Comparison of dietary assessment methods in a low-fat dietary intervention program

The food frequency questionnaire (FFQ) is commonly utilized for assessment of dietary fat intake, but its validity among individuals following a low-fat diet is unclear. We evaluated the agreement of nutrient estimates derived from FFQ, 24-h recall, and 3-day food records obtained from 104 participants in a randomized trial of a low-fat dietary intervention for women at elevated breast cancer risk. Comparisons were made for total calories, percent calories from fat, and total fat after 1 yr. Correlation was assessed using standard methods based on a null hypothesis of no agreement between instruments as well as by a methodology based on a null hypothesis that the instruments should be in agreement. With the use of standard methods, FFQ estimates for women on the low-fat diet were significantly correlated to records only for percent calories from fat (r = 0.39), whereas recall and record estimates were significantly correlated for all three dietary variables. Using the new method, we found no significant correlation between FFQ and either recalls or records for women following a low-fat diet but significant correlation between recall and record estimates for total calories (r = 0.67). Traditional correlation testing may overestimate the extent of agreement in dietary instruments among women on a low-fat diet. We found empirical support for the nontraditional method.