REPRODUCIBILITY AND RELATIVE VALIDITY OF FOOD GROUP INTAKE IN A FOOD FREQUENCY QUESTIONNAIRE DEVELOPED FOR THE GERMAN PART OF THE EPIC PROJECT

Background. For the German part of the European Prospective Investigat ion into Cancer and Nutrition (EPIC), a self-administered, optically-r eadable food frequency questionnaire (FFQ), including 158 food items a nd 87 coloured portion size photographs, was developed to assess the u sual food and nutrient intake of individuals during the past year. In 1991/1992, the reproducibility and validity of the questionnaire measu rements were studied according to the EPIC protocol. This article repo rts the results on reproducibility and relative validity of measuremen t of food group intake. Methods. A total of 104 men and women aged 35- 64 years, who are members of the local health insurance institution, A OK Heidelberg, participated in this study. Reproducibility of the ques tionnaire measurements was obtained by a repeated administration of th e FFQ to the same study subjects at a 8-month interval. The mean of 12 24-hour dietary recalls applied at monthly intervals served as refere nce method for the estimation of the relative validity of questionnair e measurements. A second version of the FFQ that integrated questions on general food consumption patterns was also investigated. Results. S pearman test-retest correlations of food group intake ranged from 0.49 for bread to 0.89 for alcoholic beverages (median = 0.70). Spearman c orrelations between food group intake values derived from the 24-hour diet recalls and the FFQ completed in the summer of 1992 varied from 0 .14 for legumes to 0.90 for alcoholic beverages (median = 0.45). Corre ction for attenuation due to within-person error in the reference meth od as well as the correction for general consumption patterns improved the correlations. Conclusion. The results indicate that our newly dev eloped FFQ gives reproducible estimates of food group intake. Large da y-to-day variation in food group intake complicated the evaluation of FFQ validity. Overall, moderate levels of relative validity were obser ved for estimates of food group intake.