Vibrio parahaemolyticus Risk Assessment – Appendix 2: Response to Public Comments
July 19, 2005
Comments on the draft Vibrio parahaemolyticus Risk Assessment were solicited in the Federal Register notice of availability (Federal Register Docket No. 99N-1075) in the following areas:
- The assumptions made
- The modeling technique
- The data used, and
- The transparency of the draft risk assessment document.
FDA received comments from a total of eight institutions or individuals, within the U.S., and abroad: The Food Marketing Institute, The New York State Department of Environmental Conservation, Flow International Corporation, Ministry of Agriculture and Forestry (New Zealand), National Fisheries Institute, PCSGA, CSPI, and Aamir M. Fazil (Health Canada). FDA thanks all of the above-mentioned for taking the time and effort to provide us with their comments. We feel that these comments helped to a great extent to improve our risk assessment. The FDA VPRA team reviewed all the comments and we have addressed them to the best of our ability and the scientific data available. Below is a summary of the key comments and FDA's response to these comments.
Comments on the Assumptions
Comment 1. The assumption of "equal virulence for all pathogenic strains of V. parahaemolyticus" is debatable.
FDA Response 1. While it is almost certain that not all pathogenic V. parahaemolyticus are equally virulent, we are unaware of any definitive data indicating the magnitude of differences in virulence among pathogenic strains that would allow us to separate them into subcategories beyond that already done in the risk assessment. The availability of such data would likely have two impacts on the risk assessment. Better data on the relative virulence among TDH+ strains would provide a better estimate of the variation among strains and thus decrease the uncertainty of the Hazard Characterization, but it would be unlikely that the additional data would greatly change the confidence intervals surrounding the Dose-Response relationship. What would have a great impact would be if there were additional or alternative virulence factors and if the prevalence of the more virulent strains varies seasonally or geographically. This would have the effect of further "concentrating " the risk within specific region/season combinations. The geographical variation of prevalence of TDH+ strains in the Pacific Northwest versus other areas of the country has already been incorporated into the assessment.
Comment 2. The assumption that all V. parahaemolyticus whether pathogenic or nonpathogenic have similar growth and survival rates is questionable.
FDA Response 2. Studies performed since the draft V. parahaemolyticus risk assessment was published comparing growth rate of pathogenic and non-pathogenic strains in broth culture, demonstrated no significant difference in growth between the different strains (Cook, 2002a). More recent data on mitigation strategies, such as mild heat treatment and ultra high-pressure treatment have shown that O3:K6 strains are somewhat more resistant to these techniques than the other pathogenic strains (Cook, 2002c). The average D value for thermal treatment of non O3:K6 V. parahaemolyticus was 47.6 seconds (ranging from 25-89), whereas that of O3:K6 isolates was 137 seconds (ranging from 108-187). When ultra high pressure was used, the average D value for non O3:K6 strains at a pressure of 36,250 MPa was 24.6 seconds, and for O3:K6 strains, it was 51.9 seconds. These differences have been noted in the revised risk assessment.
Comment 3. Consumption patterns by immunocompromised and healthy populations should not be assumed to be the same.
FDA Response 3. There is little information currently available to estimate the impact of warning labels and other consumer advice on the behavior of individuals who may be more susceptible due to compromised immune function. In the absence of such information, we continue to feel that this is the most appropriate assumption regarding consumption patterns.
Comment 4. The assumption that lag time to growth of V. parahaemolyticus in oysters after harvest appears to be negligible is conservative and may result in an overestimate of the growth rate.
FDA Response 4. The specific behavior of the V. parahaemolyticus within the oyster at the time of harvest has not been studied extensively; however, there is a wealth of information available on the behavior of the microorganisms in relation to the lag phase. Lag in growth occurs when there is a change in an organism's environment or there is a substantial temperature change. At harvest, V. parahaemolyticus remains within the oyster, and is subjected to only modest temperature change over a substantial period of time. A lag phase is therefore not expected under these circumstances, and the original assumption remains the most biologically plausible.
Comment 5. The assumption that water activity of oysters does not vary substantially is conservative and may result in an overestimate of the growth rate.
FDA Response 5. A growth study of V. parahaemolyticus in oysters that was replicated during each month of the year indicated similar growth rates when salinity ranged from 8.5 to 25 ppt (Gooch et al., 2002). Reduced growth was observed in February when the salinity was 4 ppt but in a nationwide retail survey of shellstock oysters salinity below 8 ppt was rarely observed (FDA/ISSC, 2000; Cook et al., 2002a). This narrow range of encountered salinity does not support consideration of alternative assumptions related to the importance of water activity differences. Furthermore, the risk assessment examined in detail the influence of salinity and concluded that the effect of that variable was minor in relation to the primary determinant, water temperature.
Comment 6. The assumption that growth rate in oysters is a constant fraction of the growth rate in broth at all temperatures is conservative and may result in an overestimate of the growth rate.
FDA Response 6. It is possible that the assumption of a constant fraction of the growth rate in broth may not hold at temperatures much higher or lower than the experimental temperature (26 °C) that was used by Gooch et al. (2002). However, the ambient temperature of Gulf Coast oysters prior to refrigeration is very close to 26 °C (78.8 °F) from April through October, the period when most V. parahaemolyticus cases occur. For this reason, the assumption may not be overly critical to the risk assessment, but the risk assessment will be revised when more data are available. Furthermore, in spite of the possibility that the assumption may not hold for temperatures substantially different from 26 °C (78.8 °F), the exposure levels predicted by the VPRA for other regions and seasons (with cooler temperature) based on the assumption are in relatively good agreement with those observed during the retail study (FDA/ISSC, 2000; Cook et al., 2002a). See Figures V-9 to V-12 of technical document.
Comment 7. The assumption that the temperature of oyster meat equilibrates rapidly with that of the ambient air and air temperature as a surrogate for oyster meat temperature is conservative and may result in an overestimate of the growth rate.
FDA Response 7. Along the Gulf, water and air temperatures are nearly the same with water temperature slightly higher than air temperature on average. Consequently, for the Gulf the assumption of rapid equilibration of oyster temperature to air temperature is not conservative per se. When stored in a burlap sack, evaporative cooling has been observed to result in gradual equilibration of oysters to a temperature up to 4 °C cooler than air temperature (Cook, 2002b) but the prevalence of this storage practice (during harvesting) is unknown. For other regions and seasons, where water temperature is substantially lower than air temperature, the assumption of rapid equilibration to air temperature may be somewhat conservative. However, model predictions were compared to retail measurements of total V. parahaemolyticus and no substantial differences were noted between observed versus predicted levels. On the West Coast intertidal oysters are typically in a monolayer before harvest and would probably heat up quickly when exposed on a falling tide. During sunny days oyster temperatures were observed to be 5 to 10 °C (41 to 50 °F) warmer than air temperatures (DePaola et al., 2002; Herwig and Cheney, 2001). Furthermore, in spite of the possibility that the assumption may be conservative and may result in an overestimate of the growth rate, the exposure levels predicted by the VPRA based on the assumption are in relatively good agreement with those observed during the retail study (FDA/ISSC, 2000; Cook et al., 2002a). See Figures V-9 to V-12 of technical document.
Comment 8. The assumption that the tdh gene is the principal marker for pathogenic V. parahaemolyticus is conservative and does not take into account that in certain areas where tdh-positive isolates were being found, there were no illnesses reported.
FDA Response 8. The thermostable direct hemolysin (TDH) is a proven virulence factor (Nishibuchi et al., 1992) and occurs in over 90% of clinical strains in the U.S. (Daniels et al., 2000a; Okuda et al., 1997a) (Unpublished CDC data) and internationally (Miyamoto et al., 1969; Nishibuchi et al., 1985; Wong et al., 2000). Nearly all West Coast isolates that possess trh also possess tdh (DePaola et al., 2000). FDA data from the U.S. Gulf Coast and a nationwide retail survey indicated that over 95% of all environmental V. parahaemolyticus in the U.S. that possess tdh also have trh, but these isolates account for less than 50% of the recent clinical isolates (FDA/ISSC, 2000; Cook et al., 2002a). While it is clear that a small percentage of the V. parahaemolyticus isolated from clinical cases of illnesses are strains with trh but not tdh, it is uncertain that these strains caused the illnesses. Even if they did, it also is uncertain whether a combination of these genes increases V. parahaemolyticus virulence.
Comments on the Modeling Techniques
Comment 9. It is troubling that the quantitative risk assessment and modeling is based on only one study, that of DePaola et al., 1990.
FDA Response 9. In order to address this particular concern, additional studies bearing on the estimated relationship between V. parahaemolyticus densities and water temperature have been evaluated and incorporated into the model. One of these studies, the ISSC/FDA V. parahaemolyticus harvest study, was ongoing at the time the risk assessment was initiated. In the ISSC/FDA study, samples were collected nationwide with the exception of the Pacific Northwest. Unpublished data on V. parahaemolyticus densities in the Northwest from 1997 through 2001 were also provided to the V. parahaemolyticus Team by Washington State authorities (WA State Department of Health, 2002a). These data were also analyzed to better quantify the apparent differences in the V. parahaemolyticus harvest densities in the Pacific Northwest compared to other regions of the country, particularly the Gulf Coast. The Washington State data were previously excluded from consideration due to the apparent effects of intertidal exposure on the V. parahaemolyticus densities in collected samples. Therefore, a subset of the Washington State samples corresponding to predominantly dredged areas were evaluated with respect to predicting V. parahaemolyticus levels in submerged oysters, prior to intertidal exposure effects (DePaola et al., 2002).
Comment 10. On page 32 of the draft assessment, the Pacific Coast Shellfish Growers are credited with stating that shellfish go into refrigeration post-harvest within a maximum of four hours. To clarify: while many growers on the West Coast can and do meet this standard, this should not be construed as the norm. There are situations in large bays with extended boat travel requirements (Willapa Bay, for example) or remote harvest locations where time from harvest to refrigeration may be significantly longer than this. More accurately, the assessment should reflect the growers on the West Coast meet the time/temperature requirements of the National Shellfish Sanitation Program.
FDA Response 10. Based on this information, we have remodeled the Pacific Northwest using a minimum time of 2 hours, a maximum of 11 hours and a mean of 8 hours for time-to-refrigeration that is still well within the NSSP requirements (see Table III-7 in the technical document).
Comment 11. A very significant portion of the shellfish cultured on the West Coast is harvested at low tide. However, intertidal exposure of oysters to ambient air temperatures is not reflected in the draft risk assessment.
FDA Response 11. The effect of intertidal harvest is included in our remodeling efforts for the West Coast. A collaborative study with FDA, Washington State and ISSC in August of 2001 generated data indicating significant increases in V. parahaemolyticus levels during intertidal exposure (DePaola et al., 2002). These data along with data from an ISSC funded study at University of Washington (Herwig and Cheney, 2001) were used to model the effects of intertidal exposure on V. parahaemolyticus levels. Washington State data indicate that V. parahaemolyticus levels at harvest in Willapa Bay, where most oysters are not exposed at low tide, are generally below detectable levels.
Comment 12. A decrease in the growth rate of V. parahaemolyticus during a cool-down (initial refrigeration or icing) of molluscan shellfish was modeled. This should be verified by collaborative scientific studies based on measurements of the actual growth rate of a tdh+ V. parahaemolyticus population in naturally contaminated (preferable) or inoculated oysters.
FDA Response 12. A direct measurement of the growth rate of pathogenic V. parahaemolyticus during the cooldown process was not undertaken. There is likely to be considerable variation in the temperature and storage conditions of oysters under commercial conditions. Consequently a direct measurement of the growth rate under one or several sets of specific and controlled refrigeration conditions does not fully determine variation in growth likely to occur under commercial conditions. Validation of assumptions underlying the predictions of growth during cooldown were addressed by measuring oyster temperature during cooldown and, in a separate experiment, measuring the growth rate of tdh+ and tdh- strains in broth culture at 25° C (77° F) (Cook, 2000a).
Comment 13. Remodel b-Poisson dose-response curve using b parameters to obtain a b-distribution, so that each individual eating occasion will have an individual likelihood of illness based on dose-response selecting from the b-distribution.
FDA Response 13. Although the simulation of the Beta-Poisson dose-response within the current assessment might be modified to incorporate the implied variation of risk according to the "exact " Beta-Poisson model, evaluation of the impact that this modification would have on the assessment suggests that it would be minimal. The principle outputs of the assessment are the uncertainty distributions of total number of illnesses across different region and season combinations. As a consequence of the Central Limit Theorem and the relatively large number of servings involved, the mean risk per serving (rather than variation of risk) is what is of particular relevance. Variability is important to the extent that it impacts mean risk per serving and the additional variation in risk implied by the exact Beta-Poisson model would need to be heavily asymmetric or skewed about the median in order to impact the mean risk per serving (and consequently total number of illnesses) compared to that of the approximate model. We expect that such skewness is unlikely to be substantial and would therefore have little impact on the uncertainty distributions of total number of illnesses relative to identified uncertainties (e.g., other dose-response models). This expectation was evaluated by conducting simulations.
In conducting these simulations the implications of the exact versus approximate models was made assuming that parameter estimates obtained by fit of the approximate model applied to both. This is not strictly correct, as discussed by Furumoto and Mickey (1967), but parameter estimates corresponding to the exact Beta-Poisson model per se could not be readily identified. The results of the simulations indicated that, at the (relatively high) levels of exposure estimated to occur, there is no appreciable difference between using the approximate rather than the exact Beta-Poisson model. The document has been revised to more clearly indicate that the approximate Beta-Poisson model was utilized and that this implies less variation (in individual risk) than that of the exact model.
Comments on Intervention strategies
Comment 14. The draft Risk Assessment identified several possible interventions that might be used to control or reduce the level of V. parahaemolyticus in shellfish, including reducing time-to-refrigeration, mild heat treatment, freezing, hydrostatic pressure, depuration, irradiation, and relaying. However, only three of these mitigation strategies were actually evaluated in the Risk Assessment, and none of the three interventions on which the draft Risk Assessment focuses are appropriate for use by retailers to enhance the safety of raw molluscan shellfish. The Risk Assessment, in citing a variety of studies, dismisses depuration as ineffective at reducing V. parahaemolyticus in oysters. The West Coast industry believes refrigerated wet storage should be investigated as a means of reducing V. parahaemolyticus post harvest and instead of being dismissed, become a priority for research.
FDA Response 14. The 2004 risk assessment focuses on the degree of reduction in the levels of V. parahaemolyticus in oysters. The results demonstrate that any mitigation strategy that reduces the level of V. parahaemolyticus in oysters also reduces illness (Chapter VI: What-If Scenarios). The predicted reduction in illness depends on the level of V. parahaemolyticus reduced in oysters. In general, as V. parahaemolyticus levels are reduced, there is a subsequent reduction in the predicted number of illnesses. Different intervention/ mitigation strategies produce different levels of reduction. We have provided some more commonly used mitigation strategies as examples of the different effects on the levels of V. parahaemolyticus. However, by no means do we imply that these are the only strategies that are effective.
Comment 15. It is premature to consider intervention strategies as part of the risk assessment modeling at this time.
FDA Response 15. We do not agree that it was premature to consider intervention strategies as part of the Risk Assessment. Evaluation of mitigation strategies is an important component of process pathway risk assessments. The second objective of the risk assessment is to evaluate the likely public health impact of different control measures, including the efficacy of current and alternative microbiological standards.
Comments on Data Used
Comment 16. The prevalence of tdh+ V. parahaemolyticus strains in the Pacific Northwest was based on a total of only 25 composite oyster samples from 2 studies. This sample size is small, therefore at least 2 more years of data on the percent of pathogenic V. parahaemolyticus (tdh+) and specific serotypes of tdh+ isolates should be collected in a national collaborative study like the FDA-ISSC survey (FDA/ISSC, 2000) of shellfish from each of the five geographic regions used in the risk assessment models.
FDA Response 16. We have used data from more recent studies in the Pacific Northwest and in the Gulf Coast in the current version of the model (DePaola et al., 2002; Kaufman et al., 2003). There were approximately 60 samples analyzed in each study for the prevalence of both total V. parahaemolyticus and tdh+ strains. These data are substantially more detailed than in previous studies (where isolates were typically pooled over multiple samples). The data was found to be adequate to statistically estimate both the mean relative prevalence of tdh+ and the variation of the relative prevalence of tdh+ from one sample to the next, for both the Pacific Northwest and the Gulf Coast.
Comment 17. The data in Table III-4 summarize the minimum, maximum and mean lengths of oyster harvesting in different regions during different seasons. It is unclear whether FDA assumed that the harvesting duration was a distribution of the harvest times from both the pre- and post-NSSP time-to-refrigeration requirements. Only the data from the post requirement period are relevant since these requirements are now mandatory.
FDA Response 17. Our assumptions concerning the length of harvesting times are more clearly described in the current version of the risk assessment document. The distributions do reflect the (self-reported) changes in harvesting evident in the dealer survey data after the post- NSSP refrigeration requirements took affect, but only with respect to those regions and seasons for which the mean water temperature is high enough for the requirements to be applicable. For the colder region/season combinations, not substantially effected by the post-NSSP time-to-refrigeration requirements, the dealer survey data corresponding to pre-NSSP time-to-refrigeration requirements were assumed to apply. Regarding the West Coast, it was our impression from information obtained previously that the maximum length of harvest time was 4 hours. As mentioned above, based on comments received in response to the risk assessment, we have since revised the assumptions to reflect the NSSP requirements appropriate to the West Coast.
Comment 18. The risk assessment did not appear to consider the possible immunological effects of oyster consumers' exposure to low levels of new or virulent strains over time and whether that might subsequently reduce the number and severity of illnesses over time.
FDA Response 18. We have found no evidence that eating raw oysters increases immunity to V. parahaemolyticus illnesses. FDA encourages the submission of data to support this assertion. The risk assessment is consistent with the CDC's definition of the risk group for gastroenteritis caused by V. parahaemolyticus, i.e., all persons (see Disease Information via www.cdc.gov ).
Comment 19. If consumer advisories about the risks associated with the consumption of raw molluscan shellfish are at all effective, then the population of consumers of raw molluscan shellfish should not be growing at the same rate as the general population.
FDA Response 19. Consumption of raw oysters was estimated based on oyster landings data, expert opinion on the percentage of the total landings consumed raw and estimates of the mean serving size obtained from a telephone survey conducted in Florida. A point estimate of consumption was obtained using average landings data from 1990 through 1998. Over this period of time, yearly oyster landings have fluctuated somewhat with a modest increasing trend. We have used the point estimate of past consumption as an estimate of current (and near-future) consumption. We do not have information necessary to investigate the potential effectiveness of education on the change in the number of consumers of raw oysters.
Comment 20. It is not clear how or if the effects of differing levels of virulence in particular strains of V. parahaemolyticus, may have been incorporated into the risk assessment.
FDA Response 20. A basic assumption of the risk assessment is that only tdh+ strains are virulent and that all strains possessing this characteristic are equally virulent. Although experimental studies suggest that additional pathogenic factors may modulate the virulence of tdh+ strains, these have not been incorporated into the present assessment. However, even with the assumption that all pathogenic strains are equally virulent there is structural (model) and parameter uncertainty associated with the estimated dose-response. These uncertainties are substantial and are a consequence of the limited data available with human subjects. Although differing levels of virulence associated with additional pathogenic factors potentially increase variability and the uncertainties associated with the output distributions for probable number of illnesses, the effect may be relatively small given the dose-response uncertainties already identified and incorporated into the assessment.
Comment 21. The risk assessment was not able to estimate an infective dose that might cause illness in the consumers of raw oysters.
FDA Response 21. As stated above, the dose-response model reflects the uncertainty and variability associated with an infective dose. Typically, data were used to estimate distributions rather than point estimates, and consequently, our results are in the form of distributions reflecting both uncertainty and variability. The available feeding studies in human subjects were evaluated to estimate the dose-response associated with pathogenic V. parahaemolyticus administered with antacid to healthy subjects. Epidemiological rates of illness in the U.S. population, probable rates of underreporting and model-based estimates of exposure were then considered to determine the likely effect of the food matrix and host factors on the dose-response. For the dose-response models that were considered there is no infectious dose level per se above which the rate of illness is 100% and below which the rate of illness is 0%. A step function dose-response implied by a single infectious dose level was considered implausible and was not evaluated. Moreover, it has been assumed that some V. parahaemolyticus serotypes such as O3:K6 require a lower dose to cause illness than other strains (Daniels et al., 2000b). Nevertheless, infectious dose was estimated in the sense that for each dose-response model considered an estimate of the dose associated with 50% probably of illness was obtained as well as the doses associated with other probabilities of illness. The uncertainties associated with these estimates were also determined. In a report by FAO/WHO (2003), mechanistic considerations of the probable independent action of bacterial pathogens imply dose-response relationships that are linear at low dose (i.e., no threshold levels).
Comment 22. On page 20 of the draft risk assessment, there is an apparent contradiction in the risk assessment, which estimates that the average percentage of V. parahaemolyticus that is pathogenic relative to total V. parahaemolyticus on the West Coast is ~3% and that the average percentage of pathogenic V. parahaemolyticus in the Gulf Coast and other areas of the country is 0.2 to 0.3%. This supposed high presence of virulent V. parahaemolyticus would seem to suggest the West Coast should have the highest incidence of illness, yet this appears to be contradicted on page 62 where the report finds that, based on the Beta-Poisson model, the largest numbers of projected illnesses were attributable to Gulf Coast product.
FDA Response 22. Although there is a higher percentage of pathogenic V. parahaemolyticus on the West Coast, there is a lower incidence of total V. parahaemolyticus in comparison to the Gulf Coast. The difference in total V. parahaemolyticus levels is a consequence of lower water temperatures and higher salinities on the West Coast. The low incidence of illness estimated in the original draft risk assessment was probably due to the shorter harvest time assumed in the model previously, as well as the failure to take the effects of intertidal harvesting into consideration. In the current risk assessment version, we have extended the harvest time to up to 11 hours and have included modeling of intertidal harvesting, which has resulted in an increase in incidence of illness closer to that reported to the Washington State Department of Health. However, risk is still lower than on the Gulf Coast because lower water temperatures (and total V. parahaemolyticus levels) compensate for the higher percentage of pathogenic V. parahaemolyticus. If all other factors such as prevalence of total V. parahaemolyticus, water and air temperature, harvest and post harvest practices, etc. were equivalent among the regions, then more illnesses would be expected to occur with the Pacific Northwest harvest than that of the Gulf Coast.
Comments on Transparency
Comment 23. The use of complex mathematical models prevents all but the most knowledgeable risk assessors from completely understanding the degree to which uncertainties in the assessment affect the outcome.
FDA Response 23. We agree and that is why FDA issued an "Interpretive Summary " of the risk assessment in conjunction with the technical document. This interpretive summary includes the essential elements of the risk assessment in a manner that can be understood by non-scientists. It states simply why the risk assessment was conducted, what was required of the risk assessment team and what was done to address these requirements, what the results were, and what these results signify. We have also attempted to explain the uncertainties as clearly and simply as possible. In addition, we provide more in the way of technical discussions related to the modeling and statistics in four appendices (3-6) in order to make the calculations more transparent. Some, but not all, technical discussion, figures and tables were moved from the document to the appendices to make the main text more readable.
Comment 24. The draft document, on page 38, appears to erroneously associate 23 cases of Vibrio parahaemolyticus related illnesses to the consumption of raw molluscan shellfish harvested in New York State in 1998.
FDA Response 24. We have corrected the numbers in the document
Return to Table of Contents