I.J. EAST1,
M.G. GARNER1
1 Office of the Chief Veterinary Officer, Department of Agriculture, Fisheries and Forestry, Australia
2 Faculty of Veterinary Science, University of Sydney, Australia
Emergency vaccination can be useful in eradication programs for highly pathogenic avian influenza (HPAI). However there is a concern that HPAI could circulate inapparently within vaccinated flocks, because vaccination protects against clinical disease but does not induce sterile immunity. Hence animal health organisations advise that vaccinated flocks should be actively monitored for HPAI through serological surveillance or the placement of unvaccinated sentinel chickens within flocks.
This paper describes the development of a stochastic state transition model to examine the effect of vaccination efficacy upon the ability of sentinel surveillance schemes to detect HPAI outbreaks in caged chicken flocks. Preliminary results show that even though the time to the detection of an outbreak increased for more effective vaccination strategies, the number of birds that had been infected by the time the disease had been diagnosed decreased because of slower transmission of infection. This would presumably lead to a reduction of the amount of HPAI contamination in the environment, reducing the risk of the infection spreading to other farms.
Introduction
There is controversy regarding the use of vaccination to control the spread of H5N1 highly pathogenic avian influenza (HPAI) in chickens. This is because even though vaccinated chickens have reduced susceptibility to HPAI, are less infectious to other birds, and are protected from developing clinical disease, they can still become infected and shed virus (Swayne, 2003). Hence there are concerns that infection may spread inapparently within vaccinated flocks. Additionally, levels of flock immunity after field vaccination may also vary. In a trial conducted in Hong Kong, 24.2% of 248 flocks had less than 70% of chickens with an adequate haemagglutination inhibition (HI) titre (≥ 16) after vaccination (Ellis et al., 2006). However, there is evidence that chickens with low HI titres (10-40) may have an altered clinical course of disease compared with unvaccinated birds (Kumar et al., in press).
Because of this uncertainty, animal health organisations recommend the implementation of active surveillance schemes to monitor vaccinated flocks for evidence of HPAI (FAO and OIE, 2005). One scheme is to place a minimum of 100 unvaccinated sentinel chickens in vaccinated flocks so that they can be monitored for clinical signs (i.e. mortality) and/or laboratory evidence of infection with HPAI (European Commission, 2006).
Experimental studies suggest that HPAI is spread in cage production systems through direct contact between infectious and susceptible birds in adjacent cages (Shortridge et al., 1998). This is supported by field studies which have noted that the disease may affect birds in localised areas of a shed before spreading to birds in adjacent cages (Beard, 1998).
In this paper we present preliminary results of a simulation model designed to investigate the potential impacts of varying levels of vaccination efficacy on the dynamics of HPAI transmission in a flock of caged chickens.
Methods
A stochastic state transition model was developed to simulate the spread of HPAI between individual chickens in daily time steps.
Chickens are represented individually in the model and exist in one of five disease states: Susceptible; Latent; Infectious; Immune or Dead. Three different vaccination states are also included in the model: Not Vaccinated (chickens that have not been vaccinated in a flock); Fully Vaccinated (chickens that develop adequate HI titres after vaccination); and Partially Vaccinated (chickens that develop inadequate HI titres after vaccination).
The model is designed so that disease may be spread between birds in the same cage; to birds within adjacent cages on the same tier; to birds in the cage below; and to birds in distant cages through contact with contaminated furniture. Individual chickens transition between disease states after certain trigger events occur, such as direct or indirect contact with an infectious chicken or at the end of the latent or infectious periods (Figure 1).
Parameters affecting these transitions are estimated for each chicken from pre-determined probability distributions. Case fatality ratios, levels of susceptibility and distributions of latent and infectious periods vary according to the vaccination status of each chicken. Transmission parameters are weighted according to the vaccination status of the Infectious and Susceptible birds. Parameters used in this model have been estimated from published challenge or transmission studies using H5N1 HPAI virus or from expert opinion (see Appendix).
Simulations were run using a flock of 9600 chickens kept in 2400 cages that were arranged spatially to emulate the typical structure of a chicken meat farm in Hong Kong. At the start of each iteration, 100 sentinel chickens were placed in randomly selected cages. All other birds in the flock were then probabilistically determined to be Fully Vaccinated or Partially Vaccinated, using a parameter representing vaccine effectiveness, i.e. the probability that an individual chicken would be Fully Vaccinated after vaccination (Pfv).
Infection was seeded into the model by exposing all birds in a randomly selected cage to disease. Each iteration continued until the disease transmission cycle had ended or a sentinel chicken died. Pfv was varied (Pfv = 0.5, 0.6, 0.7, 0.8 and 0.9) and the effect upon the proportion of outbreaks that were detected by sentinel surveillance, the time until detection and the number of birds infected during the outbreak were recorded. One hundred iterations were carried out for each value of Pfv.
Results
The probability of detection was inversely related to the proportion of Fully Vaccinated birds (Figure 2). As Pfv increased from 0.5 to 0.9, the proportion of outbreaks that were detected through sentinel surveillance decreased from 100% to 78%. The effect of Pfv on the distributions of the time until an outbreak was detected through sentinel surveillance is presented (Figure 3a) showing that distributions became more positively skewed as Pfv increased up to 0.9. The median number of birds infected by the time disease was detected decreased from 291 to 13 as Pfv increased from 0.5 to 0.9 (Figure 3b). Where the outbreak died out without being detected, relatively few birds were affected: only 5 of 30 outbreaks that died out spread beyond the index cases and the maximum number of birds infected was 12.
Discussion and conclusion
An ideal surveillance system for HPAI in vaccinated flocks would allow the prompt detection of incursion of the infection, before it could spread to other farms. These results show that even for very high levels of flock immunity, surveillance of sentinels appears to be a sensitive mechanism for the detection of HPAI.
Although the time until detection tended to increase as vaccination effectiveness increased, this needs to be discussed in the context of spread of infection. As vaccination effectiveness increased, there was a trend towards lower numbers of birds being infected because the disease spread more slowly. This indicates that even though the disease may have been circulating in a flock for some time before detection the actual level of virus production, and hence the level of environmental contamination, will be lower as vaccination effectiveness increases. This would presumably reduce the risk of the disease spreading to other flocks.
Caution must be exercised when extrapolating these preliminary results to the field situation because parameter values have been estimated from a range of experimental studies that used different H5N1 virus isolates and only one transmission study involving vaccinated chickens was found. Furthermore, only one introduction scenario (exposure of all birds within one cage) was used for this study and it is possible that disease may be introduced to larger numbers of birds, leading to higher incidences of infection.
Further work will include modifying the model to incorporate other methods of passive surveillance, to investigate sampling strategies for serological surveillance schemes and to undertake sensitivity analysis on the flock structure and introduction scenarios.
References
Beard, C. (1998). Foreign Animal Diseases – The Gray Book. Cited 21/9/06.
Bublot, M., Le Gros, F-X, Nieddu, D, Pritchard, N, Mickle, T., Swayne, D. (In Press) Avian Diseases http://avdi.allenpress.com/pdfserv/10.1637%2F7623-042706 Cited 21/9/06.
Ellis, T.; Sims, L.; Wong, H.; Wong, C.; Dyrting, K., Chow, K., Leung, C., Peiris, J. (2006) Developments in Biologicals. 124: 133-134. European Commission (2006). Discussion paper: Vaccination of poultry against highly pathogenic avian influenza H5N1 (DIVA strategy) SANCO/10103/2006 rev. 2.
FAO and OIE (2005). Report of the second FAO/OIE regional meeting on Avian Influenza control in Asia, Food and Agriculture Organisation of the United Nations and the World Animal Health Organisation, Ho Chi Minh City, Viet Nam.
Kumar, M., Chu, H.-J., Rodenberg, J.,Krauss, S., Webster, R. (In Press) Avian Diseases. http://avdi.allenpress.com/pdfserv/10.1637%2F7605-041706 Cited 21/9/06.
Shortridge, K. F., N. N. Zhou, Guan, Y., Gao, P., Ito, T., Kawaoka, Y., Kodihalli, S., Krauss, S., Markwell, D., Murti, K.G. (1998). Virology 252(2): 331-342.
Swayne, D. (2003). Developments in Biologicals 114: 201-212.
From Proceedings of the “19th Australian Poultry Science Symposium”, New South Wales, Australia.
