SEARCH VEGSOURCE:

 

 

Follow Ups | Post Followup | Back to Discussion Board | VegSource
See spam or
inappropriate posts?
Please let us know.
  




From: TSS ()
Subject: Spatial heterogeneity of the risk of BSE in France following the ban of meat and bone meal in cattle feed
Date: February 13, 2005 at 4:06 pm PST

-------- Original Message --------
Subject: Spatial heterogeneity of the risk of BSE in France following the ban of meat and bone meal in cattle feed
Date: Sun, 13 Feb 2005 18:06:56 -0600
From: "Terry S. Singeltary Sr."
To: Bovine Spongiform Encephalopathy
CC: ylepape@aol.com, vdedet@invivo.edu

Spatial heterogeneity of the risk of BSE in France following the ban
of meat and bone meal in cattle feed

David Abriala
,
Corresponding Author Contact Information
,
E-mail The Corresponding Author ,
Didier Calavasb
,
Nathalie Jarrigeb

and Christian Ducrota

aUnité dEpidémiologie Animale, INRA, Centre de recherche de
Clermont-Ferrand-Theix, 63122 Saint Genès Champanelle, France
bUnité Epidémiologie, AFSSA Lyon, 31 avenue T. Garnier, 69364 Lyon Cedex
07, France

Received 19 March 2004; revised 17 September 2004; accepted 4 October
2004. Available online 19 December 2004.

Preventive Veterinary Medicine


Volume 67, Issue 1

, January 2005, Pages 69-82


Abstract

In France, meat-and-bone meal (MBM) has been prohibited for cattle
feeding since 1990, but bovine spongiform encephalopathy (BSE) cases,
called NAIF, appeared in animals born after this feed ban.
Furthermore, in 1996 a new measure was taken: removal of cadavers and
specified risk materials (SRM) from the processing of MBM dedicated to
animal feed. Nevertheless, BSE cases (called super-NAIF) appeared in
cattle born after this measure was in force. We analysed the spatial
distribution of 445 NAIF and 58 super-NAIF cases detected in France
from July 1, 2001 to July 31, 2003. The detection of BSE was based both
on the mandatory reporting system (MRS) and the systematic test
screening of cattle at the abattoir and at the fallen-animal plant with
rapid tests. The background population was based on the adult-cow census.

The disease mapping of the BSE risk was based on the standardised
incidence ratio (stochastic Poisson process). A spatial component, which
takes into account the spatial dependence between the geographical units
by a notion of adjacency was used to eliminate the over-dispersion in
the risk assessment. The geographical units were defined by hexagons
with a side of 23 km (France had 1264 hexagons). The parameters were
estimated by a Metropolis Gibbs sampling algorithm using the
Markov-chain Monte Carlo methods.

The BSE cases were not randomly distributed. Furthermore, the areas at
risk for the super-NAIF matched part of the areas at risk for the
NAIF caseswhich suggests that it might be a common source of
contamination.

Keywords: BSE; Bovine; Spatial Analysis; Disease Mapping; Markov-chain
Monte Carlo


Article Outline

1. Introduction

2. Data collection

3. Statistical Methods

3.1. The disease-mapping model

3.2. Estimations of the parameters

3.3. The ordinary Kriging


4. Results

5. Discussion

6. Conclusion

Acknowledgements

References


1. Introduction

In France, the first case of bovine spongiform encephalopathy (BSE) was
described in 1991 (Gouëllo, 1991
).
Since then, the peak incidence of detected BSE cases was reached in 2001
(273 cases), and in all, 857 BSE cases had been detected up to July 31,
2003.

The first control measure, in 1990, consisted in the ban of
meat-and-bone meal (MBM) for cattle feeding. However, BSE cases appeared
in animals born after this feed ban (Ducrot et al., 2000
)
and were called NAIF (for nés après linterdiction des farines
animales). Furthermore, in June 1996 a new measure was taken: the
removal of cadavers and specified risk materials (SRM) from the
processing of MBM dedicated to animal feed (DGAL, 2003
).
Despite this complementary preventive measure, BSE cases appeared in
cattle born after it was in force and were called super-NAIF. Up to
July 31, 2003, 445 NAIF cases, defined as cases born between January
1, 1991 and July 1, 1996, and 58 super-NAIF cases, defined as cases
born between July 1, 1996 and January 1, 2001 have been detected. The
BSE cases born after January 1, 2001 correspond to another category
since the ban of MBM in the feed of all species was started.

The main hypothesis for the contamination of the NAIF and super-NAIF
cases is the cross-contamination between cattle feedstuff and
monogastric feedstuff either at the factory, during the shipping or on
the farm (BNEVS, 2000
,
Ducrot and Calavas, 2002
,
Wilesmith et al., 1988

and Wilesmith et al., 2000
).
Another possibility is the contamination with the BSE agent of different
by-products used in the food industry (such as bones and tallow). If
these assumptions are true, we expect that depending on the raw
materials and the technology used by the feedstuff industry and on the
type of food produced in the factories, the BSE infection risk varies in
space.

A preliminary descriptive spatial analysis of the NAIF cases was
carried out by Abrial et al. (2003)

on cases detected in Western France between August and December 2000.
The main conclusion was that the risk of infection with the BSE agent
was not distributed randomly in this part of France.

We extended that preliminary study to the whole French territory and to
a larger period of time, from July 1, 2001 to July 31, 2003. Moreover,
both the NAIF and the super-NAIF cases were included (but analysed
separately). The goals were to draw the disease mapping of the BSE risk,
both for the NAIF and super-NAIF cases, to highlight the geographic
areas with a significant higher risk, and to compare the geographical
distribution of NAIF and super-NAIF cases. The disease mapping was
based on the method of the standardised incidence ratio (SIR) with
Poisson inference (Mollié, 1999
).
Compared to our preliminary study in Western France (Abrial et al., 2003
),
the mapping methods were improved. The choice of the geographical unit
was reconsidered, the over-dispersion of the risk estimators was reduced
by the use of spatially structured priors (Besag et al., 1991
)
and an ordinary Kriging was used to eliminate the arbitrary of
geographical units.


2. Data collection

Epidemiological data on BSE were provided by the Agence Française de
Sécurité Sanitaire des Aliments (AFSSA Lyon, France), in charge of the
monitoring of BSE.

The analysis was restricted between July 1, 2001 to July 31, 2003 to get
precise and comparable data on BSE incidence. During this period, the
detection of BSE was based both on the mandatory reporting system and on
the comprehensive active surveillance programme based on rapid tests and
carried out on all cattle ? 2 years old, dead or slaughtered (Calavas et
al., 2001

and Calavas and Ducrot, 2003
).
These two systems were complementary because they allowed the screening
of every bovid ? 24 months old, dead or slaughtered.

BSE cases taken into account in the analysis were either clinically
suspect animals confirmed at the national reference laboratory of AFSSA
with Western blot or immunohistochemistry (i.e. cases found with the
mandatory reporting system) or test-positive animals (in the rapid tests
that passed the European Union validation (Moynagh and Schimmel, 1999
))
confirmed with the same two techniques, among the entire cattle
population tested within the active surveillance programme (Calavas et
al., 2001
).

The geographical location of the BSE cases was defined as the location
of the commune (the smallest French administrative unit) of the farm
in which the case has been raised between the sixth and twelfth months
after birth. This period corresponds to the highest risk of infection
based on modelling results (Donnelly, 2000

and Supervie and Costagliola, 2004
).

BSE incidence varies according to the production type (dairy versus
suckler cattle) (Ducrot et al., 2003
,
Morignat et al., 2002

and Wilesmith et al., 1988
).
We took into account this factor in the analysis (in particular, in the
standardisation of BSE cases in mapping disease). We defined production
type on the basis either of the breed of the BSE case, or of the
production type of the farm (when the breed was not recorded or when the
case was of a mixed breed).

The background population was assessed by the demography of the female
adult bovids, (i.e. cattle having calved), available at the level of the
canton (French administrative unit including five communes on
average; France is divided into 3705 cantons). Data were obtained from
the Agricultural Census 2000 edited by the Statistics Office of the
Ministry of Agriculture and Fisheries (Agreste 251 rue de Vaugirard 
75732 Paris, France).

The geographical data on the cantons perimeters and the communes
centroids were provided by the GEOFLA® France Métropolitaine (IGN©
Paris, version 6, 2002).


3. Statistical Methods


3.1. The disease-mapping model

Disease-mapping methods generally are linked to administrative units.
Indeed, the demographic raw data are aggregated at the canton level
(150 km2 on average).

Theses units are not homogenous in shape, size and number of neighbour
units. To overcome this problem, we built our own geographical units
redefined as hexagons of 23-km width and 450-km2 area. The French
territory thus was divided into 1264 contiguous hexagons labelled i = 1,
&, 1264. The number of BSE cases (commune level) and the population of
the female adult cattle (canton level) were aggregated at the hexagon
level.

Let yi denote the count of observed BSE cases, ei the number of expected
cases and ri the unknown SIR in hexagon i. It followed from the null
hypothesishomogenous spatial distribution of the BSE riskthat the
expected number of cases (ei) was obtained by applying the overall BSE
rate in France to each hexagon, standardised on the type of cattle breed
(dairy versus beef). The SIR ri represented the increase/decrease in the
risk of contamination compared to a standard risk evaluated on the whole
French territory.

We assumed that yi followed a Poisson-distribution (rare event and large
population size). So, the probability of observing yi cases correspond
to the first level of the model:


yinot, vert, similar Poisson(?i) with i=1,&,1264. First level

(1)

In the methodology of the generalised hierarchical model, the prior
model must be defined. Besag et al. (1991)

proposed smoothing the raw SIRs by using the intrinsic conditional
autoregressive (CAR) model with a spatial component Si. So, the log of
the ?i parameters are decomposed into two components:


log(?i)=log(ei)+Si. Second level

(2)

The Si are the spatial components which take into account the spatial
dependence between the hexagons (and use the notion of adjacency). The
distributions of these spatial parameters are:


Click to view the MathML source

(3)

Click to view the MathML source

denotes the mean of the spatial component in the set ?i of hexagons
adjacent to hexagon i and Click to view the MathML source
is
the variance inversely weighted by the number of neighbours of hexagon
i. Finally, we had the SIR by the following expression:

ri=eSi.

(4)

In Eq. (2)
,
we could add an geographically unstructured componentto account for
Gaussian white noise with unknown variance (Besag et al., 1991
).
But, with our data, the part of variability of this unstructured
component was negligible in comparison with the structured component
(Si). The variability of this unstructured component was evaluated by
the variance parameter of a normal prior assessed with the
Gibbs-sampling method (see Section 3.2). We made the choice to drop this
component to simplify the model.


3.2. Estimations of the parameters

The WinBUGS Package (URL:
http://www.mrc-bsu.cam.ac.uk/bugs/welcome.shtml) is a free software for
Bayesian inference using Gibbs-sampling for the estimations and analysis
of stochastic models using Markov-chain Monte Carlo methods. It grew
from a statistical research project at the Medical Research Council
Biostatistics Unit of Cambridge (London). We have used the Conditional
Autoregressive Model function implemented in the geographical extension
of WinBUGS version 1.4: GeoBUGS developed by a team at the Department of
Epidemiology and Public Health of Imperial College at St. Mary's
Hospital, London.

Gibbs sampling is an adaptation of the general Metropolis algorithm
(Gilks et al., 1996
).
It consists in visiting each parameter (called node) in turn and
simulating a new value for this parameter from its full conditional
distribution given the current values for the remaining parameters
(Mollié, 1999
).
In our analysis, we obtained a Markov-chain from which a sufficiently
long burn-in of samples (size 10,000 cycles) was discarded from
calculations, followed by an effective and usable chain (size 100,000
cycles). The parameters of interest were estimated from the sampling of
20,000 cycles from the chain, after the verification of the stability of
the Markov-chains. The verification was made with the HeidelbergerWelch
convergence diagnostic (Heidelberger and Welch, 1983
).

To identify the hexagons with a SIR significantly >1, we computed a 99%
confidence interval based on the method proposed by Mollié (2000)
.
The confidence interval was estimated by the 0.5 and 99.5% quantiles
calculated from the Gibbs sampling of 20,000 cycles.


3.3. The ordinary Kriging

The disease-mapping method with the hexagons produced an unusual lattice
for the French area. Even if this lattice is usable for a correct
interpretation, we preferred transforming the set of hexagons into a map
with continuous spatial SIR to eliminate the cutting into hexagons. We
attributed the SIR value of each hexagon to its centroid to obtain a
regular point pattern.

Thus, an ordinary Kriging (Cressie, 1993
,
pp. 119) could be used on this point pattern. Kriging is a
minimum-mean-squared-error method of interpolation. Data (SIR in our
case) on points where values were unobserved, were predicted from a set
of values observed in a sample of spatial points (the centroids of the
hexagons). The method uses a spatial regression model empirical
semivariogram to link the values of interest and the geography. The
empirical semivariogram provides information on the spatial
autocorrelation of the dataset. However, it does not provide information
for all possible horizontal directions and distances. For this reason,
and to ensure that Kriging predictions had positive Kriging variances,
we fit the model (i.e., a continuous function or curve) to the empirical
semivariogram. The model was based on the squared difference between the
values for all pairs of points with observed values. In our maps, we
used the spherical model for the semivariogram (Cressie, 1993
,
pp. 58). Ordinary Kriging is implemented in Arcview® software; the
output of Kriging is a set of curve levels for the SIR that are drawn on
maps (Fig. 2
).


4. Results

The epidemiological and demographic data are presented in Table 1

for the 22 regions of France. (The region is the greatest French
administrative unit, with on average 25,000 km2 and 56 hexagons.) In the
period from July 1, 2001 to July 1, 2003, 366 NAIF and 48 super-NAIF
BSE cases were detected in the dairy population of 4.39 millions cows
(8.33 NAIF and 1.09 super-NAIF cases per 100,000 dairy cows). In the
same period, 79 NAIF and 10 super-NAIF BSE cases were detected in
the beef population of 3.86 million cows (2.05 NAIF cases and 0.26
super-NAIF per 100,000 beef cows). (No super-NAIF case born after
July 1, 2001 has been identified as of this date.)

Table 1.

Numbers BSE cases (NAIF and super-NAIF; July 2001 to July 2003)
for the 22 regions of France, and cattle population (Agricultural
Census 2000)
Regions BSE cases Adult female cattle population (×103)

NAIF Super-NAIF

Dairy Suckling Dairy Suckling Dairy Suckling
Alsace 3 0 0 0 54 14
Aquitaine 11 6 1 1 146 266
Auvergne 13 6 0 0 287 456
Basse-Normandie 43 1 7 0 530 121
Bourgogne 3 5 1 0 71 459
Bretagne 88 4 5 0 791 123
Centre 13 7 2 0 76 194
Champagne-Ardenne 10 2 0 0 122 100
Corse 0 0 0 0 2.1 25
Franche-Comté 20 0 2 0 213 41
Haute-Normandie 7 0 2 0 162 61
Ile-de-France 1 0 0 0 7.5 6.7
Languedoc-Roussillon 0 0 0 0 28 62
Limousin 4 13 1 1 46 468
Lorraine 6 2 0 1 222 131
Midi-Pyrénées 10 11 1 0 200 419
Nord-Pas-de-Calais 15 1 1 0 215 61
Pays de la Loire 59 9 12 4 607 406
Picardie 12 1 3 1 151 65
Poitou-Charentes 10 7 2 2 127 212
Provence-Alpes-Côte-dAzur 0 0 0 0 13 13
Rhône-Alpes 38 4 8 0 321 153

Total 366 79 48 10 4,393 3,857

The 22 regions of France are presented in Fig. 1
.

Enlarge Image

(28K)

Fig. 1. The 22 regions of France.

The variability of the geographical structured component Si was
significantly > zero for the NAIF cases (99% confidence interval:
3.17, 14.13) and for the super-NAIF (99% confidence interval: 4.15,
16.46). This indicated that there was significant spatial heterogeneity
of the BSE incidence, for both type of cases detected between July 1,
2001 and July 31, 2003. During this period, the BSE cases were not
distributed randomly in France.

Fig. 2

shows the disease mapping of the BSE risk according to the spatial model
(Eq. (2)
.
In these maps, the SIR of the BSE risk ranges between 0 and 2.1 for the
NAIF cases (map A) and between 0 and 1.36 for the super-NAIF ones
(map B).


Enlarge Image

(82K)

Fig. 2. (A) Disease mapping of the standardised incidence ratio of
BSE risk for the NAIF cases (born between 01/01/1991 and
07/01/1996) in the period July 2001 to July 2003 (intrinsic
conditional autoregressive model without covariate, Markov-chain
Monte Carlo methods for estimations, and ordinary Kriging). (B)
Disease mapping for the super-NAIF cases (born between 07/01/1996
and 01/01/2001) in the same period.

Fig. 3

highlights the hexagons for which the SIR was significantly >1 at level
1%. For the NAIF cases (map A), 455 (36%) of the 1264 hexagons had a
SIR significantly >1, and only 240 (19%) for the super-NAIF cases (map
B).


Enlarge Image

(71K)

Fig. 3. Centroids of the hexagons with a SIR significantly >1 at
alpha = 1%. For the NAIF cases (map A) 455 hexagons were
significant and 240 for the super-NAIF (map B).


5. Discussion

The first part of the discussion is focused on the method and data used.
The disease mapping method is very well discussed by Mollié (1999)

and Diggle (2000)
.
With rare events such as BSE, disease mapping commonly is based on a
Poisson model with the use of the standardised expectations (ei)
introduced by Diggle et al. (1997)
.
The assumption that the observed cases follow a Poisson distribution is
essential but rarely verified with the observations. The main sources of
extra-Poisson variation that are linked to the geographical units are:
(1) the covariates not taken into account in the model and supposed to
be constant within each geographical unit and (2) more important, the
residual spatial variation. For the covariates, the geographical unit
must be as small as possible. In Eq. (2)
,
the residual spatial variation is the term Si, so the extra-Poisson
variation due to this residue is controlled. Prior information about the
term Si (Eq. (3)
)
was proposed by Besag (1974)
,
Clayton and Kaldor (1987)

and by Besag et al. (1991)
.
The Si follow a multivariate Gaussian distribution in which dependence
between a pair of geographical units is determined by their spatial
location. This model is specified for each Si given all other Sj (j ? i)
and such a model is complicated by the irregular spatial arrangement of
the administrative units (the canton in our analysis) commonly used.
To overcome this problem, Clayton and Kaldor (1987)

proposed adding into the risk assessment a weighted matrix of constants
(real values), which represents the strength of adjacency between the
geographical units. It is difficult in practice to decide upon a degree
of adjacency; so, we used a simple weighted matrix of binary values
(Clayton and Bernardinelli, 1992
).
Moreover, Wakefield et al. (2000)

argued that with the conditional approach, is often is unclear how to
choose the neighbourhood ?i (Eq. (3)
).
Several authors (Bernardinelli et al., 1997
,
Besag et al., 1991
,
Clayton and Kaldor, 1987
,
Richardson, 1992

and Richardson et al., 1995
),
have taken areas i and j to be neighbours if they share a common
boundary. This is right if all geographical units have a similar size
and are arranged in a regular pattern. According to these various
considerations, we used a very regular lattice based on 1264 hexagons
with a small size; each hexagon covered 445 km2 and represented only
0.08% of the overall area of France. We used also a simple
neighbourhood definition: a common boundary between hexagons. These
geographical units were homogenous in their area and boundary length, so
the neighbouring weight (strength of adjacency) naturally was fixed to
1 for each adjacency in the analysis.

The period considered, July 1, 2001 to July 31, 2003, allowed us to get
comparable data between regions, because the surveillance system has
been comprehensive during that period (including both clinical
surveillance and systematic testing on slaughtered and dead cattle). We
avoided the possible important bias due to geographical differences in
the awareness of farmers and veterinarians with the clinical
surveillance alone (Cuenot et al., 2003
).
However, following this method, we could not account for all the NAIF
and super-NAIF cases that occurred in France, so we cannot estimate
the overall incidence of BSE for these two categories; a large part of
the NAIF cases reached the clinical stage of the disease before 2001
(i.e. before the surveillance was comprehensive and fully efficient).
Besides, part of the super-NAIF cases did not occur yet at the end of
the period of interestespecially, those born between 1998 and 2000
(i.e. that did not reach the median age at clinical onset by July 2003).
The cases that moved from one farm to another during the first year of
life represented only 7.2% of the BSE cases. The type of production was
worth accounting for, because the crude ratio of dairy to beef BSE cases
(Table 1
)
was about 4 to 1; it did not vary between NAIF and super-NAIF cases
(366 dairy NAIF on a total of 445 and 48 dairy super-NAIF on a total
of 58).

The location of the BSE cases during the first year of life, detected
between July 1, 2001 and July 2003, was not distributed randomly in
France. Such a result is in agreement with the findings by Wilesmith et
al. (2000)

and Stevenson et al. (2000)

in Great Britain, Doherr et al. (2002)

in Switzerland and Abrial et al. (2003)

in western France. The unstructured variability (a Gaussian random
effect) was not significantly >0 (globally and for all individual
hexagon) and represented <1% of the overall variability. This is the
reason why this component was dropped of the model (Eq. (2)
)
and it showed that the choice of the simple neighbourhood (common
hexagon boundary) for the neighbourhood component was sufficient to
capture most of the variability.

We obtained a disease mapping of the risk of BSE separately for NAIF
and super-NAIF cases (Fig. 2
).
We emphasize that the trends can be compared between the two maps for
NAIF and super-NAIF casesbut that each map shows a geographical
relative risk compared to the average incidence for the separate
category. The maps do not provide an estimate of the exact level of the
BSE risks and hence cannot be compared between the two maps. The western
coast (where the first cases were detected) was the geographical area
the most at risk for NAIF cases, mostly in agreement with the results
obtained from a pilot study in western part of France carried out in
2000 (Abrial et al., 2003

and Morignat et al., 2002
).
Concerning the super-NAIF casesthose born after June 1996the areas
with a risk higher than average match approximately those of the NAIF
cases. Interestingly to us, however, the areas with the highest risk
were slightly different. This suggests that the area the most at risk
changed through time. If the disease mapping gives the general trend of
the BSE risk on the overall territory, the interest of Fig. 3

is to show precisely which ones of the geographical units had a BSE risk
significantly higher than the average national one (at p < 0.01). The
analysis of surveillance data has shown (Morignat et al., 2002

and La Bonnardiere et al., 2004
)
that the risk of infection decreased sharply in France for animals born
in 1995 and after. Our results suggest that the risk of BSE has been
comparatively better handled and reduced sooner in Bretagne for those
animals born after June 1996super NAIF casescompared to the other
regions, in contrast with the preceding period concerning the NAIF
cases. We do not know precisely when the animals get infected, but the
current assumptions drawn from modelling results (Donnelly, 2000

and Supervie and Costagliola, 2004
)
are that it is most of the time before the age of 2 years. If this is
correct, the change in the risk level between Bretagne and other
regions, comparing the NAIF and super NAIF cases, referred to the
years 1997 and after, when the super-NAIF animals submitted to the BSE
risk were between 1 and 2 years old. It might be useful to try to
identify more precisely when this shift between the regions the most at
risk did occur, and to try to relate this to differential control
measures between regions (either voluntary measures preceding the
regulation of June 1996 or a more or less efficient application of this
new regulation).

Indeed, the main hypothesis concerning the infection of NAIF animals
is the role of cross-contamination between monogastric feed and cattle
feed (BNEVS, 2000
,
Wilesmith et al., 2000

and Ducrot and Calavas, 2002
),
when MBM still were allowed for monogastrics (Morignat et al., 2002
).
That a spatial heterogeneity has been shown for the risk of BSE
concerning the NAIF cases, as well as the fact that we identified the
geographical areas with a significantly higher risk than average, will
allow us to test this hypothesis. Another hypothesis is the
contamination with the BSE agent of different by-products (such as bones
and tallow) used in the food industry. Should the data from the food
industry concerning the use of these by-products by location be
available, it might be possible to test this alternative hypothesis,
too. The situation for the super-NAIF is more complicated because they
were born after the implementation of reinforced measures to avoid the
contamination of the MBM with the BSE agent (mostly the removal of
cadavers and risk materials from the raw material used for the
processing of MBM). However, it was legal to import MBM from countries
where these measures had not been implemented, and the hypothesis of
contaminated by-products still holds for the super-NAIF. Our results
show that the super-NAIF cases were not randomly distributed, and that
the areas at risk for them matched different zones that were already at
risk for NAIF cases. These two elements are in favour of the same
sources of contamination for NAIF and super-NAIF cases. A temporal
analysis of the BSE cases might give more-precise results about when the
shift in the risk between regions occurred. Also, based on our results,
the next step will be to link the disease mapping to our assumptions
concerning the sources of infection for BSE, including the pig and
poultry demography, the use of MBM for monogastrics combined with
measures to control cross contamination in the feed industry, and the
use of by-products in cattle feed.


6. Conclusion

We have evidenced that the BSE cases were not randomly distributed in
France, both for NAIF and super-NAIF cases detected from July 1,
2001 to July 31, 2003. Furthermore, the areas at risk for the
super-NAIF matched part of the areas at risk for the NAIF cases,
which suggests that the source of contamination might be the same.
However, we observed a shift in the regions the most at risk for these
two categories of exposed cattle; the Bretagne (in western France) had
the highest risk for the NAIF cases but seems to have handled and
reduced the risk of risk sooner than other regions after additional
controls were mandated.


Acknowledgements

We thank the Direction Générale de lAlimentation that provided the data
of the surveillance programme of BSE, especially Daniel Lafon from the
Brigade Nationale dEnquêtes Vétérinaires et Phytosanitaires, as well
as Eric Morignat from the Agence Française de Sécurité Sanitaire des
Aliments.


References

Abrial et al., 2003

D. Abrial, D. Calavas, N. Lauvergne, E. Morignat and C. Ducrot,
Descriptive spatial analysis of BSE in western France, Vet. Res. 34
(2003), pp. 111.

Bernardinelli et al., 1997

L. Bernardinelli, C. Pascutto, N.G. Best and W.R. Gilks, Disease mapping
with errors in covariates, Stat. Med. 16 (1997), pp. 741752.

Besag, 1974

J. Besag, Spatial interaction and the statistical analysis of lattice
system, J. R. Statist. Soc. 2 (1974), pp. 192236.

Besag et al., 1991

J. Besag, J. York and A. Mollié, Bayesian image restoration, with two
applications in spatial statistics, Ann. Inst. Statist. Math 43 (1991)
(1), pp. 159.

BNEVS, 2000

BNEVS, 2000. Brigade nationale denquêtes vétérinaires et sanitaires.
Enquête épidémiologique relatives aux cas dESB survenus en France en
1999. URL: http://www.agriculture.gouv.fr/alim/sant/mala/esb.pdf.

Calavas and Ducrot, 2003

Calavas, D., Ducrot, C., 2003. LESB en France  Synthèse sur
lévolution de lépizootie à partir des données disponibles au 1er
janvier 2003. Agence Française de Sécurité Sanitaire des Aliments, URL:
http://www.afssa.fr.

Calavas et al., 2001

D. Calavas, C. Ducrot, T. Baron, E. Morignat, J.L. Vinard, A.G. Biacabe,
J.Y. Madec, A. Bencsik, S. Debeer and M. Eliazsewicz, Prevalence of BSE
on western France by screening cattle at risk: preliminary results of a
pilot study, Vet. Rec. 149 (2001), pp. 5556.

Clayton and Bernardinelli, 1992

D. Clayton and L. Bernardinelli, Bayesian methods for mapping disease
risk. In: P. Elliot, J. Cuzick, D. English and R. Stern, Editors,
Geographical and Environmental Epidemiology: Methods for Small-area
Studies, Oxford University, London (1992), pp. 205220.

Clayton and Kaldor, 1987

D. Clayton and J. Kaldor, Empirical bayes estimates of age-standardized
relative risks for use in disease mapping, Biometrics 43 (1987), pp.
671681.

Cressie, 1993

In: N.A.C. Cressie, Editor, Statistics for Spatial Data, Wiley, New York
(1993).

Cuenot et al., 2003

M. Cuenot, D. Calavas, D. Abrial, P. Gasqui, G. Cazeau and C. Ducrot,
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 (2003) (3), pp. 261272.

DGAL, 2003

DGAL, 2003. Information about BSE in France, URL:
http://www.agriculture.gouv.fr/esbinfo/esbinfo.htm.

Diggle et al., 1997

P.J. Diggle, P. Elliott, S. Morris and G. Shaddick, Regression modelling
of disease risk in relation to point sources, J. R. Stat. Soc. A. 160
(1997), pp. 491505.

Diggle, 2000

P.J. Diggle, Overview of statistical methods for disease mapping and its
relationship to cluster detection. In: P. Elliot, J. Wakefield, N. Best
and D. Briggs, Editors, Spatial Epidemiology, Oxford University (2000),
pp. 87103.

Doherr et al., 2002

M.G. Doherr, A.R. Hett, J. Rufenacht, A. Zurbriggen and D. Heim,
Geographical clustering of cases of bovine spongiform encephalopathy
(BSE) born in Switzerland after the feed ban, Vet. Rec. 151 (2002) (16),
pp. 467472.

Donnelly, 2000

C.A. Donnelly, Likely size of the French BSE epidemic, Nature 408
(2000), pp. 787788.

Ducrot and Calavas, 2002

C. Ducrot and D. Calavas, Epidémiologie de lencéphalopathie spongiforme
bovine et de la tremblante. In: B. Dodet and F. Cathala, Editors, Les
maladies humaines et animales à prions, Elsevier, Paris (2002), pp. 4760.

Ducrot et al., 2000

C. Ducrot, D. Calavas, T. Baron, A.E. Agrech, M. Coudert and M. Savey,
Epidemiological status of BSE in France update on born after the ban
cases, Eurosurveillance 5 (2000) (9), pp. 97100.

Ducrot et al., 2003

C. Ducrot, P. Roy, E. Morignat, T. Baron and D. Calavas, How the
surveillance system may bias the results of analytical epidemiological
studies on BSE: prevalence rate among dairy versus suckler cattle breeds
in France, Vet. Res. 34 (2003), pp. 185192.

Gilks et al., 1996

In: W.R. Gilks, S. Richardson and D.J. Spiegelhalter, Editors, Markov
Chain Monte Carlo in Practice, Chapman and Hall, London (1996).

Gouëllo, 1991

L. Gouëllo, Analyse épidémiologique du premier cas dencéphalopathie
spongiforme bovine en France, Epidémiologie et Santé Animale 19 (1991),
pp. 6370.

Heidelberger and Welch, 1983

P. Heidelberger and P.D. Welch, Simulation run length control in the
presence of an initial transient, Opns. Res. 31 (1983), pp. 11091144.

La Bonnardiere et al., 2004

C. La Bonnardiere, D. Calavas, D. Abrial, E. Morignat and C. Ducrot,
Estimating the trend of the French BSE epidemic over six birth cohorts
through the analysis of the abattoir screening in 2001 and 2002, Vet.
Res. 35 (2004) (3), pp. 299308.

Mollié, 1999

A. Mollié, Bayesian and empirical Bayes approaches to disease mapping.
In: A. Lawson, A. Biggeri and D. Bohning, Editors, Disease Mapping and
Risk Assessment, Wiley, Aberdeen (1999), pp. 1529.

Mollié, 2000

A. Mollié, Bayesian mapping of Hodgkin's disease in France. In: P.
Elliot, J. Wakefield, N. Best and D. Briggs, Editors, Spatial
Epidemiology, Oxford University (2000), pp. 266285.

Morignat et al., 2002

E. Morignat, C. Ducrot, P. Roy, T. Baron, J.L. Vinard, A.G. Biacabe,
J.Y. Madec, A. Bencsik, S. Debeer, M. Eliazsewicz and D. Calavas,
Targeted surveillance to assess of BSE in high risk populations in
western France and the associated risk factors, Vet. Rec. 151 (2002)
(3), pp. 7377.

Moynagh and Schimmel, 1999

J. Moynagh and H. Schimmel, Tests for BSE evaluated, Nature 400 (1999),
p. 105.

Richardson, 1992

S. Richardson, Statistic modelling of geographic and epidemiologic
variations, Rev. Epidemiol. Sante. Publique 40 (1992) (1), pp. 3345.

Richardson et al., 1995

S. Richardson, C. Montfort, M. Green, G. Drapper and C. Muirhead,
Spatial variation of natural radiation and childhood leukaemia incidence
in Great Britain, Stat. Med. 14 (1995), pp. 24872501.

Stevenson et al., 2000

M.A. Stevenson, J.W. Wilesmith, J.B.M. Ryan, R.S. Morris, D.U. Lawson,
D.U. Pfeiffer and D. Lin, Descriptive spatial analysis of the epidemic
of bovine spongiform encephalopathy in Great Britain to June, Vet. Rec.
147 (1997), pp. 379384.

Supervie and Costagliola, 2004

V. Supervie and D. Costagliola, The unrecognized french BSE epidemic,
Vet. Res. 35 (2004) (3), pp. 349362.

Wakefield et al., 2000

J.C. Wakefield, N.G. Best and L. Waller, Bayesian approaches to disease
mapping. In: P. Elliot, J. Wakefield, N. Best and D. Briggs, Editors,
Spatial Epidemiology, Oxford University, London (2000), pp. 87103.

Wilesmith et al., 2000

J.W. Wilesmith, J.B.M. Ryan, M.A. Stevenson, R.S. Morris, D.U. Pfeiffer,
D. Lin, R. Jackson and R.L. Sanson, Temporal aspects of the epidemic of
bovine spongiform encephalopathy in Great Britain: holding-associated
risk factors for the disease, Vet. Rec. 147 (2000), pp. 319325.

Wilesmith et al., 1988

J.W. Wilesmith, G.A.H. Wells, M.P. Cranwell and J.B.M. Ryan, Bovine
spongiform encephalopathy: epidemiological studies, Vet. Rec. 123 (1988)
(25), pp. 638644.


Corresponding Author Contact Information
Corresponding
author. Tel.: +33 473 624263; fax: +33 473 624548.


http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6TBK-4F2B7F0-2&_user=2035569&_handle=B-WA-A-W-Z-MsSAYVW-UUA-AAAUDEZDYB-AAUYBDDCYB-EAZEUAVZE-Z-U&_fmt=full&_coverDate=01%2F01%2F2005&_rdoc=6&_orig=browse&_srch=%23toc%235145%232005%23999329998%23566186!&_cdi=5145&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=2035569&md5=e31dd5df1b29c6af0d708d9cc3fb6ee5

TSS





Follow Ups:



Post a Followup

Name:
E-mail: (optional)
Subject:

Comments:

Optional Link URL:
Link Title:
Optional Image URL: