erations, including obligatory fetal transfer, if data are available, and on increased maternal needs related to increases in energy or protein metabolism, as applicable. For chromium, manganese, and molybdenum, the AI or EAR is determined by extrapolating up according to the additional weight gained during pregnancy. Carmichael et al. (1997) reported that the median weight gain of 7,002 women who had good pregnancy outcomes was 16 kg. No consistent relationship between maternal age and weight gain was observed in six studies of U.S. women (IOM, 1990). Therefore, 16 kg is added to the reference weight for nonpregnant adolescent girls and women for extrapolation.

It is assumed that the total nutrient requirement for lactating women equals the requirement for nonpregnant, nonlactating women of similar age plus an increment to cover the amount needed for milk production. To allow for inefficiencies in the use of certain nutrients, the increment may be greater than the amount of the nutrient contained in the milk produced. Details are provided in each nutrient chapter.

Reliable and valid methods of food composition analysis are crucial in determining the intake of a nutrient needed to meet a requirement. For nutrients such as chromium, analytic methods to determine the content of the nutrient in food have serious limitations.

The quality of nutrient intake data varies widely across studies. The most valid intake data are those collected from the metabolic study protocols in which all food is provided by the researchers, amounts consumed are measured accurately, and the nutrient composition of the food is determined by reliable and valid laboratory analyses. Such protocols are usually possible with only a few subjects. Thus, in many studies, intake data are self-reported (e.g., through 24-hour recalls of food intake, diet records, or food frequency questionnaires).

Potential sources of error in self-reported intake data include over-or underreporting of portion sizes and frequency of intake, omis-