Reproducibility and validity of major dietary patterns among Swedish women assessed with a food-frequency questionnaire

Abstract

Defining dietary patterns by factor analysis is an alternative approach to dietary assessment that has been used recently to examine diet-disease relations. However, only 1 study evaluated the reproducibility and validity of this method. Our aim was to assess both the validity and reproducibility of major dietary patterns based on data from a 60-item FFQ. We chose 2 independent random samples among over 60,000 women aged 40-74 y participating in the Swedish Mammography Cohort (SMC). In the validation study, the FFQ was compared with 4 7-d dietary records (DRs) among 129 women. For the reproducibility study, the FFQ was administered twice, 1 y apart in 212 women. By conducting factor analysis, 3 major dietary patterns were identified: healthy (high in vegetables, fruits, fish, poultry, tomato, cereal, and low-fat dairy products), Western (processed meat, meat, refined grains, sweets, and fried potatoes), and drinker (beer, wine and liquor, snacks) pattern. These 3 patterns explained 29-34% of the total variance in these 2 studies. The Spearman correlation coefficients between FFQ1 and FFQ2 (reproducibility) for healthy, Western, and drinker pattern were 0.63, 0.68, and 0.73, respectively (all P < 0.0001). Correlation coefficients between the FFQ and DRs (validity) for these patterns were 0.59, 0.50, and 0.85, respectively (all P < 0.0001). Our results indicate that identification of dietary patterns through factor analysis is a reproducible and valid method. The dietary patterns approach might be used in nutritional epidemiology as an alternative method of dietary assessment.