ABSTRACT
Epidemiologic studies based on quantitative estimation of food and nutrient intake have been limited in Japan. We evaluated validity and reproducibility of a personal computer (PC)-assisted dietary interview used in the Fukuoka Colorectal Cancer Study. The subjects were 28 participants as controls in a population-based case-control study. Four 7-day diet records were kept during a period from July 2001 to May 2002, and the PC-assisted dietary interview was administered before and after the diet records (May to July 2001 and July to September 2002). Intakes of total energy, 26 nutrients, and 19 food groups were estimated. Of 26 Pearson correlation coefficients for energy-adjusted nutrients, 18 for the first interview and 19 for the second interview were 0.40 or greater, with the highest correlation for saturated fat in the first interview (0.72) and vitamin C in the second interview (0.60). Validity scores were fairly high for most of the food groups with an exceptionally low correlation for potatoes (0.19 for the first and 0.13 for the second interview). Reproducibility was good for most nutrients and foods; but low correlation was noted for Vitamin D and nuts/seeds. Reproducibility was modest for fish and fish products, beef and pork, and processed meat. The PC-assisted dietary interview was fairly valid and reproducible regarding most of the nutrients and food groups. Validity and reproducibility of meat and fish intakes were probably attenuated because of the episode of bovine spongiform encephalopathy in September 2001.