Distribution and Impacts of Tasmanian Devil Facial Tumor Disease
- First Online:
- Cite this article as:
- McCallum, H., Tompkins, D.M., Jones, M. et al. EcoHealth (2007) 4: 318. doi:10.1007/s10393-007-0118-0
- 2.5k Downloads
The Tasmanian devil, Sarcophilus harrisii, is the largest extant marsupial carnivore. In 1996, a debilitating facial tumor was reported. It is now clear that this is an invariably lethal infectious cancer. The disease has now spread across the majority of the range of the species and is likely to occur across the entire range within 5 to 10 years. The disease has lead to continuing declines of up to 90% and virtual disappearance of older age classes. Mark-recapture analysis and a preliminary epidemiological model developed for the population with the best longitudinal data both project local extinction in that area over a timeframe of 10 to 15 years from disease emergence. However, the prediction of extinction from the model is sensitive to the estimate of the latent period, which is poorly known. As transmission appears to occur by biting, much of which happens during sexual encounters, the dynamics of the disease may be typical of sexually transmitted diseases. This means that transmission is likely to be frequency-dependent with no threshold density for disease maintenance. Extinction over the entire current range of the devil is therefore a real possibility and an unacceptable risk.
KeywordsTasmanian devilinfectious cancerextinctiondisease ecology
Signs characteristic of Tasmanian devil facial tumor disease (DFTD) were first detected in photographs taken at Mt. William National Park, Northeast Tasmania in 1996. The first case of DFTD confirmed through histological examination was collected in 1997 (Loh et al., 2006) near the photographic report. It has now spread over the majority of the range of the species, leading to major population declines. In this article, we review and update current knowledge about the distribution and spatial spread of the disease, its observed impact on affected host populations, and models to predict its consequences for host population dynamics.
DENSITY AND DISTRIBUTION OF DEVILS
It has been suggested that devil population numbers may have fluctuated substantially in the 19th century, perhaps in association with disease (Guiler, 1964, 1992). This idea was used as the basis of a model (Bradshaw and Brook, 2005), suggesting that density-dependent disease may have been responsible for these previous fluctuations. Owens and Pemberton (2005) investigated the anecdotal evidence for substantial fluctuations in devil density, and found no grounds for any firm conclusion to be drawn. Devils have unusually low genetic diversity for a mammal (Jones et al., 2004), which is consistent with a founder effect when Tasmania became isolated at the end of the last Ice Age, but is not necessarily evidence of any more recent bottleneck.
DISTRIBUTION AND SPREAD OF DISEASE
Two methods have been used to estimate the distribution and spread of DFTD across Tasmania. First, roadkill reports and trapping data were mapped with associated dates of first report of confirmed DFTD (details in Hawkins et al., 2006). These data are reliable, but spatially patchy and are likely to underestimate the disease distribution. Second, 132 10-km spotlighting transects have been surveyed annually for a range of wildlife species since 1985 (Hocking and Driessen, 1992; Hawkins et al., 2006). Further transects have been progressively added, so that 177 were available for analysis since the first DFTD report. After aggregating transects into 54 neighboring groups of one to six (accounting for topographical features), sightings of devils were visually assessed for an acute, sustained decline over time (usually over at least 2 years). In 15 of these groups, there was a clear sustained decline, and the year at which this decline commenced was treated as the year of emergence of disease. In a further 11 groups, there was a less dramatic decline; these cases were only used where the apparent year of emergence was compatible with that indicated by nearby confirmed or anecdotal disease reports. Groups that showed no decline were assumed disease-free and arbitrarily given an emergence year of 2007, (i.e., 1 year into the future.) Inevitably, this process is rather subjective. Further, the approach presumes that all the declines were due to disease. Potentially, however, it may produce a finer scale spatial representation of disease spread than can be derived from confirmed-disease reports alone.
To visualize our best understanding of the spread of DFTD across Tasmania, year of emergence based on the spotlighting data was used as a foundation for interpolation through splining. Confirmed reports were then used to validate and refine the disease spread surface (Fig. 2); confirmed reports were only included where they indicated earlier disease emergence than that inferred by spotlighting data. Overall, the two sources of data were consistent. In areas of the Northwest where there were no spotlighting data, trapping data (where at least 50 individuals had been caught and no sign of disease found) were also incorporated in the same way as spotlight data showing no decline.
At a coarse scale, the pattern of spread indicated by Figure 2 supports the hypothesis that this is an infectious disease and that a single origin, rather than multiple origins, is implied. It appears that DFTD has spread at a variable rate. Assuming DFTD emerged shortly before 1996, it has spread approximately 270 km South in 11 years through fairly continuous forested habitat, while during the same period it spread only 190 km West–Southwest though heterogeneous habitat. It appears to have spread down the East coast by as much 155 km in 3 years, from 1996 to 1999. Devil movements can be highly variable: Individuals have been known to travel by as much as 16–25 km in a single direction in one night (Guiler, 1978; Pemberton 1990) [M.E.J., personal observation], while individuals may also be recaptured less than a kilometer from the first trap site, even if several nights have elapsed between captures [Hawkins et al., unpublished data]. The slow rate of disease spatial spread relative to the movement rate of individual devils suggests that R0 is not very large.
In the Freycinet National Park and surrounds, the region with the best longitudinal data (several trapping trips each year from 1999–2006), a logistic model suggests an accelerating rate of spread as the disease moves Southwards (Likelihood ratio for distance–time interaction 11.78, P = 0.0001), with a mean rate of spread of about 7 km yr−1.
The furthest point from the known front, as indicated by confirmed reports, is the Northwest tip of Tasmania, at a distance of approximately 140 km. Given that the spread rate is variable, between 7 km yr−1 down the Freycinet peninsula to 52 km yr−1 down Tasmania’s East coast, DFTD could take 3–20 years to reach a statewide distribution, with disease likely to reach the Northwest within 3–10 years.
IMPACT OF DISEASE
Estimated annual survival of adults first marked as adults (i.e., first caught at age 2+) has declined dramatically at Freycinet from approximately 0.5 before disease arrival to very close to zero. Annual survival of adults first marked at age 1+ has also declined, but not as rapidly (Lachish et al., 2007). The slower decline in survival of animals first marked as 1+ individuals is probably because individuals marked as 1+ are, on average, younger than those first captured as 2+. In the early stages of any epidemic, when force of infection is relatively low, the average age at which disease is acquired is greater than later in the epidemic (Grenfell and Anderson, 1985). There is some empirical evidence [S.L., unpublished; C.H., unpublished] that the disease first occurs in older animals. The very low survival of adults means that females are likely to survive for, at most, one breeding season, with severe consequences for population persistence. Most females do not commence breeding until 2+, although there appears to be increased breeding of 1+ females in diseased populations [M.E.J., S.L., C.H., unpublished data]. DFTD is also associated with a marked change in age structure: Representation of older age classes continues to decrease with time since emergence, even at Mt. William, with no animals older than 3 years being trapped from many affected areas [Jones et al., in preparation; Hawkins et al., in preparation].
WILL DISEASE LEAD TO EXTINCTION?
Two forms of analysis of the Freycinet data project possible local extinction. First, a reverse-time CMR method (Nichols and Hines, 2002; Pradel, 1996) was used to estimate the finite rate of change λ of the adult population. Prior to disease arrival, the population appears to have been stable (confidence intervals for λ over the first 2 years include 1.0), with an immediate decline in growth rate following the onset of disease. The final estimate for λ suggests that the adult segment of the population declined immediately following disease arrival and is now approximately halving annually (Lachish et al., 2007). Given that the adult (2+) population size in the Freycinet study area is estimated to be less than 10 individuals and that the total population is estimated to be less than 50 (Lachish et al., 2007), this estimate of λ projects extinction within a few years.
Second, a preliminary non-spatial age- and sex-structured model was constructed, with discrete yearly age classes for 0–6-year-old animals, and simulated in discrete monthly time steps. Within each age/sex class, animals were split into three discrete disease-related subclasses—Susceptible, Exposed, and Infectious.
The model was initially constructed in the absence of disease, with population regulation occurring via density-dependent effects on juvenile survival and the proportion of females breeding in each age class (both increasing exponentially with decreasing population size). Maximum juvenile survival was set equal to that of 2–4-year-old adults, with density dependence scaled to give a carrying capacity of 1000 animals. Density dependence on the proportion of females breeding was scaled such that the proportions of 1, 2, 3, and 4–6-year-old animals breeding at carrying capacity matched observed proportions in the field [Hawkins et al., unpublished data]. All other parameters were estimated from available data.
DFTD was modeled as a contact-transmitted disease, with transmission based on probabilities of each individual biting each other individual in the population per month. Three potential modes of transmission were identified from bite-wound surveillance data (Hamede et al., 2007)—background transmission, male/male pre-mating transmission, and male/female mating transmission. Background transmission was assumed to occur all year round, with the per-capita rate being the same for both sexes and all ages. The additional per-capita probability of both male/male transmission during the pre-mating period, and male/female transmission during the mating period, was estimated from the bite wound data as twice the background probability. The only impact of disease modeled was disease-induced mortality of infectious individuals.
There is insufficient knowledge of how transmission scales with devil density. Three different transmission functions were trialed: per-capita probability of disease transmission remaining constant, and transmission increasing with decreasing density in either a linear or exponential fashion. Such increases can potentially occur when interactions among individuals are maintained at low density due to behavioral and social reasons, such as co-feeding, antagonistic behavior, and mating. For each transmission function, parameters giving the best simulated fit to the data were obtained using numerical optimization of the simulation model.
Irrespective of the predicted outcome of the disease-induced decline, the simulation approach utilized here indicates an ongoing rapid decline of the Freycinet population, in agreement with that inferred from the mark-recapture analysis. Determining whether local extinction is likely to be the ultimate outcome requires more field data. In particular, field confirmation that the exponential transmission model is valid is required (other transmission models are less likely to lead to predictions of extinction), as is more detail on the latent period of infection in the wild.
Conventional epidemiological theory suggests that a pathogen with only a single host is unlikely to drive that host to extinction (de Castro and Bolker, 2005; McCallum and Dobson, 1995). Once the disease has sufficiently reduced host population density, the disease will be unable to transmit effectively and should become extinct before the host. Indeed, in almost all of the cases in the literature in which a disease has likely been responsible for extinction or near extinction, one or more reservoir species, within which the pathogen has a limited effect on survival, exist. Reservoirs result in a strong force of infection for the endangered species occurring even as its numbers decline towards extinction (de Castro et al., 2005; Gog et al., 2002; McCallum and Dobson, 2002). However, this conventional wisdom relies on transmission occurring in a density-dependent rather than a frequency-dependent fashion. If transmission depends on the frequency of infected hosts in the population, rather than their density, there is no threshold population size for disease maintenance and pathogen-induced extinction is a possibility even in the absence of reservoir hosts (de Castro et al., 2005).
If transmission of DFTD occurs primarily from biting and such biting occurs mainly during sexual encounters, then DFTD will possess many of the characteristics of a sexually transmitted disease, one of which is the likelihood of frequency-dependent transmission. Unfortunately, determining the functional form of the transmission function is extremely difficult from empirical data (McCallum et al., 2001). However, there are indications from the monitoring reported in this article, and from the contact rate data reported elsewhere in this issue, that transmission may depend weakly, if at all, on devil population density. For example, DFTD has been confirmed in Strathgordon in the low density Southwest region, and continues to be present in Mt. William even after a 90% population decline. If this is the case, then the extrapolation of current rates of decline to produce a prognosis of extinction on the Freycinet Peninsula in less than 10 years may be valid. If trends in DFTD spread and population decline continue, there is a strong possibility of extinction of the Tasmanian devil due to the disease over a timeframe of perhaps 25–35 years.
The simulation model suggests that whether the disease will lead to extinction is strongly dependent on the length of the latent period. Indeed, such time delays in disease dynamics do commonly act to destabilize host populations (May and Anderson, 1978). Hence, estimation of this critical parameter is clearly a major priority. However, enclosure or laboratory based transmission experiments do not adequately represent the modes of transmission that may occur in the field. At present, our knowledge of the latent period relies on anecdotal observations of the time until disease emergence in a few animals that have been brought into captivity. A systematic program to bring animals from diseased areas into captivity and then to record the time until disease emergence would at least provide lower bounds on the latent period. Given information on the force of infection in the areas from which the animals were captured, it should be possible to obtain an estimate of the latent period. Some methods that have been developed to estimate incubation or latent periods in human diseases such as SARS (Kuk and Ma, 2005) or AIDS (Ronald, 2001) may be applicable.
A critical question is the value of a passive monitoring program, such as we have been applying to date. The trapping program is expensive, consuming approximately $300,000 annually. Passive monitoring has been described as:
“an inefficient use of scarce conservation funding …a form of political and intellectual displacement behavior …or worse, a deliberate delaying tactic”(Nichols and Williams, 2006).
Spotlighting surveys, carried out with the specific aim of monitoring the impact of management practices on native browsing mammal species, in which other mammal species (such as devils) were incidentally monitored, were crucial in demonstrating a significant DFTD-associated decline in the overall devil population (Hawkins et al., 2006). The resource-intensive trapping program has provided at least five key benefits. First, the monitoring was essential to determining the extent of the problem. In the absence of this systematic program, it would not have been evident that the disease was invariably fatal in the field, nor that it was spreading spatially and behaved within host populations as an infectious disease. Second, the program has shown that the disease persists within affected populations and does not disappear once populations have been sufficiently reduced. These two conclusions are essential in justifying the need for aggressive (and expensive) management interventions. Third, passive monitoring of the progression of epidemics within unmanipulated populations should enable us to estimate epidemiological parameters that can be used to build models to assist in the management of the disease. In particular, estimates of the basic reproductive rate R0 are required to determine whether or not removal of infected animals is likely to be a successful control strategy. Fourth, control of any invasive organism is best focused at the edges of the invasion front. Monitoring is necessary to identify where the front might be. Finally, information on the progression of the disease within unmanipulated populations can be used to compare with disease progression in manipulated populations (this issue) to assess whether the disease control program is effective.
For the future, our data on distribution, rate of spread, and rate of overall population decline continue to be critical to maintaining support and to management planning, especially in decisions on the timing and location of collection of individuals for insurance populations to be isolated from the disease. At our current state of knowledge of DFTD, these estimates may change significantly and so ongoing monitoring is required. At this stage, we therefore view ongoing research into DFTD outside management trials as an essential, additional tool in the rapid identification of the best disease management strategy.
H.M., M.J., and S.L. are partially funded under an Australian Research Council Linkage Grant. We are grateful to a large number of volunteers who have made this work possible. All fieldwork was carried out under animal ethics approvals to the University of Tasmania, The University of Queensland, and the Tasmanian Department of Primary Industries and Water.