Comparison of two methods for identifying dietary patterns associated with obesity in preschool children: the GENESIS...

The aim of this work was to identify dietary patterns based on reduced rank regression (RRR) and principal component analysis (PCA) and to evaluate the association of these patterns with the prevalence of childhood obesity.

A sample of 2317 toddlers and preschoolers from Greece (Growth, Exercise and Nutrition Epidemiological Study In preSchoolers) was used. In total, 12 food groups were used as predictors of RRR and PCA. Nutrients such as total fat, simple carbohydrate and fiber intake were used as response variables to apply RRR.

One factor/pattern was retained from RRR and PCA in order to ensure the comparability of the methods. The pattern derived from PCA was mainly characterized by consumption of fruits, vegetables, legumes, fish and seafood, grains and oils. This pattern explained 12.5% of the total variation in food groups. On the other hand, the pattern extracted from RRR was mainly characterized by reduced consumption of fruits, vegetables and legumes, and by increased consumption of sweets and red meat. The pattern derived from RRR explained 8.2% of the total variation in food groups. Simple and multiple logistic regression revealed that the pattern extracted from RRR is significantly associated with the prevalence of childhood obesity (OR=1.11, 95% CI: 1.00-1.28 for each unit increase of dietary pattern) as opposed to the pattern derived from PCA.

The preferable technique to derive dietary patterns related to childhood obesity seems to be RRR compared with PCA.