Dietary pattern analysis, which reflects the complexity of dietary intake, has received considerable attention by nutritional epidemiology. For a long time, two general approaches have been used to define these summary variables in observational studies. The exploratory approach is based only on the data of the study, whereas the hypothesis-oriented approach constructs pattern variables based on scientific evidence available before the study. Recently, a new statistical method, reduced rank regression, was applied to nutritional epidemiology that is exploratory by nature, but can use scientific evidence by focusing on disease-related dietary components or biomarkers. Several studies, both observational and clinical, suggest that dietary patterns may predict the risk of CHD and stroke. In the present review, we describe the results of these studies and the available evidence regarding the relationships between dietary patterns and risk of CVD and we discuss limitations and strengths of the statistical methods used to extract dietary patterns.