The reproducibility of dietary intake data in a prospective study of gastrointestinal cancer.

To test the reproducibility of the dietary questionnaire used in the Japan-Hawaii Cancer Study and to measure the consistency of dietary patterns over time, 109 men were reinterviewed with the same questionnaire after a 6-month interval and 111 men were reinterviewed after a 2-year interval. The responses for intake of the 33 food items in the questionnaire were fairly reproducible. Food items with substantial correlation coefficients in both groups of men tended to be consumed on a more habitual basis. This finding suggests that if habitually consumed foods are linked to specific cancers, it is less likely that such associations occur by chance. This is of particular importance because when many food items are considered in a dietary study, chance association between a specific food and cancer is not an infrequent event.