Sixteen RCTs (23 publications) were included (n=2,139, range 14 to 622). Twelve studies had a Jadad score of three or more. Four studies reported adequate allocation concealment. Seven studies had losses to follow up of less than 20 per cent (range six per cent to 30 per cent, where stated).
Primary outcomes
Calcitriol or alfacalcidol (with or without calcium) versus placebo or calcium: There was no statistically significant difference between the groups in the risk of new vertebral fractures (OR 0.89, 95% CI: 0.57, 1.39, 13 RCTs, n=1,396). The funnel plot for this outcome suggested significant publication bias (p=0.10). There were significantly fewer non-vertebral fractures (OR 0.51, 95% CI: 0.30, 0.88, six RCTs, n=1,014) and falls (OR 0.66, 95% CI: 0.44, 0.98, two RCTs, n=624) in the intervention group.
Withdrawals and harms
Calcitriol or alfacalcidol (with or without calcium) versus placebo or calcium: The intervention group had a significantly higher risk of hypercalcaemia (OR 3.63, 95% CI: 1.51, 8.73, 10 RCTs) and a trend to increased risk of hypercalciuria (five RCTs, unsuitable for meta-analysis). There was no statistically significant difference between the groups in the risk of total or harm-related withdrawal, death, renal stones or gastrointestinal side effects.
Calcitriol or alfacalcidol with calcium versus calcium alone: There was a significantly higher risk of hypercalcaemia in the intervention group (OR 2.85, 95% CI: 1.21, 6.70, five RCTs, n=536).
Subgroup and sensitivity analyses
Alfacalcidol (with or without calcium) was associated with a significantly lower rate of vertebral fractures than placebo or calcium (OR 0.50, 95% CI: 0.25, 0.98, five RCTs, n=410), although calcitriol was not (eight RCTs, n=986). Other analyses did not change the statistical significance of the overall results.
Statistical heterogeneity was generally low for all analyses.