Download Estimating the prevalence of infectious bursal disease

Estimating the prevalence of infectious bursal disease within the Thai poultry
industry.
K.
L. A. Kelly*1, 2, L Ratananakorn 3, A. Chaisingha3, C. Mahanchaisakul3,
Asayuth3 and . Mathayompong 3.1Centre for Epidemiology and Risk Analysis,
Veterinary Laboratories Agency, UK, 2Department of Statistics and Modelling
Science, University of Strathclyde, UK, 3 Department of Livestock Development,
Ministry of Agriculture, Thailand
Summary
This paper describes a study designed to estimate the prevalence of infectious bursal
disease (IBD) in commercial broiler chickens in Thailand. The study was undertaken
in 2 stages. Firstly, experimental studies were designed to provide quantitative data.
Secondly, the collected data were used, together with data from the literature, to
develop a Monte-Carlo simulation. Results from the simulation indicate that flock
prevalence is between 0.004 and 0.046 (90% uncertainty interval). This information,
together with the fact that IBD is unlikely to survive cooking, leads to the conclusion
that cooked Thai chicken products present a low risk with respect to IBD.
Introduction
Infectious bursal disease (IBD) is an acute, highly contagious viral infection of young
chickens. The disease was first reported in broiler chickens in central Thailand in
1973, however, since the introduction of control programmes, there has been no
report of any serious IBD outbreaks in any commercial broiler farms.
Thailand is one of the major poultry meat producers of the world and each year,
approximately 20% of all production is exported. Given the importance of the poultry
industry to the Thai economy, the prevention of infectious diseases such as IBD is
critical, particularly for commercial farms which supply the export market.
Objectives
The objective of this study was to estimate the prevalence of IBD virus in commercial
broiler farms and its associated uncertainty.
Materials and methods
Estimation of prevalence was undertaken in 2 stages. Firstly, experimental studies
were designed to provide quantitative data. Secondly, the collected data were used,
together with data from the literature, to develop a Monte-Carlo simulation
(implemented using @RISK © Palisade Corp.).
In total, 455 (n) commercial flocks were tested for infection with IBD. Twenty birds
from each flock were randomly sampled at the point of slaughter and the bursa from
these birds were pooled and tested using RT-PCR. Of the 455 flocks tested, 10 (r)
were found to be infected. These results were used together with estimates for the
sensitivity (SF) and specificity (SPF) of the testing procedure to calculate the flock
prevalence (PF).
Proceedings of the 10th International Symposium on Veterinary Epidemiology and Economics, 2003
Available at www.sciquest.org.nz
Sensitivity of the testing procedure (SF)
The sensitivity of the testing procedure was defined using equation (1)
S F = S PCR × (1 − (1 − PW ) 20 )
(1)
where PW is the within flock prevalence in infected flocks at the time of slaughter and
SPCR is the sensitivity of the PCR test, given that there is at least one infected bursa in
the pooled sample of 20 bursa.
PCR is accepted to be the most sensitive test for detection of IBD virus (Moody et al,
2000). Despite this, there are no quantitative estimates available for this parameter.
Based on the qualitative information available and discussions with microbiological
experts the sensitivity of the RT-PCR test was assumed to be between 90% and 99%.
The uncertainty surrounding SPCR was then described by a Uniform distribution.
The parameter PW is an estimate of the percentage of birds in the flock that have IBD
virus in their bursa. This value depends on the age at infection (AgeI), the duration of
infection in the bursa (DurB) and the age at slaughter (AgeS). A probability
distribution for AgeI was determined using results from a serological study
undertaken on commercial flocks in Thailand. In this study, birds from 66 flocks
were blood sampled at 1, 7, 14, 21, 28, 35 and 42 days old and the antibody levels
recorded. It was assumed that if bird has an antibody titre of less than 1000, then
infection is possible (Bruce, 1992) The cumulative conditional probability of age
given an antibody titre of <1000 was then used as the distribution for AgeI
The period of time post-infection over which IBD virus is recoverable from the bursa
was described as ranging uniformly from 1 to 28 days by MAF (1999) in their IBD
risk assessment. The same distribution was used here to characterise the variability in
DurB.
In Thailand, commercial flocks are slaughtered between 42 and 56 days of age, with
most birds being slaughtered at 48 days old. The variability distribution for AgeS was
thus described by a Pert distribution.
A separate simulation (2000 iterations) was used to estimate PW. In summary, this
simulation determined the proportion of times AgeI+DurB<AgeS ( Pw ). The
uncertainty associated PW was then described by a Normal distribution,

P (1 − PW ) 
PW = Normal PW , W

20000 


Proceedings of the 10th International Symposium on Veterinary Epidemiology and Economics, 2003
Available at www.sciquest.org.nz
Specificity of the testing procedure (SPF)
The specificity of the testing procedure is equal to the specificity of the PCR. Most
studies on RT-PCR suggest 100% specificity. However, as for the sensitivity of the
PCR, there is no quantitative data available to characterise the uncertainty in this
parameter. Microbiologists in Thailand were of the opinion that a plausible range
would be from 0.98 to 1 and that a Uniform distribution would be appropriate to
describe the uncertainty.
Estimate of PF
Bayesian inference was used to describe the uncertainty associated with PF . Firstly,
an uninformed prior was assumed for PF, that is π(PF) =1. Secondly, as the total
number of flocks sampled (n=455) was considered small in relation to the total
number of flocks slaughtered per year, a Binomial likelihood was used for the number
of positives observed (r=10). Thus,
l(n,r|PF) = Binomial(n, PFSF+(1-PF)(1-SPF))
Finally, the posterior distribution (f(PF|n,r)) was given by
f(PF|n,r) = π(PF) l(n,r|PF)
Results and discussion
The Monte-Carlo simulation was run for 10,000 iterations and the mean flock
prevalence was estimated to be 0.022. The corresponding 5th and 95th percentiles were
0.004 and 0.046. Thus, there is 90% certainty that the flock prevalence is between
0.004 and 0.046, representing between 4 and 46 infected flocks in every 1,000.
Infectious bursal disease virus persists for a short period of time in muscle and is very
likely to be killed during cooking. This information, together with the estimate of
flock prevalence leads to the conclusion that cooked Thai poultry presents a low risk
with respect to IBD. This result is particularly important for the Thai Poultry Industry
and Veterinary Services as well as potential export markets word-wide. The approach
adopted here also highlights the benefits of parallel data collection and model
development in ensuring a scientific basis for risk assessment in international trade.
References
1. Bruce L.H., Gray D.B., Richard D.M., Alfredo F.R. (1992) – Subclinical
infectious bursal disease in an integrated broiler production operation. J. Vet.
Diagn. Invest. 4, 406 – 411.
2. Ministry of Agriculture and Forestry (1999). Import risk analysis: chicken
meat and chicken meat products; Bernard Matthews Foods Ltd turkey meat
preparations from the United Kingdom. March 1999.
3. Moody,A., Sellers, S. and Bumstead,N. (2000). Measureing infectious bursal
disease virus RNA in blood by multiplex real-time quantitative RT-PCR.
J.Virol Methods, 85(1-2):55-64.
Proceedings of the 10th International Symposium on Veterinary Epidemiology and Economics, 2003
Available at www.sciquest.org.nz