Validation of a Quantitative Food Frequency Questionnaire for a Japanese Population in Hawaii

Objective: To measure the validity of a quantitative food frequency questionnaire (QFFQ). Design: A cross-sectional validation study of the QFFQ against a four-day food record (4DR) using Spearman correlation, cross-classification, kappa statistics, and Bland-Altman plotting. Setting: The Gastroenterology Department of Kaiser Permanente Hawaii. Subjects: 76 healthy Japanese American men and women, aged 40-75 years. Results: Somewhat stronger average correlations were observed between the QFFQ and the 4DR for macronutrients compared to micronutrients (Spearman rho of 0.47 vs. 0.35). Moderate correlations between the two tools were observed for macronutrients (including saturated fatty acids and dietary fibre), iron, β-carotene, vitamin C, and ethanol (rho: 0.38-0.58). Overall, stronger correlations were found among men than women between the two tools (mean rho 0.41 vs. 0.26). In a cross classification analysis, for more than 75% of the observations, a complete to relative agreement between the two methods was observed for fat, α-carotene, folate, vitamin D, and ethanol. Sex difference in agreement was minimal in cross-classification (overall extreme misclassification of 9.80% for men and 12.40% for women). Bland-Altman plots showed over-estimations of dietary fibre and α-carotene intake and an under-estimation of cholesterol intake by the QFFQ at high levels of consumption. However, the QFFQ estimation for fat, dietary fibre, folate, cholesterol, α-carotene, vitamins D and C, and ethanol intake was less than 7% different compared to the 4DR. Conclusions: The QFFQ has an adequate validity for fat, folate, vitamin D, and ethanol and can correctly categorize participants for intakes of cholesterol, dietary fibre, α-carotene, and zinc.