Authors: Dekkers AL, Verkaik-Kloosterman J, van Rossum CT, Ocké MC
Citation: J. Nutr. 2014 Dec;144(12):2083-91
PMID : 25320187, Journal: J. Nutr., 144, 12
Date created: 2014-11-21
BACKGROUND: For the evaluation of both the adequacy of intakes and the risk of excessive intakes of micronutrients, all potential sources should be included. In addition to micronutrients naturally present in foods, micronutrients can also be derived from fortified foods and dietary supplements. In the estimation of the habitual intake, this may cause specific challenges such as multimodal distributions and heterogeneous variances between the sources.
OBJECTIVE: We present the Statistical Program to Assess Dietary Exposure (SPADE) that was developed to cope with these challenges in one single program.
METHOD: Similar to other methods, SPADE can model habitual intake of daily and episodically consumed dietary components. In addition, SPADE has the option to model habitual intake from dietary supplements. Moreover, SPADE offers models to estimate habitual intake distributions from different sources (e.g., foods and dietary supplements) separately and adds these habitual intakes to get the overall habitual intake distribution. The habitual intake distribution is modeled as a function of age, and this distribution can directly be compared with cutoff values to estimate the proportion above or below. Uncertainty in the habitual intake distribution and in the proportion below or above a cutoff value is quantified with ready-for-use bootstrap and provides 95% CIs.
RESULTS: SPADE is implemented in R and is freely available as an R package called SPADE.RIVM. The various features of SPADE are illustrated by the estimation of the habitual intake distribution of folate and folic acid for women by using data from the Dutch National Food Consumption Survey 2007-2010. The results correspond well with the results of existing programs.
CONCLUSION: SPADE offers new features to existing programs to estimate the habitual intake distribution because it can handle many different types of modeling with the first-shrink-then-add approach.