Differences in precision of dietary estimates among different population subgroups

Populations vary in consistency of diets. The precision of estimates (E) of nutrient intake depends both on the ratio (R) of intraindividual to interindividual variances, as well as on the number of days of dietary intake recorded (k). We present baseline data from 872 participants in the Trial of Antihypertensive Interventions and Management (TAIM), a randomized clinical trial of diet and drug treatment for mild hypertension. There is wide variation in R among different demographic groups and for different nutrients. This has implications for the ability to estimate the correlations between the nutrient and a physiologic variable such as blood pressure, as well as for power calculations to compare group means on nutrient intake. For example, to achieve the same degree of precision (E = 0.80) in estimating the correlation between sodium intake and blood pressure for whites in New York, 7 days of food records would be required, compared to 2 days for whites in Alabama. Tables are presented indicating within- and between-person variances for different groups and different nutrients. Methods of calculating E, determining k, and calculating samples size are provided. Investigators designing nutritional epidemiologic studies should take into account ratios of intraindividual to interindividual variances of different population subgroups and different nutrients in order to determine the number of daily food records that must be obtained and the sample size for group comparisons.