Assessment of dietary intake of flavouring substances within the procedure for their safety evaluation: advantages and limitations of estimates obtained by means of a per capita method

Food and Chemical Toxicology 43 (2005) 105–116 www.elsevier.com/locate/foodchemtox

Assessment of dietary intake of flavouring substances within the procedure for their safety evaluation: advantages and limitations of estimates obtained by means of a per capita method D. Arcella, C. Leclercq

*

Human Nutrition Department, INRAN, National Research Institute for Food and Nutrition, Via Ardeatina 546, 00178 Rome, Italy Received 15 April 2004; accepted 30 August 2004

Abstract The procedure for the safety evaluation of flavourings adopted by the European Commission in order to establish a positive list of these substances is a stepwise approach which was developed by the Joint FAO/WHO Expert Committee on Food Additives (JECFA) and amended by the Scientific Committee on Food. Within this procedure, a per capita amount, based on industrial poundage data of flavourings, is calculated to estimate the dietary intake by means of the maximised survey-derived daily intake (MSDI) method. This paper reviews the MSDI method in order to check if it can provide conservative intake estimates as needed at the first steps of a stepwise procedure. Scientific papers and opinions dealing with the MSDI method were reviewed. Concentration levels reported by the industry were compared with estimates obtained with the MSDI method. It appeared that, in some cases, these estimates could be orders of magnitude (up to 5) lower than those calculated considering concentration levels provided by the industry and regular consumption of flavoured foods and beverages. A critical review of two studies which had been used to support the statement that MSDI is a conservative method for assessing exposure to flavourings among high consumers was performed. Special attention was given to the factors that affect exposure at high percentiles, such as brand loyalty and portion sizes. It is concluded that these studies may not be suitable to validate the MSDI method used to assess intakes of flavours by European consumers due to shortcomings in the assumptions made and in the data used. Exposure assessment is an essential component of risk assessment. The present paper suggests that the MSDI method is not sufficiently conservative. There is therefore a clear need for either using an alternative method to estimate exposure to flavourings in the procedure or for limiting intakes to the levels at which the safety was assessed.
2004 Elsevier Ltd. All rights reserved. Keywords: Flavourings; Per capita method; MSDI method; Safety evaluation; Exposure assessment; Intake

Abbreviations: ADI, acceptable daily intake; DDA, detailed dietary analysis; EFSA, European Food Safety Authority; FEMA, Flavour and Extract Manufacturers Association of the United States; FGE.01, Flavouring Group Evaluation 1; FGE.02, Flavouring Group Evaluation 2; FOSIE, Food Safety in Europe; FSM, flavourings stochastic model; ILSI, International Life Sciences Institute; JECFA, Joint FAO/WHO Expert Committee on Food Additives; MRCA, Market Research Corporation of America; MSDI, maximised survey-derived daily intake; SCF, Scientific Committee on Food; TAMDI, theoretical added maximum daily intake; TTC, thresholds of toxicological concern; USDA, US Department of Agriculture; UULs, upper use levels * Corresponding author. Tel.: +39 06 51494402; fax: +39 06 51494550. E-mail address: [email protected] (C. Leclercq). 0278-6915/$ - see front matter
2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.fct.2004.08.017

106

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

1. Introduction The level of chemical substances in food is an important aspect of food quality and safety. These components may constitute a health hazard when overall intakes are too high. Exposure assessment is therefore a crucial component of risk assessment. Dietary exposure to chemical substances should be assessed by combining data on concentration in all food products with data on their consumption. This ideal situation is difficult to realise and it is considered neither cost-effective nor necessary to collect detailed food consumption and chemical concentration data for every hazardous substance (Lawrie and Rees, 1996). A stepwise procedure is therefore commonly used to focus resources on the most important issues. The stepwise approach to dietary exposure assessment of food chemicals is such that as the accuracy of dietary exposure assessments increases, the cost of collecting adequate data and resources needed to undertake the assessments also increases (WHO, 1997). The methodologies adopted to assess exposure from diet must take into special consideration non-average individuals, and in particular those who consume relatively large quantities of foods containing higher concentrations of substances that may potentially lead to a health risk (European Commission, 1998). Special consideration is generally given to children due to their higher intake of food per kilogram body weight (Lowik, 1996). Thus, in a stepwise procedure for the safety evaluation of chemical substances, intake estimates at the first steps need to be truly conservative. In principle, a proper risk assessment requires that the time frame for food consumption is the same as for the safety assessment (Lowik, 1996). Therefore, when safety reference values are referred to potential chronic effects, the assessment of exposure should ideally also consider life time consumption patterns. However such data are not usually available and exposure is often assessed on the basis of short duration food surveys which provide conservative estimates of long term exposure in high consumers. At the international level, dietary exposure assessments are essential components in assuring that health requirements for food are consistent globally and protective for public health. Although international dietary exposure methods need to screen a large number of food chemicals, they must also provide the most realistic estimates of dietary exposure by making the best use of available data. All other things being equal, the international calculations should provide exposure estimates that are equal to or greater than the best estimate carried out at the national level. If the estimated international chronic dietary exposure to a chemical does not exceed its ADI or PTWI/PTDI, then the level of exposure will usually be acceptable at the national level because even

the best international dietary exposure assessments will overestimate exposure (WHO, 1997). In the European Union, a specific regulation of the European Parliament and of the Council (European Commission, 1996) lays down a procedure for the establishment of a list of authorised flavouring substances (hereafter called ‘‘flavourings’’). In application of this Regulation, a register of about 2700 flavourings used by the food industry in or on foodstuffs in the Member States was adopted (European Commission, 1999; European Commission, 2002). The procedure for the safety evaluation of flavourings adopted by the European Commission (European Commission, 2000) in order to establish a positive list of these substances is a stepwise approach which was developed by the Joint FAO/WHO Expert Committee on Food Additives (JECFA, 1997). The JECFA procedure integrates information on intake from current uses, structure–activity relationships, metabolism and toxicity. In addition, information on purity and chemical specification is assessed. In 1999, the Scientific Committee on Food (SCF) was asked to advise the European Commission on the above mentioned programme for the evaluation of flavourings and amended this procedure (Scientific Committee on Food, 1999). The only change was that flavourings would not be included in the list ‘‘with the only argument that their estimated intake is lower than the threshold of concern of 1.5 lg/person/day’’. One of the key elements in the procedure is the subdivision of flavourings into three structural classes for which human exposure thresholds have been specified that are ‘‘not considered to present a safety concern’’. The concept that there are levels of exposure that do not cause adverse effects is inherent in setting acceptable daily intakes (ADIs) for chemicals with known toxicological profiles. The threshold of toxicological concern (TTC) principle extends this concept by proposing that a threshold value, below which there is a very low probability of an appreciable risk to human health, can be identified for many chemicals, in the absence of a full toxicity database, based on their chemical structures and on the known toxicity of chemicals which share similar structural characteristics. In 2004 an expert group of the European branch of the International Life Sciences Institute—ILSI Europe—has examined the TTC principle for its applicability in food safety evaluations (Kroes et al., 2004). The group concluded that ‘‘the TTC principle can be applied for low concentrations in food of chemicals that lack toxicity data, provided that there is a sound intake estimate’’. In the procedure developed by the JECFA for the safety evaluation of flavouring substances, the maximised survey-derived daily intake (MSDI) method is used to estimate the intake of a particular flavouring and to compare it with the above mentioned thresholds

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

(JECFA, 1997; Munro et al., 1999). The MSDI method, also known as the ‘‘per capita · 10’’ approach, is based on poundage data of the flavourings industry (kg/year) adjusted to a per capita basis. Estimates obtained are often referred as ‘‘estimated daily per capita intakes’’. Manufacturers are requested to exclude the use of flavourings in pharmaceutical, tobacco, and cosmetic products. Corrections are made for incomplete survey data by industry on the basis of an estimated ‘‘survey total response rate’’. The intake is estimated for consumers only, generally assumed to be 10% of the population for all flavourings. The following formula is used: Intakeðlg=person ¼

dayÞ

annual productionðkgÞ  109ðlg=kgÞ population of consumers  survey response rate  365ðdaysÞ

The aim of this paper is to provide an updated review of scientific opinions and papers dealing with the MSDI method used to assess flavouring exposure and to check whether it provides a conservative estimate of intake.

2. Review of scientific papers and opinions dealing with the MSDI method The MSDI method was asserted to be ‘‘a conservative and practical approach to assessing exposure to flavouring substances and other food ingredients’’, based on the results of a study carried out in the United States (Hall and Ford, 1999). In this study, for 10 flavourings, the estimated intakes obtained with the MSDI method were compared with the results of a detailed dietary analysis based on 14-day food frequency data combined with mean portion sizes and presence/concentration data provided by experts from the flavour industry. The findings of this publication were more recently supported by a study in which estimates of exposure to flavourings based on the MSDI were compared with exposure estimates based on a flavouring stochastic model (FSM) (Lambe et al., 2002). In 1999, the SCF (Scientific Committee on Food, 1999) recognised that the JECFA procedure ‘‘is a pragmatic approach’’, the ‘‘most updated and systematic’’ for the safety evaluation of chemically defined flavouring substances. The Committee therefore declared to be ‘‘prepared to use this approach for chemically defined flavouring substances within the evaluation programme of the Commission’’. The Committee recommended however that ‘‘the intake estimation needs further consideration and should not be based only on production data but should also take into account other intake assessments’’. SCF highlighted the fact that ‘‘the MSDI model may underestimate the intake by certain groups of consumers— although compensated somewhat by the assumptions

107

concerning under-reporting and that only 10% of the population consume flavoured foods and beverages. The MSDI model does not take into account the consumption pattern of population sub-groups. It neither considers geographic variations in the use of the flavourings nor the fact that a specific flavouring substance can be used only in one or very few food categories, consumed only locally or by a limited number of persons. Furthermore, it is known that at least for some flavourings there may be a great variation in the periodically reported production volumes. This variation can be orders of magnitude for a specific flavouring. This would be particularly important for flavourings with intakes only slightly below the threshold criteria within the evaluation procedure’’. The SCF concluded that a ‘‘more appropriate intake estimate could be performed using a stepwise procedure’’ and suggested that the MSDI should be combined with a more conservative method, the theoretical added maximum daily intake (TAMDI). The stepwise procedure ‘‘could start with both the MSDI and the TAMDI and, when needed, the TAMDI results could be refined by exclusion of food categories in which the flavour is not present and by inclusion of maximum usage levels for specific food categories. Prerequisite for such a stepwise procedure would be that the flavour industry and the Member States provide detailed data on global and specific maximum levels of use and on the type and consumed amounts of flavoured food categories’’. In the report of the 55th meeting of JECFA (JECFA, 2000) the Committee ‘‘noted that use of the Ôper capita · 10Õ method may, in some cases, result in an underestimate of the intake of persons with high levels of consumption of specific foods’’. The Committee expected results from ongoing research on methods to predict the intake of flavourings in order ‘‘to investigate the use of MSDI and alternative methods of intake assessment at future meetings of the Committee’’. An extensive review of the techniques used for the assessment of intake of chemical substances from the diet was recently performed within the frame of the EC Concerted Action on Food Safety in Europe (FOSIE) (Kroes et al., 2002). The authors state that the MSDI method ‘‘is considered useful to observe trends over time but is also recognised to not represent adequately the exposure to chemicals that are not broadly distributed over various food categories, for instance a substance contained in a very specific exotic food consumed by a very low percentage of the population’’. The method was stated to ‘‘provide a reasonable estimate for amounts available only if the market for the relevant raw material is reasonably well assessed’’. In 1999, the Nordic Group on Food Toxicology and Risk Assessment (NNT) and the Nordic Group on Diet and Nutrition (NKE), under the Nordic Council of

108

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

Ministers, critically reviewed the MSDI method within the project ‘‘Methods for calculating the intake of flavourings’’ (Bergsten et al., 2002). The Nordic project group agreed that ‘‘MSDI method, in many cases, provides intake estimates of flavourings that are conservative in relation to actual intake in the population. However, in some cases the MSDI intake estimate may underestimate the actual intake, especially the intakes among high consumers’’. The Group concluded that the MSDI method ‘‘is not considered to be fully suitable to use in a safety evaluation procedure for flavourings’’.

3. Comparison between reported use levels and intake estimated by the MSDI method in the safety evaluation of SCF In the last SCF opinions related to flavourings, MSDI estimates were calculated on the basis of the corresponding annual production volume, as reported in a survey conducted in Europe in 1995 by the International Organization of the Flavour Industry (Scientific Committee on Food, 2003a,b,c). According to the method principles, the reported annual production volume was divided by 0.6 (assuming that 40% of the production volume is not reported), by the number of flavouring consumers (estimated to be 10% of the European population i.e. 37.5 million) and by 365 (days in a year). In these opinions upper use levels (UULS) reported by industries were indicated (i.e. for glycyrrhizinate 200 mg/kg in nonalcoholic beverages and 1550 mg/kg in candies) and could be used to assess the order of magnitude of the potential level of exposure. In the opinion on glycyrrhizinic acid and its ammonium salt (Scientific Committee on Food, 2003c), the MSDI for ammonium glycyrrhizinate and glycyrrhizinic acid were respectively 130 and 240 lg/person/day. The Committee noted that ‘‘the MSDI calculations may result in unrealistically low intake figures for individuals who select to consume certain foods, flavoured at the UULs’’. An example is provided in the opinion: ‘‘intake of 50 g of candies, flavoured with ammonium glycyrrhizinate at the UUL would lead to an intake of ammonium glycyrrhizinate of 77.5 mg, while drinking of one litre of a glycyrrhizin-flavoured drink would provide an intake of 200 mg glycyrrhizinate. Both figures exceed the calculated MSDI by 3 orders of magnitude’’. The Committee concluded that in order to complete the evaluation for both glycyrrhizinic acid and ammonium glycyrrhizinate, further information is needed to be provided and in particular ‘‘more refined usage data’’. The example of glycyrrhizinic acid was reported here because, only in this case, detailed concentration data reported by the industry are indicated in the SCF opinion.

A similar comparison between MSDI intake estimates and reported use levels can be carried out by using the data published in the opinions on the first two Flavouring Group Evaluations performed by the SCF: ‘‘Branched-chain aliphatic saturated aldehydes, carboxylic acids and related esters of primary alcohols and branched-chain carboxylic acids from chemical groups 1 and 2’’ (FGE.01) (Scientific Committee on Food, 2003a) and ‘‘Branched- and straight-chain aliphatic saturated primary alcohols, aldehydes and related esters of primary alcohols and straight-chain carboxylic acids from chemical groups 1 and 2’’ (FGE.02) (Scientific Committee on Food, 2003b). In FGE.01 and FGE.02, respectively 17 and 41 substances were evaluated. According to the flavour industry, all of the candidate substances are used in flavoured food products in most of the food categories identified in the Annex III of the Commission Regulation no. 1565/2000 (European Commission, 2000). For example, most flavourings listed in FGE.01 are used in beverages, alcoholic and nonalcoholic, dairy products, confectionery and bakery wares. The same is true for most flavourings listed in FGE.02, with the exception that these are not used in beverages. The ‘‘normal use levels’’ for the 17 candidate substances in FGE.01 were in the range of 1–20 mg/kg food, with ‘‘maximum use levels’’ of up to 200 mg/kg in ‘‘confectionery’’ and ‘‘ready-to-eat savouries’’. The ‘‘normal use levels’’ for the candidate substances in FGE.02 were reported to be in the range of 1–50 mg/kg food, with ‘‘maximum use levels’’ of up to 350 mg/kg food in ‘‘ready-to-eat savouries’’. In FGE.01, pentyl isovalerate (FL no. 09.499) was the flavouring with the highest ‘‘estimated daily per capita intake’’, equal to 11 lg/day. A quick ‘‘back calculation’’ shows that 11 lg/day would be the intake of FL no. 09.499 in a person consuming approximately • three yoghurts every month (125 g each) containing 1 mg/kg of the substance or • one glass of nonalcoholic beverage every year (200 g) containing 20 mg/kg of the substance or • 10 candies every year (2 g each) containing 200 mg/kg of the substance. Similarly, in FGE.02, butyl decanoate (FL no. 09.327) was the flavouring with the highest ‘‘estimated daily per capita intake’’, equal to 7.9 lg/day. A quick ‘‘back calculation’’ shows that 7.9 lg/day would be the intake of FL no. 09.327 in a person consuming approximately • two yoghurts every month (125 g each) containing 1 mg/kg of the substance or

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

• one piece of bakery wares every year (60 g) containing 50 mg/kg of the substance or • 8 g of ready-to-eat savouries every year containing 350 mg/kg of the substance. These levels of consumption appear particularly low when compared with those recorded in food consumption surveys. According to the Danish and Norwegian national dietary surveys, 10% of Danish 1–3 year old children consume at least 100 g of fruit yoghurt per day, whereas the average consumption of confectionery is 11 and 12 g/day in Denmark and Norway, respectively (Bergsten et al., 2002). According to the Italian Nationwide food survey INN-CA, consumers of ‘‘yoghurt’’ (29% of the population), ‘‘soft drinks’’ (31%) and ‘‘cakes, biscuits, pastries, etc.’’ (83%) eat, on average, 58.5, 113.3 and 51.9 g/day of these products respectively (Turrini et al., 2001). For the other flavourings evaluated in FGE.01 and FGE.02, ‘‘estimated daily per capita intakes’’ were even lower than that of the reported examples (up to 1500 and 2200 times lower for FGE.01 and FG.02, respectively) but the use levels were the same. The expected frequency of intake of foods flavoured with these substances at use levels reported by the industry would therefore be even lower than that calculated above. These examples show that for these two groups of substances (FGE.01 and FGE.02) the ‘‘estimated daily per capita intake’’ calculated by means of the MSDI method is not consistent with the potential exposure to flavourings that would be observed in regular consumers of products flavoured at the use levels reported by the industry. For the two substances taken as an example exposure derived from the daily consumption of some of these flavoured products would result in an intake more than 100 times higher. For other substances of these two flavouring groups, it could result to be more than 100,000 higher. There are two possible explanations for this discrepancy: the use levels reported by the industry are up to 5 orders of magnitude higher than those effectively used, or the method used to assess the intake underestimates by up to 5 orders of magnitude the intake of regular consumers of flavoured products, due to a gross overestimation of the percentage of consumers and/or a gross underestimation of production volumes.

4. Review of the studies used to support the MSDI method In view of the above mentioned considerations regarding the MSDI method a critical review of the two studies mentioned above used to support the statement that it is a conservative method for assessing exposure to flavourings was performed.

109

In the study carried out by Hall and Ford (1999), a method based on detailed dietary analysis (DDA) was used to provide the most accurate possible estimates of intake. Ten flavourings were chosen to represent substances ranging from low volume/highly specialized uses to broad use/higher volume substances. For each flavouring, the 95th percentile of intake was calculated with the DDA method and then compared with the intake calculated with the MSDI method. The ratio of the MSDI to the 95th percentile of intake calculated with the DDA method ranged from 0.73 to 63. These results were used by authors to support their conclusion that the MSDI method ‘‘errs on the side of conservatism by tending to overstate exposure’’ and ‘‘serves very well as a conservative estimate of intake’’. The authors therefore suggested that the MSDI method ‘‘should be an acceptable and practical alternative to the costly and time-consuming DDA method’’ because poundage data are much easier to obtain than dietary intake data. It is worth noting that the 95th percentile of intake calculated with the DDA method exceeded the MSDI estimate in 2 out of 10 flavourings (20%). This margin of error cannot be neglected. However, the main shortcoming of this study is that the DDA method, as was applied, is inadequate to reliably estimate high percentiles of intake. Evidence for this statement follows. Average daily intakes of each flavouring were obtained at an individual level on the basis of (1) the amount of each specific food eaten; (2) the eating frequency for each specific food; and (3) the amount of the flavouring in each food. A thorough reading of the paper suggests that the formula used could be expressed as P14 Pn I ik ¼

d¼1

j¼1 F idj

 S ij  P jk  Ljk ; 14

where Iik = 14-day average intake of flavouring k for individual i, Fidj = total number of eating occasions of food category j on day d for individual i, Sdj = mean portion size of food category j for individual i, Pjk = probability that food category j actually contains flavouring k, Ljk = weighted mean of the ‘‘usual use level’’ of flavouring k in food category j, n = total number of food categories. Data on the daily frequency of intake of highly specific food subcategories were derived from the Market Research Corporation of America (MRCA). The MRCA data were drawn from a stratified panel of more than 12,000 individuals, representative of urban, suburban and rural populations, and covering age groups from 0 to 5 months to over 65 years. The data were recorded as daily diaries covering both foods eaten at home and away from home, over a 14-day period. The collection of data from the MRCA panel was spread

110

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

evenly throughout the year in order to include seasonal variations. The amounts consumed per eating occasion were obtained from the US Department of Agriculture (USDA) food consumption survey, and averages were computed, by age groups of the eaters, for each food subcategory. The use of average portion sizes is the first possible cause of underestimation of the higher percentiles of intake. In fact, with this technique, intakes of individuals who consume higher than average portions of foods and beverages containing flavourings are not represented. A further cause of underestimation is the use of the probability that each specific food would contain each of the flavourings under study. Probabilities of containing each of the flavourings significantly varied according to the food category. Thus, ‘‘the probabilities for the foods in which menthol was used ranged from 0.01 for a particular frozen dessert on a stick to 1.0 for menthol cough drops’’. Consumers of a food category susceptible to contain a specific flavouring were always assumed to be consumers of the flavouring, even if the probability that the food category actually contains the flavouring was low. The average concentration level of flavourings was calculated among consumers of the specific food category and thus lowered by dilution. This can be illustrated with the following simple hypothetical case. If flavouring Alfa may be present only in food category Beta and • 30% of the 12,000 individuals consumed products from food category Beta; • all consumers ate products from food category Beta twice a day; • the mean portion size of products was 125 g in food category Beta; • the probability that Alfa is present in a product of food category Beta was set at 0.2; • the ‘‘usual use level’’ of Alfa in food category Beta was 100 mg/kg. In this case, a total of 3600 individuals (30%) appeared to be consumers of flavouring Alfa at a level of intake equal to 5 mg (2 · 0.125 kg · 0.2 · 100 mg/kg) for all of them. Also the 95th percentiles of intake would be therefore equal to 5 mg. In reality, if the probability that Alfa was contained in food category Beta was equal to 0.2, 80% of the items of food category Beta would not contain this substance. Assuming that each individual always consumed the same brand (full brand loyalty), only 20% of consumers of the food category Beta would actually be consumers of flavouring Alfa. Therefore only 720 individuals (6%) would be consumers of Alfa and their level of intake would be equal to 25 mg (2 · 0.125 kg · 100 mg/kg). The actual 95th percentiles of intake would be therefore equal to 25 mg, 5 times higher than the estimate obtained with the DDA method ap-

plied in the study. Also the percentage of flavourings consumers would be overestimated (30% versus 6%). Another assumption made was that any flavouring assigned a non-zero probability of use in a specific food would be used in that food at the Ôaverage usual use levelÕ for that food category determined by the Flavour and Extract Manufacturers Association of the United States (FEMA), in collaboration with the National Research Council. The authors noted that ‘‘reported level does not account for the many cases where the substance may be present at much lower levels’’ and ‘‘does not include the few cases where the substance provides a very dominant note’’, ‘‘in which cases the actual use may be closer to the average maximum levels’’. The authors asserted that ‘‘these two cases tend to balance each other’’. The use of ‘‘average use levels’’ in fact provides a good estimate of average intakes but clearly tends to underestimate high percentiles. It can therefore be concluded that assumptions made in this study in relation to the average portion size, to the probability that the food category actually contains the flavouring and to the ‘‘average usual use level’’ tend to reduce the variability of intake. The estimated intakes provided in this paper represent the flavouring intake of consumers who would ingest a ‘‘blend’’ of all flavoured foods and beverages available in the market. The fact that consumers tend to be loyal to products was not taken into consideration and this leads to an underestimation of high levels of exposure. This phenomenon was discussed by Leclercq et al. (2003) and Arcella et al. (2003) with respect to intense sweeteners. The DDA method, as it was applied in the paper, underestimates the 95th percentile of flavouring intake. This is probably due to the fact that, at the time the paper was published by Hall and Ford (1999), little research had been conducted in relation to the factors that may affect exposure at high percentiles, such as brand loyalty and portions sizes. The results of this study were used to support the statement that 10% is a conservative estimate of consumers of flavourings, as was also previously stated by Cadby (1996) and Munro et al. (1998). In reality, 10% is only an estimate of consumers of products susceptible to contain flavourings. It most likely largely overestimates the percentage of consumers for many flavourings. The overestimation of the percentage of consumers with the DDA method would be particularly high for those flavourings used in a few brands of products that belong to widely consumed food categories. The second study (Lambe et al., 2002) applied stochastic modelling, using data from food surveys and from the flavour industry, to investigate the probability of occurrence of intakes of intentionally added flavouring in excess of the MSDI estimates. Twelve flavourings were chosen to reflect broadly the large number of flavourings currently used in Europe. Over 40,000 flavour formulae, used at known levels of incorporation in 31

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

different categories of food and beverages, were examined to provide maximum concentrations, distributions of concentrations and an indication of the probability of encountering each substance in a flavoured food or beverage in any category. The values from the flavourings stochastic model (FSM) distribution showed a very low probability of exceeding MSDI estimates. For 10 of the flavourings, the MSDI covered at least 99% of the FSM distribution. For one flavouring (2,6-dimethyl pyrazine), the 97.5th percentile of the FSM was almost 10 times higher than the MSDI estimate. The authors concluded that their study confirmed the findings of Hall and Ford (1999) that ‘‘the MSDI appeared, within the limits of the studies, to be a practical and conservative method for assessing exposure to intentionally added flavouring substances’’. The results obtained from probabilistic models of dietary exposure clearly depend on many model inputs, such as the consumption and food chemical databases, the form of the setting of model parameters and the choice of modelling options. Algorithms and assumptions used in the models adopted by Lambe et al. (2002) are not easy to evaluate because few indications are given in relation to the parameters used and results of the simulation are provided only in terms of flavouring intakes. It is thus not possible to infer if the low intakes derive from low consumption of products susceptible to be flavoured or from low concentrations of flavourings in such products. However, this was the first attempt to design probabilistic models for assessing exposure to food flavourings and it clearly appears from the paper that the authors tried to take into consideration all the useful components. Moreover some of the potential limitations of the model, such as the lack of information about the market share of individual brand foods and of the market share associated with the concentrations of individual flavourings, were extensively discussed. On the other hand, data used in the model were not fully adequate to evaluate the reliability of the MSDI method for its use at the European level. Intakes are based on the observed presence and concentration of flavourings in foods and beverages available in one single market in the EU whereas the 12 flavourings were chosen to reflect broadly the large number of flavourings currently used in Europe. Moreover the consumption pattern of flavoured foods is that observed only in males aged 16–24 years, according to the Dietary and Nutritional Survey of British Adults (1988). This group was chosen because of its relatively high intake of processed foods and therefore potentially high intake of flavourings with respect to older adults. However, for each flavouring, high consumers per kg body weight will differ and could easily be found in another age class or in another Member State. It is likely that high intake of flavourings used in sweets will be found among children,

111

high intake of flavourings used in substitute meals in people on a diet, high intake of flavourings used in a traditional food in its country of origin, etc. As reported in the same paper, Hall and Ford (1999) demonstrated that in the USA 6–17 year olds were more likely to consume flavourings than those in the age range of 0–23 months, 2–5 years or >18 years. Consumption data used by Lambe et al. (2002) cannot be considered to be representative of the whole EU population. According to WHO (1997) the primary requirement for food consumption data used at the international level, for chronic dietary exposure assessments, is that it should take into account the differences in food consumption patterns, both within and among countries. The percentage of flavoured brands within food groups was estimated using the Irish National Food Ingredient Database. Thus, the chance of encountering any flavouring in a specific food category was calculated in the Irish market, similar to the UK one. On the contrary, the probability of the presence of a single flavouring in a food category was calculated using data at the European level. More than 40,000 formulae among those currently sold for use in any of the above categories of food and drink in Europe were considered. This most probably led to an underestimation of this probability at the country level. This error could be particularly important for those flavourings used prevalently in one country. Using this method, the specific flavouring will result to be present with the same probability all over Europe and this probability will be lower than the real one in a specific country due to a dilution effect. Finally, it is valuable to highlight the case of 2,6-dimethyl pyrazine, the only flavouring for which the MSDI estimate (0.026 lg/kg b.w./day) resulted to be lower than the 97.5th percentile of intake calculated using the FSM (0.24 lg/kg b.w./day). One flavouring out of 12 (8%) is a margin of error that cannot be neglected. Moreover the MSDI method underestimated the intake of 2,6-dimethyl pyrazine at the 97.5th percentile by almost 1 order of magnitude. The 97.5th percentile of intake and the probability of exceeding the MSDI estimates, reported in the paper as equal to 0.0753, can be used to approximate by defect the total amount of this flavouring consumed in the UK through a ‘‘back MSDI’’ calculation. If one assumes that all the individuals consume the lower level in their class of exposure and that their body weight was standard (60 kg), 54,637,000 people (92.47% of UK population) would have an intake equal to 0, whereas 2,972,000 people (5.03%) would have an intake equal to 569.4 lg/year (0.026 lg/kg b.w./day · 365 days · 60 kg b.w.) and 1,477,000 people (2.5%) would have an intake equal to 5,256 lg/year (0.24 lg/kg b.w./day · 365 days · 60 kg b.w.). The total amount of 2,6-dimethyl pyrazine consumed in the UK would thus be 9.5 kg/year. In spite of the fact that the calculated amount of 2,6-dimethyl

112

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

pyrazine was underestimated this would represent 86% of the overall European poundage data reported by the industry (equal to 11 kg/year). If the reliability of the FSM data is taken for granted, this result can be explained only in two ways, either this flavouring is used almost only in the UK or the poundage data reported by the industry grossly underestimates production levels at the European level. If it were uniformly used all over Europe, production volumes would be at least 5 times higher. Numerical simulation techniques provide powerful tools that take advantage of all the available knowledge in order to provide realistic estimates of exposure. The results, however, are only as good as the input data, algorithms and assumptions. The food consumption and chemical presence/concentration data used by Lambe et al. (2002) in the FSM are not representative of the EU situation and the population chosen cannot either be considered to be representative of a population with particularly high levels of flavouring exposure within Europe. Therefore, results from this study are not suitable to validate intake estimates obtained with the MSDI method, which are referred to the EU population. Another general consideration is that one may question whether the conclusions of studies conducted on 10 (Hall and Ford, 1999) and 12 (Lambe et al., 2002) flavourings can be extrapolated to the about 2700 flavourings present in the European market.

5. Discussion and conclusions The MSDI method was developed since a pragmatic approach to exposure assessment was needed within the stepwise procedure for safety evaluation of flavourings. Production volumes were the most readily available data and per capita estimates could therefore be easily calculated. The MSDI method could provide a reliable estimate of the current intake of flavouring in a ‘‘mean European consumer’’ if three assumptions were verified (1) if the percentage of poundage under-reporting was homogeneous among flavourings, (2) if industrial poundage data of 1995 were representative of current production levels, (3) if 10% was a reliable estimate of the percentage of European consumers for all flavourings. In relation to the first assumption, according to Munro et al. (1999) ‘‘numerous checks using data from independent sources such as imports show that, in general, the reported poundage in surveys accounts for only 60% of the total used’’ but no published data are available to check for the homogeneity of under-reporting among flavourings. It might therefore be that under-reporting is higher than 40% for some flavourings. In relation to the second assumption, annual volumes of produc-

tion of flavourings vary considerably up to orders of magnitude. Bergsten et al. (2002) noted that production volumes used by Hall and Ford (1999) were related to the year 1970 and searched more recent data. Production volumes related to 1987 and 1999 were found for four out of the 10 selected flavourings and ranged from 17% to 735% of the 1970 data. This variability is also clearly evidenced by comparing the US annual volumes of 1982 and 1987 which were reported by Adams et al. (1996). Thus, d-neo-menthol varied by 3 orders of magnitude (from 0.1 to 140 kg/year) and p-menth-8-en-2-one varied by 5 orders of magnitude (from less than 0.01 to 921 kg/year). This variability may reflect the actual variability in production: no production of a specific flavouring may be required in one year when stocks are sufficient. High productions volumes in one year would therefore be related to consumption over more than a year. Therefore, the use of production volumes in years of high production may overestimate by a factor of 5 the actual pro capita intake (if the volume produced were spread over five years). On the other hand, it appears from cases reported above that the use of production volumes in years of low production may lead to an underestimate of a factor of up to 1000. In relation to the third assumption, it is made clear in the present paper that 10% may be a conservative estimate of consumers of products susceptible to contain flavourings but cannot be considered a conservative estimate of consumers of single flavouring. A specific flavouring can be used only in one or few brands of a food category, consumed only locally or by a limited number of persons in Europe. In this case, the number of consumers will be orders of magnitude lower than the assumed 37.5 millions. If updated production volumes were made available and proved to be stable, the MSDI method could provide a reasonably conservative estimate of the mean intake of flavourings which are used in a large range of products consumed all over Europe. This may be the case for vanillin which is largely used in fine bakery wares. In Table 1, the critical points of the assumptions adopted by the MSDI method are summarised. Therefore, none of the three assumptions can be verified. Errors of orders of magnitude can concomitantly occur in at least two variables (production volumes and percent of consumers) and no data are accessible to quantify the uncertainty related to a third variable (poundage under-reporting). Such a situation leads to an extremely high uncertainty in estimates derived from these variables. If use levels reported by the industry are close to reality, it is likely that, for a number of flavourings, the MSDI provides a gross underestimate of the intake of mean consumers. The reality checks mentioned above support this assumption: intake estimates from the MSDI can, for certain flavourings, be up to 5 orders of magnitude lower than those calculated con-

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

113

Table 1 Critical points of the assumptions adopted by the maximised survey-derived daily intake (MSDI) method Information needed

MSDI assumptions

Critical points of the MSDI assumptions

Poundage under-reporting

Reported poundage in surveys accounts for 60% of the total

No published data are available to evaluate the uncertainty of the percentage of under-reporting and to check for its homogeneity among flavourings

Production volumes

1995 Production volumes are representative of current production levels

Year to year variability in the production volumes may be very high

Number of consumers

Consumers represent 10% of the whole European population for all flavourings

Percentage of consumers could be far lower, in particular for flavourings used in small brands or in restricted geographical areas

Step 1 Decision tree structural class

Step 2 Can the substance be predicted to be metabolised to innocuous products ?

Step A3*

Step B3*

Yes

Do the conditions of use result in an intake greater than the threshold of concern for the structural classes ?

Yes

No

Data must be available on the substance or closely related substances to perform a safety evaluation.

Yes

No

Do the conditions of use result in an intake greater than the threshold of concern for the structural class ?

Step A4 Step B4* Is the substance or are its metabolites endogenous ?

Yes

Substance would not be expected to be of safety concern.

Yes

Step A5*

No Yes

Does a NOEL exist for the substance which provides an adequate margin of safety under conditions of intended use, or does a NOEL exist for structurally related substances which is high enough to accommodate any perceived difference in toxicity between the substance and the related substances ?

Additional data required

No

Does a NOEL exist for the substance which provides an adequate margin of safety under conditions of intended use, or does a NOEL exist for structurally related substances which is high enough to accommodate any perceived difference in toxicity between the substance and the related substances ?

No

No

* Bold and underlined characters indicate the steps in which intake estimates based on the MSDI method are involved.

Fig. 1. Procedure for safety evaluation of chemically defined flavouring substances (Scientific Committee on Food, 2003b).

sidering regular consumption of foods and beverages flavoured at concentration levels provided by the industry. In the procedure for Safety Evaluation of Flavourings, the estimation of the intake of a particular flavouring is involved in four different steps of the decision tree (Fig. 1). If the intake is lower than ‘‘threshold of concern for the structural class’’ (steps A3 and B3) or if there is an adequate margin of safety between intake ‘‘under conditions of intended use’’ and NOEL (steps A5 and B4), the conclusion will be ‘‘substance would not be expected to be of safety concern’’. Exposure assessment is only one step in the process that leads to their final risk characterisation. In this last step the uncertainties of the hazard characterization and of the exposure assessment are consolidated. However, intake estimates are so essential in the procedure that the margins of safety inherent to hazard characterisation

may not provide reassurance because the uncertainty on intake estimates based on the MSDI method is so large and difficult to quantify. A review of the group evaluations of flavourings recently performed by the SCF and the JECFA shows that in a very large number of cases no further toxicology data were required for the flavourings. Thus, in the SCF evaluation of FGE.01 and FGE.02, all candidate substances (respectively 17 and 41 substances) were classified in structural class I (low order of oral toxicity) and the European daily per capita intakes were below the threshold of concern value for this class, equal to 1800 lg/person/day. Since these substances were ‘‘expected to be efficiently metabolised’’, the conclusion was that these substances ‘‘would not give rise to safety concerns at the current estimated levels of intake from their use as flavourings’’.

114

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

As far as the JECFA is concerned, only in very few cases, the intake estimates based on the MSDI were above the threshold of concern. Thus in its 55th and 59th meetings, respectively, 124 and 196 flavourings were evaluated and for all these substances the MSDI estimates were low and the Committee concluded ‘‘no safety concerns at the current estimated levels of intake’’ (JECFA, 2000; JECFA, 2003). In the 57th meeting, 202 flavourings were evaluated and the Committee conclusion was ‘‘no safety concerns at the current estimated levels of intake’’ for all substances except one (JECFA, 2002). The first steps to establish a register of flavourings permitted in the EU were taken in 1996 and the evaluation procedure is intended for completion by 2005. According to flavouring producers, this exercise has already been extremely expensive both in terms of manpower and money (Hardinge, 2003). The development and validation of a completely new method to assess intake of flavourings would require time and resources and would further delay the adoption of the positive list. However, another simple method does exist. The theoretical added maximum daily intake (TAMDI) method has been used by the SCF to assess exposure to single flavourings (Cadby, 1996) and could be used in the procedure for Safety Evaluation of Flavourings. In contrast to the MSDI, which is based on poundage data for flavourings, the TAMDI method assumes that the hypothetical consumer will day in and day out consume a fixed amount of flavoured food and beverages and that these items will always contain the specific flavouring at its specified upper use level. Cadby (1996) noted that the TAMDI estimate is usually between one thousand and one million times greater than the MSDI estimate. Doubtless, the TAMDI scenario is a worst case and may lead to a gross overestimation of flavouring intake. According to a FAO/WHO consultation, international dietary exposure methods must ‘‘provide the most realistic estimates of dietary exposure by making the best use of the available data’’ and ‘‘care must be exercised in the application of conservative assumptions so that the calculations done at the international level are as representative as possible of actual exposure’’. The same document states that ‘‘the international calculations should provide exposure estimates that are equal to or greater than the best estimate carried out at national level’’ (WHO, 1997). Since refined exposure data are available for a very limited number of flavourings, a conservative method is needed at the first steps of the procedure of safety evaluation. Thus, in its opinion on the programme for the evaluation of flavourings, the SCF suggested that the stepwise procedure ‘‘could start with both the MSDI and TAMDI’’ (Scientific Committee on Food, 1999). Also the Nordic project group proposed the use of the TAMDI method, modified in order to obtain less conservative estimates (Bergsten et al., 2002).

The use of conservative methods, such as the TAMDI, are appropriate in order to prevent consumers from being exposed to unexpectedly high intakes but, on the other hand, can produce estimates of exposure which are improbable and sometimes obscure the ability of regulators, industry and consumers to determine which scenarios present a risk that is likely to occur and therefore need to be addressed. Flavourings which do not give rise to safety concerns at the levels of intake estimated by the TAMDI method would be also safe for those who consume relatively large quantities of foods containing the highest concentrations of these substances. Due to the conservativeness of these estimates it would not be reasonable to exclude from the list all the other flavourings. Refined intake estimates may be calculated on the basis of consumption data of products that may be flavoured by the substance under study. In this case, it will be necessary to take into account geographical variation in Europe. In the mid-term, the creation of a data bank on the mean and high consumption of foods and beverages in adults and children of all EU countries would be extremely useful to this aim. The data should have a standardized format in order to allow the performing of conservative exposure assessment, taking into account geographical variation of food consumption patterns. If conservative estimates of intake are high and refined estimates that actually reflect intake in consumers of flavoured products are not available, there may be a need for risk management decisions. In order to ensure the safety of food, intake could be limited to the level at which the safety of the substances was evaluated. This is of course particularly important for substances of structural class III (‘‘that have structural features that permit no strong initial presumption of safety or may even suggest significant toxicity’’). Their intake threshold is very low (90 lg/person/day) and therefore easier to be exceeded. Maximum Permitted Level may be set in order to limit the potential intake and prevent it from exceeding threshold levels. A simple calculation described by Kroes et al. (2004) allows one to establish the concentration in food and beverages that could lead to the threshold by assuming that the compound occurs uniformly in the whole diet. Thus the authors calculated that when the compound under consideration has a threshold of concern equal to 90 lg per person per day, this intake is reached by a concentration of 30 lg/kg in the diet (considering a daily consumption of 1.5 kg of food and 1.5 kg of beverages). MPL could be set at this concentration until refined intake data or toxicological data are made available. A further research need related to flavouring exposure is the development of a method to evaluate the combined intake of added flavourings with those naturally present in food. Moreover, relevant quantities of flavourings are added to some animal feed and, if

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

proved to be significant, the possible contribution of residues of such substances ingested via food of animal origin should be included in the evaluation of human exposure. On the other hand, there may be the need to better quantify the actual use levels and the effect of food processing on flavouring concentration levels: higher amounts might be added than those present in the final product, especially for those substances which are volatile. The establishment of a EU-wide positive list of flavourings is aimed at making food safer and is an important piece of European food safety legislation (European Food Safety Authority—EFSA, 2003). In order to reach this important commitment, the procedure for safety evaluation of flavourings should be based on conservative intake estimates.

Acknowledgments Authors are most grateful to M.R. Milana and E. Testai for their critical review of the manuscript.

References Adams, T.B., Hallagan, J.B., Putnam, J.M., Gierke, T.L., Doull, J., Munro, I.C., Newberne, P., Portoghese, P.S., Smith, R.L., Wagner, B.M., Weil, C.S., Woods, L.A., Ford, R.A., 1996. The FEMA GRAS assessment of alicyclic substances used as flavour ingredients. Food and Chemical Toxicology 34 (9), 763–828. Arcella, D., Soggiu, M.E., Leclercq, C., 2003. Probabilistic modelling of human exposure to intense sweeteners in Italian teenagers: validation and sensitivity analysis of a probabilistic model including indicators of market share and brand loyalty. Food Additives and Contaminants 20 (Suppl. 1), S73–S86. Bergsten, C., Andersen, N., Gry, J., 2002. Methods for Calculating the Intake of Flavourings. TemaNord 2002, 550. Volume Nordic Council of Ministers, Copenhagen. Cadby, P., 1996. Estimating intakes of flavouring substances. Food Additives and Contaminants 13 (4), 453–460. European Commission, 1996. Regulation (EC) No 2232/96 of the European Parliament and of the Council of 28 October 1996. Official Journal of the European Communities, L 299, pp. 1–4. European Commission, 1998. Development of Methodologies for the Monitoring of Food Additive Intake across the European Union, Task 4.2—Final Report Submitted by the Task Co-ordinator 16 January 1998. Volume European Commission, Brussels, Belgium, pp. 1–45, Annex A–J. European Commission, 1999. Commission Decision 1999/217/EC of 23 February 1999 adopting a register of flavouring substances used in or on foodstuffs. Official Journal of the European Communities, L 84, pp. 1–137. European Commission, 2000. Commission Regulation No 1565/2000 of 18 July 2000 laying down the measures necessary for the adoption of an evaluation programme in application of Regulation (EC) No 2232/96. Official Journal of the European Communities, L 84, pp. 8–16. European Commission, 2002. Commission Decision 2002/113/EC of 23 January 2002 amending Commission Decision 1999/217/EC as regards the register of flavouring substances used in or on

115

foodstuffs. Official Journal of the European Communities, L 49, pp. 1–160. European Food Safety Authority—EFSA, 2003. EFSA gets into Flavourings. Brussels, 18 March 2003. Hall, R.L., Ford, R.A., 1999. Comparison of two methods to assess the intake of flavouring substances. Food Additives and Contaminants 16 (11), 481–495. Hardinge, J., 2003. The Situation of the EU Legislation on the Flavourings: Status, Recent Developments and Next Future from an Industry Expert. In: International Symposium. Flavourings: Safety use, Assessment and Status of the Legislation in the EU. 17 September 2003, Villa Erba, Cernobbio, Italy. JECFA, 1997. Safety Evaluations of Groups of Related Flavouring Agents: 46th Report of the Joint FAO/WHO Expert Committee on Food Additives. Volume IPCS, WHO, Geneva. JECFA, 2000. Evaluations of Certain Food Additives and Contaminants: 55th Report of the Joint FAO/WHO Expert Committee on Food Additives. Geneva, Switzerland. JECFA, 2002. Evaluations of Certain Food Additives and Contaminants: 57th Report of the Joint FAO/WHO Expert Committee on Food Additives. Geneva, Switzerland. JECFA, 2003. Evaluations of Certain Food Additives and Contaminants: 55th report of the Joint FAO/WHO Expert Committee on Food Additives. Geneva, Switzerland. Kroes, R., Muller, D., Lambe, J., Lowik, M.R., van Klaveren, J., Kleiner, J., Massey, R., Mayer, S., Urieta, I., Verger, P., Visconti, A., 2002. Assessment of intake from the diet. Food and Chemical Toxicology 40 (2–3), 327–385. Kroes, R., Renwick, A.G., Cheeseman, M., Kleiner, J., Mangelsdorf, I., Piersma, A., Schilter, B., Schlatter, J., van Schothorst, F., Vos, J.G., Wurtzen, G., 2004. Structure-based thresholds of toxicological concern (TTC): guidance for application to substances present at low levels in the diet. Food and Chemical Toxicology 42 (1), 65–83. Lambe, J., Cadby, P., Gibney, M., 2002. Comparison of stochastic modelling of the intakes of intentionally added flavouring substances with theoretical added maximum daily intakes (TAMDI) and maximized survey-derived daily intakes (MSDI). Food Additives and Contaminants 19 (1), 2–14. Lawrie, C.A., Rees, N.M.A., 1996. The approach adopted in the UK for the estimation of the intake of food additives. Food Additives and Contaminants 13 (4), 411–416. Leclercq, C., Arcella, D., Le Donne, C., Piccinelli, R., Sette, S., Soggiu, M.E., 2003. Stochastic modelling of human exposure to food chemicals and nutrients within the ‘‘Montecarlo project: an exploration of the influence of brand loyalty and market share on intake estimates of intense sweeteners from sugar-free soft drinks. Toxicology Letters 140–141, 443–457. Lowik, M.R.H., 1996. Possible use of food consumption surveys to estimate exposure to additives. Food Additives and Contaminants 13 (4), 427–442. Munro, I.C., Kennepohl, E., Kroes, R., 1999. A procedure for the safety evaluation of flavouring substances. Joint FAO/WHO Expert Committee on Food Additives. Food and Chemical Toxicology 37 (2–3), 207–232. Munro, I.C., Shubik, P., Hall, R., 1998. Principles for the safety evaluation of flavouring substances. Food and Chemical Toxicology 36 (6), 529–540. Scientific Committee on Food, 1999. Opinion on a Programme for the Evaluation of Flavouring Substances, SCF/CS/FLAV/TASKF/11 Final. Volume European Commission, Health&Consumer Protection Directorate-General, Bruxelles. Scientific Committee on Food, 2003a. Opinion of the Scientific Committee on Food on Chemically defined Flavouring Substances Listed in the EU Register Flavouring Group Evaluation 1 (FGE.01): Branched-Chain Aliphatic Saturated Aldehydes, Carboxylic Acids and Related Esters of Primary Alcohols and

116

D. Arcella, C. Leclercq / Food and Chemical Toxicology 43 (2005) 105–116

Branched-Chain Carboxylic Acids from Chemical Groups 1 and 2 (Commission Regulation (EC) No 1565/2000 of 18 July 2000), SCF/ CS/FLAV/FLAVOUR/42 Final. Volume European Commission, Health&Consumer Protection Directorate-General, Bruxelles. Scientific Committee on Food, 2003b. Opinion of the Scientific Committee on Food on Chemically defined Flavouring Substances Listed in the EU Register Flavouring Group Evaluation 2 (FGE.02): Branched- and Straight-Chain Aliphatic Saturated Primary Alcohols, Aldehydes and Related Esters of Primary Alcohols and Straight-Chain Carboxylic Acids from Chemical Groups 1 and 2 (Commission Regulation (EC) No 1565/2000 of 18 July 2000), SCF/CS/FLAV/FLAVOUR/42 Final. Volume Euro-

pean Commission, Health&Consumer Protection DirectorateGeneral, Bruxelles. Scientific Committee on Food, 2003c. Opinion of the Scientific Committee on Food on Glycyrrhizinic Acid and its Ammonium Salt, SCF/CS/ADD/EDUL/225 Final. Volume European Commission, Health&Consumer Protection Directorate-General, Bruxelles. Turrini, A., Saba, A., Perrone, D., Cialfa, E., DÕAmicis, A., 2001. Food consumption patterns in Italy: the INN-CA Study 1994– 1996. European Journal of Clinical Nutrition 55 (7), 571–588. WHO, 1997. Food Consumption and Exposure Assessment of Chemicals, WHO/FSF/FOS/97 5. Volume FAO/WHO, Geneva.