400 METs-minutes per week for the lumbar spine and around 600 for the femoral neck and total hip — roughly two to three hours of brisk walking. More was not better.

In the fracture analysis (26 trials, 11,132 participants), mixed aerobic exercise gave an odds ratio of 0.29 (95% credible interval 0.11 to 0.78) and mind-body exercise 0.58 (0.36 to 0.93), while other modalities were neutral. Certainty was rated low to moderate throughout, so the ranking between modalities is softer than the numbers look. The usable message is a dose and a modality a patient can actually be told.

Prescribe a dose, not an exhortation: about two to three hours of brisk walking a week, spread across most days.
Add resistance work for the spine; walking alone does less there than combined training.
Do not tell patients that more is proportionally better — the curve flattens and may turn down.
Mind-body exercise such as tai chi has a fracture signal, probably through balance rather than bone; it suits patients who cannot walk briskly.
Exercise supplements, and does not replace, calcium, vitamin D and antiresorptive therapy where those are indicated.

Why it matters

Exercise advice for bone health is usually given without a dose, and this gives one a patient can follow.

The statistics, in plain English

A credible interval is the Bayesian counterpart of a confidence interval, and these are narrow enough to exclude no effect — but the authors rate certainty as low to moderate, which reflects the quality of the underlying trials rather than the width of the interval. A bone mineral density difference of 0.013 g/cm² is small in absolute terms; it is the fracture odds ratios, not the density numbers, that would matter to a patient, and those rest on only 26 trials.