Risk assessment and predictive modelling a review of their application in aquatic animal health Workpackage 2 Deliverable 2.1

Transcription

1 Risk assessment and predictive modelling a review of their application in aquatic animal health Workpackage 2 Deliverable 2.1 The DIPNET project is funded under the EU Framework Programme 6 priority 8 Scientific Supports to Policy (SSP).

2 This report has been carried out with financial support from the Commission of the European Communities, specific RTD programme Specific Support to Policies, SSP Disease interactions and pathogen exchange between farmed and wild aquatic animal populations - A European network. It does not necessarily reflect its views and in no way anticipates the Commission s future policy in this area. Title: Risk assessment and predictive modelling a review of their application in aquatic animal health Authors: E.J. Peeler (Cefas), A.G. Murray (FRS), A. Thebault (AFSSA), E. Brun (NVI), M.A. Thrush (Cefas), A. Giovaninni (IZAM) Issued by: VESO Project number: SSP Scientific coordinator: Laurence Miossec, Ifremer Project manager: Åse Helen Garseth; VESO Date: September Sponsor: European Commission Sponsor s reference: SSP Contact person: Jacques Fuchs Availability/ISBN: ISBN Number of pages: 67 (Including appendix 1 and 2) Number of attachments: 2 Keywords:

3 Summary: In the field of aquatic animal health, risk assessment has been applied mainly to examine the risks of disease spread through international trade (import risk analysis IRA). In general the approach to risk assessment recommended by the Organisation des Épizooties (OIE) has been most frequently used. In the current report, published IRA have been reviewed. Only a minority of the studies reviewed included a consequence assessment. Few aquatic animal IRA discuss disease interaction and pathogen exchange between wild and farmed populations within consequence assessment. Apart from IRA, risk assessments of the spread of Gyrodactylus salaris have demonstrated how risk assessment can effectively support decision making. The work on Gyrodactylus salaris also provides the best illustrations of how risk assessment can examine disease interaction between wild and farmed fish. Risk assessment has been little used to study disease spread in shellfish; therefore, this review has focused on identifying questions that could be addressed by risk assessment, and corresponding data requirements. Models used in the study of terrestrial livestock disease are a useful starting point for aquatic animals, however, the transmission of disease through the water column means that the principles of terrestrial animal models need to be reconsidered. The epidemiology of sealice infections has been the main application of mathematical modelling, with a focus on transmission from wild to farmed salmonids. Modelling sealice transmission has illustrated how mathematical models can provide useful insights into disease interaction without an overarching description of the system and its interactions; however, ecological models are needed for environmental effects of disease interaction to be better assessed. A lack of data has constrained the application of mathematical modelling. Both risk assessment and mathematical modelling have potential to improve our understanding of disease interaction and pathogen exchange between wild and farmed populations. The consequence assessment part of IRA should explicitly include disease interaction. Methodologically, a great use of mathematical modelling and economic methods is required to improve the assessment of consequences of disease introduction, including between wild and farmed populations.

4 Table of Contents 1 Introduction Risk assessment What is risk assessment The application of risk analysis to animal health Qualitative versus quantitative approaches Introduction Qualitative risk assessment Quantitative risk assessment Conclusion Recommended approaches to risk assessment for animal health Introduction NAS-NRC model Covello and Merkhofer model International Plant Protection Convention (IPPC) Import risk assessment methodology Overview Risk communication Hazard identification Risk assessment Release assessment Exposure assessment Consequence assessment Risk estimation Risk management Risk management and biosecurity Hazard analysis and critical control points Risk assessments for diseases in terrestrial animals Introduction Import risk analysis Conclusion Risk assessment for diseases in shellfish Introduction Pathways of disease interaction Transmission and maintenance of a disease agent Conclusion

5 7 Risk assessments for diseases in finfish Gyrodactylus salaris Import risk assessment Published aquatic animal import risk analyses Disease emergence Modelling disease interaction between wild and farmed fish populations Introduction Models of farm-wild interactions General Principles of epidemiological modelling in aquatic environments Models of disease in farmed fish and experimental systems Models of disease in wild fish and other aquatic populations Models of disease in shellfish and other predictive modelling Discussion Applying the Covello Merkhofer disease interaction Quantitative or qualitative Consequence assessment The application of mathematical models Combining risk and mathematical models Conclusion and recommendations Acknowledgements References Appendix 1. Terrestrial Import risk analyses Appendix 2. Review of published risk analyses for the transmission of aquatic pathogens

6 1 Introduction In this review, risk assessment methodology, import risk assessments (concerning terrestrial and aquatic animal) and other applications of risk assessments in aquatic animal health have been reviewed. The application of mathematical modelling to the study of disease interaction between wild and farmed fish populations has also been considered and published literature reviewed. The potential for both approaches to further our understanding of disease interaction and pathogen exchange is discussed. 2 Risk assessment 2.1 What is risk assessment Risk analysis methods were developed by the nuclear and space industries to assess the likelihood and probability of undesirable events, known as hazards. It was subsequently adopted by the chemical and petrochemical industries. More recently it has been applied to biological systems. It is important to distinguish the two elements that are combined to make a risk: i) the likelihood and ii) the consequences of the hazard occurring. Some risks will have a very low probability of occurrence but severe consequences (e.g. an accident at a nuclear power plant). The advantages of risk analysis are that it is comprehensive, transparent and produces defensible results. It forces a thorough and logical approach to be adopted in considering the likelihood and consequences of undesirable events. Risk analysis is fundamentally a tool to assist decision-making. 2.2 The application of risk analysis to animal health In recent years, risk analysis has been applied in the field of animal health, particularly in the area of food safety (microbiological risk assessment) and import risk analysis (IRA). The main application of risk analysis in the aquatic animal health field has been IRA, which is the assessment of disease risks associated with international trade in animals and their products. The Agreement on the Application of the Sanitary and Phytosanitary Measures (SPS Agreement) of the World Trade Organization (WTO) has stimulated the application of risk analysis and resulted in significant improvements in the methodology as applied to international trade policy for animals and animal products. A key principle of the SPS agreement is that there should be no restrictions (sanitary measures) on trade in animal and animal products, unless it is likely that the trade will result in disease spread with unacceptable consequences. In the past the risk of disease introduction has been used to block trade in order to protect local industries. The SPS agreement allows the importing country to 3

7 impose measures that reduce the risk to an acceptable level; however, an IRA is required to justify measures that exceed internationally agreed standards. 2.3 Qualitative versus quantitative approaches Introduction There are two main approaches to risk analysis, qualitative or quantitative (1). Descriptive terms are used in qualitative analysis (e.g. risk of introduction is negligible ) and in quantitative analysis by risk rates (e.g. current levels of salmon imports will result in the introduction of disease X once every 1,000 years). The initial step for both approaches will be the similar: the question must be clearly stated and the construction of a scenario tree of events necessary for the hazard to occur (2). Figure 1 illustrates a generalised framework for a scenario tree. Figure 1. Generalised framework for a scenario tree (3) A scenario tree is a graphical depiction of the biological pathways by which a hazard might be introduced into an animal population and provides a useful conceptual framework. It is tool for clarifying and communicating the logic behind the risk assessment. A scenario tree assists in conveying the range and types of pathways considered (3). It is generally recommended that a qualitative analysis be undertaken prior to a full quantitative analysis. Quantitative analyses require more data time and manpower, compared 4

8 with a qualitative approach and, therefore, are generally reserved for the most important routes of introduction (frequently identified by the qualitative analysis). A decision tree can be used to determine when a quantitative risk assessment should be undertaken (see Figure 2.). Figure 2. Decision tree for determining when to undertake a quantitative risk analysis qualitative assessment reveals no significant risk no yes STOP qualitative results sufficient for decisionmaking no yes STOP resources (time, personnel, funds) are available yes data are available yes no no STOP STOP PERFORM QUANTITATIVE ANALYSIS Qualitative risk assessment A qualitative assessment is essentially a reasoned, systematic and logical discussion of the relevant contributory factors and epidemiology of a hazard in which the likelihood of its release and exposure and the magnitude of its consequences are expressed using nonnumerical terms such as high, medium, low or negligible (3). Adjectives commonly used and their definitions drawn from the Concise Oxford Dictionary are the following: High: Extending above the normal or average level Significant: Noteworthy; important; consequential Average: The usual amount, extent, rate Low: Less than average, coming below the normal level Remote: Slight, faint Insignificant: Unimportant; trifling Negligible: Not worth considering; insignificant 5

9 2.3.3 Quantitative risk assessment Monte Carlo simulation is used in quantitative risk analysis to assimilate the probability components of the import scenario. Software programmes Palisade Inc.) have been developed within a spreadsheet environment for Monte Carlo simulation. The uncertainty associated with an input, and its known variability are modelled as a probability distribution. The type of distribution used will depend on the nature of the input and the available data (1) (e.g. parameters estimates based on expert opinion are generally modelled as Pert distributions). The simulation randomly samples from the distributions of all the inputs. This process is repeated thousands of times and thus an empirical distribution that resembles the theoretical probability distribution of the model output is achieved (4). A paper by Paisley et al. (5) probably best illustrates the data requirements for a quantitative fish disease risk analysis. Quantitative risk assessment is able to model uncertainty 1 (lack of precise knowledge), true variability (inherent to all biological systems) and undertake sensitivity analysis (to determine the relative importance of variation in different inputs on outputs) Conclusion Clearly, quantitative analysis requires personnel with a high level of numerical and modelling skills. Equally, it could be argued that qualitative analysis requires the risk analyst to bring a higher level of knowledge and familiarity of the subject compared with a quantitative analysis. Qualitative risk assessments use verbal descriptors of risks and severity, and often involve the aggregation of expert opinions. An efficient mechanism to elicit the systematic collection and assessment of expert opinion for a qualitative risk analysis is the appointment of a Delphi panel (7). In this case, the methodology for qualitative risk assessment must be rigorous to ensure that the real risk, and not the arbitral risk perception, is assessed. To this end, the independence and transparency of expert opinion is important and terms used to describe risk must be defined. If the definitions used in a qualitative assessment are given a numerical range (e.g. greater than 50%) the assessment moves towards a quantitative approach (sometimes referred to as semi-quantitative). 1 Uncertainty is the assessor s lack of knowledge about the parameters that characterise the physical system that is being modelled. It is sometimes reducible through further measurement or study, or through consulting more experts [Vose p196 (6)] 6

10 A lack of data, particularly for wild fish, has been cited as a major problem for import risk analysis for aquatic diseases (8). Expert opinion can be used to fill gaps in the published data (e.g. to derive a dose response curve in the absence of scientific data). Methods have been developed to improve the collection of expert opinion. The Delphi Method is based on a structured process for collecting and distilling knowledge from a group of experts by means of a series of questionnaires interspersed with controlled opinion feedback. It has been used in aquatic animal health (9). However, a lack of data should not be considered an obstacle to risk assessment:..if we know everything, quantitative risk analysis is not useful, because we have the answer, if we know nothing, nothing can be done (D Vose, pers.com.). It can be difficult for non-mathematicians to understand the methods and results used in quantitative risk analyses; qualitative risk analyses are generally easier to follow. The importance of the distribution around the mean result can be difficult to communicate to both policy makers and the general public. If the parameters with wide confidence limits have been used, due to uncertainty, the output will also have wide confidence limits. Ultimately, it may be argued that quantitative results with very wide confidence limits are of little more use than qualitative results. From a methodological viewpoint, qualitative risk analysis has some limitations: there are no well-accepted methods from combining separate qualitative assessments of risk at the different stages of the analysis and sensitivity analysis is not possible. For both approaches, methods to combine the likelihood of risk occurrence (a probability) with consequences (often measured by economic impact) are not well developed. 3 Recommended approaches to risk assessment for animal health 3.1 Introduction Risk assessment has been widely used in a number of areas that has given rise to a different a variety of terms and definitions. However, most of the approaches subdivide risk assessment into four components: Hazard identification, Risk assessment, Risk management Risk communication. 3.2 NAS-NRC model The first risk assessment framework to be developed was the NAS-NRC model (10). The NAS-NRC system was developed to set maximum limits of chemical substances in the environment, food, etc. The risk assessments undertaken using this system were therefore 7

11 designed to answer the question: What is the maximum amount of substance (or pathogen) to which a person should be allowed to be exposed from a particular source? Thus the framework used in this model is designed as a regulatory tool for setting allowed, acceptable or tolerable levels of contaminants and pathogens in food, and is the system most frequently used by toxicologists (3). The Codex Alimentarius Commission of the Food and Agriculture Organisation (FAO) and the World Health Organisition (WHO) use NAS-NRC model (11) terminology. Figure 3 illustrates the model s structure. The risk assessment is divided in the following phases: The hazard identification is the identification of biological, chemical and physical agents capable of causing adverse health effects and which may be present in a particular food or group of foods. The hazard characterization is the qualitative and/or quantitative evaluation of the nature of the adverse health effects associated with biological, chemical and physical agents which may be present in food. For chemical agents, a dose-response assessment should be performed. For biological or physical agents, a dose-response assessment should be performed if the data are obtainable. The exposure assessment is the qualitative and/or quantitative evaluation of the likely intake of biological, chemical and physical agents via food as well as exposures to other sources if relevant. The risk characterization is the qualitative and/or quantitative estimation, including attendant uncertainties, of the probability of occurrence and severity of known or potential adverse health effects in a given population based on hazard identification, hazard characterization and exposure assessment. 8

12 Figure 3. The structure of the NAS-NRC risk analysis process (3). This scheme may be applied for both qualitative and quantitative risk assessments, but it is more suitable for a quantitative approach. The components of the risk management are: The risk evaluation is the identification of a food safety problem, the establishment of a risk profile and ranking of the hazard for risk assessment and risk management priority. The option assessment is the identification of available management options and the selection of preferred management option, including consideration of an appropriate safety standard. This step includes weighing various health risks along with economic, political and social factors. The monitoring and review is the assessment of effectiveness of measures taken and the review of risk management and/or assessment as necessary. 3.3 Covello and Merkhofer model The guidelines for import risk assessment published in the OIE terrestrial and aquatic animal health codes (12), are based on the Covello and Merkhofer model (13). It is designed to assess the actual magnitude of the risk for specified consequences in a given situation. It can then be used to decide whether the risk is acceptable as it stands, or whether sanitary measures are required to reduce the risk to an acceptable level (see section ). Risk assessments using this system are designed to answer the question: What is the likelihood of specified 9

13 consequences occurring as a result of exposure to a particular substance or pathogen that came from a defined release source? This system is more versatile than the NAS-NRC system and can be applied qualitatively or quantitatively to various risk questions, making it the system of choice for many risk assessors in animal health (3). Hazard identification is the first step and considered separately from the risk assessment. Risk assessment is further subdivided into stages i) release assessment (description of pathways necessary for the introduction of the hazard), ii) exposure assessment (description of pathways necessary for the hazard to occur following introduction, e.g. exposure of aquatic species in the importing country to the introduced exotic pathogen), iii) consequence assessment (identification of the consequences - the adverse human health, animal health, economic or environmental effects of interest- of disease introduction and establishment), and iv) risk estimation (integration of the release, exposure and consequence assessments) (14) (see Figure 4.). Risk management and risk communication combined with risk assessment and hazard identification are referred to as risk analysis. 10

14 Figure 4. Structure of the Covello and Merkhofer risk analysis process Hazard identification Risk assessment: -release -exposure -consequence -risk estimation Risk management: -risk estimation -option evaluation -implementation -monitoring & review risk communication 3.4 International Plant Protection Convention (IPPC) A third approach to risk assessment has been developed by the International Plant Protection Convention (IPPC) 2, which is referred to as pest risk assessment (PRA). The objectives of a PRA for a specified area, are to identify pests and/or pathways of quarantine concern and evaluate the risk presented, to identify endangered areas and, if appropriate, to identify risk management options. The steps in the process are similar to those described in the Covello and Merkhofer model, with the main exception being that the IPPC (like NAS-NRC) includes pest categorisation (equivalent to hazard identification) within risk assessment, rather than as a separate procedure. 4 Import risk assessment methodology 4.1 Overview IRA are usually conducted for trade or legal reasons, generally in response to a proposed importation of an animal commodity. In conducting an import risk analysis, a number of important steps must be worked through in a systematic manner, while keeping the assessment as simple as possible. According to the OIE Handbook on Import Risk Analysis for animals and animal products (3), these include: determine the scope of the risk analysis state the question to be answered clearly and explicitly assemble the team

15 develop a risk communication strategy determine the information required determine the approach o determine what information is available for each step in the assessment o identify the populations of interest o estimate the likelihood of the hazard(s) being imported o estimate the likelihood of susceptible animals or humans being exposed to the hazards o estimate the likely consequences of susceptible animals or humans being exposed to the hazards o decide whether risk management measures are warranted examine the risk management strategies available formulate a programme of risk management measures. document the assumptions, evidence, data and uncertainties for each variable consider how the data and the results should be presented to facilitate communication commission a peer review of the risk analysis, and address input 4.2 Risk communication The OIE code defines risk communication as.. process by which information and opinions regarding hazards and risks are gathered from potentially affected and interested parties during a risk analysis, and by which the results of the risk assessment and proposed risk management measures are communicated to the decision-makers and interested parties. Risk communication is an essential element of risk analysis and should be conducted throughout the risk analysis process. Risk communication provides an opportunity for groups and individuals who have an interest in the results (i.e. stakeholders) to comment before decisions are made. Guidelines on IRA recommend that the scope of the project be initially discussed with stakeholders. Stakeholders should also have an opportunity to discuss the preliminary results and recommendations (including risk mitigation measures) before a final report is produced. The responses received should be collated and a formal response made. One benefit of risk communication is that whilst some stakeholders may not agree with the outcome of the process, they should have a good understanding of the basis on which the decisions were made. An effective risk communication strategy should minimise disputes about the measures required to manage the risk. The main problems with risk communication are the perception of risk by the general public (especially unrealistic expectations and a desire for certainty), a lack of interest in technical complexities and a difficulty in understanding probabilistic information. 12

16 Effective communication between risk assessors and risk managers is essential when defining the task for the assessment, during the analytical process and when presenting the results to ensure the success of the project. This communication is important for the assessor to fully understand the task and to enable the assessor to provide supportive and clear information to the risk managers in their weighing of policy alternatives. However, close cooperation should not to hamper the objectivity of the assessment. 4.3 Hazard identification The first step in any risk analysis process is the hazard identification. According to the OIE Aquatic animal health code (12), hazard identification involves identifying the pathogenic agents that are capable of producing adverse consequences associated with the importation of a commodity. The hazards identified would be those capable of infecting or contaminating the species being imported, or from which the commodity is derived, and which may be present in the exporting country. It is then necessary to identify whether each hazard is already present in the importing country, and if so, if it is an OIE-listed disease or is subject to control or eradication in that country. Hazard identification is a categorization step, identifying biological agents dichotomously as hazards or not hazards. The risk assessment should be concluded if hazard identification fails to identify hazards associated with the importation. The OIE list provides an obvious starting point when developing the list of hazards, but pathogens not included in the OIE list should also be considered, where appropriate. Each pathogenic agent should be dealt with separately with a reasoned, logical and referenced discussion of its relevant epidemiology including an assessment of its likely presence in the exporting country (3). This often requires knowledge of not only the potential pathogens for the commodity in the country of origin, but also on a world-wide basis. This is because the health status of the commodity in the exporting country will often be poorly reported, and specific diseases are often difficult to detect due to an absence of a reliable diagnostic test or their low prevalence in the population. The sources of information should be comprehensively documented 3. The end result of the hazard identification process will be a list of hazards (pathogens) of concern (Table 1). If the hazard analysis determines that the commodity, due to its nature, 3 Manual on risk analysis for the safe movement of aquatic animals (FWG/01/2002)

17 origin, the processing methods used or another factor, contains no hazards, then the risk analysis process is terminated. 14

18 Table 1: Results of a hypothetical hazard identification exercise and additional disease information required for it in a risk analysis involving the proposed importation of Crassostrea gigas for aquaculture OIE listed and other significant pathogens of concern for Crassostrea gigas and/or Crassostrea spp. are identified as follows: Haplosporidium nelsoni Perkinsus marinus Perkinsus olseni Bonamia ostreae Herpes virus Oyster velar virus disease Other diseases considered to be significant to the importing country which are present in the exporting country and not present in the importing country, or if present, subjected to disease control measures Additional disease information: Diseases recorded from C. gigas and/or Crassostrea spp. from the exporting country Diseases recorded from C. gigas and/or Crassostrea spp. from the importing country Significant diseases of Crassostrea spp. Other diseases listed by the importing country Other diseases significant to the importing country The steps to define if a potential hazard is an actual hazard are listed in Table 2. Table 2: Steps to determine if a pathogenic agent is a hazard (3) 1. If the commodity is a potential vehicle for the pathogenic agent proceed to step 2, otherwise the pathogenic agent is not a hazard 2. If the pathogenic agent is exotic to the importing country but likely to be present in the exporting country it is classified as a hazard Note: If an exporting country claims that it is free of a particular hazard, supporting evidence must be documented. In such cases the appropriate sanitary measure to be applied is certification from the Veterinary Administration or Competent Authority in the exporting country that it is free from the hazard. 3. For a hazard reported in both the exporting country and the importing country IF (a) there are documented free zones or zones of low prevalence in the importing country, OR (b) the pathogenic agent is subject to an official control programme in the importing country, OR (c) there is a more virulent strain in the exporting country, THEN the pathogenic agent may be classified as a hazard. The final result of the hazard identification in the IRA on non-viable salmonids and nonsalmonid marine fish (15) undertaken by Biosecurity Australia is summarized in Figure 5. 15

19 Figure 5: Summary of the hazard identification linked to the import of non-viable salmonids 16

20 4.4 Risk assessment The Covello and Merkhofer risk analysis process (13), splits the risk assessment into the following steps: Release assessment Exposure assessment Consequence assessment Risk estimation Release assessment According to the OIE Aquatic animal health code (12), release assessment consists of describing the biological pathway(s) necessary for an importation activity to release (i.e. introduce) a hazard into a particular environment, and estimating the likelihood of that complete process occurring. The release assessment describes the likelihood of the release of each of the hazards under each specified set of conditions with respect to amounts and timing, and how these might change as a result of various actions, events or measures. Examples of the kind of inputs that may be required in the release assessment are: a) Biological factors: Species, strain or genotype, and age of aquatic animal, Strain of agent, host range and prevalence Prepatent period Tissue sites of infection and/or contamination, Vaccination, testing, treatment and quarantine. b) Country factors: Incidence/prevalence, Evaluation of Competent Authorities, surveillance and control programmes, and zoning systems of the exporting country, in particular: o Evaluation of diagnostic tools (sensitivity, specificity) and combinations o Frequency and importance of transfers and imports o History of regular monitoring c) Commodity factors: Whether the commodity is alive or dead, Quantity of commodity to be imported, Ease of contamination, Effect of the various processing methods on the pathogenic agent in the commodity, Effect of storage and transport on the pathogenic agent in the commodity. If the release assessment demonstrates no significant risk, the risk assessment need not continue. Each hazard should be dealt with separately with a reasoned, logical and referenced discussion. The results of a hypothetical release assessment are summarized in Table 3. 17

21 Table 3: Results of a hypothetical release assessment for two identified hazards involving the importation of live marine molluscs for aquaculture development a) Haplosporidium nelsoni Definition of populations in consideration: exportation of Crassostrea gigas for disease risk of crassostrea gigas in importing countries, with marine waters and zootechnical practices similar of French Atlantic or Mediterranean coasts Natural hosts: Crassostrea virginica, C. gigas Global distribution: USA, Japan, Korea and France Occurrence: Reported from exporting country; not reported from importing country. Release assessment: low (with good confidence); juveniles originating from indoor facility where life cycle is not known to be fulfilled; pathogen occurring at a very low level of prevalence in exporting country (<1%); pathogen reported from adults but never from juveniles. b) Herpes virus of oyster Natural hosts: C. virginica, C. gigas, Ostrea edulis, O. angasi, Tiostrea chilensis,ruditapes philippinarum, R. decussatus Global distribution: reported from France, USA, Taiwan, New Zealand and Australia. Occurrence: Reported from exporting country; not officially reported from importing country. Release assessment: high (reasonably certain); agent initially described from hatching facilities; recurrent report from exporting country monitoring in oysters. In a qualitative risk assessment, the likelihood of event occurring is expressed in descriptive terms. Consequently, the choice of adjectives used to qualify the degree of likelihood (high, moderate, low, negligible, etc.) should be very careful and refer to standardized definitions. A release assessment scenario tree for finfish carcasses imported for human consumption is provided in Figure 7 (15). 18

22 Figure 7: Scenario tree for release assessment showing pathways followed by a product imported for human consumption (15) Exposure assessment According to the OIE Aquatic animal health code (12), exposure assessment consists of describing the biological pathway(s) necessary for exposure of humans and aquatic and terrestrial animals in the importing country to the hazards and estimating the likelihood of these exposure(s) occurring, and of the spread or establishment of the hazard. 19

23 The likelihood of exposure to the hazards is estimated for specified exposure conditions with respect to amounts, timing, frequency, duration of exposure, routes of exposure, and the number, species and other characteristics of the human, aquatic animal or terrestrial animal populations exposed. Examples of the kind of inputs that may be required in the exposure assessment are: a) Biological factors: o Presence of potential vectors or intermediate hosts, o Genotype of host, o Properties of the agent (e.g. virulence, pathogenicity and survival parameters). b) Country factors: o Aquatic animal demographics (e.g. presence of known susceptible and carrier species, distribution), o Human and terrestrial animal demographics (e.g. possibility of scavengers, presence of piscivorous birds), o Customs and cultural practices, o Geographical and environmental characteristics (e.g. hydrographic data, temperature ranges, water courses). c) Commodity factors: o Whether the commodity is alive or dead, o Quantity of commodity to be imported, o Intended use of the imported aquatic animals or products (e.g. domestic consumption, restocking, incorporation in or use as aquaculture feed or bait), o Waste disposal practices. If the exposure assessment demonstrates no significant risk, the risk assessment should conclude at this step. Figure 8 is a simplified, generic scenario tree for assessment of the likelihood that susceptible hosts would be exposed to a pathogen present in an imported batch of an aquatic animal. 20

24 Figure 8: Scenario tree for assessment of the likelihood that susceptible hosts would be exposed to a pathogen present in an imported batch of an aquatic animal (5). 21

25 4.4.3 Consequence assessment According to the OIE Aquatic animal health code (12), consequence assessment consists of identifying the potential biological, environmental and economic consequences of exposure to a hazard. A causal process must exist by which exposures to a hazard result in adverse health, environmental or socio-economic consequences. Examples of consequences include: a) Direct consequences: o Aquatic animal infection, disease, production losses and facility closures, o Adverse, and possibly irreversible, consequences to the environment, including spread of disease between wild and farmed populations. o Public health consequences. b) Indirect consequences: o Surveillance and control costs, o Compensation costs, o Potential trade losses, o Adverse consumer reaction. The consequences to animals, people, the environment and the economy may be direct and indirect, and the probability of a particular outcome will be determined by factors associated with establishment and spread of the disease, assuming exposure of susceptible animals (3). Each hazard should be dealt with separately with a reasoned, logical and referenced discussion to: i) estimate the likelihood that at least one animal becomes infected ii) identify the biological, environmental and economic consequences associated with the entry, establishment or spread of the hazard, and their likely magnitude iii) estimate the likelihood of the occurrence of these consequences. Note: A causal relationship must exist between exposure to a hazard and an adverse effect. The risk analysis may be concluded at this point if direct or indirect consequences are not identified. Epidemiological results may be expressed as Risk Ration or Odds Ratio to evaluate consequence assessments (these data must be included in a risk analysis when available). In order to evaluate the likely magnitude of the consequences, and the likelihood that they will occur at any given magnitude, the risk analyst may identify and describe a small number of outbreak scenarios. The relative likelihood of each these occurring can then be estimated, 22

26 along with the likely magnitude of the consequences in each case. For example, outbreak scenarios might include: o o o o disease does not establish within the exposed population disease establishes within the exposed population, but is quickly identified and eradicated disease establishes within the exposed population and spreads to other populations before eventually being eradicated disease establishes within the exposed population, spreads to other populations (including wild fish populations) and becomes endemic. In a qualitative risk analysis, the impact at each level can be described in terms such as negligible, moderate, significant or severe. When considering the consequences of a disease outbreak, the risk analyst may need to consider the persistence of its effects (3). Once an introduced disease is established in a wild population, eradication is extremely difficult, if not impossible. A summarized consequence assessment is described in Table 4. Table 4: Summary consequence assessment for the import of a stock of imported fish. Scenario Likelihood of scenario Type of consequence Likelihood of consequence Significance of consequence at a national level 1. No infection High Not applicable as no adverse consequences of scenario of local fish 2. Infection of one local farm 3. Infection of one river system 4. Infection beyond the local river system Very low Very low Negligible Biological High Negligible Environmental Negligible Negligible Economic High Very low Biological High Low Environmental Negligible Moderate Economic High Low Not taken further as negligible likelihood of scenario Risk estimation Risk estimation consists of integrating the results of the release assessment, exposure assessment, and consequence assessment to produce overall measures of risks associated with the hazards identified at the outset (12). Thus risk estimation takes into account the whole of the risk pathway from hazard identified to unwanted outcome. In the risk estimation, each hazard should be dealt with individually, summarizing the results and/or conclusions arising from the release, exposure, and consequence assessments to estimate the likelihood of the 23

27 hazard entering the importing country, becoming established or spreading and resulting in adverse consequences. It is not sufficient to conclude that there is a possibility of entry, establishment or spread or that there may be consequences. Risk estimation requires that probabilities (from release and exposure assessments) are combined with consequences that may be estimated in monetary terms. 24

28 4.4.5 Risk management The Covello and Merkhofer risk analysis process (13), splits the risk management into the following steps: Risk evaluation Options evaluation Implementation Monitoring and review Risk evaluation and the appropriate level of protection (ALOP) According to the OIE Aquatic animal health code (12), risk evaluation is the process of comparing the risk estimated in the risk assessment with the Member Country s acceptable level of risk. The acceptable level of risk associated with the importation of a commodity is defined as the risk judged by an importing country to be compatible with the protection of public and animal (terrestrial and aquatic) health (16). The Sanitary and Phytosanitary (SPS) agreement of the World Trade Organisation (WTO) recognises that governments have the right to provide the level of protection that it deems appropriate. Setting the acceptable level of risk is, therefore, a political decision, which is appropriate since each decision will result in benefits or costs to one group over another (17). However, the SPS agreement requires that the acceptable level of risk is based on an objective and scientific assessment of the probability and likely consequences of introduction. The acceptable level of protection can be considered as the measures necessary to reduce the assessed level of risk to an acceptable level (17). The SPS agreement allows measures to be provisionally adopted if insufficient scientific information exists on which to evaluate the risk. However, additional information must be sought within a reasonable period of time. The two components of a risk analysis process represented by risk estimation (the last part of the risk assessment) and risk evaluation (the first step of risk management) are intermingled and in the case of qualitative risk assessment are carried out together. A clearer distinction between the two phases is possible (and often needed) in case of a quantitative risk assessment. A matrix can be used to combine the probability of introduction and establishment with consequences (i.e. risk estimation) and use the output to determine whether the risk is acceptable. An example is given in Figure 9 which was designed for the Import Risk Analysis on Non-viable Salmonids and Non-salmonid Marine Finfish produced by the Australian 25

29 Quarantine and Inspection Service (15) and should not be considered (as any other matrix) a general evaluation criterion applicable in all situations. Figure 9: Risk evaluation matrix (15) The risk evaluation process is performed for all the hazards identified and the summary results shown in Figure 10 are those for only one of the diseases analyzed. The very final result of the process is a summary table with the results for all hazards (Figure 11) Option evaluation and equivalence According to the Aquatic animal health code (12), option evaluation is the process of identifying, evaluating the efficacy and feasibility of, and selecting measures to reduce the risk associated with an importation in line with the Member Country s appropriate level of protection. The efficacy is the degree to which an option reduces the likelihood and/or magnitude of adverse health and economic consequences. Evaluating the efficacy of the options selected is an iterative process that involves their incorporation into the risk assessment and then comparing the resulting level of risk with that considered acceptable. The evaluation for feasibility normally focuses on technical, operational and economic factors affecting the implementation of the risk management options. The results for the option evaluation process in the example on the import of non viable salmonids to Australia (15) are summarized in Figure 11. The concept of equivalence is embedded in the SPS agreement. It is important that countries are consistent in applying the acceptable level of risk to different 26

30 commodities. Australia was forced to change import restrictions on imports of salmon products because they were found to be inconsistent with the restrictions imposed on other fish products that carried similar risk (18). Figure 10 Summary results for risk assessment and risk evaluation for salmonid diseases (15) 27

31 Figure 11: Risk management options chosen to mitigate the risk of establishment of Renibacterium salmoninarum in Australian salmonids populations (15) Implementation According to the OIE Aquatic Animal Health Code (12), implementation is the process of following through with the risk management decision and ensuring that the risk management measures are in place Monitoring and review According to the OIE Aquatic Animal Health Code (12), monitoring and review is the ongoing process by which the risk management measures are continuously audited to ensure that they are achieving the results intended. Two aspects of the review need to be considered: scientific review and updating of risk analyses. Risk analysis is based in science, and so risk analyses should be subjected to a process of peer review. To ensure the technical robustness of the analysis it should be subject to a process of: Internal scientific review within the Competent Authorities External scientific review by selected experts with specialized knowledge in risk analysis and its application to the diseases under consideration 28

32 This review process ensures that the risk assessment can withstand criticism by stakeholders opposed to importation or in favor of unrestricted importation, as well as potential challenge within the WTO dispute settlement system (3). Risk analyses need to be periodically updated. The risk associated with importation of animals or animal products is dynamic. Factors which affect the previously determined risk may vary over time. Most of this fluctuation is accounted for within the risk analysis process. However, there are several factors of major importance (such as changes in the animal disease status of the exporting country, neighboring countries, or regions, or the amount of commodity imported) that may have an immediate impact on the risk, and these should be monitored. In addition there are specific factors associated with each risk analysis that may need to be reviewed periodically because of their potential effect on the resultant estimate. Specific factors that may need to be monitored periodically can be identified through the process of risk analysis. Those steps in the importation process, which incorporate the greatest uncertainty or have the greatest impact on the risk estimate, should be monitored Risk management and biosecurity Risk assessment had mainly been applied at a national or regional level to assess the risks of disease introduction. However, risk assessment approaches can also be applied at the farmlevel (19). The risk management element of the assessment can take the form of a biosecurity programme (a management strategy to minimise the risk of disease introduction). A quantitative risk analysis will quantify the risks of disease introduction and establishment presented by different routes, whilst a qualitative analysis will lead to a ranking. Both approaches allow the priorities for biosecurity programmes to be identified. In many instances a number of aquaculture facilities use the same water resource. Many aquatic animal pathogens are able to survive outside their host and can be transmitted between farms by currents, tides or wild fish. Therefore, the development of regional biosecurity programmes is important. Outbreaks of ISA in Scotland and IHN in Canada have resulted in regional biosecurity plans. In Scotland area management plans are based on fundamental aspects of oceanographic conditions (e.g. tidal excursions) and specific local conditions (20) Hazard analysis and critical control points Hazard analysis and critical control points (HACCP) is a systematic approach to the identification and evaluation of food safety hazards, and the mitigation of associated risks. It is a quality assurance method to focus on the most important steps and/or bottlenecks in 29

33 various operations and procedures along the entire production line. Indirect or direct measures of the items when passing these critical control points work as a warning system for whether corrective action (risk management) has to be undertaken. HACCP is commonly used within food safety, but has not been more widely applied within livestock production. This may be due to difficulties in defining efficient control points. However, in many parts of the aquaculture production there are successive events influenced by temperature, oxygen, salinity and so on which may be seen as critical control points /risk factors, as well as inputs to a quantitative risk analysis of the production. Such an analysis may be useful for estimating total risk profiles in an operation or highlight the most hazardous link(s) in a chain. HACCP should be viewed as method that complements other risk management approaches. 5 Risk assessments for diseases in terrestrial animals 5.1 Introduction The promulgation of the SPS agreement led to an increased application of import risk analysis and to significant improvements in the methodology of risk analysis as applied to international trade for animals and animal products. However, there was very little development of risk analysis in veterinary fields other than international trade and management of health risks to consumers of animal products and little has been published on its use in the evaluation and choice of control or prophylaxis strategies for animal diseases (1). Nevertheless, it is apparent that the techniques of quantitative risk assessment could have a wide range of animal health applications. Risk assessment is a tool for decision-making in the face of uncertainty and provides numerical estimates of the probabilities and consequences associated with particular scenarios. Risk assessment allows a quantitative evaluation and comparison of various scenarios ranging from no safeguards to combinations of various safeguards. It facilitates the communication of risks and their consequences to stakeholders and decision makers and allows decision makers to choose appropriate safeguards in a transparent fashion. In the past few years there has been a growing interest for risk analysis including reassessment of control or prophylactic strategies for animal diseases. Since the year 2002 the control strategy for bluetongue (an infection of domestic ruminants) in Italy has been based on a risk assessment (21) as have subsequent modifications of the surveillance strategy and 30

34 movement restriction policy (21, 22). Similarly, the discussion on the modifications to the OIE Code Chapter on Bluetongue, a new version of which has been approved in the last OIE General Session, has been based on risk assessments. 5.2 Import risk analysis Most of the risk assessments available in the literature refer to the introduction of disease through the import of animals or animal products from infected to free countries (listed in Appendix 1) or refer to the evaluation of the disease status of a country in respect to a given disease (e.g. bovine spongiform encephalopathy in Canada and tuberculosis in Australia). The majority have been produced by government agencies and are available as government reports (only few have appeared in peer-reviewed journals). The large majority of reviewed IRA originate from Australia or New Zealand, the remainder come from the UK, USA or Canada. The resources devoted to IRA in Australasia reflects the perceived importance of protecting livestock industries in these countries and a willingness to make them publicly available. Of the 41 risk assessments identified only 4 were quantitative. Indeed, virtually any risk assessment is first undertaken qualitatively, and if considered necessary, a quantitative model is developed 4 (3) (see section 2.3). 5.3 Conclusion A clear example of how the qualitative and quantitative risk assessment complement each other and the sequence of the two components is given by an assessment performed to decide the control strategy for bluetongue in Italian ruminant populations (21). The overwhelming preponderance of qualitative risk assessments makes it easier to transfer and apply to aquatic animals the experience developed in terrestrial ones. 6 Risk assessment for diseases in shellfish 6.1 Introduction The OIE International Aquatic Animal Health Code (OIE, 2004) lists the most important infectious disease agents on a certain number of criteria, including likelihood of international spread. However, quantitative import risk assessments have, to our knowledge, not yet been undertaken for shellfish diseases. Hine (23) lists possible reasons: lack of reliability of detection of mollusc pathogens, lack of knowledge about life-cycle, modes of transmission, specificity between hosts (host-range), prepatent period, infective dose, survival in 4 Manual on risk analysis for the safe movement of aquatic animals (FWG/01/2002) Accessed on July 7,

35 environment or to disinfectants, lack of knowledge about the level of prevalence, temporal and spatial variation and the influence of environmental factors (24). 6.2 Pathways of disease interaction Despite the lack of data, different pathways of disease interaction between wild and cultivated shellfish can be identified. Firstly, wild and cultivated shellfish populations must be distinguished. Shellfish can be considered as cultivated if they have been physically relocated as a result of human activity. Other differences also exist: i) the density of cultivated animals is normally higher than for wild animals, and ii) the demographic structures (age) is different in general, wild shellfish populations harbour a larger proportion of old animals. Pathways of disease interaction can be conceptualised by considering a particular location and applying three epidemiological concepts. The first epidemiological concept to consider is the epidemiological triad that describes disease as the outcome of the interaction of host, pathogen and environment (Figure 12). Figure 12. The occurrence of disease depends upon the interaction of the host, pathogen and environment. Pathogen Host Disease Environment The first point is to consider the origin of a disease agent for a population. Completely new agents are rare; disease agents generally originate from the introduction (importation) of infected animals, e.g. the introduction of Bonamia ostreae to Europe (25), or emergence or reemergence from an environmental reservoir (e.g. Marteilia refringens in flat oysters introduced into Arcachon bay, unpublished data). 6.3 Transmission and maintenance of a disease agent The second point is to consider the transmission and maintenance of a disease agent following its introduction (26). This is illustrated by two examples. 32

36 Example 1: Introduction and establishment of disease A certain number of infected shellfish are introduced into a bay. The density of infected shellfish is high compare to those of wild shellfish. Risk and mathematical modeling could address a number of issues: the probability of transmission of infection from farmed to wild population; time to disease occurrence, extent of the outbreak (including host range and geographic spread, level of mortality) and important modulating factors. the probability that the wild shellfish would survive and at what density. the probability that a reservoir of infection in wild shellfish of the same species or in wild shellfish of another species. the probability that cultivated shellfish in the same bay would be affected (when and at what distance). the likely impact of control measures. the probability of transmission to another bay or another site, with or without anthropogenic shellfish movements. Example 2: Transmission from farmed to wild shellfish What is the probability that cultivated shellfish become infected with a disease that is endemic in wild shellfish? What might be long-term impact for wild and cultivated animals? The following environmental factors may be important: density of shellfish (wild and cultured), water depth, temperature, salinity (these variables must be carefully defined, for example at what depth, with or without emersion, and frequency). The information needed for risk assessments to address the questions raised in the two examples are summarised in Table 5. 33

37 Table 5. Data required for risk assessment studies. Detection, prevalence and host range Host range (known) Host range suspected Seasonality of prevalence Factors affecting host susceptibility Prevalence in wild and cultivated populations (range) Characteristics of diagnostic tools Transmission, maintenance Life-cycle and mode of transmission Infective dose Prepatent period Survival (time )in different environmental situations Biological or physical reservoir Survival time to disinfection Knowledge about hosts Biodiversity and habitats Density wild stocks Density cultivated Environmental factors Seasonal variation of temperature and salinity Geographic and seasonal variation of phytoplankton Variation of time of emersion Depth Nature of seabed Hydrological modeling 6.4 Conclusion Risk modeling has the potential to provide insights into disease spread between wild and farmed populations of shellfish to support the development of the industry and to protect wild stocks. Qualitative risk assessments can be used to address the first and second examples cited in this paper for some diseases. Secondly, a quantitative risk assessment can be attempted to validate the qualitative approach. Examples of shellfish species and diseases (in parentheses) that might be suitable candidates are: Crassostrea virginica (MSX, Perkinsus), Crassostrea gigas (MSX), Cerastoderma edule (Trematods), Ruditapes philippinarum and decussatus (Vibrio, Perkinsus), and Ostrea edulis (Bonamia, Marteilia). 34

38 7 Risk assessments for diseases in finfish 7.1 Gyrodactylus salaris Gyrodactylus salaris is a viviparous, monogenean, freshwater ecto-parasite of Atlantic salmon whose natural host are Baltic strains of Atlantic salmon. It is regarded to be one of the major threats to the North Atlantic salmon Salmo salar, both in Norway and the UK. G. salaris may cause a fatality rate of almost 100% in Norwegian parr and severe population declines (27, 28). While the UK is free of G. salaris, the parasite was apparently introduced to Norway in early 1970s and spread to different geographical regions through stocking of rivers with infected fish (29). By 2004, a total of 45 Norwegian salmon rivers had been infected, with disastrous effects on the local stocks. For the last 20 years, annually 0-3 new rivers have been infected, a spreading primarily occurring within the proximity of a regional index river (30). The early awareness of the devastating effect G. salaris-infection on Atlantic salmon, initiated basic research which over a number of years has provided essential information for risk assessments, e.g. transmission studies, survival capacity related to temperature and salinity, fecundity, and potential hosts. However, information gaps still exist, including host behaviour. Nevertheless, a number of risk assessements on the spread of G. salaris in Norway and elsewhere have been completed. Paisley et al (1999) completed a quantitative assessment estimating the risk of introducing G. salaris from potentially infected salmon smolts transferred to seacages, to a wild salmon in a nearby river (5). This assessment became part of a decision making process concerning the consequences of future salmon farming in fjord systems. The risk of spreading G. salaris from infected farmed rainbow trout in a fresh water lake in Sweden to a down stream Norwegian salmon river was also evaluated (31). Prevalence among escapees was estimated as well as the annual probability of salmon fry in the river being infected assuming a steady state situation in the different salmonid populations. After the primary introduction of G. salaris to Norway, the parasite spread to new rivers has been regarded as a result of natural migration of infected salmon or other passive carrier fish, or by movement of infected water, including wet equipment. One of the major Norwegian salmon rivers has for more than 15 years maintained its salmon stock after the introduction of Gyrodactylus salaris, by stocking with hatchery reared juveniles up-stream of the infected stretch of river. A proportion of smolts leaving the river system are infected. The authorities 35

39 requested a quantitative assessment of the risk of infected smolts carrying the parasite to neighbouring non-infected rivers (31). The general conclusions from this work supported findings by Paisley et al (1999) (5) that the estimated risk of such transmission is most of all sensitive to salinity along the migration route. The assessment estimated that a high probability exists of G. salaris spread by smolt migration from the infected river to a nearby neighbouring river. Other risk assessment has been used to identify likely routes of introduction of G. salaris into the UK (32, 33) and uninfected territories in Europe (32-34) as well potential routes of spread between catchments for contingency planning (34) The assessment by Peeler and Thrush (2004) (32) was a qualitative, restricted IRA (i.e. based on existing risk mitigation measures) and was intended primarily to identify priorities for further in depth risk assessments and other research. The OIE IRA method was followed. The authors argued that a qualitative approach was a necessary starting point to identify the most significant routes that required further in depth investigation. The risk of introduction (release) and establishment (exposure) was evaluated separately and the combined risk assessed. A scenario tree was developed for one route of entry. No attempt was made to assess the consequence of G. salaris introduction and establishment. Effectively, a severe impact on wild stocks, due to pathogen exchange with farmed rainbow trout and salmon, was assumed based on the experience in Norway and evidence of the susceptibility of UK stocks to G. salaris (35). The risk associated with most routes was found to be negligible (i.e. too low to be significant), three routes were found to carry a higher level of risk; however, the authors pointed out that the lack of data resulted in considerable uncertainty about the level of risk. The paper concluded with priorities for further risk assessments. It could have been improved by scenario trees for the 3 routes of introduction that were estimated to have higher than negligible risks of introduction and more detailed discussion of the probabilities associated with each identified step in the tree. The paper by Peeler et al (2004) was undertaken in response to changes in EU legislation that changed the restrictions on the movement of live salmonids into G. salaris free approved zones (33). Effectively EC decision 2004/453 allowed shipments from seawater sites, given various caveats. The analysis, therefore, was sharply focused on one route of entry under particular circumstances. This analysis complements the previous IRA (32) by examining live fish introductions. The OIE IRA method was followed A scenario tree was developed to 36

40 illustrate the various routes of introduction. The risk assessment successfully addressed the key question, which was whether the new routes of entry carried a higher level of risk compared with existing routes of entry. The authors argued that the qualitative approach assisted policy makers to formulate changes in legislation and a quantitative analysis would have contributed little extra to the process. Additionally, the lack of data would have severely limited the usefulness of a quantitative analysis. In this paper the authors highlight the difference between the risk associated with a unit of the commodity in question (in this case a shipment of live fish) and the absolute risk, which is dependent on the volume of trade. From the perspective of international trade and the SPS agreement of the WTO, it is the risk per unit that has to be used when comparing commodities. However, a new commodity may greatly increase the absolute risk of disease introduction because of an increase in the volume of trade. The third risk assessment takes the OIE IRA method and applies it to the transmission of the parasite between rivers in the UK (34). The purpose of the assessment was to identify and rank all identified routes of entry to direct the development of policies to minimise the risk of spread in the event of an outbreak, and to identify further research priorities. For each route an event was defined which might lead to G. salaris transmission and the event was described by a qualitative probability. This approach allowed each route to be assessed on a number of criteria: i) number of catchments at risk, ii) the number of events per year per catchment at risk, iii) the number of parasites introduced per event. The routes were ranked in order of importance. The paper demonstrated how even in a qualitative analysis the data on which the ranking was made, can be made explicit. In methodological terms, this paper extended the qualitative approach by defining an event which may lead to G. salaris transmission. The work identified the movement of live trout as the main route of spread and this has been investigated further in a quantitative analysis (36). The paper could have been improved by the inclusion of scenario trees for the more important routes of spread. A similar analysis has been undertaken in Norway (30). To evaluate possible transmission routes between Norwegian rivers, different alternatives were assessed and ranked in a qualitative analysis. Migrating salmonids in brackish water was suggested the most important route for spread, showing an historic potential of a new river being infected every 2-4 years. Different forms of passive transmission by equipment (birds, animal (fur), boats, nets, personal belongings etc. not directly exposed to infected fish) were regarded negligible. The 37

41 conclusion was based on the assumption from literature showing that G. salaris rapidly dies if the surrounding environment dries out. The importance of human-transported volumes of water was quantitatively estimated and it was concluded that this route has negligible risk, based on the estimated density of free-floating parasites in the water column of an infected river. None of the three papers included a consequence analysis that would have dealt with the spread of G. salaris between wild and farmed fish populations. Therefore, they contributed little to our understanding of how risk assessment may provide insights into disease interaction. From a methodological viewpoint, the papers demonstrate that qualitative risk assessments can be effective in answering specific policy questions. The importance of the denominator or unit of risk has been highlighted by Peeler and Thrush (2004), (36) and should be advocated for future quantitative risk assessments. Scenario trees greatly assist in understanding disease spread by breaking the routes down into their various components and their use should be strongly encouraged in risk assessments of disease interaction. 7.2 Import risk assessment Published aquatic animal import risk analyses Risk analysis methods in the field of aquatic animal health have mainly been used to assess the risks of introducing exotic diseases into a country. Appendix 1 provides a tabular review of important published risk assessments for the transmission of aquatic pathogens covering a range of different routes and commodities, ordered by category of assessment (i.e. quantitative or qualitative). This summarises scope and format (including risk assessment components completed) and indicates where studies have addressed pathogen interaction and exchange within and between wild and farmed aquatic animal populations. Fourteen of the 23 studies assessed the spread of disease between countries, other assessment investigated within country spread. Only 5 of the studies were quantitative. The commodity studied was most frequently live aquatic animals (10 of the studies). A consequence assessment was only completed for 6 of the studies. A number of the publications are reports produced by the Australian or New Zealand Ministries of Agriculture. IRA completed by the Australian authorities (AFFA) include studies of salmon, ornamental and marine fish (a IRA for prawns is ongoing). Most of these studies were undertaken for trade or regulatory purposes and have taken the commodity as the 38

42 starting point. The best example of quantitative IRA is probably the paper by McDiarmid (37) on the risk of disease introduction with the import of ocean caught Pacific salmon. The report by Kahn et al (1999) demonstrates the depth and scale of data required for a comprehensive quantitative commodity import risk analysis. Four IRA appeared as papers given at a conference sponsored by the Office des Epizooties (OIE): Risk Analysis in Aquatic Animal Health, (38) (available from OIE). In addition to the IRA papers, others dealt with the application of risk analysis methodology to aquatic animal health. LaPatra el al. (39) attempted to isolate IHNV from rainbow trout in an endemically infected area to assess the risk of virus spread with movement of processed carcasses (the paper also appears in the OIE proceedings - (38)). However, the results are not used in a formal risk analysis to assess a probability, possibly because all the samples were negative. 7.3 Disease emergence An emerging disease is defined as a new disease, a new presentation of a known disease (e.g. increased severity, appearance in a new species) or an existing disease that appears in a new geographical area (40). Emerging diseases have become an increasingly important area of research in both human (41) and animal health (42), including aquatic animals (43). Murray and Peeler (44) applied the import risk analysis (IRA) methodology recommended by the Office International des Epizooties (14) to study disease emergence in aquaculture. A four stage model was developed: i) emergence of a pathogen (release), ii) establishment in a farmed population. iii) establishment at the larger (regional) scale, and iv) development of disease and its consequences (economic, ecological, welfare). A number of processes underlie disease emergence including international movement of live animals, climate change, intensive fish production systems and interaction between wild and farmed fish. Most new diseases of farmed fish probably existed in wild populations where they caused little or no clinical disease. Murray and Peeler (44) have demonstrated how risk analysis methods developed for one purpose (i.e. import risk analysis) can be successfully applied in a different area. The risk assessment methodology provided a framework for the process of disease emergence be broken down and logically evaluated. 39

43 8 Modelling disease interaction between wild and farmed fish populations. 8.1 Introduction Wildlife can form an important reservoir of infection for domestic livestock (e.g. bovine tuberculosis in badgers) and may be impacted by pathogen released from domestic populations (e.g. canine distemper virus infection in African hunting dogs). Modelling disease interaction between wild and farmed populations has become an important area of study in terrestrial ecosystems for issues such as agriculture and conservation biology. However, in aquatic ecosystems modelling such interactions remains rudimentary. One area in which some modelling has occurred is for sea lice, although even here models are incomplete, concentrating on exchange from farmed to wild salmonids, rather than a complete epidemiology of sea lice interaction between the two populations. Owing to the paucity of models of exchange between farmed and wild fish, in this review we extend our discussion to models of the spread of disease in single populations, first looking at general principles and then at more detailed specific models of transmission within purely cultured and purely wild populations of fish. 8.2 Models of farm-wild interactions One of the few areas in which there has been modelling of disease interactions between farmed and wild fish is in the exchange of sea lice from farmed salmon to wild salmonids. Statistical analysis of the association between farmed salmon and lice loads on wild fish has been conducted in Scottish (45) and Irish (46) coastal waters. In Scotland no evidence was found of association, while in Ireland a weak but significant positive association was detected. Modelling using observed load distributions between individual fish (47) indicated the importance of patchiness in infection pressure in producing observed load patterns. Another simple modelling approach has been to estimate production of larval lice by farmed and wild salmonids in Norwegian (48) and Scottish (49) waters. Such modelling has been used to demonstrate that a very large proportion of larval lice produced in coastal waters was from farmed fish (even under low loads per farmed fish) and thus likely to substantially alter the coastal epidemiology of sea lice, compared with the pre-aquaculture situation. Models by Asplin et al.(50) and Murray and Gillibrand (51) have used particle tracking to identify where larval lice released from farms might end up, and have identified mechanisms for the formation of concentration of larval lice which accord with observations. These 40

44 concentrations may occur many kilometres from the point of release. The location and strength of these concentrations are dependent on factors such as wind and freshwater flows. An alternative approach to modelling dispersal is to assume smooth continuous dispersal (52) which also predicts effects kilometres from the source location. Butler (50) developed simple conceptual models of farm management that could be used to ensure lice release was minimised at times of greatest ecological sensitivity, e.g. sea trout spring runs. Although there is no over-arching model of sea-lice interaction between farmed and wild fish, the results from a series of modelling approaches support each other. Analysis of load observations indicated infection is more likely to occur near salmon farms, albeit in a patchy fashion. Other modelling clearly demonstrates that the large majority of lice egg production originates from infections on farmed fish. Particle tracking models show how larval lice can form geographic clusters at some distance from the farm of origin, and their location depends on wind and other conditions. This finding is consistent with the weak but significant relationship between farmed and wild fish that has been observed. The missing component of this combined modelling effort is the role of wild salmonids as reservoirs for infection of farmed fish. Other areas of modelling of transmission between wild and farmed fish are rudimentary. Murray et al (53) have developed hydrodynamic models to analyse dispersal of pathogens, specifically infectious salmon anaemia (ISA) and infectious pancreatic necrosis (IPN), but with potential application for other pathogens. The principle aim of this work is to look at spread between farms, but spread from farmed to wild populations or vice versa can be simulated by this approach. Interaction of pathogen spread between several populations may be important for the development of self-sustaining infections (54), even if no single reservoir could support infection indefinitely. The research into sea lice indicates that modelling of farmed wild-interaction is possible, even in the absence of an overarching model that describes the complete epidemiology of the system. However, exchanges of other pathogens have not been modelled in detail, and so we look at modelling in single systems. 8.3 General Principles of epidemiological modelling in aquatic environments There are two basic approaches to modelling. Statistical models use observations to derive relationships between those observations, without specific explanation (e.g. regression), and simulation models in which a theory of a system s behaviour is encapsulated in mathematical 41

45 form and the predictions of the model tested against observations. Quantitative risk analysis is a type of mathematical modelling. Models can be steady-state analytical models or dynamic models describing changes in the system with changes in forcing, or due to transient effects of forcing. Statistical models have been used mainly to describe macro-parasite loads (55). For example generalised linear modelling used to identify risk factors associated with sea lice loads on farmed salmon (56) or loads of Anisakis in wild Baltic herring (57). The general principles of simulation modelling of the spread of disease in single populations have been described using so called SIR models, in which populations are divided into susceptible (S), infected (I) and removed (recovered or dead) (R) phases. Disease spreads to susceptibles by their interaction with infected individuals (58). The application of basic principles from SIR type modelling to aquatic ecosystems has been described by Reno (59). However, there are differences in transmission of pathogens in aquatic environments which is likely to reduce the dependence on close direct contact between infected and susceptible hosts (60). This could lead to important functional differences, such as the absence of a threshold for infection and non-linear relationships between infected population and infection pressure. We have seen how infection pressure from sea lice may be focused on areas quite remote from the source (51). This means that models derived for terrestrial ecosystems cannot be applied to aquatic ecosystems without reconsideration of these basic aspects. Issues of the different natures of terrestrial and marine pathogen exchange are less important for statistical models, because no assumptions are made about the principles behind the systems functioning. However, for the same reason, such models are less informative about the processes underlying disease spread. 8.4 Models of disease in farmed fish and experimental systems One simple application of SIR modelling has been the description of the dynamics of furunculosis (Aeromonas salmonicida) in Chinook salmon (61)). In this study a simple SIR model using density dependent transmission described patterns of infection and mortality very adequately and allowed a critical host density for infection to be calculated. In contrast, analysis of IPNV dynamics in a similar experimental set-up failed to produce similar pattern, with infection described as effectively point source (62), i.e. when the density of susceptibles lies below a certain level, transmission is ineffective, but once this 42

46 density is exceeded transmission within a single population, tank, farm school, is effectively instantaneous. Such a situation might be modelled using non-linear transmission (63). At a much larger scale (transmission between farms) Murray (2005) found that a density dependent SI model (with removal of infected populations only on harvesting) gave a description of IPNV spread in Scotland at the national and regional levels that fitted the extensive inspector s observation data set. Yokota and Watanabe (64) have developed a simple model of the spread of disease between fish ponds. The statistical modelling approach has been used to examine farmed fish data, for which extensive datasets exist. For example case-control studies have been used support odd-ratio analyses to identify risk factors behind several common disease of farmed salmon, for example infectious salmon anaemia (ISA) (65), IPN (66) and furunculosis (67). Statistical analysis, generally regression, has been used to identify risks of ISA spread associated with well-boats (68) and to describe the spread of IPNV in Scottish salmon farms (69). Sea lice have been the subject of considerable modelling; process modelling such as description of sea lice development (70, 71), to assess the likely impact different times of application of treatment (72) and economic modelling of sea lice impacts (73). However, the most extensive modelling has been of the statistical analyses used to elucidate epidemiology of Lepeophtheirius salmonis (74), Caligus elongatus (75) and Chilean Caligus sea lice (76), including risk factor analysis (56) and time series analysis (77). 8.5 Models of disease in wild fish and other aquatic populations There are some models of disease affecting wild populations of fish and marine mammals. The general problem with modelling these disease interactions is, with a few exceptions, a lack of data. One group of diseases that can be modelled in wild animals are particularly severe epidemics that cause mass mortality resulting in the accumulation of carcasses that allow the progress of the epidemic to be accurately assessed. One such disease is pilchard herpesvirus whose epidemic spread has been modelled using the observed times and locations of mass stranding of carcasses on the Australian coast (47, 78). A statistical association between the origin of the disease and tuna farming sites has been demonstrated (79). Another such disease is phocine distemper virus which caused mass mortality of North Sea common seals, this has been modelled as the spread between discrete populations (80). The effects of repeated PDV 43

47 epidemics on seal populations has been modelled (81) as has PHV on pilchard populations (82) Diseases that cause substantial mortality may not always lead to mass stranding of bodies, for example Icthyophonus in herring (83). However, pathogen prevalence by age class can be used to estimate levels of mortality and hence impact on populations and potential catches. This approach was suggested by Munro et al. (83) and has been applied to Lymphocytosis in flounder (84), Icthyophonus in North Sea herring (85) and to Icthyophonus and VHS in Pacific herring (74). The model of lymphocytosis is a SEIR type model with age classes which was verified using observed prevalence (72). The VHS and Icthyophonus modelling used observed prevalences to estimate likely levels of mortality (85, 86). Pathogen levels can also be associated with environmental factors using statistical models to show association with factors such as pollution (87) and temperature (88). This pathogen monitoring approach applies best to disease with the reverse characteristics of the diseases causing rapid mass mortality, since the pathogen must persist in the host long enough to be detected. Another area of disease for which data can be collected relatively easily is the presence and distribution of larger parasites. Such data has been used to create statistical models of sea lice prevalence on wild salmonids (89) by fitting negative binomial distributions to demonstrate the highly dispersed nature of such infection. The approach has been taken a little further by Bakke and Harris (90) who showed an apparent levelling off of high loads, indicating mortality of heavily infected hosts, and Murray (91) who used a patchy infection model to simulate observed prevalence patterns. A similar modelling approach was used much earlier to analyse dynamics of plaice lice (Lepeophtheirus pectoralis). Observed distributions of two endoparasitic nematodes have been modelled Pseudoterranova decipiens (92) and Anisakis simplex (93, 94). For Pseudoterranova (92) a dynamic model, with fixed host populations was used to determine steady-state prevalence of infection under different conditions. In the Anisakis modelling the approach was statistical modelling of observed prevalence using steady-state solutions (94). Using a reverse approach, parasite accumulation can be used to estimate asymptotic growth of fish species (95). A more theoretical approach to modelling is to look at the impact on populations of mortality due to disease at different stages in the growth of an animal to determine which stages are most sensitive to disease impacts. This approach has been used for Atlantic salmon by des 44

48 Clers (96). A structured population model has been used to simulate post-epidemic recovery of pilchard populations (97). Models of the interaction of fisheries with disease have also been developed (98) for endemic disease and (99) for an epidemic disease. These models demonstrate that intensive fishing might mitigate or eradicate diseases. It should be emphasised that this finding depends on existence of density dependence in transmission which may not apply, e.g. in schooling fish where school density does not depend on school size. Disease may have indirect impacts on other components of the ecosystem, for example massive pilchard herpesvirus epidemics were associated with starvation of predators (100) and increases in competitors (101). However, there is little modelling of such effects, the exception being a model of the impact of mortality of fish on pelican dynamics for the Salton Sea in California (102). Pathogens may compete; for example, a decline was observed in seaworm in cod following PDV induced mass mortality of seals (92), this interaction of diseases has not been modelled directly in aquatic ecosystems, although des Clers and Wooton (103) did apply their model to examining the effects of changes in seal and in cod population densities to the dynamics of Pseudoterranova decipiens. 8.6 Models of disease in shellfish and other predictive modelling Two main disease were studied and modelled for the shellfish Crassostrea virginica: Perkinsus marinus (102, 103, 104, 105, 106, 111) and Haplosporidium nelsoni (107, 108, ). Furthermore, the interaction between these two pathogens is explored (112). P. marinus is a parasite that can be transmitted directly between shellfish. Haplosporidium is also a protistan parasite, but in contrast to H. nelsoni, direct transmission is not possible, and an intermediate host is suspected. The host parasite model for Perkinsus is based on separate models for the dynamics of post-settlement oyster populations (size structured) and the growth of P. marinus (102, 103). The global model is physiologically based, and the relationship between the two separate dynamic populations are based on energy requirement. Parameters included in the first models are: oyster density, temperature, salinity, food supply, turbidity and current flow (102, 103). In other words, this population can be considered as wild, except that the oyster size frequency distribution is also linked to human harvesting activities directed at commercial sized shellfish. For this approach, the long term study of numerous parameters is required, and this was achieved for C. virginica in the USA. In particular, infection intensity is needed (in number of parasites per gram of shellfish). This is 45

49 not a conventional expression in shellfish pathology, because this quantification is not easy for other parasites. The effect of transplants are taken into account in the latter model (104, 105). Competition for food by mussels and oysters, population dynamics of plankton, predation (linked to size of oysters) temperature and salinity are taken into account (104). This model is clearly linked with the transplanting of farmed stocks, and their consequences. Different strategies where tested with this model. All these models (102,103,104,105) where solved numerically (using differential equations) for each simulation. If temporal aspects are well identified, spatial approach is understood in particular situations (e.g. in discrete points of a bay, not in the continuum of the bay: there s no spatio temporal mapping of the risk). Spatial analysis of Perkinsus is addressed in a separate work (106) and is not linked to the general model of transmission. As we improve our knowledge of wild/farmed dynamics of Crassostrea gigas (113, 114), taking into account transplants, and because the susceptibility of C. gigas for Perkinsus compared to C. virginica is known (115, 116, 119), and environmental modelling is done in some large areas of production, and include water flow (117, 118), an extrapolation of the potential effect of an introduction on Perkinsus marinus, on some areas in France where neighbouring populations of wild and farmed C. gigas occur, could be done (with some heavy modelling effort). Because of the unknown intermediate host, the MSX model is not complete, but is established with the same principles of Perkinsus. Two models, one for the parasite, one for shellfish rely on basic physiological processes of both host and parasite modified by the environment to reproduce the observed annual prevalence (107,108,109,110). The next step should be to consider these two pathogens on the same host (C. Virginica) (112) and in future to apply this to C.gigas in France, where some mild MSX infections exist (H. nelsoni). This review has focused on oysters, but predictive modelling of population dynamics has also been undertaken for cockles, mussels, and clams, by ecologists, in wild systems. Because the main disease agents of farmed shellfish are not considered in these models, and because diseases are not modelled in all these studies, they were not included here, but should be taken into consideration for future work (120). For spatial modelling, spatial heterogeneity of the sea bed for suitable habitats for wild shellfish and the scale of study should be carefully determined (121,122). At a larger scale, marine ecosystems are not at present predictable enough to include pathogens as factors for modelling (123). In conclusion, the only published models for the transmission of pathogenic agents are complex and data intensive. Data availability for shellfish and their infectious agents is limited. In this case, perhaps a 46

50 simplified approach to modelling could be of some help. As physiologists and Ecologists begin to develop modelling techniques, perhaps epidemiologists should adopt simplified components. Evaluation of models developed for mammals or fish could give some examples of approaches that could be applied to shellfish. However, a simplified approach would limit extrapolation to new applications, and the evaluation of the effects of some parameters. 9 Discussion 9.1 Applying the Covello Merkhofer disease interaction The Covello Merkhofer model has been most widely used in aquatic animal health risk assessments. Disease interaction can be conceptualised as pathways between the two populations. In many aquaculture systems an important pathway may be pathogen exchange in the water column or on currents. Disease interaction may occur through escapees from farm sites or stocking of hatchery-reared fish into rivers. Similarly, wild-caught fish may be used for broodstock. These pathways can be modelled within the release assessment element of the risk assessment methodology recommended by the OIE. The introduction of a disease only becomes significant if it establishes and spreads, known as the exposure assessment with the OIE risk analysis framework. Again, pathways leading to establishment can aid our understanding of the process. Mapping out potential disease interaction through scenario trees showing pathways of pathogen spread and establishment is a fundamental step in using risk assessment. This approach clarifies and clearly communicates the key issues underlying disease interaction. The OIE risk assessment approach, based on the Covello-Merkhofer model, whilst primarily developed for IRA, has been successfully used in other contexts, including disease interaction in the case of G. salaris. It is a useful starting point from which to construct risk assessment models of disease interaction. 9.2 Quantitative or qualitative In some instances, qualitative risk assessment will not provide the information needed for decision-making. However, it is likely that for most disease interaction questions, all the data need for a quantitative assessment will not be available. Nevertheless, this should not be always considered a barrier to quantitative assessments. Quantitative analysis allows sensitivity assessment to be undertaken to determine the input variables to which the model is most sensitive, and these results can be used to prioritise future research. 47

51 9.3 Consequence assessment A minority of aquatic animal risk assessments have included a consequence assessment. Consequence assessment requires a different set of skills compared with the probabilistic release and exposure assessment. Economics as well as epidemiology is needed to properly assess the consequences of disease introduction. Methods to evaluate the impact on wild populations in economic terms are not well developed. However, it should not be forgotten that risk has two components and the value of the risk assessment is greatly diminished when consequences are not adequately evaluated. Disease interaction and pathogen exchange between wild and farmed populations has an important role to play in consequence assessment. An introduced disease may become established in a wild population that can then act as a reservoir of infection farmed stocks. Eradication of disease from wild stocks is difficult and often impossible. The direct impact on wild populations may be more important than morbidity and mortality in farmed stocks (e.g. the impact of G. salaris). IRA should routinely include a discussion disease interaction between farmed and wild stocks. 9.4 The application of mathematical models Mathematical modelling has been successfully used in studies of human and terrestrial livestock diseases to better understand and quantify aspects of disease transmission. The results have been used successfully in the design of vaccination programmes. The application of modelling to sealice epidemiology has demonstrated how quantitative approaches can improve our understanding of disease interaction. The lack of data has been a major constraint to its wider use. However, mathematical models, through sensitivity analysis, can be used to identify parameters to which the outcome is most sensitive; such results can guide future research. 9.5 Combining risk and mathematical models Both risk assessment and mathematical modelling require that a complex biological or population based system, in this case disease interaction, is reduced to as simple idealised sequence of events. The advantages of combining both approaches needs to be explored. For example, mathematical modelling has a clear role in assessing both whether disease establishment and spread takes place (i.e. exposure assessment) and the potential severity of disease interaction (i.e. consequence assessment). Methodological advances in assessing the consequences of disease spread and interaction between wild and farmed fish will be achieved by using mathematical modelling approaches. 48

52 10 Conclusion and recommendations Disease interaction is frequently a controversial subject. The application of risk assessment and mathematical modelling can assist in resolving disputed scientific questions by imposing a rigorous, logical, disciplined, systematic approach. Framing the risk question requires sets out the limits of the study at the outset. Scenario tree modelling produces a logical consistent construction of routes of disease interaction between farmed and wild populations. One of the strengths of risk analysis is the documentation of the evidence used in determining levels of risk and the clear identification of gaps in the available data. The results of a risk analysis are, therefore, transparent and defensible; and the application of risk analysis lends similar benefits to decision making. Both mathematical and risk modelling approaches clearly identify the data needed to quantitatively explore disease interaction and by doing do assist in the prioritisation of research. Risk assessment models have been successfully used in the study of disease spread between countries (IRA) and more locally (i.e. between rivers). Disease interaction and pathogen exchange between wild and farmed populations can be viewed as a form of disease spread, and there are a few studies where risk assessment has been successfully applied to disease interaction (5, 104). In these studies the application of the Covello-Merkhofer model (the basis of OIE guidelines) has proved flexibile and robust and is the most suitable approach to the use of risk assessment in the study of disease interaction between wild and farmed populations. IRA is a well-established approach to the study of international spread of aquatic animal and other diseases. The potential of risk analysis for studying the exchange of pathogens between wild and farmed aquatic animals is still rudimentary and clearly under exploited. On the basis of this report, recommendations can be made to promote the use of risk assessment and modelling in the study of disease interaction: 1. import risk assessments should explicitly include disease interaction, and especially the impact of introduced diseases on wild populations, within the consequence assessment 2. the Corvello Merkhofer approach should be modified and employed to study the risks for pathogen exchange between farmed and wild aquatic animal populations 49

53 3. mathematical disease modelling should be used to estimate consequences of disease introduction 4. the economic consequences of disease introduction, including disease interaction, should be assessed 5. teams undertaking import risk assessment should include economists and disease modellers as well as pathologists and epidemiologists 6. quantitative risk assessment should be undertaken when qualitative results are not sufficient for decision making and the required data is available 7. mathematical models of disease interaction need to be combined with ecological models to assess the knock-on environmental effects of disease interaction 11 Acknowledgements The authors would like to thank Paul Midtlyng and Laurence Meossec for their valuable comments on earlier drafts of this document. 50

54 References 1. Vose D. Qualitative versus quantitative risk analysis and modelling. In: Rodgers CJ, editor. OIE International Conference on Risk analysis in Aquatic Animal Health; February 2000; Paris: OIE;2001: MacDiarmid SC, Pharo HJ. Risk analysis: assessment, management and communication. Rev Sci tech Off Int Epiz 2003;22(2): Murray N, MacDiarmid SC, Wooldridge M, Gummow B, Morley RS, Weber SE, et al. Handbook on Import Risk Analysis for Animals and Animal Products - Introduction and qualitative risk analysis. 1 ed; Vose D. Risk analysis - a quantitative guide. Chichester: John Wiley & Sons; Paisley LG, Karlsen E, Jarp J, Mo TA. A Monte Carlo simulation model for assessing the risk of introduction of Gyrodactylus salaris to the Tana river, Norway. Diseases of Aquatic Organisms 1999;37(2): Vose D. Risk analysis: a quantitative guide. 2nd ed. New York: John Wiley and sons; Bruneau NN, Thorburn MA, Stevenson RMW. Use of the Delphi panel method to assess expert perception of the accuracy of screening test systems for infections pancreatic necrosis virus and infectious hematopoietic necrosis virus. Journal of Aquatic Animal Health 1999;11(2): Rodgers CJ. Risk assessment in aquaculture report on a workshope held at the VIII EAFP international conference. In: EAFP VIII International Conference; 1997; Herriot-Watt University: EAFP; 1997; Bruneau N. A quantitative risk assessment for the introduction of Myxobolus cerebralis to Alberta, Canada, through the importation of live salmonids. Risk analysis in aquatic animal health. In: Rodgers CJ, editor. Risk analysis in aquatic animal health; February 2000; Paris: Office International des Epizooties; 2001; Anon. National Research Council, Committee of Institutional Means of Risks to Public Health; Risk assessment in the federal government: Managing the process. Washington: National Academy Press; Anon. Risk assessment in the federal government: Managing the process. Washington: National Academy Press; O.I.E. Aquatic Animal Health Code. 7th ed. Paris; Covello VT, Merkhofer MW. Risk assessment methods: Approaches for assessing health and environmental risks. New York: Plenum Publishing; O.I.E. International Aquatic Animal Health Code - fish, molluscs and crustaceans. 6th ed. Paris;

55 15. Kahn SA, Beers PT, Findlay VL, Peebles IR, Durham PJ, Wilson DW, et al. Import Risk Analysis on Non-viable Salmonids and Non-salmonids Marine Finfish. Canberra: Australian Quarantine and Inspection Service; O.I.E. Manual of Diagnostic Tests for Aquatic Animals. 4th ed. Paris; Pharo HJ. The limits to quantification in import risk analysis. In: Anon., editor. 10th International Symposium for Veterinary Epidemiology and Economics; 2003; Vina del Mar, Chile; Trachtman JP. Decisions of the Appellate Body of the World Trade Organisation current survey. Australia - Measures affecting importation of salmon. European Journal of International Law 1999;10: Peeler EJ. The role of risk analysis and epidemiology in the development of biosecurity in aquaculture. In: Walker PJ, Lester RG, Bondad-Reantaso MG, editors. Diseases in Asian Aquaculture V; 2005; Queensland, Australia: Asian Fisheries Society; 2005; Anon. Final report of the the Joint Government / Industry Working Group on Infectious Salmon Anaemia (ISA) in Scotland: Scottish Executive; Giovannini A, Conte A, Calistri P, Di Franceco CE, Caporale V. Risk analysis on the introduction into free territories of vaccinated animals from restricted zones. Veterinara Italiana 2004;40(4): Calistri P, Giovannini A, Conte A, Caporale V. Use of Monte Carlo simulation model for the re-planning of bluetongue surveillance in Italy. Veterinara Italiana 2004;40(4): Hine PM. Problems of applying risk analysis to aquatic organisms. In: OIE, editor. Risk analysis in aquatic animal health. Paris; 2001; Thebault A. Certifying the french population of Crassostrea gigas free from exotic diseases: a risk analysis approach. In: OIE, editor. Risk analysis in aquatic animal health. Paris: OIE; 2001; Grizel H. Les maladies des mollusques bivalves: risque et prévention. Rev. Sci. tech. Off. Int. Epiz. 1988;16(1): Thrusfield M. Veterinary epidemiology: Butterworths; Johnsen BO, Jensen AJ. Infestations of Atlantic salmon, Salmo salar, by Gyrodactylus salaris in Norwegian rivers. Journal of Fish Biology 1986;29(2): Johnsen BO. The effect of an attack by the parasite Gyrodactylus salaris on the population of salmon parr in the River Lakselva, Misvaer in northern Norway. Astarte 1978;11(1): Johnsen BO, Jensen AJ. The Gyrodactylus story in Norway. Aquaculture 1991;98:

56 30. Jansen P, Hogaasen H, Brun E. An evaluation of potential routes for spreading Gyrodactylus salaris. Oslo: Norwegian Scientific Committee for Food Safety; Report No.: 05/ Høgåsen HR, Brun E. Risk of inter-river transmission of Gyrodactylus salaris by migrating Atlantic salmon smolts, estimated by Monte Carlo simulation. Diseases of Aquatic Organisms 2003;57(3): Peeler EJ, Thrush MA. Qualitative analysis of the risk of introducing Gyrodactylus salaris into the United Kingdom. Diseases of Aquatic Organisms 2004; : Peeler EJ, Thrush MA, Paisley L, Rodgers CJ. An assessment of the risk of spreading the fish parasite Gyrodactylus salaris to uninfected territories in the European Union with the movement of live Atlantic salmon (Salmo salar) from coastal waters. Aquaculture 2006;258: Peeler EJ, Gardiner R, Thrush MA. Qualitative risk assessment of routes of transmission of the exotic fish parasite Gyrodactylus salaris between river catchments in England and Wales. Preventive Veterinary Medicine 2004;64: Bakke TA, MacKenzie K. Comparative susceptibility of native Scottish and Norwegian stocks of Atlantic salmon, Salmo salar L., to Gyrodactylus salaris Malmberg: Laboratory experiments. Copenhagen (Denmark): Ices; Thrush MA, Peeler EJ. Stochastic simulation of live fish movement in England and Wales to predict potential spread of exotic pathogens. Diseases of Aquatic Organisms. In press. 37. McDiarmid SC. The risk of introducing exotic diseases of fish into New Zealand through the importation of ocean-caught Pacific salmon from Canada. Wellington: Ministry of Agriculture and Fisheries, New Zealand; O.I.E. Risk analysis in aquatic animal health. In: Rodgers CJ, editor. OIE International Conference on Risk Analysis in Aquatic Animal Health; February 2000; Paris: Office International des Epizooties; LaPatra SE, Batts WN, Overturf K, Jones GR, Shewmaker WD, Winton JR. Negligible risk associated with the movement of processed rainbow trout, Oncorhynchus mykiss (Walbaum), from an infectious haematopoietic necrosis virus (IHNV) endemic area. Journal of Fish Diseases 2001;24(7): Brown C. Emerging Diseases of Animals - an Overview. In: Brown C, Bolin C, editors. Emerging Diseases of Animals: American Society for Microbiology; p Krause RM. Emerging Infections. New York: Academic Press; Daszak P, Cunningham AA, Hyatt AD. Emerging infectious diseases of wildlife - threats to biodiversity and human health. Science 2001;287:

57 43. Harvell CD, Kim K, Burkholder JM, Colwell RR, Epstein PR, Grimes DJ, et al. Emerging marine diseases - climate links and anthropogenic factors. Science 1999: Murray AG, Peeler EJ. A Framework for Understanding the Potential for Emerging Diseases in Aquaculture. Preventive Veterinary Medicine 2005;67: MacKenzie K, Longshaw M, Begg GS, McVicar AH. Sea lice (Copepoda: Caligidae) on wild sea trout (Salmo trutta L.) in Scotland. ICES Journal of Marine Science 1998;55: Tully O, Gargan P, Poole WR, Whelan KF. Spatial and temporal variation in the infestation of sea trout (Salmo trutta L.) by the caligid copepod Lepeophtheirus salmonis (Kroeyer) in relation to sources of infection in Ireland. Parasitology 1999;119(1): Murray AG, O'Callaghan M, Jones B. Simple models of massive epidemics of herpesvirus in Australian (and New Zealand) pilchards. Environment International 2001;27: Heuch PA, Mo TA. A model of salmon louse production in Norway: effects of increasing salmon production and public management measures. Diseases of Aquatic Organisms 2000;45(2): Butler JRA. Wild salmonids and sea louse infestations on the west coast of Scotland: sources of infection and implications for the management of marine salmon farms. Pest Management Science 2002;58: Asplin L, Boxaspen K, Sandvik AD. Modelling distribution of salmon lice in a Norwegian fjord.: ICES; Report No.: CM 2004/P: Murray AG, Gillibrand PA. Modelling salmon lice dispersal in Loch Torridon, Scotland. Marine Pollution Bulletin in press. 52. Krkošek M, Lewis MA, Volpe JP. Transmission dynamics of parasitic sea lice from farm to wild salmon. Proceedings of the Royal Society Series B 2005;272: Murray AG, Amundrud TL, Gillibrand PA. Models of hydrodynamic pathogen dispersal affecting Scottish salmon production: modelling shows how Scotland eradicated ISA, but not IPN. Bulletin of Aquaculture Association of Canada submitted. 54. Haydon DT, Cleaveland S, Taylor LH, Laurenson MK. Identifying reservoirs of infection: a conceptual and practical challenge. Emerging Infectious Diseases 2002;8(12): Shaw DJ, Dobson AP. Patterns of macroparasite abundance and aggregation in wildlife populations: a quantitative review. Parasitology 1995;111:S

58 56. Revie CW, Gettinby G, Treasurer JW, Wallace C. Identifying epidemiological factors affecting sea lice Lepeophtheirus salmonis abundance on Scottish salmon farms using general linear models. Diseases of Aquatic Organisms 2003;57: Podolska M, Horbowy J. Infection of Baltic herring (Clupea harengus membras) with Anisakis simplex larvae, : a statistical analysis using generalized linear models. ICES Journal Of Marine Science 2003;60(1): Anderson RM, May RM. Population biology of infectious diseases: Part 1. Nature 1979;280: Reno PW. Factors involved in the dissemination of disease in fish populations. Journal of Aquatic Animal Health 1998;10(2): McCallum HI, Kuris A, Harvell CD, Lafferty KD, Smith GW, Porter J. Does terrestrial epidemiology apply to marine systems? Trends in Ecology & Evolution 2004;19: Ogut H, Reno PW, Sampson D. A deterministic model for the dynamics of furunculosis in chinook salmon Oncorhynchus tshawytscha. Diseases of Aquatic Organisms 2004;62: Smith G, Bebak J, McAllister PE. Experimental infectious pancreatic necrosis infections: propagative or point source epidemic? Preventive Veteinary Medicine 2000;47: Liu WM, Levin SA, Iwasa Y. Influence of nonlinear incidence rates upon the behaviour of SIRS epidemiological models. Journal of Mathematical Biology 1986;23: Yokota M, Watanabe S. (2004) Risk analysis of epidemic disease on one-dimensional cultured fish pond model. Nippon Suisan Gakkaishi/Bull. Jap. Soc. Sci. Fish. 2004;70: Jarp J, Karlsen E. Infectious salmon anaemia (ISA) risk factors in sea-cultured Atlantic salmon Salmo salar. Diseases of Aquatic Organisms 1997;28(2): Jarp J, Gjevre AG, Olsen AB, Bruheim T. Risk factors for furunculosis, infectious pancreatic necrosis and mortality in post-smolt Atlantic salmon, Salmo salar L. Journal of Fish Diseases 1994;18: Jarp J, Tangen K, Willumsen FV, Dyupvik HO, Tveit AM. Risk factors for infection with Aeromonas salmonicida subsp. salmonicida in Norwegian freshwater hatcheries. Diseases of Aquatic Organisms 1993;17(2): Murray AG, R.J. S, Stagg RM. Shipping and the Spread of Infectious Salmon Anemia in Scottish Aquaculture. Emerging Infectious Diseases 2002;8: Murray AG, Busby CD, Bruno DW. Infectious pancreatic necrosis virus in Scottish Atlantic salmon farms. Emerging Infectious Diseases 2003;9:

59 70. Revie C.W. R, C., Gettinby, G., Kelly, L. and Treasurer J.W. A mathematical model of the growth of sea lice L salmonis populations on farmed salmon in Scotland and its use in the assessment of treatment strategies. J. Fish Dis. in press. 71. Stien A, Bjørn PA, Heuch PA, Elston DA. Population dynamics of salmon lice Lepeophtheirus salmonis on Atlantic salmon and sea trout. Mar. Ecol. Prog. Ser. 2005;290: Tucker CS, Norman R, Shinn AP, Bron JE, Sommerville C, Wooten R. A single cohort time delay model of the life-cycle of the salmon louse Lepeophtheirus salmonis on Atlantic salmon Salmo salar. Fish Pathology 2002;37: Pike AW, Wadsworth SL. Sea lice on salmonids: their biology and control. Advances in Parasitology 2000;44(9): Revie CW, Gettinby G, Treasurer JW, Rae GH, Clark N. Temporal, environmental and management factors influencing the epidemiological patterns of sea lice (Lepeophtheirus salmonis) infestations on farmed Atlantic salmon (Salmo salar) in Scotland. Pest Management Science 2002;58(6): Revie CW, Gettinby G, Treasurer JW, Rae GH. The epidemiology of the sea lice, Caligus elongatus Nordmann, in marine aquaculture of Atlantic salmon, Salmo salar L. in Scotland. Journal of Fish Diseases 2002;25(7): Zagmutt-Vergara FJ, Carpenter TE, Farver TB, Hedrick RP. Spatial and temporal variations in sea lice (Copeda: Caligidae) infestations of three salmonid species farmed in net pens in southern Chile. Diseases of Aquatic Organisms 2005;64(2): McKenzie E, Gettinby G, McCart K, Revie CW. Time-series models of sea lice Caligus elongatus (Nordmann) abundance on Atlantic salmon Salmo salar L. in Loch Sunart, Scotland. Aquaculture Research 2004;35(8): Murray AG, O'Callaghan M, Jones B. A model of spatially-evolving herpesvirus epidemics causing mass mortality in Australian pilchards (Sardinops sagax). Diseases of Aquatic Organisms 2003;54(1): Gaughan DJ, Fletcher WJ, White KV. Growth rate of larval Sardinops sagax from ecosystems with different levels of productivity. Marine Biology 2001;139(5): De Koeijer A, Diekmann O, Reijnders P. Modelling the spread of phocine distemper virus among harbour seals. Bulletin of Mathematical Biology 1998;60(3): Harding KC, Harkonen T, Caswell H. The 2002 European seal plague: epidemiology and population consequences. Ecology Letters 2002;5: Gaughan DJ, Mitchell RW, Blight SJ. Impact of mortality, possibly due to herpesvirus, on pilchard Sardinops sagax stocks along the south coast of Western Australia in Marine & Freshwater Research 2002;6:

60 83. Munro ALS, McVicar AH, Jones R. The epidemiology of infectious diseases in commercially important wild marine fish. Rapp. P.-v. Réun. Cons int. Explor. Mer 1983;182: Lorenzen K, Clers S, Anders K. Population dynamics of lymphocystis disease in estuarine flounder, Platichthys flesus (L.). Journal of Fish Biology 1991;39(4): Patterson KR. Modelling the impact of disease-induced mortality in an exploited population of the fungal parasite Icthyophonus hoferi in the North Sea herring (Clupea harengus). Can. J. Fish. Aquat. Sci. 1996;53: Marty GD, Quinn TJ, Carpenter G. Role of disease in abundance of a Pacific herring (Clupea pallasi) population. Canadian Journal of Fisheries and Aquatic Sciences 2003;60(10): Wosniok W, Lang T, Dethlefsen V, Feist SW, McVicar AH, Mellergaard S, et al. Analysis of ICES long-term data on diseases of North Sea dab (Limanda limanda) in relation to contaminants and other environmental factors.: ICES; Report No.: CM 200/S Speare DJ, Beaman HJ, Daley J. Effect of water temperature manipulation on a thermal unit predictive model for Loma salmonae. Journal of Fish Diseases. 1999;22(4): Boxshall G, Defaye D. Pathogens of wild and farmed fish: sea lice. New York: Ellis Horwood; Bakke TA, Harris PD. Diseases and parasites in wild Atlantic salmon (Salmo salar) populations. Canadian Journal of Fisheries & Aquatic Sciences 1998;55(Suppl 1): Murray AG. Using observed load distributions with a simple model to analyse the epidemiology of sea lice (Lepeophtheirus salmonis) on sea trout (Salmo trutta). Pest Management Science 2002;58(6): Des Clers S, Andersen K. Sealworm (Pseudoterranova decipiens) transmission to fish trawled from Hvaler, Oslofjord, Norway. Journal of Fish Biology 1995;46(1): Horbowy J, Podolska M. Modelling infection of Baltic herring (Clupea harengus membras) by larval Anisakis simplex. ICES Journal of Marine Science 2001;58(1): Podolska M, Horbowy J. The analysis of infection of Baltic herring (Clupea harengus membras) with Anisakis simplex larvae using generalized linear models. Report; Summary. Copenhagen: International Counc. for the Exploration of the Sea; Report No.: ICES CM 2001/U: Gibson D, Jones J. Fed up with parasites - a method for estimating asymptotic growth in fish populations. Marine Biology 1993;117(3):

61 96. des Clers S. Modelling the impact of disease-induced mortality on the population size of wild salmonids. Fisheries Research 1993;17: Murray AG, Gaughan DJ. Using an age-structured model to simulate the recovery of the Australian pilchard (Sardinops sagax) population following epidemic mass mortality. Fisheries Research 2003;60: Dobson AP, May RM. Effects of parasites on fish populations - theoretical aspects. International Journal of Parasitology 1987;17: Murray AG. Managing harvesting to minimise the impact of epidemics on wild fish stocks. Natural Resources Modelling 2004;17: Dann P, Norman FI, J.M. C, Neira FJ, Chiaradia A. Mortality and breeding failure of little penguins, Eudyptula minor, in Victoria, , following a widespread mortality of pilchard Sardinops sagax. Marine and Freshwater Research 2000;51: Ward TM, Hoedt F, McLeay L, Dimmlich WF, Jackson G, Rogers PJ, et al. Have recent mass mortalities of the sardine Sardinops sagax facilitated an expansion in the distribution and abundance of the anchovy Engraulis australis in South Australia? Marine Ecology-Progress Series 2001;220: Chattopadhyay J, Bairagi N. Pelicans at risk in Salton sea - an eco-epidemiological model. Ecological Modelling 2001;136: Des Clers SA, Wootten R. Modelling the population dynamics of the sealworm Pseudoterranova decipiens. Neth. J. Sea Res. 1990;25(1-2): Høgåsen HR, Brun E. Estimating the risk of inter-river transmission of Gyrodactylus salaris by migrating Atlantic salmon smolts using Monte Carlo simulation. Diseases of Aquatic Organisms 2003;57(3): PharoHJ, MacDairmid SC. Quantitative analysis of the risk of disease importation of salmon for human consumption. In :Risk analysis in aquatic animal health. Office International des Epizooties, Paris 2001; Paisley LG. A monte carlo simulation model for assessing the risk of the introduction of G. salaris to the Tana river, Norway: a second scenario. In :Risk analysis in aquatic animal health. Office International des Epizooties, Paris 2001; Stone MAB, MacDiarmid SC, Pharo HJ. Import Health Risk Analysis: salmonids for human consumption. Ministry of Agriculture Regulatory Authority, New Zealand. 1997;269pp Anon. Import Risk Analysis: Prawns and prawn products (Draft report). Australian Quarantine and Inspection Service (AQIS), Canberra. 2000;185pp. 58

62 109. Diggles BK. Import Risk Assessment: Juvenile yellowtail kingfish (Seriola lalandi) from Spencer Gulf Aquaculture, South Australia. NationalInstitute of Water & Atmospheric Research. Wellington, New Zealand Kahn SA, Beers PT, Findlay VL, Peebles IR, Durham PJ, Wilson DW, et al. Import risk analysis on live ornamental finfish. Australian Quarantine and Inspection Service (AQIS), Canberra. 1999;172pp Travis DA, Baya AM, Hueston WD. Risk analysis for pathogen introduction into aquaculture via live fish, fish egg, or fish product importation. Proceedings Of The 9th Symposium Of The International Society For Veterinary Epidemiology And Economics, Breckenridge, Colorado, USA: (ISVEE) 2000;9: Anderson C. The risk of introducing exotic fish disease with imported fish flesh. Surveillance 1990;19(1): Edgerton BF. Hazard analysis of exotic pathogens of potential threat to European freshwater crayfish. Bull. Fr. Pêche Piscic. 2002;367: Manfrin A, Friso S, Perin R, Qualtieri K, Bovo G. Tropical fish importation from third counties: the potential risk of introducing human and aquatic animal pathogens. In :Risk analysis in aquatic animal health. Office International des Epizooties, Paris 2001; Mortensen S. Scallop introductions and transfers, from an animal health point of view. Aquaculture International. 2000;8(2-3): Anderson C. Important diseases of salmonid fish and the risk they pose to New Zealand. Surveillance 1992;17(2): Thebaut A, Berthe F, Audigé L. Certifying the French population of Crassostrea gigas free from exotic diseases: a risk analysis approach. In :Risk analysis in aquatic animal health. Office International des Epizooties, Paris 2001;

63 12 Appendix 1. Terrestrial Import risk analyses 1. Imported seropositive animals: Assurance provided by serological tests animals/risk/seropositive-ra.pdf Accessed on July 7, Import risk analysis: avian paramyxovirus type 1 in hens' hatching eggs animals/risk/avianparamyxovirus-ra.pdf Accessed on July 7, Import risk analysis: honey bee (Apis mellifera) genetic material animals/risk/honey-beegenetic-material-ra.pdf Accessed on July 7, Import risk analysis: horses and horse semen animals/risk/horse-ra.pdf Accessed on July 7, The use in New Zealand of imported semen derived from an argali (Ovis ammon polii) sheep animals/risk/argali-semen-ra.pdf Accessed on July 7, Import risk analysis: Babesia gibsoni in dogs (Canis familiaris) and dog semen animals/risk/babesia-gibsoni-ra.pdf Accessed on July 7, Diseases of Antelope: Risks of introducing live antelope into zoological gardens. animals/risk/antelope-ra.pdf Accessed on July 7, Import risk analysis: Honey bee products animals/risk/ira-beeproducts.pdf Accessed on July 7, Import risk analysis: Honey bee hive products and used equipment animals/risk/irahoney-products-and-equip.pdf Accessed on July 7, Import risk analysis: Possum fibre from Australia pestsdiseases/animals/risk/possum/possum.pdf Accessed on July 7, Import risk analysis: macropod fibre and skins from Australia Accessed on July 7, Import risk analysis: unprocessed fibre of sheep and goats animals/risk/fibre-sheepgoats-ra.pdf Accessed on July 7, Import risk analysis: importation of weed species by live animals and unprocessed fibre of sheep and goats animals/risk/weeds-seedsra.pdf Accessed on July 7, Import risk analysis: Belovo Egg Powders Accessed on July 7, Import risk analysis: chicken meat and chicken meat products; Bernard Matthews Foods Ltd turkey meat preparations from the United 60

64 Kingdom animals/risk/chicken-meat-ra.pdf Accessed on July 7, Import risk analysis: chicken meat and chicken meat products; Bernard Matthews Foods Ltd turkey meat preparations from the United Kingdom. Revised quantitative risk assessments on chicken meat from the United States; Reassessment of heat treatment for inactivation of Newcastle Disease Virus in chicken meat Accessed on July 8, Import risk analysis: Camel (Camelus dromedarius) Meat for Human Consumption from Australia Accessed on July 8, A Report on the Risk to New Zealand of Canine Heartworm (Dirofilaria immitis) and Quarantine Measures which could be considered Appropriate to reduce this Risk Accessed on July 8, The Potential Risks to Animal Health from Imported Sheep and Goat Meat Accessed on 8 July The importation into New Zealand of meat and meat products: a review of the risks to animal health Accessed on July 8, Hazard identification and import release assessment: The introduction of red imported fire ants into New Zealand via the importation of goods and arrival of craft from Australia, the Caribbean, South America, and the USA Accessed on July 8, Risk analysis for influenza viruses in poultry meat. The Impact of New Epidemiological Information on a Risk Analysis for the Introduction of Avian Influenza Viruses in Imported Poultry Meat Accessed on July 8, Determination of the acceptable risk of introduction of FMD virus in passenger luggage following the UK outbreak in Accessed on July 8, Foot-and-mouth disease: an assessment of the risks facing New Zealand Accessed on July 8, Process for conducting import risk analyses for animals and animal products Accessed on July 8, 61

65 Risk Analysis Opening the Way for Safety in Agricultural Trade Accessed on July 8, Risk. Risk assessment for the import and keeping of exotic vertebrates in Australia nce/lms/ferals/risk_assess_book.doc Accessed on July 8, Import risk analysis report on the importation of bovine semen and embryos from Argentina and Brazil into Australia. Part 1: bovine semen ss/biosecurity/animal/2000/00-003a.pdf Accessed on July 8, Import risk analysis report on the importation of bovine semen and embryos from Argentina and Brazil into Australia. Part 2: bovine embryos ss/biosecurity/animal/2000/00-003b.pdf Accessed on July 8, Importation of Sausage Casings into Australia. Import Risk Analysis ss/biosecurity/animal/2000/00-005b.pdf Accessed on July 8, Importation of Crocodile Meat from Zimbabwe into Australia. Draft Import Risk Analysis Paper ss/biosecurity/animal/2000/00-007a.pdf Accessed on July 8, Importation of dairy products into Australia for human consumption. Import Risk Analysis ss/biosecurity/animal/1999/99-076a.pdf Accessed on July 8, Import risk analysis report on the revision of import policy related to scrapie ss/biosecurity/animal/2000/00-038a.pdf Accessed on July 8, An analysis of the disease risks, other than scrapie, associated with the importation of ovine and caprine semen and embryos from Canada, the United States of America and member states of the European Union ss/biosecurity/animal/2000/00-038b.pdf Accessed on July 8, Risk Assessment on Bovine Spongiform Encephalopathy in Cattle in Canada Accessed on July 8, Risk analysis: Evaluation of risk to the United States (US) of importing Foot and Mouth Disease (FMD) Virus in fresh or frozen beef from Argentina bf0046d1e2/e8e268f74e81de bb0004f71f3/$FILE/ArgRisk 62

66 Final.FMD.pdf Accessed on July 8, Risk Analysis for Importation of Classical Swine Fever Virus in Swine and Swine Products from the European Union Accessed on July 8, Risk Assessment: Importation of Adult Queens, Package Bees and Germplasm of Honey Bees, Apis mellifera L., From Australia. - Qualitative, Pathway-Initiated Pest Risk Assessment Accessed on July 8, Analysis of Risk Update for the Final Rule: Bovine Spongiform Encephalopathy; Minimal Risk Regions and Importation of Commodities _risk_doc.pdf Accessed on July 8, Kennedy report (Chapter 10 - risks of an imported cat or dog developing rabies). 2003(?).Defra, UK - m Accessed on July 8, Risk Assessment: Bovine Tuberculosis in Australia. 1999(?). cb852564bf0046d1e2/d515964e90eb26e985256bb /$file /austtb.pdf Accessed on July 8, USDA Pest Risk Assessment: Khapra Beetle Accessed on July 8, Highly Pathogenic Avian Influenza in Thailand and.pdf Accessed on July 8, Probability of infection with FMD of GB livestock as a result of illegally imported meat ml Accessed on July 8,

67 13 Appendix 2. Review of published risk analyses for the transmission of aquatic pathogens Author(s) Reference Category of Risk Assessment Location Scope Commodity/route (Country) (Area) Bruneau (10) Quantitative - monte carlo simulation Canada Trans-province Live fish Hogasen & Brun (31) Quantitative - monte carlo simulation Norway Trans-River Live fish (migration) Pharo & (105) Quantitative - monte carlo simulation New Zealand Trans-country Non-viable fish (for human consuption) MacDairmid Paisley et al. (5) Quantitative - monte carlo simulation Norway Trans-River catchment Live fish Paisley (106) Quantitative - monte carlo simulation Norway Trans-River catchment Live fish MacDairmid (37) Qualitative New Zealand Trans-country Fish products " Quantitative - mont carlo simulation (see note > ) Stone et al. (107) Qualitative and Quantitative - monte carlo simulation New Zealand Trans-country Fish products New Zealand Trans-country Fish products Peeler et al. (34) Semi-quantitative UK Trans-River catchment Multiple routes assessed Anon. (108) Qualitative Australia Trans-country Non-viable whole prawn and prawn products (for human consumption) Diggles (109) Qualitative New Zealand Trans-country Live Fish Kahn et al. (16) Qualitative Australia Trans-country Non-viable whole fish and fish products (for human consumption) Kahn et al. (110) Qualitative Australia Trans-country Live fish Peeler & Thrush (32) Qualitative UK Trans-country Multiple routes assessed Peeler et al. (33) Qualitative unspecified Trans-territory / [area of Live fish 'unequal' disease status] Travis et al. (111) Qualitative USA Trans-country Multiple routes assessed Anderson (112) Not a formal RA (see note > ) New Zealand Trans-country Non-viable whole fish and fish products (for human consumption) Edgerton (113) Not a formal RA (see note > ) unspecified Trans-country Live crayfish La Patra et al. (39) Not a formal RA (see note > ) USA Trans-[area of 'unequal' Non-viable whole fish and fish products disease status] (for human consumption) Manfrin et al. (114) Not a formal RA (see note > ) Italy Trans-country Live fish Mortensen (115) Not a formal RA (see note > ) NA Trans-[site] (see note > ) Live shellfish Anderson (116) Not a Risk Analysis (see note > ) New Zealand Trans-country Thebaut et al. (117) Not a Risk Analysis (see note > ) France NA 64

68 Appendix 2(2) Author(s) Reference Pathogen Susceptible species Marine/ Pathogen exchange*: (or spp group) freshwater Aq > Aq Aq > W W > Aq W > W Bruneau (10) M. cerebralis Rainbow trout Freshwater Yes Hogasen & Brun (31) G. salaris Atlantic salmon Freshwater (Yes) Pharo & (105) A. salmonicida Atlantic samlon Freshwater Yes MacDairmid Paisley et al. (5) G. salaris Atlantic samlon Freshwater Yes Yes Paisley (106) G. salaris Atlantic samlon Freshwater Yes MacDairmid (37) All ecom. & env. Impt. salmonid pathogens Salmonids Both Yes Yes " A. salmonicida Salmonids Both Yes Yes Stone et al. (107) All ecom. & env. Impt. salmonid pathogens Salmonids Both Yes Yes Yes Yes Peeler et al. (34) G. salaris Salmonids Freshwater Yes Yes Yes Yes Anon. (108) All ecom. & env. Impt. penaeid pathogens Diggles (109) 11 ecom. & env. Impt pathogens Kahn et al. (16) All ecom. & env. Impt. fish pathogens Kahn et al. (110) All ecom. & env. Impt. fish pathogens Penaeid prawns Both Yes Yes Yes Yes yellowtail kingfish marine Yes Yes Finfish (multi spp.) Both Yes Yes Yes Yes Ornamental finfish Both Yes Yes Yes Yes Peeler & Thrush (32) G. salaris Salmonids Freshwater Yes Yes Yes Peeler et al. (33) G. salaris Atlantic salmon (Freshwater) Yes Travis et al. (111) unspecified Striped seabass Marine (Yes) (Yes) Anderson (112) A. salmonicida Salmonids Both Yes Yes Edgerton (113) All ecom. & env. Impt. crayfish pathogens Crayfish spp. Freshwater Yes La Patra et al. (39) IHNV Rainbow trout Freshwater (Yes) (Yes) Manfrin et al. (114) Vibrio & Salmonella spp. Humans & Ornamental fish Both Mortensen (115) All ecom. & env. Impt. Scallops Marine (Yes) (Yes) shellfish pathogens Anderson (116) All ecom. & env. Impt. Salmonids salmonid pathogens Thebaut et al. (117) Perkisus & Microcytosis Oysters Marine Yes (ornamental) * denotes pathogen exchange within and between wild (W) and aquaculture (Aq) environments brackets denote inferred results 65

69 Appendix 2(3) Author(s) Reference RA format RA Components completed: (OIE/FAO) Hazard ID Release Exposure Consequence Risk management Bruneau (10) (Conforms to) OIE Yes Yes Yes (see note >) Hogasen & Brun (31) No format defined Yes Yes Yes (see note >) Pharo & (105) OIE Yes Yes Yes MacDairmid Paisley et al. (5) No format defined Yes Yes Yes Paisley (106) No format defined Yes Yes Yes Yes (see note >) MacDairmid (37) No format defined Yes Yes " No format defined Yes Yes not really? Stone et al. (107) No format defined Yes Yes Yes Yes Peeler et al. (34) OIE Yes Yes Yes Anon. (108) FAO Yes Yes Yes Yes Yes Diggles (109) OIE Yes Yes Yes Yes Yes Kahn et al. (16) FAO Yes Yes Yes Yes Yes Kahn et al. (110) FAO Yes Yes Yes Yes Yes Peeler & Thrush (32) OIE Yes Yes Yes Yes (see note > ) Peeler et al. (33) OIE Yes Yes Yes Travis et al. (111) OIE Yes Yes Yes Yes Yes Anderson (112) None (Yes) (Yes) (Yes) Edgerton (113) No format defined Yes La Patra et al. (39) Manfrin et al. (114) None Yes Mortensen (115) None (Yes) Anderson (116) Thebaut et al. (117) b rackets denote inferred results 66

70 Appendix 2(4) Author(s) Reference Notes: Bruneau (10) Risk reduction procedures are discussed in terms of the factors included in the sensitivity analysis Hogasen & Brun (31) Risk +vely correlated to the number of fish migrating form the index river (suggested as a tool for risk management) Pharo & (105) Release and Exposure assessments combined to give a risk estimation MacDairmid Paisley et al. (5) (See second scenario, below, which assumes that the smolt plant is infected) Paisley (106) Two simulations were run: saline treatment in the smolt plant had significant effect on total risk per year - Sensitivity analysis performed on inputs MacDairmid (37) Qualiltative RA demonstrated that none of 23 diseases of salmonids present in N. America was likely to be introduced, but that furunculosis had the greatest risk of introcuction. Quantitative RAs were then conducted on 4 commodities for this pathogen (Assess " release only) Stone et al. (107) Similar approach as above. Quantitative assessments used for A. salmonicida, BKD, IHN, ISA, Ceratomyxa shasta Peeler et al. (34) Anon. (108) (as above) Diggles (109) Undertaken by private consultant for Island Aquaculture Ltd., accepted as technically sound by MAF, NZ, following their internal and external review process Kahn et al. (16) Kahn et al. (110) Both RAs evaluate risks associated with individual diseases & disease agents; identify measures appropriate to the risks presented by importation; Risk management measues are proposed. These are generic IRAs (addressing all diseases and pests) to facilitate assessment of individual access requests according to the health status of the source country Peeler & Thrush (32) Impact in wild fish populations and aquaculture (movement restrictions) discussed Peeler et al. (33) Travis et al. (111) This is a summary of unspecified work - Risk management options are based on the final estimates of risk Anderson (112) Not a formal risk assessment. Contains elements of release (lots of biophysical data) and exposure assesssment, but not systematically addressed Edgerton (113) This is a review of potential hazards - the first stage of an Import Risk Analysis (A 'companion' IRA document is described) La Patra et al. (39) Not a systematic RA, but really data to support a Quantitative RA Manfrin et al. (114) Presents results of testing imported ornamental fish for bacteria pathogenic to humans (primarily) and aquatic organisms Mortensen (115) Review article: identifies potential risk factors for the transfer of pathogens with the movement of live shellfish; clarifies the level of scientific knowledge and experience to handle shellfish diseases and discusses practial problems in scallop health control Anderson (116) Not a risk assessment - just a list of potential hazards (important diseases of salmonids) Thebaut et al. (117) to sampling strategy assessment & illustrates use of stochastic modelling for planning Aq disease surveys 67

71 IFREMER Institut Français de Recherche pour l Exploitation de la Mer Laboratoire Génétique et Pathologie, La Tremblade, France VESO Veterinærmedisinsk Oppdragssenter AS Postboks 8109 Dep Oslo, Norway FRS Marine Laboratory PO Box Victoria Road Aberdeen AB11 9DB UK CEFAS The Centre for Environment, Fisheries & Aquaculture Science Barrack Road The Nothe, Weymouth DT4 8UB, UK UZ Universidad de Zaragoza, Lab. de Ictiopatología, Fac. de Veterinaria Miguel Servet Zaragoza Spain ISBN