Reproducibility and validity of a food frequency questionnaire for nutrient intakes in the study areas of large-scale cohort studies in Japan.

The Japan Multi-Institutional Collaborative Cohort Study, the Yamagata Molecular Epidemiological Cohort Study, and the Tsuruoka Metabolomics Cohort Study use a 47-item food frequency questionnaire (FFQ) developed in central Japan in 2004. We applied regression analyses to estimate nutrient intakes in the FFQ. The regression equations, however, may not be so robust and may vary among areas, even in Japan. We aimed to evaluate the reproducibility and validity of the FFQ over an expanded area of Japan. Healthy volunteers aged 34-70 years from 13 areas of Japan provided 12-day weighed dietary records (WDRs) and completed two FFQs over 1 year. We evaluated reproducibility and validity by comparing the intakes of 27 nutrients between the two FFQs and the first FFQ (FFQ1) and WDRs, respectively. Spearman's rank correlation coefficients (SRs) between estimates from the FFQs and WDRs were calculated and corrected for intra-individual variation in the WDRs. Intakes of the selected nutrients estimated from the two FFQs were equivalent. The median energy-adjusted SRs between FFQ1 and the second FFQ were 0.66 for both men and women. Regarding validity adjusted for within-individual variation, energy-adjusted SRs for WDRs vs FFQ1 ranged from 0.13 (thiamin) to 0.79 (alcohol) for men, and the median was 0.35. The energy-adjusted SRs ranged from 0.20 (protein) to 0.71 (alcohol) for women, and the median was 0.43. The FFQ demonstrated high reproducibility and moderate validity, which suggests that it is appropriate to clarify associations between diet and health and/or disease among adults in Japan.