NUTRIENT DENSITY MODEL REVISITED

One of the common obstacles in the analysis of nutrient data is the hi gh correlation between nutrients (N) and total caloric intake (T) when used in a traditional multivariate model. This results in difficultie s in separating the effect of N from the effect of T when they are sim ultaneously used in a statistical model. Using computer simulations an d data from a case control study, we show that nutrient density N/T is weakly correlated with T even when N and T are highly correlated. As nutrient density has intuitive appeal to nutritional scientists and it is increasingly important to separate out the effect of specific nutr ients from the effect of total calories on various health outcomes, th e multivariate nutrient density model which includes both N/T and T sh ould be utilized more often in nutritional epidemiology.