Submitted to: British Medical Journal
Publication Type: Other
Publication Acceptance Date: 4/25/2011
Publication Date: 5/3/2011
Citation: Bischoff-Ferrari, H.A., Willett, W.C., Orav, E.J., Kiel, D.P., Dawson-Hughes, B. 2011. In response to Dr. Jose AP da Silva: fall prevention with vitamin D clarifications needed. British Medical Journal. 342:d2608. DOI: 10.1136/bmj.d2608. Interpretive Summary:
Technical Abstract: We agree with Dr. da Silva, and wish to clarify three issues raised by the recent IOM report concerning our 2009 meta-analysis on vitamin D and fall prevention. 1. The IOM questioned the inclusion of Broe et al., which did not pre-specify falls as a primary or secondary outcome. While this did violate our inclusion criteria, the Broe trial took advantage of a high-quality fall assessment throughout the trial. In all other regards: blinding, randomization and fall ascertainment methodology, this trial qualified for the primary analysis. The IOM also questioned the omission of Law et al. The Law study was excluded appropriately because it was not blinded as required. 2. It was stated that the dose-response relationship in our Figure 3 was inappropriately presented. Our intent was to visually identify a threshold at which vitamin D appeared efficacious (when the RR dropped below 1). Unfortunately, Figure 3 was misinterpreted as representing the actual meta-regression. We therefore present a revision of Figure 3 without the trend line through the RRs. 3. Finally, it was stated that one trial (Broe et al.) which compared 4 different doses of vitamin D versus a common placebo was incorrectly treated in our meta-regression as 4 independent meta-analysis entries. We agree that there are stochastic dependencies, and re-ran the analyses as a random effects logistic regression, allowing for between-patient correlation within all arms of the Broe study. In this re-analysis, there is a significant 27% reduction in the odds of falling overall (OR=0.73 [.62, .87]), a 34% reduction for high dose vitamin D (700 to 1000 IU vitamin D per day; OR=0.66 [.53, .82]), and no reduction with low dose vitamin D (OR=1.14[.69, 1.87]). Although the interaction term does not reach significance (p = 0.06), this does not invalidate the significantly lower risk with doses of 700 IU or larger.