fewer than about 50 or 60 individuals result in unreliable estimates of usual intake distributions (Nusser et al., 1996). Because only the replicate observations in the sample contain information about the day-to-day variability in intakes, it is important to have a moderately large number of individuals in the replicate sample, perhaps not fewer than 30 or 40, and these individuals should be representative of the full group. Each person in this sample must have at least two independent daily intake measurements or three daily intake measurements if data are collected on consecutive days.
Carriquiry and colleagues (1997) successfully applied the ISU method to adjust intake distributions and distributions of blood biochemical measurements using data collected in the Third National Health and Nutrition Examination Survey (NHANES III), even though sample sizes for some life stage and gender groups were moderately small (fewer than 70 to 80 individuals) and the proportion of replicate observations was low (approximately 6 percent). In general however, having a minimum number of replicate records in the sample is more important than having a minimum proportion of replicate observations.
The following example is based on estimated usual intake distributions for two dietary components—phosphorus and vitamin B6—for women aged 19 through 50 years who were neither pregnant nor lactating at the time the data were collected. Only intakes from food were considered (i.e., intake from supplements is not included in these examples). The dietary intake data were collected in NHANES III, so only a small proportion of individuals in the sample had a replicate observation collected several weeks after the first. Estimated Average Requirements (EARs) have been established for the two nutrients in this example (IOM, 1997, 1998b). Using the EAR cut-point method, the proportion of women at risk of nutrient inadequacy can be estimated by computing the percentage of individuals in the group with usual intakes below the corresponding EAR.
For purposes of illustration, the usual intake distributions of phosphorus and vitamin B6 were estimated by two different approaches: (1) using only the first day of intake data for each individual in the sample; and (2) using replicate intake data (whenever available) and applying the ISU method to adjust the distribution. It is anticipated that the estimate of the usual intake distribution obtained using one day of intake data will have the incorrect variance; the variance of the estimated distribution will contain an unwanted day-to-day variability component. Therefore, estimates of the prevalence of nutrient inadequacy will be biased. The two estimates of the usual