This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Transcription

1 This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier s archiving and manuscript policies are encouraged to visit:

2 Food and Chemical Toxicology 50 (2012) S699 S709 Contents lists available at ScienceDirect Food and Chemical Toxicology journal homepage: Application of the BRAFO-tiered approach for benefit-risk assessment to case studies on natural foods Bernhard Watzl a, Eva Gelencsér b, Jeljer Hoekstra c, Sabine Kulling a, Eva Lydeking-Olsen d, Ian Rowland e, Benoît Schilter f, Jakob van Klaveren c, Alessandro Chiodini g, a Max Rubner-Institute (MRI), Federal Research Institute for Nutrition and Food, Haid-und-Neu-Strasse 9, Karlsruhe, Germany b Central Food Research Institute (CFRI), Department of Food Safety, Herman Ottó út 15, 1022 Budapest, Hungary c National Institute for Public Health and the Environment (RIVM), 3720 Bilthoven, The Netherlands d Institute for Optimum Nutrition, Teglgaardstraede 4, 1452 Copenhagen, Denmark e University of Reading, Department of Food Biosciences, PO Box 226, Whiteknights, RG6 6AP, Reading, UK f Nestlé, P.O.Box 44, Vers-Chez-Les-Blanc, 1000 Lausanne 26, Switzerland g ILSI Europe, Avenue E. Mounier 83, Box 6, 1200 Brussels, Belgium article info abstract Article history: Received 21 May 2010 Accepted 15 February 2011 Available online 19 February 2011 Keywords: BRAFO Benefit-risk assessment Tiered approach Fish Soy There is evidence that consumption of fish, especially oily fish, has substantial beneficial effects on health. In particular an inverse relationship of oily fish intake to coronary heart disease incidence has been established. These beneficial effects are ascribed to fish oil components including long chain x-3 polyunsaturated fatty acids. On the other hand it should be noted that oily fish also contains hazardous substances such as dioxins, PCBs and methylmercury. Soy consumption has been associated with potential beneficial and adverse effects. The claimed benefits include reduced risk of cardiovascular disease; osteoporosis, breast and prostate cancer whereas potential adverse effects include impaired thyroid function, disruption of sex hormone levels, changes in reproductive function and increased breast cancer risk The two cases of natural foods highlight the need to consider both risks and benefits in order to establish the net health impact associated to the consumption of specific food products. Within the Sixth Framework programme of the European Commission, the BRAFO project was funded to develop a framework that allows for the quantitative comparison of human health risks and benefits in relation to foods and food compounds. This paper describes the application of the developed framework to two natural foods, farmed salmon and soy protein. We conclude that the BRAFO methodology is highly applicable to natural foods. It will help the benefit-risk managers in selecting the appropriate dietary recommendations for the population. Ó 2011 ILSI Europe. Published by Elsevier Ltd. All rights reserved. Abbreviations: AhR, Aryl hydrocarbon receptor; AMD, Age-related macular degeneration, BW; BMD, Body mineral density; BW, Body weight; CHD, Coronary heart disease; CVD, Cardiovascular disease; DALY, Disability adjusted life years; DL- PCB, Dioxin like polychlorinated biphenyl; EPIC, European prospective investigation into cancer and nutrition; FDA, Food and drug administration; JECFA, Joint FAO/ WHO expert committee on food additives; HDL, High density protein; HHL, Health conscious lifestyle; IQ, Intelligence quotient; I-TEQ, International toxic equivalent; LC-PUFA, Long chain polyunsaturated fatty acids; LDL, Low density proteins; LFC, Low fat cholesterol; MetHg, Methyl mercury; MPI, Milk protein isolate; PCB, Polychlorinated biphenyl; PCDD, Polychlorodibenzo-dioxine; PCDF, Polychlorodibenzo-furane; PSA, Prostate specific antigen; PTMI, Provisional tolerable monthly intake; PTWI, Provisional tolerable weekly intake; SCOOP, Scientific cooperation; SPI, Soy protein isolate; SURA, Coastal ocean observing and prediction; TC, Total cholesterol; TEF, Toxic equivalency factors; TEQ, Toxic equivalents; VLDL, Very low density lipoprotein; WHO, World health organization. Corresponding author. Address: ILSI Europe a.i.s.b.l., Avenue E. Mounier 83, Box 6, 1200 Brussels. Tel.: +32 (0) ; fax: +32 (0) address: publications@ilsieurope.be. 1. Introduction BRAFO is the acronym for the benefit-risk assessment of foods. This EU-funded project has developed a framework that allows for the quantitative comparison of human health risks and benefits in relation to foods and food compounds. In a recent publication in FCT we have described this new methodology allowing to integrate benefits and risks of foods and food constituents (Hoekstra et al., 2010). Fig. 1 shows the tiered approach of the BRAFO methodology. In the current publication we describe the application of the new methodology to natural foods. Natural foods usually provide a large variety of essential nutrients and contribute to a healthy nutrition. At the same time natural foods also may contain contaminants as well as antinutritive and bioactive compounds, which could impair health. Therefore, natural foods are good examples to illustrate the need for and the /$ - see front matter Ó 2011 ILSI Europe. Published by Elsevier Ltd. All rights reserved. doi: /j.fct

3 S700 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 Pre-assessment and problem formulation Tier 1 Individual assessment of risks and benefits no benefit no risk Reference scenario Alternative scenario Stop: advise reference Stop: advise alternative both risks and benefits Tier 2 Qualitative integration of risks and benefits no clear dominance risks clearly dominates benefits benefits clearly dominates risks Stop: advise reference Stop: advise alternative Tier 3 Deterministic computation of common health metric worst/bad case analysis Sensitivity analysis Increasingly assessing more and more parameters probabilistically Tier 4 Probabilistic computation relatively small uncertainties Net benefit < 0 advise reference Net benefit > 0 advise alternative large uncertainties Δ Health units Fig. 1. A flow chart of the BRAFO-tiered approach for health risk benefit assessment of different dietary scenarios (reference and alternative). For assessment of risks and benefits within a single scenario, see text. The formulation of the risk benefit question may be iteratively refined in consultation with the risk manager/policymaker as the assessment progresses, as indicated by the dashed arrows at the left side of the figure. process of benefit-risk assessment. The natural foods covered within this article are farmed salmon and soy protein. There is evidence that consumption of fish, particularly oily fish, has substantial beneficial effects on health. As a result the public is advised to increase its fish consumption as in many countries intake is considerably less than optimal. On the other hand it must be noted that oily fish contains hazardous substances as well like dioxins, poly-chlorinated biphenyl (PCBs) and methyl-mercury (methg). Excessive exposure to these substances may for example induce cancer and neurodevelopment alterations in the foetus. As a consequence oily fish is a particularly good example illustrating how consumer confusion can develop regarding dietary recommendations and therefore was selected as a case study for assessing the actual value of the BRAFO approach. Soy has been selected as the second case study of a natural food because it is recognized as a healthy natural food delivering various essential nutrients. In addition, soy intake is associated with a reduced risk for cardiovascular disease (CVD). However, phytochemicals occurring naturally in soy such as the isoflavones can have both, beneficial as well as adverse effects as demonstrated in a number of animal studies. The main purpose of the present paper is to apply the methodology developed within BRAFO (Hoekstra et al., 2010) to farmed salmon and soy protein. Evaluation of the evidence was based on published systematic reviews, meta-analyses, individual studies, and expert judgements. 2. Case study: Farmed salmon 2.1. Pre-assessment and problem formulation According to the BRAFO methodology, the first step sets the scope of the assessment and describes the benefits and risks potentially involved in a particular exposure situation. In the present exercise, focus has been on the comparison of the current situation regarding fish intake (reference scenario) with an alternative scenario consisting of a general consumption of 200 g/week of farmed salmon. In general, benefits and risks can be assessed at the level of the European population at large or at the level of individual countries. Precise intake data for the respective natural foods and consequent exposure to associated contaminants are required. Further, incidence data for those diseases which are linked to the intake of specific foods or contaminants should be available. For most European countries these detailed informations are not available except for the Netherlands. For this country intake data on different fish species have been described, and the current incidences for most human diseases are available ( Therefore, the present benefit-risk assessment for farmed salmon has been conducted mainly based on data available for the Netherlands. Besides different conditions in individual European countries affecting the outcome of the benefit-risk assessment (general fish intake, incidence of myocardial infarction), special segments of the population (young children, women of child-bearing age, pregnant women, nursing mothers, men >50 years) within each European country may show a different benefit-risk assessment outcome for farmed salmon. The literature was searched for beneficial and adverse effects potentially associated with fish intake, using a number of recent scientific reports, but mainly systematic reviews, and meta-analyses (EFSA, 2005; Mozaffarian and Rimm, 2006). The benefit identification for this case study was primarily based on these publications, and further supported by relevant new studies, which were not included in the published systematic reports/reviews. No data from animal studies were included in the benefit identification Fish intake considerations Fish consumption, especially oily fish, within Europe is low. Landings of oily fish species in Europe in 2003 were 18 g/d per inhabitant (EFSA, 2005). The total fish intake at the 99th percentile in various European countries ranged from g/d (EFSA, 2005). In the UK total fish consumption is 31 g/d with 7 g/d derived from oily fish (SACN/COT, 2004). A similar intake level was reported for Germany with a total fish intake of 26 g/d without further differentiation between white/oily fish (Max-Rubner-Institut,

4 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 S ). In the Netherlands 37% of the households reported a total fish intake of at least one serving/week (21% reported at least one serving/week of oily fish) (Pieniak et al., 2008) with an average daily fish intake of between 10 and 13 g (van Dam et al., 2003) (EFSA, 2008). This means that more than 60% of the adult Dutch population does not eat fish on a weekly basis and about 80% do not eat oily fish on a weekly basis Identification of key potentially beneficial components and positive health effects Beneficial effects are primarily caused by n-3 long-chain polyunsaturated fatty acids (n-3 LCPUFA), selenium, iodine and vitamin D. The benefit identification and characterization will be restricted to the intake of oily fish as a whole food and to the intake of n-3 LCPUFA. The mean content of n-3 LCPUFA in salmon is 2.99 g/ 100 g (range ) (EFSA, 2005). Besides oily fish there are no other dietary sources of these fatty acids. The consumption of oily fish has been associated with a number of positive health effects. These include improved eye and brain development in early life, decreased risk of coronary heart diseases (CHD), prevention of age-dependent macular degeneration, improvement of blood lipid composition, and various forms of arthritis (EFSA, 2005; Mozaffarian and Rimm, 2006; SACN/COT, 2004). Currently, the best documented positive health effects of the intake of oily fish relate to endpoints of CHD and brain development. According to a recent meta-analysis a daily intake of n-3 LCPUFA with oily fish is consistently associated with a decline in CHD risk (Mozaffarian and Rimm, 2006). Compared with little or no intake, modest consumption of n-3 LCPUFA ( mg/d) lowers relative CHD risk by 25% or more (Mozaffarian and Rimm, 2006) (Fig. 2) Identification of potential hazardous components and toxicological effects Potential adverse effects relate to the intake of contaminants including methg, PCB and dioxin. Amongst others, methg and dioxin-like (DL) compounds (dioxins and dioxin-like PCBs) revealed to be the most critical since high consumers of certain fish may exceed tolerable/safe levels even without taking into account other sources of dietary exposure (EFSA, 2005). Consequently, for the present exercise, focus has been on methg and DL compounds. Other contaminants were not considered of significant health concern since intakes did not contribute significantly to the overall exposure and/or were unlikely, even in high fish consumers, to exceed the relevant tolerable intakes (EFSA, 2005) Mercury. Hg may occur in food as inorganic and organic forms. Although there are common mechanisms of toxicity, organic Hg is considered to be the most toxic following ingestion. Studies RR fatal CHD ; 1 100; ; ; ; DHA+EPA intake mg/day Fig. 2. Relative risk of fatal CHD as a function of DHA + EPA intake (1). have shown that in fish methg is the predominant form of Hg, although the percentage with total Hg may vary according to the type of fish. For the present exercise, total Hg was assumed to be in the form of methg Hazard identification and characterization The primary target of methg is the central nervous system. In both laboratory animals and humans, neurodevelopment is considered to be the most sensitive health outcome and in utero phase the most susceptible period of exposure. In the last decade, methg has been evaluated by several expert panels (EFSA, 2004; US FDA, 2009; USNRC, 2000; WHO/JECFA, 2004). A number of human studies are available providing valuable information on dose response relationships. JECFA has established a provisional tolerable weekly intake (PTWI) of 1.6 lg/kg BW/week (EFSA, 2004) Occurrence and exposure The combination of the Scientific Cooperation (SCOOP) and the Coastal Ocean Observing and Prediction (SURA) data on fish and seafood contamination by methg allowed calculating a weighed mean content of 109 +/ 845 lg/kg food (ppb) of total Hg (EFSA, 2004). The high standard deviation reflects the wide variations in the analytical results due to the significant differences amongst methg levels in different fish and seafood. SCOOP data indicated a weighed average of methg for salmon of 40 ppb. Mean methg contents for farmed Atlantic salmon between 1995 and 2003 were found to range from 8 to 52 ppb (EFSA, 2005). Data from USA indicate values for farmed salmon significantly lower than 50 ppb (Ginsberg et al., 2009), and for the UK 40 ppb have been reported (Zeilmaker et al., 2010). No specific data on methg contamination of farmed salmon were available for the Netherlands. Based on SCOOP data, average exposure to methg of 0.1 lg/kg BW/week was estimated for the Netherlands (EFSA 2004). This figure could be used as a reference exposure Dioxins and dioxin-like compounds. In the present paper, the terms dioxins and dioxin-like compounds refer to a group of chemicals known to elicit toxicity through a similar mode of action involving the activation of the Aryl-hydrocarbon receptor (AhR). This group of chemicals comprises poly-chlorinated dibenzo-pdioxins (PCDDs, dioxins), poly-chlorinated dibenzofurans (PCDFs, furans) and non-ortho and mono-ortho substituted poly-chlorinated biphenyls (dioxin-like PCBs, DL-PCBs). In almost all food matrices, including fish, dioxins and dioxin-like PCBs are present as complex mixtures. For the purpose of risk assessment, analytical data for individual congeners are converted into toxic equivalents (TEQs). It is based on the hypothesis that the dioxins and dioxinlike compounds are acting through the AhR, producing similar qualitative effects but with different potencies. The use of this concept assumes that the doses of the different congeners are additive. The TEQs are obtained by multiplying analytical results for the individual congeners by appropriate toxic equivalency factors (TEFs) estimated in relation of the most toxic congeners (2,3,7,8- TCDD), which is assigned the arbitrary TEF of one. Different TEFs have been proposed (international TEFs, WHO-TEFs) resulting into slightly different TEQ values for some specific congeners. TEQ values based on WHO-TEFs are about 10 20% higher than those obtained with the international TEFs (I-TEFs) (DG Sanco 2000) Hazard identification and characterization It is widely accepted that for PCDDs, PCDFs and DL-PCBs, the activation of the AhR is a necessary but not sufficient requirement for the elicitation of toxicity, including cancer. The issue of the shape of the dose response at low level of exposure to chemical substances inducing receptor-mediated toxicity such as dioxins is still debated. In the present assessment, it has been assumed that

5 S702 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 for dioxins and dioxin-like compounds there is a threshold for all reported effects. In both laboratory animals and human, carcinogenicity, immunotoxicity, behavioural and reproductive toxicity have raised the most significant concerns. There is a large body of scientific literature describing the toxicological potential of dioxin-like compounds. In the last decade, dioxins and dioxin-like compounds have been evaluated by several expert panels. A number of human studies are available. There is currently no consensus on how to exploit these data in order to define pivotal effects and associated dose response relationships (EFSA, 2004; US-NRC, 2000; WHO/ JECFA, 2004). The EU Scientific Committee on Foods has established a PTWI of 14 pg WHO-TEQ/kg BW. More recently, The WHO Joint expert Committee on Food Additives and Contaminants (JECFA) allocated a Provisional Tolerable Monthly Intake (PTMI) for dioxin of 70 ng/ kg BW/day (WHO/JECFA, 2004) Occurrence and exposure Relevant information is available from the EU SCOOP initiative, which collected data from 10 European countries. In fish, dioxins and DL-PCBs exhibited a wide distribution spreading over two to three orders of magnitude (DG Sanco 2000). This has been explained by the variety of fish species considered (belonging to different levels of trophic web), including farmed, wild fish, and fish from various regions. On average, reported values of PCDD and PCDF levels were in the order of 10 pg International Toxic Equivalents (I-TEQ)/g of fat, while DL-PCBs were estimated to contribute between one to three times the TEQ contribution of PCDDs and PCDFs. Specific occurrence data for farmed fish (mainly salmon and trout) indicated average levels of 0.5 ng and 1.36 ng of WHO-TEQ/kg fresh weight for dioxins + furans and DL-PCBs, respectively, resulting in a total average level of 1.87 ng WHO-TEQ/kg fresh weight (EFSA 2004). The 95th percentiles were about twice the average values. It is usually considered that more than 90% of human exposure to dioxins and DL-PCB derives from food. Total fish intake contributes to 12% of dioxin and DL-PCBs in the Dutch diet, major contributors are dairy (38%), meat (17%), and vegetable oils (17%) (De et al., 2008). For the Netherlands the dioxin concentration (expressed as WHO-TEQ) of salmon is 1.11 ng/kg (van der Voet et al., 2007a) Pre-assessment and problem formulation: Summary With the current intake of salmon or oily fish in Europe at a rather low level, we defined no intake as reference scenario. As an alternative scenario for the benefit-risk assessment a weekly intake of farmed salmon of 200 g by the general population was chosen. Using the above pre-assessment, it was decided for the present exercise to base the risk benefit on the following parameters: Oily fish/n-3 long-chain polyunsaturated fatty acids (n-3 LCPU- FA) and their potential positive health effects on: Primary and secondary prevention of Coronary Heart Diseases (CHD) Improved cognitive development in infants (gain in Intellective Quotient (IQ) points) Retardation of cognitive decline in the elderly reduced cancer risk reduced risk of age-related macular degeneration (AMD) methg and its resulting impaired cognitive development in infants (loss in IQ points) Dioxin-like compounds and their potential increased cancer risk 2.2. BRAFO tier 1 level The objective of tier 1 is to identify health consequences associated with the change from the reference scenario to the alternative scenario through the application of benefit and risk assessment as conducted independently (Fig. 1). Table 1 summarizes the major health effects of a change from the reference to the alternative scenario, the quality of the evidence, and the population affected based on recent publications. Clearly, a number of health benefits could be induced by increasing the intake of farmed salmon. At the same time adverse effects including increased incidence of cancer as well as impaired infant development may occur. However, in order to assess the adverse effects, exposure to the contaminants in farmed salmon has to be related to tolerated (safe) intake levels. Assuming the consumption of 200 g of farmed salmon per week contaminated at an average level of 50 ppb of methg, an average exposure of 0.17 lg/kg BW/week (for a 60 kg adult) can be calculated. Focus has been on mean occurrence values based on the fact that regular consumers are unlikely to only ingest fish with the highest level of methg contamination, but more likely they would consume fish for which contamination is randomly distributed. Therefore, the alternative scenario would not exceed the PTWI of 1.6 lg/kg BW/week. For dioxin and DL-PCBs the consumption of 200 g of farmed salmon per week contaminated at an average level of 1.11 ng WHO- TEQ/kg fresh weight, an average exposure of about 3.7 pg WHO- TEQ/kg BW/week can be calculated. Again, focus has been on mean occurrence values for farmed fish based on the fact that regular consumers are unlikely to only ingest fish with the highest level of dioxin contamination, but more likely they would consume fish for which contamination is randomly distributed. The DL compounds from the fish of the alternative scenario would result in an exposure corresponding to about 25% of the PTWI of 14 pg WHO-TEQ/kg BW (and about 21% of the JECFA PTMI). However, the actual risk associated with DL-compounds must take into account the total exposure through all sources. Summing up the average basal exposure level and the additional exposure from the fish of the alternative scenario results in an intake still below the available health based guidance values, with the alternative scenario contributing to about 37% of total exposure (Table 2). Table 2 also indicates that for consumers at the highest percentiles (>P95) of basal exposures, the addition of an intake from the fish of the alternative scenario would result in total exposures equal or slightly above the available health based guidance values. For example, at the P99 of the exposure curve, an excursion above the PTWI or PTMI of about 50% is calculated with the alternative scenario. However, the actual impact of the alternative scenario on the overall DLcompound intake of consumers at the highest percentiles of the distribution curve of the background DL-compound exposure is difficult to assess. At P99, the exposure is already higher than the available health based guidance values. In this situation, the contribution of the fish intake recommended by the alternative scenario to the total exposure would be relatively limited, about 16%. Furthermore, 200 g/week of salmon would certainly displace other foods such as other fish species or meat, which may likely be more contaminated in average than the farmed salmon considered in the alternative scenario. Therefore, in general, the implementation of the alternative scenario seems unlikely to affect total exposure enough to trigger any adverse effects in consumers. This could be further documented through the application of exposure modelling approaches taking into consideration potential replacement of foods by the 200 g of farmed salmon. In summary, tier 1 strongly suggests that the alternative scenario is likely to provide some health benefits, while exposure to main contaminants would stay within safety values. No further

6 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 S703 Table 1 Case study on farmed salmon benefit-risk assessment for a change from the reference scenario to the alternative scenario at tiers 1 and 2. Tier 1 Tier 2 Health effect Change Quality of evidence based on WHO criteria Incidence of death from CHD (myocardial infarction) or stroke Healthy infant development Cancer incidence Improved cognitive functions Prevention of AMD Impaired infant development Reduced incidence of fatal myocardial infarctions or strokes (up to 36%) Increase in IQ and visual acuity (benefits averages few IQ points) No effect on cancer incidence Reduced incidence of dementia/ Alzheimer s disease (database does not allow quantitative estimates) Reduced incidence of AMD (database does not allow quantitative estimates) Decrease in IQ (impairment averages few IQ points) Magnitude of the effect Population affected Health impact (beneficial/ adverse/ None) References to the quality of evidence Convincing Strong Adults >50 Beneficial (SACN/COT, 2004; US FDA, 2009; EFSA, 2005) (Whelton et, al., 2004); (He et, al., 2004); Possible Small Infants Beneficial (Mozaffarian and Rimm, 2006) (Cohen et al., 2005) (SACN/COT, 2004) Number of individuals (current incidence in the Netherlands a ) Effect severity a (w = 0 1) The cognitive development benefit averages few IQ points per newborn baby Probable Small Adults Beneficial or adverse Maclean et al., 2006, Possible Small Adults Beneficial (Lim et al., 2006) (Issa et al., 2006) (Fotuhi et al., 2009) (Kroger et al., 2009) (Albanese et al., 2009) Possible Small Adults Beneficial (Chong et al., 2008) b Possible Small Infants Adverse (Mozaffarian and Rimm, 2006)(SACN/COT, 2004) US FDA (EFSA, 2005) (Cohen et al., 2005) When the quality of evidence for a certain health effect is considered insufficient it is advised not to go to Tier 2. a b Global burden of disease. WHO The cognitive development impairment averages few IQ points per newborn baby integrated benefit-risk assessment would therefore be necessary and the evaluation could stop already at that stage, allowing a recommendation of the alternative scenario. However, from a risk perspective, some uncertainties still remain. For DL-compounds, they mainly concern intake for the highest exposed populations. Regarding methg, it may be argued that no threshold for developmental effects exists. These uncertainties require to advance to the tier 2 of the BRAFO methodology BRAFO tier 2 level Table 2 Dioxin exposure levels (pg WHO-TEQ/kg BW/day) of average and above average consumers for the reference scenario and the alternative scenario. Reference scenario Alternative scenario Only fish consumption Exposure levels average consumers (all food items) Exposure levels above average consumers (P95) ,6 Exposure levels of extreme consumers (P99) In tier 2, the benefits/risks have to be characterized based on the severity of the effects and on the potential number of individuals affected (Hoekstra et al., 2010). Health effects with a high severity included CHD, cancer, cognitive decline, and a medium to low rank was attributed to the effects healthy infant development and age-related macular degeneration (Table 1). Since CHD is one of the leading causes of death and fish intake is inversely associated with this endpoint, a high number of individuals may benefit from the alternative scenario. A dose response relationship between the intake of n-3 LCPUFA and the relative risk for fatal CHD is shown in Fig. 2. A weekly intake of 200 g of farmed salmon would provide 854 mg/d of n-3 LCPU- FA (EFSA 2004). As indicated in Fig. 2, such an intake of n-3 LCPUFA is associated with a significant decline in the relative risk of fatal CHD. For the contaminant methg the available data indicate that the most documented adverse effect in humans is neurodevelopmental impairment resulting from prenatal maternal exposure. Other potential effects are still equivocal. Despite the great variability of settings and outcomes, substantial degree of concordance exists across neurodevelopmental studies, and the combined evidence appears as convincing in regard to the dose response relationship between methg and adverse effects. Assuming that methg has a threshold of effects, Reference Dose (RfD) and PTWI ranging from 0.7 lg/kg BW/week (US NRC, 2000) to 1.6 lg/kg BW/week (WHO/ JECFA, 2004) have been established using the data of the three major cohort studies (Davidson et al., 2006; Grandjean, 1997; Kjellstrom et al., 1989). Two analyses integrating data from the three major epidemiologic studies available (New Zealand, Faroe and Seychelles) re-

7 S704 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 sulted in the establishment of quantitative dose response relationships between methg maternal body burden and neurodevelopmental effects (Kjellstrom et al., 1989); (Grandjean, 1997). In these approaches a linear relationship between Hg body burdens and neurodevelopmental outcomes was assumed. Various methods to convert dose response slopes from cognitive tests into impact on IQ were applied. Cohen et al. (2005) reported that prenatal exposure to methg sufficient to increase the concentration of Hg in maternal hair of 1 ppm decreases IQ by 0.7 points (central estimate), with a plausible range of values for this loss of 0 to 1.5. In the discussion of the paper, it is considered that 0.2 IQ points/ppm in maternal hair maybe most appropriate. Axelrad et al. (2007) reported a central estimate of 0.18 IQ points per ppm increase of maternal Hg (95% confident interval: to 0.009). Body burden biomarkers such as level of Hg in maternal hair can be converted into external exposure (USNRC 2000), with 1 ppm corresponding to an average dietary exposure of 0.42 lg/kg BW/week. On the other hand, n-3 LCPUFA induce a gain in IQ points and the overall effect of an increased intake of farmed salmon would be a slight benefit in IQ points as calculated within the QALIBRA project (Zeilmaker et al., 2010). A recent meta-analysis indicated no significant association between higher methg exposure and risk of CHD (Mozaffarian and Rimm, 2006). At the tier 1 level we have concluded for dioxin and DL-PCBs that the available information indicates that for the general population, average exposures range within the JECFA PTMI (van der Voet et al., 2007b). High intake consumers with additional 200 g/ week of farmed salmon would possibly slightly exceed the safe level. As discussed before, more data and analysis are required to characterize more precisely the impact of the implementation of the alternative scenario to the overall exposure to DL compounds. In addition, the actual health significance of an excursion above PTWI or PTMI would be difficult to assess. As highlighted before, dose response curves associating dietary DL-compound exposure and adverse effects in humans are not readily available. More data exist in experimental animals. However, no standard methodology is currently developed to extrapolate human dose response from animal data in the context of the BRAFO benefit-risk assessment methodology. For the benefit-risk assessment of the extreme consumers, to proceed to tier 3 may be theoretically recommended, where beneficial and adverse effects would be assessed and integrated quantitatively. However, the necessary data are unlikely to be available Summary and conclusion Based on the present analysis and calculations it is concluded that the alternative scenario will result in a significant reduction of CHD incidence. Although this scenario will increase the intake of contaminants, for methg as well as for dioxin and DL-PCBs the intake will still be within the PTWI or PTMI for a significant proportion of the consumer population. Therefore, it can be concluded from the tier 1 already that the benefit outweighs the risk and that the alternative scenario can be recommended over the reference. For high consumers at the highest range of the DL-compound intake, exposures may be slightly higher than the health based guidance values, but the actual contribution of the alternative scenario is likely to be small. Additional investigations as indicated by the next BRAFO tier steps could be valuable to further characterise some specific benefit-risk aspects in consumers at the highest range of the background DL-compounds exposure. However, this was prevented by the lack of sufficient data. 3. Case study: Soy protein 3.1. Background Soybeans are a good source of high quality vegetable protein (containing adequate amounts of all essential amino acids) of vitamins and minerals (vitamins B and E, calcium, magnesium, boron). They are low in fats, which are primarily unsaturated n-6 and n-3 fatty acids. In addition they contain a range of phytochemicals including isoflavones, phytosterols and saponins, and oligosaccharides. Soybeans are almost invariably processed before consumption to produce traditional foods, such as tofu and soy milk, and more westernized products such as yoghurts and desserts. Such foods can be derived from processing the whole bean, which retains most of the original components of the beans, or by extracting the protein. The process used to produce isolates of soy protein involves several protein purification stages at various ph levels, which result in losses of certain phytonutrients, e.g., oligosaccharides (raffinose and stachyose), vitamin E, and phytosterols, although the soy isoflavones are usually retained. Many of the benefits and risks ascribed to soy consumption relate to the intake of the soy isoflavones, which are diphenolic compounds possessing weak estrogenic activity, hence the name phytoestrogens. Such activity has resulted in promotion of soy and isolated isoflavones as alternatives to hormone therapy for menopausal symptoms including hot flushes and bone loss. Furthermore, the hormonal activity associated with isoflavones has stimulated much research into benefits (and in some cases risks) for breast and prostate cancers Pre-assessment and problem formulation Consumption of soy foods and in particular the bioactive isoflavones that they contain (in varying amounts depending on source of beans, type of food and, processing) have been claimed to have human health benefits and risks in the following main areas: The potential benefits are: Reduced breast cancer risk Reduced prostate cancer risk Improved bone health (osteoporosis) Reduced cardiovascular disease risk Improved cognitive functions. The potential risks are: Disruption of sex hormone levels and changes in reproductive function Impaired thyroid function Increased breast cancer risk Impaired cognitive functions. The evidence for the above effects are derived from (a) epidemiology studies that are based on soy food consumption estimated either by dietary assessment or by urinary/blood isoflavones measurements and (b) intervention studies using soy foods or isoflavones supplements. There have also been extensive animal and in vitro studies using soy and purified isoflavones that provide additional evidence for beneficial and adverse effects and mechanisms, but this paper will focus exclusively on evidence from human studies wherever possible. It should be noted that soybeans contain a variety of bioactives in addition to isoflavones (e.g., protease inhibitors, oligosaccharides,

8 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 S705 and peptides), so there can be discrepancies between results of studies using whole soy foods and isolated isoflavones. It should also be noted that the epidemiological evidence for protective effects of soy consumption is usually stronger in Asian populations than in Western populations. This is most likely a consequence of the higher level of intake of the former rather than any ethnic differences i.e., the range of intake of soy in Western population is insufficient to elicit demonstrable biological effects. In addition, a high intake of soy in Asian populations is associated with specific dietary patterns, which differ in many ways from typical Western dietary patterns. The European Prospective Investigation into Cancer and Nutrition (EPIC) study on soy consumption has found a generally low soy product intake in 10 European countries by using a standardised reference dietary method (Keinan-Boker et al., 2002). Intake of soy products was investigated among participants aged from 34 to 74 with inclusion of subjects with a habitual health-conscious lifestyle (HHL), i.e., non-meat eaters who are fish eaters, vegetarians or vegans. Although soy product intake was generally low across all countries, the highest intake level was observed in the UK, due to over-sampling of a large number of participants with an HHL. The most frequently consumed soy foods were dairy substitutes in the UK and France and beans and sprouts among mid-european countries. For both genders, soy dairy substitutes were consumed in the highest quantities (1.2 g/day for men; 1.9 g/day for women). Participants with an HHL consumed higher quantities of almost all soy product groups. With the current intake of soy in Europe, we defined no intake as reference scenario, as an alternative scenario for the benefit-risk Table 3 Case study on soy protein tiers 1 and 2. Tier 1 Tier 2 Health effect Change Quality of evidence Changes in cognitive/ memory function Change in blood cholesterol Change in blood pressure Decreased breast cancer risk Decreased prostate cancer risk Improved bone health Increased breast cancer risk Disruption of sex hormone levels/ reproductive function Impaired thyroid function Overall change Reduction in LDL cholesterol Reduction in diastolic blood pressure Decrease in cancer incidence/ mortality Decrease in cancer incidence/ mortality Maintained bone mineral density, improved bone markers Increased mortality in BC patients or increased growth of precancerous lesions Low testosterone, reduced sperm count Magnitude of the effect Population affected Health impact (beneficial/ adverse/ none) Insufficient Very small European population Beneficial and adverse Convincing Moderate 6% reduction in LDL cholesterol Possible Moderate - 2mm reduction in diastolic presssure Probable Low at individual level (RR ). Larger effect if soy consumed in adolescence) Possible Low at individual level (RR 0.74) Possible Insufficient from animal studies indicating adverse effects. Probable from human studies indicating no effect Animal data possible, Human data possible (for no effect) Animal data possible human data indicating no effect possible Low at individual level Moderate at population level (RRs 0,64 0,79) Very low European population with hypercholesterolaemiat European population with hypertension Females (pre and postmenopausal) Beneficial Beneficial Beneficial References to the quality of evidence (Ho et al., 2007) (Kreijkamp- Kaspers et al., 2004) (Harland et al., 2008) (Hooper et al., 2008) (Trock et al., 2006) (Wu et al., 2008) (Wu et al., 2002) Males Beneficial Yan et al., 2009 Females and males Beneficial (Liu et al., 2009) Females with breast cancer/preneoplastic lesions Adverse (Shu et al., 2009) Very low Males Adverse (Hamilton- Reeves et al., 2009) Very low Subjects with congenital hypothyroidism Adverse Beneficial (Dillingham et al., 2007) Effect severity (w = 0 1) Number of individuals 0 741,000 deaths form CHD in EU 1.1 m stroke incidence in Europe 0.09 Incidence cases in EU cases in EU25 Fractures: 0266 vertebra femur * 0372 short term0,272 long term 19 mio. affected EU25 All osteoporotic fractures 3,79 mio hip fractures/ y 0.09 Incidence cases in EU25 * Hip fractures not in Global burden of disease table 2004, substituted w. femur fracture.

9 S706 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 assessment a daily intake of soy protein of 25 g was defined. The European population at large was chosen as target population BRAFO tier 1 level At tier 1, the aim is to identify benefits and risks associated with changes from the reference scenario to the alternative scenario. These adverse and beneficial health consequences are summarized in Table 3 in terms of the quality of the evidence, and the magnitude of the effect Positive health effect identification and characterization The evidence for the potential positive health effects is largely derived from epidemiological studies (including a number of meta-analyses) and dietary intervention studies with soy, soy protein and purified isoflavones, with supporting data and information on mechanisms being obtained from animal models and in vitro studies. It is noteworthy also that soy foods are a common component of the diet in Asian countries resulting in isoflavones consumption of mg/d by comparison to intake in Western societies of less than 1 mg/d Breast cancer In recent years there has been considerable interest in the relationship between soy consumption and breast cancer risk, largely because of the weak oestrogenic activity of soy isoflavones. A meta-analysis (Trock et al., 2006) of 18 epidemiologic studies that examined soy exposure and breast cancer risk, found that high soy intake was associated with a modest, but significant, reduction in breast cancer risk Odds Ratio (OR) = 0.86; 95% CI = A more recent meta-analysis by Wu et al. (2008), which included eight studies in Asian populations with high soy intake found a more protective effect of soy food intake with OR of 0.71 (95% CI = ) for those with highest soy food vs lowest intake, although no protective effects were observed in studies involving non-asians, most likely because soy food intake was too low to exert biological effects. Three studies have evaluated the impact of soy food intake during childhood and adolescence, and indicate that soy intake in early life may offer greater protection against subsequent breast cancer with relative risks (RR) and OR between 0.51 and 0.65 (Shu et al., 2001; Lee et al., 2009; Wu et al., 2002). Three epidemiological studies have investigated soy or isoflavone intake and breast cancer survival. Two (one in China, one in USA) found no effect (Boyapati et al., 2005; Fink et al., 2007). In the Shanghai Breast Cancer Survival Study, which was a 3.9 y follow up of 5033 surgically-treated breast cancer patients (Shu et al., 2009), soy food intake was inversely associated with mortality and recurrence for both ER+ and ER breast cancer Prostate cancer A recent meta-analysis (Yan et al., 2009) of cohort and case control studies that examined the relationship between intake of soy (15 studies) or soy isoflavones (9 studies) and prostate cancer risk also found a modest protective effect combined RR/OR of 0.74 (95% CI: 0.63, 0.89; P < 0.01) which was dominated by studies on nonfermented soy Cardiovascular disease The potential beneficial effect of soy on CVD risk is mainly based on extensive dietary intervention trials and meta-analyses (6 in total) investigating the effect of soy on blood lipids. In 1995, a metaanalysis of 38 controlled clinical studies concluded that substituting soy protein isolates containing isoflavones for animal protein significantly lowered total cholesterol (9.3%), LDL cholesterol (12.9%), and triglycerides (10.5%) without affecting HDL cholesterol (Anderson et al., 1995). Subsequent clinical trials have reported less potent hypocholesterolaemic activity of soy protein (Harland et al., 2008) possibly due to the use of cohorts with lower initial serum cholesterol concentrations. The mechanism involved in the lowering of LDL cholesterol by soy protein isolates has proved somewhat elusive. In general soy protein washed free of isoflavones (and possibly other components) does not exert hypocholesterolaemic effects, but neither do purified isoflavones. These observations have led to a negative opinion by EFSA, which did not consider that soy protein per se could influence plasma cholesterol. It is clear however that soy protein that has not been treated to remove isoflavones, has hypocholesterolaemic effects, albeit modest Bone health Evidence for improvements in bone health by soy is derived from epidemiology and dietary intervention trials mostly with bone mineral density as the prime endpoint. The epidemiological studies (in Asian populations) revealed soy intake to be associated with higher BMD (Ho et al., 2001, 2003; Somekawa et al., 2001) and one study found an inverse association with fracture risk (Zhang et al., 2005). Intervention studies with soy, soy protein and isoflavones have yielded variable results (Ma et al., 2008; Liu et al., 2009) Hazard identification and characterization The evidence for adverse effects of soy and soy isoflavones is largely based on animal studies or even in vitro experiments. Concern has been expressed that the estrogenic effects of isoflavones in soy may stimulate the growth of existing oestrogen-sensitive breast tumours or precancerous lesions and increase risk of women at elevated risk of breast cancer. Such concerns are based on animal studies in which genistein, or soy containing genistein, were found to stimulate the growth of mammary tumours in ovariectomized athymic nude mice that had been implanted with MCF-7 estrogen-sensitive breast cancer cells. Human studies, in contrast, both epidemiological and intervention are reassuring in terms of safety Breast cancer Four intervention studies in healthy subjects and breast cancer patients show no effect of soy consumption on breast epithelial cell proliferation, apoptosis, mitosis or Bcl-2 expression, even at soy protein intakes of 60 g/d (Hargreaves et al., 1999). A number of studies indicate that isoflavones do not affect mammographic breast density, a marker that has consistently been associated with risk for breast cancer (Powles et al., 2008) Reproductive effects The weak oestrogenic activity of soy isoflavones has raised concerns that exposure to large amounts of isoflavones may result in hormonal imbalances with consequences that include reproductive abnormalities and sexual development issues, in particular male fertility and testis development. Most of the evidence supporting such concerns derives from isoflavone studies in rodents, with routes of administration (sub cutaneous or intraperitoneal) not relevant to human exposure (Chen et al., 2004). The main concern is focused on soy milk formula fed babies, which unlike adult soy consumers, are exposed to very high levels of isoflavones per kg body weight. Tan et al. (2006) studying soy formula feeding on testicular development in marmosets which exhibit a period of neonatal testicular activity similar to that in the human male found no gross adverse reproductive effects, in fact testis weight and cell numbers were increased. Another primate study (Perry et al., 2007) confirmed the lack of effect of soy protein with or without isoflavones fed for 3 years on male fertility parameters in adult male cynomolgus macaques. Human data in this area are scarce. One retrospective study of infants fed soy formula milk (Strom et al., 2001) found no evidence for any effects on reproductive development in men or women and

10 B. Watzl et al. / Food and Chemical Toxicology 50 (2012) S699 S709 S707 a 2 months intervention with a 40 mg isoflavone supplement in healthy men from UK revealed no effect on semen quality or on blood sex hormone and gonadotrophin levels (Mitchell et al., 2001). In contrast a recent study has reported a negative relationship between soy consumption (as determined by food frequency questionnaire) and sperm concentration in male partners of couples that were being treated in a fertility clinic (Chavarro et al., 2008). Although sperm concentration was decreased, there was no significant effect on total sperm count, and other semen quality parameters (sperm motility, sperm morphology and ejaculate volume) were also not affected. A meta-analysis (Hamilton Reeves et al., 2009) of 15 placebocontrolled trials in which adult men consumed soy foods, isolated soy protein, or isoflavone extracts revealed no significant effects on circulating testosterone, free testosterone, sex-hormone binding globulin, or free androgen index. Overall therefore, it seems likely that the effects of soy protein consumption by adult males will have little or no effect on reproductive function and hormone levels Goitrogenic effects Data from studies in rodents deficient in iodine suggest that relatively high levels of dietary soy may have a goitrogenic effect, possibly through a potential interaction of phytoestrogenic isoflavones with thyroid gland. However, human dietary intervention studies indicate that soy foods, soy protein or isoflavones are unlikely to affect thyroid function in normal individuals with adequate iodine intake. For example, a review by Messina et al. (2006) identified 14 trials in healthy humans in which at least one measure of thyroid function was assessed (although thyroid function was not the primary aim of the studies). With one exception, either no effects or only modest changes were noted in these trials. These results have been confirmed recently in two randomized, placebocontrolled trials with thyroid-related measures as the primary outcome. There is some evidence (Belle et al., 2001) that soy protein preparations can increase the oral dose of T4 required to achieve therapeutic effects in hypothyroid patients, due to impairment of absorption. However the problem can be overcome by separating in time the consumption of soy and dose of T4 (Belle et al., 2001). Overall the above findings indicate that in iodine-repleted individuals with normal thyroid function, soy protein or isoflavones do not adversely affect thyroid function Effects on brain functions A cohort study in Japanese American male adults, found poor cognitive test performance, enlargement of ventricles and low brain weight associated with high mid-life tofu intake (ORs 1,6 2,0) (White, 2000). The effect was small, explaining less than 1% of the variance in performance score, (age, education and stroke history explained 28%). A cross-sectional study in Asian elderly subjects showed high tofu consumption was associated with memory decline, but high tempeh intake associated with better memory (Hogervorst et al., 2008). It has been postulated that preservation with formaldehyde as used in the past in certain areas of Asia, may have been a contributing factor for these findings (White, 2000). Intervention studies with soy protein and isolated isoflavones (some over periods of several months) do not provide supporting evidence for these epidemiology findings. One reported a small decline in working memory associated with soy milk intake (<25 g/ day) (Fournier et al., 2007), but the other studies show either no effect (three studies; (Basaria et al., 2009; Ho et al., 2007; Kreijkamp-Kaspers et al., 2004) or improvement (five studies) in various cognitive measures (Duffy et al., 2003; File et al., 2001; Gleason et al., 2009; Kritz et al., 2003; Thorp et al., 2009). Thus, the evidence for any impact on cognitive function adverse or beneficial is not strong for soy protein. From these considerations at the tier 1 level, it is apparent that the alternative scenario involves both potential benefits and potential risks compared with the reference scenario (Table 3). However, the evidence for benefits is derived mostly from epidemiology and human intervention studies. Furthermore for each of these beneficial outcomes there have been sufficient studies to conduct metaanalyses, so the quality of evidence is in many cases probable or even convincing. In contrast the data on risks are mostly derived from animal models and are usually not supported by data from human studies so the quality of evidence ranges from insufficient to possible. It is also important to note that the magnitude of the beneficial and adverse effects is considered to be at best moderate, but in most cases low or very low (Table 3) BRAFO tier 2 level In tier 2 the benefit-risk effects have been characterized on the basis of severity of the effects and on the number of individuals affected Benefit assessment The potential benefits, although modest in terms of risk reduction, are in relation to common chronic diseases such as CVD (about 740,000, deaths in EU), breast cancer (320,000 deaths) and prostate cancer (310,000 deaths) and hence have large impact on the number of individuals and thus public health in general Risk assessment The potential risks associated with soy product consumption have generally been deduced from animal or in vitro investigations and have not been substantiated by evidence from human studies. Most of the hazardous potentials of soy products have been associated with the presence of isoflavones (AFSSA, 2005; Munro et al., 2003; COT, 2003). In 1999 (FDA, 1999), the US FDA approved the health claim Diets low in saturated fat and cholesterol that include 25 g of soy protein a day may reduce the risk of heart disease. In addition, FDA concluded that isoflavones present in the protein isolate posed no safety concerns. Based on their figures of 2 mg isoflavones per g soy protein, this would correspond to 50 mg isoflavones. Other expert groups have addressed the safety of soy and soy isoflavones. The consumption of a high soy diet corresponding to isoflavone intakes up to mg/day has not been associated with adverse health effects (Munro et al., 2003). The French Food Safety Authority (AFSSA) considered that an intake of 1 mg/kg BW/day of aglycone isoflavones does not indicate any risk for the general population (AFSSA 2005). In soy protein isolates, levels of isoflavones ranging from 0.1 to 2 mg/g protein have been reported (COT 2003; FDA 1999). The consumption of 25 g of soy protein would result into an isoflavone exposure of maximum 50 mg/day, a level within the range considered by various expert groups to be safe for the general population. 4. Conclusions The tier 2 assessment indicates that the beneficial effects clearly dominate the potential risk. It is not advised to go to tier 3 since the overall health impact is negligible. The alternative scenario would result in an overall benefit for the general adult population Overall conclusions The BRAFO-tiered approach was applied to two natural foods, farmed salmon and soy protein. The available scientific evidence