Validation of a Food Frequency Questionnaire to Assess Nutrient Intake for Diabetes Intervention Trial

Abstract

Background Type 2 diabetes is an important global health issue. Since individuals with type 2 diabetes (T2D) are taught to use food exchanges when measuring their macronutrient intake, food frequency questionnaires (FFQ) using this familiar method is critical in the dietary intake assessment of this group. Aim To evaluate the ability of a FFQ developed for participants of a diabetes intervention trial in estimating nutrient intake against multiple unannounced 24-hr recalls. Methods Patients with type 2 diabetes (T2D) (n= 43; 25 females and 18 males), with mean age of 61.6 (SD 13.1) years and a mean BMI of 29.0 (SD 21.0), participated in a randomized, prospective 24-week parallel-group clinical intervention trial. Dietary intake was assessed using 6 unannounced 24-hr recalls (24HDR) composed of 4 weekdays and 2 weekend days, and a self-administered semi-quantitative FFQ with 72 hard-coded items that was developed for the study. Portion sizes in the FFQ were based on the 2014 American Diabetes Association and the Academy of Nutrition and Dietetics, Choose Your Foods food lists. Wilcoxon signed ranks test, correlation analyses with energy adjustment and deattenuation, and cross-classification were used to evaluate the agreement between the FFQ and multiple 24HDR on intake of 25 nutrients. Results Compared to the 24HDR, the FFQ has significantly underestimated most of the nutrients except for total sugar, animal protein, arachiodonic acid, EPA, dietary fiber, calcium, and potassium. Deattenuated energy-adjusted correlations ranged between 0.62 (calcium) and 0.07 (iron), with weak correlations (r<0.35) for 9 nutrients, moderate correlations (0.35≤r<0.50) for 5 nutrients and strong correlations (r≥0.50) for 10 nutrients. In the cross-classification analyses, agreement within 1 quartile ranged from 66% (energy and sodium) to 91% (animal and vegetable proteins) while gross misclassifications ranged from 0% (vegetable protein) to 11% (sodium). Conclusion The FFQ has a tendency to underestimate absolute intake of most nutrients; however, it has good ranking ability. Underreporting tendency in this group of mostly overweight/obese subjects is similar to findings in other studies. The FFQ developed for the intervention trial can be considered a reasonably valid tool in assessing intake of most nutrients.