Errors in reported dietary intakes. I. Short-term recall.

Abstract

Measures for quantifying reproducibility and between-subject variability of nutrient intake data are applied to intakes reported in two interviews (1-3 months apart) by 44 male and 17 female healthy white subjects aged 45-75 years. Intakes were assessed by three methods: a dietary history that included consumption frequency and serving size for 71 food items (dietary history method); a combination of individual consumption frequencies with sex-specific mean serving sizes (frequency method); an extrapolation from frequencies and serving sizes of all foods reported for a "typical day" in the specified time period (typical day method). Intake variation within subjects, between subjects, and between methods was assessed by analysis of variance for each sex and for each of the nutrients: total calories, protein, fat, vitamin A, and protein and fat as percentage of total calories. Dietary history-assessed intakes exceeded those assessed by the other two methods. The dietary history versus frequency excess was greater than the dietary history versus typical day excess for calories, fat, and protein, while the reverse was true for vitamin A and fat as percentage of total calories. The typical day method was unreliable for vitamin A because it occasionally produced extremely high, unreproducible intakes. The intraclass correlation coefficient was used to measure a method's ability to distinguish interpersonal variation from within-person error. The frequency method produced less within-person error than did the dietary history method for all nutrients. For absolute intakes, the frequency method produced less interpersonal variation than did the dietary history method, while for relative intakes, the reverse was true. Females reported intakes with less within-person error than did males, and the interpersonal spread of their intakes was smaller. Consequences of these findings for the power and sensitivity of studies on the role of dietary factors in the etiology of chronic disease are explored.