SEARCH

SEARCH BY CITATION

Keywords:

  • birth and death processes;
  • feline enteric coronavirus;
  • infectious disease;
  • spatial spread of disease

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

1. Deterministic models predict that susceptible-infective-susceptible (SIS) disease, where there is no immunity to reinfection following recovery, will become infinitely persistent in a host population. We explored the incorporation of stochasticity into SIS models; modelled interacting host-disease agents in metapopulations; and examined model predictions in a real system involving viral infection in domestic cats.

2. SIS models incorporating stochasticity predicted that disease persistence would be finite and dependent on the host population size, provided the host population was isolated. However, the disease may persist by dynamic spread among interacting host metapopulations.

3. Feline enteric coronavirus (FECV) dynamics in domestic cats were well predicted by stochastic metapopulation models.

4. The models we present are mathematically tractable, generalizable, and mechanistically realistic. The findings from the cat-virus system are immediately applicable to the management of cattery populations and could be adapted to inform eradication programmes for other infectious diseases in animal and human populations. The most practical methods to eradicate feline enteric coranovirus would be to remove small catteries (islands) from interactions with large catteries (mainlands) and to convert mainlands to islands by depopulation.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Host individuals are ‘islands’ of transient disease, among which islands disease agents move to survive. Disease outbreaks may terminate, in small patches of hosts, when there are insufficient susceptible hosts to maintain an epidemic. Deterministic models of epidemic disease describe this local extinction of disease using the SIRS (for Susceptible [RIGHTWARDS ARROW] Infective [RIGHTWARDS ARROW] Recovered [RIGHTWARDS ARROW] Susceptible) approach (Kermack & McKendrick 1927; May & Anderson 1984). Numerous authors have also incorporated stochasticity into epidemic models (Bartlett 1956; Bharucha-Reid 1956; Bailey 1975), although most such models are impossible or difficult to solve. Incorporation of stochasticity is critical for small host populations or when dynamics are of the SIS (Susceptible [RIGHTWARDS ARROW] Infective [RIGHTWARDS ARROW] Susceptible) form, because deterministic SIS models always provide for endemic disease above some exact threshold population size (Edelstein-Keshet 1988). In reality, thresholds for endemicity of SIS disease may vary stochastically. When stochastic models are employed to describe SIS disease in small host populations, predicted epidemics are always finite in duration. Given stochastic dynamics, persistence of disease agents depends on either an inert reservoir in the environment or dynamic spread of the agent among small local populations, i.e. a metapopulation in the broad sense. Gyllenberg, Hanski & Hastings (1997) appreciated that the deterministic dynamics of SIRS and metapopulation models are analogous, and Ferguson, May & Anderson (1997) described a simulation of measles persistence in a grid-like metapopulation. The three criteria for diseases well suited to metapopulation analysis are: (i) existing in nature in discrete host populations; (ii) spreading by colonization of these host populations; and (iii) sometimes experiencing extinction within local populations.

In this paper we model the persistence of an SIS, a viral infectious disease of felids (feline enteric coronavirus; FECV), that fulfils the three criteria given. The virus is endemic in catteries (which are small, discrete local populations of hosts; Foley et al. 1997), is highly infectious, and can become locally extinct if cat groups are small, given stochastic dynamics. Local FECV dynamics are modelled using a birth and death process model coupled with a deterministic metapopulation model. The model is fitted with parameters from empirical data. There are two important applications of this work: the appreciation that the expected extinction time of a disease can be used directly in the global metapopulation model, and as an example of how such models account for the persistence and spread of infectious disease in nature. Lastly, we give recommendations for the management of FECV based on the model and show how this system is a template for the development of eradication strategies for similar diseases in discrete, interacting host populations.

Feline enteric coronavirus natural history and epidemiology

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Domestic cats occur world-wide in feral cat colonies, as indoor–outdoor pets and in indoor-only catteries, either for humane purposes, pets or commercial breeding. They are susceptible to a relatively benign enteric coronavirus (FECV) that thrives in intestinal epithelium and produces mild gastro-enteritis (Pedersen et al. 1981). Feline coronaviruses are contracted primarily during exposure to infectious cat faeces in the environment, but also via ingestion or inhalation during cat-to-cat contact. FECV is generally fragile once outside the cat's body, but may survive for as long as 7 weeks if protected from heat, light and desiccation. The transmission of the FECV is frequent and difficult to detect clinically. Prevalence of FECV infections is related to cattery size and density: in multiple-cat homes with five or more cats, approximately 100% of the cats have been exposed (Foley et al. 1997). Once exposed (often as kittens), cats may periodically shed FECV in faeces for weeks to a few months and after recovery are not immune. Thus they are likely to become rapidly reinfected. Active FECV infection is diagnosed by detection of FECV RNA in faeces by reverse-transcriptase polymerase chain reaction (RT-PCR) (Poland et al. 1996; Foley et al. 1997). Cats develop positive antibody titres in serum about 7 days after exposure to FECV, but the titres remain elevated for months to years, even when the cat is recovered and not shedding virus.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Metapopulation model

The strict-sense Levins metapopulation comprises a set of identical, equally accessible interacting patches (Levins 1969), with dynamics given by:

  • image(eqn 1)

in which p(t) is the fraction of patches occupied at time t by (in this case) FECV, m is the migration rate among patches, and e is the per-patch extinction rate. Equivalently, p(t) is the probability that a particular host patch is infected by the disease. If the metapopulation has extremely variable patch sizes, then mainland–island models are more appropriate, in that some patches become permanent refuges (or foci) for disease (MacArthur & Wilson 1967; Hanski 1991). The dynamics are then:

  • image(eqn 2)

For FECV, endemically infected catteries are mainlands and transiently infected catteries are islands. Intermediate patch size variability is included in the model by using a structured metapopulation in which p(t) depends on the patch size, N (Gyllenberg, Hanski & Hastings 1997). The parameters that govern the dynamics of p(N,t) for all models are the extinction rate, e(N), and the colonization rate, m(N), i.e. the rate of disease introduction into susceptible local populations. The predicted Levins metapopulation equilibrium, p*, is:

  • image(eqn 3)

while the island–mainland p* is:

  • image(eqn 4)

If e exceeds m, then a strict-sense Levins metapopulation would go extinct across all patches, i.e. FECV would disappear within that metapopulation of catteries. In contrast, a mainland of endemic disease would by definition remain infected and would allow for positive occupancy probabilities on all islands accessible to the mainland.

Local population model

Local population extinction rates of FECV are obtained as the solution of a continuous time birth–death process Markov chain (Feller 1971; Karlin & Taylor 1975; Nisbet & Gurney 1982) with discrete individual accounting, demographic stochasticity (Mollison 1981) and continuous time. The number of infectives, I, can be a value from 0 to N (total population size). Then B(I)dt is the probability of an infection in some small time period dt. D(I)dt is the probability of recovering from disease in dt. If there are I(t) infectives at time t, then the probability that I will increase by 1 in time dt is (dropping terms in dt2):

  • image(eqn 5)

Here β is the rate of transmission, i.e. the number of new infections produced per day after the introduction of one infective cat into a susceptible population. The probability that I will decrease by 1 is:

  • image(eqn 6)

where γ is the rate of recovery, i.e. the number of days after infection before a cat is no longer shedding infectious virus. The probability that I will stay the same is:

  • image(eqn 7)

The expected value (or mean value) of I for this time period is:

  • image(eqn 8)

The time to extinction of FECV using this birth–death process and starting at one infective individual satisfies:

  • image(eqn 9)

where:

  • image(eqn 10)
  • image(eqn 11)

The derivation may be found in Nisbet & Gurney (1982), Karlin & Taylor (1975) or Gardiner (1985). For SIS disease dynamics, we insert values for births and deaths obtained from equations 5 and 6 to obtain the persistence time of FECV in a local host population starting with I(0) = 1,

  • image(eqn 12)

The extinction rate of FECV is 1/Te. Parameters needed to apply the metapopulation model are β, γ, N (or e), p and m.

Local population parameter estimation

Transmission rates (β) were estimated from exposures to FECV in previously uninfected cats. Groups of specific pathogen-free (SPF) domestic short-haired cats and Persian cats were maintained in viral containment facilities at the UC Davis Veterinary Retrovirology Laboratory in ratios of 1 susceptible cat: 1 infective cat (Pedersen et al. 1981) or four susceptible cats: 1 infective cat, and in Havana Brown and domestic short-hair cats in private homes in ratios of 4 : 1 and 2 : 1. After exposure to FECV, cats were determined to be infective by detection of FECV RNA within faeces by RT-PCR (Foley et al. 1997), or anti-coronavirus IgG (Pedersen 1995) in serum if the cat had been antibody-negative before the experimental manipulation. Previous experiments indicated an approximate 4-day period of latency from initial exposure to PCR-positive faeces (when cats are infective) and cats do not seroconvert until 7 days after exposure (Foley et al. 1997). Therefore the time from exposure to infectivity was adjusted by subtracting 3 days if the infection was detected by seroconversion. Latency was not incorporated into the present models because an added state (exposed, not infective) would increase model complexity significantly and the magnitude of the latent period of 4 days is small compared with the typical duration of the susceptible state and infective states (both months). Moreover, three cats were observed in the present study to become faecal PCR-positive by day 2, suggesting a very brief period of latency under some conditions. The maximum likelihood estimator (MLE) of βI is the reciprocal of the mean date till first infection, assuming an exponential distribution (Johnson, Kotz & Balakrishnan 1994). With this MLE and I = 1 (for these experiments), the most robust estimate of latency is the minimum value of the time to infection (2 days).

The recovery rate, γ, was determined by maintaining seropositive, faecal PCR-positive cats in isolation until faecal shedding ceased for 3 consecutive weeks. Then γ = 1/(last day faecal positive). The 95% confidence intervals for β and γ were obtained as described elsewhere (Johnson, Kotz & Balakrishnan 1994). The duration of immunity was evaluated by reintroducing 13 recovered cats into an endemically infected home in which 100% of the 40 resident cats were seropositive and 75–80% were shedding FECV in faeces.

Metapopulation parameter estimation

An empirical estimate of p was obtained by determining how many catteries of size N = 1, 2, … 6+ had at least one seropositive cat at time t. The extinction rate was estimated as the reciprocal of epidemic duration (Te), obtained from the birth–death process local model. Alternatively, e(N) could have been observed empirically, which would have required a large sample size for reasonable confidence.

The colonization rate (m) was estimated by subdividing colonization into four independent events, and attributing to each event a probability that this event would result in the introduction of infection into a naive population. These events were: adoption of a new cat (ma), a cat visit (to another home, veterinary office or cat show) (mv), exposure to fomites via the owner (typically because of occupational exposure to cats) (ms) and exposure to outside cats while roaming (mo). The probability or risk coefficient (α) that a single performance of each event would introduce FECV into a naive population was estimated as the ‘best guess’ by researchers, veterinarians and cattery managers familiar with FECV in the field. For example, adopting an FECV PCR-positive cat is a colonization, with α = 1. The approximate values of α used in the present model are summarized in Table 1.

Table 1.  Risk coefficients for various events leading to colonization of a population with FECV
ActivityRisk coefficient (α)
  • *

    p(N) is the observed fraction of homes of size N with FECV infections, data shown in Table 2a, b.

Adopting new cat from shelter0·9
Adopting new cat from pedigreed breeder0·9
Adopting new cat, other homep(N) of source home*
Adopting new cat, feral0·1
Adopting new cat, stray0·4 (highly variable)
Veterinary offices and cat shows visits0·05
Visits to other householdsp(N) of the home visited
Owner's occupational exposure0·05

Components of colonization were assumed to be additive, so that the overall colonization rate was:

  • image(eqn 13)

Values for ma, mv, ms, and mo were estimated from questionnaires in which cat owners were asked: the annual frequency of travel to veterinarians, cat shows, and other homes; the number and source of any new cats in the home over the last 2 years; whether cats were strictly indoors or indoor/outdoor; whether any cats had visited the home over the last 2 years; and whether members of the household worked with cats. The best estimate for each m(N) was obtained from the linear model regressing m (from questionnaires) on N.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

Parameter estimates

A decay function of the percentage of uninfected cats (Fig. 1) indicated that the waiting time until infection was exponentially distributed (d = 0·1569, P = 0·383). Thus transmission could be modelled as a Poisson process. The mean date of first infection was 23·04 days, resulting in an estimate of β of 0·043 and 95% CI of 0·065–0·032.

image

Figure 1. Exponential decay function of the proportion of cats experimentally exposed to FECV that remained uninfected over time.

Download figure to PowerPoint

Maintenance of infected cats in isolation revealed that 8·2% of cats never ceased shedding coronavirus while under observation; these cats were classed as chronic shedder cats (SI) cats and were excluded from the calculation of recovery rate. For the 13 cats observed recovering, the mean period until recovery was 37·2 days, with γ = 0·0269 and 95% CI of 0·017–0·061.

Persistence of fecv

The expected persistence time of FECV in catteries increased rapidly with cattery size (Fig. 2). In fact, there was a threshold population size of approximately five, above which no reasonable values of β and γ allowed for extinction of FECV, i.e. no management strategies to reduce infections and promote rapid recovery would allow for elimination of the endemic disease in a year or less. Conversely, in a small cattery with no chronic shedders, FECV epidemics would be unlikely to become persistent, even in the face of delayed recovery and rapid virus transmission. Very close to the threshold, changes of β and γ contribute to making Te manageable, but the most important determinant of Te far from threshold is N (Fig. 3). If any cat in a population is a chronic shedder (SI), then Te = ∞, independent of N.

image

Figure 2. Expected time to extinction (Te) of an outbreak of FECV as a function of cattery size (N), assuming a birth and death process SIS model, with β = 0·043 and γ = 0·0269.

Download figure to PowerPoint

image

Figure 3. Predicted times to extinction (days) of a stochastic epidemic birth and death process in a cattery with three, five and 10 cats. Beta is transmission rate and gamma is recovery rate. See text for model.

Download figure to PowerPoint

Given the presence of persistently infected catteries, an island–mainland metapopulation model accounts best for global FECV persistence. The empirical patch occupancy p(N) of indoor-only catteries sharply increased from about 0·1 at N = 1–100% for N ≥ 5 (Table 2a), which agrees well with the birth–death process model expectations, and confirms that catteries with ≥ 5 cats act as mainlands. Also given in Table 2 are the calculated values of e(N) from the local birth–death process model (with β = 0·043, γ = 0·027) and calculated estimates of m(N) from questionnaires for indoor-only catteries. Reported values of m(N) were obtained from the linear model regressing observed m on N. The empirical estimate of p(N) was compared with the predicted value of p*(N), assuming a Levins model and the values of m(N) and e(N) given in Table 2a. The data were also shown as numbers of observed and expected infected catteries, to simplify comparisons. In all five cases, the fit was remarkably good given the small sample size: the rounded-off estimates of expected infected catteries were equal to the observed number of infected catteries.

Table 2.  (a) Metapopulation parameter and variable values for indoor-only catteries with birth and death process dynamics and γ = 0·0269, β = 0·043; (b) metapopulation values for indoor–outdoor catteries
(a) Indoor-only catteries    
Cats per cattery12345
Number of catteries122664
e(N) (year–1)9·8125·4562·2810·6710·140
m(N) (year–1)0·6061·2121·8182·4243·03
Expected number of infected0·690·362·664·693·82
Observed number of infected10344
Expected p*(N)0·0580·1820·4440·7830·956
p(N) – observation0·080·00·50·671·0
(b) Indoor–outdoor catteries    
Cats per cattery12345
Number of catteries62423
e(N) (year–1)9·8125·4562·2810·6710·140
m(N) – quest (year–1)001·192·751·2
m(N) – model (year–1)0·0190·1671·0303·584
Expected number of infected001·371·612·69
Observed number of infected11303
Expected p*(N)000·3430·8040·895
p(N) – observation0·160·50·7501

Indoor–outdoor catteries had generally higher values of p(N) than indoor-only catteries, reflecting the increased exposure of indoor–outdoor cats to infective cats and fomites (Table 2b). The observed number of infected indoor–outdoor catteries was not as close to expected as for indoor-only catteries, probably because owners could not estimate accurately on questionnaires how often cats came into contact with infectious cats or fomites while outdoors. We compared the observed values of m(N) from questionnaires with those predicted from the Levins model equilibrium with known e and p. The results indicated that the estimate of m underestimated the predicted m, i.e. there was significant colonization not accounted for in the questionnaire. The critical value for FECV metapopulation persistence, i.e. m(N) > e(N), occurs in both indoor-only and indoor-outdoor catteries at N = 4. Therefore, the threshold cattery size to be a mainland (N = 5) was close to the critical N for persistence of a Levins metapopulation, indicating that the FECV system most closely resembles an island–mainland version of a metapopulation.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

As we have demonstrated with the FECV model, metapopulation dynamics contribute to the ongoing incidence and persistence of patchily distributed infectious diseases. In small populations, demographic stochasticity can produce local disease extinction, so that global disease persistence depends on the existence of mainland populations and frequent migration among small populations. In an unstructured metapopulation of small populations, metapopulation persistence occurs only when the colonization rate exceeds the local extinction rate. Otherwise the disease agent goes extinct globally. In an island–mainland system such as FECV in catteries, extinction on small islands occurs only if all migration ceases. In the FECV–cattery system, homes with ≥ 5 cats function as mainlands.

The output from the FECV metapopulation model may be used to develop management strategies to control and eradicate FECV in nature. From the perspective of a single small cattery, the most effective methods of controlling FECV would be to remove the cattery from the metapopulation (by eliminating all sources of immigration of FECV) and to convert the cattery from a mainland to an island, by reducing the population to four or fewer cats. We did not incorporate latency into the parameter estimates, but doing so would not significantly change expected dynamics. Rather, transmission would be slightly more efficient, which would slightly increase the expected persistence time. However, expected persistence times are already extremely high except below a stringent threshold in population size; the threshold value is not sensitive to the addition of minor latency and minor changes of beta (β) and gamma (γ). There is no sufficient increase in gamma (γ) that could allow for FECV extinction in larger catteries. In contrast, major reductions in transmission rate would be expected to reduce FECV persistence in catteries significantly over the present threshold size; such a change in the number of new transmissions [beta (β)] would probably be accomplished in the near future by a vaccine. As such technologies are developed, the expected change in dynamics could be explored using this model.

The important assumptions in the metapopulation model were that all patches are equally accessible and the assumptions implicit in the estimation of migration rate. It is probably not true for real interacting host populations that all patches are equally accessible. However, simulations show that metapopulation dynamics (at least near equilibrium) are not very sensitive to deviations from the equal accessibility assumption (Hanski, Foley & Hassell 1996). Moreover, catteries are interconnected in a more equally accessible way than many animal local populations in which geographical proximity sets limits to recolonization events.

The assumptions implicit in the calculation of m included: (i) that all sources of colonization were included in the subdivision of m; (ii) that the weights attributed to each event were reasonable; and (iii) that the component events were approximately independent of each other and interaction terms could be disregarded. There was evidence that owners of indoor–outdoor cats significantly underestimated m in questionnaires. A simpler and more robust method of estimating m would have been to observe naive populations of various sizes, and quantify the number of disease introductions in time. Such an approach requires a very large spatial and temporal scale of study because colonization events for any system are rare or at least unevenly distributed in space and time (Ims & Yoccoz 1997). This approach has rarely been performed for any studies of metapopulations in nature. A further difficulty in estimating m comes from variability in rates of colonization across populations and time. Conceptually, the effect of variable m is to reduce the equilibrium level of disease (Levins 1969). It is likely that variability in m is significant in real cat populations relative to the scale of movement of FECV. Further research estimating m and its variability is warranted.

Metapopulation models may be extended to other disease systems with high turnover, such as the common cold, gonorrhoea and phocine distemper. Humans on the island of Tristan da Cunha (Shibli et al. 1971) and in the small arctic community of Spitsbergen (Paul & Freese 1933) were regularly and discretely colonized by the common cold (an SIRS disease) when visited by ships. A structured metapopulation analysis is appropriate for the analysis of the common cold in discrete human populations. Globally endemic gonorrhoea may persist due to the presence of at least one persistently infected subgroup, i.e. a mainland (Lajmanovich & Yorke 1976). Hethcote (1976) generalized this result, showing that global disease persistence could occur among ‘subpopulations’ each with R0 < < 1 (where R0 ≥ 1 refers to the minimum initial population size that will sustain an epidemic) provided there was sufficient migration among groups. Phocine distemper, also an SIRS disease, occurred in discrete harbour seal populations in the North Sea that varied in size from 30 to 3500 seals (Dietz, Heide-Jorgensen & Harkonen 1989). Local epidemics of phocine distemper varied in duration from 42 to 115 days over 17 populations with no significant relationship between population size and epidemic duration (suggesting the possible contribution of demographic stochasticity).

In contrast to disease metapopulations, strict-sense Levins metapopulations may be uncommon in plant and animal systems (Harrison 1991; Harrison & Taylor 1997). Butterflies (Thomas & Hanski 1997) and pool frogs (Sjogren Gulve 1994) furnish the few well-studied examples we have. Most discretely distributed species have considerable variation in local population size, and are thus structured metapopulations (Harrison & Quinn 1989; Harrison 1991; Gyllenberg, Hanski & Hastings 1997). Such structured metapopulations are harder to analyse than the classical Levins metapopulations, but simplifications are possible. As shown in this paper, it is possible to identify local mainland populations that remain persistent. The smaller populations wink on and off at a much more rapid rate. In fact, disease–host systems make better examples of classical metapopulations than some previously studied large organisms.

Even in the case where host distributions are not discrete, metapopulation analysis may be useful. The alternative modelling approaches, including non-spatial models such as the original Kermack–McKendrick models, cannot capture the spatial patchwork patterns of disease persistence. Percolation models (essentially stochastic cellular automata) do allow for space and chance (Durrett & Levin 1994), but this ingenious and insightful theory is hard to apply to real population dynamics. Continuously distributed hosts may be modelled with reaction–diffusion (using partial differential equations) to predict the wave of disease advance (Mollison & Kuulasmaa 1985; Murray 1989; van den Bosch, Metz & Diekmann 1990; Mollison 1991). Reaction–diffusion models show disease flowing over a host landscape and then disappearing (in theory). In practice such disease may persist due to low densities of infectious subpopulations of hosts. The purely continuous models do not account for these persistent ‘slow fuses’ in a host landscape through which disease has already passed.

Many causes of patchy distributions of susceptible hosts may occur in nature, such as variability in vaccine usage. Widespread vaccination for measles created pronounced alterations in local and global disease dynamics in cities in England (Grenfell, Bolker & Kleckowski 1995). Metapopulation approaches using transiently patchy susceptible and infectious local populations offer the best hope of predicting disease persistence.

The metapopulation approach is an analytically tractable method to study disease dynamics in discretely distributed hosts, and should provide insight into continuously distributed hosts where susceptible, infectious and resistant patches arise during epidemics. Whenever local disease extinction occurs, we must understand global infection dynamics to account for patterns of disease persistence and emergence. This understanding will be critical in any attempts to manage or eradicate disease.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References

The authors wish to acknowledge the invaluable technical assistance of Amy Poland, Cindi Ramirez, Jeff Carlson, Kim Floyd-Hawkins, Jennifer Norman and Rosemary Panduro. Alan Hastings, Ed Caswell-Chen, and Walter Boyce contributed to the conceptual development. This research was supported by grants to Janet Foley from the Winn Feline Foundation, Morris Animal Foundation, Solvay Animal Health, the San Francisco Foundation, and the Center for Companion Animal Health, School of Veterinary Medicine, University of California, Davis.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Feline enteric coronavirus natural history and epidemiology
  5. Materials and methods
  6. Results
  7. Discussion
  8. Acknowledgements
  9. References
  • Bailey, N. (1975) The Mathematical Theory of Infectious Diseases and its Applications. Griffin, London, UK.
  • Bartlett, M. (1956) Deterministic and stochastic models for recurrent epidemics. Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Vol. IV (ed. J.Neyman), pp. 81109. University of California Press, Berkeley, CA.
  • Bharucha-Reid, A. (1956) On the stochastic theory of epidemics. Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, Vol. IV (ed. J.Neyman), pp. 111119. University of California Press, Berkeley, CA.
  • Van Den Bosch, F., Metz, A. & Diekmann, O. (1990) The velocity of spatial population expansion. Journal of Mathematical Biology, 28, 529565.
  • Dietz, R., Heide-Jorgensen, M.-P. & Harkonen, T. (1989) Mass deaths of harbor seals (Phoca vitulina) in Europe. Ambio, 18, 258264.
  • Durrett, R. & Levin, S. (1994) Stochastic spatial models: a user's guide to ecological applications. Philosophical Transactions of the Royal Society of London, Series B, 343, 329350.
  • Edelstein-Keshet, L. (1988) Mathematical Models in Biology. Random-House, New York, NY.
  • Feller, W. (1971) An Introduction to Probability Theory and its Applications, Vol. 2. Wiley, New York, NY.
  • Ferguson, N., May, R. & Anderson, R. (1997) Measles: persistence and synchronicity in disease dynamics. Spatial Ecology: the Role of Space in Population Dynamics and Interspecific Interactions (eds D.Tilman & P.Kareiva), pp. 137157. Princeton University Press, Princeton, NJ.
  • Foley, J.E., Poland, A., Carlson, J. & Pedersen, N.C. (1997) Patterns of feline coronavirus infection and fecal shedding from cats in multiple-cat environments. Journal of the American Veterinary Medical Association, 210, 13071312.
  • Gardiner, C. (1985) Handbook of Stochastic Methods. Springer, New York, NY.
  • Grenfell, B., Bolker, B. & Kleckowski, A. (1995) Seasonality, demography and the development dynamics of measles in developed countries. Epidemic Models: Their Structure and Relation to Data (ed. D.Mollison), pp. 248268. Cambridge University Press, Cambridge, UK.
  • Gyllenberg, M., Hanski, I. & Hastings, A. (1997) Structured metapopulation models. Metapopulation Biology (eds I.Hanski & M.Gilpin), pp. 93122. Academic Press, San Diego, CA.
  • Hanski, I. (1991) Single-species metapopulation dynamics: concepts, models and observations. Bulletin of Journal of Linnaean Society, 42, 1738.
  • Hanski, I., Foley, P. & Hassell, M. (1996) Random walks in a metapopulation: how much density dependence is necessary for long-term persistence? Journal of Animal Ecology, 65, 274282.
  • Harrison, S. (1991) Local extinction in a metapopulation context: an empirical evaluation. Metapopulation Dynamics: Empirical and Theoretical Investigations (eds M.Gilpin & I.Hanski), pp. 7388. Academic Press, London, UK.
  • Harrison, S. & Quinn, J. (1989) Correlated environments and the persistence of metapopulations. Oikos, 56, 293298.
  • Harrison, S. & Taylor, A. (1997) Empirical evidence for metapopulation dynamics. Metapopulation Dynamics: Empirical and Theoretical Investigations (eds M.Gilpin & I.Hanski), pp. 2742. Academic Press, London, UK.
  • Hethcote, H. (1976) Qualitative analyses of communicable disease models. Mathematical Biosciences, 28, 335356.
  • Ims, R. & Yoccoz, N. (1997) Studying transfer processes in metapopulations. Emigration, migration, and colonization. Metapopulation Biology. Ecology, Genetics, and Evolution (eds I.Hanski & M.Gilpin), pp. 247265. Academic Press, San Diego, CA.
  • Johnson, N.L., Kotz, S. & Balakrishnan, N. (1994) Continuous Univariate Distributions, Vol. 1. Wiley, New York, NY.
  • Karlin, S. & Taylor, H. (1975) A First Course in Stochastic Processes. Academic Press, London, UK.
  • Kermack, W. & McKendrick, A. (1927) Contributions to the mathematical theory of epidemics. Royal Statistical Society Journal, 115, 700721.
  • Lajmanovich, A. & Yorke, J. (1976) A deterministic model for gonorrhea in a nonhomogeneous population. Mathematical Biosciences, 28, 221236.
  • Levins, R. (1969) Some demographic and genetic consequences of environmental heterogeneity for biological control. Bulletin of Entomological Society of America, 15, 237240.
  • MacArthur, R. & Wilson, E. (1967) The Theory of Island Biogeography. Princeton University Press, Princeton, NJ.
  • May, R. & Anderson, R. (1984) Spatial heterogeneity and the design of immunization programs. Mathematical Biosciences, 72, 83111.
  • Mollison, D. (1981) The importance of demographic stochasticity in population dynamics. The Mathematical Theory of the Dynamics of Biological Populations II (eds R.Hiorns & D.Cooke), pp. 99107. Academic Press, London, UK.
  • Mollison, D. (1991) Dependence of epidemic and population velocities on basic parameters. Mathematical Biosciences, 107, 255287.
  • Mollison, D. & Kuulasmaa, K. (1985) Spatial epidemic models: theory and simulations. Population Dynamics of Rabies in Wildlife (ed. P.Bacon), pp. 291309. Academic Press, London, UK.
  • Murray, J. (1989) Mathematical Biology. Springer, New York, NY.
  • Nisbet, R. & Gurney, W. (1982) Modelling Fluctuating Populations. Wiley, New York, NY.
  • Paul, J. & Freese, H. (1933) An epidemiological study of the ‘common cold’ in an isolated arctic community (Spitzbergen). American Journal of Hygiene, 17, 517.
  • Pedersen, N. (1995) The history and interpretation of feline coronavirus serology. Feline Practice, 23, 4652.
  • Pedersen, N.C., Boyle, J.F., Floyd, K., Fudge, A. & Barker, A. (1981) An enteric coronavirus infection of cats and its relationship to feline infectious peritonitis. American Journal of Veterinary Research, 42, 368477.
  • Poland, A., Vennema, H., Foley, J.E. & Pedersen, N.C. (1996) Feline infectious peritonitis is caused by simple mutants of feline enteric coronavirus (FECV) that arise frequently during the course of primary FECV infection. Journal of Clinical Microbiology, 34, 31803184.
  • Shibli, M., Gooch, S., Lewis, H. & Tyrrell, D. (1971) Common colds on Tristan da Cunha. Journal of Hygiene, Cambridge, 69, 255262.
  • Sjogren Gulve, P. (1994) Distribution and extinction patterns within a northern metapopulation case of the pool frog, Rana lessonae. Ecology, 75, 13571367.
  • Thomas, C. & Hanski, I. (1997) Butterfly metapopulation. Metapopulation Biology (eds I.Hanski & M.Gilpin), pp. 359386. Associated Press, San Diego, CA.

Received 6 February 1998; revision received 30 April 1999