Vous êtes sur la page 1sur 8

Trends in Food Science & Technology 19 (2008) S26eS33

Review

Challenges of
quantitative and in particular the Sanitary and Phytosanitary agreement
(WTOeSPS) created a worldwide need for science-based
standards for food safety. The recent High Level Interna-
microbial risk tional Food Safety Forum (Beijing, China; 26e27 Novem-
ber 2007) reaffirmed the need for transparent food safety

assessment at EU regulations, based on risk analysis. Human dose-response


models and predictive microbiology e two important
components of any QMRA model e were proposed in the
level 1980s (Haas, 1983; Ross & McMeekin, 1994). The incorpo-
ration of variability and uncertainty e crucial in the model-
ing of any biological phenomenon e was made possible by
Arie H. Havelaara,b,*, Eric G. the increasing availability of powerful (desktop) computers
and dedicated simulation software in the 1990s. Based on
Eversa and Maarten J. Nautaa these developments, principles and guidelines for food
a
safety risk analysis were defined by the Codex Alimentarius
Laboratory for Zoonoses and Environmental Commission and specifically by the Codex Committee on
Microbiology, Netherlands Centre for Infectious Food Hygiene (CCFH) (Anonymous, 1995). Risk analysis
Disease Control, National Institute for Public Health consists of three separate, but interrelated stages: risk
and the Environment, P.O. Box 1, 3720 BA Bilthoven, management, risk assessment and risk communication.
the Netherlands CCFH is in the process of further developing and finalizing
b
Division of Veterinary Public Health, Institute for Risk guidelines for these activities. In parallel, technical work on
Assessment Sciences, Faculty of Veterinary Medicine, QMRA was started in several countries worldwide, in partic-
Utrecht University, P.O. Box 80175, 3508 TD Utrecht, ular in the industrialized world. The World Health Organiza-
the Netherlands (Tel.: D31302742826; tion (WHO) and the Food and Agricultural Organization of
fax: D31302744434; e-mail: arie.havelaar@rivm.nl) the United Nations (FAO) have coordinated technical
developments worldwide by establishing the Joint Expert
Meetings on Microbiological Risk Assessment (JEMRA).1
Quantitative microbial risk assessment (QMRA) aims to model
JEMRA has initiated a process to produce technical guide-
the fate of pathogenic micro-organisms along the food chain
lines on the separate stages of the QMRA process (notably
and the associated health risks. More importantly, it allows
hazard characterization, exposure assessment and risk
the a priori estimation of the impact on public health of inter-
characterization) and has produced several international
ventions in the food chain. The European Food Safety Author-
risk assessments, based mainly on a combination of modules
ity is increasingly asked to provide scientific advice to the
developed for national risk assessments.
European Commission based on QMRA. Its application at
Within EFSA, the panel on Biological Hazards (BIO-
the European level poses some unique challenges, both of
HAZ) deals with questions on biological hazards relating
a scientific and of an organizational nature. On the other
to food safety and food-borne diseases, including food-borne
hand, collaboration at the European level will lead to more
zoonoses, food hygiene, food microbiology, transmissible
effective use of limited expertise and resources.
spongiform encephalopathies, and waste management. The
final outcome of the work of the panel is usually reflected
Introduction in an EFSA Opinion.2 Currently, BIOHAZ opinions in the
Quantitative microbial risk assessment (QMRA) is a rela-
tively young branch on the risk analysis tree. Its develop- 1
ment was stimulated by both political and technical Information on JEMRA can be obtained from the websites http://www.
who.int/foodsafety/micro/jemra/en/index.html and http://www.fao.org/ag/
factors. The agreement on the World Trade Organization agn/agns/jemra_index_en.asp.
2
For more information, see http://www.efsa.europa.eu/EFSA/efsa_
* Corresponding author. locale-1178620753812_ScientificOpinionPublicationReport.htm.
0924-2244/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tifs.2008.09.003
A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33 S27

domains of food-borne zoonoses and food hygiene3 share assessment of the costs needed to avert a unit number
many characteristics with risk profiles as defined by (e.g. 1000) DALYs in the population (cost-utility analysis
CCFH4: the risk profile is a description of a food safety prob- [Havelaar, Mangen, et al., 2007]). Such analyses can
lem and its context that presents in a concise form, the even provide a balanced comparison between interven-
current state of knowledge related to a food safety issue, tions to control different pathogens.
describes potential risk management options that have A further advantage of the mathematical modeling in
been identified to date, when any, and the food safety policy QMRA is that it also enhances our insight in the important
context that will influence further possible actions. The processes in the food pathway, and highlighting knowledge
BIOHAZ opinions on food-borne zoonoses and food gaps. Examples of this (again from CARMA) are the obser-
hygiene also provide qualitative assessments of risk. The vations that lowering the concentration of a pathogen may
step towards QMRA is in development and will be the focus be effective and more efficient than complete elimination,
of this review. and that cross-contamination during slaughtering is basi-
cally a non-linear process. These findings for example
QMRA: what can it deliver, advantages and imply that the estimated effects of interventions on human
disadvantages health risks should not be based on prevalence data alone.
Even more than the estimation of the human health risk Mathematical modeling also helps to appreciate the
consequential to a pathogenefood product combination, complex dynamics of seemingly simple processes such as
the most important significance of QMRA is that it allows slaughtering.
an a priori assessment of the effect of intervention In general, state-of-the-art QMRA does not allow for
measures along the whole food chain, or combinations of a precise estimate of human cases for pathogenefood prod-
intervention measures, on public health. Although uct combinations, due to numerous sources of uncertainty
a farm-to-fork food chain model is the often used termi- along the food pathway and in the dose-response modeling.
nology to describe the necessary type of model to achieve Typically these estimates are much higher then expected
this, this neglects the important extension from exposure based on epidemiological estimates such as those from
to human illness by effect (dose-response) modeling. In population-based cohort studies (e.g. Nauta, Evers, Takumi,
addition, when no interventions in the first part of the & Havelaar, 2001), possibly because the dose-response
production chain (e.g. the farm) are considered, the model relationship overestimates the probability of illness because
need not incorporate this part of the chain, and the model acquired immunity is not accounted for (Nauta, Jacobs-
may, for example, become a slaughter house to fork Reitsma, & Havelaar, 2007). To address this problem
model. An example of calculating the effect of interven- requires interaction and integration of epidemiology and
tion measures is the CARMA QMRA on Campylobacter QMRA (Havelaar, Braunig et al., 2007).
in chicken (Havelaar, Mangen et al., 2007), which showed An often heard criticism of QMRA is that it is extremely
that a combination of improvement of farm hygiene, reduc- data hungry and/or time consuming. We do not agree with
ing fecal leakage during processing and decontamination of this criticism. The data need is not a consequence of the
carcasses with lactate is expected to yield a considerably QMRA method, but of complexity of the questions that
lower human incidence of campylobacteriosis. are asked and the degree of certainty that is required; in
The value of QMRA for risk management is enhanced other words of the expectations that risk managers have
when it is combined with socio-economic analyses. This of the QMRA method. A QMRA can in principle be simple
allows for selecting intervention measures that are realis- when the question is simple or when the risk manger needs
tically applicable in society, using criteria such as cost- an order of magnitude estimate for initial risk ranking pur-
effectiveness and public support (Mangen, Havelaar, poses. In such cases, point estimates can be used and simple
Poppe, & de Wit, 2007). These analyses become tools have been proposed (e.g. Ross & Sumner, 2002). But
extremely powerful when they are combined with disease the model cannot be simple when the question is, e.g., to
burden methodologies introduced by WHO for the estimate the effect on the number of human cases of, e.g.,
Global Burden of Disease project (Murray & Lopez, vaccination of production animals on the farm. This is
1996). Disability Adjusted Life Years (DALYs) are a complex problem and it demands a full QMRA model.
used to integrate the effects of morbidity and mortality The large data need is then not a consequence of the
in one common metric, enabling the comparison of method but of the complexity of the problem. QMRA is
different diseases and outcomes. When combined with a good tool to structure the problem and helps identify
risk assessment and economic analysis, they allow the the relevant data needs, but cannot be expected to give
efficiency of intervention methods to be compared by cheap and quick yet correct answers. Several recent ques-
tions presented to EFSA have highlighted the need to
3 balance the expectations of the risk managers with the
For a discussion of risk assessments of transmissible spongiform encephlo-
pathies, see the accompanying paper by Budka, Goossens, and Ru (2008). possibilities of QMRA, given current limitations of data
4
CCFH 37th session, ALINORM 05/28/13: http://www.codexalimenta availability, budget and time. Some examples will be
rius.net/download/report/638/al28_13e.pdf. discussed later.
S28 A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33

With a QMRA model, proper sensitivity analysis can targets, to promote risk-based decision making, to improve
help to address the data problem by determining which pa- risk communication, and to focus data collection efforts.
rameters are important for the model output so that data Other expected advantages were specific for its applica-
collection can be focused on these parameters. Lessons tion at the European level. These included facilitating the
on this important issue in QMRA may be learned from free movement of goods, the optimal use of scarce re-
the related field of epidemiology. Here, much time and ef- sources, strengthening the position of the EC in Codex
fort is put in large observational studies and the establish- and the World Trade Organization and knowledge transfer.
ment and maintenance of complex databases. The food Expected disadvantages of a general nature were the time
consumption surveys to support nutritional epidemiology and resource intensive nature of QMRA (but see the discus-
are a case in point. Efforts are emerging to include the sion earlier in this paper), current models not being suffi-
needs of chemical and microbial risk assessment in such ciently realistic to reflect the complicated technological
studies. There is not a lack of data collection in the EU, and biological processes, and the risk of wasting resources
as the EFSA reports on zoonotic agents illustrate. The chal- if questions were not sufficiently focused (indicating the
lenge is to achieve effective communication between those need for effective interaction between risk managers and
responsible for monitoring and surveillance programs and risk assessors). Specific disadvantages at the European level
those responsible for risk assessment. included the difficulty of adequately reflecting regional dif-
In practice, quantifying the uncertainty of the model out- ferences in food production and consumption, a lack of har-
put of QMRA is complex or even impossible. Partly, this is monization of models, many MS not yet being able to
a theoretical and model implementation problem: estimat- contribute and the possibility that too much emphasis
ing uncertainty becomes complicated when the QMRA would be placed on differences between MS.
model is complex. For example, in food chain models it Based on the results of the consultation, EFSA and the
may be necessary to include different types of variability, EC have discussed potential applications of QMRA to sup-
e.g. within and between batches of the food product. Uncer- port European food safety policy, and several mandates
tainty then needs to be superimposed on this already com- have been received since 2006. Some questions dealt with
plex model, resulting in infeasible or computationally by EFSA will be discussed in this review, together with
extremely demanding models. Uncertainty is also a data some other examples from the authors work that are rele-
problem. The uncertainty of parameter values that are fully vant for decision making at the European level.
unknown cannot be assessed and the same may hold for es-
timates based on expert opinion or estimates based on mi- Examples of QMRA at the European level
crobiological data from laboratory experiments? A similar Salmonella in meat
problem is the quantification of the uncertainty of the This opinion of the BIOHAZ panel (Anonymous, 2008a)
dose-response relationship based on data from only one was based on terms of reference from the EC to EFSA to
or a few strains and tested on a selected group of (healthy) assess:
volunteers. Realizing this, however, is not a drawback of
The relative contribution of different meat categories,
QMRA but an advantage: it clarifies the difficulties associ-
such as carcasses, fresh meat and products thereof,
ated with limited knowledge that may be ignored otherwise
minced meat and meat preparations to cases of food-
and, as a consequence, lead to incorrect management deci-
borne Salmonella infections in humans, taking into ac-
sions. Solving the uncertainty problem is one of the main
count the occurrence of the pathogen in the food chain,
challenges of QMRA in the near future.
risk factors, food production flows and food preparation
and consumption habits. A distinction between meats
EFSA and QMRA derived from different species, such as bovine, porcine,
EFSA is developing a strategy on QMRA taking into poultry (if possible separately broilers and turkeys) and
account expectations from the European Commission other possible species should be considered. In particu-
(EC), Member States (MS) and scientists, the expected lar, the impact of the intended and common use of the
advantages and disadvantages, the available resources at Eu- abovementioned meat categories derived from different
ropean and national levels and the international experience. species should be taken into account as well as the im-
A consultation among all interested parties was held in pact of cross-contamination.
2004e2005.5 It was concluded that there was broad support Basically, this is a risk attribution question. Attribution
in the EC, among MS and scientists for development of can be performed according to two approaches, bottom-
QMRA at the European level by EFSA. Some expected up or top-down. In the bottom-up approach, multiple risk
advantages were general for QMRA. These included provid- assessments are carried out for one or more agents in one
ing a better basis for objective, risk-based criteria and or more food products as the basis for comparing the con-
sumer risks. This can be done with relatively simple QMRA
5
www.efsa.europa.eu/EFSA/DocumentSet/af_qrma_16thmeet_en_3b1. models as these need not evaluate the public health effect of
pdf. measures, which allows for comparing many food products,
A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33 S29

and even transmission via direct contact with animals and fully analyzed in particular to identify the main sources
via water can be included. The technique of comparative of human salmonellosis (source attribution). Most cases
risk assessment gives on the one hand more insight in the of human salmonellosis are considered to be foodborne.
relative effect on public health of interventions in one spe- It was only possible to provide a qualitative ranking of dif-
cific transmission route, and on the other hand it can serve ferent food groups. In the EU, eggs and egg products are
as a first step of a general QMRA for a pathogen, as it can still the most frequently implicated source of human salmo-
pinpoint the most important transmission routes to focus on nellosis. Meat is also an important source of food-borne sal-
with an extensive full scale QMRA model. monellosis, with poultry and pork being implicated more
A well-known example is the US FDA/FSIS Listeria often than beef and lamb. More specific conclusions about
monocytogenes in different ready to eat foods.6 This study the relative importance of specific meat categories as re-
identified deli meats as the most risky products in relation quested by the EC, for example fresh meat, minced meat
to food-borne listeriosis in the USA. It is interesting to and products thereof, cannot be made at the present. It
note that high population risks can be related to high risks was concluded that for this question, the development of
per serving (e.g. pate and meat spreads) but also to high a bottom-up QMRA approach was the most feasible option.
consumption of foods with relatively low risks per serving To support such an approach, representative data on the
(e.g. pasteurized milk). An example that included compar- prevalence and (variability in) concentrations of Salmonella
ison of different food routes with non-food routes for in retail products must be available for all products to be
Campylobacter spp. was published by Evers, Van Der compared in specific scenarios agreed upon by both risk as-
Fels-Klerx, Nauta, Schijven, and Havelaar (2008). The sessors and risk managers. In addition, data on food stor-
results suggest that raw food consumption and direct age, handling, preparation and consumption that reflect
contact with animals are significant transmission routes. the diversity of consumer habits in the EU should be avail-
The bottom-up approach has not yet been applied to human able. The EU baseline studies on the prevalence of zoonotic
salmonellosis. pathogens in food animals are an important step in this di-
In the top-down approach, the incidence and outcomes rection and should be extended to cover other parts of the
of enteric illness are taken from epidemiological surveil- food chain.
lance. In Europe, data on the incidence of infectious disease
are collected by the Basic Surveillance Network, coordi- Salmonella in pigs
nated by the European Centre for Disease Control and Pre- In EC regulation no. 2160/2003, provisions are laid
vention (ECDC). EFSA and ECDC report data on zoonotic down for Salmonella and other food-borne agents that
pathogens annually in the Community Summary Report on pose a public health risk. The Regulation aims at setting
Trends and Sources of Zoonoses, Zoonotic Agents and An- targets for the reduction of the prevalence of zoonotic
timicrobial Resistance and Food-Borne Outbreaks in the agents at different stages of the food production chain in
European Union.7 Data in this report are mainly based on member states. Before setting a target for Salmonella in
passive, laboratory-based surveillance and represent only pigs, a cost-benefit analysis must be performed. According
a small proportion of the total illness occurring in the pop- to the Regulation, EFSA must be consulted prior to setting
ulation. This proportion varies between countries, depend- a reduction target. Therefore, the BIOHAZ Panel was re-
ing on the health care system and other factors. quested to carry out a quantitative microbial risk assess-
Relatively few attempts have been made to calibrate these ment (QMRA) of Salmonella in pigs, as this is
surveillance systems. This lack of information is one of a necessary condition in order to be able to perform
the main problems in obtaining a better insight in the bur- a cost/benefit analysis. Baseline studies on Salmonella in
den of infectious disease in Europe. pigs are executed at present, providing essential data for
Once incidence data are available, the next task is to at- QMRA modeling.
tribute the observed illness to different (food) sources. With In the framework of article 36 of its founding regulation,
regard to attribution of salmonellosis, some studies have EFSA launched a call for proposals entitled Quantitative
been performed by individual MS, using different data microbial risk assessment on Salmonella in slaughter and
sources and methods. These included outbreak investiga- breeder pigs. The awarded consortium (VLA, RIVM,
tions, analytical epidemiology (in particular case-control FOOD-DTU) will produce a QMRA, which will be used
studies), microbial subtyping and elicitation of expert opin- by the BIOHAZ Panel to formulate their opinion. This is
ion (Anonymous, 2008b). No studies have been performed the first time that the article 36 procedure is applied, using
at the European level, and data for such studies are incom- the list of expert organizations drawn up in December 2006
plete. Even where data are available, they are not always on the basis of national nominations. Being the first time,
a number of teething troubles can be identified:

6
http://www.foodsafety.gov/~acrobat/lmr2-5.pdf. - it took a long time before the call for proposal was pub-
7
The report on the year 2006 can be found at http://www.efsa.europa. lished on the EFSA website (from October 2006 to
eu/EFSA/efsa_locale-1178620753812_1178671312912.htm. April 2007). In combination with a lengthy decision
S30 A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33

and contracting procedure, this implied that the start of available data, and potentially useful interventions. This
this 18-month project had to be postponed from Octo- will be achieved by building a network of scientists, placing
ber 2007 to January 2008. Hence, preparations for the a call for data on the EFSA website and organizing work-
contract took almost as much time as the anticipated shops on QMRA and data.
duration of the QMRA (which is typically criticized A novel feature of this QMRA is that the whole of the
for being time consuming); EU will be considered, including differences between
- due to the fact that EFSA only partially funds direct member states. These differences are expected to be large,
costs and, more importantly, only pays a low percent- and related to pig production systems (conventional, large-
age of indirect costs, substantial national co-funding scale), as well as slaughtering and processing methods
had to be obtained, for which no mechanism was yet (large industrial scale, traditional methods) and eating
in place; habits (kinds of products and ways of preparation). As
- QMRA scientific expertise is a very limited resource in much as resources allow, this variability will be built into
Europe, in terms of persons and institutions. This easily the model. The way to do this is still under discussion,
leads to potential conflicts of interests between Article but grouping of countries and selecting indicator food prod-
36 consortia and EFSA panels and working groups, ucts are being considered. In this discussion, data availabil-
hampering the utilization of experts. ity also plays a role: this is expected to show large variation
between EU countries.
The QMRA for Salmonella in pigs will encompass Import and export of pigs and pork also has to be consid-
a probabilistic farm to fork (and beyond) food chain model ered in this QMRA. Intracommunity trade between EU
for the entire EU, where variability will and uncertainty member states may be less important at the EU level than
will not be quantified. The considered outputs will be prev- at the national level. Import from outside the EU does mat-
alence and/or numbers of Salmonella on pork meat and the ter: interventions undertaken on EU farms will not influ-
number of human cases of salmonellosis. The model will ence the safety of imported pork and therewith the risk to
be divided into a number of modules that are allocated to EU citizens. Export outside of the EU matters also: inter-
the different institutions. The modules will be (1) farm, ventions undertaken in the EU will have less effect on
transport and lairage, (2) slaughter and processing, (3) EU citizens when there is more export, but may impact
preparation and consumption, and (4) hazard and risk char- the competitive position of the European pork industry
acterization. The following interventions included in that needs to be considered in the subsequent cost-benefit
EFSAs Terms of Reference will be considered: analysis.
- the expected reduction of Salmonella cases in humans
(or pig meat at retail) by a reduction (e.g. 5- or 10- Campylobacter in broilers (model harmonization)
fold) of Salmonella prevalence in slaughter pigs, based Thermotolerant Campylobacter spp. are a leading cause
on bacteriology in lymph nodes or serology at slaugh- of zoonotic enteric infections in most developed and devel-
ter. It must be noted here that for QMRA it is not these oping nations worldwide. In the European Union (EU),
measurements that are essential, but Salmonella preva- Campylobacter was the most commonly reported gastroin-
lence and numbers in feces and on hides. The QMRA testinal bacterial pathogen in humans in 2004 and 2005. In
will attempt to use a dynamic model of on-farm infec- the past decade, several national quantitative risk assess-
tion (Hill, Snary, Arnold, Alban, & Cook, 2008) to ments for Campylobacter in broiler meat have been devel-
combine the different sources of information on Salmo- oped to support risk managers in controlling this pathogen.
nella infection of slaughter pigs; A first international activity where several available risk as-
- the expected reduction of Salmonella cases in humans sessments were combined was organized by FAO and
(or pig meat at retail) by a reduction of Salmonella WHO (Anonymous, 2003). Recently, within the European
prevalence in piglets from breeder farms; Network of Excellence Med-Vet-Net,8 a group of scientists
- the reduction of the prevalence in slaughter pigs by the has explored the feasibility of the construction of a Euro-
most important potential treatments or control mea- pean consensus model of Campylobacter in broiler meat,
sures at farm level (including sources of infection of based on the existing models. Ideally, such a consensus
fattening pigs); model would provide a tool for individual member states
- the expected reduction of Salmonella cases in humans to perform their own risk assessment. Also, it would allow
(or pig meat) by the most important control measures for a harmonized approach of European risk management
during transport, at lairage or during the slaughter of the problem.
process. Models developed in the United Kingdom (Hartnett,
Kelly, Newell, Wooldridge, & Gettinby, 2001), Denmark
Many national QMRAs have been completed or are un- (Rosenquist, Nielsen, Sommer, Norrung, & Christensen,
derway. The information and expertise obtained in these
QMRAs must be utilized in terms of QMRA models, 8
www.medvetnet.org.
A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33 S31

2003), the Netherlands (Nauta et al., 2007) and Germany operations, several new concepts were deemed necessary.
(Brynestad, Luber, Braute, & Bartelt, 2008) have been These include the Food Safety Objective (FSO, the preva-
compared to identify their differences and similarities lence and/or concentration of the hazard at the moment
(Nauta et al., submitted for publication). It was found of consumption), the Performance Objective (PO, the haz-
that, although the dynamics of the existing models may dif- ard level at an appropriate point in the food chain) and the
fer substantially, some conclusions are shared by all. An Performance Criterion (PC, the effect of a process on
important finding is that all risk assessments conclude a food-borne hazard). These new concepts have been de-
that the most effective intervention measures aim at reduc- fined alongside existing microbiological criteria, which still
ing the Campylobacter concentration, rather than reducing play an important role in the implementation of food safety
the prevalence. This insight is a typical result from systems in practice.
QMRA activities and broadens the potential for (more Whereas traditional microbiological criteria were based
cost-effective) interventions. on the level of hazard control that was feasible in well-op-
The group concluded that it is neither feasible nor de- erated food production systems, the new metrics (and mi-
sirable to merge the different models into one generic crobiological criteria derived from them) should explicitly
risk assessment model. This is largely a consequence of be linked to consumer risks. QMRA is the appropriate
the fact that the most important attribute of a microbial tool to provide this link between hazard control and
risk assessment is that it is fit-for-purpose (Lammerd- consumer risk.
ing, 2007). The purpose of risk assessments may vary be- Different approaches have been proposed to apply
tween individual countries and that of a generic model has QMRA in the setting of food safety criteria, and this field
yet to be defined at a European level. Also, the variety in is still actively developing. One example is presented by
farm systems, industrial processing and most importantly Nauta and Havelaar (2008), based on the model for Cam-
the large variety in practices related to consumer food pylobacter in the broiler meat chain, developed in the
preparation and consumption, complicates a unified CARMA project. The model suggests that reducing the
approach. number of Campylobacter on broiler meat is an efficient
As an alternative to the construction of a consensus way of reducing human campylobacteriosis. There are dif-
model, the scientists in Med-Vet-Net are now develop- ferent ways of achieving this goal, including improved
ing a framework that, based on the existing models, slaughter hygiene, decontamination of carcasses and sched-
may offer guidance to those who wish to develop their uling. The latter approach is based on the observation that
own risk assessment model. Ideally, this framework is in live broilers, the variation in the level of Campylobacter
maintained at the European level after its development between different flocks is higher than the variation within
in the Med-Vet-Net project. Relevant areas for further flocks. Thus, if highly contaminated flocks were detected
research to improve the quality and practicality of and diverted from the fresh meat chain (scheduling), a re-
QMRA are the integration of data- and mechanistic- duction in the level of contamination of meat and of con-
based broiler processing models, and definition of min- sumer risk is to be expected. The risk assessment model
imum requirements for models describing the pathogen was used to estimate the effects of testing at different stages
transfer and survival in the consumer phase of the farm of the food chain (at the farm, at the entrance of the slaugh-
to fork chain. ter house or after cooling) and using different test sensitiv-
European risk management of the Campylobacter prob- ities. Basically, the results present the risk manager with
lem could benefit from an improved interaction between the a tool for balancing the risk reduction for consumers against
establishment of surveillance and monitoring programs and the costs for industry (i.e. the percentage of scheduled
quantitative risk assessment activities. Here, a clear state- flocks).
ment of purpose by risk managers should be the starting
point of QMRA. The data needs are then identified by Challenges and opportunities for QMRA
QMRA, and those, once checked for practicality, will de- at the European level
fine the most effective points in the food chain for monitor- QMRA has the potential to develop into a practical and
ing and surveillance. important decision support tool for Community food safety
policy and EFSA should play a key role in further develop-
Risk-based standards ing and coordinating QMRA at the European level. To
New approaches to food safety management are being reach this goal, several challenges exist.
developed by CCFH, taking into account the Sanitary and
Phytosanitary agreement under the World Trade Organiza- Organizational challenges
tion. The Appropriate Level of Protection (ALOP) plays Organizational challenges include the need to improve
a pivotal role in the SPS agreement. It is the current level the dialogue with risk managers (primarily at the European
of consumer health protection, as offered by existing food Commission) to develop pro-active, focused questions.
safety systems. It is not some future goal (Anonymous, Current questions are usually very broad and may lead to
2002). To implement the ALOP concept in food business complex and resource demanding models. To fully profit
S32 A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33

from the power of QMRA, and in particular its iterative na- without predefined hypotheses about the system dynamics
ture, questions should be formulated well before the policy usually do not yield sufficient information.
needs to develop regulations arise. This will also make it Detailed information on food consumption and prepara-
possible that model development actually takes place tion practices is necessary to adequately model the con-
before data collection, resulting in more focused surveil- sumer phase. Current food consumption surveys do
lance and monitoring activities. It must be borne in mind provide very useful information on meal composition,
that developing a mathematical model is cheap in compar- which can directly be used for QMRA. However, the dy-
ison to collecting data, certainly across the whole of the namic nature of microbial contamination in the kitchen
EU. The potential of QMRA to focus data collection efforts (growth under suboptimal storage conditions, die-off due
is currently not used because data collection usually pre- to cooking and cross-contamination) create a need for
cedes model development instead of the reverse, which more detailed information on consumer handling practices.
would be far more efficient. Partly, this information can be obtained by extending the
There is limited capacity and expertise to develop and questionnaires of food consumption surveys; partly there
apply QMRA in Europe. Even though the capacity is grow- is a need for specific, observational studies.
ing, there is an urgent need to foster international collabo- To improve the timeliness of scientific advice based on
ration. Learning from developments in other countries by QMRA, pro-active model development is necessary. By us-
participation in joint projects at the European level is far ing a modular approach, basic elements of food chain
more preferable than duplicating the development of models should be developed in such a way that they can ef-
models in different member states. It will also lead to har- fectively be combined for a complete and case-specific
monization of QMRA in a more natural way. food chain (Nauta, 2007). It is essential to explore when
QMRA results demonstrate more explicitly than before simple (e.g. deterministic) models suffice and when more
that zero-risk is not attainable. This implies that risk man- complex (stochastic) models are necessary. The debate on
agers are increasingly faced with the need to develop more this issue is currently most active in relation to the new
explicit approaches to defining tolerable risk, in relation to metrics defined by CCFH (FSO, PO, PC, see elsewhere).
economic and social pressures and to translate this into tol- Variability between MS is a challenge for any risk as-
erable hazard levels (prevalence and concentration) at dif- sessment, and also for QMRA. It will not be possible to in-
ferent points in the food chain. clude all existing food production, processing, preparation
and consumption practices across the EU in a risk assess-
Scientific challenges ment. A series of typical products and processes should
A key need in any risk assessment is the availability of be defined. As this is partly a subjective choice, which may
appropriate data. Such data should be representative of the influence risk management decisions, these defaults should
system under study, and adequately reflect the variability of be agreed between risk assessors and risk managers.
microbial contamination in these systems. Structured, har-
monized data collection is a prerequisite and significant The role of EFSA in QMRA at the European level
progress has been made in recent years. The baseline stud- Important tasks for EFSA to stimulate the development of
ies on the prevalence of zoonotic pathogens in primary pro- QMRA at the European level were identified in the
duction for the first time provide comparable data across 2004e2005 consultation. A first task is to create a network
the EU. In addition to the prevalence, data should be avail- of European institutes for QMRA, to foster international
able on concentrations of pathogens at representative steps collaboration and to promote exchange of data and models.
of the food chain. Such data are less variable between coun- In this network, harmonization of QMRA models should be
tries and in time, and therefore study designs which are less a principal target. Whereas any risk assessment should be
ambitious than the baseline studies can be applied. Their fit-for-purpose and thus be built to answer a specific ques-
design should be based on identified needs in risk assess- tion, the efficiency of the process can be greatly improved by
ments and be developed in close collaboration between utilization of agreed modular approaches. The article 36
risk assessors, laboratory scientists and risk managers. As procedure is not very flexible for this purpose and alternative
discussed before, a pro-active approach is necessary as to- solutions may be considered. Seeking cooperation with
morrows risk assessments should define todays data existing networks (e.g. Med-Vet-Net and JEMRA) may be
collection efforts. a first step in this direction. EFSA may also seek to promote
Quantitative information on pathogen behavior (patho- the support of QMRA networks through other EU funds,
gen characteristics in combination with environmental e.g. through DG Research or DG Sanco.
characteristics) in the food chain is a crucial input in risk A second task for EFSA is to develop and maintain
assessment models. Here, even more than in other fields, databases to support QMRA. Such databases are now estab-
the models should guide data collection. Food production lished for the production of the annual zoonoses reports.
systems are highly complex and adequate study designs Whereas these reports provide a useful overview of the cur-
need to take this complexity into account. Descriptive anal- rently available data in the EU, their usefulness is limited as
yses, where samples are taken and results compared they were collected by different methods and according to
A.H. Havelaar et al. / Trends in Food Science & Technology 19 (2008) S26eS33 S33

different sampling schemes. Improving the data quality in Havelaar, A. H., Braunig, J., Christiansen, K., Cornu, M., Hald, T.,
the zoonoses reports is a stated priority for EFSA. Recently, Mangen, M. J., et al. (2007). Towards an integrated approach in
supporting microbiological food safety decisions. Zoonoses Public
the BIOHAZ panel presented opinions on the monitoring of Health, 54(3e4), 103e117.
Toxoplasma, Yersinia and Vero-toxin producing Escheri- Havelaar, A. H., Mangen, M. J., de Koeijer, A. A., Bogaardt, M. J.,
chia coli. The recommendations in these opinions will be Evers, E. G., Jacobs-Reitsma, W. F., et al. (2007). Effectiveness and
the basis for discussions with member states about monitor- efficiency of controlling Campylobacter on broiler chicken meat.
ing of these pathogens and reporting of results. Likewise, Risk Analysis, 27(4), 831e844.
Hill, A. A., Snary, E. L., Arnold, M. E., Alban, L., & Cook, A. J. (2008).
EFSA is evaluating the existing food consumption surveys Dynamics of Salmonella transmission on a British pig grower-fin-
from the needs of risk assessment, which will lead to rec- isher farm: a stochastic model. Epidemiology and Infection, 136(3),
ommendations on possible amendments and additions. 320e333.
Lammerding, A. (2007). Using microbiological risk assessment (MRA)
in food safety. Brussels: ILSI Europe. Summary report of a work-
References shop. Prague, Czech Republic, October 2006.
Mangen, M. J., Havelaar, A. H., Poppe, K. P., & de Wit, G. A. (2007).
Anonymous. (1995). Application of risk analysis to food standards is- Cost-utility analysis to control Campylobacter on chicken meat-
sues. Geneva: World Health Organization. WHO/FNU/FOS/95.3. dealing with data limitations. Risk Analysis, 27(4), 815e830.
Anonymous. (2002). Principles and guidelines for incorporating mi- Murray, C. J. L., & Lopez, A. D. (Eds.). (1996). The global burden of
crobiological risk assessment in the development of food safety disease: A comprehensive assessment of mortality and disability
standards, guidelines and related texts. Rome: FAO and WHO. from diseases, injuries, and risk factors in 1990 and projected to
Anonymous. (2003). Risk assessment of Campylobacter spp. in broiler 2020. Cambridge, MA: Harvard School of Public Health on behalf
chickens and Vibrio spp. in seafood: Report of a joint FAO/WHO of the World Health Organization and the World Bank.
expert consultation. Rome: FAO. FAO Food and Nutrition paper 75. Nauta, M. J. (2007). The modular process risk model (MPRM):
Anonymous. (2008a). Scientific opinion of the panel on biological a structured approach to food chain exposure assessment. In
hazards on a request from the European Commission on a quanti- D. W. Schaffner (Ed.), Microbial risk analysis of foods
tative microbiological risk assessment on Salmonella in meat: (pp. 99e136). Washington, DC: ASM Press.
source attribution for human salmonellosis from meat. European Nauta, M. J., Evers, E. G., Takumi, K., & Havelaar, A. H. (2001). Risk
Food Safety Authority Journal, 625, 1e32. assessment of Shiga toxin producing Escherichia coli O157 in steak
Anonymous. (2008b). Scientific opinion of the panel on biological tartare in the Netherlands. Bilthoven: National Institute for Public
hazards on a request from EFSA on overview of methods for source Health and the Environment. Report nr. 257851003.
attribution for human illness from food borne microbiological Nauta, M. J., & Havelaar, A. H. (2008). Risk-based standards for
hazards. European Food Safety Authority Journal, 764, 1e43. Campylobacter in the broiler meat chain. Food Control, 19,
Budka, H., Goossens, B., & Ru, G. (2008). BSE and TSEs: past, present 372e381.
and future. Trends in Food Science & Technology, doi: 10.1016/ Nauta, M. J., Jacobs-Reitsma, W. F., & Havelaar, A. H. (2007). A risk
j.tifs.2008.09.010. assessment model for Campylobacter in broiler meat. Risk Analy-
Brynestad, S., Luber, P., Braute, L., & Bartelt, E. (2008). Quantitative sis, 27(4), 845e861.
microbiological risk assessment of campylobacteriosis cases in the Nauta, M., Hill, A., Rosenquist, H., Brynestad, S., Fetsch, A., Van Der
German population due to consumption of chicken prepared in Logt, P., Fazil, A., Christensen, B., Katsma, E., Borck, B., & Have-
home. International Journal of Risk Management, 8(3), 194e213. laar A. A comparison of risk assessments on Campylobacter in
Evers, E. G., Van Der Fels-Klerx, H. J., Nauta, M. J., Schijven, J. F., & broiler meat. Submitted for publication.
Havelaar, A. H. (2008). Campylobacter source attribution by Rosenquist, H., Nielsen, N. L., Sommer, H. M., Norrung, B., &
exposure assessment. International Journal of Risk Management, Christensen, B. B. (2003). Quantitative risk assessment of human
8(1/2), 174e190. campylobacteriosis associated with thermophilic Campylobacter
Haas, C. N. (1983). Estimation of risk due to low doses of microor- species in chickens. International Journal of Food Microbiology,
ganisms: a comparison of alternative methodologies. American 83(1), 87e103.
Journal of Epidemiology, 118(4), 573e582. Ross, T., & McMeekin, T. A. (1994). Predictive microbiology. Interna-
Hartnett, E., Kelly, L., Newell, D., Wooldridge, M., & Gettinby, G. tional Journal of Food Microbiology, 23(3e4), 241e264.
(2001). A quantitative risk assessment for the occurrence of cam- Ross, T., & Sumner, J. (2002). A simple, spreadsheet-based, food safety
pylobacter in chickens at the point of slaughter. Epidemiology and risk assessment tool. International Journal of Food Microbiology,
Infection, 27(2), 195e206. 77(1e2), 39e53.

Vous aimerez peut-être aussi