Analysis of the NICHD Vitamin D Pregnancy Cohort on a Per-Protocol vs. Intent-to-Treat Basis: The Effect of Adherence on Trial Results

Abstract

Objective: To perform per-protocol analysis of data obtained from the NICHD vitamin D pregnancy study published by Hollis et al., which found via intent-to-treat analysis that 4000IU/day vitamin D supplementation is safe and effective in achieving sufficiency in women and neonates. This study hypothesizes that differential adherence as examined by per-protocol analysis will affect the magnitude of differences in maternal and neonatal vitamin D status between treatment groups.Study design: A double-blind, RCT of vitamin D supplementation (400, 2000 or 4000IU/day) in 350 Caucasian, African American and Hispanic women with singleton pregnancies was conducted. This study defines adherence as 75% and 85% of pills taken between visits and examines the effect of adherence on vitamin D status across treatment groups. The primary outcome, measured by radioimmunoassay, is maternal serum 25(OH)D throughout pregnancy, one month prior to delivery (PTD) and neonatal serum 25(OH)D at delivery.Results: No statistically significant difference in maternal 25(OH)D throughout pregnancy, 1-month PTD, or neonatal 25(OH)D were found between 75% adherent participants and nonadherent participants regardless of supplementation. At 85% adherence, maternal 25(OH)D throughout pregnancy, one month PTD, and neonatal 25(OH)D were significantly higher in the 4000IU group compared to nonadherent participants (p=0.0002, p=0.0074, p=0.0068, respectively). No significant differences were found with 400 or 2000IU supplementation regardless of adherence.Conclusions: Participants 85% adherent to protocol and receiving 4000IU vitamin D were the only group that demonstrated significantly higher vitamin D status for each outcome: maternal 25(OH)D throughout pregnancy, 1-month PTD and neonatal 25(OH)D. Compared to intent-to-treat, this powerful per-protocol analysis demonstrates the impact that nonadherence can have on study results and has implications for how clinical trial data are analyzed and presented.