Improved Dietary Guidelines for Vitamin D: Application of Individual Participant Data (IPD)-Level Meta-Regression Analyses

Abstract

Dietary Reference Values (DRVs) for vitamin D have a key role in the prevention of vitamin D deficiency. However, despite adopting similar risk assessment protocols, estimates from authoritative agencies over the last 6 years have been diverse. This may have arisen from diverse approaches to data analysis. Modelling strategies for pooling of individual subject data from cognate vitamin D randomized controlled trials (RCTs) are likely to provide the most appropriate DRV estimates. Thus, the objective of the present work was to undertake the first-ever individual participant data (IPD)-level meta-regression, which is increasingly recognized as best practice, from seven winter-based RCTs (with 882 participants ranging in age from 4 to 90 years) of the vitamin D intake-serum 25-hydroxyvitamin D (25(OH)D) dose-response. Our IPD-derived estimates of vitamin D intakes required to maintain 97.5% of 25(OH)D concentrations >25, 30, and 50 nmol/L across the population are 10, 13, and 26 µg/day, respectively. In contrast, standard meta-regression analyses with aggregate data (as used by several agencies in recent years) from the same RCTs estimated that a vitamin D intake requirement of 14 µg/day would maintain 97.5% of 25(OH)D >50 nmol/L. These first IPD-derived estimates offer improved dietary recommendations for vitamin D because the underpinning modeling captures the between-person variability in response of serum 25(OH)D to vitamin D intake.

Keywords: DRV; EAR; Individual Participant Data-level meta-regression analyses; RDA; vitamin D recommendations.

Figure 1

The relation between serum 25-hydroxyvitamin D (25(OH)D) concentrations (in late winter) and total vitamin D intake (i.e., from diet plus supplemental) in healthy persons aged 4–86 years living at northerly latitudes (between 51° N and 60° N) based on individual participant data (IPD) (n = 882 individuals) (A) and aggregate RCT group mean data (n = 23 arms) (B) from the same RCTs. The solid and dashed lines through the IPD data points (black circles) (A) correspond to the fitted regression lines based on the two-step (adjusted for age and baseline 25(OH)D) and one-step IPD analyses, respectively, and the corresponding 95% prediction bands are shown in grey (the lightest being the band for the one-step IPD analysis) (A). The fitted regression lines for the unadjusted and adjusted (age and baseline 25(OH)D) standard meta-regression based on aggregate data from same RCTs (black triangles) (B) are shown as dotted and dot-dashed lines with corresponding 95% confidence bands shown in grey (the lightest being the band for the unadjusted analysis).