U. S. Food and Drug Administration
Center for Food Safety and Applied Nutrition
January 2001

Draft Risk Assessment on the Public Health Impact of
Vibrio parahaemolyticus in Raw Molluscan Shellfish

The objectives of the risk assessment were two-fold: (a) to create a mathematical model and assess the current risk of becoming ill due to the consumption of pathogenic V. parahaemolyticus in raw oysters; and (b) develop a comprehensive and current scientific framework, which will assist the agency with the review of current programs relating to the regulation of V. parahaemolyticus in raw molluscan shellfish to ensure that such programs protect the public health. The risk assessment task force was also charged to evaluate: the evidence for increased risks from specific newly emerging "outbreak strains", the effectiveness of potential strategies for limiting exposure of the public to raw molluscan shellfish, particularly oysters containing pathogenic V. parahaemolyticus, the current criteria for opening and closing harvest waters, and FDA's current established guideline level of 10,000 V. parahaemolyticus/g of food.

The risks for sporadic illnesses occurring due to V. parahaemolyticus in oysters are determined by this risk assessment. Assessment of the risks associated with oyster-borne outbreaks caused by pathogenic V. parahaemolyticus will not be feasible until a later date. The data needed to model outbreak-associated risk are not as yet available. These data will be obtained from the interim control strategy for preventing outbreaks caused by V. parahaemolyticus by monitoring oyster meats for strains having the TDH gene, implemented in 1999 by the Interstate Shellfish Sanitation Conference (ISSC). Monitoring for pathogenic strains and implementation of harvest controls commenced in the spring of 2000 (some states began monitoring in 1999). To reasonably assess the effectiveness of this interim control strategy, FDA needs the following information: (a) which regions are monitoring, (b) the number of growing areas monitored in each region, (c) the number of total samples collected and the number positive for V. parahaemolyticus in each region, (d) the sensitivity of testing, (e) oyster landing data for each site and region during periods of monitoring, and (f) case data on oyster-borne illnesses caused by pathogenic V. parahaemolyticus.

The solicitation and assemblage of information and scientific data on V. parahaemolyticus from many sources produced a thorough, up-to-date compilation. This information was used in the construction of a mathematical model to produce results on the risk of illness incurred by eating raw oysters containing pathogenic V. parahaemolyticus. Three basic factors were found to be associated with this pathogen and consumer risk: the level of pathogenic V. parahaemolyticus in seafood at harvest, effect of post harvest handling and processing, and the ability of the organism to multiply to an infective dose. As a result, the risk assessment project was divided into three separate modules, which corresponded to different stages leading potentially to consumer exposure: the Harvest, the Post Harvest, and the Public Health Modules. The Harvest Module estimated the prevalence of pathogenic V. parahaemolyticus at time of harvest. The Post Harvest Module determined the role of post harvest processing and handling on the levels of pathogenic V. parahaemolyticus at consumption. The Public Health Module estimated the risk of illness caused by this organism. Because of harvesting and temperature differences, for the purpose of the model, the United States harvest areas were divided into five regions, and each region was divided into four seasons. Differences existing in oyster harvesting practices and climates in the United States were sufficiently significant to identify five separate geographic regions (Northeast Atlantic, Mid-Atlantic, Pacific Northwest, Louisiana Gulf Coast, and the remainder of the Gulf Coast) for each season, for consideration in modeling each of the modules. Factors influencing the risk of illness posed by V. parahaemolyticus were identified and incorporated in each module as appropriate. Integration of the various parameters comprising these modules into a quantitative risk assessment model has provided a more comprehensive understanding of the relative importance and interactions among these factors influencing risk. This gain in understanding should serve to facilitate several processes, including the formulation of effective guidance for the industry, regulators and consumers, the evaluations of risk mitigation strategies, and the development of options and policies for managing risk.

While providing a framework for understanding the relationship of risk to various parameters, the development of the risk assessment model necessarily required certain assumptions to fill the data gaps. The assumptions incorporated in the model were reviewed by NACMCF at a public meeting in September1999. Based on the information currently available, for the Harvest Module, it was assumed that the presence of the TDH gene be used as the basis for pathogenicity. It is not currently known what average levels of TDH-positive strains actually exist in shellfish, nationally or regionally. The estimates made in the V. parahaemolyticus risk assessment, based on observed frequency of TDH-positive isolates, were the best possible with the data currently available. However, since we do not know how this frequency may vary from one year to the next, we assumed a 2-fold up or down triangle distribution. Also, within a given year, we were unsure about the variance of percentage pathogenic in one composite of oysters to the next. For example, outside of the Pacific Coast, percentage pathogenic V. parahaemolyticus in a given year ranged from 0.1% to 0.3%; for the Pacific Northwest the range used was 2% to 4%. Furthermore, these estimates are based on older data, and may not be predictive of future years, given that frequency of percentage pathogenic V. parahaemolyticus may be changing as new outbreak strains emerge or reemerge, such as the emergence of O3:K6 or recurrence of known outbreak strains such as O4:K12.

For the Post Harvest Module, several assumptions were made based on the knowledge of current post harvest practices and information available. The time oysters are harvested to the time they are refrigerated was based on the current NSSP requirement (64) put into effect in 1997. The extent of growth that occurs during the period of time from harvest until the time that oysters are first placed under refrigeration is determined by three factors: (a) the growth rate of V. parahaemolyticus as a function of temperature; (b) the temperature of oyster meat after harvest and (c) the length of time held unrefrigerated. The growth rate of pathogenic V. parahaemolyticus in oysters was assumed to be one fourth that in broth culture at all temperatures. This rate was based on the model of Miles et al. (95), and the corresponding studies in oysters by Gooch et al. at 26°C (51). Also, since the V. parahaemolyticus organisms do not change their growth environment after harvest (within the oyster meat), it was assumed that lag time was negligible and was therefore omitted from the growth model. Regarding growth rates, preliminary studies at GCSL, showed no significant difference between pathogenic and non-pathogenic strains of V. parahaemolyticus. Since data on cooling rates of commercial oyster shellstock could not be located, the time for oysters to cool after being placed under refrigeration was assumed to be quite variable. This depended on efficiency of the cooler, quantity of oysters to be cooled and their arrangement in the cooler. A uniform distribution between 1 and 10 hours was used to model this parameter based on preliminary GCSL experiments for the time it took a single shell oyster at 30°C placed into a 3°C cooler to reach that temperature, and the time it took for 24 oysters in an uninsulated plastic container at 26°C to reach 3°C.

For the sake of simplicity of the model, we assumed that consumption patterns were the same for both the sensitive and otherwise healthy population, for all regions. It was assumed that all virulent/pathogenic strains of V. parahaemolyticus are equally virulent with the same dose-response as those strains fed to human volunteers in earlier studies. This assumption was based on personal communication with Dr. Nishibuchi, Kyoto University (105), who stated that due to lack of information, it is not known whether there are differences in virulence among different strains.

Our model clearly illustrated that air and water temperatures were the driving factor for initial pathogen loads as well as continued growth after harvesting, but there is some uncertainty due to lack of data showing correlating V. parahaemolyticus levels. It is also noteworthy that in the Pacific and Mid-Atlantic, the lower air temperatures reduce the importance of air temperature and time unrefrigerated compared to the Gulf Coast.

The risk assessment model illustrated that the most significant factor influencing probability of illness due to V. parahaemolyticus is the level of V. parahaemolyticus present in the oyster at harvest. However, the model is based on a strong correlation between total and pathogenic V. parahaemolyticus levels at time of harvest. We have also assumed that pathogenic strains of V. parahaemolyticus grow at the same rate as non-pathogenic strains. Consequently, as the level of total V. parahaemolyticus increases so does the number of pathogenic V. parahaemolyticus.

For the Gulf Coast, the second most influential factor for occurrence of illness is the duration that oysters are left unrefrigerated after harvest. For the remaining regions modeled, i.e., Mid-Atlantic and Pacific Northwest, water temperature was the second most influential parameter. For all regions, however, the amount of oysters consumed was the third most influential factor. It is interesting to note that time the oysters were left unrefrigerated was more significant for the Louisiana Gulf Coast than for the remaining Gulf Coast. It is known that many oyster harvesting areas are further offshore in Louisiana than in the rest of the Gulf Coast, and therefore it takes longer for the boats to return to the shore after harvest, leaving the oysters unrefrigerated for a longer time period.

Modeling of the Post Harvest Module demonstrated that if oysters are not refrigerated rapidly after harvest as recommended by NACMCF (102), V. parahaemolyticus rapidly multiply in oysters resulting in much higher levels. The model's simulation of mitigation strategies indicated a significant reduction in the probability of illness when the oysters are cooled immediately after harvest. Furthermore, V. parahaemolyticus densities were shown to decrease slowly during refrigerated storage, as also stated by the MSI and PCSGA in response to the Federal Register notice Docket No. 99N-1075 (43). Moreover, the use of mild heat treatment, which causes at least a 4.5 log₁₀ decrease in the number of viable V. parahaemolyticus in oysters, practically reduced to zero the probability of illness occurring. Freezing, which causes a 1 to 2 log₁₀ decrease substantially reduced the probability of illness.

Earlier human trials conducted in Japan showed an increase in the number of illnesses with increasing levels of pathogenic V. parahaemolyticus. Different dose-response models were compared for the purpose of extrapolating risk of illness estimated on the basis of human feeding trials at high levels of exposure to the lower levels of exposure associated with consumption of raw oysters. However, consideration of CDC estimates of annual illness suggested that the dose-response under conditions of population exposure was different than that observed in human volunteer studies. In other words, direct extrapolation of the dose-response under conditions of exposure in the feeding trials is not supported by the epidemiological data. The human feeding trials were conducted under conditions of concurrent antacid administration. Due to possible food matrix effects of the oyster, dose-response was shifted by 1 log₁₀ from that based on published clinical trials. Preliminary data have shown that this shift is "supported" by consideration of the CDC numbers of V. parahaemolyticus infection. Distributions of ingested dose were developed by considering the probabilistic variation of number and meat weight of oysters in a serving in addition to the expected variation of the density of pathogenic V. parahaemolyticus determined in the Harvest and Post Harvest Modules.

The outputs from this project provide estimates of risk for illness among consumers of raw oysters (average nationwide yearly incidence of 4,750 cases per year, with a range from 1,000 to 16,000 cases - for the Gulf Coast, 25 (winter), 1,200 (spring), 3,000 (summer), and 400 (fall); for the Pacific Northwest, 15 (spring) and 50 (summer); for the Mid-Atlantic, 10 (spring) and 12 (summer); and for the Northeast Atlantic, 12 (spring), 30 (summer) and 7 (fall)). Risks increase with increasing levels of total V. parahaemolyticus and therefore pathogenic strains of V. parahaemolyticus.

The model made it possible to develop a mathematical means of relating potential microbiological criteria with both the predicted percentage of illness prevented and the predicted percentage of the oyster landings that would no longer be available to consumers if the criterion could be implemented with 100% efficiency. Retail surveys of oysters, clinical studies and outbreak investigations, have shown that the guidance level of 10,000 viable V. parahaemolyticus cells/gram of oyster meat may not be relevant to safety. Simulations on the rate of illness caused by oyster-servings where the levels of V. parahaemolyticus at harvest are at or above 10,000 cells/g suggest that approximately 15% of the illnesses are associated with the consumption of oysters containing greater than 10,000 V. parahaemolyticus/g at time of harvest. This is a consequence of the fact that higher levels of pathogenic V. parahaemolyticus are more likely to occur with higher levels of total V. parahaemolyticus. Nevertheless, even at these high levels, the risk of illness per serving is still comparatively low. Comparing the number of servings that cause illness to those that don't, the simulations demonstrate that on average 0.6% of the servings result in illness when V. parahaemolyticus levels are at 10,000 cells/g or above.

The risk assessment team addressed the questions that it was charged with as described below.

• What is the frequency and extent of pathogenic strains in shellfish waters and shellfish?

  • There is a need for more information on virulence factor determination, such as invasion of the enterocytes (3), and production of an enterotoxin (61), and particularly, urease production (77, 79). Since the data on levels of pathogenicity pose a large uncertainty, the model was run with a range of estimates for TDH, from 0.0032%, to 3.2%. This produced a range of possible illness rates, with the most likely being identified (along with the reasons why it seems the most likely) as the most probable level of risk.

  • The epidemiology and pathogenicity of V. parahaemolyticus will be better understood once the ecological nuances of the microbe-host dynamics in the estuarine ecosystem are elucidated.

• What parameters can predict presence?
  • Our model clearly illustrated that air and water temperatures were the driving factor for initial pathogen loads as well as continued growth after harvesting, although some uncertainty exists due to lack of direct measurement of the relationship in regard to pathogenic V. parahaemolyticus levels per se.

  • The literature has shown that other factors not incorporated into the model may also predict presence of pathogenic V. parahaemolyticus. These include:
    • Salinity - The Texas outbreak showed a significant increase in the salinity levels and temperature in the spring (May-June) of 1998 compared with the previous five years, however these levels are very similar to normal summer temperature and salinity levels. It is also possible that the sudden increase in temperature/salinity causes the V. parahaemolyticus, both the pathogenic and non-pathogenic strains to grow quickly or multiply. However, our model suggested that salinity was not, in fact, an important variable.

    • Ballast waters - Although there was insufficient data to quantitatively model this parameter, ballast waters have been associated with Vibrio illnesses. (93).

    • Other factors, which were also not quantitatively modeled, such as immune status of the oyster, have also been shown from the literature to play an important role in the prevalence of pathogenic V. parahaemolyticus. Oyster data from the DePaola et al. study (38) on the incidence of V. parahaemolyticus in U.S. coastal waters and oysters were probably representative of different physiological states of the oysters. Once sufficient data have been accumulated to actually investigate the V. parahaemolyticus load and health of the oysters, the information can be incorporated into the model. There is a need to correlate the number of V. parahaemolyticus with the percentage oysters diseased.

• How do levels at consumption compare to initial levels?
  • Preliminary analysis of the ISSC/FDA retail study showed higher V. parahaemolyticus densities at retail than at harvest for the Gulf Coast. Model predictions suggest that the difference between densities at harvest versus those at time of consumption is largely attributable to the extent of growth that occurs after oyster harvest is landed and has not yet been cooled to no-growth temperatures. The extent of the difference varies by region and is largest during the summer ranging from 0.50 for the Pacific Northwest to 1.25 log₁₀ for the total Gulf Coast. During the spring the difference between densities at harvest versus time of consumption is approximately 0.75 log₁₀ for the Gulf Coast and somewhat less for other regions of the country. The simulation results are consistent with the V. parahaemolyticus densities observed at retail in the ISSC/FDA study.

• What is the role of post harvest handling?
  • The model demonstrated that if oysters are not refrigerated after harvest, V. parahaemolyticus rapidly multiply in oysters resulting in much higher levels. The model revealed a significant reduction in the probability of illness when the oysters are cooled immediately after harvest. Furthermore, V. parahaemolyticus densities are shown to decrease slowly during refrigerated storage.

• What intervention strategies can be used?
  • Our simulation of post harvest mitigation strategies has shown that mild heat treatment which causes at least a 4.5 log decrease in the number of V. parahaemolyticus essentially eliminates any probability of illness.

  • Freezing combined with frozen storage, which causes a 1 to 3 log decrease substantially reduces the probability of illness.

• What is known about dose-response?
  • Early Japanese feeding studies described in previous sections, have shown an increase in number of illnesses with increasing levels of pathogenic V. parahaemolyticus. However, the extent to which the estimate of the ID₅₀ under conditions of human feeding trials underestimates the "true" population ID₅₀ is unknown. In consideration of CDC estimates of annual illness rates, model predictions based on projected levels of pathogenic V. parahaemolyticus suggest that the dose-response for illness under conditions of normal population exposure may be significantly different from the conditions of the feeding trials. Possible reasons for the difference include food matrix or immunological effects of preexposure to the organism including antibodies/vaccines to the organism (88).

• How does dose-response vary for different strains of V. parahaemolyticus?
  • Based on insufficient epidemiological data and personal communication with Prof. Mitsuaki Nishibuchi who suggested that more research is needed to determine whether strains differ in virulence (105), we have assumed equal virulence for all virulent strains. The assessment of dose-response according to virulence was developed based on the organism's ability to produce TDH. This may be modified as new data become available that identify new virulence determinants. As mentioned previously, other strain characteristics, such as invasion of the enterocytes (3), and production of an enterotoxin (61) not normally investigated in the environmental or in clinical isolates, may also be important in characterizing pathogenicity. In particular, recent data from British Columbia (77) may suggest an association of urease positive strains with clinical isolates, which may or may not be TDH positive.

• How does dose-response vary among humans with different susceptibilities?
  • Epidemiological data indicate that the whole population is susceptible to infection.

  • Differences in dose-response among humans with different susceptibilities, are not as yet known. However, given infection, regardless of dose, there is a greater probability of infection leading to more severe illness, such as septicemia and death, among the subpopulation with a concurrent underlying medical condition. As mentioned in the Risk Characterization section, using case series data provided by the CDC, diarrhea in an infected person has a 12% chance of going on to septicemia if that person belongs to the sensitive subpopulation, and V. parahaemolyticus has been culture-confirmed.

• Is current knowledge adequate in assessing the risk of consuming raw oysters containing pathogenic V. parahaemolyticus?
  • Our gain in understanding of the relative importance and interactions among the factors influencing risk should serve to facilitate several processes, including the formulation of effective guidance for the industry, regulators and consumers, the evaluations of risk mitigation strategies, and the development of options and policies for managing risk.

  • However, the risk assessment also identified data gaps, which will need to be addressed by research.

• Where should future research be directed to reduce uncertainty in risk estimate?
  • Future research should be directed at the data gaps as listed below

Data Gaps and Future Research Needs

Deficiencies of the current research with respect to risk assessment were identified in order to suggest future research or further data gathering to reduce uncertainties.

• Incidence/frequency of pathogenic V. parahaemolyticus in water and shellfish.
  • Factors that affect incidence of pathogenic V. parahaemolyticus in the environment.

  • Role of oyster physiology and immune status in levels of V. parahaemolyticus. There is a need to correlate the number of V. parahaemolyticus with the percentage oysters diseased.

  • More research on the potential virulence factors of pathogenic strains other than TDH, e.g. urease, enterotoxins. V. parahaemolyticus strains that do not produce TDH, TRH, or urease have recently been found to induce fluid accumulation in suckling mice and diarrhea in a ferret model after oral inoculation in a dose-dependent manner (84). Correlation between clinical and environmental incidence of these strains is yet to be determined.

• Growth rate of V. parahaemolyticus within oysters at temperatures other than 26°C; including the issue of potential differences in the growth rate of pathogenic strains versus total V. parahaemolyticus populations.

• Rates of hydrographic flushing (water turnover) in shellfish harvest areas based on levels of freshwater flows, tidal changes, winds, depth of harvesting area and how these factors may influence pathogenic V. parahaemolyticus levels.

• Dose-response data.
  • FDA is currently funding a cooperative agreement with the University of Maryland to acquire a dose-response curve for V. parahaemolyticus in humans after extrapolation from animal studies. Volunteers will be fed raw oysters containing V. cholerae non-O1 at varying doses. Animals will be fed with V. cholerae non-O1 and V. parahaemolyticus at varying doses. After the animal doses and strains have been compared, the V. parahaemolyticus dose-response observed in animals will be extrapolated to humans, based on the data from the clinical and animal studies with V. cholerae non-O1. Better dose-response data will be available within the next two years after the completion of this study.

  • More intensive investigations of shellfish foodborne disease outbreaks in such a way as to examine the relationships between the dose of contaminated food items ingested and the severity of the resulting illness controlling for host factors.

• More data from State surveillance systems.

• Consumer handling of oysters.

• Improved global public health surveillance of V. parahaemolyticus to identify new epidemic strains as they emerge.

  • FDA is already currently involved in several collaborative efforts with the ISSC and the states, to protect the public from illness caused by V. parahaemolyticus and other seafood pathogens. Some of these efforts have provided invaluable information for the risk assessment. Data acquired in the future from the other efforts (listed below) will be used to validate and update our model.

• The V. parahaemolyticus interim control plan for closing and reopening harvest waters was adopted by the ISSC in July, 1999, for areas confirmed as an original source of oysters harboring pathogenic (TDH+) V. parahaemolyticus, that have been associated with two or more confirmed illnesses within the last three years. The outcome of modeling this plan will also enable the evaluation of the criteria for closing and reopening harvest waters. When the requisite data become available, it is anticipated that the reduction in risk afforded by the ISSC prevention strategy as well as the criteria for opening and closing harvest waters can be modeled.

• Oyster harvest monitoring survey - FDA in conjunction with the ISSC and the states is conducting a study to determine the levels of V. parahaemolyticus in oysters in the growing areas before they are harvested. This study was initiated in 1998, and some of the data obtained thus far, was incorporated into the risk assessment when determining levels of total and pathogenic V. parahaemolyticus.

• The time to refrigeration data obtained from states that had had V. vulnificus illnesses, proved to be invaluable in providing the data for the time to temperature simulations in the Post Harvest Module.

In conclusion, this risk assessment significantly advances our ability to describe our current state of knowledge about this important foodborne pathogen, while simultaneously providing a framework for integrating and evaluating the impact of new scientific knowledge on enhancing public health.

The results of this draft risk assessment on V. parahaemolyticus are influenced by the assumptions and data sets that were used to develop the exposure assessment and hazard characterization. These results, particularly the predicted estimates of risk for illness among consumers of raw oysters, and the most significant parameters, which influence the incidence of illness, could change as a result of future data obtained from the Interim Control Plan and the FDA actively seeking new information, scientific opinions, or data during the public comment period. It is anticipated that periodic updates to the risk model will continue to reduce the degree of uncertainty associated with risk estimates, and that this will assist in making the best possible decisions, policies, and measures for reducing the risk posed by V. parahaemolyticus in raw molluscan shellfish.

Hypertext updated by bap/kwg/dms 2001-JAN-30