as the AI, if based on observed mean intakes, incorporates the variability of both requirements and intake. The AI represents an informed judgment about what seems to be an adequate intake for an individual based on available information, whereas the RDA is a more data-based and statistically relevant estimate of the required level of intake for almost all individuals. For this reason, AIs must be used more carefully than RDAs.
In the derivation of the EAR or AI, close attention has been paid to determining the most appropriate criteria of adequacy. A key question is, Adequate for what? In many cases a continuum of benefits may be ascribed to various levels of intake of the same nutrient. Each EAR and AI is described in terms of the selected criterion or, in some cases, criteria. For example, the EAR, and thus the RDA, for folate for women of childbearing age is based on a combination of biochemical indicators or criteria. A separate recommendation is made for women capable of becoming pregnant to reduce the risk of a neural tube defect in the offspring if pregnancy occurs. There are many possible and equally legitimate criteria of adequacy. The criteria are discussed in each nutrient report as part of the rationale for the DRIs developed (IOM, 1997, 1998b, 2000).
The task of setting both median requirements (EARs) and ULs for apparently healthy persons of all ages and both genders in various physiological states is ambitious. Ideally, data from the target population on intakes at various levels and the functional effects of these intakes would be available. In reality the information base is often limited, and its reliability varies from nutrient to nutrient. These limitations are discussed in detail in each of the nutrient reports (IOM, 1997, 1998b, 2000). Users of these reports should recognize that the DRIs are estimates based on available data, and that even when an EAR, RDA, and a UL for a nutrient are provided for a life stage and gender group, there is considerable uncertainty about these values. The DRIs will continue to evolve as better information becomes available. When interpreting the results of assessments of individuals or groups, it is appropriate to consider possible limitations in the information base that was used to generate the relevant DRIs.