Nutritional epidemiology: forest, trees and leaves

Ioannidis has stated that the field of nutritional epidemiology has generated confusion and numerous implausible findings and is in need of radical reform. One of the reforms he proposes is to conduct analyses that take into account the "totality for all nutritional factors measured". This approach is based on isolating and reducing diet into numerous independent variables with little regard to prior knowledge or the interrelations among dietary components, and relying on a "discovery" approach. This method, akin to genomewide association studies (GWAS), would involve very large sample sizes, small associations, no prior knowledge, and multiple testing considerations. This method is contrary to the more traditional hypothesis generating and testing approach built on all types of evidence. This commentary will contrast how suitable these two approaches are to study diet and disease.