Last year (2014) saw an unusual event. Two statisticians at the University of Alberta in Ednonton, Canada (Paul Veugelers and JP Ekwaru) published a paper in the online journal Nutrients (6(10):4472-5) showing that the Institute of Medicine (IOM) had made a serious calculation error in its recommended dietary allowance (RDA) for vitamin D. Immediately, other statisticians checked the Canadians’ analyses and found that, indeed, they were right. Together with my colleagues at Grassroots Health, I went back to square one, starting with a different population entirely, and came to exactly the same conclusion. The true RDA for vitamin D was about 10 times higher than the IOM had said. Not a small error. To understand, how this might have happened and why this is important, some background may be helpful.
An RDA is technically the amount of a nutrient every member of a population should ingest to ensure that 97.5% of its members would meet a specified criterion of nutritional adequacy. For vitamin D, the IOM panel determined that the criterion for adequacy was a serum concentration of a particular vitamin D derivative (25-hydroxyvitamin D) of 20 ng/mL or higher, and that for adults up to age 70, 600 IU of vitamin D per day was the RDA.
Both of those figures provoked immediate and unprecedented dissent from a diverse group of nutritional scientists, but the disagreement centered mostly around the IOM panel’s reading and interpretation of the evidence, rather than its calculation of the RDA. The Edmonton statisticians took the dissent a step further, showing that the actual calculation was itself wrong. Here’s what seems to have happened.
Not everyone gets the same response to a given intake of any particular nutrient, i.e., some require more than others to reach the specified target, and while the average response to a certain dose of vitamin D may be above the target level, a substantial fraction of a population can still be below it. Thus, the RDA will always be higher than the average requirement, and for some nutrients, substantially so. As a consequence, ensuring that every member of a population receives the RDA guarantees that 97.5% of that population will be getting at least enough, while many will be getting more than they actually need.
The IOM panel identified a number of published studies showing the 25-hydroxyvitamin D response to various vitamin D doses. They plotted the average response in each of those studies against dose, thereby generating what is termed a “dose response curve”, i.e., a way to estimate how much of a response would be predicted for any given vitamin D intake. But, to make a long story short, because it used average responses, that curve tells us nothing about the intake requirement for the individual members of a population, and particularly those whose response to a given dose falls in the bottom 2.5 percentiles. The IOM panel surely knew that the average intake required to meet or exceed 20 ng/mL was not the same as the RDA, as it would be inadequate for all those with below average responses (about half the population). So, to catch the “weak” responders, they calculated the 95% probability range around their dose response curve, designating as the RDA the point where the bottom end of that probability range exceeded 20 ng/mL. While this might seem to have been the right approach, it was not. The panel appears to have overlooked the fact that the 95% probability range for their curve is for the average values that would be expected from similar studies at any particular dose. The dispersion of averages of several studies is, as every beginning student of statistics knows, much more narrow than dispersion of individual values within a study around its own average. And it’s the 2.5th percentile individual values from those studies, not the study averages, that should have been used to create the relevant dose response curve.
It’s this latter approach that the Canadian statisticians used. They took precisely the same studies as the IOM had used and demonstrated that the requirement to ensure that 97.5% of the population would have a value of at least 20 ng/mL, was 8,895 IU per day. Recall that the IOM figure was less than 1/10 that, i.e. 600 IU per day up to age 70 (and 800 IU per day thereafter). When my colleagues and I analyzed the large GrassrootsHealth dataset, we calculated a value closer to 7,000 IU per day, still a full order of magnitude higher than the estimate of the IOM, and not substantially different from the estimate of Veugelers and Ekwaru.
Why This Is A Problem
This is an important mistake, not simply because it shouldn’t have been allowed in a major policy document, but because IOM recommendations have important effects on a wide array of government programs. These include nutritional standards for US military, for school lunch programs, for WIC and many others, both in the United States and in Canada.
Canada, which paid one third the cost of generating the IOM report, is in a particularly difficult situation. Its First Nations peoples, living near the Arctic Circle, do not get any vitamin D from the sun, as do those of us living at more temperate latitudes. They are totally dependent upon food and supplement sources. Their ancestral diets, based largely on seals and whales, constituted a rich source of vitamin D. They are much less commonly consumed today, in part because of the ready availability of low nutrient density foods flown in from the south, and in part because environmental pollution has made seal and whale products a source of dangerous toxins (as well as necessary nutrients). The Canadian government, responsible for the health of all of its citizens, can turn only to the existing IOM recommendation (600 IU per day) to set standards for the people living in its northern territories. But, as the Edmonton statisticians noted, that number is woefully inadequate.
There is almost no public awareness of this error or its implications in the United States, but that is not true for Canada. A large nutritional health foundation located in Calgary (Pure North S’Energy Foundation) has taken out a series of half page advertisements in Canada’s national newspaper (Globe and Mail), alerting Canadians to the fact that the error was made and that they need more vitamin D than current policy indicates (http://www.purenorth.ca/?page_id=1356). The IOM, Health Canada, and the Canadian Ministry of Health have all been formally alerted to this problem. The Health Ministry has agreed to undertake an independent reanalysis of the calculation of the RDA, but the results are not yet available and the shape of the ministry’s action is still uncertain.
How It May Have Happened
It’s one thing to know how the mistake was made, and quite another to know how it could have happened. Here, one can only speculate, as the IOM processes are shrouded in secrecy. The IOM report was a massive document and it is likely that much of the background work, such as the literature search, the drafting of the report, and the statistical calculations, were done by IOM staff members who may not, themselves, have been sufficiently expert in the vitamin D field to recognize discrepancies that might have popped up. (It is noteworthy that several of the dissenting letters submitted to scientific publications following release of the IOM report had specifically cited the fact that 600 IU per day was not sufficient to guarantee a level of 20 ng/mL.) It would then have been up to the expert panel to review and adjust this staff work. To be fair to the panel, it is important to understand that the scientific members of IOM panels are not compensated for their time and effort. They do it as a public service, and they are all busy scientists with work of their own. Still, it was their job, and one must wonder how they failed to see an error that was apparent to other equally knowledgeable, but outside, scientists.
There may be a moral here. It is widely recognized that many of the panel members, before coming together to review the evidence, had already staked out a position to the effect that, while the previous (1997) recommendation for vitamin D (200 IU per day) was probably inadequate, the actual RDA was almost certainly below 1000 IU per day. Accordingly, when the statistical calculations produced a number that matched their own expectations, they may not have been inclined to question its derivation.
There is a generally held belief that science is objective, data-driven. And to a substantial extent that is so. But science and scientists are not identical. Scientists often have strongly held opinions and, like people in general, find ways to construe the evidence to support their beliefs. When those beliefs are wrong, science, as a field, ultimately abandons them. I am confident that this IOM error will be corrected sooner or later. This is partly because it is demonstrably erroneous, and partly because the related set of IOM recommendations for vitamin D has not elicited a consensus in the field of vitamin D research. If the Dietary Reference Intakes produced by the IOM are important, then it is important that they be right. I can only hope that not too much human damage will occur as we wait for the needed correction to happen.
[Author’s note: if readers are aware of factual inaccuracies in what I have written here, I would be grateful if you would call them to my attention at , so that I can correct this post.]