The technical documents on Salmonella hazard identification, hazard characterization and exposure assessment presented to the consultation were discussed in detail by working groups. The full documents are available on request from FAO or WHO and can also be found at the following Internet addresses: http://www.fao.org/WAICENT/FAOINFO/ECONOMIC/ESN/pagerisk/riskpage.htm and http://www.who.int/fsf/mbriskassess/index.htm.
The executive summaries of these documents were updated during the consultation to take into account some of the questions and comments on the papers resulting from these discussions and are presented below. These are followed by a summary of the discussions of additional points that were not directly incorporated into the executive summaries of the discussion papers.
5.1.A Executive summary
Introduction
This document focuses on evaluating the nature of the adverse health effects associated with foodborne non-typhoid and non-paratyphoid Salmonella spp. and how to quantitatively assess the relationship between the magnitude of the foodborne exposure and the likelihood of adverse health effects occurring.
Objectives
The objective and scope of the Salmonella Hazard Characterization document is to provide:
A review of the characteristics of the host, organism and food matrix.
A summary and review of available data and information on adverse health effects.
A summary and evaluation of existing dose-response models with respect to assumptions, sources of uncertainty, strengths and limitations.
A description of the use of available outbreak data to evaluate published dose-response models.
An evaluation of the outbreak data for evidence of a difference between susceptible and normal populations and between Salmonella Enteritidis and other strains.
Approach
Information was compiled from published literature and from unpublished data submitted to FAO/WHO by public health agencies and other interested parties. The first section of the document provides a description of the public health outcomes, pathogen characteristics, host characteristics, and food-related factors that may affect the survival of Salmonella in the human gastrointestinal tract.
The second section of the hazard characterization document presents a review of the background and rationale for different models that have been reported and used to estimate the dose-response relationship of Salmonella. These models mathematically describe the relationship between the numbers of organisms that might be present in a food and consumed (dose), and the human health outcome (response). There are three different models for salmonellosis that have been published or reported: the USDA-FSIS-FDA Salmonella Enteritidis model, the Health Canada Salmonella Enteritidis model, and a beta-Poisson model fit to human feeding trial data for various Salmonella species.
An extensive review of available outbreak data was also conducted, and data appropriate for dose-response estimations were summarized. The dose-response curves reviewed were then compared with the outbreak data to equate the model with observed information. Where possible, the outbreak data were also used to characterize the differences that may exist between the potential for infection in susceptible and in normal segments of the population. Finally, the outbreak data were used to estimate additional dose-response models.
Overall, the document on "Hazard identification and hazard characterization of Salmonella in broilers and eggs" provides a summary of a vast amount of literature available on this subject.
Key findings
In most people, the gastroenteritis lasts 4 - 7 days and patients fully recover without medical treatment. However, some people may develop more severe illness, including potentially fatal infections of the bloodstream or other parts of the body or long-term syndromes such as reactive arthritis and Reiter's syndrome.
Clinical manifestations of Salmonella infections in animals generally differ from the typical gastroenteritis and other sequelae produced in humans, therefore, extrapolations of disease in animals to disease in humans must be done with great caution.
In the case of Salmonella, unlike most other bacterial pathogens, there is a reasonable amount of human data. As a result, it was felt that the inclusion of additional information from animal data may contribute to increasing the uncertainty rather than improving the dose-response relationship.
Insight into the potential for some segments of the population to be more susceptible to Salmonella infection than others was provided by data extracted from two outbreaks. Assuming children under 5 years of age represented a more susceptible population, it was estimated that at the doses observed in these outbreaks (approximately 2 and 4 log CFU/g), the susceptible population was 1.8 to 2.3 times more likely to get ill.
A review of currently available outbreak data did not produce any evidence to support the hypothesis that Salmonella Enteritidis has a higher likelihood of causing illness upon ingestion than a similar dose of another serovar.
The outbreak data indicate that the dose-response relationship (or infectivity/pathogenicity) for all non-typhoid and non-paratyphoid Salmonella spp. are similar and could theoretically be characterized using a common model. Specifically, the epidemiological data does not offer any evidence to conclude that different serotypes are more or less pathogenic than others.
Complete outbreak data are sparse and important information for the calculation of dose-response assessments is often missing from outbreak reports. In particular, enumeration of organisms in the implicated food vehicle is frequently not carried out in many outbreak investigations. Valuable data for this report was provided by Japan[2], where since 1997, all large foodservice establishments have been advised to keep frozen portions of prepared foods for a minimum of 2 weeks for subsequent testing if illness is associated with the food. These data allowed significant insights to be made into the hazard characterization of Salmonella.
Five models are summarized below and in Figure 5.1. Three models are published or documented in official reports and two new models were generated from the collected outbreak data. They are:
i. Naïve human feeding trial data beta-Poisson model
The model suffers from the nature of the feeding trial data (i.e. the subjects used were healthy male volunteers) and may not reflect the population at large. The model tends to greatly underestimate the probability of illness as observed in the outbreak data, even if the assumption is made that infection, as measured in the dose-response curve will equate to illness.
ii. USDA-FSIS-FDA Salmonella Enteritidis beta-Poisson model
The model uses human feeding trial data for Shigella dysenteriae as a surrogate pathogen with illness as the measured endpoint in the data. The appropriateness of using Shigella as a surrogate for Salmonella is questionable given the nature of the organisms in relation to infectivity and disease. Compared to the outbreak data, and on a purely empirical basis, this curve does tend to capture the upper range of these data.
iii. Health Canada Salmonella Enteritidis beta-Poisson model
To date this model has not been fully documented and lacks transparency. The model uses data from many different bacterial pathogen-feeding trials and combines this information with key Salmonella outbreak data using Bayesian techniques. Using data from many bacterial feeding trials and the current lack of transparency is a point of caution. Empirically, the curve describes the outbreak data at the low dose well but tends towards the lower range of response at higher doses.
iv. Outbreak data exponential model
The exponential model fit to the outbreak data does not produce a statistically significant fit. The curve does provide an adequate description of the data at the mid- and high-dose ranges, however, it underestimates the low-dose observed data.
v. Outbreak data beta-Poisson model
Similar to the exponential function, the beta-Poisson model, when fit to the outbreak data, does not produce a statistically significant fit. The curve does produce an adequate characterization of the observed data in the low to mid-dose range. The low-dose range of the dose-response relationship is an especially important area.
Figure 5.1: Comparison of Salmonella
dose-response models.
NOTE: The points on the curves do not represent data
points and are used only for legend purposes.
Gaps in the data
Outbreak and epidemiological data, specifically indicating: concentration in the implicated food, amount of food consumed, accurate numbers on ill and exposed populations, accurate characterization of the population including age profiles, medical status, sex and other potential susceptibility factors.
Quantitative data measuring the impact of the food matrix effects on the probability of infection.
Quantitative information to facilitate estimating the probability of developing sequelae following illness.
Characterization and quantification of the relationship between susceptibility factors and increased likelihood of infection.
Conclusions
The derivation of any one of the models is based on many assumptions, such as the use of S. dysenteriae or other surrogates for Salmonella, combining results of feeding studies for different pathogens, the relevance of infection versus illness as endpoints, and the study design and health status of the test subjects in the human feeding trials. The outbreak data revealed several uncertainties and several assumptions had to be made to derive some of the outbreak estimates subsequently used to fit new dose-response curves.
At present, a single model representation for the relationship between dose and response can not be highlighted as vastly superior to any other model. Compared to the reported outbreak data, the naïve beta-Poisson model is the least desirable since it vastly underestimates the probability of illness and tends towards the lower bound even when the assumption is made that all infections lead to illness. The remaining models were relatively reasonable approximations, with different degrees of under- or over-prediction of illness based on the outbreak data described in this report. The models fit to the outbreak data appear to offer reasonable potential given that they qualitatively, though not statistically significantly, describe observations in a real world environment.
Recommendations
Consideration should be given to the inclusion of S. typhi and S. paratyphi in future hazard characterizations. A dose-response relationship for all Salmonella spp. could prove to be of great utility, and the added information from S. typhi could also serve to expand the current information.
This document did not consider a quantitative evaluation of secondary transmission (person-to-person) or chronic outcomes. In addition, the impact of the food matrix was not incorporated into the assessment. These may be considerations for future document development.
Additional data will help to refine the information currently available and ideally support the development of better risk assessments to help make more accurate predictions regarding the safety of foods contaminated with Salmonella and other pathogens of public health concern.
The importance of accurate and complete epidemiological data collection during outbreak investigations should also be communicated and encouraged.
5.1.B Summary of discussions related to hazard identification and hazard characterization of Salmonella
General Comments
The expert consultation welcomed the technical report prepared by the expert drafting group as a significant advancement towards the understanding of the hazard characterization of Salmonella. It was recommended that the title of the document should be changed to “Hazard identification and hazard characterization of non-typhoid and non-paratyphoid Salmonella” to better reflect the scope of the work.
Dose response curves
The consultation agreed that the inherent variability of the Salmonella dose-response data summarized in the report necessitated the fitting of several dose-response curves to describe the outbreak observations. Evidence does not support the selection of a single curve for summarizing the Salmonella dose-response data at this time. Given the limited number of data sets and reliability/variability of the data, it was decided to retain all candidate dose-response curves and provide commentary on the degree of fit, practical suitability, and pros and cons of each curve.
The appropriateness of using Shigella as a surrogate for Salmonella is questionable given the nature of the organisms in relation to infectivity and disease. However, compared to the outbreak data, and on a purely empirical basis, this curve does tend to capture the upper range of the outbreak data.
Specific comments related to dose response curves based on outbreak data
The dose-response data from outbreaks were highly relevant to foodborne illness but models produced using available data had a poor statistical fit. There were limitations to the outbreak data as the data were collected over several decades and the investigative methodology may have changed without including corrections to account for methodological changes. It was noted that the majority of the data were from North America, Europe and Japan. General applicability of outbreak data dose-response assessments would be improved by including data from other countries.
Much discussion focused on the advantages/disadvantages of deleting particular observations because they did not fit the general trend of the outbreak data (see Figure 5.2 - outliers). The requirement to be transparent dictates that when data are excluded the authors/modellers should be explicit regarding the reasons for this.
It was suggested that a linear regression of attack rate on the log10 number of Salmonella consumed might provide an alternative description of the dose response relationship. Such a model would have the advantage of simplicity but was not preferred to beta-Poisson type models that reflect an underlying biological basis of infection. It was noted that the linear regression approach could disguise the truly non-linear relationship between concentration and attack rate.
The present models do not account for food matrix effects nor fully account for host or pathogen variations. One example of a host effect that is not captured in the model is treatment with antibiotics that may make an individual more susceptible to infection because of changes in the composition of the gut flora. It was noted that good calibration studies would help identify how to consistently adjust specific attack rates for individual outbreaks.
It was noted that Salmonella typhi and Salmonella paratyphi outbreaks were not included in the data used to generate the present model. Experts agreed a hazard characterization that includes S. typhi and S. paratyphi data should be considered in the future.
Figure 5.2. Outliers. The three data points circled are deleted for modelling outbreak data because they did not fit the trend of the data. For further discussion of the rationale behind this, see the full document referenced below.
Note: Numbers 1 - 34 in the legend refer to outbreak data.
These outbreaks are described in detail in the text of the document on Hazard
identification and hazard characterization of Salmonella in broilers and
eggs available on the
Internet:
http://www.fao.org/WAICENT/FAOINFO/ECONOMIC/ESN/pagerisk/riskpage.htm
and
http://www.who.int/fsf/mbriskassess/index.htm
Gaps in the data
Specific information regarding incidence of disease in countries other than North America, Europe, and Japan. This may require the use of regional centres for collection of this information.
Studies on the microbial ecology to determine reservoir and quantity of organism would support hazard identification.
Recommendations
WHO should review and update specific guidelines to ensure more thorough and consistent investigation of outbreak data so that these data are available for hazard characterization. Specific guidelines are needed to address the problems of measurement error in data (including both the accuracy of bacterial enumeration and determination of the number of individuals exposed), variations in case definitions of illness for different outbreak investigations, and better identity of suspected/incriminated foods. This should also include guidelines for selecting data sets used to determine dose-response relationships including recommendations regarding sampling, laboratory methods, epidemiological methods and analysis.
FAO/WHO should strongly encourage Member Countries to publish existing and new epidemiological data and to conduct epidemiological investigations of outbreaks with data that include enumeration of pathogens and denominators for attack rates.
FAO/WHO are encouraged to facilitate the development of protocols used to measure attack rates and collection of surveillance data.
The consultation recommends to FAO/WHO that prevalence and outbreak data be compiled in a central repository.
FAO/WHO should facilitate the development of methods suitable for enumeration of salmonellae in foods including identification/development of new enumeration techniques.
Current models should be expanded to include food matrix and host effects.
5.2.A Executive summary
Introduction
A critical review of the existing exposure assessment models can contribute to the advancement of microbiological risk assessment. Discussions between the FAO/WHO secretariat and the expert drafting group determined the need for a comparison of existing exposure assessments to characterize the state of the art in the practice of risk assessment. Such a comparison would identify similarities and differences between existing models. This approach should be beneficial to future exposure assessments of this pathogen-commodity combination.
The review intends to identify those methods that were most successful in previous exposure assessments, and also recognize the weaknesses of those assessments as a result of inadequate data or methodology. Although no specific risk management direction was provided for this report, the findings should be useful for future risk management.
Objectives
The purpose of this report is to compare existing techniques and practices used to construct an exposure assessment for Salmonella Enteritidis in eggs and to provide a framework for future exposure assessments of this pathogen-commodity combination.
The scope of this analysis is limited to the probability of human exposure associated with eggs that are internally contaminated with Salmonella Enteritidis. The analysis and conclusions are similarly focussed to only apply to currently understood mechanisms and variables as used in previous exposure assessments. Therefore, caution should be exercised in interpreting this report in relation to data that has become available since these models were completed.
Approach
Five previously prepared exposure assessments of Salmonella Enteritidis in eggs were reviewed. Of these, three exposure assessments were selected for in-depth comparisons.
These were:
USDA-FSIS-FDA. Salmonella Enteritidis risk assessment: Shell eggs and egg products. Final Report. June 12, 1998.
Health Canada. Salmonella Enteritidis risk assessment model. Unpublished.
Whiting R.C. and Buchanan R.L. (Whiting). Development of a quantitative risk assessment model for Salmonella Enteritidis in pasteurized liquid eggs. International Journal of Food Microbiology 36:111-125, 1997.
Four stages of a "farm-to-table" exposure assessment were defined: production, distribution and storage, egg products processing, and preparation and consumption. The production stage considers the laying of Salmonella Enteritidis contaminated eggs. The distribution and storage stage considers the time between lay and preparation of egg-containing meals. The egg products processing stage considers commercially broken eggs that are usually pasteurized. Preparation and consumption stage considers the effects of different meal preparation practices and cooking.
The USDA/FSIS-FDA exposure assessment included all above four stages of an exposure assessment. The Health Canada model included production, distribution and storage, and preparation/consumption, but did not cover egg products processing. The Whiting model focused on egg products processing, but it also included elements of production and distribution/storage stages.
Generally, data considered in this analysis applies to either occurrence or concentration of Salmonella Enteritidis. Specific data used in previous exposure assessments are presented and analysed for each of the model stages. To provide a more complete description of available data, a summary of published and non-published research on Salmonella Enteritidis occurrence and concentration was undertaken. Although some of these data are not used in the three previously prepared exposure assessments, their inclusion in the report provides currently available data that could assist future exposure assessments.
Key findings
Accurate estimates of prevalence inputs require that surveillance data be adjusted to account for likelihood of detection and other biases. The USDA/FSIS-FDA model includes such adjustments, but the other two models did not.
In the distribution and storage stage, there is a need to separately model growth for each distinct pathway to account for different time and temperature distributions. Growth of Salmonella Enteritidis inside eggs was found to be sensitive to assumed temperature distributions at retail and consumer storage in two exposure assessments. When time and temperature inputs are similarly defined, the USDA/FSIS-FDA and Health Canada models give similar predictions (see Figure 5.3).
Figure 5.3. Comparison between USDA/FSIS-FDA and Health Canada models. On the left are predicted distributions of logs of growth for those contaminated eggs in which growth occurs. On the right are these predictions when the USDA/FSIS-FDA model temperature inputs are modified to be similar to the Health Canada model inputs.
The USDA/FSIS-FDA and Whiting models of the egg products processing stage predicted wide variability in pasteurization effectiveness. This finding substantially influences the predicted number of Salmonella Enteritidis remaining in egg products after pasteurizing.
The USDA/FSIS-FDA model predicts an increased probability of exposure associated with pooling of eggs in the preparation stage, while the Health Canada model shows a decrease in probability of exposure associated with egg pooling. This difference occurs because the Health Canada model does not include post-pooling growth and restricts pooling scenarios to those only involving scrambled egg meals. If more diverse pooling scenarios were considered in the Health Canada model, then pooling might more significantly contribute to probability of illness in that model.
Gaps in the data
Data relating to the ecology of Salmonella Enteritidis in eggs are needed. This need is seemingly universal in its application to previous and future exposure assessments.
It was recognized that estimating the number of Salmonella Enteritidis contaminated eggs at the production stage was based on data from, at most, 63 eggs. More epidemiological and enumeration data would improve modelling of egg contamination.
To adequately assess preharvest interventions, more data are needed on the prevalence of Salmonella Enteritidis in breeder and pullet flocks, as well as in feedstuffs. In particular, associations between the occurrence of Salmonella Enteritidis in these preharvest steps and its occurrence in commercial layers should be quantified.
Better data on time and temperature, specifically in relation to egg storage would serve to build confidence in modelling. The importance of time and temperature distributions in predicting growth of Salmonella Enteritidis in eggs - combined with the lack of reliable data to describe these distributions - highlights the need for these data.
The high degree of uncertainty/variability in cooking effectiveness inputs noted in the comparison of the models also highlights the need for more research on these inputs.
To predict reliably the effectiveness of regulatory standards concerning egg products, there is a need for additional data concerning the concentration of Salmonella Enteritidis in raw liquid egg before pasteurization.
The exposure assessments considered in this report primarily relied on relevant North American data. Additional data will need to be collected to conduct exposure assessments in countries where egg contamination with Salmonella Enteritidis is different to that in North America. For example, countries will probably need to assess the prevalence of Salmonella Enteritidis in their egg industry. The marketing fractions, times and temperatures of storage, and preparation and cooking practices will probably differ in other countries. Therefore, these exposure assessment inputs will need to be estimated from country- or purpose-specific data.
Conclusions
This report identifies similarities and differences between previously prepared exposure assessments of Salmonella Enteritidis in eggs. Potential pitfalls, important data analysis, and critical data needs are reported for each stage of a "farm-to-table" exposure assessment. This report does not intend to provide detailed guidelines on how to conduct an exposure assessment of this pathogen-commodity combination. Additional work is required to develop such guidelines. In addition to this report’s findings, those wishing to complete such an analysis should refer to the original papers cited in the report, as well as risk analysis texts.
Many similarities were found in the approaches used by the three exposure assessments analysed in this report. For example, the distributions for initial number of Salmonella Enteritidis per egg were derived in a similar manner. The growth equations were similar, as were the pasteurization equations. Often, the same distribution types were used to model the same inputs, although different parameters might be specified. The modelling approaches - for example the pathways considered, and the factors modelled - were very similar.
It was concluded that Salmonella Enteritidis exposure assessments should model growth and preparation/consumption as one continuous pathway. In this manner, growth and decline of Salmonella Enteritidis is explicitly modelled as dependent on the pathway considered.
Predictive microbiology should be common to any exposure assessment of Salmonella Enteritidis in eggs. Because environmental conditions differ on an international level, time and temperature distributions may be different between analyses. Yet, it was concluded that the predictive microbiology equations used in future exposure assessments could be similar.
Careful attention should be placed on areas in preparation and consumption where the product changes form or the units change. Pooling eggs into a container creates a product distinctly different from shell eggs. This product is able to support immediate bacterial growth and its storage should be modelled as a unique event.
Given the lack of published evidence on relevant egg consumption and preparation practices among populations of end-users, the preparation and consumption component of an exposure assessment is the most difficult to accurately model. Even with perfect information, this component is very complicated. Multiple pathways reflecting multiple end-users, products, practices, and cooking effectiveness levels ensure that the preparation and consumption component has many difficulties. Nevertheless, the strides taken in previous models can serve as reasonable starting points for subsequent analyses.
Limitations common to the models compared in this report include lack of consideration for possible re-contamination of egg products following pasteurization and/or cooking, and no consideration of cross-contamination of other foods from Salmonella Enteritidis contaminated eggs. Furthermore, the results and conclusions of these models are dependent on conventional assumptions regarding mechanisms of egg contamination. These mechanisms suggest that Salmonella Enteritidis contamination in eggs is initially restricted to albumen and that such contamination enters eggs during their formation inside hen’s reproductive tissues. Also, the growth kinetics estimated for these models are not necessarily representative of all Salmonella Enteritidis strains or other Salmonella serotypes.
While these models are similar to one another, and provide common stages of an exposure assessment, they may require substantial reprogramming to be useful to some countries or regions where the situation is markedly different from that in North America. Such reprogramming may be limited to changing some input distributions, but may also require eliminating or adding some variables or parameters to the models.
Recommendations
The scope of an exposure assessment should be clearly defined and its objectives clearly stated.
The existing exposure assessment models - in combination with dose-response models developed for Salmonella species - can be used to evaluate some mitigation strategies, but the applicability of their findings would most likely be limited to the countries for which the model inputs were derived.
Because the predictive microbiology of Salmonella Enteritidis in eggs is considered common to any exposure assessment of this pathogen-commodity combination, its progress and further development are encouraged. A central repository of data on the mathematical behaviour of Salmonella Enteritidis in eggs should be established. FAO and WHO should consider supporting this initiative.
In parallel with the current work to develop guidelines for hazard characterization, FAO/WHO should facilitate the development of guidelines for exposure assessments, including approaches to be used when data are limited or only semi-quantitative in nature.
5.2.B Summary discussion related to exposure assessment of Salmonella Enteritidis in eggs
General Discussion
The consultation welcomed the technical report as an important contribution on exposure assessment of Salmonella Enteritidis in eggs. It agreed that inclusion of epidemiological concepts in determining flock prevalence, within flock prevalence, and apparent prevalence was a significant feature of this study.
Current models for exposure assessment are used to predict the public health benefits of an intervention imposed at any time prior to consumption. However, data may not be available for assessing all candidate interventions. If the concern is at consumption, data prior to this are not required. In contrast, the assessment of interventions applied in the pre-harvest period demand data is available for a larger segment of the production-consumption process.
Limitations of the models presented are that they do not allow a discussion of cross-contamination and recontamination although anecdotal evidence suggests that these are important. The models are not specifically designed to evaluate the importance of vertical transmission in breeder flocks. The consultation noted the need to consider this and the geographic variation of flock prevalence.
The current models also place emphasis on the potential for growth of Salmonella Enteritidis in eggs. The consultation noted that growth was not always necessary for human infection, as very low doses can be infectious. Although published information does not indicate a difference in heat sensitivity between Salmonella Enteritidis and other serotypes, recent evidence suggests that specific strains (e.g. Salmonella Enteritidis PT4 containing a 25 mD plasmid) may have differing growth characteristics.
Robust testing and sampling methodologies are essential for exposure assessments. For example, when considering pooled eggs, small numbers of organisms present in the eggs may reduce the probability of detection.
The models assume that Salmonella in naturally infected eggs are located in the albumen outside the vitelline membrane. The models also assume there is a brief period of time immediately following lay where there may be growth of Salmonella Enteritidis in the egg. This assumption may not be valid in all cases. Anecdotal data suggests there may be conditions where Salmonella grows rapidly in intact eggs.
The breakdown of the yolk membrane is a key concept of the model. However, some experts expressed a cautionary note on the simplification introduced by modelling the lag phase in cumulative fashion. Assumptions such as this may be necessary for simplicity of the model; however, they result in a less than precise depiction of the real world situation.
Gaps in the data
Additional studies on the number of Salmonella Enteritidis in naturally contaminated intact shell eggs are required (information is currently available for 63 intact shell eggs). There is also a need for enumeration data of Salmonella Enteritidis in raw liquid egg.
Additional data are required on the duration of storage in retail stores and in homes, and on temperatures experienced during those storage periods.
Information is required on the characteristics of flocks in different countries. This information would include flock age profiles and the size of flocks.
Information is required on individual preferences for storage and consumption in different countries including information regarding the actual cooking practices used. For example, time between lay and consumption differs between the US and Canada. This important variable must be determined for each individual country.
More inoculation studies are required to assess survival time of Salmonella Enteritidis in eggs.
Additional studies are required to estimate the kinetics of growth as a function of egg composition and strain infectivity as well as heat sensitivity between various Salmonella Enteritidis strains.
Studies need to be conducted on sites of Salmonella Enteritidis infection in the reproductive tract of hens.
Studies are required to investigate the possibility that under certain conditions Salmonella can grow rapidly in intact eggs.
5.3.A Executive summary
Introduction
An understanding of Salmonella spp. in broilers is important from both public health and international trade perspectives. As a result, there is an urgency to evaluate this pathogen-commodity combination by quantitative risk assessment methodology. To date, no full quantitative exposure assessments have been undertaken in this respect. This work illustrates a way that such assessments can be developed.
Objectives
This report focuses on the development of a model framework, highlighting ideal data requirements and possible methodologies. In addition, it presents available data for developing such models and makes an assessment of their usefulness. It is not intended to present a full farm-to-fork model; rather, the content of the report can be used for guidance. Where appropriate, example models are presented to illustrate possible methodologies related to individual steps that could be included within a full model.
Considering the proposed methodologies and available data, areas of limited information are highlighted and recommendations for directing future study are made.
Approach
The report begins by presenting an overall model framework that describes the exposure pathway from the farm to the point of consumption (see Figure 5.4). The pathway consists of a number of related modules (production, transport and processing, retail, distribution and storage, preparation) that describe the changes in prevalence and concentration of organisms. If the framework is used to construct a model, the outputs can then be combined with consumption data to estimate exposure.
Figure 5.4: Modular pathway to describe the production to consumption pathway.
P: Changes in prevalence
N: Changes in numbers of organisms
Issues common to all steps on the pathway are discussed. In particular, data related factors are explored. These factors include possible data sources (published, regulatory and industry), problems associated with obtaining data from different sources, combining data from disparate sources and selecting the most valid data. Different modelling approaches are also summarized including the use of static and dynamic models, deterministic and stochastic alternatives and the appropriate incorporation of uncertainty and variability. With these points in mind, the individual modules are considered in detail.
The production module aims to estimate the prevalence of Salmonella positive broilers at the time of leaving the farm for slaughter. The number of organisms per positive bird is a required output. Ideal data requirements for this step are outlined and include source of infection, flock prevalence, within flock prevalence and full details of study methods including sampling (e.g. site, selection, timing, relationship to overall population) and microbiological methods used. The number of organisms per bird is also essential.
Processing of broilers is outlined in the second module and here the aim is to estimate prevalence and concentration at the end of processing. For this module, ideal data relates to changes in numbers and prevalence during the various steps of processing, together with details of the study as discussed previously. Such information should capture the importance of cross-contamination during this step.
Retail, distribution and storage (module 3) considers the time after processing and before preparation and consumption by the consumer. The aim is to estimate the change in the number of organisms per contaminated product. These stages can be considered as a series of time / temperature profiles to which the broiler is exposed and, therefore, growth and survival are the critical microbial processes. There are two classes of ideal data for this module. Firstly, the time / temperature data which describes the processes and secondly appropriate predictive models to describe the growth and survival processes.
Preparation is considered in module 4 to estimate the change in numbers as a result of preparation prior to consumption. Ideally, cross-contamination should be modelled in this stage and thus appropriate models and data are required for this module. In addition, when considering frozen broilers, data are needed to describe the thawing process. Finally, data relating to cooking are required for use in predictive models that describe thermal death.
In the final module consumption patterns are considered. Ideally, this requires data on consumption patterns of a population. To be useful, the population should be divided into sub-groups that could be based, for example, on age, sex or immune status, etc. Consumption data must be national - generalisation is not appropriate.
Although no full exposure assessments have been undertaken for this pathogen-commodity combination, there are models available that start later in the exposure pathway (e.g. the start of processing and retail). A full exposure assessment for Campylobacter in broilers has been undertaken. These assessments are reviewed to determine their usefulness for a full exposure assessment of Salmonella in broilers (recognising the differences between Campylobacter and Salmonella). From this review, many of the models have features that could be utilised.
Key findings
Modelling the full exposure pathway from farm-to-fork is a complex process. The individual modules of this pathway will be complex and may have high degrees of associated uncertainties which, when combined, can generate an estimate of exposure with a wide range of uncertainty. Consequently, it is important to consider the points where modelling should begin and end. This will be defined by the risk management question.
When collating data from a large number of dissimilar studies, it is important to present this information in tabulated form, considering the ideal data requirements identified prior to collection. Such presentation enables critical evaluation of the data and helps to ensure that the most valid data are selected.
With regard to models for individual stages of the exposure assessment, there is a balance between the need for accurate prediction and the simplicity of the approach taken. This should be considered during the model selection process.
Gaps in the data
The main gaps identified in the data are as follows:
There is limited prevalence data for many regions of the world and for areas where prevalence is reported information on the study design is often lacking.
For all stages of the exposure pathway, there is little recent quantitative data on the numbers of organisms per bird.
Cross-contamination data is extremely scarce and modelling of this event is in its infancy.
Often data are presented as average consumption per day and this is less useful than data that describes portion size and frequency of consumption.
There are a limited number of models to describe survival under chilling and freezing conditions.
Specific consumption data are limited for most geographical locations.
Conclusions
The technical report illustrated that modelling exposure for Salmonella in broilers from "farm-to-fork" is a realistic proposition. The framework proposed in the report presented standard modelling techniques in a modular fashion and the output from such a framework could be readily integrated into risk characterization if required. Difficulties in modelling individual stages (e.g. limited data for pathway analysis) and the complexities associated with describing biological processes (e.g. cross-contamination) were identified.
The model framework illustrated the importance of suitable data inputs to ensure a robust exposure assessment. In particular, data should be representative, of appropriate quality and sufficient to meet the purpose and scope of the risk assessment.
Recommendations
The following recommendations for directing future work can be made.
Reporting prevalence at different steps of the full exposure pathway should be encouraged in all regions of the world.
Reported data should give full details of study methodology, including sampling site, sampling time, the relationship of the sample to the overall population and microbiological methods.
Determination of quantitative data should be encouraged. If it becomes available then full exposure assessments could be developed to investigate mitigation strategies (e.g. use of chlorine in chill water) or compare alternative practices (e.g. air chilling versus immersion chilling).
Cross-contamination during processing and handling operations should be studied quantitatively and methodologies for modelling this process should be developed. Cross-contamination during these stages is a critical factor that is often associated with outbreaks.
The collection of consumption data should be promoted at the national level. The design of these studies should accommodate the data requirements for exposure assessments. These requirements include population variability, portion size and frequency of consumption.
In predictive microbiology, survival has been less well studied than growth or death. There are few predictive models that describe survival at chill and frozen temperatures. Further development of these models is essential.
5.3.B Summary of discussion related to exposure assessment of Salmonella spp. in broilers
General Discussion
The consultation welcomed the technical report prepared by the expert drafting group as a significant contribution to the exposure assessment of Salmonella spp. in broilers. Although the structure of the model was strongly supported, any exposure assessment completed would be of limited representativeness because most input data was only obtainable from a small number of countries.
Different processing techniques, including freezing of chicken meat and carcasses, are common in the international trade of broiler products. Consequently, the effects of freezing on the concentration of Salmonella were identified as an important data gap for exposure assessments addressing the international trade of poultry.
The aim of an exposure assessment is to model the dose of Salmonella consumed. When chicken products enter the kitchen they are subjected to a variety of preparation steps that introduce a wide range of opportunities for cross-contamination. Because it is difficult to identify and evaluate all of these processes, modelling of events in the kitchen is a difficult proposition.
The identification and acquisition of all potentially available data is a significant problem in conducting exposure assessments. In many cases the most desirable data for modelling is proprietary or unpublished. Commercial interests need some assurance that providing their proprietary data will not prejudice their business.
The consultation stressed the importance of clear tabulation of collected data with respect to ideal data requirements. In particular future presentation of such data should include, where possible, details of microbiological methods.
It was noted that specific sampling and enrichment methods used in studies influences the reliability and accuracy of the data, e.g. poultry rinse samples, swab samples and excised skin do not yield comparable results. Similarly, the culture of poultry litter samples may have a different sensitivity when compared with the culture of cloacal swabs.
Gaps in the data
Information on the distribution of time and temperature for storage and cooking in a variety of national environments is lacking. Similarly pathways for cross-contamination are difficult to identify and model.
Different enumeration methods vary in sensitivity. More sensitive enumeration methods are needed.
More detailed data are required describing consumption patterns to improve risk estimates.
Knowledge of survival times of Salmonella below 4ºC is essential to conduct predictive microbiological exposure assessments that are representative of international trade of poultry products.
More data on Salmonella-free feed, Salmonella-free replacement stock, fasting prior to slaughter, and scalding, defeathering and evisceration processes are needed to effectively model the benefits of control interventions at the levels of production and processing.
FAO/WHO should provide a means for the food and agriculture industry to provide proprietary data on pathogens and process variables in foods.
FAO/WHO should request the CAC to provide focused risk management questions for further development of risk assessments on Salmonella.
|
[2] In accordance with Japanese
notification released on March 1997, large scale catering facilities (> 750
meals per day or > 300 dishes of a single menu) have been advised to save
food for future examination in the case of illness being associated with the
food. Fifty gram aliquots of each raw food material and cooked dish should be
saved for a minimum of 2 weeks at temperatures below -20 ºC. Although this
notification is not mandatory, it is also applicable to smaller catering
facilities with social responsibility such as those in schools, day-care
centres, and other child-welfare and social-welfare facilities. Some local
governments also have regulations relating to food saving, but the required
duration and the temperature of storage vary. |
