View Report Details

Microbiological Risk Assessment for Norovirus infection - Contribution to the overall burden afforded by food-borne infections

Project Code: B12001

31/01/2004

HPA Porton Down
Lawrence, L ; Kerrod, E; Gani, R; Leach, S

Gastrointestinal viruses, especially noroviruses (NVs [formerly Norwalk-like viruses – NLV]) are considered to make a huge contribution to the burden of Infectious Intestinal Disease (IID) in the United Kingdom and elsewhere. A number of routes of infection are thought to be significant, including person-toperson and foodborne transmission. The relative importance of the different routes is not clearly defined and significant data gaps regarding disease transmission exist, making a full Quantitative Microbial Risk Assessment (QMRA) potentially difficult. The Food Standards Agency considers it essential that the contribution made by the food chain be assessed relative to other sources of infection. In order to achieve this the Microbial Risk Assessment (MRA) team - Health Protection Agency, Porton Down, have conducted a feasibility study for a QMRA to establish the contribution made by the food chain to NV infection relative to other pathways.

This document gives a summary account of the project outputs (taken from previously circulated reports and workshop outputs – refer to section 3.0 for list of reports) detailing the risk assessment process from the development of a conceptual framework and the discrete risk modules highlighting the risk posed from significant modes of transmission to the organisation of information sources.

In order to accomplish the aim of the study, as stated above, the key objectives (refer to section 1 for full list of project aims), have been to provide (i) the assembly of all the known sources of virus and related issues for NV disease into a coherent framework in relation to current knowledge in virology, epidemiology, food microbiology, etc., with all qualitative and quantitative information sources assembled in a dedicated database containing appropriate summaries and keywords to facilitate retrieval (ii) development of a holistic qualitative MRA framework for NV depicting an overall risk flow and flow chart breakdown by risk modules, showing linkages and input parameters, separated into rational and manageable risk modules, with inputs and outputs identified that are conceptually compatible (iii) preliminary mathematical analysis of the assembled quantitative data for NV.

The first phase of the project involved the collation and critical appraisal of all available data on NV, sourced through published literature, unpublished reports and data obtained from a multitude of sources, including both published and unpublished scientific literature and reports. All information sources have been assembled in a bespoke bibliographic database (shown below) developed using Microsoft Access. Users are able to search the database by matching search terms against either keyword(s), title or author columns. As this is a stand-alone application it can be installed on as many computers as required.

Further to this, the most relevant qualitative and quantitative data sources have been assembled in a comprehensive annotated bibliography, compartmentalised into appropriate divisions pertaining to norovirus, norovirus disease and routes of transmission, specifically highlighting in this instance, useful data related to the focussed set of transmission routes addressed in this study (person-to-person, salad vegetables and fruit, and bivalve shellfish) (refer to appendix 1). Alongside this, where insufficient data exist for norovirus, temporary alternative data (e.g. potential surrogates microorganisms) have also been sought. Each entry is accompanied by a synopsis of the key objectives of the literature and points to potentially useful quantitative data relating to norovirus or potential surrogates.

In addition to the data gathering and appraisal, great emphasis has been placed on the input of expert knowledge from the project Steering Group and Food Standards Agency representatives, facilitated from workshops, questionnaires and consultations. These participants comprised a range of national and international disease experts in the field of epidemiology, food microbiology and risk modelling. Further to this, input was also enlisted from consultations with other experts to provide specific evidence or data. Such data sourcing has facilitated the development of a holistic qualitative model framework that encompasses the significant routes of infection in the United Kingdom and identifies the associated flows of risk. The model framework has been structured in a flexible manner to allow for the provision of additional “modules” as knowledge regarding NV disease improves.

Significant data gaps have been identified during the framework development with many deficiencies largely attributed to the lack of a diagnostic technique able to detect live virus, and the absence of a useful animal model; this was further compounded, until recently, by the lack of simple and sensitive diagnostic techniques. However, a potential solution to such limitations in data availability could be the employment of survival data from surrogate cultivatable viruses that may reasonably reflect NV behaviour. For these reasons, potentially useful surrogate data have also been reviewed and assembled in the reference database.

The overall framework of NV burden has then been segregated into discrete risk modules highlighting the risk posed by other pathways, particularly person-toperson – the most commonly recognised mode of NV transmission. Furthermore, epidemiological data and expert opinion has highlighted two other significant routes of NV transmission, via the foodborne route. These food vehicles are bivalve shellfish, and salad vegetables and fruit (representing the raw, ready-to eat food group). Schematic risk flows of these three discrete elements depict the known or suspected individual risk factors (hazard identification and exposure assessment) to human health from NV and their likely outcomes (hazard characterisation), producing discrete but inter-related risk modules.

The SEIR person-to-person NV transmission model presents a basic schematic framework that depicts the dynamics of NV spread within the population. From contact with an infectious individual, those susceptible (S) individuals that have been exposed will either contract the disease (responders) moving into the exposed (E) but not yet infectious category, or for those who do not contract the disease return to the susceptible category (non-responders). After a short incubation period the exposed individuals will become infectious (IA/IS) themselves (assumed that asymptomatic [IA] and symptomatic shedding [IS] behaviour is similar). After a short duration of illness (in the symptomatic sufferer) they will recover (R) with individuals entering the immune category where they remain until they become susceptible again once immunity wanes. The model’s flexible framework means that it can be modified for different scenarios including “exogenous” inputs such as those from contaminated foods.

The salad vegetables and fruit risk flow framework addresses not only the sources of NV input(s) from farm to fork but also those elements that could affect viral persistence (i.e. rate of decline) and thus the potential risk to the consumer. These elements include contamination of crops by waters contaminated with human waste (e.g. run-off, wastewater and sludge, use of contaminated irrigation or washing water, use of unclean water in food preparation or cleaning dishes). However, a significant burden of contamination to this food group arises from NVinfected individuals, particularly those working in the latter stages of food preparation within the catering environment, such as food handlers or cooks.

Similarly the bivalve shellfish risk flow framework has been segregated into sources of NV input(s) from farm to fork; addressing the elements that could affect viral persistence. However, unlike the salad vegetables and fruit module, contamination is less likely to come from direct transmission from an infectious food handler but indirectly, principally from the contamination of human sewage pollution entering into the bivalve shellfish habitat. To compound this contamination problem, current protocols for post-harvest purification processing (depuration) are usually inadequate at purging virus from the bivalve shellfish. Moreover, contamination can also occur during seafood processing due to inadequate hygiene practices or systems. Such issues are further compounded by the habit of consuming some bivalve shellfish (e.g. oysters) raw or lightly cooked, as normal cooking practices generally inactivate the virus. This emphasises the importance of avoiding the initial contamination of bivalve shellfish in their shellfishery beds.

Following the initial qualitative assessment, all available quantitative information was collated and evaluated for preliminary mathematical parameterisation to be performed. As was anticipated at the inception of this feasibility study, the collation and evaluation of all the necessary quantitative features for even the refined set of transmission routes highlighted above has demonstrated that there are a significant number of critical gaps. Therefore, fully quantifying the flows for the risk of infection through each of these different routes would be inherently difficult and most likely require further laboratory, field and epidemiological research at a large number of points from production to consumption in order to ensure that any in-depth study of the risk implicating NV in the burden of disease is adequately addressed. Furthermore, due to the complexity of the relevant parts of the food chain in each case, it is likely to make it difficult to generalise specific details in a quantitative sense even across individual food items within each category (e.g. different shellfisheries, fruit and vegetable items).

Considering such constraints, a more focussed and useful approach of the quantitative phase of the project has been two-fold. Firstly, to parameterise the person-to-person model in order to assess the potential impact at varying degrees of foodborne transmission input on the general incidence. Secondly, to identify specific key points in the risk modules, that if parameterised from future research would yield the most direct estimates of foodborne transmission (see below). These two aspects could then be combined to show how foodborne transmission could be estimated following the completion of a set of given recommendations.

In spite of the lack of critical data in many key areas, sufficient data exist to begin to enable the person-to-person model to be parameterised. Infections in humans would appear to be essential to the long-term maintenance of the virus (since there is no convincing evidence of a significant zoonotic reservoir for human NV). Furthermore, without exogenous inputs (i.e. infected food vehicles) into an otherwise person-to-person transmission cycle, each infection in the population must result in at least one further infection if the virus is to persist (i.e. reproduction number, R₀, of the virus must be greater that 1). With this basic epidemiological constraint in mind it has been useful, therefore, to estimate the basic reproductive number of the virus (with respect to person-to-person transmission) in the UK general population with a view to clarifying the overall disease burden in the UK. This in turn has enabled an estimate to be derived to scope the possible impact from other routes of NV contamination (i.e. indirect transmission from the foodborne route) that would impact on the total incidence of NV in the general population.

A theoretical limit has been calculated for the number of exogenously generated cases required for the disease to be maintained in the population (denoted θ). If the true number of exogenously generated cases is greater than θ their elimination would result in the disease no longer being endemic, as this will imply that the true value of R₀ is less than one. If the number of exogenously generated cases is lower than θ then elimination of exogenously generated cases will decrease overall incidence in the general population but R₀ will remain greater than one and thus the disease will remain endemic. Either way, estimating the actual value of θ (or indeed R₀ more directly for person-to-person transmission) would allow us to ascertain the impact from the removal or reduction in foodborne transmission on the incidence of disease in the general population.

In order to estimate θ it was first necessary to estimate the transmission rate that was required to produce the observed incidence in the general community representing both endogenous and exogenous transmission routes, denoted R₀’. This was met by integrating a set of first-order differential equations that describe the NV disease dynamics between different states of infection (thus allowing us to calculate the number of people in each of the given states), applying Euler’s method, and then running the person-to-person model until the state variable no longer changed. Hence, using the model, the value of R₀’ that produced the observed number of NV cases per year was derived. The next step was to estimate a value of θ from R₀’ by splitting R₀’ into two parts to account for the endogenous transmission thus generating the maximum number of exogenous cases that must be generated in order for transmission to be maintained with the community without foodborne contamination.

Initial estimates of R₀ and θ estimates were derived using data taken from the IID study which suggested that the NV incidence within the general population in England was 3.7 million cases per year. Using this incidence rate, a R₀ of 1.31 and a θ value of ~230,000 cases in England per year were derived. However, expert opinion suggests that the NV incidence rate taken from the IID study has a significant degree of under-ascertainment due to the diagnostic techniques employed at that time. Preliminary results of re-analysed samples from the IID study, performed by members of the Steering Group, using more sensitive diagnostic techniques suggest the incidence rate is more likely to be ~10 million cases per year in England. These refined incidence values derived an R₀ of 1.43 and a θ of 1.8 million cases in England per year. However, uncertainty exists over the true incidence rate of NV within the population. Given this uncertainty, sensitivity analysis was performed on varied incidence rates. In addition to this, the period of immunity conferred post-infection is also under debate, and therefore sensitivity analysis was also performed to assess the potential impact of the length of immunity on these parameter estimates.

Thus, if the general incidence of NV in the population is 10 million cases per year, then the maximum annual number of foodborne cases required for endogenous transmission cycle to be maintained, (denoted θ) is 1.8 million cases in England per year. If the number of foodborne transmission events is indeed 1.8 million then their elimination would reduce R₀ for NV to <1 and conditions would exist for the disease to be no longer endemic. Whilst it is virtually impossible at this time to independently measure θ, we can at least calculate upper limits for oyster and food handler transmission and compare them to θ by following the steps outlined below. Thus if interventions are applied to change the number of foodborne cases per year, their impact on overall incidence rates can be calculated.

With the data constraints/gaps discussed in this and previously circulated reports (refer to section 3.0) in mind, this feasibility study has identified a number of endogenous transmission quantification strategies. Such strategies would assist in the derivation of estimates of the NV burden afforded from contaminated oysters (UK-produced and imported) at point-of-sale and from food handlers in the catering environment, thus clarifying the burden afforded by NV from these prominent routes of foodborne transmission. A number of potential areas for further focused work have been identified and are summarised below:

Dietary surveys on raw and cooked oyster consumption
Point-of-sale survey of oysters for NV contamination
Survey of catering staff for NV infection and determination of work practices when they are infected
Quantify the number of cases generated per infected food handler through epidemiological studies of existing outbreak data
Studies to determine levels of asymptomatic shedding of virus and infectivity
Follow clains of infection originating as outbreaks into the general population through seeking out infected contacts of cases to estimate the true community transmission rate excluding foodborne contamination i.e. an estimate of true R₀ for person-to-person transmission that is independent of estimates of the total disease burden

Main File

Some of the files on this site may be in a format that your computer can't read. However, you can download Readers and Viewers for the following document types below: