Comparison of two models for the estimation of usual intake addressing zero consumption and non-normality

Abstract

Various models exist for estimating the usual intake distribution from dietary intake data. In this paper, we compare two of these models, the Iowa State University Foods (ISUF) model and the betabinomial-normal (BBN) model and apply them to three different datasets. Intake data are obtained by aggregating over multiple food products and are often non-normal. The ISUF and BBN model both address non-normality. While the two models have similar structures, they show some differences. The ISUF model includes an additional spline transformation for improving the normality of the intake amount distribution, while the BBN model includes the possibility of addressing covariates, such as age or sex. Our analyses showed that for two of the example datasets both models produced similar estimates of the higher percentiles of the usual intake distribution. However, for the third dataset, where the intake amount distribution appear to be multimodal, both models produced different percentile estimates.

A comparison by simulation of different methods to estimate the usual intake distribution for episodically consumed foods

“…The results of both models are combined to estimate the usual intake distribution. This model has been part of MCRA from 2005 onwards and is fully described in de Boer et al (2009). A similar model is employed by Slob (2006), although Slob relates the parameter of the beta distribution to covariates rather than the mean probability / as is more usually done.…”

Oral intake of added titanium dioxide and its nanofraction from food products, food supplements and toothpaste by the Dutch population

“…Furthermore, to determine long-term intake, ideally, statistical models should be used that correct the variation in long-term intake between individuals for the within individual (between days) variation (Hoffmann et al, 2002;Nusser et al, 1996;Slob, 1993). An important prerequisite for this is that the logarithmically transformed daily intake distribution is normally distributed (de Boer et al, 2009). Since the intake data were not normally distributed for TiO 2 (not shown), the observed individual means (OIM) method was used.…”

Long-term dietary exposure to lead in young European children: comparing a pan-European approach with a national exposure assessment

“…For this the positive daily exposure distribution was logarithmically transformed in a normal distribution, an important prerequisite to use the BBN model for estimating long-term exposure which should always be checked. For this we used the normal quantile-quantile (q-q) plot as proposed by de Boer et al (2009). In those cases in which the transformed positive daily exposure distribution is markedly non-normal, the results may be misleading.…”

“…The BBN approach models separately the contaminant intake frequency and positive intake amounts as a function of age to produce the long-term exposure distribution. For this, the positive amounts distribution is logarithmically transformed into a normal distribution to remove the within-person variation that is of no interest for this type of exposure (Hoffmann et al 2002;de Boer et al 2009). After removal of the within-person variation, the logarithmically transformed positive exposure distribution is back-transformed and combined with the intake frequency to estimate the long-term exposure distribution.…”