Six RCTs with over 1,656 participants were included in the review.
Study quality was not high. None of the studies reported allocation concealment, only two reported blinding of the outcome assessor, and only three reported withdrawals from treatment. Losses to follow-up, where reported, ranged from 2 to 41%.
Three studies (involving at least 322 postmenopausal women) were identified. There was no significant difference between the intervention and control groups in the incidence of spinal fracture (3 studies; RR 0.52, 95% CI: 0.17, 1.60); statistically significant heterogeneity was detected (I-squared 57%). One trial (n=92) reported no significant difference between the exercise and control group for wrist fractures (RR 1.78, 95% CI: 0.39, 0.86) or fractures at any site (RR 0.91, 95% CI: 0.34, 2.50).
Two studies (1,076 patients aged over 65 years, 36% male, living in residential care homes or own home) were identified. Two trials reported hip fractures. Interventions were associated with a non statistically significant reduction in hip fractures compared with control (RR 0.37, 95% CI: 0.13, 1.03, p=0.06); no statistically significant heterogeneity was detected (I-squared 0%). One trial (n=674) reported fractures at any site. There was no significant difference between the groups (RR 1.00, 95% CI: 0.50, 2.00).
Exposure to sunlight.
One study (258 chronically hospitalised disabled stroke patients aged over 65 years, 40% male) was identified. There was no significant difference between the groups in the incidence of hip fractures (RR 0.17, 95% CI: 0.02, 1.35).
Withdrawal from treatment.
The majority of studies either did not report withdrawals or reported no withdrawals. The exception was the trial of exposure to sunlight in which there was no difference between the groups in the number of withdrawals (8% in both groups).