💊Clinical Bottom Line

This letter argues that the Institute of Medicine (IOM) made a critical statistical error when setting the 600 IU/day Recommended Dietary Allowance (RDA) for vitamin D. The authors claim the IOM mistakenly based its recommendation on the 95% prediction interval for study averages rather than for individuals.

The authors’ re-analysis of the IOM’s own data sources suggests that an intake of 600 IU/day only ensures 97.5% of individuals will achieve a serum 25(OH)D level of 26.8 nmol/L, not the IOM’s target of 50 nmol/L. While they caution their own data is an extrapolation, the authors estimate that a dose of 8,895 IU/day might be required to achieve the IOM’s stated goal for 97.5% of the population. This analysis strongly implies the current RDA is insufficient to meet its intended public health objective.

Results in Context

Main Results

The paper’s primary finding is a re-estimation of the vitamin D dose-response curve, which yields dramatically different results from the IOM’s 2011 report.

• IOM’s Finding: The IOM concluded that 600 IU/day would achieve a serum 25(OH)D level of 50 nmol/L or more in 97.5% of healthy individuals.

• Authors’ Re-Analysis: The authors’ new model, based on the IOM’s source data, found that 600 IU/day achieves a 2.5th percentile serum level of only 26.8 nmol/L.

• Authors’ Extrapolation: To achieve the IOM’s goal (a 2.5th percentile of 50 nmol/L), the authors’ model estimated an intake of 8,895 IU/day would be necessary.

The Statistical Critique

The authors argue the IOM’s error was confusing two different statistical concepts:

1. IOM’s Method (Figure 1): The IOM aggregated 32 study averages from 10 studies and calculated a 95% prediction interval. The authors state this interval only predicts where 97.5% of future study averages would fall.

2. Authors’ Method (Figure 2): The RDA, by definition, must cover 97.5% of individuals. To model this, the authors state one must account for the variation between individuals, which is much greater than the variation between study averages. They did this by using the standard deviations reported in the original studies to calculate the 2.5th percentile (average minus 2 standard deviations) for 23 data points. They then regressed these values (the yellow dots in Figure 2) to create a new, lower prediction limit (the red dashed line).

Participants

The data for this re-analysis was drawn from the same 10 supplementation studies used by the IOM. The IOM’s analysis used 32 study averages from these 10 studies. The authors’ re-analysis was limited to 23 study averages from 8 of those studies, as only those 8 provided the standard deviations necessary to perform their calculation.

Assertive Critical Appraisal

Limitations & Bias

The “study” being appraised is the authors’ re-analysis.

• Key Strength: The primary strength is its methodological transparency. The authors identify a specific and plausible statistical flaw in the IOM’s report—confusing the variance of study means with the variance of individuals. Their argument is clearly illustrated by comparing Figure 1 (IOM’s model) with Figure 2 (their re-analysis), which shows their calculated 2.5th percentiles (yellow dots) fall well below the IOM’s prediction line.

• Primary Limitation (Acknowledged): The estimate that 8,895 IU/day is needed to reach the target is a major extrapolation. As shown in Figure 2, the original studies’ data cluster well below 1,200 IU/day. The authors correctly state that “caution is warranted” when interpreting this specific number. However, this extrapolation doesn’t invalidate their primary critique: that 600 IU/day is statistically insufficient to reach 50 nmol/L in 97.5% of people.

• Supporting Evidence: The authors strengthen their claim by citing two Canadian studies which found that 10% and 15% of participants, respectively, had levels below 50 nmol/L despite supplementation at or above the RDA level. This provides real-world evidence that the RDA is not achieving its 97.5% target.

Reporting Quality Assessment (STROBE)

While this is a “Letter” and not a traditional observational study, its reporting is clear and its methods are explicitly stated.

• Addressing Methods: The authors are exceptionally clear about their methods. They state they “subtracted 2 standard deviations from the average” for 23 data points and “regressed these 23 values against vitamin D intake.” The “confounding variable” they address is the IOM’s statistical method itself. The use of comparative figures (Figure 1 vs. Figure 2) is a highly effective way to report their findings and makes the statistical argument accessible.

Applicability

The findings are highly relevant to clinical and public health practice. The RDA is a foundational metric for patient guidance and food fortification policy. If the authors’ critique is correct, the IOM’s 600 IU/day recommendation is based on a statistical error and is insufficient to meet its own stated goal of preventing deficiency, with “serious” public health implications.

Research Objective

The objective was to identify and correct a statistical error in the Institute of Medicine’s (IOM) 2011 estimation of the Recommended Dietary Allowance (RDA) for vitamin D.

Study Design

This is a methodological critique and re-analysis of the data used by the IOM. The authors reviewed the 10 supplementation studies the IOM used, extracted data on averages and standard deviations from the 8 studies that reported them, and performed a new regression analysis to model the dose-response relationship for the 2.5th percentile of individuals.

Setting and Participants

The “participants” in this analysis are the data points from 10 supplementation studies. These original studies were conducted at high latitudes during winter to minimize the influence of cutaneous vitamin D synthesis. The authors’ re-analysis used 23 data points from 8 of these studies.

Bibliographic Data

• Title: A Statistical Error in the Estimation of the Recommended Dietary Allowance for Vitamin D

• Authors: Paul J. Veugelers and John Paul Ekwaru

• Journal: Nutrients

• Year: 2014

• DOI: 10.3390/nu6104472