6. Risk assessment of Vibrio spp. in seafood

6.1 Summary of the risk assessments

6.1.1 Introduction: Vibrio spp. in seafood

Vibrio spp. are Gram-negative, facultatively anaerobic motile curved rod-shaped bacteria with a single polar flagellum. The genus contains at least twelve species pathogenic to humans, eight of which can cause or are associated with food-borne illness (Table 6.1). The majority of food-borne illness is caused by Vibrio cholerae, Vibrio parahaemolyticus or Vibrio vulnificus (Oliver and Kaper, 1997^[27]; Dalsgaard, 1998^[28]). Most countries have guidelines for detecting V. parahaemolyticus and V. cholerae O1 and O139 in seafood, whereas few have guidelines for V. vulnificus. Accordingly, routine microbiological analysis of seafood includes testing for V. parahaemolyticus and V. cholerae O1 and O139, but seldom for V. vulnificus.

Some species are primarily associated with gastrointestinal illness (V. cholerae and V. parahaemolyticus) while others can cause non-intestinal illness, such as septicaemia (V. vulnificus). In tropical and temperate regions, disease-causing species of Vibrio occur naturally in marine, coastal and estuarine (brackish) environments and are most abundant in estuaries. Pathogenic vibrios, in particular V. cholerae, can also be recovered from freshwater reaches of estuaries (Desmarchelier, 1997^[29]), where it can also be introduced by faecal contamination. The occurrence of these bacteria does not generally correlate with numbers of faecal coliforms and depuration of shellfish may not reduce their numbers. However, a positive correlation between faecal contamination and levels of V. cholerae may be found in areas experiencing cholera outbreaks. A positive correlation between water temperature and the numbers of vibrios has also been shown in several parts of the world. Further, according to data from the United States of America and Denmark, there is a positive correlation between water temperature and both the number of human pathogenic vibrios isolated and the number of reported human infections. This correlation is particularly striking for V. parahaemolyticus and V. vulnificus (Dalsgaard et al., 1996^[30]).

The objective of the work was to undertake a risk assessment of Vibrio spp. in seafood products that have the most impact on public health and/or international trade. Three species, V. parahaemolyticus, V. vulnificus and choleragenic V. cholerae (toxigenic V. cholerae O1 and O139 that may cause cholera) were identified as being responsible for most illnesses caused by Vibrio spp. The approach taken was to quantify those illnesses caused by Vibrio spp. in different countries following the consumption of a range of seafoods and this report documents the results of that approach.

6.1.2 Scope

The risk assessment work was undertaken on the following pathogen-commodity combinations:

• Vibrio parahaemolyticus in raw oysters consumed in Japan, New Zealand, Canada, Australia and the United States of America.

• Vibrio parahaemolyticus in finfish consumed raw.

• Vibrio parahaemolyticus in bloody clams consumed in Thailand.

• Vibrio vulnificus in raw oysters consumed in the United States of America.

• Choleragenic Vibrio cholerae in warm-water shrimp in international trade.

6.1.3 Vibrio parahaemolyticus in raw oysters consumed in Japan, New Zealand, Canada, Australia and the United States of America

6.1.3.1 Introduction

FAO and WHO aim to make optimal use of existing risk assessments in their MRA activities. As there have been large outbreaks of illness due to V. parahaemolyticus in North America following consumption of raw oysters, the United States Food and Drug Administration (USFDA) commissioned a quantitative risk assessment on the "Public Health Impact of Vibrio parahaemolyticus in Raw Molluscan Shellfish" (FDA-VPRA), one output of which was the development of a risk model. A critical component of the model was water temperature. Since high water temperatures are a factor in several countries with significant oyster industries, FAO and WHO decided to undertake a risk assessment on consumption of raw oysters in a number of different countries. As well as generating an estimate of the number of annual illnesses, a further aim was to assess the potential of the model developed in the United States of America to predict oyster-borne V. parahaemolyticus illness from oysters grown in different regions and using different production systems.

TABLE 6.1: Vibrio spp. which cause, or are associated with, human infections (after Dalsgaard, 1998^[31])

* The symbol (+) refers to the relative frequency of each organism in clinical specimens and (-) indicated that the organism was not found

** The ability of V. vulnificus to cause gastro-intestinal disease remains to be confirmed

6.1.3.2 Scope

The risk assessment covers consumption of raw oysters in five countries: New Zealand, Japan, Canada, Australia and the United States of America.

6.1.3.3 Hazard identification

V. parahaemolyticus has been recognized as a major cause of seafood-borne gastroenteritis in Japan (Twedt, 1989^[32]; Ministry of Health, Labour and Welfare, Japan, 2000^[33]) and other Asian countries. By contrast, in most countries outside of Asia, the reported incidence appears to be low, perhaps reflecting a different mode of seafood consumption. Gastroenteritis caused by this organism is almost exclusively associated with seafood consumed raw or inadequately cooked, or contaminated after cooking. In the United States of America, prior to 1997, illness was most commonly associated with crabs, oysters, shrimp and lobster (Twedt, 1989^[32]; Oliver and Kaper, 1997^[34]). Four V. parahaemolyticus outbreaks associated with the consumption of raw oysters were reported in the United States of America in 1997 and 1998 (DePaola et al., 2000^[35]). A new V. parahaemolyticus clone of O3:K6 serotype emerged in Calcutta in 1996. It has spread throughout Asia and to the United States of America elevating the status of V. parahaemolyticus to pandemic (Matsumoto et al., 2000^[36]). In Australia, in 1990 and 1992, there were two outbreaks of gastroenteritis caused by V. parahaemolyticus in chilled, cooked shrimps imported from Indonesia (Kraa, 1995^[37]) and there was also a death in 1992 associated with the consumption of oysters.

6.1.3.4 Hazard characterization

This section focuses on evaluating the nature of adverse health effects associated with V. parahaemolyticus in seafood and how to quantitatively assess the relationship between the magnitude of the food-borne exposure and the likelihood of adverse effects occurring. It included the elaboration of a dose-response curve. Infection by V. parahaemolyticus is characterized by an acute gastroenteritis. Therefore, the end-point of the dose-response curve was defined as gastroenteritis.

A review of the literature was undertaken to identify and characterize the infectivity and genetic factors of V. parahaemolyticus, which has both pathogenic and non-pathogenic forms based on the presence of specific virulence genes: tdh (thermostable direct haemolysin gene) and trh (TDH-related haemolysin gene). Relevant factors with respect to the host and food matrix have been identified and where data are available may be incorporated into the model.

The determination of the dose-response relationship was based on the best available data. Human volunteer studies were available for the construction of the dose-response curve for V. parahaemolyticus, however, these studies characterize the dose-response relationship for V. parahaemolyticus administered with a pH-neutralizing buffer rather than with a food matrix. The data were analysed using curve-fitting routines to find a best fit for the Beta-Poisson dose-response curve. Because of the limited amount of data available from human volunteer studies the resulting dose-response relationship is uncertain. This uncertainty was accounted for by representing the dose-response relationship in the form of a family of plausible data-derived dose-response curves determined using resampling techniques. Figure 6.1 shows the most probable dose-response curve for V. parahaemolyticus; however, the family of curves representing uncertainty that surrounds the curve is not shown.

FIGURE 6.1: Beta-Poisson dose-response curve for V. parahaemolyticus (endpoint modelled is gastrointestinal illness)

- Beta-Poisson
¨ Sanyal and Sen (1974)^[38]
Aiso and Fujiwara (1963)^[39]
· Takikawa (1958)^[40]

6.1.3.5 Exposure assessment

In the United States of America, during 1997 and 1998, there were more than 700 cases of illness due to V. parahaemolyticus, the majority of which were associated with the consumption of raw oysters. In two of the 1998 outbreaks a serotype of V. parahaemolyticus previously reported only in Asia, O3:K6, emerged as a principal cause of illness for the first time. It was suggested that warmer than usual water temperatures were responsible for the outbreaks.

Temperature profiles in the oyster industries of Japan, New Zealand, Australia, Canada and the United States of America were obtained, together with consumption levels of oysters and bacterial levels of V. parahaemolyticus in the oysters. The objectives were to quantify the exposure of consumers to pathogenic V. parahaemolyticus from consumption of raw oysters in these countries.

The FDA-VPRA model was used as the base to accommodate data inputs from other countries. This model incorporates all phases in the harvest - post-harvest - consumption continuum in three modules (Figures 6.2-6.4).

Figure 6.2 shows a conceptual model for the harvest module. Water temperature is the driving input with regard to the initial numbers of V. parahaemolyticus in oysters. In the way the analysis is constructed regional and seasonal temperature variations allow for a multi-year analysis that can account for long-term temperature trends. Water salinity is shown in dotted lines to indicate that for some model applications salinity may be another important input.

FIGURE 6.2: Harvest module for exposure assessment of V. parahaemolyticus in oysters.
(Vp = Vibrio parahaemolyticus)

Figure 6.3 shows the conceptual model for post-harvesting practices. The post-harvest module determines the role of post-harvest processing and handling on the numbers of pathogenic V. parahaemolyticus at consumption. The bubble denoting "V.p/g at harvest" is the output of the harvest model shown in Figure 6.2. Inputs on the time the oysters are out of the water and the air temperature are used to predict growth of V. parahaemolyticus in the oysters. Growth continues as the oysters are cooled but at a different rate. V. parahaemolyticus levels decrease during storage and the storage time is therefore an input time that affects V. parahaemolyticus numbers.

Figure 6.4 represents the consumption module. The bubble denoting "path Vp/g (numbers) at consumption" is the output of the post-harvest module. This number is multiplied by the number of oysters per serving and the weight of the oysters to yield the ingested dose. This ingested dose is used in the dose-response to calculate the risk of illness associated with the consumption of one oyster meal.

FIGURE 6.3: Post-harvest module for exposure assessment of V. parahaemolyticus in oysters.
(Vp = pathogenic Vibrio parahaemolyticus)

FIGURE 6.4: Consumption module for exposure assessment of V. parahaemolyticus in oysters
(Vp = pathogenic Vibrio parahaemolyticus)

6.1.3.6 Risk characterization

The data from the five countries were analysed for incorporation into the risk assessment model. The risk assessment model was modified to allow for a monthly analysis of data from Japan, Australia, and New Zealand. The analysis for Canada and the United States of America was done on a seasonal basis. Using the Japanese data only one simulation, consisting of 100 000 iterations, was undertaken, as multiple year temperature data were not available. Thirteen simulations, consisting of 10 000 iterations were undertaken for Australia reflecting the availability of 13 years of data. As only one year's data were available for New Zealand, one simulation, consisting of 100 000 iterations, was undertaken. For Canada 1 000 simulations, consisting of 10 000 iterations, based on United States of America Pacific Northwest data, was undertaken. The analysis for the United States of America consisted of 10 000 iterations for seasons in four regions.

6.1.3.7 Key findings

Introduction

The complete data sets required to test the applicability of the model to harvesting waters of countries other than the United States of America were not available. In particular tdh⁺ and trh⁺ data were lacking and in such cases United States of America data was used as a surrogate to allow for testing of the model.

Japan

Based on the available data set^[41], the preliminary predictions of illness are shown in Table 6.2. The model predicted low levels of illness for November to April. The model was not run for the months of May to October as oysters for raw consumption are not harvested during this period^[42].

It was difficult to compare this with epidemiological data for V. parahaemolyticus-related oyster illnesses in Japan for a number of reasons. The Japanese surveillance system focuses mainly on outbreaks of food-borne disease and therefore the number of laboratory confirmed reported illnesses may not include sporadic cases or diffuse outbreaks and the extent of under-reporting is not known (K Osaka, personal communication, 2002). In addition the food source of the illness may not always be identified. However, in cases where oysters have been identified as the food source causing illness, large variability in the annual number of V. parahaemolyticus-related oyster illnesses has been noted over the last five years^[43]. It is also worth noting that the model estimation is based on data (e.g. air and water temperature, salinity) available from only one of the major harvesting areas and therefore does not necessarily capture the situation in the different oyster growing areas in Japan.

TABLE 6.2: Preliminary predictions of V. parahaemolyticus illness in Japan associated with oyster consumption

Australia

Based on the available data set, the preliminary predictions of illness are shown in Table 6.3. The model predicted more illnesses than the number of reported cases (J. Sumner, personal communication, 2002). The application of United States of America surrogate data to a different species of oyster, specifically the Sydney rock oyster, may have a role in the overestimation of risk.

TABLE 6.3: Preliminary predictions of V. parahaemolyticus illness in Australia associated with oyster consumption

New Zealand

The model predicted more illnesses than the number of reported cases (D.J. McCoubrey, personal communication, 2002) (Table 6.4). As extensive use of surrogate data from the United States of America was necessary as inputs for some of the parameters required to run the model, the true risk may be much lower than that predicted.

TABLE 6.4: Preliminary predictions of V. parahaemolyticus illness in New Zealand associated with oyster consumption

Canada

The preliminary results (Table 6.5) indicate that model predicted cases of illness that are relatively close to the number of reported cases^[44], ^[45]). The proximity of the Canadian harvesting waters to one of the regions of the United States of America that was modelled allows greater confidence in these predictions. It should be noted that the model did not consider the mitigation to cool oysters immediately after harvest that was introduced in the Canadian oyster industry in 2000, as the data used was collected prior to the implementation of this measure.

TABLE 6.5: Preliminary predictions of V. parahaemolyticus illness in Canada associated with oyster consumption

United States of America

The predicted numbers of illness for the United States of America as shown in Table 6.6. In this case the dose-response relationship was adjusted to take into account the estimation that the actual number of cases of V. parahaemolyticus illness in the United States of America exceeds the reported number of cases by a factor of 20 to 1 (Mead et al, 1999^[46]). However, it was acknowledged that the predicted number of illnesses associated with oyster consumption is probably still an overestimation as the study of Mead et al (1999)^[46] that estimated the degree of under-reporting used statistics on the annual incidence of V. parahaemolyticus illness and not only those for which oyster was the vehicle of transmission. Evidence for validation of the model comes from the observed agreement between model predictions of V. parahaemolyticus numbers with observed harvesting and retail numbers of V. parahaemolyticus.

Table 6.6: Preliminary predictions of V. parahaemolyticus illness in the United States of America associated with oyster consumption

6.1.3.8 Limitations and Caveats

It was difficult to critically evaluate the performance of the model in harvesting waters outside of the United States of America. In many cases the raw data on which to adapt the model to local conditions were not available because:

• The data had not been accumulated because of expense or lack of need for the data.

• The data were in summary form and therefore not amenable to reanalysis.

• The data were difficult to retrieve from stored printed form and to convert into electronic format.

• The methodology used to generate the data in countries other than the United States of America was not comparable to that used in generating the data that served as a basis for the establishment of the model parameters.

Where limited data were available, judgement was needed on how to adapt this data for incorporation into the model. However, there is currently no guidance on this issue or even whether adaptation of data is desirable.

Validation of model predictions by epidemiological observations was complicated by the fact that the relationship between observed and predicted illness is generally unknown. In the United States of America the ratio of predicted to observed illness has been estimated to be 20 to 1 (Mead et al, 1999^[47]). This relationship has not been estimated for other countries and where it may differ from that in the United States of America.

Limited data has the effect of reducing the variance of the model’s prediction of risk. The reduced variance of predictions may be misinterpreted as greater confidence in a predicted risk than a predicted risk with wider variance that is based upon more extensive data.

The species of oyster may have a profound effect on the model and further research is needed to develop the oyster-V. parahaemolyticus ecology knowledge base.

Accurate model predictions may require adapting the model to parameters that are critical to harvesting areas and are different from those in the United States of America where the model was developed. For example, salinity may be a critical element in the control of V. parahaemolyticus in New Zealand and Australia. The model will be elaborated to test whether the addition of this parameter can improve model predictions.

The use of surrogate data, particularly in relation to the occurrence of tdh+ and trh+ strains, may limit the utility of the model in predicting illnesses from V. parahaemolyticus contaminated oysters harvested from waters other than those of the United States of America. Obtaining these data may be difficult, especially when few illnesses associated with oysters from certain harvesting areas lead to the fact that data (required by the model) have not been collected.

6.1.3.9 Data gaps

The risk assessment identified a number of data gaps which limited in particular the application of the model developed in the United States of America to oysters harvested in different regions of the world. Some of the main data and knowledge gaps include:

• Multi-year temperature data for seawater and air at harvesting areas.

• Data for determining the relationship between temperature and V. parahaemolyticus numbers in some harvesting areas.

• Characterization of harvesting activities in some areas.

• Information on the role of oyster ecology in altering the model parameters.

• Data for tdh⁺ and trh⁺ prevalence of the total V. parahaemolyticus in some national harvesting waters.

• Dose-response information, the lack of which results in uncertainty in the dose-response relationship and adds substantial variance to predictions of illness.

• Methodology for estimating the ratio of reported to total cases of illness. Appropriate methodology needs to be officially developed and applied in countries that wish to compare the number of illnesses predicted by risk assessment to the number of reported & recorded cases of illness.

• Data sources that can indicate whether or not the model is succeeding. Validation of the risk assessment model should be attempted at as many intermediate stages as is practical.

6.1.4 Vibrio parahaemolyticus in bloody clams

6.1.4.1 Introduction

V. parahaemolyticus has been recognized as a major cause of food-borne gastroenteritis in Japan and other Asian countries. However, the data available on V. parahaemolyticus and seafood, other than oysters, that was also suitable for the quantitative risk assessment were very limited world-wide.

A small-scale study was undertaken, based on data collected in the Songkla Province of southern Thailand. A joint Thai-Japanese team carried out the study on the prevalence and concentration of V. parahaemolyticus in non-oyster seafood. All strains of V. parahaemolyticus and pathogenic strains which have the tdh and/or the trh gene, and thus have the potential to produce TDH and/or TRH, were enumerated in the data collection process. No foodborne disease surveillance data for this area were available. However, the preliminary study showed that the strains isolated from clinical specimens in this area were identical, in terms of serotype and molecular genetics, with the strains isolated from the shellfish harvested in the area rather than other seafood such as fish and shrimps. Therefore, a popular bivalve in Thailand, the bloody clam (Anadara granosa), was chosen as the target seafood in this risk assessment. This shellfish is also traded in the Southeast Asian region.

6.1.4.2 Scope

Using state of the art techniques, the data necessary for developing the quantitative risk assessment were collected and a model was elaborated in a developing country situation where there was a lack of quantitative data.

6.1.4.3 Hazard identification

V. parahaemolyticus is considered to be an important cause of seafood-borne disease in Thailand. A survey of clinical specimens obtained from patients with diarrhoea resulted in the isolation of 294 pathogenic strains from 317 cases that were confirmed positive for V. parahaemolyticus (Table 6.7). Several seafood items were also tested for pathogenic strains of V. parahaemolyticus and in this preliminary study shellfish were the most commonly contaminated among the samples tested (Table 6.7). The profile of strains (serotype and possession of tdh/trh gene) isolated from clinical samples were consistent with that of the strains isolated from shellfish (Table 6.7). Therefore, shellfish were considered as an important source of V. parahaemolyticus infection.

6.1.4.4 Hazard characterization

The dose-response model used in the hazard characterization of V. parahaemolyticus in oysters (see section 6.1.3.4) was also used in the hazard characterization of V. parahaemolyticus in bloody clams.

6.1.4.5 Exposure assessment

The exposure assessment was divided into four stages; harvest, retail, cooking and consumption as shown in Figure 6.5.

Data on the prevalence and numbers of V. parahaemolyticus in clams was collected at each step of the exposure pathway. A single lot of the clams was taken from a boat shortly after landing at the harvest site. Following initial sampling ("Harvest" stage), the remaining clams were transported to the local market area, which was located close to the laboratory. A sample of clams were examined at this point to represent the "Retail" stage. Thereafter, the clams were maintained outside of the laboratory for a period of time to simulate the transportation step; these were subsequently examined in the laboratory. Typically, the clams are cooked in the home by boiling briefly (in some cases with insufficient heating). The "Cook (Boiling)" stage was simulated in the laboratory and the clams were tested thereafter. To obtain consumption data, local people were interviewed on the frequency and quantity of bloody clams consumed.

TABLE 6.7: Results of the study on isolation of V. parahaemolyticus from seafood and the most common strain profiles of isolates from clinical specimens and seafood

*Samples were examined over a four year period from 1998 to 2001. During the first year of the study period pathogenic V. parahaemolyticus were only isolated from shellfish. Therefore, during the subsequent years of the survey, efforts focussed mainly on detection of pathogenic V. parahaemolyticus in shellfish samples.

** V. parahaemolyticus was isolated from 317 diarrhoea specimens out of a total of 11 375 samples that were examined during a survey sporadic cases of illness with diarrhoea in 1999. Specimens came from different patients in two big hospitals in the province. Of the 317 cases confirmed positive for V. parahaemolyticus, 294 of these were confirmed to be pathogenic strains of V. parahaemolyticus.

The laboratory analysis undertaken included V. parahaemolyticus toxR gene sequencing to identify V. parahaemolyticus and group-specific PCR (GS-PCR) to detect pandemic strains carrying the tdh gene. The prevalence of total and tdh⁺ or trh⁺ V. parahaemolyticus strains was examined at harvest and retail, and after cooking.

The total numbers of V. parahaemolyticus from culture and PCR methods were assumed to have a lognormal distribution. The prevalence of total V. parahaemolyticus after boiling was estimated using the laboratory generated data. The prevalence of tdh⁺ and trh⁺ strains that possibly remain in clams after boiling was estimated by assuming that the ratio of the prevalence of total and virulent strains before heating was maintained after boiling. The same assumption was made with regard to numbers.

FIGURE 6.5: Schematic representation of the exposure model developed for the risk assessment of V. parahaemolyticus in bloody clams.

Comparison was made between the predicted and observed values of total V. parahaemolyticus numbers during transportation from harvest to the retail stage, in order to determine whether the increase in numbers could be analysed or predicted using an equation developed in the FDA-VPRA.

Although bloody clams are a popular seafood in this region there were no available data on their consumption. Therefore a small preliminary consumption survey was undertaken. Fourteen people (students and workers) at the university were selected for interview because of their accessibility. They were interviewed on how frequently they ate clams at home and how many they ate at one meal.

6.1.4.6 Risk characterization

The output of the exposure assessment feeds into the hazard characterization to produce the risk characterization output. The probability of getting ill following consumption of a single serving of clams was estimated for a defined population (i.e. people who were interviewed) by using the "dose" calculated in the exposure assessment and the dose-response equation. The probability of getting ill per year was further estimated by multiplying the frequency of clam consumption per year. The consumption data for bloody clams were used for estimating the risk of ingesting pathogenic strains of V. parahaemolyticus.

6.1.4.7 Key findings

1. The total number of V. parahaemolyticus was estimated as 6.5/clam, with a standard deviation of 2.2/clam, at harvest, and 7.8/clam, with a standard deviation of 2.0/clam, at retail.

2. After boiling, V. parahaemolyticus was detected in only one and two out of 32 samples by PCR and culture methods, respectively. Pathogenic strains were not isolated from any of the boiled samples.

3. Using the data generated from culture methods, the mean probability of illness per year due to clam consumption was estimated to be 9.18E-10 per person (approx. 1 person per 1 000 000 000 people becomes ill per year) and maximum probability was 9.34E-6 (approx. 1 person per 100 000 people becomes ill per year).

4. The observed growth rate of V. parahaemolyticus in bloody clams was found to be half the rate of growth predicted by the FDA-VPRA V. parahaemolyticus growth rate model in oysters.

5. Although time and resources were limited and there was a lack of quantitative data, this study indicated that, even when such obstacles exist, progress can still be made on data generation and risk assessment modelling.

6.1.4.8 Limitations and Caveats

The link between human illness and consumption of bloody clams was based on detection of strains of equivalent serotype and molecular genetics in both clinical samples and bivalve samples. There were no data from outbreak investigations or case control studies of sporadic cases to confirm this link or to prove that illness was indeed caused by foodborne transmission. Additional data is required to strengthen this linkage and this should ideally be included in the risk profile that is undertaken before the risk assessment is commissioned.

The results are restricted to a single food item, and the sample size may not be sufficiently large. Therefore, the data presented in table 6.7 should be interpreted with caution. Furthermore, the study on survival of V. parahaemolyticus from harvest to consumption was carried out for only a three month period in one specific area in Thailand. More data are needed for other months and other areas.

Because the cooking (by boiling) module was developed based on experimental data using fixed time/temperature values within a very limited range, scenario analysis with different time/temperature combinations are impossible. Also the consumption survey was carried out on a small group of people working within the same environment and therefore may not be representative of the region as a whole.

The cross-contamination model was not applied in this risk-assessment, because of lack of data and appropriate models for cross-contamination. Due to insufficient epidemiological data, model validation could not be undertaken.

6.1.4.9 Gaps in the data

To improve the risk assessment, the following data will be needed.

• Quantitative data on V. parahaemolyticus in bloody clams and other shellfish in various combinations of water temperature and salinity.

• The proportion of virulent strains in various shellfish, areas and seasons.

• The differences in sensitivity between virulent strain and non-virulent strain to heating and other mitigation steps.

• Data from a case control study or an outbreak investigation to strengthen the linkage between consumption of bloody clams and V. parahaemolyticus illness in humans.

6.1.5 Vibrio parahaemolyticus in finfish

6.1.5.1 Introduction

V. parahaemolyticus is a leading cause of seafood-borne illness in Japan and other Asian countries Several reports exist on the high prevalence of the organism in a variety of seafoods, in particular finfish, lobster and shrimp. Outbreaks due to V. parahaemolyticus associated with fish and shellfish other than oysters have been reported in some countries including the United States of America, Thailand, China (Taiwan) and Spain. With the globalization of Japanese cuisine and the increased practice of eating raw fish and shellfish, there is an increased possibility of V. parahaemolyticus infection as a result of consumption of these foods. A risk assessment of V. parahaemolyticus in finfish could provide useful information for reducing this risk.

An exposure assessment document was prepared and presented to an expert consultation^[48] in 2001. Although the drafting group had decided not to include this part in the final report due to the lack of quantitative data, it was noted that, although not a complete quantitative risk assessment, it still includes information that may be important for many countries and therefore should be recorded and available in the public domain.

This work could currently be described as a qualitative (descriptive) risk assessment. An effort to collect quantitative data on total V. parahaemolyticus in finfish, as well as data on virulent strains, is not yet completed. However, should it be possible to collect the necessary quantitative data a revised document, incorporating such data will be prepared.

6.1.5.2 Scope

This work focused on describing the possible contamination of finfish by V. parahaemolyticus from harvest to consumption.

6.1.5.3 Hazard identification

Published data on the prevalence and concentration of V. parahaemolyticus in finfish and other seafood were collected and collated. Literature reviews were also conducted through Medline and other resources on the world wide web.

6.1.5.4 Hazard characterization

The dose-response model used in the hazard characterization of V. parahaemolyticus in oysters (see section 6.1.3.4) was also considered to be applicable in the case of V. parahaemolyticus in finfish.

6.1.5.5 Exposure assessment

The pathway from pre-harvest to consumption was divided into four stages; pre-harvest, harvest, post-harvest and consumption. It includes a descriptive explanation of the possible risks of V. parahaemolyticus contamination at each stage. The possibility of proliferation/reduction of V. parahaemolyticus in each stage was considered through a qualitative description of the data collected.

6.1.5.6 Risk characterization

Because insufficient data were available to bring the assessment forward, no further work was undertaken.

6.1.5.7 Key findings

1. The prevalence and numbers of V. parahaemolyticus in seawater are influenced by seawater temperature and salinity. However, there may be other influencing factors such as plankton and tides.

2. Many species of finfish could be contaminated with V. parahaemolyticus though the prevalence and number of V. parahaemolyticus present vary with species. Differences in prevalence and numbers seemed to be associated with the species and their habitat (e.g. coastal or deep-sea).

3. Coastal seawaters used at landing docks and at markets were shown to be highly contaminated with V. parahaemolyticus. Therefore, the post-harvest stage may be of particular importance with regard to contamination of finfish.

4. This conceptual modelling approach could be appropriate for determining the potential effectiveness of mitigation strategies such as the use of chlorinated water and thermal processing.

5. The fluctuation of time and temperature during transportation and storage may be less important for finfish than raw oysters as V. parahaemolyticus was shown not to proliferate significantly on finfish samples up four hours at 25^oC.

6. Washing the visceral cavity after evisceration of the intestine reduced the numbers of V. parahaemolyticus on the fish fillet compared to eviscerated fish in which the visceral cavity had not been washed.

7. Food preparation in the home, including the time prior to washing out the visceral cavity, was identified as an important step in relation to cross-contamination and reducing the numbers of V. parahaemolyticus.

6.1.5.8 Limitations and Caveats

This is a qualitative (descriptive) risk assessment and quantitative data on the prevalence and concentration of V. parahaemolyticus in targeted seafoods are needed to undertake quantitative risk assessment.

6.1.5.9 Gaps in the data

The lack of quantitative data prevented the completion of this risk assessment. Primarily data in the following areas are needed.

• Number and proportion of pathogenic V. parahaemolyticus cells in various species of finfish.

• Frequency of consumption and quantity of raw fish consumed.

• Transportation practices (time and temperature).

6.1.6 Vibrio vulnificus in raw oysters

6.1.6.1 Introduction

The general approach to undertaking this assessment and many of the parameters were adopted from the FDA-VPRA and the FAO/WHO V. parahaemolyticus risk assessment, which are the only available quantitative risk assessments for a Vibrio spp. in raw oysters. Due to the lack of appropriate data from outside of the United States of America for many of the model inputs this assessment relies almost totally on data from this country. The approach for determining dose-response used exposure and illness frequency. For this reason some elements of hazard characterization were included in the exposure assessment.

The choice of the United States of America data was intended only to provide an example on how to apply the exposure model to a different national situation. This model could be further tested and modified when appropriate data from other countries or situations become available.

6.1.6.2 Scope

The main objective of this risk assessment was to determine the usefulness of adapting the FDA-VPRA model to assess the risk from V. vulnificus associated with the consumption of raw oysters. In addition it aims to identify the most appropriate data as well as the data gaps and limitations for modelling V. vulnificus in oysters, conduct a risk characterization of V. vulnificus in raw oysters using available data and evaluate targeted mitigation levels aimed at reducing the risk of V. vulnificus illness

6.1.6.3 Hazard Identification

V. vulnificus has been associated with primary septicaemia in individuals with chronic pre-existing conditions, following consumption of raw bivalves. This is a serious, often fatal, disease. In the United States of America, it carries the highest death rate of any food-borne disease agent (Mead et al, 1999^[49]). To date, V. vulnificus seafood-associated disease has almost exclusively been associated with oysters (Dalsgaard et al, 2001^[50]; Oliver and Kaper, 1997^[51]). In addition to the primary septicaemia that follows ingestion, V. vulnificus is known to infect wounds of otherwise healthy individuals, although the majority of patients with serious wound infections have an underlying disease. Such wound infections occur most often as a result of contamination of pre-existing wounds with seawater or after contact with fish or shellfish. V. vulnificus has in a few cases been isolated from patients with gastrointestinal disease, however, its role as a primary cause of gastrointestinal disease remains to be determined. Recently, cases of primary septicaemia associated with V. vulnificus infections seem to have been related to consumption of a variety of raw seafood products in Korea and Japan (S. Yamamoto, personal communication, 2001).

6.1.6.4 Exposure assessment

A schematic representation of a conceptual model of the V. vulnificus risk assessment model showing integration of all the modules is outlined in Figure 6.6. This includes the exposure assessment modules for harvest, post-harvest and consumption that were derived from the FDA-VPRA. The exposure assessment examined the appropriateness of transferring inputs from the V. parahaemolyticus risk assessment to that of V. vulnificus. Where this was not possible alternate approaches were developed. The predicted exposure was validated with data from a survey of V. vulnificus numbers in raw oysters at retail.

FIGURE 6.6: Schematic diagram of the V. vulnificus conceptual risk assessment model showing integration of all the modules.

6.1.6.5 Hazard characterization

V. vulnificus can occasionally cause mild gastroenteritis in healthy individuals, but for specific subpopulations V. vulnificus can cause a serious septicaemia that frequently leads to death in susceptible people. Therefore, the endpoint for the dose-response curve is defined as septicaemia. There was not adequate information to differentiate between virulent and avirulent strains of V. vulnificus. Therefore, all V. vulnificus strains were considered to be equally pathogenic.

While data from human volunteer studies were available for the construction of dose-response curves for V. parahaemolyticus and V. cholerae O1, no such data were available for V. vulnificus. Therefore, an alternate approach is being attempted. The dose-response relationship can be estimated by fitting a Beta-Poisson model using monthly data on the numbers of V. vulnificus in United States of America Gulf of Mexico oysters and the estimated consumption of raw oysters together with the monthly reported cases of V. vulnificus-associated septicaemia in that country. With further research, this risk relationship will be applied in the V. vulnificus risk assessment and validated. Preliminary results from this work were used in the risk characterization. However as this is a new approach to developing a dose-response relationship it is currently being fine-tuned and therefore the dose-response curve is not shown here.

6.1.6.6 Risk characterization

The risk characterization linked the exposure assessment and the dose-response to predict V. vulnificus illness rates. The predictions were compared to the observed illness rates. The model was then used to evaluate targeted mitigation levels for risk reduction.

6.1.6.7 Key findings

1. The FDA-VPRA provided a useful framework to model the risk of V. vulnificus septicaemia from consumption of raw oysters.

2. The model predictions of V. vulnificus exposure were validated using independent data from a survey of raw Gulf Coast oysters from the United States of America. Averaging water and air temperature over a ten year period, as was done in the V. parahaemolyticus risk assessment, could cause substantial deviations from observed levels in unusual climatic conditions such as the La Niña that occurred in 1998. However, there was good agreement between observed and predicted numbers of V. vulnificus using observed temperatures during this period as shown in Figure 6.7.

3. Preliminary results for the dose-response model indicated that closer agreement between predicted and reported illness rates can be obtained by either eliminating data associated with unusual climatic conditions or by using temperature, exposure and illness data individually for each month of each year available (1995 to 2001) without averaging.

4. An aggregate population dose-response could be approximated using available data on the differences in exposure to V. vulnificus from consumption of United States of America Gulf Coast oysters and reported illness frequency during warm and cold months.

5. The approach for determining dose-response circumvents the lack of data on frequency of virulent strains in raw oysters and uncertainty concerning the susceptible population by assuming that these do not vary from month to month.

6. Preliminary results giving the predicted illness rate using the averaging of years approach produced good agreement with observed illness rates except during the winter.

7. Preliminary evaluations of mitigations aimed at reducing V. vulnificus numbers in oysters to 3, 30 and 300 CFU/g indicate 60% to almost total reduction in predicted numbers of illness per year in the United States of America. Although some further refining of the modelling is required it appeared that this approach was appropriate for determining the potential effectiveness of specific mitigations to reduce V. vulnificus illness associated with consumption of raw United States of America Gulf Coast oysters.

6.1.6.8 Limitations and Caveats

This risk assessment framework and in particular the dose-response relationship developed using data for consumption of raw Gulf Coast oyster in the United States of America could be applied to oysters and other molluscan shellfish species in other regions of the United States of America and perhaps other countries. However, the use of data from this model as surrogate data for other regions should be carefully considered, especially in conditions such as temperature and salinity that are substantially different from those used to construct this model, and with shellfish species, culture conditions or industry practices that are different than the submerged bottom culture that is typical of the United States of America Gulf Coast. A modified temperature vs. V. vulnificus relationship for high salinity (30-35 ppt) may be more appropriate for many parts of the world where shellfish production is predominantly taking place in highly saline waters as the current model does not incorporate salinity and overestimates V. vulnificus densities at high salinities. The model framework is quite flexible and most model inputs could be easily adapted to fit specific situations if appropriate data were available.

The dose-response approach used in this assessment was a curve-fitting of the data to a beta-poisson model. While the beta-poisson model was selected any other empirical model that fits the data could be used. Extrapolation of the beta-poisson parameters used in this analysis beyond the range of normal consumption would be inappropriate.

The application of this model to predict the risk of V. vulnificus illnesses from seafoods other than molluscan shellfish was limited as the ecology of this bacterium differs considerably as do industry practices and consumer handling. However, the dose-response relationship could be useful in determining the risk of other seafoods if V. vulnificus levels in these products were known at the point consumption. The accuracy of these assessments would depend on the extent of matrix effects on the dose-response.

FIGURE 6.7: Predicted and observed numbers of V. vulnificus per gram oyster on a seasonal basis.

The model does not account for variation of strain virulence. Regional or seasonal variation in V. vulnificus virulence could alter the current dose-response and affect illness estimates.

This assessment was based on the distribution of at risk individuals^[52] in the population of the United States of America and this parameter would need to be redefined on a country by country basis depending on the size and characterization of the at risk population.

Since the dose-response data was generated in part using monthly illness rates in the United States of America, this data, in its current format, cannot be used to validate the model. However, illness data from that country may be useful for validation of the risk characterization in a different format (i.e. retrospective analysis of annual illness rates before and after specific mitigations). In 1997 the Interstate Seafood Sanitation Conference (ISSC) in the United States of America adopted a time-temperature matrix for reducing the time to first refrigeration of oysters from 20 to 10h under certain circumstances. The model could be used to analyse the effect on exposure and predicted illness and these could be compared to illness rates before and after adoption of the time-temperature matrix.

6.1.6.9 Gaps in the data

In the course of this work the following data gaps were identified.

• There was sufficient exposure data available for modelling the risk of V. vulnificus illness from consumption of raw Gulf Coast oysters in the United States of America using the proposed approach but this data especially numbers of V. vulnificus at harvest is also lacking in most other countries.

• There is a lack of reliable markers for virulence determination of V. vulnificus, thus requiring the assumption that all strains are virulent.

• The incidence of specific risk factors in the population consuming a seafood of interest and exposure associated with this seafood are the primary data needed for applying this model to other countries.

• Validation of the model in a given region or country would require epidemiological data on the incidence of primary septicaemia caused by V. vulnificus on a monthly basis.

6.1.7 Choleragenic Vibrio cholerae O1 and O139 in warm-water shrimps for export

6.1.7.1 Introduction

The justification for undertaking a risk assessment of this product-pathogen combination was that shrimp is an important commodity in international trade and is occasionally suspected to be involved in transmission of cholera, although there is little or no evidence that imported shrimp are actually the vehicle of transmission. The total world shrimp production in 1999 was about four million tons, of which 1.3 million tons were traded internationally with three quarters of this originating from developing countries (FAO, 1999^[53]). Shrimp exports are negatively affected, particularly when there are cases of cholera in shrimp producing countries.

A risk assessment of choleragenic V. cholerae O1 and O139 in shrimp for domestic use had also been initiated. This was discontinued as shrimp consumed domestically does not appear to be an important vehicle for transmission of cholera. Also, major difficulties and uncertainties exist in defining handling and storage practices; possible routes of faecal cross-contamination and consumption practices of domestic shrimp.

6.1.7.2 Scope

To assess the health risk of cholera associated with the consumption of imported warm-water shrimp.

6.1.7.3 Hazard identification

Toxigenic V. cholerae O1 and O139 are the causative agents of cholera, a water- and food-borne disease with epidemic and pandemic potential. Non-O1/non-O139 strains may also be pathogenic but are not associated with epidemic disease. Non-O1/non-O139 strains are generally nontoxigenic, usually cause a milder form of gastroenteritis than O1 and O139 strains, and are usually associated with sporadic cases and small outbreaks rather than epidemics (Desmarchelier, 1997^[54]). Outbreaks of cholera have been associated with consumption of seafood including oysters, crabs and shrimp (Oliver and Kaper, 1997^[55]). The largest outbreak was a pandemic in South America in the early 1990s when V. cholerae O1 caused more than 400,000 cases and 4,000 deaths, in Peru (Wolfe, 1992^[56]). Contaminated water used to prepare food, including the popular, lightly marinated fish ceviche, was associated with the outbreak.

Cholera occurs in areas with inadequate sanitary conditions and infrastructure and is associated with faecal contamination of water and foods. V. cholerae is widely distributed in coastal and estuarine environments all over the world and there exists over 170 serotypes of V. cholerae.

According to WHO definitions^[57], only serotypes O1 and O139 are the causes of cholera. Ability to produce cholera toxin (CT) is the determining virulence factor for causing cholera. However, environmental strains of V. cholerae O1 have often been shown to be non-toxigenic. Though some strains of non-O1/O139 V. cholerae may also cause gastroenteritis, the disease is of a mild to moderate severity^[58]. Choleragenic V. cholerae is susceptible to inactivation by cooking. Most of the risk associated with choleragenic V. cholerae comes from the consumption of raw seafood or from cross-contamination of the foods by food handlers or contaminated water.

Accordingly, this risk assessment considers only choleragenic V. cholerae O1 and O139.

6.1.7.4 Hazard characterization

V. cholerae O1 and O139 have both pathogenic and non-pathogenic forms based on the presence of specific virulence genes, ctx (cholera toxin gene). Infection by choleragenic V. cholerae O1 and O139 is characterized by an acute gastroenteritis. Therefore, the end-point of the dose-response curve was defined as gastroenteritis.

Human volunteer studies were available for the construction of the dose-response curves for V. cholerae O1. Reasonable Beta-Poisson dose-response parameters were obtained from data sets; however, the human volunteer studies characterize dose-response relationships for pathogens administered with a pH-neutralizing buffer rather than for pathogens administered with a food matrix. In normochlorohydric adult volunteers, doses of up to 10 ¹¹ choleragenic V. cholerae given without buffer or food did not reliably cause illness, whereas doses of 10 ⁴ to 10 ⁸ organisms given with NaHCO₃ (sodium bicarbonate) resulted in diarrhoea in 90% of individuals. Dose-response curves (Figure 6.8) show that a high dose of V. cholerae O1 (10 ⁶) was normally needed to cause illness when V. cholerae are consumed in food. In populations not exposed to choleragenic V. cholerae all age groups are equally susceptible. Immunity seems to be serotype specific.

FIGURE 6.8: Beta poisson dose-response curve for Vibrio cholerae

6.1.7.5 Exposure assessment

This risk assessment includes aquacultured and wild-caught warm-water shrimp. Choleragenic V. cholerae O1 and O139 generally occur in waters with salinity's between 0.2 to 20 ppt. Therefore, water and shrimp from offshore waters have not been found to contain choleragenic V. cholerae. Thus, it is assumed that any presence of choleragenic V. cholerae in offshore wild caught shrimp is caused by post-harvest cross-contamination. Though the presence of choleragenic V. cholerae in aquaculture environments is very rare, the model assumes that such choleragenic V. cholerae strains could be present in shrimp at levels similar to those found in coastal waters of cholera endemic countries.

The model developed was based on shrimp handling, processing and storage practices in units approved for export of shrimp (Figure 6.9). Such approval is based on sanitary requirements as described in Good Manufacturing Practices (GMP) and Good Hygienic Practices (GHP). Shrimp intended for export are generally iced immediately after harvest and transported in ice to certified processing units that meet GHP/GMP requirements. However, a worst-case scenario of shrimp being processed in non-approved units was also considered.

The major factors influencing the numbers of choleragenic V. cholerae in shrimp are time and temperature during handling, processing and storage. In the absence of available data it was necessary to make assumptions on distributions of time and temperature under such conditions. Adequate data were available on the effect of washing, freezing and cooking on the numbers of choleragenic V. cholerae in shrimp. In particular, the duration of frozen storage before consumption will cause a significant reduction in numbers of choleragenic V. cholerae. Limited information was available on the levels of faecal cross-contamination during handling and multiplication of choleragenic V. cholerae in raw shrimp. The model takes into account the pronounced reduction in the levels of choleragenic V. cholerae that would occur during cooking of shrimp either before export or before consumption. It also assesses the risk if shrimp were consumed raw or inadequately cooked in the importing country.

6.1.7.6 Risk characterization

The risk characterization was undertaken by combining the dose-response model with the estimated exposure to choleragenic V. cholerae via shrimp. Based on the available data, including additional information that was identified by the expert consultation, but not yet considered in the risk assessment, it will be feasible to progress with a semi-quantitative risk assessment. Further, epidemiological data on cholera cases reported to the WHO from major countries importing warm-water shrimp is available and this, together with data on shrimp imports and consumption, will be collected to validate the model.

6.1.7.7 Key findings

1. Following an extensive literature review it was noted that while V. cholerae is widely distributed in the environment, only strains producing cholera toxin and belonging to serotypes O1 and O139 are causative agents of cholera.

2. Contamination of shrimp, either wild-caught offshore or aquacultured, with choleragenic V. cholerae could happen during handling and processing, but there is very little opportunity for the multiplication of V. cholerae in shrimp processed in units meeting GMP/GHP requirements.

3. Major log-reductions in numbers of choleragenic organisms occur during washing, freezing and cooking.

4. Dose-response curves (Figure 6.8) show that a high dose of choleragenic V. cholerae O1 (10 ⁶) is normally needed to cause illness when choleragenic V. cholerae O1 are consumed in food.

5. The qualitative (descriptive) risk assessment showed that there was not a public health problem associated with the consumption of imported warm-water shrimp.

6.1.7.8 Limitations and caveats

There was limited or negative data available on the level of choleragenic V. cholerae O1 and O139 in shrimp at harvest. Estimations were based on reported levels found in waters.

Only one 20-year old reference was available on the lack of multiplication of choleragenic V. cholerae in raw shrimp.

There was no data on the level of choleragenic V. cholerae that may be transmitted by shrimp handlers, e.g. on fingers. Therefore, an assumption had to be made on transmission of V. cholerae by faecal cross-contamination.

The dose-response data for choleragenic V. cholerae O1 consumed with food were available for the classical biotype but not for the El Tor biotype. The El Tor biotype is the more common form and dose-response information on this form when administered with food rather than acid neutralizing vehicles is desirable.

6.1.7.9 Gaps in the data

The following data gaps were identified in the course of the work.

• Data on the levels of choleragenic V. cholerae O1 and O139 in natural waters and aquaculture environments.

• Data on the multiplication of choleragenic V. cholerae in cooked and raw shrimp.

• Data on levels of faecal cross contamination during handling of shrimp.

• Data to clarify the dose-response when the El Tor form is ingested in food.

FIGURE 6.9: Conceptual model for risk assessment of choleragenic V. cholerae in warm water shrimp for export.

6.2 Review of the Risk Assessments

The expert consultation undertook a technical review of the draft risk assessment document entitled "A Draft Risk Assessment of Vibrio spp. in Seafood." The expert consultation evaluated the risk characterization, as well as the underlying data, assumptions, and associated uncertainty and variability. The expert consultation recognized the extensive work undertaken by the drafting group, provided additional references to meet some of the specific data needs of the risk assessment and made recommendations on how to improve the document.

6.2.1 Introduction: Vibrio spp. in seafood

The expert consultation noted that it was desirable to state that the pathogenicity of V. parahaemolyticus is associated with TDH and/or TRH production. With regard to V. cholerae, it is necessary to state clearly that epidemic cholera is only associated with cholera toxin-producing strains of serogroups O1 and O139. It was recognized that not all professionals may be aware that virulence of V. parahaemolyticus, and O1 and O139 V. cholerae is associated with certain toxin-encoding genes, and that diagnostic tests can be used to separate these strains from others.

The expert consultation expressed concern that testing of seafood for Vibrio spp. was sometimes based on inappropriate markers (e.g. genus, species and/or non-consideration of pathogenic factors) that do not reflect the potential to cause human illness. It was agreed that the risk assessment should include a section outlining the Vibrio spp., types and virulence factors that may be included in the examination of different types of seafood in order to protect public health.

6.2.2 Vibrio parahaemolyticus in raw oysters consumed in Japan, New Zealand, Canada, Australia and the United States of America

The expert consultation identified that the risk assessment should include further consideration of the oyster industry practices in different parts of the world as these could have a significant effect on the appropriateness of the present model. In particular, this applied to the uncommon use of refrigerated transport and storage in many countries. The caveats that applied to the assessment due to these differences should be specified. It was determined that modelling should also consider salinity as a parameter, since in some areas of the world, salinity remains high throughout the year and exerts an effect in addition to that of seawater temperature. This might be addressed by having separate models for areas of relatively constant high salinity; relatively constant low salinity; and varying salinity. It was also necessary to consider that different oyster species might behave differently with regard to both the concentration of V. parahaemolyticus in the harvest area and to the growth of the organism during periods of temperature abuse or cool-down. Several of these considerations could have led to the model over-predicting the incidence of illness due to V. parahaemolyticus in oysters in Australia and New Zealand.

It was highlighted that the model did not currently incorporate consideration of TRH-producing strains of V. parahaemolyticus, nor did it presently encompass possible variability in the prevalence of TDH-producing strains seen in other countries. The model could be amended in the future to address these aspects.

The risk assessment should include fuller consideration of uncertainty and variability and it would be useful to include an outline of the use of the @Risk model. It would be necessary to specify as fully as possible the data needed, as well as any key caveats, if the model were to be applied or modified to be used in other regions and for other species.

6.2.3 Vibrio parahaemolyticus in bloody clams

The expert consultation congratulated the drafting group and the associated researchers for undertaking a valuable targeted risk assessment in a very short period of 5 months, going from literature review and capture of novel data to modelling and drafting of the assessment. The process itself could form an example for future targeted assessments.

The expert consultation recommended that the drafting group revise the assessment itself to ensure clarity with regard to the data and methods used (both microbiological and modelling) and to ensure that the figures and other outputs were sufficiently explained in the text. There was the potential to include more about the uncertainty and variability in the model and to look in more detail at the modelling aspects of the mitigation process (boiling). These recommendations did not detract from the quality of the work that has already been done.

6.2.4 Vibrio parahaemolyticus in finfish

An exposure assessment had been prepared previously and presented to the first expert consultation in Geneva in 2001^[59]. FAO and WHO, in conjunction with the drafting group, subsequently decided that there were insufficient data to take the assessment forward at the present time and therefore no further work had been undertaken. The expert consultation recognized the useful content included in the exposure assessment and determined that it should be included in the final report on the Vibrio risk assessments and could then form the basis for possible future work and as support to FAO/WHO member countries and Codex.

6.2.5 Vibrio vulnificus in raw oysters

The expert consultation noted the success in extending the use of the V. parahaemolyticus risk assessment and model to another species. It was discussed that the models are currently based on United States of America Gulf Coast data sets, and that the illness data used for validation are from the same country. The development of an additional salinity-temperature model will assist in extending this tool to other environments where high salinity may be a factor in limiting exposure. The stipulations noted for the V. parahaemolyticus assessment also apply to this assessment and appropriate data needs and caveats for application to other regions also need to be emphasized. The assessment should also include further discussion as to how the tool can be used in regions where shellfish -associated V. vulnificus infection may be of importance.

6.2.6 Choleragenic Vibrio cholerae O1 and O139 in warm-water shrimp for export

According to WHO definitions^[60], only serotypes O1 and O139 are the causes of cholera and the ability to produce cholera toxin (CT) is the determining virulence factor. However, environmental strains of V. cholerae O1 have often been shown to be non-toxigenic, therefore, according to the expert consultation, seafood products should only be analysed for cholera toxin - producing V. cholerae O1 and O139.

The adaptation of the V. parahaemolyticus model in oysters (based on temperature-V. parahaemolyticus numbers predictor) to a V. cholerae O1/O139 choleragenic model in warm-water shrimp was recognized as difficult because of the absence of a good predictor for V. cholerae numbers in shrimp, and it would therefore require a large amount of work to develop a completely new model.

The export of warm-water shrimp constitutes an important aspect of world trade. It was generally agreed that the available qualitative (descriptive) risk assessment showed that there was not a public health problem associated with the consumption of imported warm-water shrimp; however a semi-quantitative risk assessment should be undertaken with the available data in order to assist risk managers to better understand this.

It was decided not to proceed with a risk assessment of shrimp consumed by domestic markets in tropical countries. Since shrimp are consumed cooked, any illness will be the result of cross-contamination and this is not specific to any single food commodity.

6.3 Utility and applicability

The expert consultation emphasized the need for the risk assessments to be distributed as widely as possible and in forms that are appropriate to the target groups. As well as the planned Executive Summary and full Technical Report, an Interpretative Summary should be produced that explains the use and limitations of these risk-based tools without great emphasis on details of the models. FAO and WHO should consider the publication of entire consultation reports in respected, widely read journals. Some issues that need to be considered here are FAO and WHO policies on copyright, the need for peer review of already heavily reviewed documents and the timing of publication. In addition, the drafting group should be encouraged to submit relevant aspects of the work for publication in peer-reviewed journals to reach a greater audience. It will be important to consider publication of elements prior to publication of the full report in order to contribute to general development of microbiological risk assessment for foods.

It will be necessary for presentations to be made to relevant stakeholders, including Codex, in order to ensure that the relevant aspects were fully emphasized. Provision of appropriate PowerPoint presentations by FAO and WHO would assist in this. It was also perceived that the results of the work would be of value to the seafood industry and thus simplified summaries of the assessments should be submitted to trade periodicals. The benefits of these risk assessments will be realized through the use of trainers who are skilled in communicating with diverse types of audiences. FAO and WHO should make the software models available together with guidance on their use. They should also provide technical assistance to developing countries wishing to extend the assessments to local needs.

These risk assessments could be used to support relevant risk management decisions. The output from the assessments should also be used to formulate and target research needs, e.g. to fill identified data gaps.

6.4 Response to the specific questions posed by the Codex Committee on Fish and Fishery Products

A number of questions were posed to the expert consultation in relation to management strategies for food-borne illness due to V. parahaemolyticus and V. vulnificus. These were posed to the members of the consultation in their roles as experts in the microbiology of vibrios and/or seafood technology. The four questions posed by the CCFFP and the response from the expert consultation are outlined below.

Question 1: Whether the following pre-harvest control measures (testing/monitoring the following parameters and consequential closure of the harvesting area) are effective in the control of Vibrio parahaemolyticus and Vibrio vulnificus in bivalve molluscs:

• Testing of bivalve mollusc meat for Vibrio parahaemolyticus and Vibrio vulnificus

• Temperature monitoring of the growing water

• Water testing for Vibrio parahaemolyticus and Vibrio vulnificus

• Salinity monitoring

Response from the consultation: The concentrations of V. parahaemolyticus and V. vulnificus in shellfish may be measured directly or predicted by monitoring temperature and salinity. There will not necessarily be a direct relationship between these surrogate variables and the measured concentrations of pathogenic vibrios for a particular area as there is uncertainty and variability in the current models. The predictive abilities of the models would be improved by incorporating local data and considering additional factors such as hydrodynamic effects and sunlight. The effectiveness of these measures in controlling illness would depend on the instigation of an appropriate mitigation (or multiple mitigations) and this is not confined to closure of a harvesting area.

The current models do not include modules relating to the concentration of the two pathogens in seawater and thus the utility of measuring this cannot be estimated. If the appropriate data were gathered then the models could be extended accordingly.

Question 2 Are the following post-harvest treatment technologies, alone or in combination, effective in the reduction or elimination of V. parahaemolyticus and V. vulnificus in bivalve molluscs:

• hydrostatic pressure

• rapid cooling

• irradiation

• mild heat treatment (pasteurization)

• freezing and thawing

• depuration

Response from the consultation: These may all have the effect of reducing the numbers of pathogenic vibrios but the effectiveness will vary according to the conditions of use, and there may be a need to balance between obtaining the maximum possible reduction in bacterial content and retaining consumer-acceptance of either the product or the process. Reports on the effectiveness of depuration vary greatly and this may again depend on the conditions of use - some reports indicate that proliferation of vibrios may occur during this process. The general opinion of the expert consultation is shown on a qualitative/semi-quantitative basis in table 6.8.

The current models could be adapted to enable estimates to be obtained of the effectiveness of the mitigations in reducing illness. With regard to the mitigation of the closure of harvesting areas, estimates could also be obtained of the proportion of harvest lost by application of a particular scenario.

Some of the listed mitigations are also used in combination, e.g. hydrostatic pressure and freezing; depuration and hydrostatic pressure or pasteurization.

TABLE 6.8: The comparative effectiveness of a number of mitigation strategies in reducing Vibrio spp..

- no effect
+ some reduction
++ moderate reduction
+++ significant reduction

References and further reading relating to inactivation strategies

Calik, H., Morrissey, M.T., Reno, P.W. & An H. 2002. Effect of high pressure processing on Vibrio parahaemolyticus strains in oysters. Journal of Food Science, 67: 1506-1510.

Cook, D. W., and A. D. Ruple. 1992. Cold storage and mild heat treatment as processing aids to reduce the numbers of Vibrio vulnificus in raw oysters. Journal of Food Protection, 55:985-989.

Cook, D. W., and A. D. Ruple. 1989. Indicator bacteria and Vibrionaceae multiplication in post-harvest shellstock oysters. Journal of Food Protection, 52:343-349.

Cook, D. W. 1999. Effect of heat and freezing treatment on Vibrio parahaemolyticus O3:K6. Unpublished data.

Eyles, M. J., and G. R. Davey. 1984. Microbiology of commercial depuration of the Sydney rock oyster, Crassostrea commercialis. Journal of Food Protection, 47:703-706.

Gooch, J. A., A. DePaola, C. A. Kaysner, and D. L. Marshall. 1999. Postharvest growth and survival of Vibrio parahaemolyticus in oysters stored at 26° C and 3° C. Abstracts of the 99th General Meeting of the American Society for Microbiology, Abstract # P52.:521.

Greenberg, E. P., and M. Duboise. 1981. Persistence of Vibrio parahaemolyticus and Vibrio harveyi in hardshell clams. Abstracts of the 81st General Meeting of the American Society for Microbiology, Abstract No. Q93.:216.

Johnson, H. C., and J. Liston 1973. Sensitivity of Vibrio parahaemolyticus to cold in oysters, fish fillets and crabmeat. Journal of Food Science, 38:437-441.

Johnson, W. G., Jr., A. C. Salinger, and W. C. King. 1973. Survival of Vibrio parahaemolyticus in oyster shellstock at two different storage temperatures. Applied Microbiology, 26:122-123.

Motes, M.L. and DePaola, A. 1996. Offshore suspension relaying to reduce levels of Vibrio vulnificus in oysters (Crassostrea virginica). Applied and Environmental Microbiology, 62, 3875-3877.

Richards, G. P. 1988. Microbial purification of shellfish: A review of depuration and relaying. J. Food Protect. 51:218-251.

Son, N. T., and G. H. Fleet. 1980. Behaviour of pathogenic bacteria in the oyster, Crassostrea commercialis, during depuration, re-laying, and storage. Applied and Environmental Microbiology, 40:994-1002.

Question 3 For Vibrio parahaemolyticus - Are food-borne illnesses caused by the heat resistant toxin produced by the pathogen or by the pathogen itself?

Response from the consultation: The illness is caused by the toxin but only if this is produced in the intestine following colonization by a strain producing TDH, TRH or both toxins.

Question 4: What is the availability of methods of analysis for Vibrio parahaemolyticus toxin gene (tdh)?

Response from the consultation: Both tdh and trh genes can be detected using PCR with relevant primers and by membrane filtration-hybridization methods with non-isotopic oligonucleotide or PCR-generated probes. For quantification, PCR methods can be applied in an MPN format whereas membrane filter-hybridization can be used for direct colony enumeration. PCR and colony hybridization procedures are also available for the thermolabile haemolysin gene (tlh) for determining V. parahaemolyticus species. As with conventional methods, there is scope for standardization and/or the determination of the relative performance of current methods.

6.5 Response to the needs of the Codex Committee on Food Hygiene

A CCFH drafting group has prepared a document titled "Discussion Paper on Risk Management Strategies for Vibrio spp. in Seafood." In that paper, some risk assessment needs and questions for risk assessors were identified, which include an evaluation of the impact of several potential interventions on the risk of V. parahaemolyticus infection. The current risk assessment of Vibrio spp. in seafood is addressing many of these potential interventions through the models that have been developed including the influence of temperature on growth and the impact of various target reduction levels of Vibrio spp. in oysters on the risk of illness. The effects of various mitigation strategies are also described in section 6.4.

6.6 Conclusions and recommendations

The drafting group has progressed towards improving the risk assessments since the last expert consultation^[61]. The majority of the previous recommendations have been addressed in the current draft report.

The vibrio risk assessments should include clear advice as to the species, serogroups, serotypes and genotypes that should be considered as being of public health significance with respect to the trade and consumption of seafood. Interpretative summaries of the risk assessments should be produced in order to maximise the understanding of the work by other professionals. Where appropriate, these summaries should have limited descriptions of the mathematical modelling in order to reach a wider audience.

It was recommended that the risk assessment on "Vibrio cholerae in shrimp" be developed further on semi-quantitative basis only and the opinions of risk managers should be sought before any work is undertaken towards a fully quantitative risk assessment.

The exposure assessment on "Vibrio parahaemolyticus in finfish" should be included in the final report on the vibrio risk assessments as this contains information that may be useful to a number of countries.

The procedures used to undertake the risk assessment on bloody clams should be recognized as a way to expedite the development of pathogen-commodity risk model. The draft of the risk assessment itself should be revised to ensure greater transparency of the modelling approaches.

The risk assessment on "Vibrio parahaemolyticus in oysters" should include further consideration of the oyster industry around the world (practices, species effects, harvest water salinity, etc). Where the model cannot be modified appropriately, the potential constraints and/or the additional data needs should be clearly identified. It should identify the potential shortcomings of the assumption which precluded the consideration of TRH-producing strains as pathogenic and also the possible consequences of the high prevalence of TDH-producing strains in some parts of the world. Further consideration should be given to uncertainty and variability, and it should also include an description of the Excel-based model in order to assist understanding its potential applications and limitations.

The risk assessment on "Vibrio vulnificus in oysters" should include consideration of the potential constraints and/or the additional data needs to extend it to other geographical areas outside of the United States of America. It should identify more clearly those areas of the world and the seafood products that were currently known to be associated with food-borne V. vulnificus infection.

In order to undertake further risk assessments it will be necessary to obtain additional information on the proportion of strains of each Vibrio spp that possess pathogenic traits. Virulence factors of V. parahaemolyticus (TDH and TRH) and V. cholerae O1 and O139 (CT) should be identified where appropriate so as to specify the strains that are relevant to human illness. The expert consultation believed that additional information on dose-response relationships with respect to food contaminated with Vibrio spp. could be obtained by investigation of outbreaks.