A dietary method for an epidemiologic study of gastrointestinal cancer.

The validity of a dietary recall questionnaire was tested among 50 men of Japanese ancestry, participating in a prospective epidemiologic study of gastrointestinal cancer in Honolulu. The questionnaire included 33 food items consumed in Hawaii and suggested as possibly carcinogenic or possibly protective. Estimated intakes of frequencies and amounts during the previous day and previous 7 days were obtained. The validity study consisted of a comparison of a 7 day diary of the frequencies and amounts of the 33 items with a subsequent recall of the same items. The sensitivity (proportion of men recording an item who subsequently recalled it) was better for 24 of the 33 items during the 7-day period than the 1-day period. The overall agreement between the two methods on whether the item was eaten or not was only marginally less for the 7 days than for the 1 day. Estimated intakes tended to be higher for the diary than the recall method. The Spearman rank correlation coefficients among persons recording the items generally were good (greater than 0.7) for items eaten habitually and for those associated with specific events. Further, rank order correlations for quantitative recell of particular food items has considerable validity and is a reasonable choice of method for studies of large groups concerning the role of particular food items in the etiology of cancer.