Assessment of the reliability of a diet history questionnaire

The purpose of this study was to estimate the reliability of a self-administered food frequency questionnaire developed at the National Cancer Institute (NCI) (Bethesda, MD). Food intake information was collected from 48 women with breast cancer and 50 women who had a familial breast cancer risk. These women were participants in a long-term investigation of diet and breast cancer at the Ohio State University (Columbus, OH). Pearson correlation coefficients were obtained for 29 nutrients. The effect of time between questionnaires, age, weight, and caloric intake was also examined. The results showed moderate reliability coefficients for most of the nutrients (r = 0.5-0.8). For the majority of nutrients, the women with breast cancer had higher coefficients than did the high-risk women. High-risk women with shorter elapsed time periods between the first and second questionnaire had higher reliability coefficients as did women who consumed fewer calories. Age and weight (standardized by height) showed no effect on reliability. Recommendations for improving the NCI questionnaire in addition to using it as a standard method of measuring food intake are discussed.