Seventeen RCTs (4,305 participants, range 53 to 1,090) were included in the review. Four RCTs had unclear risk of bias for at least three criteria and seven RCTs were at high risk of bias in at least one domain. Trial follow-up durations ranged from six to 30 months.
Fall-prevention exercise interventions statistically significantly reduced risk for all injurious falls (RR 0.63, 95% CI 0.51 to 0.77; 10 RCTs), falls resulting in medical care (RR 0.70, 95% CI 0.54 to 0.92; eight RCTs), falls resulting in serious injuries (RR 0.57, 95% CI 0.36 to 0.90; seven RCTs) and falls resulting in fractures (RR 0.39, 95% CI 0.23 to 0.66; six RCTs). There was some evidence of significant statistical heterogeneity for overall risk of injurious falls (I²=50%) and falls resulting in serious injuries (I²=46%).
Other subgroup analyses and sensitivity analysis did not significantly alter the findings. Adverse events were rarely reported.
A funnel plot gave no clear evidence of publication bias.