Reproducibility and biomarker-based validity and responsiveness of a food frequency questionnaire to estimate protein intake.

Abstract

This study, conducted in the Netherlands in 1994-1996, assessed the reproducibility, validity, and responsiveness of a food frequency questionnaire (FFQ) used with older subjects with type 2 diabetes. During a period of 6 months, 93 subjects had a stable protein intake according to their urinary urea excretion (UUE). An FFQ was completed before and after this period. Reproducibility was evaluated on the basis of the correlation (0.80) between duplicate FFQ estimates of protein intake. At the group level, the smallest detectable difference was 3 g. For 354 subjects, average protein intake according to the FFQ was similar to that from UUE estimates. Validity was evaluated on the basis of the crude and deattenuated correlations, which were 0.31 and 0.39, respectively. The crude correlation was highest for former smokers (r = 0.39) and for patients with a higher body mass index (r = 0.37), and it was very low for current smokers (r = 0.14, p > 0.20). The correlation did not vary with age or gender. Responsiveness was judged on the crude correlation between changes in intake according to the FFQ and UUE; the correlation was 0.22 (n = 134). The responsiveness ratio was 0.73. The authors concluded that FFQ estimates are well reproducible. At the group level, protein intake was estimated correctly, whereas individual protein intake and changes in intake were reflected to a moderate extent only.