An understanding of Campylobacter spp., and specifically Campylobacter jejuni in broiler chickens is important from both public health and international trade perspectives. As a result, there is an urgent need to evaluate this pathogen-commodity combination by quantitative risk assessment methodology.
The objective of this work was to undertake the first steps of a risk assessment that will ultimately provide estimates for 1) the risk of illness from pathogenic Campylobacter spp. in broiler chickens consequential to a range of levels in raw poultry for the general population and various susceptible population groups and 2) the change in exposure and illness likely to occur for different interventions in primary production, processing, handling and preparation of poultry. This report describes the first three steps of the risk assessment. It will be concluded at a future date with the completion of the risk characterization. The final report will also include discussion and evaluation of potential intervention strategies and uncertainty analysis. Methods of validating such a model will be explored.
Campylobacter is the leading cause of zoonotic enteric infections in most developed countries. Human cases are usually caused by C. jejuni or to a lesser extent by Campylobacter coli. Most human Campylobacter infections are classified as sporadic cases or as part of small family related outbreaks. Identified outbreaks are not common. Information on the burden of human Campylobacter infections for the developing countries is very limited. However, it is likely that the rate of campylobacteriosis is especially high among children below 2 years of age causing substantial morbidity and eventually mortality.
The principal reservoir of pathogenic Campylobacter spp. is the alimentary tract of wild and domesticated mammals and birds. Campylobacter is commonly found in poultry, cattle, pigs, sheep, wild animals and birds, and in dogs and kittens. C. jejuni is predominantly associated with poultry and C. coli is predominantly found in pigs.
It is well recognized that poultry products can be contaminated with Campylobacter. However, Campylobacter is also found in beef, pork, other meat products, raw milk and milk products, fish and fish products, fresh vegetables and modified atmosphere packaged foods such as unsmoked bacon and salad vegetables.
Campylobacter may be transmitted from the reservoirs to humans by direct contact with contaminated animals or animal carcasses or indirectly through the ingestion of contaminated food or water. Case-control studies conducted worldwide have repeatedly identified handling raw poultry and eating poultry products as important risk factors for sporadic campylobacteriosis. Other food related risk factors that have repeatedly been identified include consumption of other meat types, undercooked or barbecued meat, raw seafood, unpasteurized milk or dairy products and drinking untreated surface water. Eating meat cooked outside the home (at restaurants) and not washing the kitchen cutting board with soap (indicating cross-contamination) have also been identified as risk factors. Other risk factors include exposures when travelling abroad, contacts with pets and farm animals, and recreational activities in nature. Person-to-person transmission is apparently infrequent, because infected humans constitute a minor reservoir for C. jejuni, and asymptomatic excretion of Campylobacter is uncommon.
This section focuses on evaluating the nature of adverse health effects associated with foodborne Campylobacter spp. and how to quantitatively assess the relationship between the magnitude of the foodborne exposure and the likelihood of adverse health effects occurring. In this document, human infection refers to the status of pathogen persistence and multiplication within the gastrointestinal tract with or without symptoms. Illness refers specifically to the state where overt symptoms occur as a result of the infection.
The objective and scope of the Campylobacter hazard characterization is to provide:
a review of the characteristics of the host, organism and food matrix effects;
a summary and review of available data and information on adverse health effects;
a dose-response model based on human feeding study data.
Information was compiled from published literature and from unpublished data submitted to FAO and WHO by public health agencies and other interested parties. The first section of the document provides a description of the public health outcomes following infection and including sequelae, pathogen characteristics influencing its ability to elicit infection and illness, host characteristics that influence the acquisition of infection, and food-related factors that may affect the survival of C. jejuni in the human gastrointestinal tract.
The second section of the hazard characterization document presents the data and methods available to derive a dose-response relationship for C. jejuni. The ultimate goal was to derive a dose-response model that mathematically describes the relationship between the numbers of organisms that might be present in a food and consumed (the dose), and the human health outcome (the response). In order to achieve this, human feeding trial data for two strains of C. jejuni were used. The data were used to derive estimates for the probability of infection as well as the probability of illness.
The current document has attempted to synthesize and summarize the information that was available either in the literature, or through the call for data to describe the factors that influence the likelihood of an individual becoming infected, ill, and developing sequelae. The quantification of the importance of most of these factors requires a substantial amount of additional research. Nevertheless, the information and analysis conducted does allow some advances to be made in estimating the risk from Campylobacter and putting into context the apparent importance of other contributory factors.
The probability of any pathogen initiating an infection is influenced to various degrees by three factors. These include the pathogen and host characteristics and the matrix or conditions of ingestion. The influence of specific components within these three factors was qualitatively described based on current thinking. Unfortunately, there is currently insufficient information on which to base a detailed analysis, that would allow fine distinctions to be drawn between, for example, the probability of illness upon ingestion of any one strain versus any other, or ingestion in milk versus water, or for an individual who is taking medication versus a very young child.
The adverse effects that can occur following infection with C. jejuni were also summarized. They included acute gastroenteritis and non-gastrointestinal sequelae such as reactive arthritis, Guillain-Barré syndrome (GBS), and Miller-Fisher syndrome. Reactive arthritis has been estimated to occur in approximately 1% of patients with campylobacteriosis. Guillain-Barré syndrome is a serious paralytic condition, which has been estimated to occur once in every 1000 cases. Finally, Miller-Fisher syndrome, which is considered to be a variant of Guillain-Barré syndrome is also reported to occur, however, there are no estimates on the frequency of the occurrence of this condition following campylobacteriosis.
The hazard characterization also describes dose-response models that can be used to mathematically describe and estimate the probability of infection following the ingestion of a dose of C. jejuni. The dose-response equations used were based upon the single-hit hypothesis, fit to human feeding trial data conducted using healthy volunteers and two strains of C. jejuni (Figure 5.1). It was proposed that pooling the infectivity data for the two strains from the feeding trial study may be appropriate, and this offers a new interpretation of the available information.
FIGURE 5.1 Beta-Poisson dose-response relationship for the probability of infection for C. jejuni based on human feeding trial data and two strains (A3249 and 81-176) and (model parameters, a = 0.21, b = 59.95)
LCL - Lower confidence limit
UCL - Upper confidence limit
The probability of illness is conditional upon the probability of infection. Using the data from the human feeding trial study, there does not appear to be a clear trend for the behaviour of this conditional probability. When the data for both strains are pooled, the conditional probability tends to exhibit a dose-independent relationship (Figure 5.2). It is important to recognize that although the conditional probability is dose-independent, the ultimate probability of illness increases with ingested dose.
In conclusion, the probability of infection upon ingestion of a dose of C. jejuni can be estimated with the caveat that the data are from a feeding study involving healthy volunteers, and using a milk matrix and a limited number of Campylobacter strains. The impact of population immunity, sub-population susceptibilities or other factors cannot be quantified from the data. The probability of illness following infection can also be estimated using a dose-independent probability. Some researchers have proposed a decreasing conditional probability based on consideration of only one of the two Campylobacter strains. Again, the impact of other factors, such as susceptibility, on the probability of illness cannot be quantified due to a lack of adequate epidemiological data and resolution to this level. Finally, the progression of the illness to more serious outcomes and the development of some sequelae can be crudely estimated from the approximate proportions reported in the literature.
FIGURE 5.2 Hypothetical probability of illness curves, influenced by three alternative conditional probabilities. The conditional probability assumption is shown in the inset curves.
(A)Conditional probability independent of dose;
(B) Conditional probability decreasing with dose;
(C) Conditional probability increasing with dose.
Gaps in the data
Data on strain variability regarding virulence/pathogenicity.
Studies on the mechanisms of infectivity, virulence/pathogenicity in the human host.
Quantitative information about infection and illness rates at low doses and with other strains of C. jejuni ranging from 100 to 109 organisms.
Complete epidemiological data from outbreak studies including enumeration of Campylobacter in suspected food items or in drinking water, numbers of people exposed, attack rates, and demographics of those exposed, particularly immunocompromised population groups and, children under the age of five.
Data describing the impact of acquired immunity resulting from recent exposure to Campylobacter.
Studies to elucidate the true number of human infections caused by Campylobacter, including GBS etc., and to determine the burden of disease attributable to different sources of Campylobacter.
To utilize an alternative approach to dose-response for management purposes in a country that lacks enumeration data in their system, several pieces of country specific information are required. Data are needed on the number of Campylobacter illnesses associated with chicken in that country in a time interval (e.g. per year). Coupled with this, the prevalence of contaminated chickens at a point in the farm-to-table chain is also required. The further down-stream the prevalence estimation (i.e. closer to the consumer) the more utility in the approach for risk mitigation strategies.
To evaluate the risk posed to the human population by the presence of Campylobacter spp. in broiler chickens, an exposure assessment model was developed. The aim of this assessment was to estimate the likelihood and magnitude of exposures to Campylobacter as a result of consumption of a chicken meal. Exposure assessments that consider Campylobacter spp. in broiler chickens have been developed independently by Canada, Denmark and the United Kingdom. However, each of these models specifically focused upon the localized situation.
The objective of the exposure assessment was to develop a model that details the prevalence and numbers of Campylobacter throughout the production chain from farm-to-table. However, the model presented to the expert consultation only considered whole fresh broiler chickens, prepared for consumption by oven roasting in the home.
The approach taken was described in detail in the background document on Campylobacter spp. in broiler chickens that was prepared for and presented to the expert consultation. In its preparation, it was felt that the most efficient means of facilitating discussion at the expert consultation would be to limit the detailed descriptions within the exposure assessment section to five key stages; rearing, transport, processing, cross-contamination and cooking. These stages were chosen according to their level of development and their predicted importance in contributing to the overall confidence in the final risk assessment results. Other stages were described only superficially and will be included for discussion at future expert consultations.
The exposure assessment was approached in a farm-to-table manner (Figure 5.3). The model framework is modular in nature and each stage of the supply chain is described by a distinct mathematical model. This provides a flexible tool for risk managers, which may be used to estimate the risk to public health and investigate the impacts of potential interventions. Only meals prepared in the home and cooked using oven-roasting are currently considered but this could be expanded in the future.
The exposure assessment begins by considering the Rearing and Transport module of the supply chain. The aim of this module is to estimate the probability that a broiler chicken will be colonized and the probability a bird will be contaminated on the exterior at the point of slaughter. The levels of colonization and external contamination associated with any given bird are also considered. From this point, the slaughter of the birds and subsequent stages of processing are investigated, corresponding to the second module of the overall framework, Slaughter and Processing. The output from this module is the probability that a chicken carcass is contaminated with Campylobacter at the end of processing, and the associated level of contamination on such a product. The exposure assessment concludes with the final module, Preparation and Consumption. This module addresses the preparation of a product in the home environment and subsequent cooking. The result is an estimate of the probability an individual is exposed to at least one Campylobacter, along with a measure of the numbers of Campylobacter cells ingested. Each of these models is stochastic and incorporates the inherent uncertainty and variability associated with the model through the use of Monte-Carlo simulation. The integration of the modules outlined above then feeds into hazard characterization. Each of these modules is described in detail below.
FIGURE 5.3 Model framework for the risk assessment of Campylobacter spp. in broiler chickens.
Rearing and transport
To estimate the colonization and contamination status of a broiler chicken at the point of slaughter two parameters are used. These are a measure of the national prevalence of flocks that contain at least one colonized bird and the within-flock prevalence of such a flock. The probability that a bird is colonized at slaughter is then the product of these two factors. Here, a colonized bird is defined as a positive bird, and a flock that contains at least one positive bird is defined as a positive flock.
Data are often available to estimate the flock prevalence; however, data may not be available to estimate the within-flock prevalence of positive flocks. Therefore, a dynamic model describing the colonization process of a flock following exposure has been developed. In brief, the model assumes that the transmission of Campylobacter within a flock is initiated by an exposure event, which results in the colonization of a single bird. Once this first bird is successfully colonized, transmission ensues amongst the birds with which the first colonized bird makes contact on a daily basis, that is, the birds social cluster. This continues until a threshold is reached where the level of contamination in the feed, and water supply is sufficient to result in the colonization of an exposed bird. From here onwards colonized birds appear randomly throughout the entire flock. This process continues until either all the birds become colonized or depopulation occurs and the birds are removed for slaughter.
The within-flock transmission dynamics may depend upon the source of Campylobacter. The above description applies to the colonization of the first bird as a result of some point source. However, if one considers exposure as a result of contaminated feed or water, under such circumstances a large proportion of the flock will be exposed and colonization of birds is likely to occur randomly throughout the flock. Further, if vertical transmission occurs it may be that initially there are several birds that are colonized and hence initiate the process of flock colonization. Therefore, the model is developed such that the dynamics of within-flock transmission are dependent upon the source of the organism. Colonization levels are estimated by use of experimental data.
A consequence of the colonization of a flock is the external contamination of the birds in that flock. This occurs either by self-contamination for a bird which is colonized and hence likely to become contaminated as a result of the excretion of Campylobacter in the birds faeces or, for a bird not necessarily colonized, by contact with faeces containing Campylobacter. This contamination is then magnified during the transportation of the flock to the slaughter facility. This is as a result of the dispersal of contaminated faeces throughout the vehicle.
To predict the extent to which birds become contaminated on their exteriors, it is assumed that the colonization process of a flock can be described spatially on a lattice structure. Each bird has an associated location on the lattice. The colonization process is modelled as described above, such that at depopulation the location of each colonized bird is known on the lattice. In this way, the location of the colonized birds in the transport vehicles is also known and hence the consequence of the shedding of Campylobacter in the faeces and the impact this has upon the contamination of the exterior of a given bird can be predicted. Estimation of the levels of contamination are based upon experimental data, however, only one data set is currently available and hence does not allow for variability in transport times that may occur. As such, the model does not currently incorporate the length of time of transport, although this may be an important factor in predicting the impact of transport on contamination levels. If such information should become available, it can be incorporated in future model development.
The contamination of the exterior of birds is not unique to positive flocks. Experimental studies suggest that birds from negative flocks can become contaminated on their exteriors at some point prior to slaughter. However, the frequency and extent of such contamination is currently unknown due to lack of data. An assumption is made that a bird from a negative flock can become contaminated with 1% of the contamination on a random bird from a positive flock. However, in the absence of data the impact this has upon the probability and levels of contamination on the exterior of a random bird at slaughter, as predicted by the model, and the validity of these predictions is unknown.
Slaughter and processing
The stages considered by the model in this module are stun and kill, scald, de-feathering, evisceration and wash and chill. The focus of the processing of broiler chickens is the level of external contamination on the bird/carcass and the manner in which this changes through processing. These changes occur due to reduction as a direct result of the processing stage itself, cross-contamination from other birds and self-contamination from the caecal contents of the bird.
To estimate the impact of the processing stages upon the level of contamination on a bird/carcass and the prevalence of contaminated carcasses a simulation model has been developed. This model utilizes the outputs from the rearing and transport module and generates a profile for a random bird. More specifically, a bird is assigned a colonization status, a contamination status and associated levels of colonization and contamination. The stochastic effects of each of the processing stages upon the level of contamination of a bird and hence the prevalence of contaminated birds is then predicted.
To model the stochastic effect of each stage quantitative measurements are required for the number of Campylobacter on a carcass before and after each of the processing stages considered. These data represent the variability between birds that occurs at each stage. Available data are sparse and only small data sets are currently available. The true extent of the inter-bird variability is uncertain. Further, published data sets often report only mean values of a number of samples. Numerous combinations of effects could have occurred to produce these mean values so further uncertainty is present. To quantify the level of uncertainty, second-order modelling is adopted through non-parametric methods that make no assumptions regarding the true form of the variability. The level of uncertainty in any model results as a consequence of the absence of data can be visualized. A further complication is the data reported do not sample the same bird before and after each stage, but random birds from the flock. Small sample sizes were used. For example, in one study only four birds were sampled. The extent of the effect a stage has on an individual bird level is therefore difficult to assess. The model development process has highlighted the importance of the availability of appropriate data, reported in a manner appropriate for use in quantitative risk assessment modelling.
The data incorporated into the model is taken from a variety of sources and hence involves a number of processing plants and production methods. In this way variability that exists between processing plants is incorporated in the model framework.
As a direct result of the nature of the data available, the model currently does not explicitly consider the evisceration process. However, a conceptual model has been developed for this stage of processing but was not presented to the expert consultation. Should more data become available this can easily be incorporated in to the current framework. The model considers different methods of chilling, these are air chilling, and water chilling with and without the addition of chlorine. In this way the model is adaptable to different processing systems.
Post-processing transportation and storage
These steps have been considered in the model development, but were not included in the document presented to the expert consultation. Future exposure assessment descriptions will include this module.
Preparation and consumption
Cross-contamination in the home
Two models for cross-contamination were developed. One describes exposure via fluid "drip" exuded from the uncooked broiler carcass and ingested via some pathway, for example on fingers, or via contact with other foods. This model is a mechanistic approach related to the water that a chicken gains through processing and is subsequently released. Loosely attached cells will enter into this fluid and may be inadvertently ingested.
The second is a "contact transfer" model that quantifies the number of Campylobacter cells transferred from the raw chicken to preparation surfaces (cutting board, utensils, etc.) or hands, and subsequently from the preparation surface to a prepared meal. The organisms may also be ingested directly by for example licking on fingers.
Three modelling approaches were examined to describe the fate of Campylobacter during thermal heating by oven-roasting of whole carcasses. Once an appropriate modelling approach has been determined, the model can be extended to other cooking styles.
The first approach, the "internal temperature approach", was based on the calculation of thermal death through a sequence of time-steps during roasting of chicken. The temperature, which determines the thermal death in each step through standard D-value calculations, was based on observations of the time-temperature profile in the centre of the drumstick portion of a roasted chicken. The final temperature achieved during the cooking process was based on observed internal temperatures achieved in domestic kitchens.
The second approach, the "protected areas approach", was based on the designation of areas in the carcass where the least heat treatment is achieved, presumably due to increased thermal insulation from the heat source. In this approach, it was assumed that any Campylobacter outside of these designated areas are killed. The approach then required assumptions regarding the proportion of pathogens that are found in these areas, and the maximum temperature achieved in this area. The thermal inactivation was estimated by D-value calculation for the final temperature in this area and the assumed time spent at this temperature.
The third approach, the "heat transfer approach", was designed to predict the time-temperature profile at various depths below the surface of chicken based on a simplified thermodynamic model of heat transfer through chicken meat. This allows for prediction of thermal inactivation as a function of depth and time. The final consumer exposure was highly dependent upon the final temperature achieved at each depth and the assumed proportion of cells that were located at each depth in the carcass.
At the current stage of development, the models are being evaluated with respect to the validity of required assumptions, the degree of conservatism which is implicit in the approach and the relative value of complex versus simple models given the amount of uncertainty in the location of cells in or on carcasses with respect to thermal insulation.
The transportation of flocks to slaughter may be a crucial stage in predicting microbial levels at slaughter. However, this is an area often not considered by researchers and more data are needed.
It was recognized that much of the data that is reported in the literature is difficult to use in risk assessment due to the methods employed and the style of the reporting. As an example, data on changes in pathogen load on carcasses is frequently reported as the mean of log concentrations for only a few carcasses. The research community should consider the statistical power of such studies and be more critical of the ability of such study designs to give definitive results for the purpose of assessing risk and intervention efficacy.
At present two models have been developed, both of which consider the overall probability of exposure to Campylobacter during a food preparation event. Comparison of the models shows that, despite the different approaches taken, they seem to be mathematically equivalent with an appropriate choice of assumptions. One of the two models may be preferable at a later date depending on the data that will become available in future.
It is difficult to model cross-contamination based on the information currently available. Further improvement of the model and model validation may be extremely difficult given the complexity of cross-contamination, the many possible pathways by which it can occur, and the variability in the behaviour of individuals in the kitchen.
Based on thermal inactivation calculations, it is difficult to reconcile the assumed importance of undercooking as a cause of human exposure with the assumption that contamination of broiler carcasses with Campylobacter is on the external or internal surface of the carcass (or very close to the surface). Resolution of this inconsistency requires the allocation of some amount of contamination to various places within the carcass where Campylobacter are significantly insulated from heat. While it is possible to demonstrate and to hypothesize that Campylobacter will, on occasion, be found in such a place, it is very difficult to quantify the frequency and extent of this particular mode of contamination. It may also call into question the importance of surface contamination with respect to consumer exposure due to undercooking. Clearly surface contamination will remain a key component for exposure via cross-contamination.
Gaps in the data
Data on the routes of Campylobacter infection of broiler chicken flocks.
Survey data on the prevalence of Campylobacter in slaughtered flocks and within flocks.
Studies on the dynamics of within flock transmission.
Data on the probability of contamination of a bird during transport.
Data on production systems in different countries and regions.
Prevalence and enumeration data for poultry before and after various processing steps such as scalding, defeathering, evisceration, washing and chilling.
External contamination levels of broilers at slaughter from both positive and negative flocks and the relationship between this and time of transport.
Prevalence and enumeration data comparing various methods of chilling - air chilling, water chilling, water chilling with chlorine, etc..
Data describing the actual cross-contamination between positive and negative flocks and within positive flocks during the different slaughter processes.
Prevalence and enumeration data comparing different scalding temperatures and different packaging methods.
Data on the relationship between the concentration on neck skin samples and the concentration on the whole chicken in order to calculate a conversion factor.
Data on the microbial implications of carcass deboning.
Survey data and direct observational data on consumer practices in preparation and handling of chicken that especially detail to which extend different contamination pathways may contribute to exposure.
Research data detailing quantities of Campylobacter that are transferred to and from surfaces and hands and from preparation surfaces to the final meal during preparation of chicken. Data on the concentration of Campylobacter in the fluid attached to the chicken.
Survey data and direct observational data on preparation and handling practices of chicken in restaurants and other retail establishments
Survey data on consumption patterns.
Data on the number of Campylobacter cells in the fluid attached to the chicken.
Distribution of pathogens in places other than the surface of chicken carcasses before and after cooking.
Final temperatures achieved in oven roasting in different areas of the carcass.
Importance of the cooling phase after removal from the oven.
Relationship between observational definitions of undercooking (e.g. self-assessed, or "pink chicken") and the actual heat treatment experienced.
The quality of data on the human incidence of Campylobacter infection varies across countries, reflecting differences in surveillance systems and microbiological techniques used. Most of the available information was summarized in the document presented to the expert consultation but it was recommended that newly available sentinel surveillance data on Campylobacter infections in the Netherlands and the United Kingdom also be included. The expert consultation noted that there appears to be differences in the pathogenicity of Campylobacter species and that for example the apparent low pathogenicity of C.. lari should be mentioned. The expert consultation agreed that Campylobacter is an important source of human foodborne illness.
The expert consultation noted the importance of distinguishing between the reduction in prevalence and reduction in levels of contamination of Campylobacter on retail products. In discussing the infectivity of so called "viable-but-non-culturable" Campylobacter the expert consultation noted that there are conflicting data in the literature concerning this issue and felt that in some published work, the techniques used to assess "culturability" were not sufficiently sensitive.
There are several sources of infection with Campylobacter spp. but the main one is believed to be poultry. However, this may differ from country to country or region to region. Therefore, it was suggested that additional data is obtained if possible or otherwise it should be generated. The relative importance of risk factors is not fully addressed and the expert consultation recommended the inclusion of epidemiological data in this section. For example, the results of recent and ongoing intervention studies such as those in Belgium and Iceland should be included as they become available. Risk factors for campylobacteriosis in developing countries may be different from those in developed countries and this should also be considered. Furthermore, the epidemiology of campylobacteriosis in developing countries and the role of acquired immunity needs to be addressed. The effect of seasonality on the prevalence of Campylobacter was noted and should be considered when discussing reservoirs of this pathogen.
There is a need for consistency in the use of terminology (e.g. birds are "colonized" but humans are "infected" with Campylobacter). In addition, it should be stated that infection precedes clinical symptoms and does not necessarily result in disease.
Although acquired immunity is likely to play a role in the risk of human infection, the consultation agreed that 15-25 year olds may be more susceptible and or more frequently exposed. Age definitions need to be made clearer for the purposes of identifying population groups of increased susceptibility or exposure.
The consultation recognized that antimicrobial resistance might compromise treatment in patients with diarrhoea and bacteremia. It recommended that existing risk assessments on antimicrobial resistant Campylobacter be considered in the future development of this work.
Dose-response data were only available from one feeding study on young, apparently healthy males, using only two strains of C. jejuni, both of which were clinical isolates. The limitation of developing a dose-response curve from such limited data was recognized. Although limited, the data showed a positive correlation between the exposure dose and the probability of infection. This correlation was not evident for disease, possibly due to the small sample size. Given the current data limitations, the expert consultation concurred with the decision to pool data from the two strains. However, more data were needed to establish a sound dose-response correlation in relation to illness. Some of those identified by the expert consultation included:
data on dose-response for other sections of the population including the more susceptible human population groups;
data for other strains of Campylobacter and the differences between strains in relation to dose-response; and,
data on the effect of food matrices on dose-response.
The fact that milk was used as the food carrier for Campylobacter in the abovementioned feeding trial raised concern about the protective role played by the food matrix. However, in the absence of other data, it was assumed that the protection afforded by milk was at least as good as that of chicken.
The consultation agreed that the dose-response model developed was a reasonable one but should be applied with caution. It may overestimate the frequency of illness in developing countries due to acquired immunity, and may likewise underestimate the frequency of illness due to differences in susceptibility of a population group. If other strains of Campylobacter act with more or less efficacy than those used in the feeding trial the dose-response model may need to be modified accordingly.
The strong need to develop a model for assessing exposure to Campylobacter was recognized. It was the opinion of the expert consultation that a model such as the one currently being developed may help to inform risk management decisions and to assess the risk to human health.
Colonization on the farm
An overview of data made available from case-control studies on the factors involved in the introduction of Campylobacter to poultry flocks on the farm should be presented. In addition, data are required on the major cause(s) of seasonality of Campylobacter colonization in broilers. This is currently unknown although some of the assumed risk factors are reduced in winter (e.g. numbers of wild birds are reduced, airflow is out of the houses, temperature is reduced, snow on earth etc.). There is a need to include information from some countries that are succeeding in reducing the prevalence of positive flocks (e.g., the United Kingdom, Sweden and Denmark). Farmers are becoming more successful in either excluding Campylobacter from the flock or delaying colonization. This means not all birds will be Campylobacter positive pre-slaughter. Information should be provided on the possibilities to limit the introduction of Campylobacter into a flock (biosecurity) and to mitigate its spread if introduced (immunization, feed additives, etc.).
Transportation of broiler chickens may further spread contamination within the flock due to spreading of faecal material over the birds. Where the distance between the farm and the processing plant is short, contamination is largely restricted to the outside of the birds. It was the opinion of the expert consultation that this contamination is easily reduced during scalding and further processing. As such, it may be neglected as a source of Campylobacter on the final carcass. However, longer periods of transportation could influence cross-contamination, gut colonization and excretion levels. The model on transport needs to take this into consideration. In addition, the model should consider the potential effects of feed withdrawal on the amount of faeces produced and the levels of Campylobacter in faeces.
Other preharvest intervention strategies
The consultation recognized that the model should include intervention strategies. These may include:
organic acids in feed;
possible competitive exclusion or vaccination.
Slaughter and processing
During processing of chickens, the bacterial flora present on the outer surface of the broiler chickens will fluctuate. During scalding a proportion of the Campylobacter present will be washed off and a proportion killed by heat. In the succeeding processes (defeathering and evisceration), further contamination may occur. The predominant source of contamination is faecal spread during these processes. Once carcasses are contaminated, reduction of bacterial flora is limited. Therefore, in most Hazard Analysis Critical Control Point (HACCP) plans, faecal contamination during evisceration is a critical control point. Although the slaughtering techniques and equipment are steadily improving, faecal spread cannot be completely avoided.
It was recommended that the statement in the background document beginning Welfare of the live birds and carcass quality are top priorities... be replaced to reflect the opinion of the expert consultation that economics are the main controlling factor in poultry processing. The model requires data on changes on Campylobacter numbers on poultry carcasses at critical points during processing. The consultation recommended that the section on poultry processing should include discussion on the survival of Campylobacter and consider that the strains may differ in their survival abilities.
The current model assumes that the effects of stunning and killing are negligible. This statement may need to be qualified in the light of information from the United States that indicates that water in the electric stunning bath can be Campylobacter positive. Birds may inhale this water and this may lead to systemic contamination.
Available data indicate that the processes that have been used during poultry slaughter have not significantly reduced Campylobacter contamination rates on chickens. It is also observed that when a Campylobacter positive flock is processed, contamination levels vary between 102 to 104 per gram of skin. As a consequence of this observation and the absence of processes that significantly reduce contamination, it was advised to consider simplifying the models.
Changes due to defeathering
Water usage and pressure force applied by the plucking fingers are considered important parameters affecting recontamination and cross-contamination during defeathering. For example, the present model does not take account of faecal material that may be introduced onto bird surfaces by the expulsion, during defeathering, of faeces from the gut as a result of physical pressure. However, such data may not be currently available.
Changes due to evisceration
The current model does not include the effects of evisceration. Although it was acknowledged this might have an impact if the viscera rupture, data on the evisceration step are currently only available for birds where the viscera did not rupture.
Effects of washing and other treatments
The model should allow for additional interventions geared at reducing the bacterial load after processing of carcasses including treatment with organic acids, irradiation and hot water baths.
Effects of chilling and freezing
Broiler chickens comprise whole birds or portions that can be either fresh or frozen. There are several methods for cooling carcasses. These include immersion cooling in a spin chiller and spray or air chilling. These processes influence both the water content of the meat and the bacterial load present. Data showed that the effect of air chilling on Campylobacter present was not significant. However, new developments in air chilled technologies demonstrate that a reduction in Campylobacter organisms is achievable. The relative proportions of contaminated chickens cooled under air compared to water chilled systems is country specific and can be handled in the model. The model can also represent effects of intervention measures such as use of chlorinated, ozonized or electrolyzed water. The expert drafting group informed the consultation that a model was also available on the effects of freezing but that this model was not included in the background document.
There is a need for the model to account for the period between the processing plant and the home. The model as presented to the expert consultation did not include a retail component and it was recommended that this be included in the further development of the model.
As there are many differences in the preparation of broiler chicken products, data on this area would be relevant for assessing the final risk to the consumer and in identifying risk factors during preparation.
Differences in preparation between countries comprise, among others, the proportion of chicken prepared at private homes and restaurants and the method of preparation (conventional oven, fan oven, microwave, boiling, frying, barbecuing). The model should be adapted so that it is able to examine risk factors associated with preparation methods and commercial preparation of chicken for consumption in restaurants, hotels and institutions.
Model for cross-contamination
The consultation recommended that the term "drip" fluid should be explicitly defined in the context of the cross-contamination model developed. It was agreed that the concepts used in deriving both the "drip" and "contact transfer" models were plausible and, therefore, should be retained for further elaboration. However, validation of the models would be difficult due to lack of data. It was noted that the "contact transfer" model provides the possibility to model interventions.
The current "drip" model uses volumes of fluid between 0.5 and 1.5ml as the potential volume of liquid released from the chicken. Given that many air-chilled chickens are comparatively dry it was suggested that the lower limit should be reduced to 0.1ml for these types of chickens.
The consultation also proposed changing the name of the "drip" fluid model due to concerns that in countries where air chilling is used, and the drip from chicken is very low, this may lead to misinterpretation of the model.
There was also a need to acknowledge that some chickens are sold as portions and not whole carcasses. The model does not currently address this and this is a limitation. A further limitation is that the model deals with only one pathway, that of a whole chicken coming into the home and being roasted.
Exposure via cooked chicken
In light of epidemiological studies implicating "undercooked" chicken as a risk factor for human campylobacteriosis, it is reasonable to include a component for human exposure via this vehicle in the model. Of the three approaches presented, "internal temperature", "protected areas" and "heat transfer", it was felt that conceptually the latter was the most reasonable. However, because of lack of data for both the "heat transfer" and "protected areas" approaches, the consultation felt that it was important that all the approaches be retained in the model. This may also help to account for the many different ways in which chicken may be cooked. The consultation recommended that the collection of data was needed to further develop and validate the heat transfer model. A microbiological survey to determine the frequency of contamination of cooked chicken was discussed, however, the consultation felt that such a study would be impractical.
Due to differences in consumption patterns between countries and regions, specific surveys are needed in this area.
The expert consultation commended the expert drafting group for the enormous amount of work done both before and during the expert consultation. In continuing their work, the consultation acknowledged and welcomed the drafting groups' plans to carry out uncertainty analysis of the final model and explore ways to validate such a model.
Hazard characterization and dose-response
The experts concluded that the present dose-response model is the best that can be produced with the existing limitations in data and should be put forward for public debate and for future validation. Validation of the model may come about through analytical epidemiological investigations and descriptive epidemiological studies.
Epidemiological studies that can serve to validate the dose-response model should be carried out and the data made available to the drafting group. Such studies may include carefully conducted outbreak investigations, intervention studies and other epidemiological approaches. To be of value such studies should collect information on attack rates among exposed persons, the amount of food ingested, the level of contamination within that food, and sampling strategies.
The strong need to further develop the model for assessing exposure to Campylobacter was recognized and the expert consultation was of the opinion that such a model will help to inform risk management decisions and to assess the risk to human health.
The expert consultation felt the early farm-to-table model was valuable in identifying data gaps and sampling strategies that can stimulate relevant research on Campylobacter in the different stages of the farm-to-table continuum. It is likely that components of the model will be capable of producing useful information within the next year. However, given the extensive nature of some of the data gaps, it is also likely that the development and validation of the full model will require longer than one year.
The expert consultation identified the following areas in the risk assessment as needing particular attention in the next year's work:
The pre-harvest module
- Introduction of Campylobacter into poultry flocks
- Reducing levels of Campylobacter in the gut of poultry
- Effects of cooling and freezing, chlorination, lactic acid treatments and irradiation on levels of Campylobacter
Distribution and processing from slaughterhouse to home (wholesale, retail)
- Effects of storage conditions on colony forming unit (CFU) counts, cross-contamination, etc.
Preparation of meals
- Meals prepared outside the home should be considered in the model
- Data to validate the cross-contamination and cooking elements in the model are needed.
The situation in developing countries
- The risk assessment does not to a very large extent take the situation in developing countries into account, primarily due to data gaps. These data gaps urgently need to be filled.
In identifying Campylobacter in chicken as a priority area in which it required expert risk assessment advice the CCFH outlined two risk management questions as follows:
1) Estimate the risk from pathogenic thermophilic Campylobacter in chicken (broilers) consequential to a range of levels in raw poultry for the general population and for various susceptible population groups (elderly, children, and immuno-compromised patients).
2) Estimate the change in risk likely to occur for each of the interventions under consideration including their efficacy.
Reduce the prevalence of positive flocks
- Destruction of positive breeder and chicken flocks (broiler) flocks
- Vaccination of breeding flocks
- Competitive Exclusion
Reduce the prevalence of positive birds at the end of slaughter
- Use chlorine in water chilling of chicken (broilers)
- Water chilling vs. air chilling for chicken (broilers)
Evaluate the importance of various routes for introduction of pathogenic Campylobacter into flocks including feed, replacement birds, vectors and hygiene
The CCFH furthermore mentioned that a risk profile could be carried out to focus the work before embarking on a risk assessment.
The expert consultation noted that the risk management questions for the risk assessors were not very well tailored to the particular problem. A risk profile could have helped in identifying relevant risk management questions in particular in relation to interventions. Because of the lack of a risk profile, specific interventions were not identified at the outset. Nevertheless the drafting group has taken into consideration a range of different relevant interventions during their model development.
The expert consultation felt that the approach taken by the risk assessors to answer the risk managers' questions, the development of an integrated farm-to-table mathematical model, was a useful one, and potentially the best approach. The consultation furthermore acknowledged that major research gaps need filling to complete and validate the model. These data gaps are not likely to be filled in the short term.
The expert consultation recommended that FAO and WHO promote the harmonization of methods used in both surveillance of human illness and food monitoring.
Due to the very limited data sets available for modelling dose-response and the difficulties of conducting further human feeding trials for Campylobacter, it was recommended that FAO and WHO promote the collection of quantitative data from outbreak investigations in member countries.
In relation to exposure assessment, FAO and WHO should promote the generation and collection of quantitative data throughout the food chain.
 ALINORM 01/13A Report of
the thirty third session of theCodex Committee on Food Hygiene Washington DC,
23 -28 October 2000|