Development of vaccination strategies for the management of rabies in African wild dogs

Abstract

The standard objective of a vaccination strategy is to reduce the reproductive ratio, R₀, defined as the number of secondary host infections arising directly from introduction of a single infected individual into an otherwise fully susceptible population, to below 1. This ensures that only very small outbreaks are likely to arise. However, this objective usually requires a high level of vaccination coverage that is often expensive and logistically difficult to achieve. For the purposes of conserving rare species that are threatened by outbreaks of infectious disease, population persistence may be assured by a vaccination strategy designed to suppress only the largest outbreaks of disease that reduce the population to below a minimum viable population size. Such strategies targeting only a viable minimal ‘core’ of the population are likely to be logistically less demanding. Here we explore how these core vaccination strategies might be designed for African wild dogs (Lycaon pictus), an endangered canid species whose remaining populations are threatened by rabies outbreaks. We develop and analyse a population viability model with an explicit epidemiological model embedded within it. The model predicts that core vaccination campaigns, using vaccines that provide two years of immunity, and targets 30–40% of individuals within a population every 1–2 years would be successful in ensuring persistence of small populations, and required coverage levels can be even lower in larger populations. This strategy appears robust to synchronized introduction of rabies into packs, possible Allee effects, and the use of vaccines providing only one year of immunity.

Introduction

The reproductive ratio, R₀, of an infectious agent in a particular population is defined as the number of secondary infections arising directly from the introduction of a single infected individual in an otherwise fully susceptible population. Standard epidemiological theory indicates that in a homogenous well-mixed population it is necessary to vaccinate a proportion 1 − 1/R₀ of individuals in the population to bring the reproduction rate in the population as a whole, R_v, to below 1, and prevent outbreaks of disease (Anderson and May, 1992). This does not prevent individuals amongst the remaining unvaccinated fraction 1/R₀ of the population from acquiring infection from an external source, but if they do, they will on average pass it on to less than one additional member of the population. For diseases like rabies in populations of domestic dogs, where R₀ is thought to range between 2 and 4, the vaccination coverage (here defined as the percentage of a target population that is vaccinated) required to bring R_v to below 1 is anticipated to be around 70% (Coleman and Dye, 1996), a value empirically confirmed by past vaccination campaigns (Cleaveland and Dye, 1995, Coleman and Dye, 1996, Kitala et al., 2001, Cleaveland et al., 2003).

The objective of vaccination strategies has largely centred on reducing R_v to below one, and the consequent elimination of all but the smallest outbreaks of disease. While this is certainly a requirement of a policy seeking regional or global elimination of an infectious disease, it may not be a requirement of a program with an objective of ensuring the persistence of a rare or endangered species that is threatened by infectious disease. Persistence of these host populations does not necessarily require the elimination of disease, but could be accomplished by eliminating the largest outbreaks of disease that bring populations below a minimum viable population size (MVPS) beyond which risk of extinction is unacceptably high. Core vaccination strategies, defined here as strategies that vaccinate a sub-set of the population equivalent to the MVPS, should ensure that population recovery occurs following disease-induced population crashes. Core vaccination requires substantially lower vaccination coverage levels than policies that aim to eliminate all outbreaks. The challenge is to identify a safe MVPS for different populations so that required vaccination coverage levels can be determined.

In this study, we aim to provide an overview of the arguments in favour of developing rabies vaccination strategies for African wild dogs (Lycaon pictus), and go on to present the results from a model that could inform how such core vaccination strategies might be implemented for this species.

Although disease may not be the most serious current threat to the persistence of all wild dog populations, rabies (RABV), canine distemper virus (CDV), and canine parvovirus (CPV) are all believed to impact on wild dog demography. The trophic position of wild dogs, their intense competition with other wild carnivores (especially lions and spotted hyenas, Creel and Creel, 1996), high degree of sociality, and close taxonomic relationship to the domestic dog may all increase exposure to, and transmission rate of infectious disease (Woodroffe et al., 2004). Woodroffe and Ginsberg (1997) suggest that wild dogs’ social organization might not only facilitate transmission of disease through intense contact between pack members, but might also hamper selection for disease resistance. Survivors of local epidemics can only pass on genes for disease resistance following dispersal and new pack formation, a process that entails high mortality risks (Ginsberg et al., 1995) and results in the formation of mostly small packs that may not be viable (Courchamp et al., 2000). However, this effect may be limited to certain diseases; evolution of disease resistance following epidemics of rabies, for example, has not been documented in any wild canid populations. In several carnivore species, including wild dogs (Gascoyne et al., 1993b), hyena (East et al., 2001) and domestic dogs (Cleaveland et al., 1999), rabies antibodies have been detected in healthy animals, suggesting that some individuals mount an immune response that can clear virus before progression to the central nervous system and development of clinical signs (‘aborted’ infection). Whether an individual develops clinical rabies (and hence dies) after having been bitten by a rabid animal depends on many factors (such as the site and severity of the bite injury, inoculating dose, and virus strain), but the role of host genetic factors is still largely unknown.

Rabies has resulted in declines in several African wild dog populations. The most severe impacts have been recorded in the Serengeti–Mara ecosystem of Tanzania and Kenya, where rabies outbreaks in the late 1980s and early 1990s caused successive pack losses, resulting in the local extinction of the population in 1991–1992 (Gascoyne et al., 1993a, Kat et al., 1995, Woodroffe, 2001). Although other factors may have driven the longer-term decline of this population, disease undoubtedly played a major role in its final disappearance. Rabies has also resulted in two successive outbreaks in wild dog packs in Madikwe, South Africa (Hofmeyr et al., 2000, Hofmeyr et al., 2004), and has been identified as the cause of the loss of five packs in Botswana (McNutt cited in Woodroffe et al., 2004) and contributed to the failure of three re-introduction attempts in Namibia (Scheepers and Venzke, 1995). Rabies is also known to have killed wild dogs in the Central African Republic (Turkalo, cited in Woodroffe and Ginsberg, 1997), and is believed to have killed wild dogs in Zimbabwe (Foggin, cited in Kat et al., 1995) and Zambia (Buk, cited in Woodroffe and Ginsberg, 1997).

Vaccines have been used to protect wild dogs against rabies on a number of occasions, with variable success (reviewed by Woodroffe et al., 2004). There is growing evidence that single-dose administration of current inactivated rabies vaccines may be insufficient to protect free-ranging wild dogs against outbreaks, with rabies confirmed in vaccinated wild dogs on at least three occasions (Woodroffe et al., 2004). However, in Madikwe, South Africa, adult wild dogs that had received multiple doses of rabies vaccine survived a rabies outbreak whereas unvaccinated pups did not (Hofmeyr et al., 2004), demonstrating a beneficial impact of rabies vaccination, but reinforcing the impression that multiple doses may be needed to ensure protection.

Given the difficulties of administering injectable vaccines to free-ranging wild dogs (which requires either capture or dart-inoculation of individuals), interest has focused on the potential use of oral vaccines in disease management. Although oral rabies vaccines have been deployed with great success in the control of rabies in wildlife reservoirs such as red foxes, coyotes, and raccoons as reviewed by the European Commission, 2002, World Health Organization, 2004, oral vaccines have not yet been widely used for disease prevention in endangered species. However, recent trials in South African wild dogs using SAG-2 vaccine demonstrate seroconversion with no apparent negative side-effects following administration of oral rabies vaccines (Knobel et al., 2003), and have identified effective baiting strategies for wild dog packs (Knobel et al., 2002).

Canine rabies in sub-Saharan Africa is widely considered to be an exotic legacy of European exploration and colonisation, with the canid strain introduced by infected dogs and becoming established in domestic dog populations throughout the continent (Smith et al., 1992). Over the past three decades, canine rabies appears to have been increasing throughout the continent, both in incidence and range of species affected (Perry, 1995, Cleaveland, 1998). In Africa, the growth rate of domestic dog populations in rural areas typically ranges between 5% and 10% per annum (Cleaveland, 1996, Laurenson et al., 1997b, Kitala et al., 2001), exceeding that of human population growth. With continuing demand for dogs for guarding, herding, and (in some areas) hunting, the growth in African dog populations is set to increase. As a result, there is growing concern that dog populations which previously could sustain only sporadic, short-lived epidemics, such as the pastoral populations of northern Tanzania (Cleaveland and Dye, 1995), may now act as reservoirs of rabies. As human and domestic dog populations continue to expand and encroach into former wildlife areas, the likelihood of contact between domestic dogs and wildlife will inevitably increase (Butler et al., 2004). High rates of contact are likely to occur throughout East Africa, where wildlife-protected areas are generally unfenced and wildlife migratory routes often extend into areas settled by people.

The effects of a population crash could quickly reduce the population to below MVPS, from which extinction is the most likely outcome. In reality MVPS for any population is best thought of as a probability expressed as a (usually declining) function of absolute population size (Fig. 1), and it follows that a proportionate loss in a small population is much more likely to result in extinction than the same proportionate loss in a larger population. The concept of MVPS for social and pack-structured species is complex. Not all individuals are equivalent, and the particular combination of individuals surviving a population crash (i.e. their sex, and status) is likely to determine MVPS. Reduced packs that retain alpha-females and dominant males might survive with greater likelihood than those that do not. Furthermore the manner of population reduction may influence MVPS. Sudden and traumatic population crashes might induce a level of social dysfunction causing pack break up. Evidence from another canid species, Ethiopian wolves, suggests that it is adults (Randall et al., 2004), and possibly the most dominant adults that are at greatest risk of exposure to rabies from their closer contacts with infected prey items, other canid species, and encounters with individuals from other packs.

The Allee effect (Allee, 1931) – which describes a process of inverse density dependence such that per capita population growth rates decline at very low densities serves to increase the MVPS (Fig. 1). Allee effects are well established in wild dog populations and are thought to arise from variation in the hunting efficiency of packs of different sizes, their subsequent ability to defend kills from other predators such as lions and hyaenas, and the need for helpers to successfully raise pups (Courchamp et al., 2000). The demographic manifestation of an Allee effect is on the number of pups that are recruited to yearling age, and the rate at which small packs break up into single sex groups. In species that exhibit Allee effects, epidemics are more likely to result in the extinction of populations.

At a small enough spatial scale, localized extinction is a regular occurrence, with animals dying within or moving out of certain areas. As long as individuals in neighboring areas can recolonize voided areas the long-term demographic consequences may be negligible. Recolonizing individuals may originate from more secure and populous ‘source’ populations perhaps residing in more protected areas (Pulliam, 1988), or the population may be structured as a metapopulation in which a (dynamically changing) fraction of habitat tends to be occupied at any single point in time (Hanski, 1991). Provided that recolonisation can be realistically assumed, local extinction need be of no particular concern. However there are two reasons to be sceptical of this assumption: First, variation in mortality rates may fluctuate synchronously across the metapopulation, in which case colonists from neighboring habitat may not be available (Ovaskainen et al., 2002). Synchronous occurrence of disease-induced mortality in different packs could arise through transmission directly between wild dog packs, or through increases in the common exposure rate experienced by different packs that might result from changes in prevalence in a shared reservoir (as a result of a wider epidemic, or seasonality). Second, neighboring populations may not be able to colonize vacant habitat because they are unable to (or incur very high mortality attempting to) move between habitat patches as a result of fragmentation or severance of natural corridors (Ginsberg et al., 1995, McNutt, 1995) in which case we would expect the occupancy rates of these now isolated habitat patches to ratchet inevitably downwards, with consequent wider-scale regional extinction.

Population viability analysis (PVA, Morris and Doak, 2002) has been used in the past to examine the effects of various sorts of demographic disturbance on the persistence of wild dog populations. In these previous studies, disease outbreaks are not modelled explicitly as infectious-dynamic processes but represented as short-term reductions in adult survival and fecundity. For example, Woodroffe and Ginsberg (1997) used the population viability model VORTEX (v. 7) to measure cumulative probability of extinction per decade over 50 years, in order to assess the impact of various threats to African wild dogs. Their results indicated that even relatively small increases in pup mortality could be sufficient to drive some populations to extinction if new causes of mortality act in addition to existing ones. The authors concluded that introduction of diseases that kill pups but rarely adults (e.g. CPV) could contribute to the extinction of even relatively large wild dog populations. The authors acknowledged significant limitations: VORTEX does not have the capacity to adequately incorporate some critical components of canid biology, for example a pack based-single female breeding system or social structure. Furthermore, VORTEX assumes that stochastic effects such as catastrophes influence each sub-population independently making the modelling exercise highly conservative in that it underestimates the extinction risks threatening wild dogs.

Vucetich and Creel (1999) constructed an individual-based model to simulate the behaviour of an African wild dog population (based on data from a 6-year field study in the Selous Game Reserve in Tanzania), in which lion population dynamics constitute a strong influence on wild dog dynamics. Simulations showed that outbreaks of infectious diseases that kill adults, such as rabies, reduced population persistence if they increased annual mortality by 30% and occurred at intervals of less than 10 years. In contrast, diseases killing only pups such as CPV had weaker effects on persistence. These results are supported by similar modelling approaches to the persistence of Ethiopian wolves (Mace and Sillero-Zubiri, 1997, Haydon et al., 2002) which suggest that CDV is not a serious threat to persistence, but that sufficiently frequent rabies outbreaks can be a problem particularly in small populations between which there is little exchange of individuals. Vucetich and Creel (1999) noted that future models would benefit from improved data on the degree to which human-caused mortality is additive or compensatory, on age-specific mortality due to infectious diseases, on transmission rates among species, on the demography of dominant competitors and on the mechanisms by which they affect wild dogs.

Because wild dogs have a complex pack structure that determines to a large extent, the annual rate of recruitment, there is a strong justification for the use of models that include an individually based representation of packs that includes some age and sex structure. Furthermore it is apparent that while we know little about inter-pack transmission, and the causes of synchronous outbreaks across a population of wild dog packs, such synchrony in the occurrence of outbreaks has been observed in the Serengeti–Mara (Woodroffe, 2001) and can be an important determinant of population persistence time. One way to represent this synchrony is to treat rabies as an explicitly infectious process, and include its dynamics as a stochastic susceptible–infected–removed/vaccinated (SIR-V) process within the PVA model. To do so confers a number of additional advantages: (1) it permits realistic variation in rabies outbreak size that varies as a function of population size (including disease fade-outs); (2) rates of rabies introduction can be synchronized across packs representing seasonality or between pack transmission; (3) age-related susceptibility or exposure rates are easily included; (4) only by explicitly modelling the SIR-V dynamics can the effects of different sorts of vaccination program be explored.

In what follows we describe and apply an individually based, stochastic, age and sex structured model of wild dog demography that includes an SIR-V process integrated into this demographic process. We use this model to explore the relative extinction risks over 20-year time horizons of wild dog populations comprised of different numbers of packs and exposed to different rates of rabies introduction from an implicit reservoir. We then explore the ability of different forms of vaccination program to reduce extinction risk in the model assuming a safe and effective vaccine to be available. We explore the effectiveness of different vaccination strategies in reducing extinction risk from rabies in populations of different sizes, when infection is introduced synchronously across packs, and when Allee effects are inherent in the population dynamics.