School of Biodiversity, One Health & Veterinary Medicine

Introduction

This project proposes to combine insights from population and epidemiological modelling with those from landscape ecology to develop improved methods for the design of programs to control infectious disease in wildlife by vaccination. Considerable progress has been made in developing the theory of infectious disease control with the use of simplified and transparent models. However, there is a clear gulf between the simplified scenarios in which epidemiological theory is developed, and the spatial complexities of real landscapes and the idiosyncrasies of real populations in which particular disease control programs need to be applied. Substantial resources are invested in the research and development of disease control products, and in their purchase by health and veterinary organizations, however, the operational theory of how to deliver control has been largely neglected. The result has been only a limited improvement, and in some places a worsening, of the impacts of disease, particularly in the developing world . There is therefore an urgent requirement to combine insights from the dynamics of simple epidemiological models with the increasingly available quantities of spatial data describing real landscapes, and the distribution of populations that live on them, in order to improve the design and implementation of disease control programs. This proposal brings together a uniquely qualified and well-balanced team of ecological and epidemiological modellers, with veterinary epidemiologists possessing a wealth of applied experience in conservation science in order to meet this requirement.

Epidemiological context

Pathogens that cause infectious disease can only persist in populations that maintain a sufficiently high number of individuals that are susceptible to infection . This threshold is known as the critical community size (CCS). Habitat destruction and fragmentation have led to rare and endangered species surviving in only small and isolated populations well below the CCS for pathogens that afflict them. However, the persistence of rare and endangered species is increasingly threatened by infectious disease caused by pathogens that can infect multiple host species. Such pathogens can persist in large populations of an alternative host, and occasionally be transmitted into much smaller populations of rare and endangered species. These 'spill-over' infections give rise to sporadic epidemics in populations that would otherwise remain free from infectious disease, and can reduce populations to critically low numbers below which extinction might arise through demographic stochasticity or the action of other mortality processes. Past definitions of infection 'reservoirs' have been ambiguous in some respects, but here we define an infection reservoir as comprising one or more connected populations of a host species that collectively exceed the CCS for the host-pathogen interaction. We refer to the 'target' population as one in which disease control is ultimately required, and into which infection may spill-over intermittently from a reservoir.

There are examples of wildlife populations constituting reservoirs for target populations of domestic livestock (for example, it is likely that badgers constitute a reservoir for TB, with spill-over into domestic cattle), but our focus here is on reservoirs of domestic animals that constitute sources of infection for wildlife populations. The threat from spill-over infections is increasing as wildlife populations become smaller, and more isolated, and as domesticated reservoir populations become larger, and transgress with greater frequency into remaining areas of natural habitat. This threat to the target population can be countered by three broadly different strategies, disease can be managed: 1) in the reservoir thereby reducing the frequency of spill-over infections; 2) directly in the target population thereby reducing the impact of spill-over infections when they arise; or 3) control measures can be directed at preventing infection passing from the reservoir to the target population at all. This proposal will develop epidemiological models to study how vaccination can be used to support implementation of each of these strategies in order to control infectious diseases maintained in reservoirs of domestic dogs. Specifically, we will focus on domestic dog populations around the edge of the Serengeti National Park (SNP), Tanzania, and the Bale Mountains National Park (BMNP), Ethiopia.

In both countries, the threat posed to wildlife by infectious disease maintained by domestic dogs is increasing . This is because the number of people around the park perimeters is increasing, the average number of dogs per household is increasing, and the buffer zones that separate the park from areas of human settlement are being reduced in size, thereby bringing wildlife populations into closer contact with reservoir populations. Within this context, we propose to study the design of vaccination strategies against three canid viral infections often endemic to local domestic dog populations in the developing world.

Rabies virus. Rabies epidemics initiated from spill-over infections from domestic dogs have sharply reduced populations of the Ethiopian wolf in the BMNP on at least two occasions in the recent past, and has been implicated in losses of African wild-dogs in the SNP.

Canine Distemper Virus (CDV). CDV has been responsible for devastating outbreaks in lions in the SNP, is likely an important source of mortality for wild-dogs , and has recently been diagnosed for the first time in Ethiopian wolves in the BMNP.

Parvovirus (PV). PV does not usually cause high rates of mortality in adult canids, but is thought to impact significantly on juvenile survivorship of wild canids, and therefore limits the ability of populations to rebound following reductions caused by other agents of mortality.

This project is timely for a number of reasons: 1) It is clear that domestic dog populations are growing rapidly and pose a steadily increasing threat to wildlife; 2) This threat is increasingly recognized by park authorities and relevant regional political authorities who are reviewing the use of vaccines in both domestic dogs and wildlife; 3) New generations of vaccines are under development that provide longer periods of protection against infection, and use new and more efficient delivery methods.

The simplest epidemiological theory indicates that the proportion of a population that must be vaccinated to eliminate an infectious disease (pc) is equal to 1-1/R₀, (where R₀, the reproductive ratio, is the number of secondary cases arising from a single infected animal introduced into a fully susceptible population). A vaccination coverage of p c ensures that the reproductive ratio is brought below 1, and that therefore infected individuals transmit disease to less than one additional susceptible individual, and so causing incidence to decline.

For rabies in domestic dog populations, R₀ is thought to be between 2.5 and 4 suggesting that pc should be between 60-75% . Current vaccination programs around our study sites aim to vaccinate at least 60% of dogs in all targeted populations. However, in common with so many other infection-reservoir situations, there are features of our particular disease control problem that are not readily addressed by this basic theory, and to which an optimal vaccination strategy is likely to be sensitive:

• Simple theory assumes the population is one single 'well-mixed' entity. In reality populations are usually comprised of multiple sub-populations of varying sizes, and transmission is much more likely to take place within sub-populations than between them. The optimal vaccination strategy in these more complex 'city and village' situations is much less straightforward, and depends on the precise details of the spatial configuration of the host populations, and the presence of natural barriers to transmission that may exist between them.
• A vaccination program may be able to combine natural features of the reservoir population's spatial distribution with targeted vaccination effort in selected parts of the landscape so as to sub-divide the dog populations into closed compartments each of which is below CCS. This could increase the frequency of stochastic fadeout of infection from significantly sized fractions of the dog population.
• Since the required vaccination coverage depends on R₀, it is also important to understand how optimal vaccination strategy might be affected by different relationships between R₀ and dog population size and density, particularly in populations that are growing.
• In principle, vaccination may be delivered on a continuous or intermittent basis. For logistic reasons, vaccination is usually conducted intermittently - a strategy known as pulse-vaccination. The required frequency of re-vaccination in sub-populations will depend on the rate of recruitment of susceptible dogs (through birth or immigration), on the life-span of vaccinated dogs (which may be longer than unvaccinated dogs, and depend on what they are vaccinated against), on the mean and variance of the duration of immunity provided by vaccination, and on the fraction of dogs in a village that are never presented for vaccination, regardless of how often vaccination services are offered.
• Vaccination theory is often developed in the context of human disease control, where the objective is eradication of infectious disease (vaccination to reduce R₀ < 1). However, for the purposes of conservation, the objective of disease control is to ensure the persistence of endangered populations. This objective could be secured by vaccinating a minimally viable subset of the target population (so-called 'core vaccination'), rather than eliminating infection from the reservoir, and this may be logistically less demanding.

While there are strong arguments for the use of simplified, transparent and analytically tractable models, experience has shown that when quantitative judgments must be made about specific epidemiological processes and where real management decisions have to made, these approaches are insufficiently realistic to address questions with confidence and precision. For example, during the 2001 foot-and-mouth disease virus (FMDV) outbreak in the UK the model that provided the most useful insights into control options, and the most confidence, used a stochastic Susceptible-Exposed-Infectious-Removed (SEIR) process imposed on a very detailed spatial distribution of host densities. It could be argued that deficiencies in our understanding of how to apply the abundant stocks of FMDV vaccine contributed significantly to the decision to control the 2001 outbreak by other means. Since 2001, detailed computer models conceptually similar to those we intend to develop and explore in this proposed research, have been developed to study how best to use vaccination to combat FMDV outbreaks in the UK, and - similarly - anti-viral drugs to control Avian Influenza in Thailand . The high scientific profile these studies have acquired indicates that these are real knowledge deficiencies that can and should be addressed with state-of-the-art epidemiological science.

This project aims to capitalize on the existence of long established field programs in two study areas, each led by co-applicants of this proposal. Each study area comprises a national park of internationally recognized biodiversity value containing key populations of threatened carnivores, and a surrounding buffer-zone under increasing pressure from human settlement. The field programs have already implemented large-scale, long-term controlled vaccination programs of domestic dogs in the buffer-zones that have generated vast quantities of unique and valuable epidemiological, immunological, and dog demographic data, and from which considerable expertise regarding the empirical practicalities of control programs is available. Additional data are available on wild carnivore abundance (from long-term transect studies), human bite injuries (from hospital data and uptake of post-exposure treatment), movement of wild-canid species and domestic dogs (from behavioural studies and radio-collaring projects), and patterns of human population settlement (from Government census data, and questionnaire studies). Upwards of $2 million has been spent on implementing these programs but resources have not previously been available to develop and explore epidemiological models of these systems. Data from these programs provide us with a unique opportunity to study long-term disease control in the context of two similar information-rich systems of global conservation significance.

Recommendations for improving the effectiveness and efficiency of vaccination programs that arise from the modelling work will be incorporated into on-going field operations across these study sites at appropriate spatial scales, and on-going monitoring work will enable potential improvements in performance to be assessed.

The BMNP sustains half the world’s population of the world's rarest canid, the endangered Ethiopian wolf (Canis simiensis). Research activities started in 1986 and conservation actions initiated in 1996 under the Ethiopian Wolf Conservation Programme, of which Dr Karen Laurenson is co- founder, means that detailed data on the demography, distribution, and social structure of the Ethiopian wolf population are available. Rabies outbreaks traced back to domestic dogs have caused sharp declines in this population in 1991-2 and 2003. The 2003 outbreak was controlled by the use of parenteral vaccination in wolves, the first time this has been permitted in wildlife in Ethiopia. Use of oral vaccination (currently not permitted) is now under review by the Ethiopian government and placebo trials have already been authorised.

In 2006, mortality due to CDV was identified for the first time in this species (Laurenson, unpublished data), although a previous outbreak was suspected. Approximately 41,000 people are located within the park and a further 200,000 people in a buffer zone within 10km of the park boundary. The number of dogs per household is estimated at 1.53 in rural areas The dog population is estimated to be growing at 8% per year, and human and livestock use of the park is rising steadily, with consequent increase in interactions between domestic dogs and wolves. Domestic dog pulse-vaccination has been conducted within the park and in some areas of the buffer-zone since 1996. Vaccination coverage has ranged between 30-50%, but is higher immediately after annual pulse vaccinations, and long-term monitoring work has demonstrated the beneficial impact of this vaccination on the prevalence of canid diseases in the domestic dog population. Moreover, large-scale but detailed research and monitoring work between 1997 and 2001 has resulted in the collection of considerable additional data on the demography, serology and epidemiology of canid diseases in the reservoir population, and most importantly – the spatially heterogeneous distribution of domestic dogs in the buffer-zones.

The SNP is a world heritage site of unique conservation value. Infectious disease has impacted significantly on wildlife populations on at least two occasions in the recent past. Infectious diseases are widely believed to have been a major contributory factor to the demise of Serengeti wild dogs in the early 1990s; and a CDV outbreak in 1994 resulted in over a 30% reduction in the parks lion population. Epidemiological evidence supports the view that domestic dogs are the most likely sources of these outbreaks. More recently, virus recovered from carcasses of hyena’s, bat-eared foxes, and felids have been shown to share almost identical genotypes to those found in domestic dogs in the buffer-zones surrounding the park (Cleaveland, unpublished data).

Like the BMNP, human settlement is increasing around much of the park boundary, particularly on the western side, adjacent to Lake Victoria, and with this settlement comes a population of domestic dogs estimated to be growing at 5-10% per year. Vaccination of wildlife inside the park is currently not permitted by the SNP authorities. The Canivore Diseases Project (led by Dr Sarah Cleaveland) has undertaken parenteral pulse-vaccination of domestic dogs in various administrative districts around the park perimeter since the late-1990s, achieving coverages of around 60-70% in as many as 145 villages, and, as in the BMNP, the positive impact on disease prevalence has been unequivocally demonstrated. However, rabies elimination has not yet been achieved, and many questions still remain about how best to optimise delivery strategies and to design sustainable programmes. Concurrent monitoring activities have resulted in the collection of large quantities of data on demography, spatial distribution, serology and epidemiology of the reservoir population.

 Figure 1. A) Map of the Bale Mountain National Park, Ethiopia. The park boundary is delineated by the outer polygon, wolf habitat is indicated in light-gray, and the dark-gray shading indicates the distribution of wolf-pack territories. Villages are indicated by solid dots. B) Map of the Serengeti-Mara ecosystem. The protected area boundaries are delineated by the polygons: i, Serengeti National Park; ii, Maswa Game Reserve; iii, Ngorogoro Conservation Area; iv, Loliondo Game Reserve Area; v, Masai-Mara Game Reserve, Kenya. Villages are indicated by solid dots.

• We will develop epidemiological models that use high resolution spatial information on the distribution and demography of host populations and that includes explicit capacity to manipulate numbers of susceptible individuals through realistic vaccination programs.
• We will use pre-existing long-term data from on-going field programs to enable application of this model framework to the control of canid viral diseases circulating in reservoir and target populations in and around national parks in Ethiopia and Tanzania.
• We will analyse these models to identify specific recommendations for the design of vaccination programs in and around these two sensitive areas of conservation concern.
• We will synthesize and generalize these findings to improve our understanding of the principles governing the control of infectious disease within spatially distributed and poorly-mixed populations at a landscape scale so that they can be related to control of disease in other reservoir systems.

We will develop epidemiological models that we will apply and parameterise to each of our two study systems to seek quantitative answers to the following questions:

• What is the relationship between vaccination coverage and disease incidence in the reservoir population?
• How should vaccination coverage be optimally allocated between villages and towns containing different numbers of dogs?
• Can features of the landscape (such as rivers and hills) and socio-economic infrastructure (such as roads, bus routes, and markets) be accounted for to improve the efficiency with which disease incidence in the reservoir is reduced by vaccination?
• How does the optimal pulse-vaccination frequency depend on the mean duration of immunity provided, and on the shape of variation around this mean?
• How should the optimal pulse-vaccination frequency and vaccination coverage be modified in response to increasing populations of domestic dogs, and to account for the effects of the vaccination program itself on dog demography?
• What is the expected range of variation of disease incidence for a particular vaccination program expected to arise from purely natural demographic stochastic processes? (this is essential knowledge required to judge the effectiveness of program implementation).
• Can vaccinating on a more continuous basis through time, or organizing pulse-vaccination campaigns simultaneously over large areas lead to improvements in the efficiency of vaccination campaigns?
• How important is the possibility that dogs missed at one pulse vaccination event may bea non-random sample of those missed at the next?
• Is it feasible to use vaccination strip-zones to partition the reservoir into small populations that are below the critical community size required to maintain endemic infection?
• In the BMNP, direct responsive core vaccination of the target population (the Ethiopian wolf) to the confirmed presence of the disease is a potential control policy option. How many individuals (and of what age, sex, and social status) should constitute this core? How should they be distributed within and between packs? How does the optimal location for the core population change as a function of where in the population disease arises?

This research is of primary importance for the conservation of rare and endangered species. Without a thorough, practical understanding of disease control in real, complex landscapes, there is little hope of protecting many populations from the decimating effects of infectious diseases that pose a grave and imminent threat to these populations. This study brings together leading modellers and wildlife epidemiologist to provide on-going disease control programs of primary global conservation importance with a novel theoretical and quantitative foundation. By doing so, real threats posed to species of conservation concern (particularly wild dogs in the SNP, and Ethiopian wolves in the BMNP) by canid viral diseases will be significantly mitigated.

The pervasive threat posed to small target populations by diseases endemic to reservoirs of alternative host species lends considerable generic value to this research. The methods and findings of this research will address important but neglected practical questions that fall between the disciplines of epidemiological theory and veterinary epidemiology. The answers will improve the efficiency with which hard-won disease control resources are allocated and distributed and be relevant to the control of disease in other disease reservoir systems. This project will directly fuel future research opportunities with the Ethiopian Wolf Conservation Program, the Carnivore Disease Project, and other disease control initiatives to implement trials of resulting vaccination protocols. An important additional benefit will be improved cost-effectiveness of rabies vaccination for reducing the human health and economic burden of rabies in the developing world, where domestic dogs are the principal reservoirs and source of infection for humans.

These project goals can only be met by combining the diverse skills of a number of uniquely qualified and experienced researchers, who have a past history of productive collaboration:

Hawthorne Beyer is a highly experienced spatial systems analyst with specialist skills in modelling spatial dynamic processes. He will bring to this project advanced computing, programming and image analysis skills, together with a wealth of experience concerning the representation of spatial ecological processes and how to use information from GIS in these analyses.

Prof Daniel Haydon (PI) is a mathematical biologist with wide-ranging research interests at the interface of epidemiology, population and landscape ecology. He has substantial experience of epidemiological and population modelling, particularly at landscape scales.

Prof Sarah Cleaveland is a veterinary epidemiologist who has conducted field-work in Tanzania for over 10 years. As the scientific coordinator of the Tanzanian Carnivore Disease Project she has extensive empirical experience of conducting domestic dog vaccination campaigns around the Serengeti.

Dr Karen Laurenson is a veterinary epidemiologist and conservation biologist who now works for the Frankfurt Zoological Society. Dr Laurenson is a founding member of the Ethiopian Wolf Conservation Programme and has organized vaccination programs both in domestic dogs around the Bale Mountains National Park, and in populations of Ethiopian wolves directly (in response to a rabies epidemic in 2003).

Haydon, D.T. , Cleaveland, S., Taylor, L.H., and Laurenson, M.K. (2002). Identifying reservoirs of infection: a conceptual and practical challenge. Emerging Infectious Diseases 8, 1468-1473.

Haydon, D.T. , M.K. Laurenson, and C. Sillero-Zubiri. (2002). Integrating epidemiology into population viability analysis: managing the risk posed by rabies and canine distemper to the Ethiopian wolf. Conservation Biology 16, 1372-1385 .