Trend watching and food safety control

Symposium SciCom 2015Brussels

Friday, 27 November 2015

Edited by the Scientific Committee and the Staff Direction for Risk Assessment of the Belgian Federal Agency

for the Safety of the Food Chain (FASFC)

Editors Etienne Thiry, Chair Scientific Committee FASFC Xavier Van Huffel, Director Staff direction for risk assessment FASFC Herman Diricks, CEO FASFC Federal Agency for the Safety of the Food Chain (FASFC) CA-Botanique Food Safety Center Boulevard du Jardin botanique 55 B-1000 Brussels Lay-out Claire Verraes, Expert Staff direction for risk assessment FASFC The contents reflect the views of the authors and not necessarily the views of the FASFC nor of the Scientific Committee. Reproduction is authorized provided the source is acknowledged.

5

TABLE OF CONTENTS Table of contents ................................................................................................................................................................ 5

List of abbreviations ......................................................................................................................................................... 7

Welcome ............................................................................................................................................................................. 11

Session 1. Trend watching in the food chain: methodology and general aspects ................................ 13

General introduction on trend watching and food chain control ........................................................... 15

Public health trends of foodborne diseases: Salmonella, Campylobacter and Listeria in Belgium as examples ................................................................................................................................................................... 21

New trends in society and food systems with possible impact on the safety of the food chain 27

Trends related to the phytosanitary situation in Europe and implications for national control programs........................................................................................................................................................................ 33

Session 2. Application of trend watching in food chain control .................................................................. 41

Trends related to chemical hazards/risks in the food chain: examples from the Belgian control plan................................................................................................................................................................................... 43

Application of trend watching in self-checking systems ........................................................................... 55

Application of trend analysis in food chain control: viewpoint of the Food Standards Agency (FSA) ................................................................................................................................................................................ 75

Application of trend watching in food chain control: expectations of the risk manager ............. 81

Conclusions ........................................................................................................................................................................ 91

7

LIST OF ABBREVIATIONS AA Acrylamide ADI Acceptable Daily Intake ALARA As Low As Reasonable Achievable AMCRA Antimicrobial Consumption and Resistance in Animals AMS Division for Agricultural Policy Analysis of the Department of

Agriculture and Fisheries ANSES Agency for Food, Environmental and Occupational Health & Safety

(France) BelVet-SAC Belgian Veterinary Surveillance of Antibacterial Consumption BMDL Benchmark Dose Lower Bound BRC British Retail Consortium bw body weight CI Confidence Interval EC European Commission EU European Union EFSA European Food Safety Authority FAFVAC Fédération des Associations francophones des vétérinaires

d’Animaux de Compagnie FAO Food and Agriculture Organization of the United Nations FASFC Federal Agency for the Safety of the Food Chain (Belgium) FBD Foodborne Disease FDA Food and Drug Administration (USA) FEVIA Federation of the Food Industry (Belgium) FSA Food Standards Agency (UK) FWO Research Foundation - Flanders (Belgium) HACCP Hazards Analysis and Critical Control Points ILVO Institute for Agricultural and Fisheries Research (Belgium) LOQ Limit Of Quantification MIRA Flanders Environment Report MRL Maximum Residue Limit MOE Margin Of Exposure PAS Publicly Available Specification PESTLE Political, Economic, Social, Technological, Legal and Environmental PPP Plant Protection Products RASFF Rapid Alert System for Food and Feed SCAR Standing Committee on Agricultural Research SciCom Scientific Committee of the Federal Agency for the Safety of the

Food Chain (Belgium) STEC Shiga-toxin producing Escherichia coli TACCP Threat Assessment Critical Control Points WHO World Health Organization WIV-ISP Scientific Institute for Public Health (Belgium)

6

7

LIST OF ABBREVIATIONS AA Acrylamide ADI Acceptable Daily Intake ALARA As Low As Reasonable Achievable AMCRA Antimicrobial Consumption and Resistance in Animals AMS Division for Agricultural Policy Analysis of the Department of

Agriculture and Fisheries ANSES Agency for Food, Environmental and Occupational Health & Safety

(France) BelVet-SAC Belgian Veterinary Surveillance of Antibacterial Consumption BMDL Benchmark Dose Lower Bound BRC British Retail Consortium bw body weight CI Confidence Interval EC European Commission EU European Union EFSA European Food Safety Authority FAFVAC Fédération des Associations francophones des vétérinaires

d’Animaux de Compagnie FAO Food and Agriculture Organization of the United Nations FASFC Federal Agency for the Safety of the Food Chain (Belgium) FBD Foodborne Disease FDA Food and Drug Administration (USA) FEVIA Federation of the Food Industry (Belgium) FSA Food Standards Agency (UK) FWO Research Foundation - Flanders (Belgium) HACCP Hazards Analysis and Critical Control Points ILVO Institute for Agricultural and Fisheries Research (Belgium) LOQ Limit Of Quantification MIRA Flanders Environment Report MRL Maximum Residue Limit MOE Margin Of Exposure PAS Publicly Available Specification PESTLE Political, Economic, Social, Technological, Legal and Environmental PPP Plant Protection Products RASFF Rapid Alert System for Food and Feed SCAR Standing Committee on Agricultural Research SciCom Scientific Committee of the Federal Agency for the Safety of the

Food Chain (Belgium) STEC Shiga-toxin producing Escherichia coli TACCP Threat Assessment Critical Control Points WHO World Health Organization WIV-ISP Scientific Institute for Public Health (Belgium)

11th Symposiumof the Scientific Committee

of the Belgian Food Safety Agency

TREND WATCHINGAND FOOD SAFETY CONTROL

Friday 27 November 2015Auditorium Pacheco

Pacheco Center – Finance TowerPachecolaan 131000 Brussels

9:00 RECEPTION

9:30 WelcomeEtienne THIRYProf. Université de Liège – Chair of the FASFC Scientific Committee

Chairs

Session 1. Trend watching in the food chain: methodology and general aspectsLieve HERMAN (Head of division ILVO-T&V – Member of the FASFC Scientific Committee)Claude SAEGERMAN (Prof. Université de Liège – Member of the FASFC Scientific Committee)

9:45 General introduction on trend watching and food chain controlDirk BERKVENSProf. Institute of Tropical Medicine – Member of the FASFC Scientific Committee

10:30 Public health trends of foodborne diseases: Salmonella, Campylobacter and Listeria in Belgium as examplesNiko SPEYBROECKProf. Université catholique de Louvain – Member of the FASFC Scientific Committee

11:00 COFFEE BREAK

11:30 New trends in society and food systems with possible impact on the safety of the food chainErik MATHIJSProf. KU Leuven

8

12:00 Trends related to the phytosanitary situation in Europe and implications for national control programsLars CHRISTOFFERSENHead of Sector Plant Health and GMOs Food and Veterinary Office European Commission

12:30 WALKING DINNER

Chairs

Session 2. Application of trend watching in food chain controlBruno DE MEULENAER (Prof. Ghent University – Member of the FASFC Scientific Committee)Katelijne DIERICK (Head of service Food borne pathogens WIV-ISP)

14:00 Trends related to chemical hazards/risks in the food chain: examples from the Belgian control planWendie CLAEYSExpert Staff direction for risk assessment FASFC

14:30 Application of trend watching in self-checking systemsClaire VERRAESExpert Staff direction for risk assessment FASFC

Koen DE REUSenior researcher & team leader ILVO-T&V

15:10 COFFEE BREAK

15:40 Application of trend analysis in food chain control: viewpoint of the Food Standards Agency (FSA)Terry DONOHOEHead of Emerging Risks FSA

16:10 Application of trend watching in food chain control: expectations of the risk managerHerman DIRICKSChief Executive Officer FASFC

16:40 ConclusionsEtienne THIRYProf. Université de Liège – Chair of the FASFC Scientific Committee

9

10

11

WELCOME Prof. dr. Etienne Thiry Prof. Université de Liège Chair of the Scientific Committee of the FASFC, Belgium E-mail: etienne.thiry@ulg.ac.be

BIOGRAPHY

Etienne Thiry was graduated as doctor in veterinary medicine in 1980 and in veterinary sciences (PhD) in 1985. He was recognized in 2001 as diplomate of the European College for Veterinary Public Health. He is professor of veterinary virology and viral diseases, Université de Liège, Belgium, and part time professor of veterinary virology at the free university of Brussels. He won the International Pfizer award in 1996, the Gaston Ramon award by the French Academy of Veterinary Medicine in 2008 and the Prix de la Francophonie by the Fédération des Associations francophones des vétérinaires d’Animaux de Compagnie (FAFVAC) in 2011.

He was recognized by the European Society for Veterinary virology as honorary member in 2009. Etienne Thiry is chairman of the Scientific Committee (SciCom) of the Belgian Federal Agency for the Safety of the Food Chain (FASFC) and chairman of the expert committee for animal health and well-being at the French Food Safety Agency (ANSES). He is also member, previously acting- and vice-chairman, of the European Advisory Board on Cat Diseases. He was also coordinator of the European Union “Better Training for Safer Food” project on “Animal Health Prevention and Control of Emerging Animal Diseases” (2012-2013). His research interests cover several aspects of animal virology, especially the study of animal virus-host interactions and the short term evolution of viral populations through genetic recombination and reassortment in herpesviruses, noroviruses, hepeviruses and orbiviruses. These scientific activities generated more than 450 papers in specialized scientific journals.

12

Trend watching and food safety control is the topic of this XIth annual symposium of the Scientific Committee of the Belgian Food Safety Agency. A huge cloud of data is already available and covers all the segments of the food chain in Belgium, in Europe and worldwide. Instead of darkening our knowledge, it may open new areas of investigation. Provided a good organization and a correct and intelligent exploitation of these data, trend watching can improve the safety of the food chain. Taking advantage of the knowledge of the past, we can better orientate the efforts for the future. Several methodologies are already enforced to do such as trend observation which may evolve to a trend analysis when a statistical analysis is applied. Trend observation and analysis can, among others, be used to forecast with the best precision the incidence of food chain hazards with the intention to assist risk managers and to provide recommendations about prioritization and allocation of resources with the best expected efficacy. Our Scientific Committee is a board of experts who gives independent scientific advice to the Agency and the Minister in all matters related to the competencies of the Agency, especially in risk assessment, regarding food safety but including also animal and plant health. We enforce a policy of impartial and independent scientific consultation and transparent communication and management of conflicts of interest. We acknowledge the strong support of the Agency through its Chief Executive Officer, the Director-General of Control Policy and the Staff direction for risk assessment. It is my great pleasure to thank the workgroup members Prof. D. Berkvens, Prof. B. De Meulenaer, Dr. K. Dierick, Dr. L. Herman, Prof. A. Legrève, Prof. C. Saegerman, Prof. N. Speybroeck, Dr. X. Van Huffel and Ir. C. Verraes for their invaluable contribution in the scientific organization of this symposium. These acknowledgements are extended to the Agency and the complete Staff direction for risk assessment for providing the needed human and financial resources. A highly significant causal relationship can be found between the quality of this symposium and the top scientific level of the speakers who agreed to share their knowledge and expertise with us today. They are warmly acknowledged. Your presence as participants reveals that the selection of this topic by the Scientific Committee was more than appropriate and meets the concerns of scientists, risk assessors, risk managers and stakeholders involved in the analysis of risks in the food chain. Indeed we all pursue the same objective: to provide safe food to our society. I wish you a very fruitful symposium.

13

SESSION 1. TREND WATCHING IN THE

FOOD CHAIN: METHODOLOGY AND GENERAL ASPECTS

14

15

GENERAL INTRODUCTION ON TREND WATCHING AND FOOD CHAIN CONTROL

D. Berkvens*, C. Verraes1, X. Van Huffel1, N. Speybroeck2,5, L. Herman3,5, K. Dierick4

1 FASFC, Belgium 2 Université catholique de Louvain, Belgium 3 ILVO, Belgium 4 WIV-ISP, Belgium 5 Scientific Committee FASFC, Belgium * Speaker: Prof. dr. ir. Dirk Berkvens Prof. Institute of Tropical Medicine, Belgium Member of the Scientific Committee of the FASFC, Belgium E-mail: dberkvens@itg.be

BIOGRAPHY

Dirk Berkvens was employed by the Belgian Development Cooperation Agency between 1976 and 1990 and lived and worked in Zambia, first as teacher at the national agricultural graduate school, afterwards as researcher in a veterinary project. He joined the Institute of Tropical Medicine in 1991, where he is head of the Unit of Veterinary Epidemiology within the Department of Biomedical Sciences.

16

Trend monitoring is traditionally subdivided into two domains: trend watching (trend observation) and trend analysis (EFSA, 2010). Trend analysis is defined in a very specific and restricted way as “involving some form of statistical analysis”. Trend watching then becomes a broad collection of formal and informal means and ways to look at data, being possibly a precursor for a full statistical analysis. EFSA (2010) refer to the latter as “passive monitoring of data for possible trends” and state that trend analysis means “statistical analysis and epidemiological interpretation of occurrence/prevalence data in order to detect significant changes over time”. It is not obvious whether a clear-cut division of trend monitoring into two well-defined, mutually exclusive subdomains is always feasible or even desirable. One should definitely avoid giving the impression that trend observation is something entirely subjective as opposed to trend analysis, considered entirely objective. Both in fact require a decision based on interpretation of information. The term “trend watching” somehow appears to have been hijacked and it has acquired a more societal, behavioral meaning, possibly being more interested in trendsetters than in trends. We therefore prefer the slightly less zestful trend observation for the general non-statistical component of trend monitoring, knowing very well that this term also does not necessarily covers the full panoply of approaches, means and methods available for the trend observer. The online version of the Oxford Dictionaries defines trend as 1) a general direction in which something is developing or changing, 2) a fashion and 3) a topic that is the subject of many posts on a social media website within a short period of time. It is obviously the first item in the above list that concerns us, but at the same time it should be noted that trend monitoring is not confined to trends of “How many times did something go wrong?”. Trend monitoring can also be applied to trends of characteristics of food/feed production/consumption, ranging from risks arising because of new habits or shifts in the foodstuff basket (e.g. the consumption of insects) to risks arising from the application of new technologies (e.g. nanotechnology) or changes in food/feed preparation (e.g. changes in process parameters in the production of meat-and-bone meal). Societal trends with respect to food safety will specifically be dealt with in one of the presentations (“New trends in society and food systems with possible impact on the safety of the food chain”). Although it is not always possible to produce an unambiguous division, it could be argued that the above examples pertain to the second item of the triplet definition of risk (Kaplan & Garrick, 1981): 1) “What can go wrong?”, 2) “What is the chance that it will go wrong?” and 3) “If it goes wrong, what are the consequences?”. The most common type of trend monitoring could then be said to deal with the second item of this definition, i.e. what is the incidence/prevalence of things gone wrong. The actual purpose of trend monitoring depends on the situation. The primary aim of the exercise is of course to ascertain whether or not a trend exists. In many

17

cases, this is where the analysis ends: a trend has been observed and no explanation is obvious. Alternatively, the next step is to explain the (presence or absence of a) trend. This consists of demonstrating a (ideally causal) relationship between one or more factors and the observed trend. Factors can be (e.g.) altered habits, control measures, … Lastly, a trend may be used to guide the establishment of sampling frameworks, both to target specific sub-populations or to adjust sample sizes because of prior knowledge. In the domain of food chain control, trend monitoring can be employed to:

Track the evolution of incidence or prevalence of norm infractions, possibly as a prediction or early warning system to pro-actively initiate control measures. Examples discussed during the symposium are (e.g.) “Trends related to the phytosanitary situation in Europe and implications for national control programs”; “Application of trend analysis in food chain control: viewpoint of the Food Standards Agency (FSA)”; “Application of trend watching in food chain control: expectations of the risk manager”. Semi-automatic data-mining of the results of the FASFC control program is achieved by means of an in-house developed interface in R, guiding the users through the formal regression analysis and offering assistance in the interpretation of the results.

Evaluate the evolution of a norm infraction after installing control measures as an aid to assess the efficacy of the control measures in question. Examples discussed during the symposium are (e.g.) “Public health trends of foodborne diseases: Salmonella, Campylobacter and Listeria in Belgium as examples”; “Trends related to chemical hazards/risks in the food chain: examples from the Belgian control plan”.

Optimize sampling frames. Knowledge about trends can in theory be used to reduce sample sizes. An example can be found in the SciCom advice 21-2012 (SciCom, 2012).

The remainder of this introduction is devoted to technical matters. Rather than subdividing trend monitoring into two distinct groups it is probably more correct to note that it is actually a continuum of approaches and techniques. The bottom end of the continuum consists of the impression that something may show a trend, without any numbers or other evidence to back it up. This may form the starting point of a more formal approach, attempting to prove or disprove the impression. Next comes eyeballing numbers and graphs, without formal statistical analysis. This is the first (and often necessary) quantitative step in the inquiry, but it is imperative to understand the limits and possible pitfalls when applying visual

18

inspection to raw data or their graphical representation: quality of the data, weighting, unknown population-at-risk, possible bias, … At the other end of the continuum stand the statistical analyses, there again existing a quasi continuous sequence from non-parametric approaches to the most complex autoregressive models. This is not the place to provide an overview of the various statistical approaches, the speakers will mention and explain the methods used in the respective case studies. Attention should however be drawn once more to the fact that statistical tests are meant to deal with random errors and that they are not able to uncover systematic errors. Formal statistical analysis is in fact on a par with visual inspection in that respect and a statistically significant result is no guarantee for the absence of confounders or other types of bias. It still is the end-user’s responsibility, after having obtained the necessary clarifications of the statistician, to make sure the statistical analysis could be carried out in the first place. The presence of repeated measures in time series deserves special attention. As already pointed out by Granger & Newbold (1974), many so-called significant trends, revealed by traditional regression models, are in fact simple random walks and the significant results are a consequence of ignoring the autocorrelation present in the sequential data. Last but not least, a significant statistical result does not automatically entail a trend that is biologically meaningful. It is the end-user of the analysis results who has to attach the biological, chemical or epidemiological explanation to the trend. This is obviously not relevant when the end-user has observed the trend and wanted a statistical confirmation, but it is especially important when the trend is discovered as the result of a data mining effort.

REFERENCES EFSA, 2010. Technical specifications for monitoring Community trends in zoonotic agents in foodstuffs and animal populations. EFSA Journal 8, 1530. Granger, C., Newbold, P., 1974. Spurious regressions in econometrics. Journal of Econometrics 2, 111–120. Kaplan, S., Garrick, B., 1981. On the Quantitative Definition of Risk. Risk Analysis 1, 11–27. SciCom, 2012. Advice 21-2012 of the Scientific Committee on the optimization of the methodology of the control program: sampling size for the study of trends (dossier Sci Com 2011/01: self-tasking initiative). Available online: http://www.favv-afsca.fgov.be/scientificcommittee/advices/_documents/Advice21-2012_000.pdf (English summary); http://www.favv-afsca.fgov.be/wetenschappelijkcomite/adviezen/_documents/ADVIES21-

19

2012_NL_DOSSIER2011-01_000.pdf (Dutch version); http://www.favv-afsca.fgov.be/comitescientifique/avis/_documents/AVIS21-2012_FR_DOSSIER2011-01_001.pdf (French version).

20

21

PUBLIC HEALTH TRENDS OF FOODBORNE DISEASES: SALMONELLA, CAMPYLOBACTER AND LISTERIA IN BELGIUM

AS EXAMPLES N. Speybroeck*, B. Devleesschauwer1, L. Diallo2, D. Berkvens3,9, S. Bertrand4, O. Vandenberg5, J. Haagsma6, A. Havelaar1, L. Vanholme7, S. Quoilin4, Y. Dupont4, P. Brandt8, C. Maertens de Noordhout2

1 University of Florida, USA 2 Université catholique de Louvain, Belgium 3 Institute of Tropical Medicine, Belgium 4 WIV-ISP, Belgium 5 Saint-Pierre University Hospital, Belgium 6 Erasmus University Medical Centre, The Netherlands 7 FASFC, Belgium 8 University of Texas, Dallas, USA 9 Scientific Committee FASFC, Belgium * Speaker: Prof. dr. Niko Speybroeck Prof. Université catholique de Louvain, Belgium Member of the Scientific Committee of the FASFC, Belgium E-mail: niko.speybroeck@uclouvain.be

BIOGRAPHY

Prof. dr. Niko Speybroeck is professor at the Université catholique de Louvain, with responsibility for teaching epidemiology and statistics. His main research interests are quantifying and modeling disease burden. He has worked in Zambia, for the African Union in Malawi, at the Institute of Tropical Medicine in Belgium and between 2004 and 2006 at the World Health Organization (WHO) in Switzerland. He is currently involved in field-studies in a variety of countries, such as Peru, Ethiopia, Vietnam, Uganda and Zambia.

N. Speybroeck is a member of the SciCom of the FASFC, of the Medical Panel of the Research Foundation - Flanders (FWO) and of the working group on screening programs of the Flemish Agency of Public Health. He further is active within the EU (European Food Safety Authority (EFSA), member of the BIOHAZ panel) and at the WHO (member of the Foodborne Disease Burden Epidemiology Reference Group). N. Speybroeck has published about 200 peer reviewed papers and is a member of the editorial board of PloS ONE. He was awarded a Marie Curie fellowship and was bestowed the Merial award from the Netherlands Society for Parasitology. Expertise: Biostatistics, Epidemiology, Modelling, Global Health, Vector-borne and Food-borne diseases.

22

ABSTRACT

This study reports the current and future incidence of campylobacteriosis, salmonellosis and listeriosis in Belgium. Based on predictions obtained from time series models, the results indicate that in Belgium the incidence of campylobacteriosis may almost double by 2020 as compared to 2012, while the incidence of salmonellosis and listeriosis may remain stable, if conditions remain the same as in 2012. Knowing the current and future incidence may help policy makers to prioritize existing hazards and as a consequence helps to decide where to use the available resources.

INTRODUCTION

Campylobacteriosis, salmonellosis and listeriosis are important foodborne diseases in Belgium and across Europe. Despite it being difficult to obtain exact number of deaths and illnesses caused by these foodborne illnesses, campylobacteriosis and salmonellosis were the most frequently reported illnesses of foodborne origin in Belgium in 2012 (WIV-ISP, 2012). Listeriosis shows a lower incidence but a relatively high mortality and hospitalization rate as compared to the two former pathogens. The real current and future public health impact of these foodborne diseases (FBD) remains however unreported in Belgium. This study fills this gap by providing estimates of present and future incidence of campylobacteriosis, salmonellosis and listeriosis in Belgium using advanced time series models. The objectives of this study are: 1) to select appropriate time series models to forecast the number of cases of campylobacteriosis, salmonellosis and listeriosis in Belgium until 2020 and 2) to calculate current and future incidence of campylobacteriosis, salmonellosis and listeriosis in Belgium using predictions obtained with the models. This bacterial FBD incidence study in Belgium generates valuable information for decision-makers and researchers, and might consequently assist in improving resource allocation, targeting interventions and monitoring possible impacts.

METHODOLOGY

Data were obtained from the Belgian Scientific Institute for Public Health (WIV-ISP) that collects data on laboratory-confirmed campylobacteriosis, salmonellosis and listeriosis human cases in Belgium. The time series of laboratory-confirmed cases ranged from 1993 to 2013 for Campylobacter, from 2001 to 2012 for Salmonella, and from 2011 to 2013 for Listeria.

23

The variety of patterns of the studied pathogens asks for a careful selection of suitable time series models. Indeed, time series require appropriate analytical tools instead of using regression tools that do not account for the seasonal patterns and serial correlation in an optimal manner. An in-depth visual inspection (e.g. investigating auto-correlations and partial auto-correlations) is used to guide an appropriate model choice. Simulated forecasts were used to obtain 95 % confidence intervals (CI).

RESULTS AND CONCLUSIONS

The results of this study are presented and discussed at the symposium on “Trend Watching and Food Safety Control”. These results are considered as a first step in estimating and predicting the current and future incidence of campylobacteriosis, salmonellosis, and listeriosis in Belgium.

(a)

(b)

200

400

600

800

1000

Cas

es C

ampy

loba

cter

Jan 1992

Jan 1994

Jan 1996

Jan 1998

Jan 2000

Jan 2002

Jan 2004

Jan 2006

Jan 2008

Jan 2010

Jan 2012

Jan 2014

Month

0

500

1000

1500

2000

Cas

es S

alm

onel

la

Jan 2001

Jan 2002

Jan 2003

Jan 2004

Jan 2005

Jan 2006

Jan 2007

Jan 2008

Jan 2009

Jan 2010

Jan 2011

Jan 2012

Month

24

(c)

Figure 1: Observed number of cases of campylobacteriosis (a),

salmonellosis (b) and listeriosis (c) in Belgium. Figure 1 shows the observed trends of campylobacteriosis (a), salmonellosis (b) and listeriosis (c). The Campylobacter trends needed a dynamic linear model because the data showed a stochastic drift, varied locally and suffered from parameter stability issues around the seasonal component of the data. The salmonellosis trends asked for a breakpoint model for time series because of the changes in Salmonella spp. transmission and control in the past (Collard et al., 2008). The listeriosis trends asked for a model that can deal with small numbers of observed cases. The selected models needed to forecast the trends were a dynamic linear model for campylobacteriosis (Durbin & Koopman, 2001), a Bai-Perron two breakpoints model for salmonellosis (Bai & Perron, 1998, 2003) and a Poisson autoregressive model for listeriosis (Brandt & Williams, 2001). The salmonellosis forecast showed a continuing decline after 2005, while an upward trend until 2020 is seen for campylobacteriosis. The salmonellosis decline coincides with EU wide measures to control salmonellosis, indicating a success story. With the limited number of observations, there is no evidence for an upward or downward trend for Listeria. The selected models resulted in a predicted monthly number of cases for 2020 of 1081 (95 % CI: 430-1741) for campylobacteriosis, of 212 (95 % CI: 43-381) for salmonellosis and of 6 (95 % CI: 2-11) for listeriosis compared to the last observed number of cases of respectively 633, 264 and 6 in 2012. These forecasts could be used to calculate the burden for 2012 and 2020. Assuming a constant environment, the results indicate that the numbers of cases of campylobacteriosis may almost double by 2020 as compared to 2012, while the numbers of salmonellosis and listeriosis will remain stable. We notice that this conclusion holds if conditions will remain the same as in 2012. For example, given that most listeriosis cases are seen in the

05

1015

Cas

es L

iste

ria

Jan 2011

Apr 2011

Jul 2011

Oct 2011

Jan 2012

Apr 2012

Jul 2012

Oct 2012

Jan 2013

Apr 2013

Jul 2013

Oct 2013

Month

25

elderly (i.e. > 70 years old, especially in patients with cancer or immunocompromised persons), an ageing population may result in an increasing incidence of listeriosis. This study shows some limitations, including the use of reported data, the exclusion of covariates in the time series models, and the short time series for listeriosis. It is practically impossible for the public authorities to consider each individual hazard, given the limitations in financial resources, and therefore choices have to be made. Knowing the burden helps to prioritize existing hazards and as a consequence helps to decide where to use the available resources. From this study it appears that additional efforts may be needed to reduce the risk of food contamination with Campylobacter spp.

REFERENCES Bai, J., Perron, P., 1998. Estimating and testing linear models with multiple structural changes. Econometrica 66, 47-78. Bai, J., Perron, P., 2003. Computation and analysis of multiple structural change models. Journal of applied econometrics 18, 1-22. Brandt, P. T., Williams, J. T., 2001. A linear Poisson autoregressive model: The Poisson AR(p) model. Political Analysis 9, 164-184. Collard, J. M., Bertrand, S., Dierick, K., Godard, C., Wildemauwe, C., Vermeersch, K., Duculot, J., Van Immerseel, F., Pasmans, F., Imberechts, H., Quinet, C., 2008. Drastic decrease of Salmonella Enteritidis isolated from humans in Belgium in 2005, shift in phage types and influence on foodborne outbreaks. Epidemiology and infection 136, 771-781. Durbin, J., Koopman, S. J., 2001. Time Series Analysis by State Space Methods. Oxford University Press. WIV-ISP, 2012. Données de surveillance du Centre National de Référence des Salmonella et Shigella, Belgique 2012, Rapport 2012. Available online: http://bacterio.iph.fgov.be/reporting/reportspdf/SalmonellaFR.pdf.

26

27

NEW TRENDS IN SOCIETY AND FOOD SYSTEMS WITH POSSIBLE IMPACT ON THE SAFETY OF THE FOOD CHAIN

Prof. dr. ir. Erik Mathijs Prof. KU Leuven, Belgium E-mail: erik.mathijs@ees.kuleuven.be

BIOGRAPHY

Erik Mathijs is Professor of Agricultural and Resource Economics at the Department of Earth and Environmental Sciences of the University of Leuven, Belgium. He holds an MSc in Bioscience Engineering (University of Leuven, 1991) and a PhD in Agricultural Economics (University of Leuven, 1998). He teaches agribusiness management and introductory principles in agricultural and resource economics. His research focuses on the transformation of the agricultural and food system towards sustainability and resilience and more specifically on the role of niche innovations, such as agro-ecology, organic farming, community-supported agriculture, insect rearing, biofuel crops, etc. He is coordinator of the FP7 project TRANSMANGO and the Horizon2020 project

SUFISA – both on the transformation of the European agricultural and food system. He acted as reporter and chair of the expert group of the 3rd (2011) and 4th (2015) Foresight Exercise for the EU’s Standing Committee on Agricultural Research (SCAR).

28

ABSTRACT

The bioeconomy goes beyond a substitution of fossil-fuels with bio-based materials. It embraces concepts such as recycling – giving rise to a circular economy – which challenges existing food safety practices and legislation. This trend may be exacerbated by new developments such as the sharing economy and new insights in nutrition science.

INTRODUCTION

While forecasts aim to predict the future as good as possible – thus building on projecting certain trends into the future, foresight focuses on future uncertainties. As developments are uncertain, it cannot be predicted whether these developments will actually break through or not. This paper highlights one trend and four signals that may or may not have an impact on the safety of the food chain. These insights are based on a recent European foresight exercises (Mathijs et al., 2015) and a systems analysis of the Flemish agricultural and food system (VMM, 2012).

TREND: THE RISE OF THE BIOECONOMY

In 2012, the EC launched the strategy for “Innovating for sustainable growth: A bioeconomy for Europe”. The strategy, together with its Action Plan, aims “to pave the way to a more innovative, resource efficient and competitive society that reconciles food security with the sustainable use of renewable resources for industrial purposes, while ensuring environmental protection” (EC, 2012). According to the European Bioeconomy Strategy, the bioeconomy or bio-based economy “… encompasses the production of renewable resources and their conversion into food, feed, bio-based products and bio-energy. It includes agriculture, forestry, fisheries, food and pulp and paper production, as well as parts of chemical, biotechnological and energy industries” (EC, 2012). However, the bioeconomy is more than a simple addition of sub-sectors. It can be seen as the set of existing relations between society and the biosphere in several aspects: provision of goods and services, the emission of pollutants and negative externalities but also the production of positive externalities to ensure that the biosphere continues to be functional for future generations. The bioeconomy concept is built on two premises. First, current biomass is being underexploited, as many waste streams are not used in an optimal way. More materials and more energy could be extracted from current biomass streams. Second, the biomass potential can be upgraded by increasing current yields by closing yields gaps, increasing productive land, introducing new or improved species that may or may not be generated by various biotechnological advances, and introducing new and improved extraction and processing technologies (Mathijs et al., 2015). The use of living matter (biomass) for economic purposes

29

has been part of society’s development for millennia. The bioeconomy has contributed to society ever since (McCormick & Kautto, 2013). Why, then do we refer to a transition to a bioeconomy? The reason is mainly related to the tremendous advancement in scientific knowledge and in technologies that have opened unprecedented possibilities of creating more value from living matter, including the development of new chemicals, materials, etc. Is the bioeconomy a miracle solution that generates less externalities than the fossil-based economy? Experience from the past warns us away from accepting uncritically the rhetoric of technological miracles. The application of a technology always has consequences that the inventors of that technology did not intend and that are often not foreseen. To anticipate unintended consequences of breakthrough technologies it is necessary to be aware that their impact depends very much on how people will organize around the opportunities and threats opened up by them and how legal and social rules will regulate their use (Mathijs et al., 2015).

SIGNALS SIGNAL 1: THE CIRCULAR ECONOMY: SYNERGY OR TRADE-OFF BETWEEN

SAFETY AND ENVIRONMENT? Part of the bioeconomy is the circular economy. The cascading approach, based on the principle that any matter can be reused or recycled, addresses the dilemma of the best use of biomass, but it does not address the issue of waste reduction per se. The concept of waste is inherent in the costs of “reuse or recycle”. Waste is generated where the (economic and ecological) costs of “reuse and recycle” are higher than the value created. To address this problem, the concept of the circular economy has been developed. A circular economy is “… an industrial system that is restorative or regenerative by intention and design. It replaces the end-of-life concept with restoration, shifts towards the use of renewable energy, eliminates the use of toxic chemicals, which impair reuse and return to the biosphere, and aims for the elimination of waste through the superior design of materials, products, systems and business models” (MacArthur Foundation, 2014). According to the MacArthur Foundation, a circular economy is based on three principles. The first principle is that, in an ideal circular economy, waste does not exist, as products are designed for a cycle of disassembly and reuse. The second principle implies a strict distinction between consumable and durable components of a product. Consumables should be returned to the biosphere without harm after a cascading sequence of uses, contributing to its restoration. Durables are designed to maximise their reuse or upgrade. To encourage the circularity of durables, these products are leased, rented or shared rather than sold, so that the owner will have the responsibility of retiring them after use and

30

starting a new cycle. The third principle is the use of renewable energy to fuel the process. The concept of circularity links to the principle of durability of material goods. The higher the number of cycles of repair, reuse or remanufacturing, the lower the ecological footprint of a product. At the same time, the longer the time of each cycle, the lower the demand for resources to create new products. In a circular economy, processing plants placed in adjoining layers of the cascading ladder are located close to each other, and firms are encouraged to collaborate to explore synergies in the respective material flows. “New generation” biorefineries process multiple feedstocks to produce multiple products. Industrial clustering, designed to adapt the logistics to the opportunities offered by the circular approach, may reduce considerably the costs of biomass management and would radically reduce waste (Mathijs et al., 2014). The circular economy is an official concept in the EU. In its Communication on a circular economy (EC, 2014), the EC pledges to further analyse the major market and governance failures which hamper the avoidance and reuse of material waste; establish a reinforced partnership to support research and innovative policies for the circular economy; facilitate the development of more circular models for products and services, encourage the cascading principle in the sustainable use of biomass; further integrate circular economy priorities into EU funding; set targets for reuse and recycling of waste. Using waste and by-products as input for food production is strictly regulated for reasons of food safety. For instance, compost from private households cannot be used in professional production. The implementation of the bioeconomy – partly addressing the increasing scarcity of key resources – may put pressure on these regulations. Also, emerging initiatives described in signal 2 and 3 are already experimenting with these concepts.

SIGNAL 2: ECOLOGICAL INTENSIFICATION OF AGRICULTURAL PRODUCTION Ecological intensification entails increasing primary production by making use of the regulating functions of nature. Its practices range from the substitution of industrial inputs by ecosystem services to the landscape-level design of agroecosystems. Ecological intensification entails a shift from the study of individual species in relation to their environment to the study of groups of organisms or polycultures in relation to each other and their environment (Tittonnel, 2014). It is based on the synergetic effects of combinations of ecosystem service processes, rather than the current focus on how single service processes work in isolation (Bommarco et al., 2013). Functional ecology and community ecology are key scientific disciplines that are emerging to support what could be called precision ecology. These disciplines are strengthened by

31

recent advances in digital technologies and approaches both at the molecular level (supported by the various –omics platforms) and at the landscape level. Ecological intensification entails using natural compounds instead of synthetic compounds, which is in line with the bioeconomy strategy. However, also natural compounds may have food safety implications that may not yet be fully understood.

SIGNAL 3: URBAN AGRICULTURE, SHORT SUPPLY CHAINS AND THE SHARING ECONOMY

While short supply chains have been around for some time now in the form of on-farm sales and farmers’ markets, in the last decade new initiatives have emerged in which the consumer takes a much more prominent role. As a result, the distinction between the producer and the consumer is becoming less and less strict, giving rise to the prosumer – a term coined by futurist Alvin Tofler in 1980 in his book The Third Wave (see VMM (2012) for a more comprehensive list of new developments). These initiatives usually take place in urban areas, but may take various forms, such as collective gardens, community-supported agriculture (in which consumers become member of the farm and take on certain activities such as harvesting), etc. An interesting food safety case has been the Spilvarken initiative in the city of Ghent (since 2014). Its aim is to reintroduce pigs into the city and to feed these pigs only with food waste from households. Another interesting case is the Thuisafgehaald.be, an online platform to enable the sharing of food surpluses (since 2012). These are all expressions of the sharing economy or peer-to-peer economy. Consumer-to-consumer interactions challenge many existing regulations related to tax laws, labour laws, but also food safety regulations.

SIGNAL 4: CHANGING PARADIGMS IN NUTRITION SCIENCE Increasingly nutrition scientists prescribe to avoid as much as possible processed and fast foods and to eat as diverse as possible. The reason is that obesity and many diseases are in fact related to the composition of our gut microbes (Spector, 2014). A recent study published in Science revealed that farm dust protects children against allergy and asthma (Schuijs et al., 2015). So, the cure is to get exposed more to microbes, not less. These insights question our understanding of and ask for a redefinition of the concept of “hygiene” (Spector, 2014).

REFERENCES Bommarco, R., Kleijn, D., Potts, S. G., 2013. Ecological intensification: harnessing ecosystem services for food security. Trends in Ecology and Evolution 28(4), 230-238.

32

EC, 2012. Innovating for sustainable growth: A bioeconomy for Europe. Commission staff working document. EC, Brussels, Belgium. EC, 2014. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. Towards a circular economy: A zero waste programme for Europe (COM/2014/0398 final/2). EC, Brussels, Belgium. MacArthur Foundation, 2014. Towards the Circular Economy: Accelerating the scale-up across global supply chains. Prepared in collaboration with the Ellen MacArthur Foundation and McKinsey & Company. World Economic Forum, Geneva, Switzerland. Mathijs, E., Carus, M., Griffon, M., Last, L., Gill, M., Koljonen, T., Lehoczky, E., Olesen, I., Potthast, A., 2015. Sustainable Agriculture, Forestry and Fisheries in the Bioeconomy - A Challenge for Europe. 4th Foresight Exercise for the SCAR, Brussels. McCormick, K., Kautto, N., 2013. The bioeconomy in Europe: An overview. Sustainability 5, 2589-2608. Schuijs, M. J., Willart, M. A., Vergrote, K., Gras, D., Deswarte, K., Ege, M. J., Madeira, F. B., Beyaert, R., van Loo, G., Bracher, F., von Mutius, E., Chanez, P., Lambrecht, B. N., Hammad, H., 2015. Farm dust and endotoxin protect against allergy through A20 induction in lung epithelial cells. Science 349, 1106-1110. Spector, T., 2015. The Diet Myth: The Real Science Behind What We Eat. Weidenfeld & Nicholson, London. Tittonnel, P., 2014. Ecological intensification of agriculture – Sustainable by nature. Current Opinion in Environmental Sustainability 8, 53-61. VMM, 2012. Transition to a sustainable agro-food system in Flanders: a system analysis. MIRA Topic Report in collaboration with AMS, Department of Agriculture and Fisheries.

33

TRENDS RELATED TO THE PHYTOSANITARY SITUATION IN EUROPE AND IMPLICATIONS FOR NATIONAL CONTROL

PROGRAMS Mr. L. Christoffersen Head of Sector Plant Health and GMOs, Food and Veterinary Office, European Commission, Ireland E-mail: lars.christoffersen@ec.europa.eu

BIOGRAPHY

Lars Christoffersen graduated in 1990 from the Agricultural and Veterinary University of Copenhagen as an agronomist in the line of plant production. He has worked in Denmark first as an agricultural college teacher and since in the Danish farmers’ extension services, both as a local extension officer in plant production and at the national headquarters in the department for development of IT farm management tools. He has also worked in Malawi, Africa with EU funded development aid, in particular in the areas of agriculture and natural resource management. In 1999, he joined the EC’s Food and Veterinary Office in Ireland where he now heads the sector for plant health and GMOs.

34

ABSTRACT

Two important sources of information for monitoring phytosanitary trends in the EU are Member States’ reporting of 1) interceptions made of non-compliant consignments in the trade (mainly imports) and 2) outbreaks and findings of harmful organisms on the EU territory. Selected data and trends are presented and discussed. Their implications for national control programs are also briefly addressed.

TRENDS IN INTERCEPTIONS MADE IN THE TRADE FOR PHYTOSANITARY REASONS.

Commission Directive 94/3/EC stipulates how EU Member States shall notify the Commission and other member States of interceptions of consignments from 3rd countries. Council Directive 2000/29/EC stipulates in Article 12(4) that Member States must also notify if they find non-compliant consignments originating in other Member states. The tool used for these notifications has for more than 10 years been EUROPHYT, which is an on-line web-based notification and rapid alert system. Since its inception, EUROPHYT has been hosted, managed and continuously developed by DG Health and Food Safety of the EC via a dedicated group of specialized personnel and IT staff ensuring day-to-day monitoring and management of the system and database, as well as coordinating on-going system maintenance and upgrades. The total number of notifications in the system now approaches 100,000 with 6,500 to 7,000 being added each year. The Commission has since 2013 published annual reports. The following is based on the latest such report and includes data up to the most recent complete year (2014). Of the 6,662 interceptions received in 2014, 96 % concerned imports, while 4 % were internal EU interceptions. These proportions have been fairly stable over the years with only a slight shift of 1-2 % over the last couple of years towards imports (see Figure 1).

Figure 1. Number of interceptions per year

0

2.000

4.000

6.000

8.000

2010 2011 2012 2013 2014

From TC

From MS

35

There are three main reasons for interceptions of consignments from third countries: 1) presence of harmful organisms (37 %), 2) non-compliant wood packaging material accompanying all kinds of goods (30 %) and 3) non-compliant or absent phytosanitary certificates (25 %). Figure 2 below shows the distribution on these reasons since 2010. It is noteworthy that the proportion of interceptions with harmful organisms has increased over the years, so that even if the total number of interceptions has remained somewhat stable, the proportion of interceptions with the highest risk, i.e. with presence of harmful organisms has shown an upwards trend.

Figure 2. Reasons for interceptions of imports The increase in harmful organism interceptions has mainly taken place in two categories of commodities, fruit/vegetables and wood packaging material. For fruit and vegetables, the interceptions have increased from 1,123 interceptions in 2010 to 1,734 in 2014. They furthermore have shown significant shifts in the origin. This typically happens when the EU enters into dialogue with exporting countries with many interceptions and restrictions on exports are imposed, either by the country itself or by the EU. Other countries in the region of export then pick up the gap in the market, perhaps based on an even poorer production and export certification system than in the countries where problems with interceptions were first identified. Figure 3 shows examples of such developments. Thailand had until 2010 a very high number of interceptions. Action was taken to address this, which had a negative impact on the possibilities for export for a while. Then Vietnam started having interceptions and when this was addressed, Cambodia started having interceptions. On the other hand, action taken against one country can in some cases encourage a country with similar exports and problems to improve their situation. A ban on import of mango and selected other commodities from India at the beginning of 2014 had significant

0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

100%

2010 2011 2012 2013 2014

Other reasons

PC> problems with declaration PC: illegible, fake, expired PC: absent

Non-compliant WPM

Non-compliance with derogation Prohibited goods

HO presence

36

positive impact on the interceptions, not only from India, but also from Pakistan whose mango exports are even more significant than that of India and who responded very quickly with solutions when approached by the EU to avoid a similar ban.

Figure 3. Harmful organism interceptions from selected countries

Increases can also be caused by new commodities, not yet regulated and therefore not subject to systematic control, being identified by one or more Member States as a risk. As other Member States follow suit, interceptions increase. The best recent example of this is peppers (Capsicum), which was subsequently regulated in 2014. Figure 4 shows this development together with increases for other selected commodities, which are not yet regulated.

Figure 4. New commodities with harmful organism interceptions Wood packaging has over the last decade and a half emerged as a major phytosanitary risk. It is the most likely cause of the pinewood nematode (Bursaphelenchus xylophilus) outbreaks in Portugal and Spain and of many outbreaks of longhorn beetles (Cerambycidae) across Europe. Systematic treatment according to International Standard for Phytosanitary Measures No. 15

0

100

200

300

400

500

600

2010 2011 2012 2013 2014

GHANA

CAMBODIA

INDIA

UGANDA

THAILAND

VIETNAM

PAKISTAN

0

50

100

150

200

250

2010

2011

2012

2013

2014

37

became a requirement for all imported wood packaging material from 2005. However, this did not solve all problems and over the past years, there has been an increasing number of interceptions with harmful organisms from certain countries, in particular China. There are significant differences between Member States in the number of interceptions they make and this does not always seem to be directly correlated to the amount of imports. There are also significant changes from year to year in some Member states. Figure 5 shows the statistics for the ten Member states that had the highest number of interceptions with harmful organisms the last five years.

Figure 5. Member States with most harmful organism interceptions

TRENDS IN HARMFUL ORGANISM OUTBREAKS FOUND IN THE EU

Export certification of risk material in exporting countries and import control in Member States are the main line of defense against the introduction into and spread within the EU of new harmful organisms. However, like elsewhere in the world, such systems do not guarantee a 100 % exclusion of harmful organisms and each year there are numerous findings and outbreaks on EU territory reported by Member States to the Commission, as required in Article 16 of Council Directive 2000/29/EC. While for interceptions there is a longstanding tool, EUROPHYT for handling and analyzing of data, such a system is only just being developed for outbreaks. The Commission has nevertheless over the years carried out some analysis of the data and intends this year to publish the first annual report with such analysis. The number of annual notifications is much smaller than for interceptions, but the work involved for the Member State and resources required for each is generally much, much higher. Some outbreaks, when they are discovered are too

0

200

400

600

800

1000

1200

UK NL DE FR ES CH SE IT BE AT

2010

2011

2012

2013

2014

38

large for immediate eradication and control would have to be maintained for many years, perhaps indefinitely to prevent further spread. Figure 6 shows the number of notifications, which has been fairly stable in the period 2010-2014.

Figure 6. Notifications of outbreaks and findings of harmful organisms in the EU

The numbers cover a range of harmful organisms of varied nature and seriousness. Figure 7 shows the distribution by taxonomic groups. While over 90 % of the harmful organism interceptions at import are insects, the proportion in the outbreaks is less than 50 %. The high proportion at import is not surprising, since the controls are based on visual inspections where insects are usually the easiest to detect. In the outbreaks, the diseases caused by fungi, bacteria, viruses, viroids and phytoplasmas are more prominent. They have had time after introduction to develop symptoms in their hosts and thus be detected visually. Often, as mentioned above, this occurs so late that further spread has taken place and eradication is difficult. The most serious recent example is Xylella fastidiosa in Puglia (Italy) and now also in Corsica (France). Until outbreaks were found in the EU, the bacteria had never been intercepted in imports.

2010 2011 2012 2013 2014

213 266

200 203 200

16

5

5 20 19

2 Update and New finding

Updates

Notifications of harmful organism

39

Figure 7. Distribution of outbreaks on taxonomic groups

In spite of the relative success of import inspections in detecting insects, some nevertheless escape detection and enter and establish in the EU. In 2014, these included two particularly serious pests. One is Trioza erytreae (citrus psyllid). It is not very harmful in itself, but it is the vector of the extremely serious bacterial disease, citrus greening. The insect’s presence in Northern Portugal and Spain constitutes a serious risk. Other parts of the world have experienced that the insect vector first spreads into an area and increases in number; a few years later, citrus greening is detected, then more detections occur every few years until the disease is widespread. Another is Popilia japonica (Japanese beetle), found in Northern Italy. It is a polyphagous pest with potential for serious damage in maize, fruit trees, nurseries, etc.

DISCUSSION AND IMPLICATIONS FOR NATIONAL CONTROL PROGRAMS

Databases like EUROPHYT and the parallel system being developed for outbreaks are extremely useful for collecting and analyzing data for purposes of risk analysis and management. However, they do require a high degree of maintenance and supervision in order to remain efficient and provide data of a quality that can be compared, compiled and analyzed. Furthermore, while they record and draw attention to risks, including new risks, the data do not necessarily reflect precisely the real risk or a real trend; to some extent they reflect how well Member States inspect imported consignments and survey their territory and what their priorities are – where you look, you find and the better you look, the more you find! Nevertheless, some general conclusions can be drawn from the presented data. It is also essential that Member States themselves regularly analyze the EU interception data in order to identify the shifts in risk origins identify new commodities of risk check that their interception statistics are at par with those

Bacterium; 49

Fungus; 21

Mite; 5 nematode;

2 phytoplasma 4

Insect; 94

virus; 22 Viroid; 3

40

of other Member States for the same product/origin combinations and adapt their import control program accordingly. The Commission has recently started publishing monthly an alert list with the most intercepted commodity/origin combinations. This is intended as a warning for the countries that find themselves on the list and for the traders, exporters and importers, but it is equally intended to help the plant health services of the Member States identify the above mentioned shifts and trends. The continued high number of interceptions of non-compliant wood packaging material and the increasing number of interceptions of harmful organisms in such material is worrying. It is only possible to inspect a very small proportion of the incoming material. Therefore it is essential that Member States target the inspections according to risk commodity/origin combinations. Furthermore, when a non-compliance is found, the most dissuasive measures possible should be taken in order to prevent future re-occurrences in the same trade. The low prevalence of disease detection in the import controls raises the question as to whether the current system, largely based on visual inspection, is adequate, especially for material for planting or further propagation, where diseases have both time and potential to develop and spread after introduction of the material. Experience has shown that early detection and action is paramount for eradicating or controlling outbreaks of harmful organisms. It is also sometimes necessary with early support from the EU, be it regulation, enforcement or economic support, to facilitate the often drastic measures needed at local level. It is anticipated that the new system for outbreak notifications will facilitate rapid reporting, analysis and follow-up of outbreaks, both at EU and national level.

DISCLAIMER

The text is published under the author’s own responsibility and the views presented are not intended to reflect the views of the European Commission or DG SANTE.

REFERENCES EC. Food and Veterinary Office - Europhyt Annual reports. Available online: http://ec.europa.eu/food/plant/plant_health_biosecurity/europhyt/annual_reports/index_en.htm. EC. Non-EU trade alert list. Available online: http://ec.europa.eu/food/plant/plant_health_biosecurity/non_eu_trade/alert_list/index_en.htm

41

SESSION 2. APPLICATION OF TREND

WATCHING IN FOOD CHAIN CONTROL

42

43

TRENDS RELATED TO CHEMICAL HAZARDS/RISKS IN THE FOOD CHAIN: EXAMPLES FROM THE BELGIAN CONTROL

PLAN W. Claeys*, J.-F. Schmit1, C. Matthys2,4, W. Steurbaut3,4 1 FASFC, Belgium 2 KU Leuven, Belgium 3 Ghent University, Belgium 4 Scientific Committee FASFC, Belgium * Speaker: Ir. Wendie Claeys Expert Staff direction for risk assessment, DG Control Policy, FASFC, Belgium E-mail: wendie.claeys@favv.be

BIOGRAPHY

Wendie Claeys holds a PhD in Applied Bioscience Engineering (KU Leuven, Belgium). After a position as postdoctoral researcher at the Centre for Food and Microbial Technology (KU Leuven, Belgium), she joined the Staff direction for risk assessment of the FASFC in 2005, where she is mainly responsible for food chain related chemical risk evaluation.

44

ABSTRACT

Whereas surveillance data of the national food safety control program offer a large pool of information over a large time frame, the observation of trends is not straightforward as it depends on assumptions made and the methodology chosen. By means of two case-studies, namely acrylamide (AA) and plant protection products (PPP), the possibilities and constraints of using such surveillance data for identifying trends related to chemical hazards in food are illustrated.

INTRODUCTION

Trends related to chemical hazards in food can be explored at different levels, namely at a “prevalence” level, such as the detection or conformity percentage, at a concentration level or at the exposure level. The level chosen will influence the interpretation of possible underlying relationships (e.g. causes), patterns or trends observed as well as inferences or predictions that might be made. In order to reliably detect and monitor trends, an adequate number of samples as well as a sufficiently long monitoring period are essential starting points. An additional prerequisite relates to the homogeneity of the data in the time series to be compared. In the case of the surveillance of a chemical contaminant in different food categories, this would include the number of samples in annual data collections, the accuracy of food categorization throughout the years, possible preparation of the samples for analysis and documentation of the analytical methods used and of all relevant sample information (e.g. preparation conditions, ingredients, etc.). In what follows, some of the possibilities and constraints of using surveillance data for observing possible (temporal) trends are briefly illustrated by means of two relevant chemical hazards in the food chain, namely AA and residues of plant protection products. For more detailed information, it is referred to two recent scientific opinions of the SciCom of the FASFC in which these case studies were thoroughly elaborated (SciCom, 2014; 2015).

ACRYLAMIDE IN FOOD

Acrylamide (AA) is a food process contaminant with carcinogenic and genotoxic properties. It is spontaneously formed at temperatures above 120 °C (and low moisture), principally via the Maillard reaction and mainly in foods containing free asparagine and reducing sugars. Intensive research resulted in various mitigation options and several are implementable in the food chain (FDE, 2014). From the onset of the AA issue around 2002, the FASFC monitors the AA content of different foods. Based on these monitoring results, the potential progress at the industrial and food service level in minimizing the AA level of food products was

45

evaluated in terms of the AA content and in terms of the dietary exposure (SciCom, 2014). TRENDS RELATED TO ACRYLAMIDE LEVELS IN FOOD ON THE BELGIAN MARKET The different foods sampled within the FASFC surveillance program were classified in different categories (Table 1). Most samples concern ready-to-eat foods. Given the heterogeneity of samples within a food category (e.g. the category of breakfast cereals contains samples with and without chocolate pieces, with toasted or with puffed cereals) and the fact that not every year the same number and types of products are analyzed within a food category, it was chosen to evaluate possible changes over time by comparing two larger time periods (2002-2007 versus 2008-2013). By merging the data over a larger period of years, it is assumed that the heterogeneity in terms of sampling is more or less leveled. The decrease in heterogeneity is shown in the yearly comparison of the AA levels of breakfast cereals compared to the pooled average levels of the 6-year periods (Figure 1 (a)). The yearly comparison does not show any downward trend in contrary to the period analysis. However, due to the diversity within the food category of breakfast cereals, it is unclear whether the decline of the pooled average level is genuinely due to efforts undertaken by the industry or to the basket of samples. Table 1. Acrylamide levels (µg/kg) of food on the Belgian market between 2002 and 2013

Food category

2002-2007 2008-2013 n° of

samples (< LOQ)

mean(a) P50 P95 max n° of

samples (< LOQ)

mean(a) P50 P95 max

breakfast cereals 71 (13) 185 ± 167 120 518 674 89 (39) 145 ± 144 63 456 670 (*) potato crisps 97 (5) 609 ± 555 485 1500 3200 54 (1) 375 ± 229 310 725 1300 (*) French fries 137 (28) 236 ± 326 170 604 3300 136 (43) 268 ± 325 218 630 2500 coffee 73 (8) 293 ± 375 200 957 2522 108 (3) 548 ± 626 330 1730 3800 (*)

roasted coffee 52 (5) 285 ± 59 170 1303 2522 56 (0) 269 ± 33 220 488 1800 (*) instant coffee 21 (3) 313 ± 45 290 792 810 52 (3) 847 ± 106 612 3170 3800 (*)

coffee substitute 29 (0) 2621 ± 895 2600 3920 4700 55 (0) 2915 ± 1111 2956 4652 5400 bread & rolls 71 (53) 38 ± 30 25 83 230 121 (97) 32 ± 44 25 66 400 (*) toast 39 (12) 130 ± 102 120 312 430 26 (5) 129 ± 120 80 390 460 biscuits (b) 50 (13) 167 ± 244 116 316 1514 53 (18) 142 ± 190 70 524 1113 sweet spiced biscuits 17 (0) 346 ± 187 270 694 760 10 (1) 339 ± 273 297 806 860

gingerbread 47 (1) 689 ± 568 450 1770 2100 59 (9) 225 ± 150 240 454 530 (*) chocolate 21 (6) 198 ± 202 112 700 750 26 (8) 74 ± 57 57 210 249 (*) cereal bars 20 (14) 61 ± 49 50 181 190 37 (27) 104 ± 138 50 264 820 popcorn 45 (7) 229 ± 242 150 802 1100 57 (8) 212 ± 130 180 422 470 baby biscuits 54 (17) 240 ± 303 135 1022 1217 123 (58) 117 ± 155 50 362 1200 (*)

(*) Significant differences (p < 0.05) between pooled 2002-2007 and 2008-2013 data based on the Kruskall-Wallis test; (a) mean ± standard deviation (with AA levels < LOQ = ½ LOQ); (b) sweet spiced biscuits are excluded

46

A total of 771 and of 954 food samples were analyzed between 2002 and 2007 and between 2008 and 2013, respectively. The most important changes observed are a significant decrease of the AA content in potato crisps and gingerbread (respectively -38 % and -67 % of the 6-year average), and a significant increase (almost twice as high) in coffee (Table 1). In contrast to potato crisps, the AA content of French fries did not change much. A yearly comparison even suggests an upward (although not significant) trend of AA levels (Figure 1 (b)). Remarkably, the variance on the AA levels measured in French fries is relatively small between 2005 and 2009, but widens from 2012 onwards. Given that the fries analyzed were primarily sampled at the level of catering (e.g. chip shops, community kitchens), the question arises whether operators have adopted a lax attitude about minimizing the AA content of their fries. The increase observed of the AA level of coffee seems to be mainly due to a rise of the AA level of instant coffee and not so much of roasted coffee. In general, higher AA levels are measured in instant coffee compared to roasted coffee (Table 1). The average AA content of coffee substitute remained essentially unchanged, although an annual comparison of the AA levels shows a gradual upward trend of both the average and the variance of results (Figure 1 (c)). The samples concern mainly coffee substitutes based on chicory. European monitoring results show higher AA levels in chicory based coffee surrogates (on average 2,942 µg/kg) than in cereals based coffee surrogates (on average 510 µg/kg) (EFSA, 2015).

(a) (b)

(5)

(20)

(10)

(11)

(10) (15)

(12)

(16) (16)

(15)

(15) (15)

0

200

400

600

800

AA (µ

g/kg

)

(5) (33)

(20)

(21)

(39) (19)

(21)

(20)

(19)

(35)

(20)

(21)

0

200

400

600

800

AA (µ

g/kg

)

47

(c) (d)

Figure 1. Yearly comparison of AA levels (µg/kg, mean + 95 % confidence interval) measured in (a) breakfast cereals, (b) French fries, (c) coffee substitute and (d) chocolate on the Belgian market between 2002 and 2013 (with between brackets the number of samples analyzed each year; AA levels < LOQ = ½ LOQ) A comparison of the two time periods, shows additionally a downward trend for the AA content of breakfast cereals, bread and rolls, chocolate and baby biscuits (Table 1), but these changes appear to be less pronounced. Regarding bread & rolls, the decrease observed for the average AA level after 2008 is probably largely artificial and merely the result of a modified method of analysis. Given that > 99 % of the AA of bread is situated in the crust and that for some 80 % of the samples the AA level was below the limit of quantification (LOQ), it was decided around 2009 to analyze the AA content solely in the bread crust with conversion of the result to the full bread, in order to obtain more accurate results. With respect to chocolate, it is remarked that this food item was not sampled every year (Figure 1 (d)). The comparison concerns the merged data of 2002-2003-2004 with those of 2009-2011-2012, and particularly in 2004 higher AA values were measured (an average of 312 µg/kg in 2004 compared to less than 100 µg/kg in the other years) which may have introduced some bias.

TRENDS RELATED TO THE ACRYLAMIDE INTAKE OF THE BELGIAN POPULATION

In a next step, it was evaluated if also at the level of dietary exposure a reduction is seen. Given the large time frames of 6 years that are compared, possible modified eating patterns should preferentially be accounted for as well. However, at present only one food consumption survey is available for adults, children and adolescents. Changes related to the intake therefore mainly reflect the changes

(9)

(10) (10)

(8) (10) (10)

(10)

(10) (7)

0

1000

2000

3000

4000

5000 AA

(µg/

kg)

(6) (5)

(10)

(7) (9)

(10)

0

200

400

600

AA (µ

g/kg

)

48

observed in AA levels, and only a slight, but insignificant decline of the overall dietary exposure is observed (Table 2). Table 2. Acrylamide intake of the Belgian population (µg/kg bw per day)(a)

estimated probabilistically from the FASFC control results and Belgian food consumption surveys(b)

2002-2007 2008-2013 Mean P50 P95 Mean P50 P95

Children 0.87 0.53 3.70 0.72 0.39 3.21 (0.74 - 1.03) (0.45 - 0.62) (2.79 - 4.99) (0.56 - 0.85) (0.34 - 0.47) (2.56 - 4.12) Adolescents 0.64 0.37 2.86 0.48 0.27 2.09 (0.55 - 0.76) (0.31 - 0.44) (2.22 - 3.71) (0.42 - 0.56) (0.23 - 0.32) (1.72 - 2.69) Adults 0.35 0.20 1.57 0.33 0.19 1.47 (0.31 - 0.42) (0.17 - 0.22) (1.26 - 2.01) (0.29 - 0.39) (0.17 - 0.21) (1.19 - 1.86)

(a) 95 % confidence intervals are given between brackets; (b) Devriese et al., 2015; Vereecken et al., 2008; Huybrechts & De Henauw, 2007; AA levels < LOQ = ½ LOQ Given that AA is a possible genotoxic carcinogen to which the ALARA principle (as low as reasonable achievable) applies, the margin of exposure (MOE) approach might give an idea of the risks associated with its presence in food. Based on the pooled 2008-2013 data, the average and P95 intakes correspond to MOEs ranging between 515 and 236 and between 155 and 71, respectively, when based on the endpoint for neoplastic effects (BMDL10 of 0.17 mg/kg bw per day) (NTP, 2012). Such low MOE values indicate that AA remains an issue for public concern.

MAIN CONCLUSION Although some food sectors have undertaken efforts for reducing the AA content of their products and although in some cases there are no single solutions due to the complexity of factors to be considered, results indicate that a renewed attention is needed, requiring more drastic efforts from both the food industry and the policy to tackle the AA issue.

PLANT PROTECTION PRODUCTS IN RAW FRUIT & VEGETABLES ON THE BELGIAN MARKET

Each year, the FASFC monitors around 2000 samples of fruits and vegetables, cereals and other products of plant origin (e.g. tea) for the presence of 400 to 500 residues of plant protection products (PPP).

GENERAL TRENDS Figure 2 presents the control results of PPP residues in raw vegetables and fruit, cereals and other products of vegetable origin that were reported to the EC and the EFSA between 2000 and 2013. In this period, the number of analyzed samples increased more than one third and also the number of analyzed residues more than doubled.

49

Interpretation of such general trends should however be done with care. For instance, the approach for programming the controls as well as the format of reporting changed. Additionally, analytical methods improved with for instance a higher detection sensitivity and the further development of multi-residue methods, which may explain to a great extent the increase in the detection frequency observed (and the decrease of the percentage of samples in which no residues were detected). Moreover, several maximum residue limits (MRL) were adjusted between 2000 and 2013, with as a main change the harmonization of MRL values in Europe in September 2008 (cf. Regulation (EC) No. 396/2005) and the Belgian anticipation of this measure. Figure 3 gives a more detailed representation of the percentage of compliant samples of fresh fruit and vegetables between 2008 and 2013. Independently of their origin, the samples show a high level of conformity and the percentage of compliant samples shows overall an increasing trend.

Figure 2. Output (% and # samples) of the FASFC surveillance program for PPP, as reported to the EC and the EFSA between 2000 and 2013 for raw fruits and vegetables, cereals and other products of plant origin (source: FASFC 2004-2013)

Figure 3. Percentage of samples of raw fruits and vegetables that were checked for the presence of residues of PPP and found compliant between 2008 and 2013 (the total number of samples analyzed is indicated by origin and by year above the columns)

0

500

1000

1500

2000

2500

0

10

20

30

40

50

60

70

80

# sa

mpl

es

% s

ampl

es

number of samples analysed number of pesticides sought LOQ< & ≤ MRL (% samples) < LOQ (% samples) > MRL (% samples)

80%

85%

90%

95%

100%

BE EU outside EU

origin unknown

889

388

382

98

538

327

535

13

763

409

600

15

840

351

572

21

895

524

443

19

873

499

415

27

% c

onfo

rmity

2008 2009 2010 2011 2012 2013

50

TRENDS AT THE RESIDUE LEVEL: QUANTITY SOLD – DETECTION FREQUENCY - EXPOSURE

From the 400 to 500 residues analyzed within the surveillance program, 34 were selected based on their detection frequency (i.e. the residue was generally detected in at least 2 % of the samples), their representativeness (i.e. a sufficient number of samples was analyzed) and their toxicity (i.e. low toxicological reference value) (SciCom, 2015). The estimated exposure of the average (adult) consumer to the majority of these selected residues was up to 100 times lower than the toxicological reference value, namely the “acceptable daily intake” or ADI. In the SciCom advice 18-2015 (SciCom, 2015), possible trends during the period 2008-2013 were discussed based on a combined overview of the estimated exposure to these selected residues, their detection frequency in raw fruit and vegetables on the Belgian market and the quantity of the active ingredient sold in Belgium. Similar to the general tendencies, trends observed for these “individual” PPP substances should be interpreted with caution. Besides the relatively short time period considered, also the sampling basket, containing a wide range of both domestic and imported products and not being the same each year, might influence the observations. Additionally, it is remarked that there is no direct correlation between the quantity sold of an active substance, the detection frequency of and the risk related to the exposure to a given residue. For instance, there seems to be a gradual increase of the quantity sold of chlorpyrifos (Figure 4 (a)), but a similar trend is not observed for the detection frequency or for the exposure to this insecticide. Fairly recently the toxicological reference values for chlorpyrifos were revised to 10 à 20 times lower values. A revision of toxicological reference values implies a reevaluation of MRL values, which in turn could have an effect on the number of authorized applications and as such on the volume sold. The highest average exposure was observed for dimethoate, namely up to 10 % of the ADI. However, due to the relatively low detection frequency in combination with the proposed residue definition for the chronic risk assessment of dimethoate (the total level is namely expressed as “the sum of the dimethoate level and 3 times the level of the metabolite omethoate”), the risk related to the exposure to dimethoate should be nuanced, as is fully explained in the SciCom advice (SciCom, 2015). From Figure 4 (b), a decrease of the quantity sold and the detection frequency is observed in 2010, which could be partly due to the fact that at that time the use of a number of dimethoate containing PPP was restricted. The exposure to dimethoate shows a small dip in 2011.

51

(a) Chlorpyriphos (a)

(b) Dimethoate (b)

(c) Dithiocarbamates (c) (d) Bifenthrin (d)

Figure 4. Amount of the PPP’s active ingredient sold in Belgium (“sold V”; relative representation without units), detection frequency of the residue in raw fruit and vegetables on the Belgian market (“det freq”, expressed as %), and chronic average exposure (“% ADI av”; expressed as % of the ADI) of the Belgian population to pesticide residues, residues through the consumption of fresh fruit and vegetables (deterministic approach, middle bound scenario with results < LOQ = ½ LOQ; the lower limit of the error bars correspond to the “best case” scenario, i.e. results < LOQ were set equal to 0, while the upper limit shows the worst case scenario, i.e. results < LOQ were set equal to the LOQ; ADI values = (a) 0.01 (S1) and 0.001 mg/kg bw day (S2); (b) 0.001 mg/kg bw/day; (c) 0.05 (S1) and 0.006 mg/kg bw/day (S2) of maneb/mancozeb and ziram resp., adjusted for CS2; (d) 0.015 mg/gk bw/day) Figure 4 (c) shows the results for the dithiocarbamates. Although they have a different toxicity, they are considered as one group since the analytical method used cannot distinguish between the different dithiocarbamates. This hampers the risk evaluation as well as the interpretation of results. While the quantity of dithiocarbamates sold in Belgium mainly concerns mancozeb, the actual risk of the exposure to the dithiocarbamates is situated between scenario S1 (exposure expressed in terms of maneb/mancozeb) and S2 (in terms of ziram) with S2 being a large overestimate. Notice that the apparently higher exposure value observed in 2009 is most probably related to the higher LOQ reported that year in combination with the use of the middle bound scenario (i.e. results < LOQ are replaced by LOQ/2) for estimating the exposure.

52

Figure 4 (d) illustrates the effect of a restriction on the use of a certain active substance in PPP formulations. Both the quantity sold and the detection frequency of bifenthrin show a sudden drop around 2011-2012 related to the fact that the use of bifenthrin containing PPP is since the second half of 2011 no longer permitted in Belgium and most other EU countries (decision of 2009). The impact on the exposure is minimal, which can be explained by an already low detection frequency and exposure before 2011. (In 2012, the use of bifenthrin is as yet approved at the EU level, but there are as yet no new admissions in Belgium.)

MAIN CONCLUSION Not only the amount of samples controlled but also the number of residues analyzed, increased significantly during the last decade. Irrespectively of some trends observed for individual PPP residues, overall the level of conformity is high and shows an increasing trend, both for EU as for non-EU products. Moreover, based on the current scientific knowledge, it can be assumed that the exposure of the average Belgian consumer to PPP residues through the consumption of raw fruits and vegetables is very low and does probably not present a health risk at long-term.

POSSIBILITIES & CONSTRAINTS

National surveillance programs offer a large source of data. Exploration of these data over time helps to identify where additional efforts are needed and whether certain inventions or incentives taken have a positive outcome. As is illustrated by the two case studies given above, the analysis of trends, or at least the observation of trends is hampered by assumptions made (e.g. food classification) and by several factors related to the methodology used and to the input data (see Table 3). Table 3. Main constraints related to trend watching of chemical contaminants based on surveillance data Source Limitation Input data When data cover an insufficiently large time period, it is difficult to

distinguish random fluctuations from real trends. Inconsistent sampling of the same (type of) and/or collection of an

insufficient number of samples per food group, as well as little detailed background information, such as information on product formulation, on a possible pre-treatment, or on the origin of collected samples, hampers delimiting the sample pool and the interpretation of observed trends.

The analytical methods used and the sensitivity of the methods might have evolved during the monitoring period which can influence the interpretation/results of the trend observation.

53

Methodology The conclusions that can be drawn from the trend observation depend on: - the level of comparison: When based on quantities sold, detection frequency or compliance percentages (i.e. “prevalence”), a part of the information will be lost than when the trend watching is based on contaminant levels. When based on dietary exposure, an additional factor, namely consumption and as such possible changing dietary patters, is accounted for. - the considered time frame, the time series that are compared: comparing years, seasons, etc. to account for e.g. seasonal variation of AA precursors, PPP applications, etc.

CONCLUSIONS

Surveillance programs are not specifically developed to monitor trends. Nevertheless, by making assumptions, the data from the national control program can be used for such purpose. A large time frame for the surveillance or monitoring of a sufficiently large basket with a stable number of similar samples, additionally strengthens the applicability of surveillance data for the monitoring of trends. In order to explore possible underlying relationships (e.g. causes), patterns and trends and/or to make inference (e.g. on the effectiveness of a policy measure) or predictions, relevant background information of the sample pool needs to be available as well.

ACKNOWLEDGEMENTS

The authors would like to thank the SciCom of the FASFC, and more particularly the members of the ad hoc working groups of the SciCom advices 18-2014 (SciCom, 2014) and 18-2015 (SciCom, 2015).

REFERENCES Devriese, S., De Backer, G., De Henauw, S., Huybrechts, I., Kornitzer, K., Leveque, A., Moreau, M., Van Oyen, H., 2005. The Belgian food consumption survey: aims, design and methods. Archives of Public Health 63, 1-16. EFSA, 2015. Scientific opinion on acrylamide in food. EFSA Journal 13(6), 4104. Available online: http://www.efsa.europa.eu/en/efsajournal/pub/4104.htm. FDE – Food Drink Europe, 2014. AA Toolbox 2013 (version 13 of 10/01/14), Brussels, Belgium 58 pp. Available online: http://www.fooddrinkeurope.eu/S=0/publication/fooddrinkeurope-updates-industry-wide-acrylamide-toolbox. Huybrechts, I., De Henauw, S., 2007. Energy and nutrient intakes by pre-school children in Flanders-Belgium. British Journal of Nutrition 98, 600-610.

54

NTP – National Toxicology Program, 2012. NTP Technical report on the toxicology and carcinogenesis studies of acrylamide (CAS No. 79-06-1) in F344/N rats and B6C3F1 mice (feed and drinking water studies). NTP TR 575. NIH Publication No. 12-5917. National Institutes of Health. Public Health Service. U.S. Department of Health and Human Services. July 2012. Research Triangle Park: US, 236 pp. Available online: http://ntp.niehs.nih.gov/ntp/htdocs/lt_rpts/tr575_508.pdf. SciCom, 2014. Advice 18-2014 on the acrylamide intake of the Belgian population – revision (dossier SciCom 2013/27). Available online: http://www.favv-afsca.fgov.be/scientificcommittee/advices/_documents/Advice18-2014.pdf (English summary); http://www.favv-afsca.fgov.be/wetenschappelijkcomite/adviezen/_documents/ADVIES18-2014_NL_DOSSIER2013-27.pdf (Dutch version); http://www.favv-afsca.fgov.be/comitescientifique/avis/_documents/AVIS18-2014_FR_DOSSIER2013-27.pdf (French version). SciCom, 2015. Advice 18-2015 on the exposure of the Belgian population to residues of plant protection products between 2008 and 2013 through the consumption of fruit and vegetables (dossier SciCom 2011/02: self-tasking initiative). Vereecken, C. A., Covents, M., Sichert-Hellert, W., Alvira, J. M., Le Donne, C., De Henauw, S., De Vriendt, T., Phillipp, M. K., Béghin, L., Manios, Y., Hallström, L., Poortvliet, E., Matthys, C., Plada, M., Nagy, E., Moreno, L. A., HELENA Study Group, 2008. Development and evaluation of a self-administered computerized 24-hour dietary recall method for adolescents in Europe. International Journal of Obesity 32 (5), S26-S36.

55

APPLICATION OF TREND WATCHING IN SELF-CHECKING SYSTEMS

SURVEY AMONG FOOD BUSINESS OPERATORS IN BELGIUM C. Verraes *, X. Van Huffel1 1 FASFC, Belgium * Speaker: Ir. Claire Verraes Expert Staff direction for risk assessment, DG Control Policy, FASFC, Belgium E-mail: claire.verraes@favv.be

BIOGRAPHY

Claire Verraes graduated as master at the faculty of Bioscience Engineering at the University of Ghent, specialized in Food Science and Nutrition. Since 2011 she works at the Federal Agency for the Safety of the Food Chain (FASFC, Belgium) in the Directorate General Control Policy in the Staff direction for risk assesment. Her main domain of expertise is food microbiology and microbiological risk assessment.

56

INTRODUCTION

The goal of this symposium is to provide a state of art of trend observation and trend analysis applied to control data from the food chain. By applying these trend watching techniques, it should be possible to get more valuable information from available data. Next to the presentations on the application of trend observation and trend analysis by the authorities, also food business operators have a share in this symposium. Therefore, a questionnaire was send to food business operators in Belgium in order to get a general view of the extent to which trend watching is used in self-checking systems.

METHODOLOGY

The Staff direction for risk assessment prepared with the help of FEVIA (Federation of the Food Industry) a questionnaire regarding the application of trend observation and trend analysis on the results of the self-checking system of food business operators. In this questionnaire, definitions for trend observation and trend analysis were provided. Trend observation was defined as the observation of data in order to visually observe possible trends in time. An example of this is the presentation of data in a line graph followed by the visual evaluation of the graph in order to determine whether a trend is present or not. Trend analysis was defined as performing a mathematical analysis on a time series of data in order to confirm the relevance of trends statistically. An example thereof is the performance of a linear regression on data with a certain significance level (e.g. a confidence interval of 95 % with a p-value of 0,05) followed by the conclusion whether a trend is statistically significant or not. The survey consisted of ten questions. Questions were asked about which sector the company belonged to, whether operators had a validated self-checking system and whether they participated in a sectoral sampling plan. Furthermore, it was asked if operators applied trend observation and/or trend analysis on the results of their self-checking system and why. Next to this, it was asked how, to which data and over which time period the trend observation and/or trend analysis was applied, as well as how frequently these data were collected. Finally, operators were asked if they were of the opinion that trend observation and/or trend analysis could offer an added value for evaluating the results of their self-checking system. The questionnaire was prepared in SurveyMonkey in Dutch and in French and was launched on 6 June 2015 via the sector organizations that are member of the Advisory Committee of the FASFC. The sector organizations were asked to distribute this questionnaire among their members in order to reach the maximum number of operators that are active in the food chain in Belgium. The deadline for completing the questionnaire was set at 15 September 2015 which was also communicated while spreading the questionnaire. Two reminders were

57

send between 6 June and 15 September. At 15 September 2015, the survey was closed and the results were gathered and analyzed.

RESULTS

Food business operators participated anonymously at the survey. 280 respondents turned in the questionnaire; 214 respondents from Flanders and 66 respondents from Wallonia. Most (77 %) of the respondents were active in the transformation sector, whereas the others belonged to the sectors of distribution (13 %), agricultural supplies (5 %), primary animal production (4 %) and primary plant production (1 %). Respectively 78 % and 30 % of the respondents have a validated self-checking system and participate in a sectoral sampling plan. Trend observation is applied by 64 % of the respondents and trend analysis is applied by 32 % of the respondents. Reasons for applying trend observation and/or trend analysis vary from gaining more insight into the evolution of the company (31 %), optimizing the self-checking system (25 %) and the early captation of alert signals (25 %). Other reasons stated by the operators are the optimization of certain parameters or the process itself, avoiding problems and waste, pest control as well as verification of the HACCP plan. From the results of the survey it seems that the data used for trend observation and/or trend analysis concern production efficiency (22 %), hygiene parameters (21 %), product quality parameters (19 %), microbiological hazards (16 %) and chemical hazards (9 %). Other data used by operators are linked to complaints (by consumers), pest control, use of antibiotics and temperatures. In the context of the self-checking system, data are collected daily (29 %), weakly (16 %), monthly (22 %) or annually (13 %). Trend observation and/or trend analysis are performed on these data on a short time period (one year) (60 %), on a medium-long period (up to three years) (13 %) and on a long period (five years or longer) (3 %). Trend observation is performed through a visual evaluation of the raw data (10 %), through a visual evaluation of processed data such as tables, averages, numbers, etc. (26 %) or a visual evaluation of a graphical representation of the data such as histograms, line charts, boxplots, Shewhart charts, etc. (20 %). Trend analysis is carried out with the aid of Excel (28 %) or other statistical programs (10 %) such as ERP, Lims, Minitab, etc. In the last question of the survey, operators were asked if they are of the opinion if the application of trend observation and/or trend analysis could offer an added value for the evaluation of the results of their self-checking system. This question was positively answered by 87 % of the respondents. Reasons were visualization

58

of trends which gives an overview of results of the self-checking system as well as insight into evolutions that can be followed on a long term. Also, trends allow operators to evaluate the achievement of goals or improvements in the company. Next to this, standards or specifications can be adjusted. On a short term, deviations can be detected and therefore quick action and adjustments can be realized. Trend observation and/or trend analyses can also allow an operator to identify trends that are otherwise overlooked. Furthermore, in a pro-active way one can search for root-cause relationships in order to anticipate on deviations or points of attention by taking action or implementing corrective measures. Trend observation and trend analysis can as well be used as a justification of the actions taken. Reasons from the other 13 % of the respondents stating that there is no added value for the evaluation of the results of their self-checking system were mostly the small scale of the company and the lack of data, parameters or case studies to observe or analyze.

DISCUSSION AND CONCLUSION

As the questionnaire was distributed via the sector organizations, it is not known how many food business operators received the invitation for participating to the survey. However, it is assumed that a relatively large response rate was obtained. When comparing the distribution of the sectors (agricultural supplies, primary animal production, primary plant production, transformation, distribution) that participated at the survey with the distribution of the operators that are registered at the FASFC, an overrepresentation of the transformation sector and an underrepresentation of the distribution sector was observed. This may introduce bias while analyzing the results of the questionnaire. It was observed that about three quarters of all participants apply trend observation and/or trend analysis on the results of their self-checking system. Reasons for this vary but are in most cases optimization of the self-checking system was reported. The nature of the used data also varied with the microbiological parameters being the most often analyzed. The data are collected mostly on a daily basis and are observed or analyzed mostly over a short time period. Trend watching is most often used with the aid of Excel. The majority of the participants of the questionnaire apply trend watching and are of the opinion that it can offer an added value to the evaluation of the results of their self-checking system. The questionnaire showed that trend watching is perceived as a useful technique by food business operators to evaluate the results of their self-checking system and to make adjustments in their business model when necessary.

59

ACKNOWLEDGMENTS

We would like to thank all the food business operators that participated in this survey as well as the sector organizations for distributing the questionnaire and especially FEVIA for their feedback during the preparation of the questionnaire.

60

61

APPLICATION OF TREND WATCHING IN SELF-CHECKING SYSTEMS: CASE STUDIES

K. De Reu *, Lieve Herman1 1 ILVO, Belgium Speaker: Dr. ir. Koen De Reu Senior researcher & team leader Unit Technology and Food, ILVO, Belgium E-mail: koen.dereu@ilvo.vlaanderen.be

BIOGRAPHY

Koen De Reu graduated at the Faculty of Bioscience Engineering of the University of Ghent (Belgium) and obtained in 2006 a PhD in Applied Biological Science on ‘Bacteriological contamination and infection of shell eggs in the production chain’. Currently he is working as Senior Researcher and Food Microbiologist at the ‘Institute for Agricultural and Fisheries Research’ or “ILVO”, in Melle, Belgium. His research group works in the broad field of food microbiology with special attention for poultry microbiology, shiga-toxin producing E. coli or STEC, cleaning and disinfection, biocide and antibiotic resistance and biofilms. He is author of more than 50 peer review international scientific papers and has at this moment the supervision on 4 PhD students.

62

ABSTRACT

At different stages in the agro-food processing chain and in the production process, trend watching of microbiological and chemical analytical results offers an important tool to optimize the internal quality management system. Trend watching of analytical results can be applied through trend observation and through trend analysis. In trend observation, data series are observed to discover visually possible evolutions. Trend analysis implies the mathematical analysis of data series to detect statistically significant changes over time (based on EFSA, 2010). Different practical examples of trend watching focussed on microbiological results obtained in self-checking systems will be discussed during the conference. Examples of 1) primary production, 2) feed industry, 3) food ingredient industry, 4) food production, 5) catering and finally the role of 6) service labs will be presented. The examples also include the different stages during food processing (delivery of ingredients, in-line sampling during processing, end product testing, contamination of equipment and environment, …). Examples were collected from a dozen companies in the agro-food processing chain. It is striking that a wide variation of trend watching methods are applied in the chain, ranging from very simple data sheets to the use of highly advanced software packages. Although the latter ones are applied less frequently. The study showed that the trend watching methods applied fit the needs of the different processes in the production chain and of the different production plants. They are used for early warning and source identification of possible contamination problems but also, mostly in larger companies, for comparison of different product types, production lines, plants, incidents, complaints, … In all contacted companies trend watching was used for detecting and identifying problems and for improvement of the quality and the safety of their production. In conclusion, the input from the concerned companies showed that trend watching is a very useful tool extracting a lot of extra information out of analytical results.

EXAMPLES OF TREND WATCHING 1. TREND WATCHING IN THE PRIMARY PRODUCTION

THE LAY-INSIGHT SOFTWARE USED FOR TREND ANALYSIS

Modern farming operations generate huge amounts of information thanks to automated data collection. Unfortunately, the farmer is often not sufficiently able to structure and interpret the data for practical use. The last years, cloud based software tools are becoming available to ensure automatic data collection analysis, interpretation and visualization. The collected data are benchmarked with the expectations based on previously collected data gathered under optimal conditions. As soon as the status indicators deviate from the control data, the farmer is notified by smartphone, PC, laptop or tablet, and will immediately see in which shed, animal group, … the deviated data are obtained which could indicate

63

a possible problem. Notwithstanding the complex statistical processing of the results, easy trend warnings are communicated to the farmer. One of the many examples is the Lay-Insight software of Porphirio (http://www.porphyrio.com/). In this example, the possibilities of trend watching of the body weight of laying hens as indicator to monitor animal health will be discussed (Figure 1). For example, an abnormality in the expected weight gain of laying hens, according to their age, is a reliable indication of possible animal health problems (e.g. infectious bronchitis virus, heat stress) and can be considered as an early warning system.

Figure 1: Real-time trend analysis of body weight at laying hen farms (http://www.porphyrio.com/). Red circles indicate abnormality in the expected weight of laying hens.

THE PRESENCE OF SALMONELLA ENTERITIDIS ENVIRONMENTAL CONTAMINATION IN PERSISTENTLY CONTAMINATED LAYER FARMS

A simple visual graphical presentation of the percentage of Salmonella Enteritidis positive environmental samples per hen house and in the egg collecting area during consecutive laying rounds gives farmers the opportunity to follow up the contamination problems and persistence on their farms. Thanks to the visual trend observation as shown in Figure 2, the farmer gets easily informed about the degree of contamination per location and the effect of control actions in time. Tests with this trend observation system on seven farms, resulted in a complete elimination of Salmonella Enteritidis on most farms and indicated the egg collecting area as the most risky place for persistence and cross-contamination (Dewaele et al., 2012).

64

Figure 2: Detailed prevalence of Salmonella Enteritidis contaminated samples on a laying hen farm during successive laying periods for each henhouse as well as for the egg collecting area with the corresponding month and year of sampling. C&D = cleaning and disinfection; CC = conventional cage; AV = aviary (Dewaele et al., 2012)

TREND OBSERVATION IN THE HATCHERY Monthly trend observations of Salmonella results for down feather samples in a hatchery will be shown and discussed during the conference.

2. TREND WATCHING IN THE FEED INDUSTRY In feed industry beside trend observations on Salmonella and Enterobacteriaceae also a parameter related to moisture content is monitored, since the latter can have an impact on microbiological growth. An example will be shown and discussed during the conference. 3. TREND WATCHING ON FOOD INGREDIENTS AND IN THE FOOD INGREDIENTS

INDUSTRY

PASTEURIZED MILK QUALITY OBSERVATION IN DAIRY INDUSTRY

One of the parameters controlled on the purchased pasteurized milk by a dairy company is the total aerobic bacterial flora. The concerned factory observes graphically the monthly average and standard deviation of the results. Higher values as obtained in April 2014 were discussed with the supplier and corrective actions were appointed (Figure 3).

65

Figure 3: Monthly average with standard deviation of total aerobic flora of pasteurized milk delivered to a dairy company (example delivered by Lavetan)

STERILIZED SEMI-FINAL PRODUCTS IN THE INGREDIENT INDUSTRY In an ingredient industry, the sterilization of a semi-final product is immediately followed by production steps as pumping, concentration, filtration, drying, … After these steps, each batch of the semi-final product is tested for the presence of thermophilic Gram-positive bacteria, Enterobacteriaceae, E. coli and coliforms. Trend observations of weekly results showed higher presence for Enterobacteriaceae and thermophilic Gram-positive bacteria during successive weeks in April 2015 compared to previous periods (Figure 4). These observations have led to a thorough investigation of the possible contamination source occurring after the sterilization process. Insufficient cleaning and disinfection (CIP) during several days of a newly installed and other type of pump after sterilization was pointed out as the reason of the contamination and corrective actions were undertaken.

66

Figure 4: Trend observation of percentage of compliant microbiological results on a sterilized semi-finished ingredient product

END PRODUCTS OF THE FOOD INGREDIENT INDUSTRY Salmonella and coliform analyses are performed on each batch of a certain type of end product in the food ingredient industry and corrective actions are taken based on the observed trends in the obtained results. Besides, batches exceeding the criteria for coliforms are used for technical applications and not for pharma or food applications. Further, the success rate in % for 5 microbiological parameters on end products is calculated per plant and compared on yearly basis. The concerned company also makes quality performance scorecards based on the combination of a score for the production results and the microbiological quality. The results are plotted in chards allowing trend observations between locations worldwide. Examples will be shown and discussed during the conference.

PASTEURIZED WHOLE EGG AS INGREDIENT FOR FOOD INDUSTRY AND CATERING Finally, an example from the egg product industry will be discussed. Each container of 1000 L pasteurized whole egg, used as ingredient in the food industry is controlled on the total aerobic bacterial flora. The target value is 1,000 cfu/g with less than 10 % of the batches on monthly basis having a number between 1,000 and 10,000 cfu/g as also still acceptable. The concerned factory makes basic but useful overviews in Excel mentioning the % of still acceptable results per month and per recipe. For results above 10 %, the source of contamination is searched and corrective actions are taken and registered.

67

4. TREND WATCHING IN THE FOOD INDUSTRY

MICROBIOLOGICAL QUALITY PARAMETERS IN A SAUCE PRODUCING PLANT

Different sauces (14) are randomly produced on different production lines within the same sauce producing plant. In the self-checking system each batch of sauce is analyzed on total aerobic flora, lactic acid bacteria, yeasts and moulds. The trend in amount and % of batches exceeding specifications is then analyzed using different graphical representations. Trends on the non-compliant microbiological parameters, per quarter, per year, per production line and per product type are studied individually and combined. The discussed examples show that in the first and last quarter of 2014 the microbiological specifications exceeded more frequently compared to other periods in 2014 and 2015. The results of 3 of the 11 production lines (AL 2+3, AL3 and AL11) were also worse compared to the other lines (Figure 5). One product gave in 2014 more problems compared to the other sauce types independent of the line on which it was produced, resulting in the conclusion that one of the ingredients of the concerned sauce type might be the cause. Trend observations of the concerned industry also showed problems in 2014 with moulds while in the first half year of 2015 lactic acid bacteria dominated the contaminations.

Figure 5: Percentage of batches exceeding microbiological specifications per production line for 2014 in a sauce producing plant

68

MICROBIOLOGICAL AND CHEMICAL PARAMETERS IN THE MEAT INDUSTRY

In meat industry trend watching on microbiological and chemical parameters is performed and findings are compared. Trend observations on a certain product point out a possible slight improvement (expressed in lower % of non-compliant results) in the microbiological quality while this is not observed for the chemical results (Figure 6). This could be explained by the fact that processing has only an impact on the microbiological parameters and not on the chemical parameters tested.

Figure 6: Trend observation in non-compliant microbiological and chemical results on end products in meat industry

QUALITY PARAMETERS IN THE DAIRY INDUSTRY The first example concerns trend observations showing a higher prevalence of yeast contamination in end products and the associated product losses at the start-up of the production after cleaning and disinfection compared to sampling later in production. These parameters were also compared between the different production lines, resulting in the identification of all production lines with similar problems at start-up and in the identification of the source of the problem: a cross-contamination occurring from the drain for those specific line constructions. The weekly incident trends per production line, production type and complaints per production type and area are also used in the concerned dairy company to optimize production per product type, line, area, … Those incidents and complaints are not restricted to microbiological parameters but are also related to open packaging, product losses, …

69

5. TREND WATCHING IN THE ENVIRONMENT OF THE FOOD INDUSTRY

ENVIRONMENTAL SAMPLING, HYGIENIC SCORES AND LISTERIA DETECTION IN MEAT PROCESSING PLANTS

Typical environmental sampling is based on agar contact plates and sponge or swab samples. As microbiological analyses using agar contact plates sometimes do not result in clear quantitative results and to make calculation and interpretation easy, hygiene scores are mostly used. In Table 1 an example of the principle of hygiene scores is mentioned. Table 1: Example of use of environmental hygiene scores Counts (CFU/area) Individual scores Average scores Evaluation 0 – 2 3 – 9 10 – 29 30 – 90 >90

0 1 2 3 4

0 – 0.5 0.5 – 1.1 1.1 – 1.6 1.6 – 2.1 2.1 – 2.5

>2.5

Very good Good

Pretty good Quite bad

Bad Very bad

In a meat factory individual hygiene scores are calculated to weekly average hygiene scores based on the table above and displayed graphically. Concerned charts are also used to compare individual zones and low and high care zones, indicating clear differences between both zone types. The concerned factory also studies the trend of the environmental hygiene per plant by making a graphical view of the monthly % results exceeding the objective values (Figure 7). Agar contact plate results and Listeria spp. and Listeria monocytogenes detections are taken into account. If deviating results are obtained, more detailed trend observations specifically focused on results obtained with the agar contact plates and Listeria swabs are used to clarify the source of the contamination. Constant differences of the hygienic scores between plants could be explained by the different type of meat product processed or different production type.

70

Figure 7: Percentage of non-compliant environmental hygiene results per plant from a meat factory In another company of meat and chicken preparations no hygiene scores are used but basic trend observations using Excel overviews of agar contact environmental sampling for total aerobic flora showed higher enumerations during one month compared to previous months. Further investigations revealed that there were problems with the dosing pump of the disinfectant used during cleaning and disinfecting.

MOULDS IN CHEESE INDUSTRY Another interesting trend observation case comes from the cheese industry and concerns a combination of trend observation in the environment and on the end product. Some years ago the factory encountered mould problems on their finished product. To combat the problem, a structured environmental air sampling for moulds was established. Trend observations on mould presence in the air on weekly and monthly basis, per area and even trend observations on types of moulds per area were used to identify the sources and to evaluate the corrective actions taken (esp. airflow). In the example the parallel decrease of the mould pressure in the air and on the cheeses will be presented. Trend during 4 years of mould numbers present in the air per floor are shown in Figure 8.

3,44,2

0,7

3,2 3,1 2,8

0,8 0,82,1

1,1

3,0 2,8

12,2 11,9

18,7 18,4

26,2

17,0

10,0 9,4

4,63,1

5,3

12,0

7,8 8,19,7

10,8

14,6

9,9

5,4 5,13,3

2,1

4,1

7,4

13,6 13,6 13,6 13,6 13,6

10,1 10,1 10,1 10,1 10,1 10,1 10,1

0,0

5,0

10,0

15,0

20,0

25,0

30,0

35,0

aug/14 sep/14 okt/14 nov/14 dec/14 jan/15 feb/15 mrt/15 apr/15 mei/15 jun/15 jul/15

% n

on co

nfor

m re

sults

hyg

iene

% non conforming results hygiene

Plant 1

Plant 2

Total

Objective

71

Figure 8: Evolution on mould numbers in the air per floor on yearly basis in cheese industry

TREND OBSERVATIONS FOR LISTERIA SPP. AND LISTERIA MONOCYTOGENES Trend observations on the presence of Listeria spp. and Listeria monocytogenes in the environment are also regularly applied in food industry. This happens on the basis of both simple non graphical reviews in Excel as by using graphics. Additionally some food industries study the trend of the mapping per area and in time of the Listeria results and even perform trend observations on the Listeria typing results of the different isolated strains per area to obtain additional information on the evolution in cross-contamination and persistence.

6. TREND OBSERVATIONS IN THE CATERING

BACTERIAL PARAMETERS OF BUFFET MEALS Buffet meals from the different locations of a catering company are analyzed for total aerobic flora, E. coli, B. cereus and sulfite-reducing anaerobic bacteria. Obtained results are evaluated as good, acceptable and unacceptable. Trend observations are intended to study evolutions over time, to compare the microbiological quality of the meals between locations (Figure 9) and to determine the microbiological parameters giving most problems overall and per location.

72

Figure 9: Evaluation of microbiological results of buffet meals from 10 different locations of a catering company (green=good; yellow=acceptable and red=unacceptable) (example delivered by Lavetan)

7. ROLE OF SERVICE LABS IN TREND WATCHING A lot of analyses in the framework of the self-checking systems are performed by external service labs. The increased capabilities of IT allow service labs to play an important role in facilitating trend watching for their clients. If recurrent samples of the individual clients are analysed by the lab, their results can be exported from the LIMS database of the service lab and analysed, using some simple data treatment and visualisation with graphs. In some cases even real time applications are already available. Doing so, service labs can easily support their clients in their trend watching. This offers the possibility to the service laboratories to give their clients an extra service and help them with their problem solving work.

SUMMARY AND CONCLUSIONS

The study showed that in every stage of the agro-food industry trend watching is used to control microbiological quality and safety of the food product. Mostly trend observation is applied, ranging a wide variety of data handling. Trend analysis using sophisticated data processing packages, allowing simple data input and delivering simple output is much less used. Observed watching fits the needs of each chain in the production process and each individual plant. In all contacted companies trend watching was used for detecting problems and for improvement of the quality and the safety. They are used to compare different types of products, production lines, plants, incidents, complaints, … In conclusion, the input from the concerned companies showed that trend watching is a very useful tool giving a lot of extra information out of analytical results.

ACKNOWLEDGEMENTS

The different companies in the agro-food processing chain are acknowledged for explaining and providing the practical examples of trend watching. Also Kristof

73

Mertens (Prophirio), Jan Robrechts and Raf Peeters (Lavetan), Jef De Smedt (Micro-Smedt) and Saskia Desnyder (LIEMAQS bvba) are acknowledged.

REFERENCES Dewaele, I., Van Meirhaeghe, H., Rasschaert, G., Vanrobaeys, M., De Graef, E., Herman, L., Ducatelle, R., Heyndrickx, M., De Reu, K., 2012. Persistent Salmonella Enteritidis environmental contamination on layer farms in the context of an implemented national control program with obligatory vaccination. Poultry Science 91, 282-291. EFSA, 2010. Technical specifications for monitoring Community trends in zoonotic agents in foodstuffs and animal populations. EFSA Journal 8(3), 1530.

74

75

APPLICATION OF TREND ANALYSIS IN FOOD CHAIN CONTROL: VIEWPOINT OF THE FOOD STANDARDS AGENCY

(FSA) Mr. Terry Donohoe Head of Emerging Risks, Science Evidence and Research Division, FSA, UK E-mail: terry.donohoe@foodstandards.gsi.gov.uk

BIOGRAPHY

Terry Donohoe is Head of Emerging Risks in the Food Standards Agency’s Science, Evidence and Research Division. He joined the Agency in 2002 and was initially responsible for developing, refining and optimizing procedures and protocols for the full range of incidents likely to affect food safety. He also led the development of the Agency’s Food Defence Programme building

and maintaining links with national and international partners, including the US, Canada, Australia and New Zealand. Since 2010 he has been responsible for developing and implementing the Agency’s systems for the identification and characterization of new and re-emerging risks. This work underpinned the development of an “Intelligence Hub”, within the FSA, to facilitate the sharing and analysis of intelligence related to food safety and integrity and was a key response to the Troop and Elliott reviews of the 2013 horsemeat incident. The “Hub” began operation in December 2013. From April 2014 Terry assumed overall responsibility for managing FSA’s work on food incidents, fraud investigations and for ensuring the Agency's resilience to deal with emergencies. He returned to heading the Emerging Risks Programme in January 2015. Terry is also a member of EFSA’s Emerging Risks Exchange Network. He has given presentations on the FSA’s approach to identifying emerging risks and intelligence sharing at national and international meetings. The Food Standards Agency is an independent government department responsible for food safety and hygiene across the UK. It works with businesses to help them produce safe food, and with local authorities to enforce food safety regulations. Everything the FSA does, reflect its vision of “Safer food for the nation”. It aims to ensure that food produced or sold in the UK is safe to eat, consumers have the information they need to make informed choices about where and what they eat and that regulation and enforcement is risk-based and focused on improving public health.

76

SCOPE

If history repeats itself, and the unexpected always happens, how incapable must Man be of learning from experience. (George Bernard Shaw) The identification of new and re-emerging risks is a challenge for both Regulators and Industry alike. A wide range of data sources are available and an increasing range of analytical tools are becoming available to enable these to be accessed and analyzed. Increasingly, a global chain approach is being taken to these analyses, using trended historical data to identify potential weaknesses in supply chains and horizon scanning approaches to attempt to predict how the pattern of vulnerabilities may change in the future. This presentation will outline the Food Standards Agency’s (FSA) approach to the identification of new and re-emerging risks and how trend analysis has underpinned this work.

HISTORY

FSA began developing an emerging risks programme in 2010. Our aim was to complement the work begun by EFSA in 2008 but to tailor our program to the FSA’s needs and resources. An important distinction is that the work of FSA encompasses risk assessment, risk management and risk communication hence the scope of the FSA programme was broader than EFSA’s. The programme was based on the National Intelligence Model, used by law enforcement bodies to maximize the utility of intelligence and resources. The challenge was the vast range and quality of potential data sources. Working with EFSA, we made use of the work of the Data Collection project employing the methodology to characterize and prioritize many hundreds of potential data sources. On reflection, it was decided that FSA’s own incidents data would be used as initial baselines. A number of factors influenced this decision: the FSA employed a broad definition of an incident; the data were of known provenance characterized and validated by the Agency and; the data had a farm to fork scope. Since 2006, the FSA has published an Annual report of incidents showing trends in reported incidents and including case studies highlighting investigations of

77

note. From 2006 the data showed a more stable and consistent pattern and statistical analysis was carried out on each of the Agency’s incident categories.

From the statistical analyses control charts were developed for each of the individual incident categories. Specific distributions were derived, tailored to encompass, for example, seasonal or cyclical effects, a feature of microbiological incidents.

The FSA’s Emerging Risks programme became operational from 2012 but was revised and extended a year later as part of a wider intelligence hub project to take greater account of authenticity issues.

FINDINGS

Generally, monthly incident statistics remained within expected limits but incidents involving microbiological contamination of paan leaves and norovirus outbreaks showed significant variance from expected norms. Further variances

1.344

1.312

1.298

1.208

1.505

1.714

1.604 1.562

1.645

1.000

1.100

1.200

1.300

1.400

1.500

1.600

1.700

1.800

2006 2007 2008 2009 2010 2011 2012 2013 2014

0

10

20

30

40

50

Numb

er of

incide

nts pe

r mon

th

Number of incidents Centre line Lower limit Upper limit

Norovirus outbreak

78

highlighted an annual increase in allergy issues, a trend that has continued since the full implementation the Food Information Regulations in the last year. Effective stakeholder engagement has been critical in the development and operation of the Programme. FSA set up a number of local, national and international discussion groups with membership drawn from Regulators and Industry. Levels of engagement have improved year on year. EFSA’s Emerging Risks Exchange Network has proved to be a particularly useful forum, bringing together membership from across the EU and also including international partners such as WHO, Food and Agriculture Organization of the United Nations (FAO) and the US Food and Drug Administration (FDA).

FURTHER DEVELOPMENTS

It was always intended that use of incidents’ baselines was an initial phase and the FSA therefore developed Global Chain Analysis and Root Cause Analysis methodologies to enable the Agency take a more global and proactive approach to the detection of new and re-emerging issues. A further project, commissioned by the FSA took a broader look at data sources, where they were held, their ownership and issues around sharing of data. This work gained increased interest following the horsemeat incident of 2013 and a number of recommendations were made concerning the development of so-called Safe Havens to provide opportunities to share information and intelligence. Horizon Scanning techniques to identify key Drivers and Indicators of vulnerabilities within global supply chains were part of the global chain approach and offer the potential for FSA be more predictive with respect to possible futures and enable scenario planning exercises.

CURRENT STATUS

Building on the portfolio of existing work the FSA is currently developing an econometric model based on PESTLE (Political, Economic, Social, Technological, Legal and Environmental) drivers with a view to developing a toolkit of indicators of emerging issues. The FSA’s Global Chain approach has underpinned current investigations into contamination of herbs and spices and work is continuing in the area of fresh produce, which is also being aligned with initiatives from EFSA and the EC.

79

Building on existing stakeholder links, the FSA is looking to Industry Stakeholders, in particular, to elicit supply chain data and information on the dynamics of fresh produce supply chains and known trends in relation to food safety and authenticity.

VULNERABILITY ASSESSMENT

The horsemeat incident of 2013 highlighted concerns over the potential for fraud in the increasingly complex global supply chains and a potentially increased vulnerability to cyber-crime. The FSA, in conjunction with Defra and the British Standards Institute therefore developed Publicly Available Specification (PAS) 96 Guide to Protecting and defending food and drink from deliberate attack to help Industry and others use a systematic approach Threat Assessment Critical Control Points (TACCP) to identify potential vulnerabilities. TACCP complements existing Hazards Analysis and Critical Control Points (HACCP) systems. The document was intended to have global application and is now being increasingly employed to demonstrate compliance the British Retail Consortium (BRC) Global Standards Version 7.

THE NATIONAL FOOD CRIME UNIT

The establishment of a National Food Crime Unit, housed within FSA, was a key recommendation following the horsemeat incident of 2013. The Unit became operational from December 2014 and is currently carrying out its first Food Crime Annual Strategic Assessment to identify food crime threats and risks facing the UK.

LOOKING FORWARD

Over the last five years there has been an improving trend towards sharing information and increased levels of trust. However, this continues to be a major challenge and requires the active and continued involvement of stakeholders throughout the food chain. FSA has had some success but much remains to be done. Improvements to testing methodologies such as whole genome sequencing and improved analytical techniques for manipulating big data sets provide us with enhanced capability to identify trends in issues affecting food safety and authenticity.

80

The technologies offer the potential for rapid, cost effective testing regimes but are likely to increase the resource demands on Regulators and others are more outbreaks are identified and reported. However, the results from new testing or data analysis techniques will require careful handling to ensure appropriate conclusions are drawn and appropriate and proportionate actions are taken. Do increased trends mean increased exposure or simply a better elucidation of what is normal in supply chains?

CONCLUSIONS

Trend analysis has and continues to inform FSA’s Emerging Risks Programme. New analytical techniques offer us the opportunity to understand better the dynamics of supply chains, highlighting challenges and opportunities. A trend towards increased collaborative working with stakeholders is welcome and the fostering of better sharing of information should ultimately elicit benefits for consumers. In today’s environment hoarding knowledge ultimately erodes your power. If you know something very important the way to get power is by actually sharing it. (Joseph L. Badaracco, Professor of Business Ethics at Harvard Business School, 1948)

REFERENCES BSi, FSA, Defra, 2014. PAS 96 Guide to protecting and defending food and drink from deliberate attack. Available online: http://shop.bsigroup.com/Browse-by-Sector/Food--Drink/PAS-96-2014/. EFSA, 2011. Data Collection for the identification of emerging risks related to food and feed. EFSA Journal 9(8), EN-185. Available online: http://www.efsa.europa.eu/en/supporting/pub/185e. FSA. Annual Report of Incidents June 2014. Available online: http://www.food.gov.uk/about-us/data-transparency-accounts/busreps/miscbusrep. FSA. Food incidents: advice for businesses. Available online: http://www.food.gov.uk/business-industry/food-incidents.

81

APPLICATION OF TREND WATCHING IN FOOD CHAIN CONTROL: EXPECTATIONS OF THE RISK MANAGER

H. Diricks*, M. Raemaekers1, X. Van Huffel1, Vicky Lefevre1 1 FASFC, Belgium * Speaker: Ir. Herman Diricks Chief Executive Officer FASFC, Belgium E-mail: ceo@favv-afsca.be

BIOGRAPHY

Herman Diricks is the Chief Executive Officer of the FASFC. Trained as a bioscience engineer Herman Diricks has been working in the areas related to Public Health, Food and Feed safety and technology since 1984, first in the Ministry of agriculture, as an independent consultant on food safety and finally in the FASFC as Director-General of the corporate services and as Director-General of Control Policy.

82

INTRODUCTION

The mission of the Belgian Food Safety Agency is to preserve the safety of the food chain and the quality of our food in order to protect the health of humans, animals and plants. To accomplish this mission, the Agency integrates all control services that are competent for the entire food chain (from farm to fork). Control of the food chain is the core activity of the Agency. It is implemented through a control program which consists of two parts: the annual analysis program and the multi-annual inspection program. The control program is translated into a sampling plan. The control program is based on a systematic yearly assessment of the risks by in-house experts from the Directorate-General Control Policy, hereby taking into account updated information on the safety of the food chain originating from multiple sources such as the results of the analysis program of the Agency of the previous year, the RASFF (Rapid Alert System for Food and Feed) reports, the scientific opinions published by national or European scientific bodies, stakeholder information, etc. The Agency aims to apply a preventive and pro-active food safety policy. In this regard, trend watching is one of the instruments that deserves special attention of risk assessors and risk managers. In this article, the expectations of the risk manager in regard to trend watching in food chain control are discussed based on existing applications or applications under development such as:

trend observation of contextual factors, application of trend watching in the annual update of the control

program of the Agency, application of trend watching to support management decisions, trend watching as a tool for indirect surveillance of the safety of the

food chain, trend watching and communication, trend watching and early identification of emerging issues.

SOCIETAL, ECONOMICAL AND BUDGETARY CONTEXTUAL FACTORS (TRENDS) AFFECTING THE CONTROL OF THE FOOD CHAIN

It is very important for a risk manager to take account of trends in contextual factors that may affect the safety of the food chain. Several socio-economical trends demand our special attention. They may occur at macro-economical and global level (climate change, migration, growing

83

complexity of the food chain, increased price fluctuation, international conflicts and terrorism, trade embargos, invasive species and emerging diseases, food fraud, new technologies, etc.) or at national scale (steady decline of the number of operators in the primary sector and of independent operators in the B2C sector, marginal profitability of primary production, increase in number of small supermarkets, increase of internet sale, aging population). Consumer habits are also evolving. Consumers show a growing interest in product authenticity and in short supply circuits, in consumption of insects, ready-to-eat food, outdoor dining, etc. The opinions of national and international scientific advisory bodies, the results of ongoing research programs and the horizon scanning by in-house experts are particular useful to attract the risk manager’s attention to new trends that may affect the food chain and deserve to be considered in the elaboration of the control policy of the Agency. As all public authorities, the Agency has recently been confronted with drastic budgetary restrictions (4 % reduction in staff expenditure, 20 % reduction of government funding in 2015 and additional reductions of 2 % in the upcoming years). Under these conditions, it represents a great challenge for the Agency to pursue its dynamic approach and it obliges the management to strive towards maximal efficiency and efficacy. The strategic and operational objectives for the upcoming period have recently been described in the 2015-2017 business plan. Priority will be given to the maximal execution of the control program to protect the safety of the consumer and to the administrative support of trade and export.

APPLICATION OF TREND WATCHING IN THE ANNUAL UPDATE OF THE CONTROL PROGRAM

The Agency has a robust risk based control program in which more than 500,000 analyses per year are foreseen. It enables the Agency to cover the whole food chain and to have a good idea of the food safety status. Up till now, trend watching is not yet routinely established in the methodology of the yearly update of the control program of the Agency. Some recent initiatives and new developments show that there is a growing interest in applying trend watching in the work method. At one hand, simplified statistical tools are used by experts to compare results (frequencies of conformity/non-conformity for specific matrix-contaminant combinations) of the control program between two periods. When the results are significantly different, the expert may decide to adapt the sampling frequency in the control program. On the other hand, a business object trend watching tool was developed by the Crisis prevention and crisis management department to “automatically” monitor the results of the control program in a more continuous way. The tool is based on

84

identifying evolutions in matrix-contaminant combinations during three consecutive time periods. The tool consists of three modules in regard to the follow-up of:

1. the frequency of unfavourable or non-conform results, 2. the frequency of detections of contaminants, 3. the medians of contaminant concentrations.

If two relevant consecutive increases are detected, the subset of underlying matrix-contaminant combinations is automatically selected. After extraction of the selected data subsets, a regression analysis has to be performed in order to examine the significance of the increase. Separate modules in R (Shiny) have been developed for this purpose. Various data subsets can be uploaded and processed at once. The slope and significance of trends are automatically calculated. Statistical results are shown and can be downloaded for further processing (e.g. making graphs). In principle, any expert of the Agency is able to carry out this trend watching by means of the regression tools. In practice, data subsets need to be pre-processed before they can be uploaded and statistically processed. So far, trend watching is not part of the routine risk assessment realised by the experts of Directorate-General Control Policy, but it is carried out by the Crisis prevention and crisis management department. When a significant increasing trend is detected, it can be the start of the revision of the risk assessment by the experts of the Directorate General Control Policy.

APPLICATION OF TREND WATCHING TO SUPPORT MANAGEMENT DECISIONS

Trend watching may be used to support management decisions. The following examples are given.

EXAMPLE 1: SALMONELLA SURVEILLANCE IN PIGS During many years a Salmonella action plan in pigs was realized. This plan was launched in 2007 and determined that blood samples from pigs taken within the framework of the Aujeszky’s disease surveillance program, had also to be examined for presence of Salmonella antibodies. Pig farms with unfavourable results were designated as being at risk and were further professionally assisted to improve the sanitary situation. In 2012, the Scientific Committee showed in its advice 03-2012 (SciCom, 2012) that the Salmonella Action Plan had had little effect on the Salmonella Typhimurium infection rate of pig farms (see Figure 1). In the light of budgetary restrictions, the risk managers decided in 2015 to

85

abandon the Salmonella Action Plan in pigs and to reflect on alternative Salmonella reduction plans together with the sector and scientific experts.

Figure 1: Evolution of the Salmonella serotypes in pigs in Belgium (Source: CODA-CERVA)

EXAMPLE 2: SURVEILLANCE OF ANTIBIOTIC CONSUMPTION IN ANIMALS

The reduction of irresponsible antimicrobial consumption in veterinary and human medicine has been under the attention of the authorities since many years in the face of the major public health concern in regard to the worrying global trend of increase of antimicrobial resistance. In Belgium a Center of expertise on Antimicrobial Consumption and Resistance in Animals (AMCRA) was founded as a consortium between the authorities, the sectors and the academic community. Ten precise objectives and action points were defined in the AMCRA 2020 vision statement (AMCRA, 2015). It should lead to a rational use of antibiotics in animals and limit the spread of resistance. The 2020 plan is based around the central objective of a 50 % reduction in the total consumption of veterinary antibiotics by 2020. This year, the sixth BelVetSac (Belgian Veterinary Surveillance of Antibacterial Consumption) report (BelVet-SAC, 2014) showed however a disappointing evolution of veterinary antibacterial consumption in 2014 compared to previous years. After a favourable trend of reduced consumption between 2010 and 2013, this report showed an unexpected increase (+1.1 %) in the overall veterinary antibiotic consumption per kg of animal biomass in 2014. This example shows

86

the importance of trend watching in the pursuit of precise management objectives. Appropriate actions are meanwhile being undertaken by the sector and the authorities as a reaction on this recent unfavourable trend observation.

Figure 2. Evolution of the total veterinary antibacterial consumption in reference to 2011 and to the goals put forward in the AMCRA 2020 plan (Source: J. Dewulf)

TREND WATCHING AS A TOOL FOR INDIRECT SURVEILLANCE OF THE SAFETY OF THE FOOD CHAIN

MONITORING CONSUMERS COMPLAINTS Consumers that are dissatisfied about the quality of food or a food business operator can introduce a complaint at the Agency. All complaints are recorded and thoroughly investigated. A vast Oracle database stores all complaints filing them according to different keywords. Complaints can be sorted by sector and by type of the operator, by product category and by product description. After extraction, the data can be introduced in a business object report which shows the trends of various groups of complaints. If the frequency of one or more of these groups exceeds a threshold, this group of complaints is studied in more detail. When a common origin (place, operator, product brand, etc.) for several complaints is detected, further tracing back investigations can be started in the field.

0,0 20,0 40,0 60,0 80,0

100,0 120,0 140,0 160,0

2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 realisation goal retrospective extrapolation

87

MORTALITY SURVEILLANCE IN CATTLE AND PIG HOLDINGS In this project trend watching is used as a preventive instrument. Data from the single carcass rendering company in Belgium are used to detect cattle and pig holdings with high mortality rates. The identification of holdings at risk is done by the Crisis prevention and crisis management department according to two methods. At one hand, the evolution of mortality rate of individual cattle and pig holdings is monitored. Holdings with extremely high mortality rates are identified. On the other hand, monthly relative mortalities are benchmarked per holding type and holdings with the highest relative mortalities (above the 99th percentile) are selected. In holdings at risk the sanitary responsible and the herd health contract veterinarian is asked to participate in a health survey in order to raise awareness about the abnormal situation and to make them evaluate it. On average, about 95 % of the surveys have been answered so far.

TREND WATCHING AND COMMUNICATION

Trend watching can be used to communicate about the evolution of the safety of the food chain. A striking example is the food safety barometer developed by the SciCom of the Agency. The food safety barometer consists of a basket of 30 carefully chosen, measurable food safety indicators which together reflect the food safety situation. These indicators include all links of the food chain, i.e. from suppliers to consumers for both the Belgian production and for imports. The control of products (on the presence of chemical and biological hazards) and of processes (inspections and audits) are both included in the basket. The preventive approach (self-checking, compulsory notification, traceability) and foodborne outbreaks are also covered by the barometer. The majority of these indicators are measured within the context of the control program of the Agency, allowing a simple annual monitoring. Based on the results of the food safety indicators and the weighting of the relative importance of these indicators a food safety barometer may thus be defined. This barometer measures the state of the food safety in Belgium on an annual basis, and this with respect to the previous year. From 2007 on, data were gathered for all 30 food safety indicators and the difference between results of two successive years was calculated in percentage. Taking into account the relative importance of each indicator, the average of these differences in terms of percentage gives the value of the barometer which reflects the food safety situation. Since the beginning of measurements in 2007, the food safety barometer shows a positive trend. This positive trend is mainly due to an increased number of operators having a validated self-checking system and to a reduced number of cases of human salmonellosis. The product control results have only a limited

88

impact on the barometer, although a high degree of conformity is observed each year. Figure 3 shows the evolution of the food safety barometer since 2007.

Figure 3. Evolution of the food safety barometer since 2007.

TREND WATCHING AND EARLY IDENTIFICATION OF EMERGING ISSUES

Every risk manager is dreaming of having a tool at his disposition for the early identification or prediction of emerging issues affecting the safety of the food chain. Many attempts have already been made especially in the animal health sector but up till now no practical functioning system for early warning or early captation of (holistic) signals is available.

CONCLUSIONS

Trend watching is a tool with many possible applications within a business context. It can be used in a simple way (trend observation) or in a more sophisticated way (trend analysis) based on statistical tools. Although the

89

systematic application of trend watching by the Agency is still in its infancy, some examples developed in this presentation have shown its possibilities and usefulness. Risk managers are especially eager to identify the latest trends in the food chain and therefore promote the development of new and practical instruments for early captation of signals in order to identify and anticipate on emerging issues. Risk managers expect that trend watching gets well established as a tool in food safety control and that it is not only applied by authorities but also by sector organizations and food business operators. The FASFC encourages food business operators to apply trend observation and trend analysis on the results of their self-checking systems in order to improve the safety of their products. Trend watching could also be applied by sector organizations on results of sectoral monitoring plans and make this tool better accessible for smaller companies. Risk managers expect also that trend watching will provide valuable information to experts to adapt the control program in function of emerging risks and to adapt the control policy to new socio-economical trends such as the interest of the consumer for the consumption of insects, the reduction of food waste, etc. In the context of the current budgetary constraints, trend watching can be used to further guarantee the effectiveness of the control of the food chain.

REFERENCES AMCRA, 2015. AMCRA 2020. An ambitious yet realistic plan for veterinary antibiotic policy until 2020. Available online: http://www.amcra.be/sites/default/files/files/AMCRA%202020%20finaal_EN%281%29.pdf. BelVet-SAC, 2014. Belgian Veterinary Surveillance of Antibacterial Consumption National consumption report 2014. Available online: http://www.belvetsac.ugent.be/pages/home/BelvetSAC_report_2014%20finaal.pdf. SciCom, 2012. Advice 03-2012 on the Salmonella action plan in pigs (dossier Sci Com 2011/05: self-tasking initiative). Available online: http://www.favv-afsca.fgov.be/scientificcommittee/advices/_documents/Advice03-2012.pdf (English summary); http://www.favv-afsca.fgov.be/wetenschappelijkcomite/adviezen/_documents/ADVIES03-2012_NL_DOSSIER2011-05_000.pdf (Dutch version); http://www.favv-afsca.fgov.be/comitescientifique/avis/_documents/AVIS03-2012_FR_DOSSIER2011-05_000.pdf (French version).

90

91

CONCLUSIONS E. Thiry*, C. Verraes1, X. Van Huffel1 1 FASFC, Belgium Speaker: Prof. dr. Etienne Thiry Prof. University de Liège Chair of the Scientific Committee of the FASFC, Belgium E-mail: etienne.thiry@ulg.ac.be This symposium provides a challenging definition of trend observation and trend analysis, both components of trend monitoring. These approaches can be used for the food chain by researchers, authorities as well as by food business operators as a valuable tool to assess past evolutions in food safety or product quality and predict future evolutions if conditions remain the same. New trends in society may have a possible impact on the safety of the food chain and should be carefully observed as well. Cornerstone for performing trend watching however is to dispose of reliable data which are regularly and systematically collected and assessed using the appropriate tools. Many interesting examples of trend monitoring were discussed during this symposium as well on a European scale, on a national scale or at the level of the business operator. Trend watching can be performed over short time periods or over a longer term provided enough quantitative data are available. However, the relevance of observed trends or statistically significant trends can be critically evaluated by experts because many peripheral factors can influence the identification of a trend and could lead to misinterpretation such as changes in the sampling plan, modifications of the measurement methodology or of standards during the observational period. A trend can be considered as being ‘real’ when the trend is the result of biological, epidemiological, climatological, socio-economical and/or management related factors which are biologically meaningful. Once trends are identified, response actions can be taken. Trend watching allows to better identify problems and take the necessary corrective measures. On a longer term, one can look for trends indicating root-cause relationships and in this way anticipate on problems. Food business operators can adapt their production process in time and improve the quality as well as the safety of their products. Food safety control agencies can adapt sampling plans in order to evaluate the effectiveness of certain management measures taken or to put other accents if trends go in the desired direction.

92

Trend watching can also be used to contribute to the identification of emerging risks in the frame of early warning systems. However, its effectiveness needs to be further studied as well as the development of appropriate instruments. After this symposium we hope that trend watching will be better understood and will be even more widely used in the future to contribute to food safety at all levels of the food chain.

93