The validity of a food frequency questionnaire (FFQ) was evaluated against four 7-d diet records among 2 3 men aged 27-70 y. Volunteer women recorded the amounts of all foods and beverages consumed by their husbands or fathers over a period of 7 d in four consecutive seasons in 1997 and 19 9 8, and a self-administered FFO was answered by each man after completion of the fourth diet record. The diet record was kept in accordance with the method used for Japan's National Nutrition Survey. The FFO ascertained frequencies and amounts of consumption of rice, bread, noodles, green tea, and coffee; consumption frequency of 35 food items; and consumption frequency and amount of five alcoholic beverages. As for most nutrients and foods, mean intakes estimated by the FFQ were lower than those estimated from the diet record. Crude intake of nutrients showed fairly good agreement between the two methods in terms of Pearson's correlation coefficient (r>0.5), except for total energy (r=0.23), fat (r=0.36), and vitamin A (r=0.29). The adjustment for energy intake generally reduced the correlation; the reduction being fairly large for fat intake and vitamin A (r=0.19 for both). There was generally good agreement for the intake of foods and beverages with exceptions for fats/oils (Spearman correlation coefficient, r(s)=0.30) and other vegetables (r(s)=0.35). The highest correlation was observed for alcoholic beverages (r(s)=0.91), bread (r(s)=0.80), and fruits (r(s)=0.77). An FFO covering a limited number of common foods may be useful in assessing the relative position of an individual's habitual consumption of foods and nutrients; however, the instrument generally underestimates absolute intake.
