Obtaining concentration data 651 Introduction

Underground Fat Loss Manual

Best Weight Loss Programs That Work

Get Instant Access

The actual concentrations of any substance migrating from the packaging into each and every foodstuff are uncertain. In order to evaluate the exposure to any substance, it is necessary to determine the amount of each and every foodstuff consumed, which may have been in contact with the substance, and the concentration of the substance in each and every foodstuff consumed. Surveillance surveys do not measure the concentration of a substance in every foodstuff, but are typically more targeted towards those foodstuffs in which the substance(s) being surveyed are considered to have their highest levels.

6.5.2 Different approaches for obtaining concentration data

In many cases concentration data in real foodstuffs are unavailable, particularly for new substances. Thus simulant data for migration is typically used. Foodstuffs are assigned a chemical mixture (a simulant), which is believed to represent the foodstuff, in both Europe and the USA. In Table 6.1, the simulants considered to represent different foodstuffs according to EU Directives 2002/72/EC and its amendments, 82/711/EEC and 85/572/EEC are given. Note that solid foods may or may not need to be tested depending upon whether they are fatty in nature. The simulant must be used under conditions which will simulate the most extreme case the packaging is likely to encounter, during the processing of its contents (e.g. sterilisation, pasteurisation). For further details on which foodstuff is represented by which

Table 6.1 Food simulants specified in the EU for testing plastics for migration (CEU 2002, 2003)


Composition of the

Foodstuffs that the simulant






Aqueous foods with a pH > 4.5


3% w/v acetic acid

Acidic foods pH < 4.5


10% v/v ethanol

Alcoholic foods and beverages


Ethanol at concentration

Concentration of ethanol

in foodstuff

(v/v) actually present


Olive oil or alternative fat simulants

Fatty foods


Test value divided by reduction factor X, in the range 2-5, because olive oil is considered too aggressive for many fatty foods


No testing required

For example, fruits with peel or dry foodstuffs

simulant and under what conditions (times and temperatures) consult these Directives.

It should be noted that at this point in time, the value for the migration into the fatty food simulant 'D' may be subject to a reduction factor (D/X) with 'X' varying from 1 to 5 for different fatty foods in recognition that olive oil is overly aggressive for real foodstuffs in many instances. The simulants for different foodstuffs may change in light of new knowledge, as is the case for milk which is being changed to 50% ethanol instead of water. In the USA the simulants used are similar, but as always there are slight differences between EU and USA. For further details of the simulants used by the US FDA consult http://www.cfsan.fda.gov/.

One of the main issues with concentration data is how the non-detectable (ND) values are treated. In many instances the substance(s) of interest is non-detectable in either food simulants or real foodstuffs. In a UK FSA survey (2000) for BADGE (bisphenol A diglycidyl ether) in canned foodstuffs, in more than 95% (105 of 111 targeted samples tested) of the foodstuffs tested the levels were non-detectable. Using targeted foodstuffs in any surveillance will always skew any results to a higher level, in that only foodstuffs considered most likely to contain the substance will typically be analysed.

If the migrant species of interest is not present in particular food packaging materials, the foodstuffs consumed in that packaging will have a true zero concentration data set or they can be excluded from any estimate of exposure. However, if it is possible that the migrant species could be present in the packaging of that foodstuff, even if it cannot be detected in the foodstuff, then it is necessary to make allowances for its presence. It is clear that if the value is ND it cannot be assumed that the value is zero and on the other hand, it cannot always be at the limit of detection (LOD). Therefore it is necessary to try to use more realistic values, of which there are many approaches. Normal or Gaussian distributions between zero and the LOD are often used. In reality this gives a mean value for the ND of half the LOD. Assuming a normal distribution, between 0 and the LOD, will effectively give similar results as a Gaussian one. When it is plausible that the migrant may be present and the migrant is toxic or when levels below the LOD may make a significant contribution to the overall exposure then it is necessary to use a value between zero and the LOD (Kroes et al. 2002). The emerging use of probabilistic (stochastic or Monte Carlo random number generation) modelling of exposure to migrants enables the NDs to be handled in a number of different ways, with values between zero and the LOD being statistically generated according to whatever distribution is required.

Improving the detection limit in many cases is one of the most efficient ways to demonstrate lower exposure. For example the exposure to BADGE in canned foodstuffs (food and beverages) was estimated (Oldring et al. 2006) using a stochastic model (probabilistic - Monte-Carlo approach) with two different LODs of 0.3 mg/dm2 and 0.5 mg/dm2 and the exposure was effectively halved, primarily because many of the foodstuffs consumed were acidic, aqueous or alcoholic where the concentrations of BADGE and its regulated derivatives were non-detectable.

Food surveillance surveys may contain a range of values for the concentration data for a given food or group of foods. Some may be ND whilst others are above the LOQ. For exposure assessments, it is possible to use various approaches to utilise this data, from assuming that the migrant is always present at the highest level recorded or the average level or the mean, etc. Another approach is to fit the actual data to a statistical distribution, for example, normal or lognormal. This enables a more representative value for migrant concentration to be used. An arguably improved treatment would be to use probabilistic modelling to randomly select concentration values in the statistical distribution range, with possibly weighting around the mean, in order to better represent the realistic concentration of the migrant in a food or range of foodstuffs.

Parmar et al. (1997) consider that in the absence of statistical tools, fewer than 20 samples is unlikely to give a sufficient range of results even if the contaminant is present in all samples, whereas more than 100 could be considered as being wasteful of resources. Fifty to 100 samples should normally be adequate, but not necessarily if the contaminant is unevenly distributed with the bulk of the samples being ND. The FSA BADGE survey would be one such example of where more than 100 samples were taken, but they were targeted towards those foodstuffs where there was a probability that they contained BADGE, even though the bulk of these were ND.

Food surveillance surveys give concentration values in either mg/kg (ppb) or mg/kg (ppm). However, concentration data derived using simulants normally give results in mg/dm2 or mg/dm2, therefore in order to relate these values to concentrations in foodstuffs it is necessary to know the actual surface to weight (volume) of the packaged foods. In practice this is seldom known and in the EU the factor typically used is 6 dm2/kg. Data from Bouma et al. (2003), Holmes et al. (2005) and ILSI (1996) indicate that in practice 6 dm2/ kg is too low by a factor of at least two. However, this under-assumption is compensated by the over-assumption of 1 kg of the foodstuff always being packaged in the same material of 6 dm2. To compound this dilemma, is the apparent growth in single person consumption.

In many EU member states, there is a growth in the number of people living alone. This impacts their consumption habits. Previously they may have lived with two or more persons and food items purchased would most likely have been such that they could be divided between them, thus the pack size would most likely have been much larger than that for the single person. For example, it could have been common to purchase a 200-500 g piece of cheese to share, with a surface area to volume ratio which could have been in the range ten to one (10 dm2/kg). In contrast the single person consumer may purchase two slices of cheese weighing 25 g with (say) 2 dm2 of packaging, which would give a surface to volume ratio of twenty to one. However, the amount consumed is the critical factor and, in all probability, the amount of cheese actually consumed by that individual will not have changed, but the exposure to any migrating species from the packaging could have increased if the amount of migration remained the same and the same packaging material were used.

It is generally recognised that values measured in simulants are normally worst case as the simulant normally extracts more of the substance than the foodstuff. Yet another approach is to use a strong solvent, such as acetonitrile, and extract all of the substance which could potentially migrate. If the estimate of exposure does not give cause for concern then it could be argued that there is no need to conduct simulant studies. An area which still needs resolution is the lag-time for multilayer packaging. Species which could migrate and are not in the food contact layer may over a period of time diffuse through the layers and eventually enter the foodstuff. For products with long shelf lives, this is a possibility. Guidelines are still being developed at the EU Commission level as to how lag times can be simulated.

Another approach to obtain migration data particularly for some plastic materials is the use of modelling. Today this approach is only suitable for certain materials but is accepted by the EU Commission. Diffusion within, and migration from, food contact materials are predictable processes that can be described by mathematical equations. Mass transfer from a plastic material, for instance, into food simulants obeys Fick's laws of diffusion in most cases. Physico-mathematical diffusion models have been established, verified and validated for migration from many plastics into food simulants and are accepted in the USA and in the EU.

Because of the complex, heterogeneous and variable nature of foods, compared to simple food simulating liquids, no general tool for modelling migration into foods is yet available. An EU project with the acronym

'FOODMIGROSURE' (www.foodmigrosure.com) was initiated with the objective to develop a migration model for estimation of mass transfer from food packaging plastics into foodstuffs by extension of the existing model for food simulants to more complex foodstuffs as contact matrices. These models probably represent the only practical way that the complete combination of relevant parameters, including variable food composition, in-pack processing and storage times and temperatures can be encompassed when compiling concentration datasets large enough to accurately describe the foodstuffs as eaten by European consumers. For further details on the use of modelling to predict migration from food contact materials consult Brandsch et al. (2002), Reynier et al. (2002) and O'Brien and Cooper (2002) (see section 6.9.1).

When concentration data for the chemical(s) of concern are considered it is necessary to relate them to their actual relevance. For example, do we use values at the highest level permitted, highest reported, mean or median or, most importantly, in the foodstuffs consumed? In addition if the consumer purchases one or more pre-packed foodstuffs and then prepares a meal from them how are any potential migrants from the processing considered?

6.5.3 Packaging of foodstuffs containing the migrant(s) of interest

Closely related to concentration data is the type of packaging from which the migrating species can originate. For the purposes of estimating exposure to migrants from food packaging, the focus has to be on the primary packaging. This would be, for example, the packet for a packet of crisps, whereas secondary packaging would be the bag containing the 12 packets of crisps. Whilst some dietary surveys may contain detailed descriptions of food items and who consumed what, very few have information on the packaging of the foodstuffs consumed; for example, in four NDNS (UK) surveys the packaging of some items (e.g. beer canned or beer bottled) is sometimes described, though this is not the norm. The primary focus of dietary surveys to date is, understandably from a nutritional point of view, on what was eaten rather than its packaging. In the case of raw foodstuffs with a thick outer covering, such as bananas or oranges, the packaging is most likely irrelevant, but with raw fish or meat or processed foods knowledge of the packaging is paramount in order to determine the potential exposure to a given migrant.

In order to obtain an estimate of exposure it is necessary to combine data derived from surveys of the food consumption with data derived from surveys of food packaging. Even then there could be issues as the food packaging may be identified as plastic without identifying the plastic. In some instances the packaging may be multilayer, and expert knowledge or analysis would be the only certain methods of determining the food contact layer. This is further compounded by the growth in mixed packages. For example, it is not uncommon in the UK to purchase fresh (raw) meat or fish in a mixed package consisting of an expanded polystyrene tray with a paper bottom insert and a plastic (but which plastic?) film over-wrap. Yet another complication is that for a definitive packaging description, such as 'a bottle of', the nature of the closure is frequently unknown and, furthermore, unless the bottle is described as glass, PET, PVC, etc. uncertainty still remains despite a valid description from a consumer's viewpoint.

One of the most straightforward approaches with the lack of packaging data is to use the total production of packaging materials for different foodstuffs, with corrections for imports and exports, and divide by the population. This is in essence the per-capita approach which is discussed later (section 6.7) and this was undertaken for canned foods and beverages (Dionisi and Oldring 2002). This has the disadvantage that it will under-estimate exposure due to the non-consumer.

The UK NDNS (National Diet Nutritional Surveys) probably have the most information about the packaging of the foodstuffs consumed by the whole of the population, but it is restricted to UK dietary habits. In one of the most recent food consumption surveys (Duffy et al. 2006b), the actual items of food packaging were collected and identified, thereby becoming the first food nutritional survey to determine the packaging of the food consumed, albeit for a limited number of children (ca. 600). A project sponsored by the FSA (project A03051) with Newcastle University is nearing completion and packaging of the foodstuffs consumed by children of different ages has been identified wherever possible. European industry is undertaking projects to improve the knowledge of the packaging of the foodstuffs consumed.

Bouma et al. (2003) undertook a survey in the Netherlands of the packaging of 606 foodstuffs, mainly retail, and analysed the food contact layer using FTIR, as well as determining the surface area to weight ratio. Polyolefins (polypropylene (27%) and polyethylene (34%)) accounted for the majority of the packaging, with polyvinylchloride, polystyrene, polyethylene terephthalate and paper and board being the next most frequent forms of packaging. However, even knowing that the packaging is derived from a particular polymer may still be inadequate for a more refined exposure assessment. Whilst it is adequate for assessing the exposure to the monomers of the polymer, it will not necessarily help with the additives. For example, different additives may be present in structurally different forms of the polymer, such as low or high density polyethylene, polystyrene or high impact polystyrene, polypropylene or orientated polypropylene. Nonetheless, this is a major improvement to the previous situation. The surface area ratios ranged from 6-95 dm2/kg; however, the higher values were for herbs, etc. For bakery products, meat, fish, fruit, vegetables, salads, microwaveable meals, nuts and sauces the typical range was from 10-30 dm2/kg.

The Food Standards Agency commissioned a PIRA International study of packaging materials used for dietary staples. This report refers to data from Mintel Food and drink reports (see below). It also gives a possible stepwise approach to allocate packaging to different staple foodstuffs, which should be considered as an appropriate protocol. Duffy et al. (2006a) summarised the situation regarding some of the available sources of information on packaging of foodstuffs.

In 1995, the EU commissioned Maurice Palmer Associates (MPA) to undertake a survey of packaging of foodstuffs across Europe. They primarily focused on the UK and Italy, but the data were not correlated with foodstuff consumption. The data collected from the UK and Italy were extrapolated to the remaining 15 EU Member States, but because many assumptions were necessary, the confidence in the results varied from about 60% for Ireland, Finland and Sweden to a high of 90% for the UK market (ILSI 1996). MPA concluded that the different databases for food packaging materials for 15 EU Member States had a considerable amount of detailed information and probably contained far more detail than would be needed to derive food consumption factors, although the information could if necessary be refined on a country by country market share or packaging type basis. It is believed that this data has not been put to many, if any, uses to date as far as packaging of foodstuffs consumed is concerned.

ILSI (1996) derived some pseudo food consumption factors for the EU based on the MPA data. The food contact area for all packaging was 20.1 dm2/person/day. As plastics accounted for 62% this equated to 12.4 dm2/ person/day. Whilst at first sight these may be seen as being significantly higher than the EU assumption of 6 dm2/person/day, if the USFDA assumption of 3 kg/person/day instead of 1 kg/person/day is used there is much better agreement (18 dm2 vs. 20.1 dm2/person/day). The use of polyethylene in Benelux and Ireland is about 9-10.4 dm2/person/day compared to 4-4.6 dm2/person/day for Spain, Portugal and Greece. This illustrates the difficulty of attempting pan-European treatments for exposure to migrants from food contact materials. PVC, PS/ABS were <1 dm2/person/day, with France having the highest usage. These results seem to tie in with the Bouma et al. study.

There are databases that contain amongst other data some information on packaging of foodstuffs, ranging from crude (e.g. box) to more specific (polypropylene), but they are not ideally suited to estimating exposure to migrants. They may provide supplementary information which could be of use when other sources of the migrant needs to be considered. Examples are the Dutch Grootverbuik Product Informatie database (www.gpi.nl) which covers foodstuffs supplied to the catering, hospital and restaurant industries. Whilst the Dutch EAN DAS (www.eandas.nl) contains information on packaging of fast moving consumer goods (FMCG), only 35% are food products. Other sources of food packaging information are commercial food and consumer databases that monitor trends in the consumer product market. The Mintel Global New Products Database (www.gnpd.com) monitors worldwide consumer packaged goods markets and covers the food, beverage and non-food sectors. The Innova Food and beverage database also collects information on ingredients, packaging and formulation of foods and beverages (www. innova-food. com).

Again, the level of packaging information in these databases ranges from crude to more specific. The German Association for Packaging Market Research (GVM) collate data on the packaging types used for foods at the retail rather than consumer level. The Fraunhoefer Institute undertook a survey of packaging of foodstuffs for a single supermarket chain based in Munich. Market power and Euromonitor, for example, collate packaging information across some or many of the EU Member States, although this is not their primary purpose. Overall these databases are a useful source of general packaging information. However, information is not routinely recorded on the actual food contact layer used for the food, the polymer type if the packaging is plastic or if the packaging is a multilayer. It is this level of detail that is needed to refine exposure assessments to chemical migrants from packaging materials. Also, these databases were not created to directly link with food consumption data and therefore give an overview only of the types of packaging used for foods available on the market but not the packaging used for foods that are actually purchased and consumed. In the absence of any other data they are an invaluable source of information in order that more realistic estimates of exposure can be made.

Was this article helpful?

0 0
Diet And Exercise Expertise

Diet And Exercise Expertise

Get All The Support And Guidance You Need To Be A Success At Dieting And Exercise. This Book Is One Of The Most Valuable Resources In The World When It Comes To Better Physical Personal Development Through Better Living.

Get My Free Ebook

Post a comment