Statistical approaches to establish mineral element recommendations.

When a committee develops a statement of requirements, the members first look to laboratory studies, and often to animal models, to identify the nature of the function of the nutrient and likely markers of this function. Then the committee looks for evidence from human studies. There are several types of evidence that they could, and should, draw upon; each has particular strengths and weaknesses. Given that all available evidence points to the fact that nutrient requirements differ among individuals, and given also the fact that unless explicitly controlled in study design, nutrient intakes vary among individual subjects, we must deal with distributions of requirements and of intakes. How we deal with distributions in different designs can lead to apparent differences in what we estimate. This paper illustrates some approaches that should be used in reviewing evidence pertaining to human nutrient requirements; it attempts to offer unifying concepts that can help us take advantage of these several types of evidence. The final message is that in estimating nutrient requirements, "congruence of evidence is happiness"; we can no longer be satisfied with a single line of evidence.