Reproducibility of recall of adolescent diet: Nurses' Health Study (United States).

Although diet and alcohol consumption at a younger age have been hypothesized to play an etiologic role in adult malignancy, few studies have addressed these relations and few data exist on the ability to measure recalled diet during adolescence. We undertook a study to determine whether the recall of diet during high school 22 to 47 years later was reproducible within a prospective cohort of women. In 1986, participants in the Nurses' Health Study (United States), aged 40 to 65 years, reported their diet during high school using a 24-item food frequency questionnaire (FFQ). On the 1988 questionnaire, the participants recorded their alcohol consumption between the ages of 18 and 22. In 1994, a random sample of 275 women were asked to record again their high school diet and alcohol consumption. The average of Spearman correlations between the two recalls of high school diet was r = 0.57 (range = 0.38 to 0.74). For alcohol intake, the correlation for the time period queried was 0.66. The correlation between the two recalls of high school diet was higher than the correlation between the first recall and current diet in 1986 (average r = 0.25). There was no effect of age on reproducibility of recall. These data suggest that recall of diet from the very distant past, during a distinct time period such as high school, is reasonably reproducible. Although further studies of validity are needed, such information may be sufficiently precise to assess the influence of remote diet in epidemiologic studies.