Modelling the koi herpesvirus (KHV) epidemic highlights the importance of active surveillance within a national control policy

Authors


Correspondence author. E-mail: nick.taylor@cefas.co.uk

Summary

1. Koi herpesvirus can cause serious disease in carp Cyprinus carpio populations globally. Populations of carp exposed to the virus are already widespread across England and Wales, and there is a need to determine whether and how to control its spread. This study evaluates potential management options and provides recommendations applicable to many infected countries.

2. The influences of the main drivers of the epidemic were investigated using simple compartment based models, and the effectiveness of several potential control options were evaluated. Models were parameterized using recorded fish movement, field and experimental data.

3. Experimental studies suggested the risk of transmitting the virus between waters on angling equipment was low. Data from previous studies suggested live fish movements between fisheries, and the introduction of imported ornamental fish to be the most likely routes by which a fishery could be infected.

4. The models suggest that fish movements between fisheries alone could not have led to the number of exposed sites known to exist in 2007. An additional external infection pressure such as the introduction of imported ornamental fish would have been required, and is likely to have been the main driver of the epidemic in its early stages.

5. Predictions of future scenarios suggest that fish movements between fisheries have taken over as the main driver of the epidemic, and consequently restricting imports to reduce the external infection pressure is unlikely to have much impact on its own.

6. Due to the small proportion of infected waters currently detected, increasing the duration of movement restrictions placed on infected sites from four years to permanent was predicted to have little effect on the epidemic.

7.Synthesis and applications. Given the current stage of the koi herpesvirus epidemic, reducing the spread of the virus between fisheries is likely to be challenging, but may be possible by conducting an active surveillance programme and placing permanent movement restrictions on exposed sites. However, this will only be effective if the external infection pressure can also be reduced, possibly through restrictions on the import of fish from koi herpesvirus infected countries.

Introduction

Koi herpesvirus (KHV) can cause serious disease in carp Cyprinus carpio L. populations (Perelberg et al. 2003). In the UK carp are predominantly found in England and Wales, where, to date clinical cases of koi herpesvirus disease (KHVD) have only occurred in fisheries (angling waters) and carp intended for the ornamental trade. Taylor et al. (2010a) investigated the distribution of KHV in English and Welsh fish farms and fisheries, with a view to determining which categories of disease status were available to the UK with regard to this disease under the European aquatic animal health directive (2006/88/EC). This work showed that KHV exposed fish are widespread and prevalent in fisheries, but due to the relatively disease-free status of English and Welsh fish farms, there may be prospects for the UK to enter the category of control and eradication (category IV) under the directive. This current study investigates the KHV epidemic in English and Welsh fisheries with a view to developing an informed national controls policy, should the UK opt to attempt control and ultimately eliminate the virus.

KHV is predominantly thought to be introduced to fisheries by the movement of live fish from other fisheries or farms (Taylor et al. 2010b). The stocking of ornamental fish by hobbyists and anglers, or transfer on angling equipment or other vectors or fomites are potentially alternative routes of introduction. Taylor et al. (2010a) found that many clinical and antibody positive cases of KHV in fisheries were associated with the presence of ornamental fish, and demonstrated that a high proportion of ornamental carp imported to the UK are positive for KHV antibody. Consequently, the stocking of ornamental carp to a fishery by hobbyists or fishery owners is thought to be the most likely alternative route of KHV introduction to a fishery (after introduction of fish from fisheries). Although transmission via angling equipment and other vectors and fomites are possible routes of introduction, no published work is available to demonstrate the level of risk posed by these pathways. However, as the virus is rapidly inactivated in environmental water (Shimizu et al. 2006) these routes may be considered comparatively low risk.

Once a fishery is infected, disease is generally only observed between a permissive temperature range of 16 and 25 °C (Hedrick et al. 1999; Denham 2003; Perelberg et al. 2003; Sano et al. 2004; Terhune et al. 2004; Tu et al. 2004). There is evidence to suggest that in fish surviving infection, the virus may persist or enter a latent state, and thus remain present in a population after any clinical signs have subsided (St-Hilaire et al. 2005, 2009; Yuasa et al. 2007; Uchii et al. 2009). If this is the case, it is likely that fisheries will only clear an infection through the culling of stock and subsequent disinfection, as such fish may infect others if the virus becomes reactivated.

A legal instrument to control KHV has only been available in England and Wales since the introduction of Council Directive 2006/88/EC and the enactment of ‘The Diseases of Fish (England and Wales) Order 2007’; prior to this there were few means by which to control KHV. The policy adopted by the competent authority since 2007 is effectively a passive surveillance programme where any suspected cases of KHV must be reported to the official service. In the UK, this is the Centre for Environment, Fisheries and Aquaculture Science (Cefas), who will investigate further. If clinical signs of disease are observed, samples will be taken and if the presence of the virus is confirmed by polymerase chain reaction (PCR), the site will either be required to cull and disinfect, or fish movement restrictions will be placed on the site for at least four years and until the fish population is demonstrated to be clear of the infection. Controls are not currently placed on waters that test positive for KHV antibody or do not show clinical signs of disease.

Given the current status of the pathogen within England and Wales, it is important to know if there are still prospects to reduce its spread, and control KHV. If so, is the current monitoring and control policy sufficient, and if not, what is likely to be the most effective approach to control? This study aimed to investigate the role of the different types of infection pressure on the epidemic, and to determine the effectiveness of movement restrictions and different levels of detection in controlling the epidemic through a mathematical modelling approach.

Materials and methods

Simple deterministic compartment based models (Anderson & May 1979) were used to examine the behaviour of the epidemic under different management options. Four models were developed and run in ModelMaker v.4.0 (ModelKinetix, Oxford, UK) to investigate the main drivers of the epidemic. Each model assumed two forms of infection pressure: (i) transmission occurring due to contacts between fisheries (β), which is predominantly thought to be due to the movement of live fish, but also includes transfer via fomites and vectors (e.g. angling equipment), and (ii) a constant external infection pressure (c), which is thought to predominantly be attributed to the stocking of imported ornamental fish.

Estimating fishery-to-fishery (β) and constant (c) transmission

Estimates of the level of contact occurring between sites due to fish movements (β) was assessed based on data held in the live fish movements database (LFMD) and presented in Taylor et al. (2010). In the UK, the LFMD is a database jointly owned by the Department for Environment, Food and Rural Affairs, Welsh assembly government, Cefas and the Environment Agency (EA), and holds information on live fish movements within England and Wales. To determine the feasibility of transmission of the virus on anglers nets, and whether it was necessary to incorporate this route of transmission into the calculation of β, a laboratory experiment was conducted. Twenty-five 50 g carp were injected (intra-peritoneal route) with cell culture grown KHV and held in a 150 L circular tank fitted with a standpipe. Once KHVD signs were observed, ninety naïve 50 g carp were added to the tank to ensure the virus could be transmitted via a natural route (i.e. cohabitation). Once these fish demonstrated loss of appetite they were transferred into two fine mesh keep-nets and re-immersed in the tank. The carp were held in these nets and monitored until mortalities attributed to KHVD were observed. At this point, nets were removed from the tank and cut into large pieces. These were either transferred immediately (0 h) into a 30 L transmission tank holding ten 25 g carp, or placed into a plastic bag and held for 24, 48 or 72-h prior to being placed into a similar transmission tank. Fish in each transmission tank were then monitored for signs of KHVD. The trial was terminated when 33% (3) of carp in a tank showed disease signs or after 28 days if no mortality was observed. Immediately after termination or death, carp were tested for KHV by single-round PCR following the methods described in the Office International des Epizooties (2009) manual of diagnostic tests for aquatic animals.

Exposure via the constant infection pressure (c) could not be assessed and was therefore estimated using the model 1 (described below). The model was run with this parameter set to zero and assuming the virus was first introduced to England and Wales in 1996 (when KHV DNA was detected in an enclosed warm-water recirculation system) or 2000 (when the first clinical case was observed in an ornamental wholesaler). The value of c was then calculated that gave rise to 1610 (31%) sites being infected in 2007 (an estimate based on work by Taylor et al. 2010a) for each introduction date.

Model 1: determining the roles of the infection pressures

Model 1 is described by eqns 1 and 2 and was designed to investigate the epidemic after the initial introduction of the virus, and to determine the relative roles of infection pressures 1 and 2.

image(eqn 1)
image(eqn 2)

where S is the total number of susceptible sites, I is the total number of infectious sites, t represents time (years), β represents the rate of transmission occurring through contacts between fisheries via live carp movements = 0·000079 (see Results section for calculation), and c is the rate at which uninfected sites become infected via a constant infection pressure, such as the introduction of imported ornamental fish. All parameters were calculated on a per year basis. Starting conditions: S = 5191, I = 1.

In order to determine whether the observed 2007 infection level could be achieved through fish movements alone when random processes were included, this model was also programmed in r version 2.11.1 (R Development Core Team 2010) using a Gillespie type algorithm (Gillespie 1977) that incorporated demographic stochasticity. The constant infection pressure (c) was excluded from this model.

Model 2: reducing the constant infection pressure

Model 2 is described by eqns 3–5 and was used to investigate the situation from 2007 onwards when control measures for the disease were implemented (four years of movement restrictions on fisheries confirmed as having clinical disease, as detected via a passive surveillance programme), and to determine the effect on the future course of the epidemic of reducing c whilst maintaining this control policy. The model assumed removing imports of carp from countries known to be infected with KHV would eliminate the constant infection pressure that is c = 0. Also assumed was a worst-case scenario in which all sites remained infected after movement restrictions were lifted.

image(eqn 3)
image(eqn 4)
image(eqn 5)

where T represents the time delay relating to the duration of movement restrictions placed on infected sites = four years, C represents the number of controlled infected sites, that is infected sites with movement restrictions enforced, α is the rate at which infectious sites are detected and movement restrictions are applied = 0·0062 based on Taylor et al. (2010), which reported only 10 of a possible 1610 infected sites (0·0062) were detected in 2007. Starting conditions: S = 3582, I = 1600, C = 10.

Model 3: increasing surveillance and the duration of movement restrictions

Model 3 is described by eqns 6–8 and was designed to investigate the influence of a more comprehensive control programme that involved import restrictions, an active surveillance programme and permanent movement restrictions on fisheries exhibiting both clinical and non-clinical signs of KHV. In this model detection of positive sites (α) was increased to establish its effect on the epidemic. Detected positive sites were assumed to have permanent movement restrictions placed on them.

image(eqn 6)
image(eqn 7)
image(eqn 8)

Here, α is the rate of detection of both clinically infected sites and infected sites showing no signs of disease. Starting conditions: S = 3582, I = 1600, C = 10.

Model 4: the influence of recovery from infection on the effectiveness of control options

Model 4 was designed to establish whether the behaviour of the epidemic under the different management recommendations studied in the previous three models held true if the assumption regarding latency was not true, and a proportion of sites recovered from infection. This model is described by eqns 9 to 12, and assumed four years of movement restrictions were applied to sites testing KHV positive. The model assumed that detection was specific to the infected group and did not detect individuals in the recovered category. No data was available to inform the values of rate of recovery from infection (γ), or the rate at which an antibody response was lost after recovery (δ); arbitrary values that the authors deemed biologically plausible were therefore chosen. The model was first run assuming a year 2000 introduction and assuming no controls were placed on infected waters in order to estimate α and c and the starting conditions in 2007 when controls were introduced.

image(eqn 9)
image(eqn 10)
image(eqn 11)
image(eqn 12)

where R represents fisheries that have recovered from infection, γ represents the rate of recovery = 0·125, c = 0·017, α = 0·0071, δ represents the rate at which immunity is lost = 0·5, and T = 4. Starting conditions: S = 3582, I = 1396, C = 10, R = 204.

Results

Estimating the fishery-to-fishery transmission coefficient (β)

Over 50% of the carp used to contaminate the nets showed clinical signs of KHV disease and tested positive for KHV by single-round PCR. The virus was not subsequently transmitted from the contaminated netting to the naïve carp. No mortality attributed to the virus was observed in these fish and none tested positive for the virus by PCR at the end of the 28-day trial. Transmission of the virus via angling equipment was therefore deemed a low risk route of transmission and was not considered further in the calculation of β. The estimate was therefore based on the probability that if two sites were selected at random, one would supply live fish to the other within the course of a year. Taylor et al. (2010a) reported 5192 fisheries to have received carp between 2001 and 2007; this was assumed to be the total population size. Data pertaining to consented carp movements revealed that on average 464 sites each supplied carp to 2·3 sites per year. The probability of a site supplying carp in a year = 464/5192, and the probability of a site receiving carp from a site in that year = 2·3/5191. However, based on EA and Cefas Fish Health Inspectorate estimates, only 50% of fish movements are consented, the estimate of β was therefore doubled accordingly. All contacts between infected and uninfected fisheries were assumed to successfully transmit the virus. Therefore; β = 2 × [(464/5192) × (2·3/5191)] = 0·000079. Models were re-run using this value of β ± 50% to ensure this did not change any conclusions arising from the model outputs. Although changes in β influenced the magnitude of responses to changes in other parameters, and the timescales over which the epidemics occurred, differences were not sufficient to alter the conclusions made in the following sections.

Model 1: determining the roles of the infection pressures

Whether it was assumed that the virus was first introduced into England and Wales in 1996 or 2000, model 1 demonstrates that it is unlikely that 31% of sites could have become exposed to KHV by 2007 through movement of live fish between fisheries alone (Fig. 1). The results of the stochastic model further confirm this, showing that although some simulations were closer to the observed data than the deterministic model result, they still did not achieve the level of infection observed in 2007, even if the virus was assumed to have been introduced as early as 1996 (Fig. 2). To achieve observed level of infection in 2007, either β would need to be in the order of 1·78 times greater than estimated, or the virus would have had to have been introduced to the UK as early as 1988, or, an additional (assumed constant) infection pressure (c) would be required. The magnitude of the pressure required varies depending on the year the virus was introduced, but from the model it was estimated that a site would have between a 0·19% and 1·04% chance per annum of being infected by means other than a fish movement if the virus is assumed to have been introduced to England and Wales between 1996 and 2000.

Figure 1.

 Results of model 1, investigating the likely epidemic if the virus was introduced in either 1996 or 2000 and transmission was based on movements between fisheries alone or a combination of fisheries movements and constant infection pressure (c). * represents the predicted proportion of KHV antibody positive sites in England and Wales in 2007 according to Taylor et al. (2010a).

Figure 2.

 Results of 500 stochastic simulation runs of model 1 based on transmission by movements between fisheries alone. a and b represent observed infection level if the virus was assumed to have been introduced 12 or eight years prior to the survey conducted by Taylor et al. (2010a) in 2007. Grey lines show stochastic simulation results, and black the deterministic model result.

The constant infection pressure appears to have been the main driving force in the early stages of the epidemic. Had it not been present it would have taken 20 years to reach the fishery level prevalence observed in 2007, and 45 years for all sites to become infected. Based on the estimated value of β, if the constant infection pressure is assumed to be between 0·0104 (assuming a year 2000 introduction) and 0·0019 (assuming a 1996 introduction) it would take eight to 12 years (60–40% reduction c.f. c = 0) to get to the situation observed in 2007, and 30–34 years for all sites to become infected (33–24% reduction in time c.f. c = 0). Between 6·4% and 21% of infected sites present in 2007 can be attributed to the constant infection pressure based on the 1996 and 2000 estimates respectively. Under both 1996 and 2000 starting conditions, c is the dominant infection pressure for the first three years of the epidemic after which point β takes over.

Model 2: reducing the constant infection pressure

Model 2 assumed a detection rate (α) of 0·006, and that any site confirmed as having KHV had four years of movement restrictions enforced. Predictions were based on a year 2000 introduction, as constant infection pressure (c) had the greatest influence under this scenario. Figure 3 demonstrates that based on the model predictions, removal of the constant infection pressure would have very little effect on the future course of the epidemic, and suggests that fish movements between fisheries have become the main driver in terms of spreading the virus.

Figure 3.

 Results of model 2, investigating the likely effect of removing the constant infection pressure (c) after year 2007.

Model 3: increasing surveillance and the duration of movement restrictions

On its own, extending the duration of fish movement restrictions on sites confirmed as having experienced a clinical outbreak of KHV from four years to permanent had little influence on the future course of the epidemic (Fig. 4).

Figure 4.

 Results of model 3, investigating the likely effect of introducing permanent movement restrictions on waters detected as having a clinical case of KHV from year 2007 onwards.

Introducing an active surveillance programme and placing permanent movement restrictions on infected sites showing no signs of clinical disease, in addition to those exhibiting clinical disease had the greatest influence in reducing the magnitude of the future epidemic (Fig. 5). Model 3 also demonstrated that although on its own the influence of imports on the epidemic is likely to be low, the epidemic cannot be stopped without removing this constant infection pressure, and this is likely to be necessary for a control programme to be effective.

Figure 5.

 Results of model 3, investigating the likely effect of placing permanent movement restrictions on KHV antibody positive waters assuming different levels of case detection and the interaction with the constant infection pressure (c) from year 2007 onwards.

Model 4: the influence of recovery from infection on the effectiveness of control options

In order to estimate the constant infection pressure required to produce 1610 antibody positive sites in 2007 (as observed by Taylor et al. 2010a), and determine how many of these sites were infected and how many had recovered, the model was first run assuming one fishery became infected in year 2000 and that no controls were in place. The constant infection pressure (c) was estimated at 0·017 (39% higher that in the models without recovery), and the rate of detection (α) at 0·0093. This gave rise to 1406 (87%) infected and 204 recovered sites in 2007.

Despite the higher estimate of the constant infection pressure, removal of this post-2007 still had little effect on the epidemic on its own. Again increased surveillance was required to reduce the scale of the epidemic; however, under this scenario permanent movement restrictions were not required and the epidemic could be controlled using the current policy of four-year restrictions (Fig. 6). The removal of the constant infection pressure was however necessary for the control policy to be efficient, otherwise a sustained reduction in the number of infected sites present in the population could not be achieved. When this pressure was removed there was a continual decrease in the number of infected sites. Increased detection increases the number of waters with movement restrictions; however, by removing the constant infection pressure the final number of sites with restrictions was reduced by 70% compared to the situation where this pressure remains.

Figure 6.

 Results of model 4, which assumes that infected sites can recover from infection and that four years of movement restrictions are placed on sites detected as being KHV positive. The figure shows the predicted situation from year 2007 onwards under different levels of case detection and constant infection pressure (c).

Discussion

KHV causes serious disease problems in fisheries and fish farms globally. In England and Wales, it is recognized as an important pathogen due to its impact across different sectors within the ornamental trade and coarse fish (sport fish) communities (Taylor et al. 2010a). Through a simple compartment based modelling approach this study aimed to evaluate the drivers behind the epidemic experienced in England and Wales, and to assess the likely effectiveness of potential management options available to control the pathogen’s spread. Use of alternative methods such as simulation based network models that have been used to study the potential spread of other aquatic pathogens (Thrush & Peeler 2006; Green, Gregory & Munro 2009) was not possible within the scope of this study due to lack of appropriate data. The use of compartment based models to study between site spread of pathogens in other aquatic systems has been demonstrated to be effective, and has been validated where greater data has been available (e.g. Murray 2006).

As with any modelling exercise it was necessary to make a number of assumptions regarding the pathogen and population being modelled. The main assumption of the models regards the degree of contact occurring via fish movements between sites (β). Estimates of β assume that all transmission between fisheries is due to movements of live carp. Although other routes, such as transfer via angling equipment were considered, based on the experimental data presented in this study this route was deemed low risk and was therefore excluded from the estimate. However, due to the high frequency of occurrence there is still likely to be a risk. This could however be mitigated relatively easily through educating anglers about the importance of disinfecting equipment, and the provision of disinfectant baths at fisheries. Other routes of transmission between fisheries that may have led to β being underestimated include spread by other fomites and connectivity via the river network. Due to difficulties in transmitting the virus on angling equipment, other fomites were deemed low risk. Spread via the river network was also deemed low risk, as many fisheries are either fully enclosed (i.e. isolated from the river network) or only have a weak connection between them. Additionally, Shimizu et al. (2006) demonstrated that KHV cannot persist in environmental water for long periods. The main limitation in estimating β was the inability to quantify movements for which consent has not been issued. Adjustments were therefore based on the expert opinion of fisheries inspectors who regularly visit sites and have a good knowledge of what occurs in the field.

Model 1 suggested that fish movements between fisheries alone were unlikely to have resulted in the number of KHV exposed sites observed by Taylor et al. (2010a) in 2007. As mean field equations were used that assumed random mixing, it would be expected that transmission between sites would be over-exaggerated compared to transmission through a network. The fact that the models still under predicted the number of KHV exposed sites in 2007 adds further strength to the hypothesis that there is an additional infection pressure influencing the epidemic. As routes of transmission between fisheries other than live fish movements were deemed low risk, it was assumed that the stocking of ornamental fish into fisheries was the main source of additional infection pressure. This was assumed to be a constant external pressure (c). Based on best evidence available to date, the stocking of imported ornamental fish appears to be the alternative route of highest risk of introducing KHV to fisheries (Taylor et al. 2010a). The stocking of such fish has also been linked to outbreaks of aquatic diseases in other countries (Whittington & Chong 2007). Restricting imports from known KHV infected countries would mean that hobbyists could only obtain fish from countries free of notifiable pathogens, thus reducing the risk such fish pose if released into a fishery.

In the initial years after introduction of the virus, the model results were very sensitive to the constant infection pressure, and it is therefore assumed that in the early phase of the epidemic, external infection pressures such as the stocking of ornamental fish drove the process. This is also likely to be the case for other non-native pathogens introduced to the UK in the future. As the epidemic progressed, sensitivity to c decreased and the models became more sensitive to changes in β and detection rate. This suggests that once a sufficient number of fisheries become infected, fish movements take over as the main driving factor, and is similar to the situation observed by Ruane et al. (2009), for Infectious Pancreatic Necrosis Virus (IPNV) in farmed Atlantic salmon Salmo salar in Ireland. They found in the early stages of an IPNV epidemic in marine salmon sites, introduction of the virus from freshwater sites was the most important route of infection, but that transmission between marine sites rapidly took over as the main driver. Based on the model results and the observations of Taylor et al. (2010b) who found 66% of KHV exposed waters could be linked to a live fish movement from a KHV antibody positive water, the KHV epidemic in England and Wales has now reached a stage where live fish movements have taken over as the main driving force. Consequently reducing the constant infection pressure on its own at this stage in the epidemic is likely to have little effect on the future situation.

An alternative hypothesis to an additional infection pressure being responsible for the 2007 observed level of exposure is that the virus was introduced to the UK earlier than estimated. For this to be the case the virus would have had to have been introduced around the 1980s. This is prior to the first known case of KHV in the world (Perelberg et al. 2003). However, it is possible that a non-virulent form or virus that causes production of cross reacting antibodies exists that could have produced false positive results in the Taylor et al. (2010a,b) study. Although extensive validation was conducted on the enzyme-linked immunosorbent assay (ELISA) method (St-Hilaire et al. 2005, 2009; Taylor et al. 2010a), this possibility cannot be excluded without further surveillance to look for the existence of such viruses.

As movement restrictions are currently only placed on those sites exhibiting clinical signs of disease, only a small proportion of the potentially positive sites have had restrictions placed on them. This has left a large reservoir of antibody positive sites, which if we consider them to be carriers of the virus, still pose a risk to other sites through the fish movement route. As such a small proportion of the potentially positive sites currently have movement restrictions on them, changing the period over which the restrictions lasts makes little difference to the epidemic unless the number of positive sites detected can be increased. Fraser et al. (2004) propose that pathogens that have the ability to transmit for extended periods prior to showing overt clinical signs are difficult to control due to difficulties in detection. In agreement with this, the models used in the current study show the most effective way to slow the epidemic is to actively increase detection of KHV and place long-term movement restrictions on sites testing KHV positive whether they show clinical signs of disease or not. However, this alone is unlikely to stop the epidemic as the constant infection pressure will still be present, and it would be expected that eventually uninfected sites would be exposed to KHV via this route if it remains in place.

Active surveillance has been demonstrated to be more effective in the detection of many diseases and health conditions than passive surveillance (e.g. Piriyawat et al. 2002; Childs et al. 2007; Lesher et al. 2009). However, it can be labour intensive and therefore expensive, and before embarking on such a policy a cost-benefit analysis should be conducted. There are ways in which the effort involved in such a programme could be reduced. In England and Wales, the stocking of fish into inland fisheries requires consent under section 30 of the Salmon and Freshwater Fisheries Act (1975). Such consents are issued by the EA. As part of the consent a health check is required to screen for, and prevent the spread of specific listed pathogens. If KHV were added to this list, active surveillance would be achieved without the need for inspectors to visit and sample the thousands of carp fisheries, which is unlikely to be a practical or cost effective option. This strategy would have the potential to increase the detection rates from the current estimate of 0·6% to the 50% used in the models (assuming 50% of fish movements seek consent). The difficulty in this strategy would be determining the test method to be used. If the ELISA method is used and waters can recover from infection, a high number of false positives could occur, causing controls to be placed on uninfected sites. Use of the PCR (Office International des Epizooties 2009) may have the opposite problem in that false negatives may occur when testing latently infected fish harbouring only low levels of virus in specific tissues. Additionally, disadvantages are that the cost of moving fish between sites could increase if the cost of testing had to be covered by the supplying site, and this policy would only influence those sites responsible enough to apply for movement consents.

Even with the introduction of active surveillance control will be challenging. In addition to active surveillance the duration of controls placed on fisheries needs to be extended if it is assumed that waters do not become free from infection without active intervention. This is because if the water is still infected when controls are lifted it will increase the force of the infection pressure to which the uninfected population is exposed. Unless waters can recover from infection such controls would only reduce further spread of the virus and not reduce the number of infected waters. To reduce the number of infected waters, controlled sites would need to have stock culled, be disinfected and restocked with pathogen free fish. Models 1 to 3 assumed that all antibody positive sites identified by Taylor et al. (2010a) are carriers of the virus, and that once infected they will remain so for life. Although this is unlikely to be the case for all sites, there is growing experimental evidence and field-based observations that suggest this scenario may be true for many sites (St-Hilaire et al. 2005, 2009; Yuasa et al. 2007; Uchii et al. 2009). Under model 4 which assumed that infected fisheries can recover from infection, the above recommendations regarding surveillance and import restrictions holds true, however permanent controls are not required on infected sites. In contrast to the previous models, under this scenario it also possible to achieve a reduction in the number of infected waters present in the population if sufficient levels of detection can be achieved as opposed to just limiting the number of sites that get infected. This model is, however, sensitive to the recovery rates used and under the proposed estimates the pathogen does not die out. It is not known whether sites can recover from infection and further research is required to determine this and the rates at which recovery may occur, however, unless recovery rates are higher than those used in this model, movement restrictions in excess of four years or additional culling would also be required to eradicate the virus.

In summary, this modelling exercise suggests that control of KHV in England and Wales is likely to be challenging, but may be possible with the introduction of intensive measures. The scenarios run using all of the above models demonstrated that in addition to an active surveillance programme, removal of the constant infection pressure is also required for a control policy to be effective. Although removal of this on its own was shown to have little influence on the epidemic, the progression of the epidemic cannot be stopped without its removal. Additionally, to be effective, a control programme would require movement restrictions to be placed on all infected sites detected and not just those expressing clinical disease. If KHV has the ability to cause latent infections and sites do not recover from infection, movement restrictions would need to be permanent (or at least long-term), and culling of infected sites is also likely to be required if the aim were to eradicate the virus from England and Wales. In addition to the above recommendations, anglers, hobbyists and fishery owners require information about KHV that highlights the importance of disinfecting equipment and not introducing fish to a fishery unless their origin is known and the movement is consented.

This study was only able to model the ‘higher-risk’ type of fishery as also studied by Taylor et al. (2010a), which regularly introduce fish to site. It was not possible to study what are often regarded as specimen waters which contain low numbers of high value fish. Such sites rarely move fish on to site and consequently the constant external infection pressure is likely to be the main route of introduction. Restrictions on the import of ornamental fish from countries known to have KHV may therefore have more benefit to reducing the likelihood of such sites becoming infected.

Acknowledgements

The authors would like to thank Claire Joiner for her excellent technical assistance and Defra for funding this study under projects FB001 and FC1167. Gratitude is also extended to Dr Olivier Restif (University of Cambridge) for his advice and insights.

Ancillary