Relative Validity of a Food Frequency Questionnaire Used to Assess Food Intake During a Dietary Intervention Study

Abstract

To develop a cost-effective alternative for evaluating dietary intake in large-scale intervention trials of cancer and cardiovascular disease outcomes, we designed and validated a semiquantitative food frequency questionnaire (FFQ). We collected 6 to 8 of the 24-hr dietary recalls from 87 adults (ages 30–72 yr) who were randomly assigned to a walnut-supplemented diet or a control diet in a 6-mo dietary intervention trial. Relative validity of a 171-item FFQ in assessing intake of selected foods and the prescribed intervention (intake ≥25 g/day or intake < 2 g of walnuts) was determined using 24-h dietary recalls as the reference. De-attenuated correlations between FFQ and dietary recalls were .82 for walnuts, .80 for fruits, .79 for grains, .77 for vegetables, .63 for water, .44 for sweets, and .36 for dairy/eggs. High within-person variation did not allow de-attenuation for the remaining foods, but uncorrected correlations were high (> .7) for the beverage variables. The FFQ correctly classified 86 out of 87 subjects in the 2 prescribed intervention groups. The FFQ can provide an accurate measure of a food-based intervention (i.e., walnut supplementation) in a trial setting and can also accurately estimate a number of other food groups consumed during the trial.

ACKNOWLEDGMENTS

We acknowledge Karen Jaceldo-Siegl, DrPH, for her invaluable help in the development of the FFQ and the research nutritionists who collected the dietary recalls. Gina Siapco was the recipient of a postdoctoral research fellowship from the California Walnut Commission for conducting this research. Funded by the California Walnut Commission.

Notes

^a Abbreviation is as follows: BMI, body mass index.

^b N = 87.

^c n = 48, walnut intake ≥25 g/day.

^d n = 39, walnut intake < 2 g/day.

^a n = 87.

^b Food grouping adapted from the Nutrition Coordinating Center Food Grouping Scheme.

^c Significantly different from dietary recall estimates at P < 0.005, paired t-test.

^d Significantly different from dietary recall estimates at P < 0.0001, paired t-test.

^a n = 87.

^b P < 0.05 for the correlation.

^c Corrected (de-attenuated) correlation could not be computed due to high within-subjects variance for this food.

^a n = 48, kappa = .99.

^b Prescribed intervention is ≥25 g of walnuts for those in the walnut-supplemented diet group and < 2 g of walnuts in the control diet group.

^c n = number of participants.

^d n = 39, kappa = 1.