Validity of nutrient estimates by food frequency questionnaires based either on exact frequencies or categories

PURPOSE: The study was designed to assess if category-based Food Frequency Questionnaires (FFQ) could be substituted for exact-frequency based questionnaires. METHODS: Nutrient estimates on 203 subjects from Toronto, Canada, derived from recoding an exact frequency based 132 item food frequency questionnaire (E-FFQ) into categories were evaluated for validity against a 7-day food record (FR). RESULTS: Among men, the median Pearson's correlation coefficient (for energy adjusted nutrients) between FR and E-FFQ was 0.53, ranging from 0.26 for thiamin to 0.72 for calcium. The median correlation dropped to a value of 0.40 (range: 0.18 for thiamin to 0.56 for carbohydrate) when a category-based FFQ (C-FFQ) and FR were compared. Data on women showed a similar pattern; a median correlation of 0.50 from E-FFQ and 0.42 from C-FFQ. CONCLUSIONS: While the results varied with individual nutrients, overall it is concluded that when validated against food records, the use of exact frequencies in FFQs yielded higher correlation coefficients for most nutrients, and slightly better agreement within quartiles of categories than recoded frequency categories. Ann Epidemiol 2000;10:354-360. (C) 2000 Elsevier Science Inc. All rights reserved.