Fracture risk across a wide range of physical activity levels, from sedentary individuals to elite athletes.

Objective: To examine how physical activity is associated with risk of different fracture outcomes across the full range of physical activity.

Methods: By combining information from three cohort studies and using generalized structural equation modelling, we estimated a continuous unitless latent variable reflecting physical activity that ranged from sedentary through elite athlete levels. Associations between physical activity and fracture outcomes were assessed with proportional hazards regression using restricted cubic splines with the mean physical activity (corresponding to 20-40 min walking or bicycling/day or 2-3 h exercise/week) as reference.

Results: Among 63,980 men and women (49-68 years) and during 13 years of follow-up, 8506 fractures occurred, including 2164 distal forearm, 779 proximal humerus, 346 clinical spine, and 908 hip fractures. Both lower and higher physical activity was associated with higher risk of any fracture compared to the mean. Physical activity at 1 standard deviation (SD) below the mean, corresponding to walking/bicycling <20 min/day or exercising <1-1 h/week, was associated with a lower risk of distal forearm fracture (hazard ratio [HR]: 0.92, 95% confidence interval [CI]: 0.85-0.99) and higher risk of hip fracture (HR: 1.24, 95% CI: 1.13-1.37), but no associations were seen above the mean physical activity level for these fractures. Physical activity was not associated with proximal humerus fracture but had a possible U-shaped association with clinical spine fracture.

Conclusion: Physical activity was non-linearly associated with fracture risk and the association differed across fracture sites. Up to 2-3 h weekly exercise is beneficial for the prevention of hip fracture but may increase the risk of distal forearm fracture.