Reliability and validity of a self-administered food frequency questionnaire in a chemoprevention trial of adenoma recurrence.

Various chemoprevention trials have assessed dietary intake by means of food frequency questionnaires. However, it is important to assess the degree to which such questionnaires can measure diet. We conducted reproducibility and validity analyses of our Arizona Food Frequency Questionnaire (AFFQ) in our recently completed, randomized, Phase III chemoprevention trial testing the effects of a wheat bran fiber supplement on colorectal adenoma recurrence. A total of 139 individuals provided a baseline and year 1 AFFQ and a set of 4-day dietary records collected over a period of 1 month. The reproducibility analyses of the AFFQ administered 1 year apart showed a mean intraclass correlation of 0.54 for unadjusted nutrients and 0.48 for energy-adjusted nutrients. The relative validity of the AFFQ, as compared with the average of the 4-day diet records, showed a mean deattenuated correlation of 0.49 (range, 0.22-0.65) for the baseline AFFQ and 0.49 (range, 0.25-0.67) for the year 1 AFFQ. When data from both AFFQs were combined and compared with the diet records, there was a slight improvement in the overall deattenuated correlations (mean, 0.56; range, 0.33-0.71). The correlations we observed for macro- and micronutrient intake were within the overall range of those reported in the literature. Reliability and validity studies of dietary instruments are feasible in the setting of a chemoprevention trial and should be conducted when the instrument's performance has not been previously assessed in the target population.