Dietary patterns derived from principal component- and k-means cluster analysis: Long-term association with coronary heart disease and stroke

• DOI: 10.1016/j.numecd.2012.02.006
• Publication Date: 2012-05-28T15:03:03Z
• Authors: M.D. Stricker; N.C. Onland-Moret; J.M.A. Boer; Y.T. van der Schouw; W.M.M. Verschuren; A.M. May; P.H.M. Peeters; J.W.J. Beulens
• Keywords: Stroke; Association (psychology)

Studies comparing dietary patterns derived from different a posteriori methods in view of predicting disease risk are scarce. We aimed to explore differences between dietary patterns derived from principal component- (PCA) and k-means cluster analysis (KCA) in relation to their food group composition and ability to predict CHD and stroke risk.The study was conducted in the EPIC-NL cohort that consists of 40,011 men and women. Baseline dietary intake was measured using a validated food-frequency questionnaire. Food items were consolidated into 31 food groups. Occurrence of CHD and stroke was assessed through linkage with registries. After 13 years of follow-up, 1,843 CHD and 588 stroke cases were documented. Both PCA and KCA extracted a prudent pattern (high intakes of fish, high-fiber products, raw vegetables, wine) and a western pattern (high consumption of French fries, fast food, low-fiber products, other alcoholic drinks, soft drinks with sugar) with small variation between components and clusters. The prudent component was associated with a reduced risk of CHD (HR for extreme quartiles: 0.87; 95%-CI: 0.75-1.00) and stroke (0.68; 0.53-0.88). The western component was not related to any outcome. The prudent cluster was related with a lower risk of CHD (0.91; 0.82-1.00) and stroke (0.79; 0.67-0.94) compared to the western cluster.PCA and KCA found similar underlying patterns with comparable associations with CHD and stroke risk. A prudent pattern reduced the risk of CHD and stroke.