Within- and between-person variations in portion sizes of foods consumed by the Japanese population.

There has been considerable controversy about the utility of asking separate questions on portion sizes in addition to items on consumption frequencies in food frequency questionnaires. One argument against this type of inquiry is that, for most foods, within-person variation of portion size is greater than between-person variation, making it difficult to characterize "usual" or long-term average serving sizes. The authors studied the relative effects of within- and between-person variances in portion sizes of 69 food items using three-day food record data collected from 335 Japanese men and women. Total variance in the portion size for each food was partitioned into within- and between-person components by analysis of variance. Although the ratios of within- to between-person variances in log(e)-transformed portion sizes were > 1.0 for the majority of foods (50 of 69 items, median = 1.7), they were smaller than or close to unity for subgroups of foods that were the major sources of nutrient variables of relevance to diet-cancer associations, such as total energy (rice, calcium (milk), sodium (pickled vegetables), and alcohol (beer, shochu, whiskey). These results indicate that the relative contributions of within- and between-person variations in portion size may vary among foods, and therefore investigators should consider the balance between the advantages and disadvantages of obtaining portion size data separately from intake frequencies when designing food frequency questionnaires.