The evidence that exercise during growth or adulthood reduces the risk of fragility fractures is weak.

There has never been, and will never be, a randomized double-blind placebo-controlled trial demonstrating that exercise in youth, adulthood or old age reduces fragility or osteoporosis-related fractures in old age. The next level of evidence, a randomized, controlled but unblinded study with fractures as an end-point is feasible but has never been done. The basis for the belief that exercise reduces fractures is derived from lower levels of 'evidence', namely, retrospective and prospective observation cohort studies and case-control studies. These studies are at best hypothesis generating, never hypothesis testing. They are all subject to many systematic biases and should be interpreted with extreme scepticism. Surrogate measures of anti-fracture efficacy are the next level of evidence, such as the demonstration of a reduction in risk factors for falls, a reduction in falls, a reduction in fractures due to falls, an increase in peak bone size and mass, prevention of bone loss in midlife and restoration of bone mass and structure in old age. Copyright 2001 Harcourt Publishers Ltd.