spatial analysis of bovine spongiform encephalopathy in galicia, spain (2000–2005)
TRANSCRIPT
Spatial analysis of bovine spongiform
encephalopathy in Galicia, Spain (2000–2005)
A. Allepuz a,*, A. Lopez-Quılez b, A. Forte b,G. Fernandez c, J. Casal a
a Centre de Recerca en Sanitat Animal (CReSA)/Departament de Sanitat i Anatomia Animals,
Edifici V, Facultat de Veterinaria, Universitat Autonoma de Barcelona,
08193 Bellaterra, Barcelona, Spainb Grup d’Estadıstica espacial i temporal en Epidemiologia i medi ambient (GEeit�E�ma)/
Departament d’Estadıstica i Investigacio Operativa, Universitat de Valencia, Burjassot, Spainc Departamento de Sanidad Animal, Facultad de Veterinaria, Universidad de Santiago
de Compostela, Lugo, Spain
Received 16 May 2006; received in revised form 10 November 2006; accepted 28 November 2006
Abstract
In Spain, the first bovine spongiform encephalopathy (BSE) case was detected in 2000 in a cow
born in the Galicia region (Northwestern Spain). From then and until October 2005, 590 cases were
detected, 223 of them in Galicia.
In 1994, meat and bone meal (MBM) was banned on ruminant feed and, in 1996, an EU decision
mandating an overall change in MBM processing was implemented. This decision was gradually
applied in the territory and not enforced before July 1998. The objective of this study was to explore
clustering of BSE cases and estimate the standard incidence ratio (SIR) of BSE in Galicia. Our study
was based on the BSE cases detected during the surveillance period 2000–2005 in the Galicia region.
These cases were divided, based on birth date, into two periods: animals born from 1994 to July
1998, and those born after July 1998. We tested the role of cross-contamination on the geographical
SIR distribution for both periods. Hierarchical Bayesian models were used to model the over-
dispersion and lack of independence of the SIR estimates. The geographical distribution of the
standard incidence ratio of BSE between both periods was different. In the second period, the SIR
was reduced in some areas. The reduction in these areas could be attributable to the changes in the
www.elsevier.com/locate/prevetmed
Preventive Veterinary Medicine 79 (2007) 174–185
* Corresponding author. Present address: Unit of Veterinary Epidemiology, Centre de Recerca en Sanitat Animal
(CReSA), Campus UAB, Edifici CReSA, 08193 Bellaterra, Barcelona, Spain. Tel.: +34 93 581 4557;
fax: +34 93 581 44 90.
E-mail address: [email protected] (A. Allepuz).
0167-5877/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.prevetmed.2006.11.012
processing of MBM. We did not find any statistical link between the poultry population and the
standard incidence ratio, but pig population had a positive effect.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Bovine spongiform encephalopathy; Spatial analysis; Bayesian model; Cross-contamination
1. Introduction
Bovine spongiform encephalopathy (BSE) is a progressive neurological disorder of
cattle. It was first diagnosed in United Kingdom in 1986 (Wells et al., 1987). Different
studies suggest that the disease was originated from feed containing meat and bone meal
(MBM), contaminated by a scrapie-like agent derived from sheep or cattle (Willesmith
et al., 1988; Anderson et al., 1996). Based on these results, MBM was banned in ruminant
feed in Europe in 1994 (Decision 94/381).
Despite this ban, many cases of BSE appeared in animals born after 1994. Cross-
contamination between feedstuff for ruminants and other species has been suggested as a
hypothesis to explain the cases of animals born after MBM ban (Hoinville, 1994; Denny
and Hueston, 1997). The implementation of the July 1996 EU-Decision, mandating an
overall change in the rendering system to 133 8C, 3 bars and 20 min, was progressively
ensured during 1997 and 1998. For this study we have considered that the rendering system
was ineffective for inactivating the BSE agent until July 1998. Improperly processed MBM
was produced and sold to the feed industry until mid-1998. Therefore, until July 1998,
cattle feed could have been contaminated with low levels of MBM (Anon.; http://
europa.eu.int/comm/food/fs/sc/ssc/out126_en.pdf). In 2001, MBM were banned for all
domestic species (2000/766/CE).
The aim of this work is to explore BSE clustering, study the spatial distribution of the
standard incidence ratio (SIR) of BSE in the Galicia region (Northwestern Spain) and test
the role of cross-contamination in its geographical distribution.
2. Material and methods
2.1. Area under study
The study was conducted in the autonomous community of Galicia, which is in the
northwest of Spain, north of Portugal. It has an area of 295.75 km2 and is divided in 4
provinces and 314 municipalities.
In this region, the cattle population is 1,128,500 animals; 693,800 of them are cattle
aged more than 15 months (population at risk) and 444,000 dairy animals.
Lugo and Coruna provinces, with 290,658 and 283,550 head of cattle, respectively, are
the ones with highest cattle population in the Galicia region. In Coruna province the dairy
population represents 72% of the total cattle population, 204,084 out of a total of 283,550.
In Lugo province there is not such a big difference between beef and dairy population;
there are 171,658 dairy animals out of a total population of 290,658 animals.
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185 175
2.2. Data
Date and place of birth and date of detection of BSE cases were obtained from the
Galician government. Our study was based on cases detected by the surveillance system
from 2000 to June 2005. During this period, the detection of BSE cases was based on the
Mandatory Reporting System of clinical suspicions and the active surveillance programme
based on rapid tests carried out on every cow aged over 24 months. Because most of the
infections occur during calfhood (Willesmith et al., 1988) the period of infection for the
BSE cases was established as the year of birth, and they were located at the farm where the
animals were born. We included in the analysis BSE cases diagnosed in cattle born in
Galicia after 1994, and we divided them in two periods; between 1994 and July 1998 and
after July 1998. In the first period, cross-contamination is believed to have happened
because improperly processed MBM was produced and sold to the feed industry. In the
second period, after July 1998, when the change in MBM processing was made effective,
cross-contamination is not believed to have happened.
Data of the population at risk (cattle over 15 months), pig and poultry population were
obtained at municipality level from 1999 agricultural census published by Galicia
autonomous government. Taking into account that BSE incidence varies according to the
production type (Willesmith et al., 1988; Ducrot et al., 2003), the cattle population was
divided into dairy and beef cattle.
2.3. Cluster analysis
To explore the spatial distribution of the BSE cases we used SatScan1 v5.1 (http://
www.satscan.org) described by Kulldorff (1997). Geographical coordinates for most of
the farms were unknown. All the cases were located at the centroid of the municipality,
the smallest administrative unit, where the animal was born. The aggregation level
of the analysis was the municipality. We assumed that, under the null hypothesis,
the number of cases followed a Poisson distribution and we ran a purely spatial
analysis.
The spatial scan statistic can locate the site and test the significance of specific clusters.
It searches for clusters by using a variable circular window size to detect spatial clusters in
large areas, while controlling for the underlying population (Kulldorff, 1997). The circle is
centred on each of the points. For each point, the size of the circle varies continuously from
zero to some upper limit specified by the user (20% population at risk in our case). For each
location and size of the scanning window, the alternative hypothesis is that there is a higher
rate within the circle than outside. If a cluster is identified, the statistical significance is
obtained by Monte Carlo hypothesis testing (Kulldorf, 2003). We run two purely spatial
analyses for each period (animals born between 1994 and July 1998 and after July 1998)
one with the dairy and the other with the beef population.
2.4. Disease mapping
The aggregation level used in the analysis was the municipality (i = 1, . . ., 314). For
each municipality we estimated the standardised incidence ratio (SIRi), which represents
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185176
an increase/decrease in the risk of infection compared to a standard risk evaluated in the
whole Galicia region:
SIRi ¼Oi
Ei
being Oi and Ei the observed and expected BSE cases, respectively, in each municipality.
As the incidence varies according to the production type we differentiated the cases in
each municipality by breed. The expected cases in each municipality were estimated by
applying the overall BSE ratio in the whole of Galicia to the cattle population in each
municipality.
Ei ¼ Rdairy� Dairyi þ Rbeef � Beefi
Being ‘‘Rdairy’’ and ‘‘Rbeef’’ the overall BSE ratio in dairy and beef, respectively, while
Dairyi and Beefi are the dairy and beef populations in each municipality.
For an uncommon disease, a map of standardised rates suffers because the rates are
over-dispersed. Their variability only partly reflects genuine geographical heterogeneity,
the remaining variance being due to Poisson variability. The most extreme rates tend to
occur in the areas with the smallest populations. On the other hand, geographically close
areas tend to have similar disease rates (Clayton and Bernardinelli, 1992) so independence
of risk between close areas cannot be assumed. To model the over-dispersion of the
parameter of interest (SIRi) and take into account the risk dependence between close areas,
spatial hierarchical Bayesian models were used.
To implement the model we used WinBUGS, Software Version 1.4.1 (http://www.mrc-
bsu.cam.ac.uk/ bugs), Windows version, for Bayesian analysis using Gibbs sampling
(Lawson et al., 2003).
2.5. Construction of the Bayesian model
As BSE is a non-contagious and rare disease, we assume it occurs independently in
individuals with equal probability and that the observed number of BSE cases (Oi) in each
of the 314 municipalities can be modelled as a Poisson distribution centred on m.
miðmi ¼ SIRiEiÞ:
Oi� PoissonðmiÞTo take into account the lack of independence of the risk between contiguous
municipalities, a random component (bi) with spatial structure can be used. This
component is based on a matrix of contiguities between geographical units. This spatial
component takes into account the spatial dependence between adjacent municipalities. The
prior distribution of this parameter follows a normal distribution:
bi�Nð]i; tniÞwhere \i is the mean of the spatial component in the set of municipalities adjacent to
municipality i (neighbours) and t is the precision, inverse of variance, weighted by ni, the
number of neighbours of municipality i.
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185 177
To model the over-dispersion of the parameter of interest, the model also included a
random component (ai) without spatial structure. For the prior distribution of ai we chose a
normal distribution centred on zero:
ai�Nð0; nÞwhere n is the precision of the prior distribution. In order to test the hypothesis of cross-
contamination between cattle feedstuff and feedstuff for other species (pigs and
poultry) we added the population of these species as covariates in the model. Both
effects and the two covariates were linearly added to the prior distribution of the
logarithm of SIRi:
log SIRi ¼ b0 þ b1pigi þ b2poultryi þ ai þ bi
To compare the adjustment of the different models we used the deviance information
criterion (DIC). DIC is a measure of model fit penalised by the complexity of the model and
its value is calculated by adding the effective number of parameters to the posterior mean
deviance of the model.
The ‘‘best fit’’ model is the one with the smallest DIC value (Abrial et al., 2003). In
order to fully manage the statistical modelling prior distributions were established for
the parameters. Non-informative prior knowledge was considered with normal
distributions of mean 0 and variance 10,000 for both covariate parameters and gamma
distributions of mean 1 and variance 10 for t and n, the precision parameters of the two
random effects. The elicitation of hyperprior distributions in this type of models is a very
challenging issue in Bayesian statistics that is generating very fruitful discussions and
results (Gelman, 2006). In this scenario, Gamma models for precision parameters have
been the most popular proposal because of conditional conjugancy. We have been careful
with these prior distributions and we have not worked with extremely small
hyperparameter values.
The posterior distribution of the parameters was obtained from 25,000 Markov chain
Monte Carlo simulations (MCMC) after a burn-in of 5000 simulations. Conformity
analysis was made from the percentile posterior distribution given by the quantiles of the
usable chain, making appropriate interpretations in the context of disease mapping
(Richardson et al., 2004).
3. Results
3.1. Descriptive analysis
In Spain, the first BSE case was detected in 2000 in Galicia (Northwestern Spain). In
Galicia, 223 BSE cases have been confirmed, representing 36.5% of all the BSE cases
detected in Spain until October 2005. In Table 1, a descriptive analysis of the BSE cases
detected in Galicia is given. The incidence in the dairy and beef populations was different,
3.65 and 2.5 cases per 10,000 cattle in the dairy and beef cattle populations, respectively.
Of the 223 cases detected in the surveillance period 2000–2005, 11 cases were born before
the ban on meat and bone meal in cattle feed, 164 of them were born between the ban and
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185178
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185 179
Table 1
Frequency of BSE cases stratified by province, year of detection, age of animals and animal type (beef, dairy)
Region Age Year of detection Cases
Beef Dairy 2000 2001 2002 2003 2004 2005 Beef Dairy
Mean Min–Max Mean Min–Max
Lugo 7.0 5–10 6.2 4–11 1 14 7 26 24 16 55 33
Coruna 6.4 5–9 6.5 4–11 1 7 10 17 13 14 52 10
Pontevedra 6.7 5–9 6.2 4–9 0 2 14 18 15 11 47 13
Orense 6.6 5–9 6.7 5–9 0 3 3 1 5 1 4 9
Total 6.67 5–9.25 6.4 4.2–10 2 26 34 62 57 42 158 65
Fig. 1. Distribution by semesters of BSE cases of animals born in Galicia until October 2005 by birth cohort (total
223 cases). (a) Ban on MBM for ruminant feed. (b) Effective implementation of rendering system to inactivate the
BSE agent (July 1998).
Fig. 2. Representation of the location and size of the cluster of BSE cases in dairy (a) and beef (b) cattle between
1994 and July 1998 (circles) on a choroplethmap of the dairy and beef cattle population (municipalities with
higher populations are those with higher colour intensity). (a) Dairy cattle: observed, 43; expected, 17 ( p-
value = 0.001). (b) Beef cattle: observed, 15; expected, 3 ( p-value = 0.002).
July 1998, when the rendering system was ineffective to inactivate the BSE agent, and 49
cases were born after July 1998 (Fig. 1).
3.2. Cluster analysis
In the first period, between 1994 and July 1998, the SatScan detected significant clusters
in the dairy and beef cattle populations. The size of the clusters is quite different but they
are both located in the central part of the area under study. In Fig. 2, the location and size of
clusters are represented with a choropleth map of the beef and dairy populations. In the
second period, after July 1998, the SatScan did not detect any statistically significant
cluster.
3.3. Disease mapping
The spatial hierarchical Bayesian model with both random effects suggested that the
risk of BSE infection over Galicia was not homogeneous in both periods. In the first period,
from 1994 to July 1998, municipalities located in the central and southeastern areas were
exposed to a risk of infection between 2 and 4 times higher than in the rest of the territory.
In the second period, after July 1998, the higher-risk municipalities were located, mainly in
the northwestern part of the region (Fig. 3). We calculated the DIC value of the different
combinations of the model with the covariates. In both periods, the smallest DIC value was
the one with the two random components and the pig population as covariate. There is a
90% probability that there is a positive relationship between the BSE risk and the pig
population for both periods. In the first period (animals born between 1994 and July 1998)
the regression coefficient of the pig population was centred on 0.2083 (�0.03, 0.45). This
means an increase of between 0.97 and 1.57 times in the BSE risk for 10,000 pigs in each
municipality. In the second period, animals born after July 1998, the BSE risk was
increased by between 0.96 and 1.77 times, a mean value of 1.304 for 10,000 pigs in each
municipality (Tables 2 and 3).
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185180
Fig. 3. Geographical distribution of the standardised incidence ratio of BSE for both periods (SIR values are
multiplied by 100). (a) 1994–July 1998. (b) After July 1998.
A.
Allep
uz
eta
l./Preven
tiveV
eterina
ryM
edicin
e7
9(2
00
7)
17
4–
18
51
81
Table 2
Estimate of the parameters of the different combination of the models for period 1, when the rendering system was ineffective for inactivating the BSE agent (1994–July
1998)
Model Model (log mi) Estimate MC error Percentile distribution DIC
5% 10% 50% 90% 95%
Two random components,
pigs and poultry as covariates
log Ei + ai + bi + b0 + b1pigi +
b2poultryi
b0: �0.409 0.005 �0.73 �0.64 �0.39 �0.16 �0.1 398.774
b1: 0.306 0.003 0.03 0.09 0.29 0.5 0.56
b2: �0.904 0.008 �2.1 �1.81 �0.86 0.01 0.23
Two random components,
pigs as covariate
log Ei + ai + bi + b0 + b1pigi b0: �0.464 0.006 �0.79 �0.71 �0.45 �0.22 �0.16 397.447
b1: 0.208 0.004 �0.03 0.01 0.2 0.39 0.45
Two random components,
poultry as covariate
log Ei + ai + b0 + b1pigi b0: �0.326 0.007 �0.64 �0.56 �0.32 �0.1 �0.04 399.537
b1: �0.338 0.012 �1.32 �1.08 �0.32 0.39 0.6
Without spatial component,
pigs as covariate
log Ei + ai + b0 + b1pigi b0:�0.502 0.004 �0.78 �0.71 �0.49 �0.3 �0.25 397.705
b1: 0.330 0.002 0.11 0.16 0.33 0.5 0.56
DIC: Deviance information criterion.
A.
Allep
uz
eta
l./Preven
tiveV
eterina
ryM
edicin
e7
9(2
00
7)
17
4–
18
51
82
Table 3
Estimate of the parameters of the different combination of the models for period 2, when the rendering system was effective for inactivating the BSE agent (after July 1998)
Model Model (log mi) Estimate MC error Percentile distribution DIC
5% 10% 50% 90% 95%
Two random components,
pigs and poultry as covariates
log Ei + ai + bi + b0 + b1pigi +
b2poultryi
b0: �0.512 0.01 �1.5 �0.91 �0.48 �0.14 �0.05 189.474
b1: 0.319 0.003 �0.04 0.03 0.31 0.6 0.7
b2: �0.48 0.016 �2.42 �1.89 �0.39 0.79 1.1
Two random components,
pigs as covariate
log Ei + ai + bi + b0 + b1pigi b0: �0.536 0.01 �1.08 �0.94 �0.5 �0.16 �0.08 188.470
b1: 0.266 0.002 �0.04 0.02 0.26 0.5 0.57
Two random components,
poultry as covariate
log Ei + ai + b0 + b1pigi b0: �0.386 0.01 �0.89 �0.75 �0.36 �0.03 0.03 190.05
b1: 0.28 0.008 �1.02 �0.7 0.31 1.22 1.48
Without spatial component,
pigs as covariate
log Ei + ai + b0 + b1pigi b0: �0.473 0.009 �0.96 �0.84 �0.44 �0.14 �0.06 191.274
b1: 0.263 0.002 �0.02 0.04 0.26 0.48 0.54
DIC: Deviance information criterion.
4. Discussion
Several studies have been carried out in order to study the spatial distribution of BSE
(Doherr et al., 2002; Stevenson et al., 2000, 2005; Abrial et al., 2003, 2005a) in all of them
the geographical distribution of the disease has been not homogeneous and clusters have
been detected in specific locations.
In France, Abrial et al. (2003, 2005a) concluded that the risk of infection with the BSE
agent in Western France and in the whole French territory was not randomly distributed
after the ban of meat and bone meal in cattle feed. Poultry density did not significantly
influence the risk of BSE, whereas the pig density was significantly associated with an
increase in the risk of 2.4% per 10,000 pigs (Abrial et al., 2005b).
In Great Britain, Stevenson et al. (2005) observed that after the feed ban there were
spatially aggregated areas of unexplained risk in the northern and eastern regions of Great
Britain. Area-level BSE risk was additionally associated with greater number of pigs per
area, relative to cattle. In Switzerland, the clusters were located in regions with a high pig
density, but there were other regions with higher pig and poultry densities that had no
clusters (Doherr et al., 2002).
In Spain, MBM was banned for ruminant feeding in 1994, but not until 2001 in other
species. The July 1996 EU-Decision, mandating an overall change in the rendering system,
was applied gradually over the country and until July 1998 the rendering system was not
considered effective to inactivate the BSE agent. The distribution of BSE cases by birth
year (Fig. 1) suggested that after mid-1998 the number of BSE cases began to decrease.
This distribution could indicate that the change on the rendering system of meat and bone
meal from mammalian animals for animal feed was an effective measure, reducing the risk
of cross-contamination and therefore the number of cases. This is supported by the change
in the spatial distribution of the standardised incidence ratio of BSE between both periods.
In the first period, there are some areas in the central and southeastern part where the SIR
value has been between 2 and 4 times higher than in other areas. In the second period some
of these areas present a value near to 1, indicating a reduction of the risk which could be
attributable to the changes in the processing of MBM (Fig. 3).
Despite the reduction of risk in the second period, the model suggested a positive
relationship between pig population and BSE risk for both of them. The influence of pig
population on the standard incidence ratio in the first period, when rendering system was
not effective, and the second period was not different. The gradual application over the
region of the change in feed processing could partly explain the fact that pig population had
a positive effect on BSE risk in the second period. On the other hand, local delivery of feed
has been assumed in this study and this is a practice that could always not been carried out.
Big feed companies may supply feed to farms situated in distant municipalities.
In our study, we did not find a statistical link with poultry population. This could be
explained by the high level of autonomy in the poultry industry. In many cases, the same
company is the owner of the farm and the feed processing plant, so it is not common to find
feed processing plants delivering feed for poultry and other species.
Finally, there could be other factors influencing the spatial distribution of the SIR of BSE.
Regional differences in the processing and use of MBM by rendering plants could make
regional differences in the transmission and amplification mechanisms of the infective agent
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185 183
(Stevenson et al., 2005). Different rates of cross-contamination with contaminated MBM
between areas (Sheridan et al., 2005), the heterogeneity in the level of vigilance between
regions (Cuenot et al., 2003), regional differences in the levels of compliance with the feed
ban and the identification and reporting of cases between areas (Doherr et al., 2002), or
different risk factors characterising the feedstuff factories and the potential for cross-
contamination on farms or at the factories, might produce different BSE risks (Abrial et al.,
2005a). However, it is not easy to have reliable information associated with all these factors,
and because of that it was not possible to include them in the statistical analysis.
5. Conclusion
The spatial distribution of the risk of infection with the BSE agent is not homogeneous
in both periods under study. In the first period, there are some areas in the central and
southeastern part of Galicia where the risk of infection has been between 2 and 4 times
higher than in other areas. In the second period, these areas show a standard incidence ratio
near to 1, indicating a reduction of the risk. This reduction could be attributable to the
changes in the processing of MBM in the region.
References
Abrial, D., Calavas, D., Lauvergne, N., Morignat, E., Ducrot, C., 2003. Descriptive spatial analysis of BSE in
western France. Vet. Res. 34, 749–760.
Abrial, D., Calavas, D., Jarrige, N., Ducrot, C., 2005a. Spatial heterogeneity of the risk of BSE in France following
the ban of meat and bone meal in cattle feed. Prev. Vet. Med. 67, 69–82.
Abrial, D., Calavas, D., Jarrige, N., Ducrot, C., 2005b. Poultry, pig and the risk of BSE following the feed ban in
France—a spatial analysis. Vet. Res. 36, 615–628.
Anderson, R.M., Donelly, C.A., Ferguson, N.M., Woolhouse, M.E.J., Watt, C.J., Udy, H.J., MaWhinney, S.,
Dunstan, S.P., Southwood, T.R.E., Wilesmith, J.W., Ryan, J.B.M., Hoinville, L.J., Hillerton, J.E., Austin, A.R.,
Wells, G.A.H., 1996. Transmission dynamics and epidemiology of BSE in British cattle. Nature 382, 779–787.
Anon., 4/12/2000. European Decision 2000/766/EC.
Anon., 8/7/1996. European Decision 1996/449/EC.
Anon., 27/6/94. European Commission Decision 1994/381/EC.
Anon., 2000. Report on the Assessment of the Geographical BSE Risk of Spain (http://europa.eu.int/comm/food/
fs/sc/ssc/out126_en.pdf).
Clayton, D.G., Bernardinelli, L., 1992. Bayesian methods for mapping disease risk. In: Elliott, P., Cuzick, J.,
English, D., Stern, R. (Eds.), Geographical and Environmental Epidemiology: Methods for Small-Area
Studies. Oxford University Press, Oxford, pp. 205–220.
Cuenot, M., Calavas, D., Abrial, D., Gasqui, P., Cazeau, G., Ducrot, C., 2003. Temporal and spatial patterns of the
clinical surveillance of BSE in France, analysed from January 1991 to May 2002 through a vigilance index.
Vet. Res. 34, 261–272.
Denny, G.O., Hueston, W.D., 1997. Epidemiology of bovine spongiform encephalopathy in Northern Ireland 1988
to 1995. Vet. Rec. 140, 302–306.
Doherr, M.G., Hett, A.R., Rufenacht, J., Zurbrigen, A., Heim, D., 2002. Geographical clustering of cases of bovine
spongiform encephalopathy (BSE) born in Switzerland after the feed ban. Vet. Rec. 151, 467–472.
Ducrot, C., Roy, P., Morignat, E., Baron, T., Calavas, D., 2003. How the surveillance system may bias the results of
analytical epidemiological studies on BSE: prevalence among dairy versus beef suckler cattle breeds in
France. Vet. Res. 34, 185–192.
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185184
Gelman, A., 2006. Prior distributions for variance parameters in hierarchical models. Bayesian Anal. 1, 515–534
(comment on article by Browne and Draper).
Hoinville, L.J., 1994. Decline in the incidence of BSE in cattle born after the introduction of the feed ban. Vet. Rec.
134, 274–275.
Kulldorff, M., 1997. A spatial scan statistic. Commun. Stat. Theor. M 26, 1481–1496.
Kulldorf M., 2003. SatScan User Guide for Version 4.0.3. http://www.satscan.org.
Lawson, A.B., Browne, W.J., Vidal Rodeiro, C.L., 2003. Disease mapping with WinBUGS and MLwiN. Wiley,
Chichester.
Richardson, S., Thomson, A., Best, N.G., Elliott, P., 2004. Interpreting posterior relative risk estimates in disease
mapping studies. Environ. Health Perspect. 112, 1016–1025.
Sheridan, H.A., MacGrath, G., Fallon, R., Shoukri, M.S., Martin, S.W., 2005. A temporal-spatial analysis of
bovine spongiform encephalopathy in Irish cattle herds, from 1996 to 2000. Can. J. Vet. Res. 69, 19–25.
Stevenson, M.A., Wilesmith, J.W., Ryan, J.B.M., Morris, R.S., Lawson, A.B., Pfeiffer, D.U., Lin, D., 2000.
Descriptive spatial analysis of the epidemic of bovine spongiform encephalopathy in Great Britain to June
1997. Vet. Rec. 147, 371–384.
Stevenson, M.A., Morris, R.S., Lawson, A.B., Wilesmith, J.W., Ryan, J.B.M., Jackson, R., 2005. Area-level risks
for BSE in British cattle before and after July 1988 meat and bone meal feed ban. Prev. Vet. Med. 69, 129–144.
Wells, G.A., Scott, A.C., Johnson, C.T., Gunning, R.F., Hancock, R.D., Jeffrey, M., Dawson, M., Bradley, R.,
1987. A novel progressive spongiform encephalopathy in cattle. Vet. Rec. 121, 419–420.
Willesmith, J.W., Wells, G.A.H., Cranwell, M.P., Ryan, J.B.M., 1988. Bovine spongiform encephalopathy:
epidemiological studies. Vet. Rec. 123, 638–644.
A. Allepuz et al. / Preventive Veterinary Medicine 79 (2007) 174–185 185