Background: Assessment of habitual diet is important in investigations of diet-disease relationships. Many epidemiological studies use the food frequency questionnaire (FFQ) to evaluate dietary intakes but few studies validate the instrument against biological markers. The aim of this study was to assess the validity and reproducibility of a previously validated 70-item food frequency questionnaire (FFQ) that was expanded to 120-items to assess diet-cancer relations.

Methods: Relative validity of the FFQ was assessed against twelve 24-hour recalls administered over 12 months in 70 subjects. The FFQ was repeated after one year (FFQ2) to assess reproducibility. The validity of the FFQ was evaluated by comparing nutrient and food group intakes from 24-hour recalls with the first and second FFQ. In addition, FFQ validity for cholesterol and folate were determined through correlation with biomarkers (serum cholesterol, serum folate and whole blood folate) in 159 control subjects participating in a case-control prostate cancer study.

Results: Compared to recalls the FFQ tended to overestimate energy and carbohydrate intakes but gave no differences in intake for protein and fat. Quartile agreement for energy-adjusted nutrient intakes between FFQ2 and recalls ranged from 31.8%-77.3% for the lowest quartile and 20.8%-81.0% in the highest quartile. Gross misclassification of nutrients was low with the exceptions of protein, vitamin E and retinol and weighted kappa values ranged from 0.33 to 0.64 for other nutrients. Validity correlations for energy-adjusted nutrients (excluding retinol) were moderate to high (0.38-0.86). Correlation coefficients between multiple recalls and FFQ1 ranged from 0.27 (fruits) to 0.55 (red meat); the second FFQ gave somewhat higher coefficients (0.30 to 0.61). Reproducibility correlations for the nutrients ranged from 0.50 to 0.84.Calibration of the FFQ with biochemical markers showed modest correlations with serum cholesterol (0.24), serum folate (0.25) and whole blood folate (0.33) adjusted for age, energy, body mass index and smoking.

Conclusions: The expanded FFQ had good relative validity for estimating food group and nutrient intakes (except retinol and vitamin E) and was a reliable measure of habitual intake. Associations with biomarkers were comparable to other studies.