Thirty-four RCTs were included in the review. Fourteen RCTs (n=11,510) compared native vitamin D with placebo or calcium, 18 RCTs (n=3,103) compared vitamin D analogues with placebo or calcium, and 2 RCTs directly compared vitamin D analogues and native vitamin D (n not reported).
The quality scores ranged from 40 to 100 for studies used in indirect comparisons (the methods used to transform scores on the 5-point Jadad scale into those reported were not described).
Indirect comparisons.
Patients not exposed to corticosteroids.
Indirect comparisons suggested that vitamin D analogues were associated with less bone loss at the hip and lumbar spine than native vitamin D after a median of 24 months; the ES for BMD was 0.36 (P<0.0001) for vitamin D analogues versus control and 0.17 (P<0.0001) for native vitamin D versus control; the difference in ES for analogue versus native vitamin D was 0.19 (P<0.05). The difference between analogue and native vitamin D remained significant for BMD of the lumbar spine (ES difference 0.22, P=0.047). There was no significant difference between analogue and native vitamin D at other sites.
Reductions in fracture rates were greater for vitamin D analogues than for native vitamin D: the RDs were 10% (95% confidence interval, CI: 2, 17) and 2% (95% CI: 1, 3), respectively. The RD difference between analogue and native vitamin D was 8% (95% CI: 1, 16). The RD difference between analogue and native vitamin D remained significant for spinal fracture: RD 13.4% (95% CI: 7.7, 19.8, P<0.001).
Patients taking corticosteroids.
Indirect comparisons showed no significant difference in bone loss at 24 months between vitamin D analogues and native vitamin D: the ES for BMD was 0.38 (P<0.0001) for vitamin D analogues versus control and 0.41 (P=0.002) for native vitamin D versus control; the ES difference for analogue versus native vitamin D was -0.03 (ANOVA P=0.84). Statistical heterogeneity was found for the meta-analysis of vitamin D versus control and a random-effects model was used; the study responsible for the heterogeneity was identified.
Neither vitamin D analogues nor native vitamin D significantly reduced the risk of fractures after 24 to 36 months. There was no significant difference between analogue and native vitamin D (RD 3.8%, ANOVA P=0.39).
Direct comparisons (2 studies, n was not reported).
Head-to-head studies of patients taking corticosteroids showed that vitamin D analogues significantly reduced bone loss (ES 0.31, P=0.02), bone loss from the femoral neck (ES 0.34, P=0.003) and spinal fractures (RD based on 1 RCT 15%, 95% CI: 6.5, 24.8) compared with native vitamin D.
The logistic regression showed no significant association between any of the variables tested and fracture rate.
The funnel plot of ESs showed slight asymmetry, suggesting the possibility of publication bias, but this did not reach statistical significance (P=0.12).