Nutrition research has traditionally focused on single nutrients in relation to health. However, recent appreciation of the complex synergistic interactions among nutrients and other food constituents has led to a growing interest in total dietary patterns. Methods of measurement include summation of food or nutrient recommendations met, such as the United States Department of Agriculture Healthy Eating Index; data-driven approaches — principal components (PCA) and cluster analyses — which describe actual intake patterns in the population; and, most recently, reduced rank regression, which defines linear combinations of food intakes that maximally explain intermediate markers of disease. PCA, a form of factor analysis, derives linear combinations of foods based on their intercorrelations. Cluster analysis groups individuals into maximally differing eating patterns. These approaches have now been used in diverse populations with good reproducibility. In contrast, because it is based on associations with outcomes rather than on coherent behavioral patterns, reduced rank regression may be less reproducible, but more research is needed. However, it is likely to yield useful information for hypothesis generation. Together, the focus on dietary patterns has been fruitful in demonstrating the powerful protective associations of healthy or prudent dietary patterns, and the higher risk associations of Western or meat and refined grains patterns. The field, however, has not fully addressed the effects of diet in subpopulations, including ethnic minorities. Depending on food group coding, subdietary patterns may be obscured or artificially separated, leading to potentially misleading results. Further attention to the definition of the dietary patterns of different populations is critical to providing meaningful results. Still, dietary pattern research has great potential for use in nutrition policy, particularly as it demonstrates the importance of total diet in health promotion.
