Using a Correction Factor to Correct for Overreporting in a Food-Frequency Questionnaire Does Not Improve Biomarker-Assessed Validity of Estimates for Fruit and Vegetable Consumption

Abstract

To correct for overreporting of fruit and vegetable (FV) consumption in a food-frequency questionnaire, summary questions about consumption of main FV groups are often used to calculate correction factors. This study compared the ability to rank people according to their FV intake of those summary questions and the sum of questions on individual FV items within categories, and of corrected or uncorrected estimates of specific sorts of FV. Healthy middle-age women (n = 161) completed a food-frequency questionnaire about FV consumption during the previous month and gave a single fasting blood sample. Correction factors were calculated as the reported frequency on a summary question divided by the summed frequencies of all items in a category. Plasma carotenoids and vitamin C served as biomarkers of FV consumption. Significant correlations between FV consumption and biomarkers were observed (e.g., Spearman's correlation coefficient r with total carotenoids/vitamin C: 0.32/0.34 for vegetables, 0.30/0.25 for fruits). Summary estimates of cooked, raw and total vegetable consumption correlated higher with biomarkers than sum estimates. For fruits no differences in correlations between sum and summary estimates were observed. Applying a correction factor on the consumption of carrots and total cabbage resulted in lower correlations with relevant biomarkers. For broccoli/cauliflower, Brussels sprouts and citrus fruits, correlations with biomarkers did not change after correction. We conclude that summary questions may suffice to rank individuals according to their intake of main FV categories, and that correction for overreporting of individual FV items is probably not advisable when ranking individuals according to intake of these items.