Waldo J. de Boer

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

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.

The MCRA model for probabilistic single-compound and cumulative risk assessment of pesticides

Pesticide risk assessment is hampered by worst-case assumptions leading to overly pessimistic assessments. On the other hand, cumulative health effects of similar pesticides are often not taken into account. This paper describes models and a web-based software system developed in the European research project ACROPOLIS. The models are appropriate for both acute and chronic exposure assessments of single compounds and of multiple compounds in cumulative assessment groups. The software system MCRA (Monte Carlo Risk Assessment) is available for stakeholders in pesticide risk assessment at mcra.rivm.nl. We describe the MCRA implementation of the methods as advised in the 2012 EFSA Guidance on probabilistic modelling, as well as more refined methods developed in the ACROPOLIS project. The emphasis is on cumulative assessments. Two approaches, sample-based and compound-based, are contrasted. It is shown that additional data on agricultural use of pesticides may give more realistic risk assessments. Examples are given of model and software validation of acute and chronic assessments, using both simulated data and comparisons against the previous release of MCRA and against the standard software DEEM-FCID used by the Environmental Protection Agency in the USA. It is shown that the EFSA Guidance pessimistic model may not always give an appropriate modelling of exposure.

New approaches to uncertainty analysis for use in aggregate and cumulative risk assessment of pesticides

Risk assessments for human exposures to plant protection products (PPPs) have traditionally focussed on single routes of exposure and single compounds. Extensions to estimate aggregate (multi-source) and cumulative (multi-compound) exposure from PPPs present many new challenges and additional uncertainties that should be addressed as part of risk analysis and decision-making. A general approach is outlined for identifying and classifying the relevant uncertainties and variabilities. The implementation of uncertainty analysis within the MCRA software, developed as part of the EU-funded ACROPOLIS project to address some of these uncertainties, is demonstrated. An example is presented for dietary and non-dietary exposures to the triazole class of compounds. This demonstrates the chaining of models, linking variability and uncertainty generated from an external model for bystander exposure with variability and uncertainty in MCRA dietary exposure assessments. A new method is also presented for combining pesticide usage survey information with limited residue monitoring data, to address non-detect uncertainty. The results show that incorporating usage information reduces uncertainty in parameters of the residue distribution but that in this case quantifying uncertainty is not a priority, at least for UK grown crops. A general discussion of alternative approaches to treat uncertainty, either quantitatively or qualitatively, is included.

Probabilistic acute dietary exposure assessments to captan and tolylfluanid using several European food consumption and pesticide concentration databases.

Probabilistic dietary acute exposure assessments of captan and tolylfluanid were performed for the populations of the Czech Republic, Denmark, Italy, the Netherlands and Sweden. The basis for these assessments was national databases for food consumption and pesticide concentration data harmonised at the level of raw agricultural commodity. Data were obtained from national food consumption surveys and national monitoring programmes and organised in an electronic platform of databases connected to probabilistic software. The exposure assessments were conducted by linking national food consumption data either (1) to national pesticide concentration data or (2) to a pooled database containing all national pesticide concentration data. We show that with this tool national exposure assessments can be performed in a harmonised way and that pesticide concentrations of other countries can be linked to national food consumption surveys. In this way it is possible to exchange or merge concentration data between countries in situations of data scarcity. This electronic platform in connection with probabilistic software can be seen as a prototype of a data warehouse, including a harmonised approach for dietary exposure modelling.

A European model and case studies for aggregate exposure assessment of pesticides

Exposures to plant protection products (PPPs) are assessed using risk analysis methods to protect public health. Traditionally, single sources, such as food or individual occupational sources, have been addressed. In reality, individuals can be exposed simultaneously to multiple sources. Improved regulation therefore requires the development of new tools for estimating the population distribution of exposures aggregated within an individual. A new aggregate model is described, which allows individual users to include as much, or as little, information as is available or relevant for their particular scenario. Depending on the inputs provided by the user, the outputs can range from simple deterministic values through to probabilistic analyses including characterisations of variability and uncertainty. Exposures can be calculated for multiple compounds, routes and sources of exposure. The aggregate model links to the cumulative dietary exposure model developed in parallel and is implemented in the web-based software tool MCRA. Case studies are presented to illustrate the potential of this model, with inputs drawn from existing European data sources and models. These cover exposures to UK arable spray operators, Italian vineyard spray operators, Netherlands users of a consumer spray and UK bystanders/residents. The model could also be adapted to handle non-PPP compounds.

Uncertainty in Intake Due to Portion Size Estimation in 24-Hour Recalls Varies Between Food Groups

Portion size estimation is expected to be one of the largest sources of uncertainty in dietary assessment of the individual. Therefore, we demonstrated a method to quantify uncertainty due to portion size estimation in the usual intake distributions of vegetables, fruit, bread, protein, and potassium. Dutch participants of the European Food Consumption Validation study completed 2 nonconsecutive 24-h recall interviews. In short, the uncertainty analysis consists of Monte Carlo simulations drawing values for portion size from lognormal uncertainty distributions. The uncertainty of the usual intake distribution and accompanying parameters (IQR and the shrinkage factor) were estimated. For the food groups, portion size uncertainty had the greatest effect for vegetables and the least for fruit: the relative 95% uncertainty interval (UI) of the IQR of the usual intake distribution was 0.61-1.35 for vegetables, 0.77-1.24 for bread, and 0.99-1.10 for fruit. For protein and potassium, the resulting relative width of the UI of the IQR for portion size uncertainty are similar: 0.88-1.14 for protein and 0.86-1.14 for potassium. Furthermore, a sensitivity analysis illustrated the importance of the specified uncertainty distributions. The examples show that uncertainty in portion sizes may be more important for some foods such as vegetables. This may reflect differential quantification errors by food groups that deserve further consideration. In conclusion, the presented methodology allows the important quantification of portion size uncertainty and extensions to include other sources of uncertainty is straightforward.

Detection of residues using multivariate modelling of low-level GC-MS measurements

The chemometric analysis of low-level analytical data is hampered by the common presence of interfering compounds, by the frequent absence of measurement signals and by a non-constant measurement variability which is related to concentration level in a non-linear way. A model is presented to handle this type of data in the context of the practical problem of multivariate detection from gas chromatography/mass spectrometry (GC-MS) data. The model, based on log ratio modelling, is compared with previous approaches to parts of the problem. The basic idea behind the model is to define for the multivariate detection problem a null hypothesis for the values of log ratio measurements and to estimate variability as a function of total measured intensity. In practice it is often impossible to anticipate all kinds of interference which may occur. Therefore we propose to use expert assessments of the probability that certain expected peak ratios are generated by the analyte rather than by interferences. These expert assessments can then be used to define a proper null hypothesis for the multivariate detection test. The application of the model is illustrated for the detection of the illegal growth promoter clenbuterol in urine by selected ion-monitoring GC-MS.

Optimizing the balance between false positive and false negative error probabilities of confirmatory methods for the detection of veterinary drug residues

GC-MS data on veterinary drug residues in bovine urine are used for controlling the illegal practice of fattening cattle. According to current detection criteria, peak patterns of preferably four ions should agree within 10 or 20% from a corresponding standard pattern. These criteria are rigid, rather arbitrary and do not match daily practice. A new model, based on multivariate modeling of log peak abundance ratios, provides a theoretical basis for the identification of analytes and optimizes the balance between the avoidance of false positives and false negatives. The performance of the model is demonstrated on data provided by five laboratories, each supplying GC-MS measurements on the detection of clenbuterol, dienestrol and 19 beta-nortestosterone in urine. The proposed model shows a better performance than confirmation by using the current criteria and provides a statistical basis for inspection criteria in terms of error probabilities.

A statistical method to base nutrient recommendations on meta-analysis of intake and health-related status biomarkers.

Nutrient recommendations in use today are often derived from relatively old data of few studies with few individuals. However, for many nutrients, including vitamin B-12, extensive data have now become available from both observational studies and randomized controlled trials, addressing the relation between intake and health-related status biomarkers. The purpose of this article is to provide new methodology for dietary planning based on dose-response data and meta-analysis. The methodology builds on existing work, and is consistent with current methodology and measurement error models for dietary assessment. The detailed purposes of this paper are twofold. Firstly, to define a Population Nutrient Level (PNL) for dietary planning in groups. Secondly, to show how data from different sources can be combined in an extended meta-analysis of intake-status datasets for estimating PNL as well as other nutrient intake values, such as the Average Nutrient Requirement (ANR) and the Individual Nutrient Level (INL). For this, a computational method is presented for comparing a bivariate lognormal distribution to a health criterion value. Procedures to meta-analyse available data in different ways are described. Example calculations on vitamin B-12 requirements were made for four models, assuming different ways of estimating the dose-response relation, and different values of the health criterion. Resulting estimates of ANRs and less so for INLs were found to be sensitive to model assumptions, whereas estimates of PNLs were much less sensitive to these assumptions as they were closer to the average nutrient intake in the available data.