LITERATURE REVIEW OF QRA FOR FOODBORNE VIRUSES
Published quantitative risk assessment dealing with viruses are given in the Table 3. The aim of this list is not to be exhaustive but to include main papers about virus risk assessment. There’s no probabilistic risk assessment for shellfish viral risk, no complete risk assessment using second order risk modeling.
The few number of QRA for viruses is probably due to the lack of reliable dose-response for the main pathogen involved in food (NoV-HAV), the lack of the reliable and inexpensive way to measure the dose of pathogen (not indicator or surrogates) (RT-PCR technique were expensive), and the lack of feasibility of measurement of infectivity for Nov and HAV (EFSA, 2011, 2012; FAO/WHO; 2008).
Before the publication of Teunis (2008), the dose-response for norovirus or all viruses was substituted the use of Rotavirus dose-response (Regli et al., 1991, Rose and Sobsey, 1993), and for HAV by the use of the surrogate Echovirus 1, 2 (see part III.2).
The question of aggregation or clumping of viruses is treated in some studies, more often with non homogenous distribution, negative –binomial use to describe the contamination in food matrix (Westrell et al., 2006).
Then our work proposed in the quantitative risk assessment for HAV in shellfish is the second one dealing with shellfish since 1993 (Rose and Sobsey, 1993), using probabilistic, second order, risk modeling. Moreover, the proposed work includes the quantitative estimate of efficiency of different monitoring /management strategies of shellfish areas of production, in particular to avoid viral contamination in oysters, on the relative risk for human consumers, which was not done before.
PRESENTATION OF THE CONTEXT OF PAPER
The objective of the work is to investigate the relative efficiency of monitoring and management systems in case of contamination of shellfish by HAV. The method used is second order risk assessment, in order to propagate separately variability and uncertainty.
This subject emerged in France after two shellfish borne (highly suspected) outbreaks that occurred in the same area in Brittany in 1999 and 2007 (Guillois-Becel et al., 2009). In order to prevent possible next outbreaks, a regular monitoring, based with HAV genomes, was done in the area of shellfish-fishing of the shore. Without any specific regulation, no particular monitoring was done for shellfish production, except the regular microbiological exams. In this context the signification of HAV genome detection and its effectiveness for monitoring and management purpose was asked by the ministry of agriculture to ANSES (ANSES, 2010).
The work realized for the group of expert was an opportunity to begin and illustrate the usefulness of the PhD work. However, this kind of question is not specific to the particular context in France, and in particular in Brittany, but could also be raised in other detected contaminated areas, in particular with HAV, such as in Australia or in the South of Italy (Puglia area) (Conaty et al., 2000; Lopalco et al., 2005).
Baseline model parameters for a Quantitative Risk Assessment approach.
Four steps are necessary to assess a risk: (1) estimation of the contamination in food products; (2) estimation of consumption patterns; (3) exposure assessment by combining (1) and (2); (4) dose-response assessment. Parameter definition and the distribution and modeling of each variable are summarized in Table 4. For one individual and for a given day of the year, dose exposure can be evaluated by the product of oyster contamination times the total amount of edible oyster tissue eaten. The lapse of time between harvesting and consumption is most often short in coastal areas close to the production area and was not taken into account.
Estimation of contamination with a realistic number of genome copies per gram shellfish digestive tissue was done using two hypothetical scenarios, because observed quantified data of contamination of shellfish by HAV are rare. Values taken from an outbreak linked to coquina clams in Valencia, Spain, corrected for extraction and enzyme efficiencies, were around 230-1800 copies per gram of digestive tissue (23). The maximum uncorrected value observed in Brittany with monthly sampling, is around 1970 copies. Values corrected for extraction efficiencies (between 50% and 10%) and for enzyme efficiencies (around 90%) lie between 4,400 (1970*100/50*100/90) and 22,000 (1970*100/10*100/90) copies in grams of digestive tissue (1). Monitoring of E. coli in either scenario gives results in agreement with European directives for Class B areas.
Oyster contamination scenarios.
The first scenario considers very short-term contamination, such as sudden and heavy rainfall and contaminated coastal land run-off, brief episodes involving the treatment of contaminated raw sewage, sewer overflows, or occasional contamination linked to tourist activities. Initial contamination, tides, currents, and environmental factors all contribute to the final concentration in oysters and are specific to each coastal area and situation. We chose not to set values of contamination from the land-based source, but set them directly in oysters with range values in the same order of magnitude as real observations (1, 23). We used initial contamination and maximum values of 25,000 HAV genome copies per gram digestive gland (Figure 6(A)). For E. coli, initial contamination and maximum values as tolerated for rare incidents in Class B areas are 46,000 genomes copies per100 g of oyster flesh (Figure 6(B)). Two incidents of less than 24 hours with fecal HAV contamination of the shellfish production area were set to occur during a winter and a summer period. We assumed that the maximum contamination level in shellfish is reached in 24 h or less, in particular for HAV (1). Then both values decay in agreement with their T90 (i.e. time necessary to inactivate 90 % of the original amount) of 28 days for HAV (17) and two days for E. coli (24). For E. coli, five days after the incident, concentrations per 100 g of edible flesh are described by a uniform distribution between 230 and 4,600 genome copies. This contamination scenario is plotted in Figure 6(A) (HAV) and in Figure 6(B) (E.Coli).
Model development and simulations
For each simulation of uncertainty, one sample value of the uncertain distribution of parameters was sampled. One thousand simulations were done for uncertain parameters: alpha, beta (parameter of dose-response) and the first date of sampling during the first period (month) of sampling. Then, depending on management strategy, every 30 days or 15 days, a sampling day was chosen for each month of the year. For each simulation of uncertainty, the annual probability of illness for each individual was calculated according to the formulae given in Table 1. Variability was taken into account for the population of 1000 consumers as was variability of contamination during the year. For each uncertainty simulation, the mean annual risk for this population was evaluated.
If a management strategy is applied, the theoretical annual exposure and risk of each individual might change. Depending on the risk management strategy chosen, if there is abnormal level of E. coli (monitoring E. coli) or detection of HAV (monitoring HAV), there is no consumption of contaminated shellfish after a given period, for a given period (no exposure) of the year, for the entire exposed population. For two log10 units of lower contamination of oysters or relaying 15 days in a clean area, oysters are less contaminated, therefore, exposure is lower. We assumed that closure and reopening of the shellfish production area had no effect on shellfish consumption after the date of re-opening. The benefit of human intervention can be demonstrated by the risk reduction, 1 minus the ratio of mean risk with intervention to mean risk without intervention. The mean of the relative risk reduction is expressed in percentage, i.e. % of cases avoided, with the median and the 95% credible interval. Producers’ cost is contingent upon the duration of closure of the area and can be simulated for each risk management option. All calculations and simulations were made in R language (version 2.12.2, R Foundation for Statistical Computing).
The dose–response was taken from published paper (Pinto et al., 2009). Dose in outbreaks is evaluated in this paper, with consumption of 60g, and for values of observed concentration corrected for extraction and enzyme efficiency of Real-Time RT-PCR, expressed in infectious dose (1 genome among 60 is considered to be infectious) and after light cooking. Different predicted attack rates were evaluated with mode values of parameters of some viral dose-response, with an approximate Beta-Poisson model (for theoretical aspects of dose-response see part III). Those attack rates were compared to observed attack rates.
Reproducing results of the paper (Pinto et al., 2009), the comparison of observed and predicted attack rates were plotted on the Figure 11.
Table of contents :
Chapter I: GENERAL INTRODUCTION
I.1. Public health impact of foodborne viruses
I.1.1. Identification of foodborne viruses
I.1.1.Estimate of number of cases linked to specific viruses
1.1.2. Attributing illness to food source
1.1.3. Public health impact estimates of foodborne viruses
I.2 Transmission of foodborne viruses
1.2.1. Pathways of transmission of foodborne viruses
1.2.2. Relative importance of different food products in the foodborne pathways
1.3. Biological characteristics of norovirus and hepatitis a virus.
1.3.1 Biological Characteristics of norovirus
1.3.2. Biological characteristics of hepatitis a virus
1.3.3. Comparison between HAV and NoV biological characteristics
1.4. Conceptual Framework and modeling approach
1.5. Overall aim of this study
Chapter II: QUANTITATIVE RISK ASSESSMENT OF FOOD BORNE VIRAL CASES
II.2. Literature review of qra for foodborne viruses
II.3. Qra for hepatitis A virus
II.3.1. Presentation of the context of paper
ii.3.2. Published paper
II.3.3. Complement to the paper
Chapter III: DOSE-RESPONSE, CRUCIAL STEP OF FOODBORNE TRANSMISSION
III.1. Main dose-response models for qra purpose
III.1.1. Dose-response models: theoretical and biological meanings
III.1.2. Key parameters estimate
III.1.3. Limitations of dose-response modeling
III.2. Literature review of dose-response for foodborne viruses
III.3. Norovirus dose-response based on shellfish outbreaks data
III.3.1. Submitted paper
II.3.2. Complement of the paper
Chapter IV: DYNAMIC MODEL FOR COMPREHENSIVE ANALYSIS OF THE FOODBORNE AND NOT FOODBORNE TRANSMISSION PATHWAY
IV.1. Dynamic models used in epidemiology: biological and theoretical meanings
IV.1.2. Example of a stochastic process with continuous time: Gillespie algorithm
IV. 1.3. Parameter estimates and sensitivity analysis
IV.2. Literature review of published dynamic models with food-borne transmission
IV.3.Dynamic model of norovirus cases in a coastal area
4.3.2. model structure
V/ GENERAL DISCUSSION-CONCLUSION
5.1. Major findings
5.2. Limitations of this work and perspectives