RESUMO
BACKGROUND: To examine the associations of dietary patterns with odds of elevated Blood Pressure (BP) among Lebanese adult males using principal component analysis (PCA), and compare the results to two other data reduction methods, including reduced rank regression (RRR) and partial least-squares (PLS) regression. METHODS: Data from the National Nutrition and Non-Communicable Disease Risk Factor Survey conducted in Lebanon between years 2008 and 2009 were used. Dietary intake data were collected by a 61-item food frequency questionnaire (FFQ). In addition, anthropometric and blood pressure measurements were obtained following standard techniques. For the purpose of this study, data of males older than 20 years with no history of chronic diseases were selected (n = 673). Elevated BP was indicated if the systolic blood pressure was > = 130mm Hg and/or the diastolic blood pressure > = 85 mm Hg. Dietary patterns were constructed using PCA, PLS and RRR and compared based on the performance to identify plausible patterns associated with elevated BP. For PLS and RR, the response variables were BMI, waist circumference and percent body fat. Multiple logistic regression was used to evaluate the associations between the dietary pattern scores of each method and risk of elevated BP. RESULTS: Three dietary patterns were identified using PCA: Western, Traditional Lebanese, and Fish and alcohol. Both the Western and the Traditional Lebanese patterns were associated with higher odds of elevated BP in the study population (OR = 1.23, CI 1.03, 1.46; OR = 1.29, CI 1.09, 1.52 respectively). The comparison among the three methods for dietary patterns derivation showed that PLS and RRR derived patterns explained greater variance in the outcome (PCA: 1.2%; PLS: 14.1%; RRR: 15.36%) and were significantly associated with elevated BP, while the PCA dietary patterns were descriptive of the study population's real dietary habits (PCA: 23.6%; PLS: 19.8%; RRR: 11.3%). CONCLUSIONS: The Western and Traditional Lebanese dietary patterns were associated with higher odds of elevated BP among Lebanese males. The findings of this study showed that, compared to PCA, the use of RRR method resulted in more significant associations with the outcome while the PCA-derived patterns were more related to the real habits in the study population.
