6. RISK ASSESSMENT OF LISTERIA MONOCYTOGENES IN READY-TO-EAT FOODS

6.1 INTRODUCTION

Foodborne listeriosis represents a relatively rare but clinically serious disease, with high case-fatality rates (20-30%) that largely affects specific segments of the population with increased susceptibility. The microorganism is widely dispersed in the environment and foods. Despite the fact that a wide variety of foods may be contaminated with L. monocytogenes, outbreaks and sporadic cases of listeriosis appear to be predominately associated with ready-to-eat (RTE) products. RTE food is a large, heterogeneous category of foodstuffs and can be subdivided in many different ways. According to the Codex definition (CAC/GL 22-1997), ready-to-eat food includes any food (including beverages) which is normally consumed in its raw state or any food handled, processed, mixed, cooked, or otherwise prepared into a form in which it is normally consumed without further processing. Furthermore, RTE foods differ in different countries according to local eating habits, availability and the integrity of the chill chain and regulations regarding, for example, the maximum temperature at retail level.

6.2 SCOPE OF THE RISK ASSESSMENT

The current risk assessment was undertaken in part to determine how previously developed risk assessments done at a national level could be adapted or expanded to address concerns related to L. monocytogenes in ready-to-eat foods at an international level. In addition, after initiation of the risk assessment, the risk assessors were asked by the 33^rd session of the CCFH, through FAO and WHO, to consider three specific questions related to ready-to-eat foods in general. These questions were:

• Estimate the risk for consumers in different susceptible population groups (elderly, infants, pregnant women, and immunocompromised patients) relative to the general population.

• Estimate the risk from L. monocytogenes in foods that support growth and foods that do not support growth under specific storage and shelf-life conditions.

• Estimate the risk from L. monocytogenes in food when the number of organisms ranges from absence in 25 grams to 1000 colony forming units (CFU) per gram, or does not exceed specified levels at the point of consumption.

Considering the resources available and the time constraints placed on the risk assessors, it was impossible to consider all ready-to-eat foods that could be contaminated with L. monocytogenes. Accordingly, it was decided to limit the risk assessment to a finite range of ready-to-eat foods that have been selected to represent various classes of product characteristics, in order to determine if the risk of these foods serving as a vehicle for human foodborne listeriosis can be estimated. These foods were selected to provide examples of how microbiological risk assessment techniques can be used to answer food safety questions at an international level^[3]. This educational component is a stated goal of the FAO/WHO microbiological risk assessment programme.

It was also decided to limit the scope of the risk assessment to foods at retail and their subsequent public health impact at time of consumption. This was done for two reasons. Firstly, such a scope was sufficient to address the charge provided by the CCFH within the time frame and resources made available to the risk assessors. Secondly, most of the exposure data for L. monocytogenes that is currently available relates to the frequency and extent of contamination at the retail level. More detailed examination of factors contributing to the levels of L. monocytogenes found at retail as a result of manufacturing parameters would have required that a much smaller range of foods be evaluated or that a substantially greater amount of resources and data be made available. Accordingly, the assessment did not evaluate the risks associated with different means of manufacturing these products. However, the risk assessment does consider several post retail factors that could influence the consumers’ risk of acquiring foodborne listeriosis, such as the temperature and duration of refrigerated storage. In addition to invasive listeriosis, L. monocytogenes can also cause mild febrile gastroenteritis in otherwise healthy individuals. The public health significance of this type of listeriosis is uncertain at this time and was not considered in the current risk assessment.

In its request to FAO and WHO in 1999 for expert risk assessment advice (ALINORM 01/13), the CCFH indicated that a farm-to-table risk assessment would provide the broadest range of risk management options. However, the expert drafting group considered only the retail to consumer part of the food chain for the reasons explained above. The expert consultation agreed that there may be a need for CCFH and national risk managers to commission new risk assessments for specific foods or product categories if pre-retail considerations are to be addressed.

6.3 APPROACH TAKEN

The drafting group split into two working subgroups addressing the hazard characterization and the exposure assessment components. The exposure assessment subgroup further split into two subgroups. This appears to be reflected in both the writing styles within the document provided to the consultation and the multiple approaches used in the exposure assessment.

While it was recognized that the exposure assessment and hazard characterization components were part of a complete risk assessment, the expert consultation recommended that these sections be written in a manner that would allow their use as stand alone documents. In particular, it would be beneficial if the hazard characterization could be written in a manner that would allow it to be used as a stand-alone training document. The expert drafting group was made aware of this and will further amend the document.

Stochastic approach

The stochastic approach, as opposed to the deterministic approach, was used for the risk assessment. Stochastic means that inputs for a model are obtained by sampling from probability distributions. This allows uncertainty (which can be reduced if more data are gathered) and variability (a pervasive feature of biological data) to be propagated through the model and reflected in the model output. Other approaches, such as point estimates, interval modelling, etc. have their own value but are less versatile and often less adequate for demonstrating the impact of uncertainty and variability. Point estimates are useful to provide a quick estimate of the magnitude of risk.

The expert consultation agreed with this approach but encouraged the incorporation of an appropriate general reference to other potential methodologies in the final report of the risk assessment.

6.3.1 Hazard identification

L. monocytogenes is widely distributed in the environment and has been isolated from a variety of sources including soil, vegetation, silage, faecal material, sewage and water. There is evidence to suggest that it is a transitory resident of the intestinal tract in humans, with 2 to 10 % of the general population being carriers of the organism without any apparent adverse health consequences. The bacterium can grow at refrigerator temperatures. It is more resistant to various environmental conditions than most other non-spore-forming, foodborne pathogenic bacteria, allowing it to survive longer under adverse conditions. Most cases of human listeriosis are sporadic and the source and route of infection are usually unknown, however, contaminated food is considered to be the principal route of transmission. Foods most often associated with human listeriosis are ready-to-eat products that support growth of L. monocytogenes. Typically, these foods have long refrigerated shelf lives, and are consumed without further listericidal treatments (e.g., cooking).

Invasive listeriosis (i.e., severe L. monocytogenes infections) is a relatively rare but often severe disease with incidence rates typically of about 4 to 8 cases per 1,000,000 individuals and fatality rates of 20 to 30 % among hospitalized patients.

Most strains of L. monocytogenes appear to be pathogenic but their virulence, as defined in animal studies, varies substantially. Listeriosis is an opportunistic infection that most often affects those with severe underlying disease (e.g. immunosuppressive therapy, AIDS, and chronic conditions, such as cirrhosis, that impair the immune system), pregnant women, unborn or newly delivered infants and the elderly. The bacterium most often affects the bloodstream, the central nervous system or the pregnant uterus. Manifestations of listeriosis include, but are not limited to, bacteremia/septicemia, meningitis, meningo-encephalitis, encephalitis, miscarriage, stillbirth, premature birth, neonatal disease and prodromal illness in pregnant women. Incubation periods range from a few days up to three months.

6.3.2 Exposure assessment

The objective of this phase of the risk assessment was to determine the frequencies and extent of ingestion of L. monocytogenes in RTE food meals using examples of food types, and population groups. Initially, existing exposure assessments were reviewed as a background in developing a modelling approach.

Seven examples to be modelled for exposure assessment were selected such that several food related attributes could be considered; different food commodities, lightly processed and highly processed RTE foods, potential for growth or not during long-term storage, cold-chain integrity, potential for inactivation, e.g., pasteurization, post-process contamination, expected high contamination load of final RTE foods, high consumption rates, and use in international trade. Modelling was done from retail to the point of consumption.

The steps taken were:

• estimation of prevalence and cell numbers of L. monocytogenes at retail level for each of the selected RTE foods;

• identification of conditions allowing or preventing growth of L. monocytogenes between retail and consumption, especially time and temperature of storage of the product;

• estimation of frequency and size of meal servings of the specific RTE foods for both less susceptible and more susceptible populations;

• determination of cell numbers of L. monocytogenes in contaminated servings consumed by less susceptible and more susceptible populations.

The influence of climate and season in different regions in the world could not be determined and was not considered.

An influence diagram to compare consumption frequency and amount based on gender, age, and susceptible population was incorporated into the exposure assessment.

The RTE foods chosen were raw and pasteurized milk, ice cream, soft mould-ripened cheese (information related to cheeses made from raw milk and from pasteurized milk were combined), minimally processed fresh vegetables, cold-smoked salmon, and semi-dry fermented meats. The aim of these examples is to illustrate the effect on risk of:

a) product categories;
b) low contamination levels in products that do not permit growth of L. monocytogenes;
c) long-term storage on L. monocytogenes concentration;
d) consumption patterns and dose ingested.

Due to time constraints, the exposure assessment on cold-smoked salmon was not presented and the expert consultation recommended that this example be completed. Specifically, it was pointed out that the processing of hot and cold smoked fish differs and this should, if possible, be reflected in the exposure assessment.

6.3.3 Hazard characterization

As agreed by the experts, severe listeriosis i.e. infected individuals suffering from life-threatening, systemic infections such as perinatal listeriosis, meningitis or septicaemia, was used as the biological endpoint.

No single previously developed dose-response model was fully able to meet the needs of the current risk assessment in relation to the parameters examined and the simplicity of calculations. For these reasons, alternate approaches were developed and evaluated for the current risk assessment. The general approach was to take advantage of the epidemiological data and detailed exposure assessment available in the FDA/USDA-FSIS risk assessment, but simplify the modelling by describing the dose-response relations using an Exponential dose-response model based on epidemiological data with the best estimation of L. monocytogenes contents of the foods.

The Exponential dose-response model was chosen because of its acknowledged applicability for modelling severe listeriosis, its simplicity as a single parameter model, and its log linear nature at the dose ranges of interest. The equation is:

P = 1 - e^-RN

where P is the probability of severe illness, N is the number of L. monocytogenes consumed, and R is the parameter that defines the dose-response relation for the population being considered. The Exponential model is a non-threshold model, which implies that there is no “minimum infectious dose.” Instead the model assumes that a single L. monocytogenes cell has a small but finite probability of causing illness. The use of the Exponential model meets the recommendations of the draft WHO/FAO Guidelines on Hazard Characterization for Pathogens in Food and Water for the selection of dose-response models for infectious microorganisms. A key attribute of the model is its loglinearity (log dose vs. log probability of illness) at low doses. This implies that if the dose is reduced ten-fold, the probability of illness is reduced ten-fold. In addition, it implies that at low doses, a single serving with a specified level of contamination has the same public health impact as ten servings with ten-fold fewer organisms.

Specific R-values were derived in the current risk assessment for both the less susceptible and more susceptible populations. This was achieved using the consolidated food contamination distribution from the FDA/USDA-FSIS model in conjunction with their annual estimated number of listeriosis cases as a percentage of the total population of either more or less susceptible groups within the U.S. population. This provided values for P and N, so that the R-value could be calculated by rearranging the above equation.

The accuracy of the R-value is dependent on the size and definition of the population being considered, the accuracy of the annual disease statistics, and the reliability of data on the frequency and extent of L. monocytogenes contamination of foods. The uncertainty based on these parameters was estimated and incorporated into the derived R-values. The effect of maximum levels of L. monocytogenes in foods on the calculation of the R-value was evaluated in detail since there is a degree of uncertainty and/or variability related to the actual maximum levels of L. monocytogenes observed in foods. Several different means of calculating the R-values were explored in order to examine the impact of basing the calculations on multiple doses or on the highest contamination level only. This was found to have minimal effect. The R-values selected for subsequent use in the risk assessment were R = 1.06 x 10^-12 with a 5% to 95% range of 2.47 x 10^-13 to 9.32 x 10^-12 for the more susceptible population and R = 2.37 X 10^-14 with a 5% to 95% range of 3.55 x 10^-15 to 2.70 x 10^-13 for the less susceptible population.

The consideration of the variability in virulence among L. monocytogenes isolates and its impact on the dose-response relations for this microorganism have been addressed in detail within the hazard characterization section of the risk assessment. This includes a discussion of how this is handled in different manners in the risk assessments that have been done previously assuming the presence of the most virulent strain. In selecting the approach used in the current risk assessment’s dose-response model, strain virulence is implicitly considered. The drafting group responded to the expert consultations questions related to the consideration of strain virulence by including both a table from the FDA/USDA-FSIS risk assessment and some additional text which describe how virulence variability is handled by the various risk assessments including the current one (Table 6.1).

Differences in the susceptibility exist among the various subpopulations with increased susceptibility to L. monocytogenes. It was decided to consider all persons with increased susceptibility as a single group for the purposes of the subsequent risk characterizations. However, the hazard characterization did consider how the dose-response relations for various subpopulations with increased susceptibility could be modelled individually by using epidemiologically derived relative susceptibility data. A detailed explanation and examples of this alternative approach is provided in section 6.7 where the response to the question from the CCFH on the risk to consumers in different susceptible population groups is addressed.

6.3.4 Risk characterization

The 33^rd session of the CCFH requested an estimation of the risk for consumers in different susceptible population groups (elderly, infants, pregnant women, and immunocompromised patients) relative to the general population. This was addressed by estimating the relative risk of different susceptible subpopulations based on epidemiological data assuming similar consumption patterns in these groups. The estimates of relative susceptibility were then used to estimate dose-response relationships for different susceptible subpopulations using an Exponential model.

Risk characterization was done by combining the dose-response models for the general population and the more susceptible population with the estimated exposure for six of the ready-to-eat products.

TABLE 6.1 Comparison of the characteristics of the dose-response models selected for use in the current risk assessment with the dose-response models developed in other studies.

6.4 KEY FINDINGS OF THE RISK ASSESSMENT

6.4.1 General

The risk assessment document:

• demonstrated that questions pertaining to international food safety issues can be addressed by expanding and/or adapting components of risk assessments done at a national level.

• showed that pre-existing models and data sets can serve as the basis for subsequent quantitative risk assessment efforts. This finding highlights the need to develop archives or clearinghouses where models and data sets can be accessed and used by others.

• identified a number of areas where data gaps exist and clearly demonstrated the need for improved data acquisition for prevalence and growth of L. monocytogenes in food and the incidence of foodborne listeriosis on a global basis.

• recognized that using data acquired from around the world has an underlying assumption that there are no regional differences, unless those differences are explicitly considered.

6.4.2 Exposure assessment

• Exposure assessments were developed for six of the RTE foods from initial prevalence and concentration at the retail level to final concentration in contaminated servings.

• The modelling chosen was to determine final concentration of L. monocytogenes in contaminated servings based either on a time-temperature distribution or by adding other factors of importance to model growth rates.

• Four home refrigerator temperature distributions were chosen based on studies done in three countries to include in the exposure assessment. These affected the predicted growth rate of L. monocytogenes in foods stored for long periods in home refrigerators. These were chosen to demonstrate the importance of different storage temperature distributions in different regions of the world.

• The exposure assessment component of the risk assessment was quite extensive providing an assessment of the exposure to L. monocytogenes from six RTE foods. However, due to the identification of an error in the simulation model the expert consultation decided that these findings should not be reported until the models used in assessing exposure have been subjected to a more extensive review and revised if necessary.

6.4.3 Hazard characterization

• The document critically reviewed the different data sources and mathematical models that have been used for developing microbial dose-response relations, providing practical advice on the strengths and weaknesses of each.

• Available dose-response models for L. monocytogenes were critically reviewed for effectiveness and for fulfilling the criteria established in the draft WHO/FAO Guidelines on Hazard Characterization for Pathogens in Food and Water. The strengths and limitations of each model were evaluated. This includes dose-response models developed using (a) expert elicitation, (b) surrogate animal data, (c) epidemiological (annual disease statistics) plus food survey data, and (d) mixed models.

• The current risk assessment developed for the first time dose-response models that were based on the use of outbreak investigation data in conjunction with the Exponential model. This included dose-response models for both severe listeriosis and for listerial gastroenteritis.

• Two simplified models based on the Exponential model and epidemiological data on severe listeriosis, one for the less susceptible population and a second for the population at increased susceptibility, were developed for the current risk assessment (Figure 6.1). These models were then used for the risk characterization phase of the risk assessment. In addition relative susceptibility data from France and the United States were used to develop dose-response curves for specific susceptible subpopulations including transplant patients, AIDS patients, individuals undergoing dialysis, patients with pulmonary and related cancers, patients with bladder cancer and related cancers, patients with gynaecological cancer, individuals suffering from diabetes, hepatitis or alcoholism, elderly (France and United States), and prenatal and neonatal cases. The alternative dose-response approach is discussed in more detail in section 6.7.1 on the risk from L. monocytogenes to susceptible population groups compared to the general population.

6.4.4 Risk characterization

• Risk characterizations based on the exposure profile of L. monocytogenes at consumption and the dose-response models were used to attempt to estimate predicted cases of listeriosis per serving for each of the six foods. However, the identification of an error in the simulation model precluded the final calculation of risk and further work will be required.

• An alternative, deterministic approach was used on an interim basis to characterize the risk based on the dose-response relations/curve from the current risk assessment and the exposure data from the FDA/USDA-FSIS model. This approach indicated that eliminating high levels of L. monocytogenes contamination from foods at the point of consumption would have a positive public health impact.

6.5 LIMITATIONS AND CAVEATS INCLUDING UNCERTAINTY AND VARIABILITY

6.5.1 Modelling

The risk characterization results are subject to uncertainty associated with a modelled representation of reality involving simplification of the relationships among prevalence, cell number, growth, consumption characteristics and the adverse response to consumption of some number of L. monocytogenes cells. However, modelling is appropriate to quantitatively describe uncertainty and variability related to all kinds of factors.

The risk characterizations attempt to provide estimates of the uncertainty and variability associated with each of the predicted levels of risk. This is one of the strengths of this type (stochastic) of modelling, but also leads to confusion in the interpretation of data. Furthermore, these estimates of uncertainty are themselves uncertain and dependent on the methods and assumptions used to make these calculations. This is often exacerbated by the limited data sets available describing the levels of L. monocytogenes and may overestimate the maximum extent of contamination and thus the risk associated with a specific food.

6.5.2 Prevalence and cell number

It was emphasized that the amount of quantitative data available on L. monocytogenes contamination was limited and restricted primarily to European foods.

Prevalence was determined by combining results available in the published literature, government surveillance reports and industry reports. Estimates of prevalence were subject to the variability and uncertainty associated with the published results and the methods used here to combine those results into a single distribution.

Cell number distributions were also derived from articles in the published literature. The estimated distributions are subject to uncertainty and variability associated with small sample sizes in many of those articles and assumptions that those separate data sets can be pooled into a single distribution. Moreover there are assumptions on the redistribution of pathogens in products when repackaged that contribute to uncertainty in the distribution of cell numbers.

The data used for prevalence and cell numbers may not reflect changes in certain commodities that have occurred in the food supply during the past ten years.

6.5.3 Consumption characteristics

The consumption characteristics used in the risk assessment were primarily those for Canada or the United States.

6.5.4 Modelling of dose-response

The R-values and their distributions were developed using epidemiological data on the current frequency of L. monocytogenes strain diversity that are observed with their associated virulence. If that distribution of virulence were to change (as reflected by new epidemiological data), the R-values would have to be recalculated. There is uncertainty associated with the form of the dose-response function used, and with the parameterization.

It should also be understood that the dose-response section of the hazard characterization is entirely a product of the shape of the distribution of predicted consumed doses in the FDA/USDA-FSIS exposure assessment. Further, since the underlying phenomenon of consumed doses is essentially unmeasurable, the validity of the dose-response model is dependent on the validity of the FDA/USDA-FSIS exposure assessment, which would be very difficult to externally validate. Note that the dependency is not limited to the parameter, r, but includes the choice of the best functional form (i.e., Exponential, beta-Poisson). Changes to the FDA/USDA-FSIS exposure assessment should lead directly to changes in the parameter, R. As a result it may not be possible to validate the exposure assessment and dose-response model as separate entities.

6.5.5 Modelling of microbial growth

Predictive modelling was used to model the growth of L. monocytogenes in RTE foods, between the point of retail and the point of consumption. Exposure assessment was based on distributions of the numbers of bacteria derived from those models. It is known that models may overestimate growth in food, as there are examples of foods in which maximum levels are much lower than those predicted by models. Reliance on such a model will result in an overestimation of the risk. An example is cold smoked fish. Another example is frankfurter where the occurrence and extent of growth on the surface is different to that in the core of product. Nevertheless, it has been observed that even in such foods, there were examples of very high levels e.g. 10¹¹ per serving (e.g. chocolate milk). Therefore, the approach taken was purposely conservative and all details can be found in the complete risk assessment document, which will be made available on the FAO/WHO webpages. There is uncertainty in the assumed distributions for the growth rates (derived from data sets available in the published literature), storage temperatures (from four published sources), and storage times.

6.6 GAPS IN THE DATA

This section identifies gaps for this and future risk assessments.

• L. monocytogenes prevalence in potential environmental sources, including i) agricultural environments, e.g. ground and well water; cultivated soil used for different crops or uncultivated soil for grazing pasture at different times of the year, silage, fresh and composted manure, farm equipment and farm workers; ii) aquatic environments, marine and freshwater where fish or shellfish are harvested including the effects of sewage or agricultural runoffs into water, fishing equipment, and commercial and recreational fishers.

• L. monocytogenes prevalence and cell numbers in production including i) primary production: animals, fish, and crops; ii) secondary production, e.g. initial preparation, cleaned carcasses, gutted and stored fish, shucked shellfish, washed produce.

• Prevalence and cell numbers of L. monocytogenes in finished packages of RTE foods from around the world on which risk assessment is being or will be undertaken.

• Information on the frequency of naturally occurring very high levels of contamination.

• Product formulation information, e.g. pH, water activity, humectants and preservatives (e.g. nitrite and organic acids, lactic acid bacteria) and their distributions to enable best estimates of microbial growth, survival or death.

• Better knowledge on the growth of L. monocytogenes in naturally contaminated food.

• Data for evaluating the validity of predictive models for L. monocytogenes in specific products, recognizing the effect of prior history of the culture and potential differences between naturally contaminated products and those deliberately inoculated in challenge tests.

• Identification of virulence markers for L. monocytogenes so that exposure to those strains can be assessed specifically.

• Identification of sources and levels of contamination and recontamination, at the point of processing, retail and the consumer are needed, with information on frequency, and microbial load transferred.

• Impact of microbiota, including spoilage organisms on the growth and survival of L. monocytogenes in RTE foods or their ingredients, and on the shelf-life of products.

• Retail and consumer handling practices, in particular, storage time and temperature and including more accurate measurements of home storage conditions including refrigeration temperatures by country or region.

• Specific RTE product consumption data for meal servings and frequency by specific populations of individuals, including in developing countries, and particularly for those who are immunocompromised or otherwise susceptible.

• Epidemiological data that distinguish serious or life threatening (usually systemic) vs. mild disease.

• Data on the degree of compliance to regulations established for control of L. monocytogenes cell numbers in food and the influence these regulations have on prevalence and cell numbers. This information will influence estimates of the efficacy of the control programme.

• Data on the capacity to which industry can control the prevalence and numbers of L. monocytogenes in foods and meet different tolerance levels (e.g. negative in 25 g, 0.1 g, 0.01g).

6.7 RESPONSE TO THE SPECIFIC RISK MANAGEMENT QUESTIONS POSED BY THE CODEX COMMITTEE ON FOOD HYGIENE

6.7.1 Evaluation of the risk from Listeria monocytogenes to susceptible population groups compared to the general population

This question was addressed by inspection of the epidemiological data on the number of cases in more susceptible subgroups. However, in taking this approach it was assumed that consumers in the different subgroups have similar consumption patterns.

If the risk for listeriosis is the same for more susceptible consumers as for the general, or healthy, population the proportion of cases in a specific subgroup should reflect its proportion of the total population. Conversely, if the risk for a subgroup is larger the actual number of cases in the subgroup is a reflection of the increased risk. This risk may be expressed as a relative risk^[4] based on the healthy population or, using the Exponential model derived in the hazard characterization section, as an R for a specific subgroup, e.g. R_elderly.

In the present analysis, epidemiological data from France (that do not include perinatal cases) and from the United States were used.

Under the assumption that the Exponential dose-response model is an accurate model to describe the relationship between dose and the probability of illness the relative risks (RR) presented for different subgroups may be used to derive dose-response models for specific susceptible subgroups. The Exponential dose-response model is described by:

P = 1 - e^-R*N

where P is probability of illness and N is the ingested dose. The R-value is a reflection of the host/microorganism interaction probability, and is the probability of the ingested organisms being individually capable of causing an infection to a specific consumer. The relative risk describes the ratio between the probabilities of illness for the subgroup and the healthy population and thus;

RR_subgroup = P_subgroup/P_healthy = 1 - e^(-Rsubgroup^*N) / 1 - e^(-Rhealthy^*N)
R_subgroup = [- ln(1-RR + RR*e^(-Rhealthy^*N))] / N

From this relationship R_subgroup can be estimated since the R-value for the healthy population (derived in the Hazard Characterization section) and the relative risks are known. Since P, the probability of illness and R are related by an Exponential function, the value of R_subgroup depends to some extent on both the dose and the magnitude of the relative risk. The influence of dose on the estimated R-values for the different subgroups was investigated by calculating R for a low (1 CFU) and a high dose (the assumed maximum doses). The influence of dose was most significant at higher doses, close to the assumed maximum dose, and when the probabilities of illness were larger than 0.01. In view of the overall uncertainty the variation with dose was considered insignificant and in Table 6.2 the average of the estimated R´s at the two doses are presented for the different subgroups.

It should be remembered that the R-values derived from calibration of the Exponential model to the exposure assessment results and the epidemiological data represents an implicit consideration of the virulence of L. monocytogenes strains, the vulnerability of the subgroups, the consumption and exposures etc..

The estimated R-value varied within a particular susceptibility subpopulation depending on assumed maximum dose. Thus for the most susceptible group (transplant patients), the estimated R-values varied from 5.8x10^-10 (log dose 7.5) to 2.3x10^-11 (log dose 10.5). In comparison, similar R-value estimates ranged from 2.23x10^-13 to 7.45x10^-15 in the healthy population.

Relative susceptibilities for individuals have been estimated from French epidemiology data using the sizes of these groups and the numbers of cases. A similar calculation for perinatal and elderly populations was made using data from the United States. The R-values were then calculated using the healthy population (less than 65 years of age and no known health conditions) for reference.

TABLE 6.2 Relative susceptibilities, R- value, and estimated log dose for immunocompromised and nonimmunocompromised populations

^a R-value 5.33 x 10^-11 from maximum dose at 10^8.5 used for reference population

6.7.2 Risk from Listeria monocytogenes in food when the number of organisms does not exceed a specified level at the point of consumption.

There are a number of different means of answering this question using various degrees of sophistication and assumptions e.g. the extent of deviation from a criterion that could be anticipated. However, it is worthwhile considering the simplest form of the answer to this question, that is the best case scenario of what could be anticipated if different criteria were successfully implemented. This best case scenario can be easily estimated by using the dose-response relationship derived in the hazard characterization in conjunction with a “global contamination distribution”. As an example of this type of analysis, the distribution for the total number of contaminated servings of food from the FDA/USDA-FSIS risk assessment was used in conjunction with the R-value (5.85 x 10^-12) from the current risk assessment assuming a maximum dose of 10^7.5 CFU/serving for the susceptible population. This was the most conservative dose-response curve used in the current risk assessment. The total predicted number of cases per year in this example was 2130 (susceptible individuals).

TABLE 6.3 Baseline number of cases predicted by the dose-response model.

* the number of cases is predicted based on the dose and the number of servings containing that dose

Using the serving numbers listed above, the upper range of cell numbers were limited to dose values equal to or less than values between 1.5 log and 4.5 log CFU per serving. Then assuming 100% realisation of these limits, the number of cases that would be anticipated was calculated for seven scenarios (Table 6.4). In the scenario calculations the number of servings at dose values higher than that of the criterion being considered were added to the highest dose level. Thus when a dose limit of 4.5 log was considered, the number of servings from the baseline data (Table 6.3) for 5.5, 6.5, and 7.5 log were added to the number of servings for 4.5 log. It is important to note that these values are in terms of CFU per serving. To calculate what this would be in terms of CFU per gram of food, the values in the table below would have to be divided by the serving size in terms of grams.

TABLE 6.4 The number of cases predicted if various criteria for CFU/serving could be realized at 100% effectiveness.

^a. This depicts the current number of predicted cases based on the observed distribution of L. monocytogenes depicted in Table 6.3.

It is obvious from the table provided above that eliminating the higher dose levels at the time of consumption has a large impact on the number of predicted cases, i.e., an approximate 99% reduction in cases could be potentially realised by implementing even the highest criterion. However, it is important to note that this is based on cell numbers at time of consumption. Consideration of cell numbers at time of retail would have to be corrected to take into account the potential increases in L. monocytogenes that would occur as a result of growth in those foods that will support replication of L. monocytogenes. Likewise, this does not take into account the reality that there would likely be some incidence where the criteria analysed above would not be realized. Consideration of these factors requires a more rigorous evaluation of the risk posed, using more sophisticated modelling techniques. This advanced modelling was not completed in time for the expert consultation, but is anticipated shortly.

6.7.3 Risk from Listeria monocytogenes in foods that support growth and do not support growth under specific storage conditions and shelf-life.

The question concerning the relative risk associated with foods that do and do not support growth can also be considered broadly by using the example above. The key consideration is whether a correction factor needs to be applied when comparing levels at time of retail versus at time of consumption. For foods that support growth, increases in L. monocytogenes cell numbers between retail and consumption would have to be assumed and there is a significant likelihood that the hypothetical criteria analysed above would be exceeded. However, this would not be the case for foods that do not support growth. Thus, for foods that do not support growth of L. monocytogenes, the predicted number of cases in relation to maximum dose level at retail would be the same as those depicted above for doses at time of consumption. Again, more rigorous modelling of other factors that could influence the differential in risk of severe listeriosis between foods that do and do not support the growth of L. monocytogenes are currently underway and the results of that activity are expected shortly. However, these are not likely to alter the large differential in risk between food that do and do not support the growth of L. monocytogenes to high levels that is suggested by the current “best-case” analysis.

6.8 CONCLUSIONS AND RECOMMENDATIONS

The L. monocytogenes risk assessment represents a major achievement and presents novel approaches to solve the problems of dose-response and exposure assessment modelling for L. monocytogenes. The expert consultation recognized that this risk assessment will provide a very valuable contribution not only relative to the process in CCFH, but also in general as a resource document for FAO and WHO member countries, academia and other interested parties.

The expert consultation acknowledged the enormous amount of work done but suggested some reorganization of the risk assessment document and an editorial review to make reading easier. It also suggested that the exposure assessment provides further explanation of the different approaches used therein. Furthermore, the expert consultation strongly recommended that the risk assessment, when completed and edited by the drafting group, be sent for an international peer review.

The consultation identified problems related to the statistical basis applied in the exposure assessment of L. monocytogenes, specifically in relation to the representation of events with very low probability that could have a very large impact on human health. These problems may limit the ability of models to produce accurate absolute numerical estimates of listeriosis. This is an issue with general implications for modelling in microbiological risk assessment. It was also concluded that there is a wider array of probabilistic techniques available from other disciplines to deal with this issue. In relation to this, the consultation suggested to FAO and WHO that further work be urgently initiated in order to reach agreement on the modelling approach to be taken to finalize the L. monocytogenes exposure assessment and risk characterization.

Risk characterization was done by combining the dose-response models for the general population and the more susceptible population with the estimated exposure for six ready-to-eat foods. The expert consultation asked that the drafting group further expand the risk characterization section of the risk assessment document. The drafting group proposed that the risk characterization would be developed for each of the six foods. An additional section to address the questions on the risk from L. monocytogenes in foods that do and do not support its growth and the risk from L. monocytogenes in foods when the number of organisms does not exceed a specified number at the point of consumption as posed by CCFH would also be included.

Quantitative data on levels of L. monocytogenes contamination of foods and prevalence of listeriosis should be obtained in various regions of the world. Similarly, this information should be developed to determine if seasonality and/or regional differences exist and the influence of climate and season in different regions in the world.

The output of any risk assessment depends greatly on the tails of distribution^[6], notably the number of bacterial cells at the time of consumption. Any management options that reduce the uncertainty associated with these tails of distribution would be helpful.

The expert consultation agreed with the statement of CCFH (ALINORM 01/13A) that one of the important future work issues for the joint FAO/WHO programme of activities on microbiological risk assessment would be to estimate the change in risk of listeriosis from food likely to occur when specific interventions are introduced.. In doing so it would be important to include in the modelling exercise data from the parts of the production chain going back to and including processing and in some cases the steps before processing.