Trends in Analytical Chemistry, Vol. 28, No.

6, 2009

Trends

Combining analytical techniques, exposure assessment and biological effects for risk assessment of chemicals in food
J.L.C.M. Dorne, L.R. Bordajandi, B. Amzal, P. Ferrari, P. Verger
Humans today are exposed to a plethora of chemicals in food whether of anthropogenic or natural origin, and public-health agencies have developed risk-assessment methods to derive safe levels of exposure and to prevent adverse health effects. This review highlights analytical techniques used to measure chemical contaminants in food, how human exposure to such contaminants is assessed, and how contamination and exposure are combined with biological (toxicological) effects for risk assessment. We illustrate the whole process using recent examples of importance in public health and give perspectives on the future. 2009 Elsevier Ltd. All rights reserved.
Keywords: Analytical chemistry; Biological effect; Chemical contaminant; Exposure assessment; Food contamination; Genotoxicity; Metabolism; Public health; Risk assessment; Toxicity

1. Introduction
J.L.C.M. Dorne* Institute of Human Nutrition, School of Medicine, University of Southampton, UK J.L.C.M. Dorne, L.R. Bordajandi European Food Safety Authority (EFSA), Largo Palli 5/A, Unit on Food Contaminants, 43100 Parma, Italy B. Amzal EFSA, Assessment Methodology Unit, 43100 Parma, Italy P. Ferrari EFSA, Data and Exposure Assessment Unit, 43100 Parma, Italy P. Verger Met@risk, INRA-AgroParis Tech, 16 rue Claude Bernard, 75231 Paris, France

Corresponding author. Tel.: +39 0521 036472; Fax: +39 0521 0361472; E-mail: jean-lou.dorne@efsa. europa.eu

Modern man is exposed to a wide range of microorganisms and chemicals, the uptake of which by the human body is mainly through food, water, air and dermal contact. The European Food Safety Authority (EFSA) was created in 2002 to assess the risk of these hazards to human health when they are ingested via food. In theory, assessment of exposure to microbes is quite similar to that for chemicals with acute adverse effects but the former topic is beyond the scope of this article, which deals exclusively with chemicals or xenobiotics occurring in food. Xenobiotics can be classied into broad categories, according to their relevance in terms of food safety, namely contaminants and chemicals that have been intentionally added to food or raw commodities. Man-made contaminants of importance include persistent organic pollutants [i.e. dioxins, polychlorinated biphenyls (PCBs), brominated ame retardants (BFRs)], melamine, phthalates, peruoroalkyl acids [i.e. peruoro-octanoic acid (PFOA) and sulfonate (PFOS)], a large number of

pharmaceuticals, and natural toxins (i.e. mycotoxins, marine biotoxins and plant toxins). Other contaminants in food are produced from the Maillard reaction during frying and cooking at high temperature (i.e. acrylamide) or as a reaction product between ethanol and precursors (cyanide) (i.e. ethyl carbamate produced mainly in stone spirits). Food additives, avorings and food-contact materials constitute the largest classes of chemicals that are intentionally added to food, whereas chemicals resulting from intentional treatment of raw commodities include mostly pesticides or biocides (e.g., herbicides, fungicides and insecticides) and veterinary residues (e.g., cocciodiostatics and antibiotics). On a worldwide perspective, the World Heath Organization (WHO) is leading an initiative to provide reliable, accurate estimates of the global burden caused by all food-borne diseases caused by chemicals, parasites, and enteric infections by 2012 [1]. This program is expected to estimate and to compare on a common scale the respective burden for human health of various hazards from different origins. 695

0165-9936/$ - see front matter 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.trac.2009.03.008

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

The United Nations Childrens Fund (UNICEF) and the WHO have already estimated that over 1 billion people are currently deprived of safe water supplies and adequate sanitation. In 2005, they launched the International Decade for Action: Water for Life and the Millennium Development Goals (MDGs) to halve the proportion of the worlds population without sustainable access to safe drinking water and sanitation by 2015 [2]. In such a context, standardized, powerful tools are needed in order to protect consumers from potential adverse health effects that may arise from exposure to chemicals. Scientists and public-health agencies have developed risk-assessment methods to derive human safe levels of exposure. Risk assessment has been divided into four sequential steps: hazard identication, hazard characterization, exposure assessment and risk characterization [3]. In practice, once a chemical has been identied, its content in food measured through validated analytical techniques, its biological (toxicological) effects characterized and a safe level derived, one can relate exposure to biological effects for human risk assessment. This review focuses on bridging the gap between chemical exposure and biological effects using analytical chemistry as a vehicle between the two. First, we review analytical techniques and exposure assessment applied to food contaminants together with biological and toxicological bases for setting safe levels of exposure in humans using recent examples of importance in public health. We then highlight the state of the art and the future of combining exposure and biological/toxicological effects with particular emphasis towards a multidisciplinary approach involving analytical chemistry, biology, toxicology and quantitative modeling.

2. Assessment of exposure to chemicals in food Human dietary exposure to chemicals is mostly through ingestion of food and water, and reects an external dose of the chemical entering the body. Assessing such exposure to chemicals is complex and depends on the availability of analytical techniques to provide the concentration of the chemical in a particular food item (occurrence data) and the corresponding patterns of food consumption in humans [4]. 2.1. Analytical techniques The occurrence data used for risk assessment are usually obtained from routine monitoring programs conducted at the level of a specic country to check compliance of contaminants for which maximum levels are laid down in European Union (EU) legislation. In addition, the implementation of the Rapid Alert System for Food and Feed (RASFF) in Europe has provided a helpful tool to perform systematic monitoring of specic notications regarding either regulated contaminants or xenobiotics 696
http://www.elsevier.com/locate/trac

that may be above maximum levels or emerging contaminants or xenobiotics occurring in food and feed. To obtain reliable occurrence data on contaminants in food, the availability of analytical methods for their determination is of utmost importance. The complexity of food samples, together with the low concentrations at which contaminants occur, requires highly sensitive, selective and reliable analytical techniques, which can also be applied to biological and environmental samples. Typically, the determination of contaminants in such matrices involves a number of steps (e.g., sampling, sample preparation, separation and detection, identication and quantication of the target compounds). Well-established separation techniques [e.g., gas chromatography (GC), liquid chromatography (LC) and capillary electrophoresis (CE)] are usually employed for the determination of food contaminants. Detection systems play a key role and should be sensitive and selective enough for the unequivocal determination of the target analytes. In particular, mass spectrometry (MS), coupled to GC, LC and CE, has become an essential tool to provide valuable structural information for identication of the compounds. The use of tandem MS (MS2) is becoming frequent with the introduction of different mass analyzers that enhance MS2 capabilities {i.e. improved designs of the triple-quadrupole (QqQ) and hybrid systems [e.g., the quadrupole-linear ion trap (QqLIT) and quadrupole time-of-ight (Qq-TOF)]}. The use of these systems allows not only for better sensitivity, but also provides further conrmation of the identity of the analytes [5,6]. While GC-MS is quite well established as a technique in many routine laboratories for the analysis of non-polar, semi-polar, volatile and semi-volatile contaminants, LC-MS or LC-MS2 has experienced great development in recent years, with an increased number of applications in both food-safety and environmental elds. It has become a powerful tool for detection and quantication of polar and non-volatile contaminants, including pharmaceutical residues, veterinary drugs, as well as metabolites and degradation products of food contaminants [7]. However, despite all the advances in instrumental techniques and detection systems, the complexity of matrixes requires, in most cases, an extensive, time-consuming, sample-preparation step, which is often still the bottleneck of the whole analytical procedure [8]. The objective of this step is to extract the target compounds from the matrix and clean up the extract to remove possible interferences (i.e. matrix components or compounds co-eluting with the analyte) that might hinder the nal instrumental determination. The selection of the sample-preparation methods depends on the matrix and the analyte. Currently, sample-preparation methods tend to move towards more environmental friendly approaches (less consumption of organic solvents), miniaturization, automation and, ideally, on-line coupling with the nal instrumental

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Trends

determination. This will lead to extracts that are manipulated less by the analyst, so decreasing the probability of experimental errors. Solid-phase extraction, pressurized liquid extraction, microwave-assisted extraction and solid-phase micro-extraction fulll some of the above-mentioned requirements and offer high throughput [9]. For a number of food contaminants, European legislation establishes the maximum levels permitted in different food commodities and also lays down the methods of sampling and analysis that should be used (e.g., for dioxins and dioxin-like PCBs [10] and ethyl carbamate). For the latter, Commission Regulation No 761/1999 [11] prescribes GC-MS in single-ion-monitoring mode as the method of choice for the determination of ethyl carbamate in wines, and provides detailed information about the whole analytical procedure to be followed [12]. For other compounds, detailed performance criteria to be fullled by the methods of analysis used by the laboratories are laid down (e.g., benzo[a]pyrene, cadmium, lead and mercury) [13]. These performance criteria include recovery ranges, limits of detection (LODs), limits of quantication (LOQs) and precision requirements. However, for other food contaminants, no specic methods of analysis or performance criteria are prescribed. This is especially the case for a number of newly emerging contaminants, for which the analytical procedures are often based on previous protocols for the determination of similar families of contaminants. The development and the validation of dedicated methods, including sample preparation and optimization of the nal instrumental determination, and implementation of quality-assurance and quality-control measures, are then of utmost importance to obtain reliable occurrence data on contaminants and decrease the uncertainty of the measurements. For metals (e.g., arsenic and mercury), speciation has become an essential tool that provides information on the chemical form present in the samples, knowledge of which is crucial to assess the toxicity accurately. In mercury speciation, a number of analytical methods have been proposed for the determination of methylmercury, including GC coupled to atomic uorescence spectrometry (GC-AFS) and LC coupled to inductively coupled plasma-MS (ICP-MS) [14]. In addition, for efcient routine monitoring of food contaminants to support risk assessments, there is a need to develop cost-efcient, reliable and rapid analytical methods. This is the case for dioxins and furans, for which the analysis is time consuming and expensive, due to the multi-step approach to sample preparation and the use of GC coupled to high-resolution MS (GCHRMS) as the only accepted conrmatory analytical method [10]. A number of alternative methods, such as GC-MS2 and multi-dimensional GC techniques, are being evaluated as more cost-effective, fast alternatives, with-

out compromising the required sensitivity and selectivity [15,16]. The implementation of those methods in routine laboratories will not only increase the availability of occurrence data but also allow contamination episodes to be detected at an early stage. Table 1 illustrates a number of examples of food contaminants or xenobiotics together with the analytical techniques that are most commonly used for their determination and recent reviews, papers or EU legislation to which the reader is referred for more detailed information. Such a list does not aim to be exhaustive but to provide the reader with examples regarding major classes of chemical contaminants in food, selected on the basis of their occurrence, toxicity and inclusion in EU legislation. 2.2. Assessing dietary exposure to chemicals Assessing dietary exposure to chemicals combines dietary consumption data with occurrence data (i.e. the concentration of a given chemical contaminant obtained through analytical methods). Ideally, such concentrations are available in an exhaustive, consistent list of food categories, but, in practice, these conditions are rarely met. The most commonly used information on consumption data is derived from dietary surveys, usually conducted at the national level on a representative sample of individuals. In general, these surveys provide estimates of consumption on a limited time frame, and not lifetime consumption. The various dietary assessment instruments can focus on a short-term diet, usually covering a period that ranges from one day to a few days, in the case of one administration vs. replicate administrations of 24-hour dietary recalls, food records or weighed records [4]. Such data are compiled into the Concise European Food Consumption Database established by EFSA to support exposure assessments in the EU [40]. Currently, 19 countries provided national food-consumption data in the adult population, and, to maximize the degree of comparability between these dietary estimates, consumption data have been aggregated in 15 broad categories of food and 29 sub-categories. Other surveys are based on food-frequency questionnaires, dietary-history questionnaires or household purchases that cover longer periods of time in terms of dietary habits. They are often dened as providing information on habitual diet [41], but make it difcult to quantify individual consumption [42]. For contaminant concentrations, analytical data available for risk assessment are customarily produced to check for regulatory compliance to specic norm values. Occurrence data should ideally be homogeneous in terms of analytical methods employed, and provide information on the average and/or extreme occurrence of chemical contaminants in a list of food groups representative of the diet in a population. Departures from
http://www.elsevier.com/locate/trac

697

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Table 1. Common classes of chemical contaminants in food and analytical techniques used for their determination Chemical contaminants in food Agrochemicals Pesticide residues (e.g., herbicides, insecticides and fungicides) Pharmaceuticals Pharmaceutical and veterinary drug residues Environmental contaminants Industrial chemicals Polychlorinated biphenyls (PCBs) Brominated ame retardants (BFRs) Peruorinated alkylated compounds (PACs) Industrial by-products Polychlorinated dibenzo-p-dioxins/furans (PCDD/Fs) Polycyclic aromatic hydrocarbons (PAHs) Metals Cadmium (Cd), Lead (Pb) Arsenic (As) Mercury (Hg), Methyl-mercury (CH3-Hg) Contaminants in food processing Heating Acrylamide Fermentation Ethyl carbamate Materials in contact with food Melamine Phthalates Natural toxins Mycotoxins Phycotoxins Analytical techniques GC-MS, GC-MS2, LC-MS, LC-MS2 LC-MS, LC-MS2, GC-MS Ref. [1719] [2022]

GC-HRMS, GC-MS, GC-MS2 GC-MS, GC-MS2, LC-MS GC-MS, LC-MS, LC-MS2 GC-HRMS, GC-MS2, GCGC GC-MS, LC-Fl AAS, ICP-OES, ICP-MS AAS, LC-ICP-MS GC-MS, LC-ICP-MS

[10,23] [24,25] [26,27] [10,15,16,28] [13,29] [13,30,31] [31,32] [13,14,31]

GC-MS, LC-MS2 GC-MS

[33] [11]

LC-UV, LC-MS2, GC-MS GC-MS

[34] [35]

LC-FLD, LC-MS MBA, LC-MS

[36,37] [38,39]

GC, Gas Chromatography; LC, Liquid chromatography; MS, Mass spectrometry; MS2, Tandem mass spectrometry; UV, Ultraviolet detector; GCGC, Comprehensive two-dimensional gas chromatography; FLD, Fluorescence detector; ICP-OES, Inductively coupled plasma-optical emission spectrometry; ICP-MS, Inductively coupled plasma-MS; MBA, Mouse bioassay.

these requirements are likely to raise concerns on the accuracy of exposure-assessment calculations. In international investigations, the comparability of gures produced at the country level needs to be carefully evaluated. In addition, special efforts need to be invested to handle non-detect values (i.e. samples for which the chemical concentration is below the LOD or LOQ). Data of this nature are typically left out [43]. The approach applied can have a great impact on dietary estimates for the chemicals under assessment. Deterministic and probabilistic approaches can be introduced to handle challenging aspects for the treatment of laboratory data [44] with the degree of performance varying according to a set of scenarios in terms of sample size, frequency of non-detects and departure of empirical values from known statistical distributions [45]. In the eld of food safety, the most commonly used method is currently substitution of results below LOD or LOQ by half the LOD or LOQ value [46]. For quantitative exposure assessment, data on food consumption and chemical occurrence are usually

combined using either a deterministic approach, also called point estimate, or a probabilistic approach [4]. The point-estimate approach is based on the selection of a xed level in the distribution of consumption multiplied by a xed value chosen from the distribution of concentration. The value of contamination could be the 95th percentile or the maximum authorized levels in the regulation (food additives) or an average summary value (mean or median) of the occurrence data (contaminants or pesticide residues). Values of the same nature are used from consumption distributions, so that often combinations of different average or high values from the consumption and the concentration sides are used to evaluate various risk scenarios, as shown in Fig. 1. This method does not reect the exposure of the population, but it is often considered the most appropriate for screening purposes [47]. In practice, the xed levels utilized to calculate point estimate are generally chosen assuming a conservative scenario, so as to be on the safe side when determining the absence of safety concern (e.g., ochratoxin A, organotin compounds, ethyl

698

http://www.elsevier.com/locate/trac

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Trends

CONTAMINANT

Consumption data

Occurrence data

1) Deterministic (point estimate) approach:

Fixed value:
Average (Mean or median) or High (95th, 97.5th percentile, maximum)

Fixed value:
Average (Mean or median) or High (95th, 97.5th percentile, maximum)

2) Probabilistic approach

X
Consumption Concentration

Figure 1. Exposure assessment of chemical contaminants in the human population.

carbamate, polycyclic aromatic hydrocarbons and PFOA) or a favorable ratio between toxicological risk and nutritional benet (e.g., contaminants of wild and farmed sh, and nitrate). Such a conservative approach is also specically promoted for the estimation of dietary intake of pesticides residues by FAO/WHO guidelines [48,49]. An estimate of a Theoretical Maximum Daily Intake (TMDI) is then calculated by multiplying the established or proposed Codex maximum residue levels (MRLs) by the estimated average daily regional consumption for each food commodity, and then summing the products: X TMDI MRLi Fi where MRLi is the MRL of given food commodity i, and Fi is the per capita average daily regional consumption of commodity i. When information is available, this formula can be rened into an International Estimated Daily Intake (IEDI), which incorporates some correction factors (e.g., to account for commodity processing and cooking). By contrast, probabilistic approaches use the whole distributions of occurrence and consumption data, thus

exploiting the variability in both quantities. These methods result in more realistic pictures, often expressed in terms of a range of possible exposure values, thus carrying out an evaluation of the uncertainty associated with exposure estimates. A variety of empirical, semiparametric and parametric models have been described in this context, depending on the opportunity of using actual data (non-parametric approach), or rather data imputed after estimation of parameters (parametric approach) of a theoretical statistical distribution (often from the log-normal, Weibull, exponential). The resulting exposure-assessment distribution is then compared with health-base guidance values, set on the toxicological prole of the compound of interest, so that the risk of exceeding such levels can be characterized in a probabilistic manner.

3. Biological and toxicological effects of chemicals in the food chain Once contaminants have been identied and quantied in food, and human exposure (external dose) is known,

http://www.elsevier.com/locate/trac

699

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

the biological activity or toxicity of chemicals arises from two basic processes:  what the body does to the chemical toxicokinetics (TK); and,  what the chemical does to the body toxicodynamics (TD). In other words, TK involves the translation of a chemical external dose to an internal dose leading to overall elimination from the body (i.e. absorption from the gastro-intestinal tract, distribution in body uids or tissues, metabolism to biologically inactive or active metabolites, and, ultimately, excretion in the urine or feces). Potential bioaccumulation in tissues of either parent compound or metabolites is an important aspect for TK and relates to absorption, distribution, metabolism and excretion of the compound. The half-life of the compound and lipophilicity in the body provide good descriptors as to whether the compound will bioaccumulate or not. Biological (toxicological) effects may occur once the toxic species, as either parent compound or an active or toxic metabolite, reaches its target (receptor, cell or organ) to express its biological activity or toxicity (TD). For biologically-active substances (e.g., pharmaceuticals), the terms pharmacokinetics (PK) and pharmacodynamics (PD) are used, and their biological effects often refer to a therapeutic property that involves substancespecic or class-specic mechanisms [i.e. inhibition of enzymes (e.g., non-steroidal anti-inammatory drugs or angiotensin-converting enzyme inhibitors), receptor agonism or antagonism (e.g., serotonin-specic reuptake inhibitors)]. Often, the borderline between pharmacology and toxicology can be thin, as the old adage by Paracelsus stipulates Sola dosis fecit venenum It is only the dose which makes a chemical a poison. In practice, PK or PD and TK or TD analyses also differ in their objectives and in their levels of precision. PK or PD evaluations are usually a step in the development of a drug (e.g., supporting evaluation of the optimal dose evaluation or the target populations). They are based on clinical trials in which both efcacy and safety are monitored. However, evaluations of TK or TD are usually performed within the context of risk assessment or pure safety assessment, using data from epidemiological or toxicological studies. Evidence gathered from such studies is far less strong and far less specic than that from clinical trials, but it is meant to identify early safety signals at the larger scale of the human population. Traditionally, the safety of pharmaceuticals has been estimated using the margin of safety corresponding to the ratio between the therapeutic dose and the lowest dose for adverse effects in humans. Ideally, such a ratio should be large enough so that the drug can be used safely. A well-known aspect of drug safety is consideration of genetic polymorphisms in drug-metabolizing 700
http://www.elsevier.com/locate/trac

enzymes [e.g., those identied for major isoforms of cytochrome P-450 enzymes (CYP) (CYP2D6, CYP2C9, CYP2C19)]. For such polymorphisms, subjects have been classied according to whether they express the enzyme functionally or not (i.e. extensive metabolizers and poor metabolizers [50,51]). Consequently, human variability in the elimination and the effect of pharmaceuticals and chemicals is large, so that a single value for a margin of safety covering the whole human population is no longer applicable, and clinical pharmacologists are moving towards the concept of personalized medicine, so that the dose of a drug can be adjusted according to individual phenotypes. For that purpose, statistical modeling, accounting for inter-individual variability, is routinely applied in PK and PD analyses [52]. Risk assessors and toxicologists are also moving towards analysis of population variability in TK and TD using statistical modeling for risk-assessment purposes [53,54]. When considering toxicological effects from a regulatory perspective, two basic mechanisms have been retained for chemical-risk assessment, namely:  whether or not the chemicals are genotoxic carcinogens; and,  the derivation of health-based guidance values for humans that differ according to this difference in mode of action (MOA). The term genotoxic is used if a substance or its active metabolite affects cellular DNA through a direct DNAreactive MOA involving covalent binding in target cells to cause pre-carcinogenic mutations and leading to neoplastic transformation and neoplasm induction. For this MOA, it is assumed that there is no threshold in the dose-response relationship, so there is no dose without a potential effect [55]. For such genotoxic carcinogens, the Margin of Exposure (MOE) approach has recently been applied to a number of contaminants (i.e. aatoxins, ethyl carbamate, polyaromatic hydrocarbons and acrylamide) by the Joint Food and Agricultural Organization of the UN/WHO (FAO/WHO) Expert Committee on Food Additives (JECFA) and EFSA [12,56]. Such MOEs are derived using dose-response or dose-effect data using the model-based evaluation of a benchmark dose (BMD) and its lower condence limit (BMDL) [57,58]. MOE values of 10,000 or more, are considered of low concern from a public health point of view and might reasonably be considered as a low priority for risk management actions [56,49]. For non-genotoxic carcinogens, the MOA is epigenetic in nature and does not involve covalent binding to DNA but effects in target cells can either lead indirectly to neoplasms or facilitate their development from cryptogenically-transformed cells. The general assumption is that there is a threshold level of exposure below which no signicant effects are induced. This implies that homeostatic mechanisms are able to counteract any

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Trends

perturbation produced by low levels of intake, and that structural or functional changes leading to adverse effects, which may include cancer, would be observed only at higher intakes [55]. For these chemicals with a toxicological threshold, a standard uncertainty factor (UF) of a 100-fold has been applied to a surrogate for the threshold from sub-chronic/chronic toxicity data [e.g., the no-observed-adverse-effect-level (NOAEL) or a BMD] to determine safe levels without appreciable health risk when consumed every day or weekly for a lifetime [e.g., Acceptable and Tolerable Daily Intakes (ADI/TDI)] or provisional Tolerable Weekly Intake (TWI) in Europe and the Reference dose in the USA. The UF of a 100-fold takes into account interspecies differences (10-fold) and human variability (10-fold). Recently, research efforts have been made to rene these UFs taking into account TK and TD aspects and replace them with pathway-related UFs, when metabolic routes are known, or, ideally, with chemical-specic adjustment factors [3]. For thresholded chemicals with acute toxic effects, the acute reference dose approach (ARfD) has been developed with one of the rst references from the FAO/WHO, spelling out that consideration should be given to the potentially acute toxic effects that are not normally considered in the assessment of the ADI. [59]. Such an approach has been applied to pesticides and dened as by the Joint FAO/WHO Meeting on Pesticide Residues (JMPR): The ARfD of a chemical is an estimate of the amount a substance in food and/or drinking water, normally expressed on a body weight basis, that can be ingested in a period of 24 h or less without appreciable health risk to the consumer on the basis of all known facts at the time of the evaluation [60]. The following sub-section illustrates these concepts using recent toxicological assessment for a genotoxic carcinogen (ethyl carbamate) and non-genotoxic carcinogens with chronic effects (methylmercury) and acute effects (okadaic acid). 3.1. Ethyl carbamate: a bioactivated genotoxic carcinogen Ethyl carbamate is naturally occurring in fermented foods (e.g., bread, soy sauce and yoghurt) and beverages (e.g., wine, beer, spirits and, particularly, stone-fruit brandies). It is the ethyl ester of carbamic acid, mostly biosynthesized from the reactions between ethanol and urea or ethanol and cyanate. Ethanol and urea are major end-products of hexose fermentation and arginine catabolism, respectively, and are often produced by the yeast Saccharomyces cerevisiae, whereas, in stone-fruit spirits, cyanate is the oxidation product of cyanide resulting from the hydrolysis of cyanogenic glycosides in the fruit stone. Ethyl carbamate is rapidly and completely absorbed from the gastrointestinal tract after ingestion, evenly distributed and metabolized via hydrolysis, and three

oxidation reactions mediated by the CYP2E1 isoform of the CYP system (i.e. N-hydroxylation, C-hydroxylation and side-chain oxidation) [12]. In mammals, the CYP2E1 isoform is highly conserved and is involved in the metabolism of endogenous compounds (i.e. vitamins, fatty acids and prostaglandins) and small hydrophobic xenobiotics, including many proximate carcinogens (e.g., ethyl carbamate, N-nitrosamines, benzene, butadiene, styrene, aniline, vinyl chloride and ethanol) [61]. The hydrolysis pathway of ethyl carbamate is mediated by esterases and leads to ethanol, carbon dioxide and ammonia. N-hydroxylation and C-hydroxylation form N-hydroxycarbamate and a-hydroxy ethyl carbamate, which are conjugated and excreted in urine and metabolized to ammonia and carbon dioxide, respectively. The side-chain oxidation of ethyl carbamate by CYP2E1 generates two genotoxic metabolites (i.e. vinyl carbamate and its epoxide). Vinyl carbamate and its epoxide have been shown to be mutagenic, clastogenic and carcinogenic in mouse. In terms of toxicological assessment, the JECFA has derived a BMDL10 of 0.3 mg/kg b.w./day based on a 10% increase in alveolar and bronchiolar neoplasms in male and female mice in a 2-year chronic study [62,63]. The International Agency for Research on Cancer (IARC) has recently reclassied ethyl carbamate from possible carcinogenic to humans (Group 2B) to probably carcinogenic to humans (Group 2A) [64]. Recent data also showed that the prevalence of tumors in the liver (i.e. hemangiomas and hemangiosarcomas), lung (i.e. nodules and adenomas) and Harderian gland (i.e. adenomas) of transgenic mice expressing the functional CYP2E1 gene (CYP2E1 +/+) were higher than for mice decient in CYP2E1 (CYP2E1 -/-). In addition, diallyl sulfone from garlic, a known CYP2E1 inhibitor, reduced by half the frequency of mutations in target tissues (i.e. lung and small intestine). Fig. 2 summarizes the metabolic activation of ethyl carbamate to its proximate carcinogen and the toxicological consequences of such bioactivation [12,63,65]. 3.2. Methylmercury: a thresholded neurotoxin Mercury is an environmental contaminant present in sh and seafood, mainly as methylmercury, which is also the most toxic form of mercury. Methylmercury is highly toxic to the nervous system, especially at the development stage of the brain. Risk assessments of methylmercury have been performed by JECFA [66], US National Research Council (US NRC) [67] and EFSA [68]. Both JECFA and EFSA approached the determination of a Provisory TWI (PTWI) for methylmercury using the NOAEL paradigm, whereas US NRC used the BMD approach. Toxicity data could be gathered from two epidemiological cohort studies, hence with high uncertainty and potential confounding factors (e.g., the presence of other pollutants, such as PCBs, or the inuence
http://www.elsevier.com/locate/trac

701

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

O O NH 2

Ethyl carbamate

CYP2E1
O O NH2
O S O
Diallyl sulfone

Vinyl carbamate

CYP2E1

CYP2E1
Inhibition

O S O
H

O O

O NH2

Diallyl monoepoxide

Vinyl carbamate epoxide

DNA

Mutations A:T-->G:C transitions A:T-->T:A transversions

Multi-site carcinogenicity
Lung Harderian gland Liver Stomach
Figure 2. Bioactivation of ethyl carbamate by CYP2E1, mutagenicity and carcinogenicity of vinyl carbamate and inhibition by diallyl sulfone.

of nutrition). JECFA and EFSA concluded with a PTWI of 1.6 lg/kg b.w., whereas US NRC concluded with a PTWI of 0.7 lg/kg b.w. This illustrates differences in risk assessment occurring from the use of different paradigms and methodologies. Interestingly, US NRC further investigated the reference dose by performing a metaanalysis of four epidemiological studies [69], using a Bayesian hierarchical model. A similar approach using Bayesian meta-analysis was recently used by EFSA for the evaluation of a PTWI of cadmium [54]. 3.3. Okadaic acid and its analogs: setting an acute reference dose Okadaic acid and its structural analogues, the dinophysis toxins (DTX1, DTX2, and DTX3), form the group of 702

okadaic-acid toxins that are produced by dinoagellates. Lipophilic and heat stable, okadaic-acid toxins are mostly found in lter-feeding bivalve molluscs (i.e. oysters, mussels, scallops and clams) and cause diarrheic shellsh poisoning through inhibition of serine/threonine phosphoprotein phosphatases. Symptoms of human intoxication include diarrhea, nausea, vomiting and abdominal pain, which can occur shortly after consumption of contaminated bivalve mollusks, so justifying the need to set an ARfD based on human case reports. Hence, EFSAs Panel on Contaminants in the Food Chain (CONTAM Panel) identied a lowest-observed-adverseeffect level (LOAEL) for human illness around 0.8 lg okadaic acid equivalents/kg b.w. in adults to derive an ARfD of 0.3 lg okadaic acid equivalents/kg using a UF of

http://www.elsevier.com/locate/trac

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Trends

3, taking into account uncertainties in the estimated exposure and extrapolation from LOAEL to NOAEL. No additional UF was applied to the LOAEL value taking into account human variability, since the data were based on a large number of shellsh consumers from various countries, who were considered to comprise the most sensitive individuals [38].

4. Combining exposure, and biological or toxicological effects for risk assessment For genotoxic and non-genotoxic carcinogens, providing a full risk assessment requires four steps: hazard identication, hazard characterization, exposure assessment and risk characterization. These four steps constitute the cornerstone of the multidisciplinary framework, illustrated in Fig. 3, to integrate exposure assessment (occurrence data gathered through analytical methods, dietary exposure assessment and food consumption data), and quantitative surrogates regarding the toxicity

of the compound subdivided according to acute and chronic effects (i.e. NOAEL, BMD and AfRD). In the case of ethyl carbamate, the relevant health effect is lung carcinogenicity in rodents and the risk characterization was performed using the MOE approach. The EFSA CONTAM Panel compared the BMD level of 0.3 mg/kg b.w./day with several scenarios for dietary exposure to ethyl carbamate. Ethyl carbamate can be present in a variety of foods, but it occurs at a very high level in quite specic alcoholic beverages. Some 33,000 analytical results for ethyl carbamate in alcoholic beverages were submitted to EFSA by EU Member States. However, estimates for ingestion of these categories of beverages are quite scarce and potentially underestimated for obvious reasons, so a deterministic approach based on scenarios was preferred to a probabilistic one. When considering the dietary exposure from all food, excluding alcoholic beverages, the margin of exposure of about 20,000 is sufcient to conclude with an absence of safety concern. By contrast, scenarios including food consumed together with a high con-

CHEMICAL CONTAMINANT

EXPOSURE ASSESSMENT

TOXICOLOGICAL EFFECTS

ANALYTICAL CHEMISTRY

Occurrence data (Concentration in food)

Food consumption

Genotoxic

Non-genotoxic

Chronic

Acute

Toxicokinetics / Toxicodynamics

QUANTITATIVE MODELLING

Probabilistic / deterministic approach

Benchmark dose/ NOAEL

Health-based guidance value

MOE

ADI / TDI

ARfD

RISK ASSESSMENT
Figure 3. A multidisciplinary framework for combining exposure assessment and toxicological effects for the risk assessment of genotoxic and non-genotoxic food contaminants.

http://www.elsevier.com/locate/trac

703

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

sumption of fruit brandy result in a margin of exposure less than 600, which could be of health concern [12]. In the case of methylmercury, the most relevant health end-point is the effect on the central nervous system, particularly during the developmental period, and the objective of the dietary exposure assessment was to estimate fetal exposure from the mother. Because sh and other seafood were shown to be the main sources of methylmercury, the assessment was based on women eating sh. This group of consumers represents a fraction of the whole population. For both JECFA and EFSAs CONTAM Panel, the risk was characterized by a comparison between the dietary exposure of women of childbearing age and the lower bound of a BMD equal to 14 lg/g of womens hair. This level in hair measured in epidemiological studies was extrapolated to the PTWI for methylmercury of 1.6 lg/kg b.w./week. In the EFSA opinion, both a deterministic (rst step) and a probabilistic (second step) assessment were performed. The rst step was based on sh consumption for adults including women of child-bearing age in six European countries and on a large compilation of contamination data (14,912 samples of sh and other seafood aggregated into 196 analytical results). The mean level of contamination was multiplied by the mean and the high levels of consumption of sh in the countries considered. The resulting dietary exposure to methylmercury was in the range 0.13.5 lg/kg b.w./week. The second step, based on the same occurrence data combined with the distribution of sh consumption in France, resulted in a probability of 1.2% for adults, including women of childbearing age, to exceed 1.6 lg/kg b.w./week. Overall, the EFSA opinion concluded that: Population groups who frequently consume large predatory sh, such as swordsh, tuna, and halibut, may have a considerably higher intake of methylmercury and exceed the PTWI. Acute toxicity is more uncommon than chronic toxicity for food contaminants, but, for those substances that exhibit such effects (e.g., okadaic acid and its structural analogues), dietary exposure assessment can be based on deterministic or probabilistic approaches. The acute effects of intoxication due to okadaic-acidgroup toxins can be mild to very severe, but their common characteristic results from a single ingested portion of bivalve molluscs. The risk characterization involves comparing, for each of the toxins, the ARfD with the amount of toxins potentially ingested within a single large portion reported in food-consumption surveys. This deterministic approach is useful to establish maximum levels of toxins in food and aims to protect high consumers, even when they are consuming a portion contaminated at the highest authorized level. However, in a number of cases, such a level of protection is difcult to achieve, and risk managers should balance major economical impacts with a certain risk of poisoning. To quantify the sanitary impact of such an acute adverse 704
http://www.elsevier.com/locate/trac

effect on the population, a probabilistic model combining the distributions of occurrence, portion sizes and frequency of eating occasions should be constructed. At this stage, important assumptions on these distributions need to be made because none of them is fully known. For example, the duration of individual surveys of 17 days does not allow accurate estimation of the yearly frequency of consumption of seafood so the estimated probability of adverse effects within the population of consumers during one year is uncertain. For okadaic acid, the deterministic approach combining a high portion of molluscs consumed and the 95th percentile of the concentration for okadaic-acid-group toxins results in an intake of about ve-fold the ARfD. The EFSAs CONTAM Panel concluded that it is therefore likely that such levels will result in diarrheic shellsh poisoning. A probabilistic approach, based on the distributions of both concentration and consumption data, showed that there is a chance of approximately 20% of exceeding the ARfD for a 60-kg adult, when consuming shellsh containing levels of okadaic-acid-group toxins that could be present in shellsh currently available on the European market.

5. Conclusions and perspectives on the future Combining state-of-the-art analytical techniques, exposure and biological or toxicological effects to protect consumers from chemical exposure constitutes one of the primary goals of risk assessment. In recent decades, the impact of the public perception of chemical risk and the rising number of chemicals in human food and environment have led to an increase in science-based risk assessments. There is a trend towards more quantitative risk assessments, which is also promoted by guidelines of international organizations [42]. It echoes the wish for risk assessors and managers to measure more objectively the uncertainty and the variability associated with any ofcial statement on public health. In this context, quantitative tools (e.g., modeling and statistical techniques) have sensibly arisen to support chemical-risk assessment in food and the environment. Aside from physiologically-based TK models and doseeffect models mentioned above, advanced statistical modeling has been used to enable integration of all the complexity from analytical techniques to exposure assessment. It includes quantitative structure-activity relationship (QSAR) modeling, population and hierarchical models and statistical methods for inference of exposure and toxicological data. With respect to hazard identication, animal-based toxicity testing is still widely used but its limitations are increasingly acknowledged [55]. QSAR modeling and structure-activity analyses can assist early identication of structures in structural alert for a wide range of

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

Trends

chemicals that may have high reactivity prior to any toxicological data being available. It is therefore a powerful tool for early, large-scale screening of chemical compounds, as explored by the US Environmental Protection Agency (US EPA) [70]. Such methodologies can generate a number of models that need specic datamining tools to visualize and to analyze them. Development of such specic quantitative methods gave birth to so-called cheminformatics (also known as chemoinformatics). With respect to hazard characterization, the modern mechanistic approach in toxicology has been critical to the identication of critical end-points of relevance to assist derivation of health-based guidance values for both genotoxic and non-genotoxic carcinogens. Understanding the TK and the TD of a particular chemical is a key step to assessing the toxicity of a compound and to bridging exposure to biological effects. As illustrated for genotoxic carcinogen ethyl carbamate, the TK and particularly bioactivation through metabolism can have dramatic effects on the biological dynamics and the TD of a chemical. Studying the toxicological perturbations caused by chemicals and other stressors at gene, protein and metabolite levels in an interactive, integrative way is the principal aim of systems biology, and such complex biological information can be arranged into a quantitative model using analytical and bio-informatics tools [71]. Specically, applications of systems biology have been a consequence of the human-genome project, and technological advances in the omics sciences (e.g., genomics, proteomics, metabolomics and metabonomics) are now generating such quantitative data related to our understanding of toxic mechanisms at the levels of populations, individuals, cells and molecular targets. The use of MOA for toxicology and risk assessment has been increasing, and genome-wide (i.e. global) measurements provide an unbiased check of the biological basis of MOA. For this purpose, identication of biomarkers in humans provides a viable way to extrapolate from disease outcomes measured at high exposure levels to outcomes at low exposure levels, so providing the opportunity to reduce considerably in vivo toxicity testing using animals [72]. In addition, intelligent or integrated hierarchical testing strategies are being developed in Europe to implement the legislation for Registration, Evaluation, and Authorization of Chemical Substances (REACH). These strategies include use of human cells or tissues potentially to eliminate interspecies extrapolation, to increase efciency in testing and to reduce the use of animals. Importantly, validation of in vitro testing methods can specically address particular MOAs for specic toxicological end-points [73,74]. For exposure assessment, the availability of reliable analytical techniques is critical to detect and to quantify chemicals in environmental and food samples (occur-

rence data) with a high level of accuracy. Advances in the instrumental techniques (especially use of MSdetection systems) together with the validation of the methods and the implementation of quality-assurance and quality-control measures improve the quality of occurrence data, leading to more robust food-safety risk assessments and to better protection for the consumer. Although sample preparation is still the bottleneck in many analytical procedures, more environmentalfriendly extraction techniques, which allow for miniaturization and on-line coupling, are currently being developed to attain faster, less labor-intensive protocols. It would be desirable for methods of analysis not only to cover detection and quantication of target contaminants down to the legislative limits, but also to allow for detection of banned chemicals that might still be in use, detection of metabolites or degradation products, and prompt detection of unintentional contamination [8]. The development of multi-residue methods for the simultaneous detection of a number of contaminants is a relevant eld, where the use of multi-dimensional separation techniques (GC GC, LC LC) and MS detection will have an essential role [5,75]. The risk-characterization step integrates exposure to a chemical in a given population with toxicological effects on which a health-based guidance value is based, and concludes with the likelihood of adverse effects. As explained above, quantifying population variability is becoming a central step in the whole risk-assessment process. Improvements to this quantication include development of rened TK models for population [76], collection of data matching dose and response [77], and use of chemical-specic UFs [3]. Such model renements go along with improved or more suitable techniques for dealing with statistical inference. For example, hierarchical models have been powerful tools to quantify variability of kinetics and response. Bayesian statistics have also been particularly suited to t such models [76]. The use of meta-analysis also constitutes a way forward to combine and to aggregate data from a range of studies for a particular compound and specic sub-populations to provide databased estimates of variability across populations and studies. Meta-analysis of human TK data for substrates metabolized via polymorphic CYP isoforms (CYP2C19 and CYP2D6) have shown that the UF allowing for variability in TKs (3.16) does not include such polymorphic and pathway-related UFs, which were suggested to allow for the incorporation of in vivo metabolism and TK data in the risk-assessment process. In addition, this approach provides a exible, intermediate option between the default UF and chemicalspecic adjustment factors (CSAFs) derived from physiology-based PK models [50,51,78]. A new method for meta-analysis of TK data based on full Bayesian inference was developed recently. It
http://www.elsevier.com/locate/trac

705

Trends

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

allowed use of hierarchical modeling with covariates, combined with compartmental models to account for inter-individual, inter-compound and inter-study variability. In addition, the use of sensitivity analysis with respect to model assumptions gave the advantage of quantifying both population variability and uncertainty [79]. Finally, since the majority of chemical hazards are characterized by potentially long-term effects on human health, time is a crucial parameter for risk characterization, but it has been widely neglected. Particularly for chemicals with a long half-life, integration of time as a factor inuencing both ingestion and elimination of the chemical by the body is important, but implies the need to estimate a new health-based guidance value [namely, a Tolerable Body Burden (TBB) based on the TDI). The TBB would correspond to the full TWI of a chemical ingested every week and eliminated as a function of its specic kinetics [80]. Such assessment needs more sophisticated mathematical modeling than classical Monte-Carlo analysis [81], particularly to estimate the probability of very rare events. A future challenge for analytical chemists, toxicologists and modelers is the development of methods to assess the risk of chemical mixtures from an ecological and human health perspective. EU 6th Framework project NOMIRACLE (http://viso.jrc.it/nomiracle/) aims to develop methods to improve the risk assessment of mixtures, including taking into account TK and TD interactions in deriving UFs for chemical mixtures and to develop options to harmonize the use of UFs for both human and ecological risk assessments using mechanistic descriptors [82]. Recently, the magnitude of TK interactions in humans between therapeutic drugs metabolized by the polymorphic CYPs were quantied and showed that the current UF for TK does not cater for such interactions. Such data were based on therapeutic doses higher than current contaminant levels in food, and further research is needed to characterize the potential effects of contaminants on individual CYPs at current levels of exposure. Recombinant technology and TK assays obtained routinely in the laboratory can provide useful tools to screen mixtures of environmental health importance to identify whether each component of the mixture has the potential to interact or not [83].

References
[1] World Health Organization, Initiative to Estimate the Global Burden of Foodborne Diseases, Foodborne Diseases a Growing Risk, 2008 (http://www.who.int/foodsafety/foodborne_disease/ ferg/en/index.html). [2] J.L.C.M. Dorne, J. Appl. Toxicol. 27 (2007) 411. [3] J.L.C.M. Dorne, K. Walton, A.G. Renwick, Food Chem. Toxicol. 43 (2005) 203. [4] R. Kroes, D. Muller, J. Lambe, M.R. Lowik, J. van Klaveren, J. Kleiner, R. Massey, S. Mayer, I. Urieta, P. Verger, A. Visconti, Food Chem. Toxicol. 40 (2002) 327. [5] R. Schuhmacher, M. Sulyok, R. Krska, Anal. Bioanal. Chem. 390 (2008) 253. [6] S.J. Lehotay, K. Mastovska, A. Amirav, A.B. Fialkov, P.A. Martos, A. de Kok, A.R. Fernandez-Alba, Trends Anal. Chem. 27 (2008) 1070. [7] Y. Pico, D. Barcelo, Trends Anal. Chem. 27 (2008) 821. [8] K. Ridgway, S.P.D. Lalljie, R.M. Smith. J. Chromatogr., A 1153 (2007) 36. [9] N. Fidalgo-Used, E. Blanco-Gonzalez, A. Sanz-Medel, Anal. Chim. Acta 590 (2007) 1. [10] European Commission, Commission Regulation (EC) No 1883/ 2006 of 19 December 2006 laying down methods of sampling and analysis for the ofcial control of levels of dioxins and dioxinlike PCBs in certain foodstuffs, Off. J. Eur. Commun. L364 (2006) 32. [11] European Commission, Commission Regulation (EC) No 761/ 1999 of 12 April 1999 amending Regulation (EEC) No 2676/90 determining Community methods for the analysis of wines. Off. J. Eur. Commun. L99 (1999) 4. [12] European Food Safety Authority, Opinion of the Scientic Panel on Contaminants in the Food Chain on a Request from the European Commission on Ethyl Carbamate and Hydrocyanic Acid in Food and Beverages, EFSA J. 551 (2007) 1. [13] European Commission, Commission Regulation (EC) No 333/ 2007 of 28 March 2007 laying down the methods of sampling and analysis for the ofcial control of the levels of lead, cadmium, mercury, inorganic tin, 3-MCPD and benzo(a)pyrene in foodstuffs. Off. J. Eur. Commun. L88 (2007) 29. [14] D.-Y. Yang, H.-Y.T. Truong, Y.-W. Chen, N. Belzile, Anal. Chim. Acta 633 (2009) 157. [15] C. Danielsson, K. Wiberg, P. Korytar, S. Bergek, U.A.Th. Brinkman, P. Haglund, J. Chromatogr., A 1086 (2005) 61. [16] J. Malavia, F.J. Santos, M.T. Galceran, J. Chromatogr., A 1186 (2008) 302. [17] A.R. Fernandez-Alba, J.F. Garca-Reyes, Trends Anal. Chem. 27 (2008) 973. [18] C. Soler, Y. Pico, Trends Anal. Chem. 26 (2007) 103. [19] G. Crnogorac, W. Schwack, Trends Anal. Chem. 28 (2009) 40. [20] A.A.M. Stolker, U.A.Th. Brinkman, J. Chromatogr., A 1067 (2005) 15. [21] C. Blasco, C.M. Torres, Y. Pico, Trends Anal. Chem. 26 (2007) 895. [22] A. Gentili, Trends Anal. Chem. 26 (2007) 595. [23] A. Beyer, M. Biziuk, Crit. Rev. Food Sci. Nutr. 48 (2008) 888. [24] A. Covaci, S. Voorspoels, L. Ramos, H. Neels, R. Blust, J. Chromatogr., A 1153 (2007) 145. [25] J. de Boer, D.E. Wells, Trends Anal. Chem. 25 (2006) 364. [26] S.P.J. van Leeuwen, J. de Boer, J. Chromatogr., A 1153 (2007) 172. [27] A. Jahnke, U. Berger, J. Chromatogr., A 1216 (2009) 410. [28] K. Srogi, Anal. Lett. 40 (2007) 1647. [29] European Food Safety Authority, Scientic Opinion of the Panel on Contaminants in the Food Chain on a Request from the European Commission on Polycyclic Aromatic Hydrocarbons in Food, EFSA J. 724 (2008) 1.

Acknowledgments and disclaimer One of the authors (JLCMD) is grateful to the European Commission (2004-December 2006) under the NO MIRACLE project (Number 003956) for funding part of this work. The views reflected in this review are the authors only and do not reflect the views of the European Food Safety Authority.

706

http://www.elsevier.com/locate/trac

Trends in Analytical Chemistry, Vol. 28, No. 6, 2009

[30] M.G.A. Korn, J.B. de Andrade, D.S. de Jesus, V.A. Lemos, M.L.S.F. Bandeira, W.N.L. dos Santos, M.A. Bezerra, F.A.C. Amorim, A.S. Souza, S.L.C. Ferreira, Talanta 69 (2006) 16. [31] S.G. Capar, W.R. Mindak, J. Cheng, Anal. Bioanal. Chem. 389 (2007) 159. [32] C. Dietz, J. Sanz, E. Sanz, R. Munoz-Olivas, C. Camara, J. Chromatogr., A 1153 (2007) 114. [33] Y. Zhang, J. Jiao, Z. Cai, Y. Zhang, Y. Ren, J. Chromatogr., A 1142 (2007) 194. [34] T.M. Vail, P.R. Jones, O.D. Sparkman, J. Anal. Toxicol. 31 (2007) 304. [35] A. Fankhauser-Noti, K. Grob, Anal. Chim. Acta 582 (2007) 353. [36] G.S. Shephard, Chem. Soc. Rev. 37 (2008) 2468. [37] N.W. Turner, S. Subrahmanyam, S.A. Piletsky, Anal. Chim. Acta 632 (2009) 168. [38] European Food Safety Authority, Opinion of the Scientic Panel on Contaminants in the Food Chain on a Request from the European Commission on Marine Biotoxins in Shellsh Okadaic Acid and Analogues, EFSA J. 589 (2008) 1. [39] B. Christian, B. Luckas, Anal. Bioanal. Chem. 391 (2008) 117. [40] European Food Safety Authority, Food Consumption Database, 2008 (http://www.efsa.europa.eu/EFSA/ScienticPanels/efsa_ locale-1178620753812_DATEX.htm). [41] W.C. Willett, Nutritional Epidemiology, 2nd Edition., Oxford University Press, New York, USA, 1998. [42] World Health Organization, Exposure Assessment for Chemicals in Food, Report of the FAO/WHO Workshop, Annapolis, Maryland, USA, 26 May 2005. [43] R.W. Hornung, L.D. Reed, Appl. Occup. Environ. Hyg. 5 (1990) 46. [44] D.R. Helsel, Nondetects and Data Analysis: Statistics for Censored Environmental Data, John Wiley & Sons, Inc., New York, USA, 2005. [45] P. Hewett, G.H. Ganser, Ann. Occup. Hyg. 51 (2007) 611. [46] World Health Organization, Reliable Evaluation of Low-Level Contamination of Food, Report on a Workshop in the Frame of GEMS/Food-EURO, Kulmbach, Federal Republic of Germany, 1995. [47] B. Parmar, P.F. Miller, R. Burt. Regul. Toxicol. Pharmacol. 26 (1997) 44. [48] World Health Organization, Guidelines for predicting dietary intake of pesticide residues (revised), 1997 (www.who.int/foodsafety/publications/chem/en/pesticide_en.pdf). [49] C. Svendsen, A.M.J. Ragas, J.L.C.M. Dorne, Health Benets of Organic Foods: Effects of the Environment, CABI Publishing, Wallingford, Oxon., UK, 2008, Chapter 6, p. 119. [50] J.L.C.M. Dorne, K. Walton, W. Slob, A.G. Renwick, Food Chem. Toxicol. 40 (2002) 1633. [51] J.L.C.M. Dorne, K. Walton, A.G. Renwick, Food Chem. Toxicol. 41 (2003) 225. ` [52] F. Mercier, G. Lefevre, H.L. Huang, H. Schmidli, B. Amzal, S. Appel-Dingemanse, Curr. Med. Res. Opin. 23 (2007) 3199. [53] P. Bernillon, F. Bois, Environ. Health Perspect. 108 (2000) 883. [54] European Food Safety Authority, EFSA J. (2009) (in press). [55] S.M. Barlow, J.B. Greig, J.W. Bridges, A. Carere, A.J.M. Carpy, C.L. Galli, J. Kleiner, I. Knudsen, H.B.W.M. Koeter, L.S. Levy, C.

Trends
Madsen, S. Mayer, J.-F. Narbonne, F. Pfannkuch, M.G. Prodanchuk, M.R. Smith, P. Steinberg, Food Chem. Toxicol. 40 (2002) 145. European Food Safety Authority, Summary of an Opinion of the Scientic Committee on a Request from EFSA related to A Harmonised Approach for Risk Assessment of Substances Which are Both Genotoxic and Carcinogenic, EFSA J. 282 (2005) 1. K.S. Crump, Fundam. Appl. Toxicol. 4 (1984) 854. S. Sand, K. Victorin, A.F. Filipsson, J. Appl. Toxicol. 28 (2008) 405. IPCS INCHEM. Environmental Health Criteria 104 (1990) 81 (http://www.inchem.org/documents/ehc/ehc/ehc104.htm). UN Food and Agricultural Organization (FAO), Joint FAO/WHO Meetings on Pesticide Residues (JMPR), Report of the JMPR, FAO Plant Production and Protection Paper 172 (2002) 4. B.I. Ghanayem, U. Hofer, Curr. Drug Metab. 8 (2007) 728. National Toxicology Program (NTP), NTP Rep. Ser. 510 (2004) 1, US National Institutes of Health, Bethesda, MD, USA. UN Food and Agricultural Organization (FAO)/World Health Organization (WHO), WHO Food Addit. Ser. 55 (2006) 205. IARC Monograph 96 (2007) 1. U. Hofer, B.I. Ghanayem, Drug Metab. Dispos. 33 (2005) 1144. UN Food and Agricultural Organization (FAO)/World Health Organization (WHO), WHO Food Addit. Ser. 52 (2004) 565. US National Research Council, Toxicological Effects of Methylmercury, National Academy Press, Washington, DC, USA, 2000. European Food Safety Authority, Opinion of the Scientic Panel on Contaminants in the Food Chain on a Request from the Commission related to Mercury and Methylmercury in food, EFSA J. 34 (2004) 1. B.A. Coull, M. Mezzetti, L.M. Ryan, J. Agric. Biol. Environ. Stat. 8 (2003) 253. C.J. Moudgal, D. Young, T. Nichols, T. Martin, P. Harten, R. Venkatapathy, G. Stelma, S. Siddhanti, C. Baier-Anderson, M. Wolfe, Environ. Res. 19 (2008) 579. W. Slikker, M.G. Paule, L.K. Wright, T.A. Patterson, C. Wang, J. Appl. Toxicol. 27 (2007) 201. S.W. Edwards, R.J. Preston, Toxicol. Sci. 106 (2008) 312. B. Meek, J. Doull, Toxicol. Sci. 108 (2009) 19. C.J. Van Leeuwen, G.Y. Patlewicz, A.P. Worth, Intelligent testing strategies, in: K. Van Leeuwen, T. Vermeire (Editors), Risk Assessment of Chemicals, Kluwer Academic Publishers, Dordrecht, The Netherlands, 2007, pp. 467509. L.R. Bordajandi, J.J. Ramos, J. Sanz, M.J. Gonzalez, L. Ramos, J. Chromatogr., A 1186 (2008) 312. S. Micallef, B. Amzal, V. Bach, K. Chardon, P. Tourneux, F.Y. Bois, Clin. Pharmacokinet. 46 (2007) 59. A. Akesson, B. Julin, A. Wolk, Cancer Res. 68 (2008) 6435. J.L.C.M. Dorne, L. Skinner, G.K. Frampton, D.J. Spurgeon, A.M. Ragas, Anal. Bioanal. Chem. 387 (2007) 1259. J.L.C.M. Dorne, B. Amzal, Toxicol. Lett. 180S (2008) S90. P. Verger, J. Tressou, S. Clemencon, Regul. Toxicol. Pharmacol. 49 (2007) 25. P. Bertail, S. Clemencon, J. Tressou, Math. Biosci. Eng. 5 (2008) 35. J.L.C.M. Dorne, A.M. Ragas, H. Lokke, Toxicology 226 (2006) 75. J.L.C.M. Dorne, A. Papadopoulos, Toxicol. Lett. 180S (2008) S90.

[56]

[57] [58] [59] [60]

[61] [62] [63] [64] [65] [66] [67] [68]

[69] [70]

[71] [72] [73] [74]

[75] [76] [77] [78] [79] [80] [81] [82] [83]

http://www.elsevier.com/locate/trac

707