Objective: To validate the use of cluster analysis for characterizing population dietary patterns.
Design: Cluster analysis was applied to a food frequency questionnaire to define dietary patterns. Independent estimates of nutrient intake were derived from 3-day food records. Heart disease risk factors were assessed using standardized protocols in a clinic setting.
Setting: Adult women (n = 1,828) participating in the Framingham Offspring-Spouse study.
Statistical analyses: Age-adjusted mean nutrient intakes were determined for each cluster. Analysis of covariance was used to evaluate pairwise differences in intake across clusters. Compliance with published recommendations was determined for selected heart disease risk factors. Differences in age-adjusted compliance across clusters were evaluated using logistic regression.
Results: Cluster analysis identified 5 distinct dietary patterns characterized by unique food behaviors and significantly different nutrient intake profiles. Patterns rich in fruits, vegetables, grains, low-fat dairy, and lean protein foods resulted in higher nutrient density. Patterns rich in fatty foods, added fats, desserts, and sweets were less nutrient-dense. Women who consumed an Empty Calorie pattern were less likely to achieve compliance with clinical risk factor guidelines in contrast to most other groups of women.
Conclusions: Cluster analysis is a valid tool for evaluating nutrition risk by considering overall patterns and food behaviors. This is important because dietary patterns appear to be linked with other health-related behaviors that confer risk for chronic disease. Therefore, insight into dietary behaviors of distinct clusters within a population can help to design intervention strategies for prevention and management of chronic health conditions including obesity and cardiovascular disease.