Skip to main content
European Commission logo
français français
CORDIS - Résultats de la recherche de l’UE
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary
Contenu archivé le 2024-06-18

Quality and Safety of Feeds and Food for Europe

Final Report Summary - QSAFFE (Quality and Safety of Feeds and Food for Europe)

Executive Summary:
A safe, healthy and high-integrity food supply for the European consumer is of paramount importance. In relation to foods that come from animal origin, the rearing of healthy European livestock is highly dependent on the provision of high-quality and safe feeds for these animals as this has major impact on the safety of the entire animal-based food chain.
The major goals of QSAFFE were to deliver better, faster and more economically viable means of ensuring the quality and safety of animal feeds in Europe.
These goals have been delivered by a number of different means.
Firstly we have developed and advanced the implementation of highly cost-effective tests that can detect any animal feed contamination incident very early, before the feeds have been consumed by farm animals, thus preventing the serious consequences that might otherwise result.
Secondly the project developed robust risk management techniques and processes based on scientific measurements to identify fraudulent practices occurring in the supply chain, e.g. the import of feed materials that have false claims associated to them.
Thirdly the wealth of research conducted through the project to try and predict what future problems may arise in the feed supply chain caused by changes in global economics and climate change. This information will inform both industry and regulators of future testing requirements.
A fourth important research topic was the development of a better understanding of how contaminants present in feed materials can transfer into foodstuffs. This information will play a major role in advising regulators about what the real risk of contamination incidents are to the consumer.
While a huge amount of important research has been conducted in the project it is of equal importance that the knowledge gained is passed onto the appropriate stakeholders. Thus in QSAFFE we delivered multiple training activities to the end-users of the tests developed and provided industry and regulators with important information to make decisions based on scientific evidence.
During the project the QSAFFE co-ordinator (Prof Chris Elliott) was appointed by the UK Government to lead an independent review of the UK's food chain, which was further recognition of the relevance and importance the work of the project. The report 'Elliott Review into the Integrity and Assurance of Food Supply Networks - Final Report' is available to download on https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/350726/elliot-review-final-report-july2014.pdf
QSAFFE has been as an important research project that delivered vital methods and information to a wide range of stakeholders. The practices developed in the project will not only have a major impact in Europe but will do so on a global scale.

Project Context and Objectives:
QSAFFE: TOOLS FOR IMPROVING THE QUALITY, SAFETY AND AUTHENTICITY OF ANIMAL FEEDS
Quality and SAfety of Feeds and Food for Europe (QSAFFE) was a European collaborative research project which was funded by the European Commission under the 7th Framework Programme (contract no. 265702). 11 leading institutes from Europe and China, representing universities, research institutes, EC and SMEs were members of the project consortium.
A global safe supply chain of feed materials into Europe is required to meet the needs of the feed industries to supply the farm animal production industries. Without such supply chains food prices in Europe would soar and could ultimately lead to the collapse of the agriculture sector in the EU. The quality and safety of feed materials produced within the EU or outside has been recognised as of paramount importance to securing safe and wholesome food to the European consumer. To this end EC-regulations such as the General Food Law (EC) No. 178/2002, the Feed and Food Control Regulation (EC) No. 882/2004, Reg. (EC) No. 669/2009 as regards the increased level of official controls on imports of certain feed and food of non-animal origin and Reg. (EC) No. 767/2009 on the placing on the market and use of feed have been introduced to ensure a high level of consumer protection with regard to food and feed safety and full traceability of feed/food through all stages of production. The Rapid Alert System on Food and Feed (RASFF) has been established and many alerts deal with unsafe feed materials.
At the start of the project, in the preparatory phase, the QSAFFE consortium conducted an intensive consultation with EU regulators and the Europe-wide feed industry to help understand and identify the major risks that needed to be addressed. It became clear from the consultation process that potential solutions developed had to consider the cost implications to allow their implementation in a highly competitive industry. As a result of this process, QSAFFE was designed with the goal to deliver better, faster and more economically viable means of ensuring the quality and safety of animal feeds in Europe. Four scientific Work Packages, each with a different focus but each linked to the core theme of providing a framework for improving the quality and safety of animal feeds in Europe, were identified. A fifth WP was aimed at ensuring widespread transfer of knowledge and maximising the impact the project can have in terms of delivering better quality and safer feeds. In summary, major objectives were identified for the project, each delivered through a number of different means.
With the ultimate goal in mind, of 'delivering better, faster and more economically viable means of ensuring the quality and safety of animal feeds in Europe' the project identified key areas for investigation.
Work Package 1 was given the task of providing the feed industry with a comprehensive strategy, for use at ports, feed mills and laboratories, to detect contamination at the earliest point in the chain before being consumed by animals and the potentially serious consequences.
Through Work Package 2, the project set out to authenticate the origin (geographic and ingredients) of feed materials, key to identifying fraudulent practices within the feed supply chain. With ever increasing economic and environmental demands across the globe, Work Package 3 set out to investigate and identify potential new sources of animal feed materials and devise strategies for use by industry and regulators to ensure the quality and safety of each new animal feed component and source of material
The Project also set out to provide regulators with a better understanding of the real risk to the consumer from contaminated food. The partners carefully selected a number of key problems, e.g. dioxins, polychlorinated biphenyls (industrial products or chemicals, commonly referred to as PCBs), melamine and Salmonella spp. Work Package 4 used data (new and existing) to show how contaminants in feed materials can transfer into foods.
As with any major research project it is important to ensure that important information is passed on to stakeholders. A carefully designed dissemination plan was devised to ensure that end users of newly developed technologies received appropriate training and that regulators were provided with substantiated evidence to help make informed decisions.
The plan also included making information accessible to a non-scientific audience. The project website, www.qsaffe.eu and the e-Newsletters provided the mechanisms for consumers to be kept informed of developments.

Project Results:
Work Package 1 (WP 1)
The major aim of WP1 was to study the adaption of different analytical methodologies for the early control of non-conformity at three different levels, at laboratory level, at feed plant level and at port of entry level. Different matrices were studied: N-adulterants, veterinary drugs, pesticides, mycotoxins and plant toxins in soybean meal, persistent organic contaminants such as polychlorinated biphenyl (PCBs), flame retardants and mineral oil as well as fats and oils. The different techniques used were split into two main large groups of i) analytical methods, i.e. fingerprint methods like Near-infrared spectroscopy (NIRS), Raman Spectroscopy, Near-infrared (NIR) microscopy and hyperspectral imaging; and ii) mass spectrometry-based methods like Liquid chromatography/time-of-flight/mass spectrometry (LC-TOF-MS), LC-Orbitrap-MS, ambient mass spectrometry and Gas Chromatography Time-Of-Flight Mass Spectrometry (GC-TOF-MS). A complete strategy has been developed for the early quality and safety assurance in the feed chain. This strategy is mainly based on the methods developed in WP 1 and has been elaborated for two specific cases: (i) adulteration of feed materials that are included in animal feed diets as protein sources with melamine and its by-products (cyanuric acid, ammelide and ammeline) and with other types of adulterants/contaminants; soybean meal has been selected as a model matrix; (ii) adulteration of vegetable oils and fats with mineral oil or transformer oil or with alternative oil/fat sources with a lower economic value.
For the primary screening of adulteration of soybean meal with melamine and other types of adulterants/contaminants, NIRS or Raman showed to be the most suitable methods. NIRS has an advantage compared to Raman because it is already used in many feed companies. Within the QSAFFE-project, an on-line NIR sensor was successfully installed in a feed mill. Samples suspected for the presence of melamine can be confirmed by liquid chromatography-high-resolution mass spectrometry (LC-HROrbitrap-MS). Other suspect samples can be confirmed with LC-HRMS and LC-MS/MS multi-(class) methods for mycotoxins, pesticides, pyrrolizidine alkaloids and veterinary drugs.
For the primary screening of adulteration of vegetable oils with mineral oil, transformer oil or other oils, Raman is the method of choice. Within the QSAFFE project, a Raman instrument was successfully installed in a feed mill to screen feed oils as they arrived on-site. Samples suspected for mineral oil contamination can be confirmed by means of GC-FID. For samples where screening resulted in transformer oil contamination suspicions, a fully automated comprehensive Gas Chromatography analysis (GCxGC-ToF-MS) method can be applied as a post-screening technique to detect polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers (PBDEs), and polycyclic aromatic hydrocarbons (PAHs) and emerging flame retardants. Confirmation can be performed with more exhaustive sampling procedures and/or specific GC (or LC) MS-based techniques. Samples that have screened suspect for transformer oil can also be tested for dioxins, applying Calux screening and GC-HRMS or GC-MS/MS confirmation. If deviant Raman spectra are observed, the samples can also be checked for adulteration with other, e.g. lower priced or inferior quality oils/fats by means of DART-MS.
Comprehensive gas chromatography coupled to mass spectrometry has been demonstrated to be a proper tool for the early detection of non-conformity of multiple classes of persistent organic contaminants in fat resources used for feed, at a laboratory level. The method here presented comprises of a fast and simple sample treatment, GCxGC-ToFMS detection and automated data processing with semi non-targeted detection of chlorinated/brominated-containing compounds.
The clean-up procedure, which was based on Solid-phase extraction (SPE) with silica, was proven to be suitable and robust for matrices such as vegetable oils, fish oil and animal fats. The configuration of the GCxGC system, which consisted of reverse-type column combination: Rxi®-17Sil MS (30 m # 0.25 mm I.D. 0.25 mm film thicknesses) x Rxi®-5Sil-MS (1m# 0.18 mm, 0.18 mm film thicknesses), accomplished for the baseline separation of all critical PAHs groups, which was especially of concern when applying automated data processing with MetAlignID.
MetAlignID considerably reduced the time needed for data processing from 20 to 5 min/file. The method has been successfully validated with respect to false negatives for the six PCB markers (PCB 28, 52, 101, 138, 153 and 180), the seven PBDEs markers (BDE 28, 47, 99, 100, 153, 154 and 183), benzo[a]pyrene and the PAH4 markers (benz[a]anthracene, benzo[b]fluoranthene, benzo[a]pyrene and chrysene) and for six emerging brominated flame retardants. A screening detection limit (SDL) equal to or lower than the maximum regulatory level was always achieved, which would suffice to test the safety of feed material as regards PCBs or PAHs contamination.
The scripting data processing to identify spectral patterns was presented as a promising filter that allowed a rapid identification of chlorinated/brominated compounds in oils of different nature.
Another important objective of WP1 was to evaluate the added value of the newly developed and/or optimized methods in the frame of the QSAFFE project against methods currently used in official control. The evaluation was performed both from a quantitative and a qualitative point of view, and classifying the methods as screening or confirmatory/quantitative methods for the different analyte/matrix combinations were addressed within the QSAFFE project.
From a quantitative point of view, none of the identified existing methods could be directly applicable at the port of entry or feed mill level. The methods developed in QSAFFE are comparable to those used in official control with fit for purpose performance characteristics whilst providing added value by targeting a larger number of analytes for screening, quantification and confirmation.
From a qualitative angle, the thorough evaluation was only possible for veterinary drugs in Dried Distillers Grains with Solubles (DDGs) and mycotoxins in corn silage due to the data available. For veterinary drugs, the QSAFFE method appears to be competitive, in terms of solvent consumption, easiness of use, number of target analytes per run and cost per analysis.
Similarly for the mycotoxins in corn silage, the QSAFFE method is also competitive in terms of cost and easiness of use while also being a multi-method fit for the purpose of screening, quantifying and confirming 51 analytes in one run as opposed to only one of the existing ones.
For all other contaminants addressed in the study, namely melamine, pesticides, mineral oil and PCBs, PAHs, PBDEs, (brominated) flame retardants in the respective targeted matrices, the information provided is not sufficient to allow a qualitative evaluation on similar criteria.
The last objective of WP1 was to establish a comprehensive and innovative strategy for the early detection and screening of feed materials at the port of entry level (entrance level of importation in Europe), at the feed plant level and at the laboratory level. Questionnaires concerning NIR and Hyperspectral NIR methodologies have been completed and submitted. A paper has been created and submitted for publication.
Work Package 2
The major aim of work package 2 was to develop analytical tools and approaches to verify or to determine the origin of feed materials. In case of an embargo (e.g. within the EU) of products from certain regions or countries, due to an identified or non-identified hazard, the proof of the geographical origin is known to be essential. As an example for feed materials, DDGS (Distillers Dried Grains and Solubles) were chosen to demonstrate provenance authentication.
Sampling and Applicability of Analytical Techniques to DDGS
It was agreed that the focus of WP 2 would be set on authenticity testing of DDGS. This feed matrix was selected because of its worldwide and increasing production. In addition, feed deriving from bio-fuel production was identified by WP 3 as one of the emerging products of the worldwide production.
Together with the sampling plan, established at an early stage in the project, "DDGS samples/DDGS materials" had been defined according to the relevant numbers of the EU feed catalogue,
Commission Regulation (EU) No 575/2011: a) number 1.12.10 (Distillers' dried grains) and b) number 1.12.11 (Distillers' dried grains and solubles/Distillers' dark grains). It was aimed to collect 2-3 DDGS materials with different botanical origins (15 samples for each species) and corn DDGS from 2-3 geographical origins (50 samples for each origin). Finally, until October 2013, 196 DDGS samples were collected and distributed in six batches to WP 2 partners, comprising DDGS samples from Canada, China, India, USA, and Europe, produced from corn, wheat and other raw materials, as well as from the industrial process origin of either bio-ethanol or alcoholic beverage production. Different analytical techniques have been considered in the design of WP 2 and were under investigation for their suitability to authenticate DDGS: Near-infrared (NIR), Fourier transform mid-infrared (FT-MIR) and Raman spectroscopy, proton transfer reaction mass spectrometry (PTR-MS), direct analysis in real time-high resolution mass spectrometry (DART-HRMS) and stable isotope ratio analysis (by isotope ratio mass spectrometry, IRMS). In addition to the techniques originally planned for analysing DDGS, near infrared (NIR) spectroscopy and liquid chromatography-time of flight mass spectrometry (LC-QTOF-MS) were also applied.
In the initial stage of the project, it turned out that all analytical techniques involved in WP 2 indicated suitability to authenticate DDGS. All techniques were successfully adapted to the matrix DDGS and showed promising results, however, Raman spectroscopy could only be applied to the extracted fat fraction, but not to the solid DDGS samples directly. Specific analytical procedures have been developed and carried out by the WP 2 co-workers and have already been published as research articles (cf. www.qsaffe.eu).

Authentication of DDGS - Different Approaches using Chemometrics
Most of the analytical methods developed were untargeted (so called fingerprinting) techniques recording a large amount of data which are combined to dedicated statistical data analysis. A broad variety of statistical methods is in use, depending on the analytical technique, and often individual dedicated statistical approaches are required. With regards to authentication issues, it is imperative that the developed statistical models are validated in order to produce realistic and continuous models. However, for the relevant techniques neither an official validation procedure nor a scientific agreement are available so far. In order to provide guidance for the validation a proposal was developed, accepted by the WP 2 partners, and subsequently used for statistical data analysis. This validation guidance allowed and provided a maximum of flexibility to individual experts involved in WP 2, but also offered a certain degree of harmonisation for the evaluation of the applicability of the approaches.
In view of the initial motivation of the work the following questions were taken into consideration: Geographical origin was investigated using the corn DDGS samples, in particular the sample sets for the differentiation of China, EU and USA. Further models were built in order to classify the DDGS according to their botanical origin (here: corn, wheat and others). In addition, the method of production (bio-ethanol versus alcoholic beverages production) or differences within one continent of origin (e.g. different production facilities) were of interest and were evaluated with some of the analytical techniques and their dedicated statistical data analysis. Generally, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), soft independent modelling of class analogy (SIMCA) or the detection of single discriminatory variables were carried out to create the relevant statistical models for DDGS authentication.
The established models were typically validated by cross-validation using the leave-one-out method, where each sample is left out of the model calculation and subsequently predicted once (internal validation). Thereafter, these models were tested with separate data sets in order to perform further validation. For this, each model was tested by prediction of samples initially isolated from the calibration data set (external validation).
Furthermore, an additional sample set -called System Challenge- of 50 DDGS was used for prediction into the respective model (botanical origin, geographical origin, process). This test challenged the constructed models by a completely independent and external sample set and was conducted to test the developed models and analytical approaches in a "real life" situation, which will be the same when the techniques will be applied in laboratories in future. This procedure, according to our knowledge, shows a unique standing in the field of validation and proof of authenticity studies. Authentication of DDGS - Results
Encouraging results were obtained in terms of identification of the geographical and botanical origin of DDGS. All analytical techniques involved demonstrated a clear differentiation of the botanical origin, like corn DDGS vs. wheat DDGS. In addition, the geographical origin could be distinguished for corn DDGS coming from China, the European Union and USA by the techniques applied. The proof of the process origin remained the most challenging task, as most of the approaches were not able to distinguish between bio-ethanol and alcoholic beverage production (whiskey, vodka) correctly.
It was pointed out that in case of a decision cascade (e.g. when the approaches are used in the field), firstly the botanical origin should be identified before the proof of the geographical origin takes place within one botanical origin. Clearly, the overlap of geographical and process origin has to be considered within the data evaluation, however only few techniques were able to classify the process origin correctly.
Generally, the capability of stable isotope ratios as a useful tool for the identification of the geographical origin - already established in the food area - was confirmed by the results also for the feed material investigated. But also other analytical techniques, which were applied to authenticate DDGS, indicated very suitable results for an initial proof of the geographical origin; especially NIR and FT-MIR spectroscopy showed potential for relatively fast and cost-efficient applications. However, as these techniques are based on statistical analysis of the spectroscopic data, exchangeability of results and databanks between different facilities remains one future challenge. Especially the results after testing the models with an independent and blind sample set (called System Challenge) indicated that the developed strategies for DDGS authentication could be of valuable use in "real-life" situations. The respective results obtained for the assignment of the geographical origin (correct classification exceed 80 % for 6 of 10 approaches) underpin the future possibilities of the methods developed in WP2. However, for future use it will be necessary to continuously update the models by analysis of new authentic DDGS samples, from further locations and of further origins. The authentication approaches/models developed in WP2 could primarily be used to verify the origin of DDGS, and not to draw final forensic conclusions on the provenance, i.e. geographical origin. Whenever discrepancies from paper documentation systems are observed, further traceability procedures could be initiated.

Databases containing Analytical Data
Data of the collected and investigated DDGS samples have been compiled in one EXCEL master table, however, at the moment not available to the public. This includes meta-data, the results of discrete analytical parameters, particularly stable isotope ratios of bulk DDGS samples and their respectively extracted fat fractions, as well as the spectroscopic and mass spectrometric data. Also, basic nutritional data for crude protein, crude fibre, crude fat, crude ash and the moisture content, obtained by NIR spectroscopy and results of gravimetric fat content analysis are included. Easy sample identification by unique sample numbers defined in the EXCEL master table is guaranteed. The databases constructed could be used in future for verification of DDGS origin, e.g. during crises. However, the maintenance and extension of the generated databanks by analysis of new authentic samples from the single locations is essential in order to monitor developments and processes that may affect the authentication. If authentication approaches for DDGS were put into practice, it is recommended to extend the constructed databases accordingly. By this, the databases could be used to verify the origin of a DDGS sample/batch and could initiate traceability procedures in case of suspicion.

Applicability of Analytical Approaches for Feed Authentication (Industry Information)
In the course of the project it has been demonstrated that various analytical techniques coupled with either univariate or multivariate data analysis can be used to authenticate the botanical and geographical origin of an animal feed material (DDGS). By extension, this methodology should therefore be applicable to the authentication of other materials used in the animal feed sector, in the same way that has been proven for DDGS. All analytical techniques have therefore been evaluated with regard to their applicability and future use. Crucial points for a future application of the techniques tested for feed authentication have been taken into account: applicability, limitations, possibilities, costs, laboratory efforts/costs and transferability were evaluated. The results of this evaluation will also be published in a journal related to the field and will then be available to the industry and interested stakeholders.
Although some of the techniques are not currently routinely used in the industry and therefore would be difficult to implement, NIR spectroscopy is already an accepted technique in the animal feed sector. It would therefore be the first choice technique to apply the developed strategies into an industrial setting, especially for feed mills, who routinely run NIR spectra of their raw materials, to help authenticate botanical and geographical origin of these materials. However, to increase confidence in the results, an extensive database of spectral data would need to be built and maintained as well as using the most appropriate data processing from the available chemometric software packages.
Work Package 3
Horizon Scanning and Development of FeedRisk Register
A report and database for industry and governments on future trends in animal feed materials and for identifying and reducing the microbiological and/or chemical risks was produced. In the first part future trends in animal feed ingredients, how these trends were calculated from historical data and predictions up to year 2020 were reported. The second part is a prototype database (FeedRisk Register) based on data contained in the EU Rapid Alert System for Food and Feed (RASFF). The report presented an overview of the historical trends and future forecasts in the availability of a number of raw materials for animal feed. The overall trend showed the major raw material for animal feed into the future appeared to be corn, the use of which is predicted to rise for the next decade. Wheat is also predicted to rise but only modestly, barley and sorghum are predicted to remain relatively stable, while oats, rye and millet were predicted to have declining trends. All the oilseeds were associated with increasing production. Soybean meal, rapeseed meal, sunflower seed meal and palm kernel meal are predicted to continue their steady increases into the next decade. Peanut meal and copra meal also indicate modest increases in production for the next decade. Fish meal appeared to have declined in recent years, however it is forecast to remain relatively stable until the end of 2020.
A prototype database was constructed by QUB detailing the reported feed ingredients notifications generated during 2009, 2010 and 2011 from the RASFF. The six materials highlighted were Corn (maize), Corn by-products, Feed Wheat, Soya, Rapeseed and Fishmeal. However, for this prototype database, the principle can be extended to other important commodities used within the feed sector. A total of 267 notifications were identified for these commodities and were assigned where possible to one of twelve hazard subheadings; Salmonella, Veterinary drugs (legal & illegal), Mycotoxins, Process animal proteins (PAPs ), Pesticides (legal & illegal), Dioxins/PCBs, Mineral oils, Heavy metals, Melamine & derivatives, Enterobacteria, GMO, & 'Others'. The prototype FeedRisk register database was constructed in Excel. Using the frequency of the risk based on notifications and the country of origin of risk, a scoring system based on magnitude of the risk and a priority testing schedule based on the frequency and scoring system combination was developed. Based on these parameters and by multiplying the frequency of risk (total) by risk score, a value is obtained that can be used to prioritise analysis of each commodity. An outline of the priority testing ranking of each risk associated with individual commodities highlighted that the major primary risk for the commodities is Salmonella apart from corn which has a primary risk related to mycotoxins. Secondary risks associated with the commodities include illegal pesticides, GMO and Enterobacteria. Development of rapid, low cost screening methods for chemical risks Dioxins, dlPCBS and PAHs Within WP3, the issue of drying processes as a particular source of contaminants like dioxins (PCDD/Fs) was addressed. Drying with unsuitable materials possibly causes the sometimes increased levels in coconut fats and fatty acids, for this reason a set of dried copra samples were tested. Both the CALUX-bioassay (a rapid screening method) and GC/HRMS were applied and polycyclic aromatic hydrocarbons (PAHs) were also determined. Results demonstrated an elevated level of dioxins in one sample, just above the maximum level of 0.75 ng Toxic Equivalent Quantity (TEQ)/kg in the EU. This sample also showed an elevated response in the CALUX bioassay, using the routine acid silica clean-up. The sundried copra sample showed very low dioxin and PAH levels, indicating that the drying process is the main source of these compounds. Twenty samples of crude coconut oil were analysed with the DR CALUX bioassay using a specific clean-up for dioxins and dl-PCBs. Since many samples showed an elevated response, it was decided to analyse all samples with GC/HRMS. CALUX could easily predict the TEQ-level and correctly identify all ten samples with a level higher than the Maximum Limit (ML) of 0.75 ng TEQ/kg. DDGS samples were tested with DR CALUX and the levels estimated by comparison with a set of reference samples. Two samples showed an elevated response and were estimated to contain levels of 0.7 and 0.9 ng TEQ/kg. These samples were analysed by GC/HRMS and shown to contain non-detectable levels of dioxins and dl-PCBs well below the ALs. The results showed drying processes are a serious source of dioxins, PCBs, PAHs and potentially other contaminants. They should be clearly identified in HACCP evaluations of production processes and there should also be more focus on fuels used for drying of feed ingredients. Based on the samples of DDGS analysed within this project, there seems to be no reason for concern. The CALUX bioassay is capable of predicting the dioxin content in coconut oil, however there is still a need for more rapid methods to identify potentially contaminated samples.

Development and validation of analytical method(s) for simultaneous determination of free and conjugated mycotoxins in feed and bio-ethanol co-products (DDGS)
Two laboratories carried out mycotoxin analysis, Vysoká škola chemicko-technologická (VSCHT) and Food Environment and Research Agency (Fera). A modified Qu ick E asy Ch eap E ffective R ugged S afe (QuEChERS) procedure followed by Liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) was used for the analyses of a range of free and conjugated mycotoxins in feed and DDGS samples by VSCHT. The developed analytical method was accredited (according to ISO 17025) for the analysis of mycotoxins in cereal-based matrices. This method was fully quantitative and had lower Limit Of Quantification (LOQs) than the UPLC-MS/MS method developed and established in Fera. The Fera method was a 'dilute and shoot' method, after dilution and filtration the extracts are analysed by LC-MS/MS. This method can measure a total of 72 mycotoxins, and there are huge differences in response and therefore Level of Detection (LODs) between analytes. There are differences in toxicity and levels of interest for the analytes. For example for aflatoxin B1 the lowest legal limit for animal feed is 5mg/kg, while for others such as fumonisin it can be as high as 60mg/kg, and for other compounds there is no limit at all.
The method was in-house validated by analysis of spiked and blank samples of wheat. Samples were spiked both before and after extraction to allow the level of any matrix effects to be assessed.
Average apparent recoveries ranged from 61% (ergometrine) to 167% (ergotamine).The coefficient of variation for replicate analyses ranged from 1.7 to 23.3%. Signal suppression/enhancement was estimated to range from as little as 1% to as much as 80%. Matrix effects could be reduced by the use of matrix matched calibrants.

Survey of Feed Materials for mycotoxins
The work plan for this task was modified to include a survey of animal feed materials for mycotoxins. A wide range of animal feed materials samples were provided by partners John Thompson & Sons Ltd (THOM), China Agricultural University (CAU) and PROVIMI (Partners 11, 8 & 9) including novel products such as material from bioethanol production, from all regions of the world. In total 173 samples were analysed.
Mycotoxins were detected in the majority of products, but the profiles of toxins present varied by commodity. For example soya products contained cytochalasins, beauvericin and enniatins, while corn contained fumonisins, fusaric acid, beauvericin, deoxynivalenol (DON), low levels of deoxynivalenol-3 -glucoside (DON-3-G), HT-2 and T-2 toxin, moniliformin and zearalenone.
Wheat contained enniatins, as well as small amounts of beauvericin, DON and DON-3-G.
DDGS samples were mostly from bio-ethanol production (EU, US and China) and were maize-based. DDGS contained beauvericin, moniliformin, fumonisins and fusaric acid. The levels of beauvericin ranged from 485 to 1170mg/kg; all DDGS samples contained beauvericin. The highest level of fumonisin B1 measured was 2910mg/kg. Low levels of aflatoxin B1 (5-10mg/kg) were also observed in most samples, probably due to the concentration effect that occurs during the DDGS production process. This data can contribute to risk assessments for emerging mycotoxins. A publication describing the results of this survey is in preparation for submission (further information on publication date is not yet available, as of Dec 2014).
Monitoring of mycotoxins in DDGS and description of their fate within the DDGS production chain Input corn, 11 production intermediates and final DDGS collected across the technological chain of ethanol production were analysed for mycotoxins. The transfer of mycotoxins from input corn to the final DDGS samples was investigated. DON, glucoside-DON (D3G) and fumonisin B1, were detected in input corn, final DDGS and all DDGS production intermediates. Due to the known mass balance of the ethanol/DDGS production process, it was possible to deduce the changes of mycotoxins' levels among particular production steps. The significant increase of D3G and also slight increase of DON levels were observed during the liquefaction, saccharification and within the first part of fermentation stages, and were probably caused by the release of these compounds from the starch matrix. The decrease of DON and D3G levels in the second part of the fermentation phase were probably due to the applied yeasts, which are capable of absorbing some mycotoxins to some extent. This phenomenon was already described in the literature. After centrifugation, the main part of mycotoxins was transferred into the syrup obtained by evaporation of thin stillage. Significant decrease of detected mycotoxins occurred in the final step of DDGS production, probably as a result of using of hot drying air in the rotary dryers. Mycotoxins levels found in dry weight DDGS were comparable to those detected in input corn with only one exception, represented by fumonisin B1. Its levels were approximately two times higher in comparison to those measured in initial corn. Altogether 5 batches of samples were collected over the project period to investigate the transfer of mycotoxins from input raw corn to the final DDGS generated as a by-product of biofuel production. Although relatively low mycotoxin levels were detected in input corn used for production of DDGS, processing intermediates and final product, contamination was high enough to allow evaluation of mass balance of the particular DDGS production process for DON, DON-3-Glc and fumonisin B1 in the case of the first batch and DON in the case of the second batch (sampled in June and July 2012, respectively). In case of all three remaining series, the contamination of input corn was very low, mycotoxins levels were either below or close to the limit of quantification, therefore, it was impossible to confirm previous findings and compare mass balance for these series.
Generally, the transfer of DON from raw material across production intermediates into the final DDGS found in the batches no. 1 and 2 was very similar. Based on mass balance, the heat-induced DON decrease accounted approximately for one third of its original content. The influence of temperature on heat-induced DON content decrease in the last phase of the ethanol/DDGS production process was further investigated. To confirm the effect of temperature on DON degradation DDGS were tested for the presence of DON degradation products. The degradation products (which are not of toxicological concern) were found in the final product. These results confirm the assumption that the decrease of DON content in the final product is temperature-induced. This finding can contribute to reduction of mycotoxins content during optimised DDGS production. Development of rapid, low cost screening methods for microbiological risks A review of novel feed materials to identify the main microbiological hazards highlighted the main organism of concern as Salmonella spp. This concurred with the evidence found in the development of the FeedRisk register. Salmonella was selected as the target organism for which to develop a rapid and low cost screening assay. A loop-mediated amplification (LAMP) assay for detection of Salmonella enterica was selected from published studies, and two key modifications were made to it. The first modification involved incorporating a thermostable polymerase enzyme, newly-available from the manufacturer of the technology, to allow whole bacterial cells to be added directly to the reaction. This modification eliminated the need for prior nucleic acid extraction, and allows the assay to be harnessed directly to conventional culturing sample treatment, which will facilitate the development of a highly user-friendly analytical method. The second modification was to develop an internal amplification control, or IAC, for the LAMP assay, in order to be able to avoid false-negative result interpretation. This is the first time that the principles of IACs have been applied to LAMP. The IAC concentration was optimised, and the detection limit of the modified assay was defined as approximately 10A2 cells per reaction.
The method was fully validated in-house based on the principles as set out in ISO 16140:2003 "Microbiology of food and animal feeding stuffs - Protocol of the validation of alternative methods". Due to the success of the method it was agreed to extend the work programme to carry out a full international multi-laboratory validation study (MVS) of the method. This MVS was conducted according to the ISO validation protocol as a "Paired" trial. In practical terms, the ISO 6579:2002 procedure was followed simultaneously with the LAMP-based alternative method. Ten laboratories participated, the LAMP -based method gave exactly the same results as the ISO method. Thus, QSAFFE's alternative method for Salmonella detection in animal feedstuffs was demonstrated to be fully equivalent to the ISO method. The significant improvement is that the QSAFFE method can screen negative samples within 2 days as opposed to 5 days required by the ISO method. The validated method was subsequently used to conduct a survey of feed materials, finding 12 positive samples, demonstrating the applicability and robustness of the method in real testing situations.
Work Package 4
Carry-over of dioxins and PCBs in sheep
Levels of dioxins and dl-PCBs in sheep fat and in particular livers tend to be high. Several RASFF-alerts on sheep liver demonstrate this problem and also EFSA, the European Food Safety Authority, concluded that the frequent consumption, especially by women and young children, presents a potential risk. Therefore, a carry-over study was performed to study the relationship between intake of dioxins and dl-PCBs and the levels in meat and liver. Contaminated (2-3x Maximum Limit (ML)) and clean dried grass pellets were obtained. The contaminated grass came from a flood plain with contaminated soil. The sheep already contained elevated dioxin levels at the start of the treatment, but nevertheless experienced a strong increase in the dioxin levels in the liver, exceeding the old lipid based ML but also the new increased and product-based ML This is remarkable since the observed grass levels are quite common in the winter time, when sheep are also outside. When switching to clean grass, the liver levels decreased strongly within several weeks, only partly explained by the growth of the animals. This offers a potential solution for decreasing the levels in sheep livers. Data on the PCBs were less easy to interpret, partly due to the much lower levels in the grass and partly to the presence of PCBs in the straw bedding provided to the animals during the whole experimental period.
Levels in kidney fat were much lower and did not exceed the ML for dioxins. Also these levels decreased when switched to clean grass, indicating that there was no redistribution of dioxins from liver to kidney fat. Levels in meat, expressed on lipid base, were very similar to those in kidney fat. This is important for checking compliance, which is more easily done in the fat. Congener patterns in liver and kidney fat were quite different, showing increased storage in livers for the higher chlorinated dioxins.
Carry-over of dioxins and PCBs in dairy cows
A new study as replacement of the already performed melamine study was proposed to the advisory board and approved. The study deals with the carry-over of dioxins formed during fires or by a drying process, which are important sources of contamination of feedstuffs. Two contaminated materials were obtained, a dried beet pulp from Germany and maize silage that became contaminated by a fire where a substantial amount of PVC was burned. Dioxin levels were 2.5 and 1.1 ng TEQ/kg dm respectively. Both materials were fed for 5 weeks to dairy cows (3 for each feed), beet pulp at 4.6 kg/day, and maize silage at 14.6 kg/day. After 5 weeks the cows switched to clean feed. It is clear that milk levels approach the maximum levels for milk. Data will be used by f for congener-specific modelling and for calculation of the carry-over rates. An important conclusion is that the absorption and carry-over to milk of dioxins from these materials is much higher than that from fly ash. This is very relevant for estimating the potential impact from future incidents with drying or fires. PBK modelling of dioxins and PCBs in various species
PBK models that were thus far developed are based on the total TEQ level, i.e. the total sum of the dioxins and/or dl-PCBs after correction for their relative toxic potencies expressed in TEFs. In practice the 29 dioxins and dioxin-like PCBs show very different kinetics. This may have consequences when using the models for predicting the effects of a specific incident, since congener patterns vary depending on the source of the contamination. Data from previous studies and the studies described above were used to improve or built PBK-models for laying hens, sheep and cows. When data allowed, a congener specific modelling was applied in addition to the default based on the sum-TEQ. Data was examined from two different dioxin mixtures, one with a major TEQ contribution of lower chlorinated congeners (mix A), one with higher chlorinated congeners (mix D). It is clear that for mix A there is little difference between data and fitted model, since this mixture resembled the one used for building the model. However, when higher chlorinated congeners contributed most to the TEQ level, there was a clear deviation. This can be explained by different absorption in the GI-tract and excretion into the eggs. Similar behaviour was observed for dairy cows and the transfer to milk. This can be relevant when dealing e.g. with a pentachlorophenol contamination.
For laying hens also the ndl-PCBs from the previous study were modelled in a congener specific fashion: The time-related changes in congener patterns of both dioxins, dl-PCBs and ndl-PCBs were also studied based on the model.
The data on dairy cows and sheep generated in the QSAFFE project will receive further examination. The PBK models for the sum-TEQ for dioxins in laying hens and dairy cows were translated into formulas that allowed the inclusion in a platform built by RIKILT. This allows a more user friendly application of the model by interested parties. The model will be made available through the website. It includes an option for probabilistic modelling of data.
PBK Modelling of melamine in laying hens
The animal study was already performed by BfR before QSAFFE started. During an exchange between BfR and Rijksinstituut voor Volksgezondheid en Milieu (RIVM), a preliminary Physiologically-based Kinetic Modelling (PBK model) for melamine was built. This was extended based on the full dataset which was transferred from Bundesinstitut fuer Risikobewertung (BfR) to RIVM. There was a clear difference between the egg white and egg yolk levels, also in terms of kinetics. Data on cyanuric acid were initially not available due to contamination problems of the feed. When they arrived time did not allow their inclusion in the model.
Challenge testing of key pathogens in feed
Based on an inventory on the microorganisms of concern, it was decided to focus only on Salmonella. Three different feeds were selected, sterilized and then spiked with two levels of Salmonella. Subsequently the feeds were stored at 5, 15 and 300C for up to 3 months. It was shown that under all conditions Salmonella was able to survive but there was no growth of the microorganisms. However, the survival rate was much higher at 50C than at 300C.

Potential Impact:
A safe and healthy food supply for the European consumer is of paramount importance. In relation to foods that come from animal origin the rearing of healthy European livestock is highly dependent on the provision of high quality and safe feeds for these animals as this has major impact on the safety of the entire animal based food chain.
With regard to this, QSAFFE was designed to deliver following impacts across the broad range of stakeholders:
1. A full food chain approach and promoting safe feed for safe food production with support of scientific data
2. Production of safe feed of high quality to improve food safety and food quality.
3. Research that will support feed and food safety policies (at both industrial and governmental levels).
As a result of the efforts described in section on 'Description of main S & T results/foregrounds', analytical tools and strategies are available which allow delivery of better, faster and more economically viable means of ensuring the quality and safety of animal feeds in Europe. All planned impact demonstrators have been successfully delivered as described in next paragraphs. QSAFFE thus provides better ways of preventing contamination and fraud, identifying and assessing new risks and providing scientific evidence of the risks of transfer of microbiological and chemical contaminants from feed to food.
Strategies for early quality and safety assurance in the feed chain have been developed using existing testing methods and emerging technologies such as fingerprinting to deliver the 'Comprehensive and innovative strategy for the early detection and screening of feed materials at a port of entry, at a feed plant and at a laboratory level'. The strategy can be further adopted by importers of feed materials and feed manufacturers for early and cost-effective screening of feed materials for (i) adulteration of soybean meal with melamine and other types of adulterants/contaminants and (ii) adulteration of vegetable oils with mineral oil, transformer oil or other oils. The strategy will be also made available to the EC's Directorate-General for Health and Consumers (DG SANCO) and EFSA Contam Panel to assist in the development of policies aimed at safer and higher quality feed and food.
In addition, the report on 'Considerations for industry' focused on recommendations for industry for early detection of contaminants and adulterants in various feed materials by means of a range of spectroscopic and spectrometric methods has been prepared. This report, that summarizes key facts on (i) Safety and Assurance of the Feed Supply chain, (ii) Implementation of Surveillance testing based on HACCP plans and Risk Assessments, and (iii) potential of a range of analytical techniques for early detection of contaminants and adulterants in various feed materials at ports of entry, at the feed mills and in the laboratory with regards key characteristics such as of LOD/LOQ, sample throughput, analysis time, type of analyses (screening / confirmatory / quantitative), costs etc., will be made available to industry to enable early detection of problems (contamination or adulteration) at the entrance level to the feed supply chain that will help to assure better quality and safety of the feed chain.
Employing developed analytical tools, a feed mill pilot study was designed for trucks containing soybean meal intentionally contaminated during the unloading with certain types of contaminants. The application of appropriate analytical methodologies combined with chemometric tools and decision rules allowed the detection of the contaminants during the unloading of the truck, before processing. The QSAFFE beneficiary Provimi will continue the study and the real implementation of In-Line NIR (Near InfraRed Spectroscopy) inside the Piglet Feed Production process for the evaluation of raw materials at the entrance of the factory and to check the quality using the recognised Finished Products Quality system.
The traceability and authenticity of feed materials have been improved by determining which tests, conventional and fingerprinting, are the most useful in tracing origins of feed materials including those derived from biofuel co-products.
All involved analytical techniques, such as fast and economic spectroscopic techniques, as well as more sophisticated mass spectrometric approaches, demonstrated a clear differentiation of the botanical origin of model material, Distillers Dried Grains with Solubles (DDGS). Discrimination of the geographical origin of DDGS (China, EU, USA) could be shown as well. The capability of stable isotope ratios as a useful tool for the identification of the geographical origin, already established in the food area, was also confirmed for animal feed sector. The proof of the processing origin remained the most challenging task, as most of the approaches were not able to distinguish between bio-ethanol and alcoholic beverage production correctly.
Knowledge on authentication of animal feed material generated by the project underpins the suitability of the proposed analytical strategies (isotopic analysis, spectroscopic and spectrometric methods combined with a subsequent statistical analysis of data) for the 'proof of origin'. Developed analytical strategies will be made available for further implementation to control labs, industry and all other interested stakeholders. These strategies and respective statistical models would have to be adapted continuously in the future for further application. Maintenance and extension of the generated databanks by analysis of new authentic samples from the single locations/industries is essential in order to monitor developments and processes that may affect the authentication of certain sample; exchangeability of results and databanks between different facilities remains one future challenge.
As an outcome of this research the 'recommendations for industry for the implementation of isotopic and fingerprinting methods to improve feed materials traceability and authenticity' have been built up, assessing application potential of various analytical strategies coupled with appropriate chemometric tools (costs/laboratory efforts, applicability, limitations, transferability, etc.) to authenticate the botanical and geographical origin of an animal feed material. This report (Deliverable 2.6) is available to industry with the information which analytical technique could be applied for the authentication of an animal feed material to improve the traceability from the sites of harvesting and production to the feed mills. The introduction of all analytical strategies will have a major impact in countering illegal practices relating to fraudulent activity of supply of feed materials both in and from outside the EU.
Emerging chemical and microbiological risks have been identified from new types and sources of animal feed materials and new production processes. These directed the development of rapid, low cost screening tests to enable high quality and safety standards to be met.
A review of the historical trends, future trends and the influence of factors on animal feed have been completed and a database has been developed to analyse data from the EU RASFF. This will assist decision makers and control authorities prioritise testing of animal feeds for identified hazards. The CALUX bioassay, dioxins and PAH methods have been applied to different samples (coconut, DDGS). Drying processes applied to feed materials remain one of the most important sources of dioxins, although there appears to be little risk from the DDGS drying process highlighting which control measures can be introduced to reduce dioxin formation.
UPLC-MS/MS methods for the determination of free and conjugated mycotoxins have been developed and used to analyse feed samples as well as monitoring the fate of mycotoxins during bio-ethanol and DDGS production (changes were observed and mass balance calculated). The results from the drying studies aid the understanding of the fate of DON in dried DDGS, proving some mycotoxins are reduced in final products. These results will be published and can assist in providing advice for best practice in DDGS production and can be used to inform industry of any risks and to produce specifications for starting materials.
The results of the mycotoxin survey of various feed materials will provide information on the occurrence of a range of mycotoxins reviewed or under review by EFSA for which very little data is available. This will add to the current state of knowledge and allow assessment of potential risk from these mycotoxins. Data may be submitted to EFSA so it can be used to inform their future risk assessments.
The LAMP-based assay for detection of Salmonella spp. is a novel development which will advance this detection methodology towards practical application in routine analysis. The validation of this method by inter-laboratory validation is a significant output from the QSAFFE project. Publication of the method with validation data, and submission to the European Committee for Standardisation (CEN) will ensure it is accepted and used to reduce screening times for Salmonella, from 5 to 2 days. The video of the LAMP assay for Salmonella is on the QSAFFE website. The video on applicability of a LAMP-based method for detection of Salmonella in soya meal, is available from the QSAFFE website on http://www.qsaffe.eu/QSAFFE_InternationalEvent.html. The video allows dissemination of this novel detection method to a wide audience, demonstrates its simplicity, and makes it accessible to control laboratories that may not have considered using a DNA based method.
The transfer of chemical contaminants such as melamine and dioxins and micro-organisms (Salmonella spp.) from feed to food has been studied using pharmacokinetic models and animal studies to provide risk assessments for regulators.
The work aimed at studying and modelling the transfer of food contaminants (dioxins, polychlorinated biphenyls (PCBs), melamine) in food producing animals. Two specific issues were addressed in relation to dioxins and PCBs, being the transfer in sheep related to observed high levels in livers, and the transfer in dairy cows when fire / drying was the source. Two animal studies were successfully carried out, using material from incidents. The sheep study confirmed the relatively high accumulation of specific congeners in livers but also revealed that a washout period with clean feed resulted in a marked decrease. The cow study showed that the absorption of dioxins formed by a fire is higher than previously thought and that fires may be a serious threat to food production.
Both datasets were used for building and adapting physiologically based pharmacokinetic (PB-PK) models, which only to some extent and only in cows could be made congener specific in addition to toxic equivalency (TEQ)-based. For sheep, only a TEQ-based model could be developed.
Data from a previous study with laying hens were successfully used for congener specific modelling. As in cows, such a model gives a better prediction when the dioxin mixture is dominated by higher chlorinated congeners, due to a different absorption and excretion in animals.
The TEQ-based models for laying hens and dairy cows were successfully integrated with a model that allows an easier application in terms of different scenarios, including probabilistic modelling. This allows a more user friendly application of the model by interested parties that will be made available through the website.
The planned study on melamine in laying hens was already performed, but data were successfully modelled within QSAFFE. This included the different kinetics in egg white and yolk.
Challenge tests were carried out with Salmonella in feed stored at different temperatures and time periods. This revealed that there is a stronger decrease in the number of bacteria at higher storage temperatures.
All of these tests will provide methods to predict impact of hazards on food safety.
The technology transfer via dissemination activities such as presentations, peer-reviewed publications, public website, e-newsletters, international scientific conference and training program consisting of dissemination workshops, series of training events and hands-on trainings, providing knowledge transfer to relevant stakeholders and end-users, allowed the direct implementation of developed methods and approaches in practice.
The QSAFFE project public website (www.qsaffe.eu) providing information on the progress in the project activities, including dissemination activities and events, has been developed and updated regularly.
Three e-Newsletters have been issued summarising project goals and considerable progress achieved by individual WPs, including information on a wide range of training and dissemination activities organised by the project. All e-Newsletters have been distributed to the large number of stakeholders and are available from the project website.
An internal training program was completed by 9 trainings for 15 trainees. In addition, 5 training events for a broader scientific audience on various approaches for feed authentication and feed safety and quality control have been organised.
Consortium members presented project results at 38 international scientific events in the form of 59 lectures and 48 posters; moreover, 15 scientific papers, 2 papers in conference proceedings (total number of papers = 17) plus other contributions, articles, press release, video, interviews have been delivered until now (other are under preparation). All these dissemination activities enabled excellent visibility of the QSAFFE project to the international scientific community.
Two dissemination workshops, for industry representatives and regulatory labs, and a QSAFFE "International session on quality and safety of feed and food in Europe", for a wide range of stakeholders, were organised resulting in worldwide dissemination of the project achievements. Widespread dissemination of the knowledge generated by the project to a broad scientific and general public and industry through the range of dissemination tools enabled knowledge transfer to all interested stakeholders. A comprehensive training program consisting of intra-project training and an exchange program, a series of training events, dissemination workshops and international conference facilitated the knowledge exchange among both consortium members and all potential end-users. The training program resulted in capacity building of all involved participants and knowledge transfer of developed approaches and procedures for feed authentication and feed safety and quality control to the target audience.
QSAFFE established an effective cooperation among academics and government scientists with substantial experience in animal feed research and industrial companies, large and small, resulting in recommendations on improving feed authenticity, quality and safety for the feed industry. In this way, QSAFFE enabled dialogue resulting in identification of feed industry requirements on one hand and approaches and procedures for feed authentication and feed safety and quality control available or under development on the other hand.
All the deliverables of QSAFFE will form the basis of technological advances and will open new possibilities for further national and international research outside the scope of the project and thus largely contribute to the consolidation of the ERA (European Research Aea) and the increase of research investments. Some of the QSAFFE results are planned to be exploited and developed further in H2020 project(s) that are currently under preparation.

List of Websites:
QSAFFE Project website - www.qsaffe.eu
Co-ordinator - Queen's University Belfast. Contact person - Prof Chris Elliott