predicting west nile virus seroprevalence in wild birds in senegal
TRANSCRIPT
VECTOR-BORNE AND ZOONOTIC DISEASESVolume 9, Number 6, 2009© Mary Ann Liebert, Inc.DOI: 10.1089/vbz.2008.0130
Predicting West Nile Virus Seroprevalence in Wild Birds in Senegal
Veronique Chevalier,1 Pierre Reynaud,2 Thierry Lefrançois,3 Benoit Durand,4
Francois Baillon,5 Gilles Balança,1 Nicolas Gaidet,1 Bernard Mondet,5 and Renaud Lancelot6
Abstract
West Nile fever epidemiology is complex, and the role of birds in the maintenance, amplification, and dissemi-nation of the West Nile virus (WNV) remains partially unknown. In 2003, a serological study was performed inSenegal, where West Nile infection is considered endemic. The goal was to identify potential reservoirs of WNVamong bird species present in the Ferlo area (northern Senegal) and the Senegal River Valley, and to screen theecological factors possibly related to West Nile infection. Serological data were analyzed using a generalized lin-ear model. Statistical association between ecological factors and the risk of infection were then modeled to de-rive a species-specific risk. A cross-validation was conducted. The overall observed prevalence rate was 5.5%(n � 422). Thirteen bird species were found positive, from which five were migrating: Lanius senator, Anthus triv-ialis, Hippolais opaca, Jynx torquilla, and Cercotrichas galactotes. The nesting type in resident birds was positivelycorrelated with the risk of infection (odds ratio [OR] � 11, p 5 0.0003); the gregariousness level of birds appearedas a protective factor (OR � 0.3, p � 0.01). The predicted prevalence varied between 1% and 39% for residentspecies and between 1% and 7% for migrating species. Results of model internal validation were satisfactory atthe individual and species level. However, more field and experimental investigations are needed to confirmthese preliminary results and help target the future research and surveillance in Senegal.
Key Words: West Nile virus—Wild avifauna—Senegal—Risk factors—Prediction.
589
Introduction
West Nile fever is an emerging arbovirosis caused bythe West Nile virus (WNV) (Flavivirus, Flaviviridae).
The epidemiological cycle involves wild birds and Culicidaemosquitoes (mainly Culex genus). Horse and man are dead-end hosts but the disease is a veterinary and human public-health issue. The role of birds in the maintenance, amplifica-tion, and dissemination of the virus remains partiallyunknown. Recent studies suggested that the bird communitystructure, diversity, and presence of particular species maydetermine the mosquito infection rate (Kilpatrick et al. 2006).Nestling as well as hatch-year birds may be important factorsof amplification (Scott and Edman 1991, Hamer et al. 2008).
In Senegal, the WNV was isolated several times from Aedesand Culex mosquitoes in the Senegal River Valley (Traore-
Lamizana et al. 1994). Different serological surveys, under-taken in the Ferlo area (northern Senegal) in humans (Mur-gue et al. 2002), horses (Chevalier et al. 2006), and sentinelchickens (Chevalier et al. 2007), suggested that the WNV trans-mission regularly and intensively occurred in this country.
In Senegal, the avian biodiversity is high, especially in theSenegal River Valley, which is a major wintering area for Eu-ropean migratory birds.
To target future research and surveillance, we sought toidentify ecological factors linked to the risk of West Nile in-fection. A serological survey was conducted to estimate theserological point prevalence of anti-WNV antibodies in res-ident and local or Palearctic migratory bird species. Statisti-cal association between ecological factors and the risk of in-fection was then evaluated and modeled to derive aspecies-specific risk.
1International Centre of Research in Agronomy for Development (CIRAD, UR AGIRs), 34398 Montpellier Cedex 5, France.2Institut de Recherche pour le Développement (IRD), Département Société et Santé, les Orteaux, 05700, Trescléoux, France.3International Centre of Research in Agronomy for Development (CIRAD, UMR 15), Domaine de Duclos, Prise d’eau, 97170 Petit Bourg,
France.4Agence Française de Sécurité Sanitaire des Aliments (AFSSA), 94706 Maisons-Alfort Cedex, France.5Institut de Recherche pour le Développement (IRD), UR178- Conditions et Territoires d’Emergence des Maladies (CTEM), Unité des
Virus Emergents, Faculté de Médecine de Marseille, 13005 Marseille Cedex 05, France.6International Center of Research in Agronomy, (CIRAD, UMR 15), Montpellier Cedex 5, France.
Material and Methods
Study areas
Two study sites were selected based on previous studies(serology) and expert knowledge (ornithology). The first sitewas in the close vicinity of the Barkedji village, in the Ferloarea (Fig. 1). The Ferlo is a Sahelian region characterized bytemporary ponds that fill up in July and remain flooded un-til November–December. When flooded, ponds constitute afavorable biotope for Aedes and Culex mosquitoes. Manyspecies of endemic birds nest around these ponds, and mi-gratory birds are attracted by the water and abundant food(Morel and Morel 1978).
The second site was located in the Djoud’j National Park(PNOD; Fig. 1). The PNOD is a large wetland located in theSenegal River delta, made of a large lake surrounded bystreams, ponds, and backwaters.
Bird trapping
Birds nesting in Barkedji leave in December when pondsdry up. Birds were trapped from September 28 to October 7,2003 in Barkedji to guarantee the trapping of both residentand migrating birds. The trapping period was similar in thePNOD. Ground-level mist nets (12 � 2.8 m, 36-mm mesh)were used to focus on passerine birds, which are supposedto play a prominent role in the epidemiological cycle of WNV(Komar et al. 2003). Nets were operated from sunrise to 11A.M. and from 5 p.m. to sunset. Birds were sampled from
the wing or jugular vein. Sera were centrifuged and storedat 4°C until they were transported to Dakar.
Serological analysis
Sera were analyzed using an epitope-blocking enzyme-linked immunosorbent assay using the antigen used byBlitvich et al. (2003).
This test was performed using the WNV-specific mono-clonal antibody (MAb) 3.1112G (Chemicon, Rosemount, IL).The original technique was slightly modified: WNV antigen(100 mL, 1/4000 in 50 mM NaHCO3, 50 mM Na2CO3, pH 9.6)was coated overnight at 4°C, followed by 1 hour at 37°C, onMaxisorp 96-well plates (NUNC, Thermo Fisher Scientific,Roskilde, Denmark). Plates were washed four times withphosphate-buffered saline (PBS; pH 7.4) containing 0.1%Tween 20, blocked with 200 mL of blocking buffer (PBS con-taining 0.1% Tween 20 and 0.2% bovine serum albumin) at37°C for 45 minutes, and washed again. Fifty microliters ofsera diluted 1/10 in blocking buffer were added to each well,and incubated at 37°C for 2 hours. After four washes, 50 mLof MAb, diluted 1/4000 in blocking buffer, was added toeach well and incubated for 1 hour at 37°C, followed by 1hour at 4°C. After four more washes, 50 mL of horseradishperoxidase labeled rabbit anti mouse immunoglobulin G(Serotec; Oxford, UK), diluted 1/500 in blocking buffer, wasadded to each well and incubated for 1 hour at 37°C. Afterfour washes, 200 mL of tetramethylbenzidine substrate(Sigma Aldrich, St. Louis, MO) were added. After 30 min-
CHEVALIER ET AL.590
FIG. 1. Location of the Barkedji village and the Djoud’j National Park (PNOD) in Senegal.
utes, optical densities were measured at 630 nm. The abilityof the test sera to block the binding of the MAbs to WNVantigen was compared to the blocking ability of chickenserum without antibody to WNV (Vector Laboratories,Burlingame, CA). Data were expressed as relative percent-ages. An inhibition value �30% was considered to indicatethe presence of viral antibodies.
Data analysis
Seven factors possibly related with the exposition of birdsto mosquito bites were included as explanatory variables: thetrapping location (Barkedji, PNOD), the migratory status (res-ident in Africa, or migratory breeding outside Africa), thefeeding behavior (flying or settled), the resting site (ground,bush, canopy), the type of nest (on the ground, platform, cupor cavity nest—so-called “medium, or bulky” nest), the herdinstinct level (high, low), and the affinity with urban areas(high, moderate, low). Among these variables, the nest typewas treated specifically because of its particular meaning: asthe exposure level of birds during the breeding season de-pends on nesting location, the nest type effect cannot be dis-sociated from the nesting site effect (i.e., the migrating sta-tus). Therefore in the analysis, the nesting type variable wasstudied separately in resident and in migratory birds.
A bivariate analysis using Fisher exact test was performedto assess the relation between the observed prevalence rateand each variable. Variables associated with serologicalpoint prevalence (p value � 0.20) were selected for inclusionin a logistic regression model where the individual serolog-ical status (positive or negative serum) was the response vari-able. For the reasons detailed above, the nest type variablewas not included alone but only in interaction with the mi-grating status variable. A stepwise procedure was performedto select the best model according to the Akaike informationcriterion (Burnham and Anderson 2002). Odds ratios (ORs)and their confidence intervals (CIs) were computed for theexplanatory variables of the resulting model, and the modelwas used to predict the prevalence rate and its CI for eachcombination of ecological characteristics corresponding tothe tested species.
A cross-validation was conducted (Stone 1974): half of thetested birds were randomly chosen for fitting the model,which was then used to predict the serological status of theremaining birds. Prediction error was evaluated at the indi-vidual level and at the species level. Individual-level errorwas the usual proportion of incorrect predictions. Forspecies-level cross-validation, two prevalence classes wereconsidered to distinguish low prevalence species (�10%)and high prevalence species (�10%). Prediction error wasthe proportion of misclassified species, weighted by thenumber of tested animals.
Results
Serological results
A total of 422 birds were sampled, belonging to 16 fami-lies and 49 species. One hundred seventy birds were trappedin Barkedji, belonging to 8 families and 16 species, most ofthem being resident (89%). In the PNOD, 252 birds were sam-pled, belonging to 13 families and 33 species. Most of themwere migratory (80%). The most frequently trapped birds be-
longed to the Muscicapidae family (29%, n � 122), followedby the Ploceidae (24%, n � 100), the Columbidae (12%, n � 51),and the Alcedinidae (8%, n � 35).
The overall observed prevalence rate was 5.5% (n � 422).Anti-WNV antibodies were detected in 23 birds from 7 fam-ilies and 13 species, among which five were migrating:Woodchat Shrike (Lanius senator), Tree Pipit (Anthus trivialis),Western Olivaceus Warbler (Hippolais opaca), Eurasian Wry-neck (Jynx torquilla), and Rufous Scrub-robin (Cercotrichasgalactotes).
For 19 species, the number of trapped birds was �5, rep-resenting 87% of the totality of tested birds (n � 369). Amongthese species, the prevalence rate was �10% for each of thethree Columbidae species, as well as in two Muscicapidaespecies. The prevalence rate was �10% for H. opaca and Plo-ceus cucullarus. No positive bird was found for the 13 otherspecies with �5 trapped birds (Table 1).
The 13 remaining species were weakly represented, with�5 trapped animals per species. Positive birds were foundin six of these species (Table 2): both tested Laniidae species,Urocolius macrourus (Coliidae), Jynx torquilla (Picidae), Ploceusvelatus (Ploceidae), and Anthus trivialis (Motacillidae).
Bivariate analysis
Prevalence rate was higher in resident (8.4%, n � 203) thanin migrating birds (2.7%, n � 219; p � 0.02), whereas theserates were similar for birds trapped in the PNOD (4.8%, n �252) and in Barkedji (6.5%, n � 170; p � 0.51).
The prevalence rate was higher in species showing a lowherd instinct (11.5%, n � 78) than in gregarious species (4.1%,n � 344; p � 0.02).
In resident birds, a high prevalence rate was observed inspecies building medium nests (19.2%, n � 73), whereas lowrates were observed for the two other types: 2.9% for birdsbuilding bulky nests (n � 102) and 0.0% for birds nesting onthe ground (n � 28; p � 0.0001). In migratory birds, nestingtype was not associated with prevalence rate; none of the mi-gratory birds belonged to species building bulky nests;prevalence rate was similar for the two other nesting types:medium nests (3.6%, n � 138) and nests built on the ground(1.2%, n � 81; p � 0.41).
Birds feeding on flight were as frequently infected (4.8%,n � 167) as birds feeding settled, in trees, or on the ground(5.9%, n � 255; p � 0.67). Prevalence rate was high in birdsresting on the ground (10.2%, n � 88), with lower values be-ing observed in birds resting in bush (4%, n � 125) or in thecanopy (4.3%, n � 209; p � 0.11). Last, affinity with urban ar-eas was not associated with the prevalence rate, ranging from3.7% (n � 107) for birds living in villages, 8% (n � 160) forperidomestic birds, and 3.9% (n � 154) for birds living farfrom urban areas (p � 0.25).
Logistic model
The explanatory variables selected for inclusion in the lo-gistic regression model were: the migration status, the herdinstinct, the resting site, and the interaction between the mi-gration status and the nest type. As none of the migratorybirds belonged to a species building bulky nests and as sim-ilar prevalence rates had been observed in species buildingbulky nests and in species nesting on the ground, the nesttype variable was recoded in two categories: medium nests
WEST NILE INFECTION ON WILD BIRDS IN SENEGAL 591
Ta
ble 1.
Ec
olo
gic
al C
ha
ra
cter
istic
s a
nd
O
bser
ved
S
er
olo
gic
al P
rev
alen
ce o
f W
est N
ile F
ev
er
in
19
Bir
d S
pec
ies
of W
hic
h �
5In
div
id
ua
ls W
er
e T
ested
in
T
wo
D
iffer
en
t A
rea
s (F
er
lo
, S
en
eg
al R
iv
er
B
asin
) o
f S
en
eg
al in
20
03
Pre
dict
edM
igra
ting
Nes
tR
esti
ngH
erd
Feed
ing
Urb
anO
bser
ved
prev
alen
ceFa
mily
Spec
ies
stat
usa
type
bsi
tec
inst
inct
dbe
havi
ore
affin
ityf
prev
alen
ce(%
)95
% C
Ig
Alc
edin
idae
Alc
edo
cris
tata
RG
BL
FL
0/17
�6
0–0.
03H
alcy
on s
eneg
alen
sis
RM
CL
FL
0/15
�7
0.00
6–0.
14C
hara
driid
aeC
alid
ris
min
uta
MG
GH
SL
0/9
�1
0–0.
03A
ctit
is h
ypol
euco
sM
GG
HS
L0/
13�
10–
0.03
Col
umbi
dae
Oen
a ca
pens
isR
MG
HS
M2/
915
0.06
8–0.
24St
rept
opel
ia s
eneg
alen
sis
RM
CH
SH
2/8
150.
068–
0.24
Stre
ptop
elia
vin
acea
RM
BH
SM
4/31
150.
068–
0.24
Hir
undi
nida
eR
ipar
ia r
ipar
iaM
GC
HF
L0/
13�
10–
0.03
Mer
opid
aeM
erop
s pe
rsic
usM
GC
HF
L0/
21�
10–
0.03
Mot
acill
idae
Mot
acill
a al
baM
MG
LS
M0/
6�
70.
006–
0.14
Mot
acill
a fla
vaM
GG
HS
L0/
18�
10–
0.03
Mus
cica
pida
eH
ippo
lais
opa
caM
MB
HF
L1/
172
0–0.
04P
hyllo
scop
us t
roch
ilus
MM
BH
FL
0/45
�2
0–0.
04Sy
lvia
can
tilla
nsM
MB
HF
M0/
29�
20–
0.04
Cer
cotr
icha
s ga
lact
otes
MM
BL
SL
1/6
70.
006–
0.14
Cer
cotr
icha
s po
dobe
RM
BL
SM
4/9
390.
17–0
.61
Pho
enic
urus
pho
enic
urus
MM
BL
SM
0/5
�7
0.00
6–0.
14P
loce
idae
Plo
ceus
cuc
ulla
tus
RB
CH
SH
2/53
20–
0.04
Plo
ceus
mel
anoc
epha
lus
RB
CH
SH
0/45
20–
0.04
Prev
alen
ce is
pre
dic
ted
for
the
corr
espo
ndin
g co
mbi
nati
ons
of h
erd
inst
inct
, nes
t typ
e an
d m
igra
tion
sta
tus;
con
fid
ence
inte
rval
s (C
Is) r
efer
to th
ese
com
bina
tion
s.a R
, res
iden
t in
Sene
gal;
M, m
igra
tory
, bre
edin
g ou
tsid
e Se
nega
l.b B
, bul
ky n
est;
M, m
ediu
m (p
latf
orm
, cup
or
cavi
ty) n
est,
G, n
est b
uilt
on
the
grou
nd.
c G, g
roun
d; B
, bus
h; C
, can
opy.
dL
, low
; H, h
igh.
e S, s
ettl
ed; F
, fly
ing.
f L, l
ow; M
, mod
erat
e; H
, hig
h.g C
I of t
he p
red
icte
d p
reva
lenc
e.
Ta
ble 2.
Ec
olo
gic
al C
ha
ra
cter
istic
s a
nd
O
bser
ved
S
er
olo
gic
al P
rev
alen
ce o
f W
est N
ile F
ev
er
in
30
Bir
d S
pec
ies
of W
hic
h �
5In
div
id
ua
ls W
er
e T
ested
in
T
wo
D
iffer
en
t A
rea
s (F
er
lo
, S
en
eg
al R
iv
er
B
asin
) o
f S
en
eg
al in
20
03
Pre
dict
edM
igra
ting
Nes
tR
esti
ngH
erd
Feed
ing
Urb
anO
bser
ved
prev
alen
ceFa
mily
Spec
ies
stat
usa
type
bsi
tec
inst
inct
dbe
havi
ore
affin
ityf
prev
alen
ce(%
)95
% C
Ig
Ala
udid
aeG
aler
ida
cris
tata
RG
GL
SM
0/1
�6
0–0.
12A
lced
inid
aeH
alcy
on l
euco
ceph
ala
MG
CL
FL
0/3
�4
0–0.
12B
ucer
otid
aeT
ocku
s er
ythr
orhy
nchu
sR
MC
HS
L0/
1�
150.
068–
0.24
Cap
rim
ulgi
dae
Cap
rim
ulgu
s cl
imac
urus
RG
GL
FL
0/2
�6
0–0.
12C
hara
driid
aeC
alid
ris
alpi
naM
GG
HS
L0/
1�
10–
0.03
Cha
radr
ius
hiat
icul
aM
GG
HS
L0/
1�
10–
0.03
Ros
trat
ula
beng
alen
sis
RG
GL
SL
0/1
�6
0–0.
12T
ring
a oc
hrop
usM
MG
LS
L0/
1�
70.
006–
0.14
Van
ellu
s sp
inos
usR
GG
HS
L0/
2�
10–
0.04
Col
iidae
Uro
coliu
s m
acro
urus
RM
CH
FL
1/3
150.
068–
0.24
Col
umbi
dae
Col
umba
gui
nea
RM
CH
SH
0/2
�15
0.06
8–0.
24T
urtu
r ab
yssi
nicu
sR
MB
HS
M0/
1�
150.
068–
0.24
Cor
aciid
aeC
orac
ias
abys
sini
cus
RM
CH
FL
0/1
�15
0.06
8–0.
24La
niid
aeLa
niar
ius
barb
arus
RM
BL
SM
1/1
390.
17–0
.61
Lani
us s
enat
orM
MB
LS
L1/
37
0.00
6–0.
14La
rida
eLa
rus
cirr
ocep
halu
sR
GG
HS
M0/
1�
20–
0.04
Mer
opid
aeM
erop
s pu
sillu
sR
GC
HF
M0/
4�
10–
0.04
Mot
acill
idae
Ant
hus
triv
ialis
MG
GH
SL
1/2
10–
0.03
Mus
cica
pida
eT
erps
ipho
ne v
irid
isR
MC
HF
L0/
1�
150.
068–
0.24
Cam
arop
tera
bra
chyu
raR
BB
LS
M0/
2�
60–
0.12
Hip
pola
is p
olyg
lott
aM
MB
HS
L0/
2�
20–
0.04
Sylv
ia a
tric
apill
aM
MB
HS
M0/
4�
20–
0.04
Sylv
ia b
orin
MM
BH
SM
0/1
�2
0–0.
04Sy
lvia
hor
tens
isM
MB
LS
M0/
1�
70.
006–
0.14
Pic
idae
Jynx
tor
quill
aM
MB
LS
L2/
37
0.00
6–0.
14M
esop
icos
goe
rtae
RM
CL
FL
0/2
�39
0.17
–0.6
1P
loce
idae
Plo
ceus
lut
eolu
sR
BB
HS
M0/
1�
20–
0.04
Plo
ceus
vel
atus
RB
BH
SM
1/1
20–
0.04
Stur
nida
eLa
mpr
otor
nis
caud
atus
RM
CH
SM
0/1
�15
0.06
8–0.
24La
mpr
otor
nis
pulc
her
RM
GH
SM
0/3
�15
0.06
8–0.
24
Prev
alen
ce is
pre
dic
ted
for
the
corr
espo
ndin
g co
mbi
nati
ons
of h
erd
inst
inct
, nes
t typ
e an
d m
igra
tion
sta
tus;
con
fid
ence
inte
rval
s (C
Is) r
efer
to th
ese
com
bina
tion
s.a R
, res
iden
t in
Sene
gal;
M, m
igra
tory
, bre
edin
g ou
tsid
e Se
nega
l.b B
, bul
ky n
est;
M, m
ediu
m (p
latf
orm
, cup
or
cavi
ty) n
est,
G, n
est b
uilt
on
the
grou
nd.
c G, g
roun
d; B
, bus
h; C
, can
opy.
dL
, low
; H, h
igh.
e S, s
ettl
ed; F
, fly
ing.
f L, l
ow; M
, mod
erat
e; H
, hig
h.g C
I of t
he p
red
icte
d p
reva
lenc
e.
and other nest types (bulky nests and nests built on theground). The stepwise model selection procedure led to theexclusion of the resting site variable, and the resulting modelthus included the migration status and its interaction withthe nest type, and the herd instinct (Table 3). The largest ef-fect was attributed to the nest type in resident birds with anOR of 11.0 (p � 0.0003) for medium nests (reference: bulkynests or nests built on soil), whereas the nesting type effectwas not significant in migratory birds. High herd instinct(reference: low) had a protective effect with OR of 0.3 (p �0.01).
Predicted prevalence varied between 1% and 39% for res-ident species and between 1% and 7% for migratory species.This predicted prevalence (and its CI) refers to combinationsof ecological characteristics (herd instinct and interaction be-tween nest type and migration status); therefore, speciessharing the same ecological characteristics have identicalpredicted prevalence and CIs (Tables 1 and 2). Cross-vali-dation procedure yielded an individual prediction error of5%: when half of the data set was left out when fitting themodel, predicted serological status for the other half was cor-rectly predicted for 95% of the birds. At the species level,prediction error was 11%: 89% of the species were correctlyclassified either as low prevalence (�10%) species or as highprevalence (�10%) species.
When considering the 19 species for which �5 birds weretrapped (Table 1), the predicted status (low versus highprevalence) was correct in all species. High prevalence sta-tus was predicted (and observed) for four resident speciesand none of the migratory species. Low prevalence statuswas predicted for 11 migratory species and four residentspecies.
Model predictions were correct for two-thirds of the 30weakly represented species (�5 tested birds; Table 2). Ahigh-prevalence status was predicted for 10 resident species,among which, Urocolius macrourus and Laniarius barbaruswere found positive with prevalence figures of 1/3 and 1/1,respectively. Conversely, the model predicted a low preva-lence level for nine resident species and for 11 migratoryspecies, positive results being observed in one residentspecies (Ploceus velatus: 1/1) and in three migratory species(Anthus trivialis: 1/2, Lanius senator: 1/3, and Jynx torquilla:2/3).
Discussion
The overall serological prevalence rate (5.5%) confirmedthat WNV was transmitted to wild birds in Senegal either inthe Ferlo or in the Senegal River basin. Senegal, where the
WNV circulation is endemic, is also a major area of winter-ing for Palearctic migrating birds. To explain the WNV pres-ence in Palearctic areas, current epidemiological knowledgeassumes that the virus may be regularly transported fromAfrica to Europe by migrating birds (Malkinson and Banet2002). The higher prevalence rate observed in our study inresident birds (8.4%)—compared with migratory species(3%)—is consistent with this assumption: the exposure to thevirus was longer in the former than in the latter species.
In migratory birds, the prevalence rate of 2.7% was closeto observations made in birds sampled in southern Francein 2004 possibly coming from sub-Saharan Africa (Jourdain2006). Birds’ serological status is not related to their epi-demiological role in WNV transmission. A large number ofbirds should be ringed in Senegal and trapped and tested forviral presence when arriving in Europe to accurately assessthe risk of introduction of the virus in Europe by these birds.
Large variations of the observed point prevalence are re-ported in the literature. In the West Indies in 2002, it rangedfrom 3.3% in Jamaica (n � 542; Dupuis et al. 2003) to 15%(n � 33) in Dominican Republic (Komar et al. 2003). In theAfrican continent, point serological prevalence rates was40% in the Nile delta, Egypt, in 1950 (n � 420; Taylor et al.1956) and 12.5% in South Africa in 1962–1965 (Jupp 2001). Inthe Middle East, seroprevalence rate was 11.5% and 11.3%on resident birds in Israel in 1965–1966 (Nir et al. 1969, 1972).However, comparisons are difficult because the ecosystemsand epidemiological situations differ. Within the same area,large differences were also observed between species: inEgypt, the point prevalence was 13% in domestic chicken(n � 15) and 65% in hooded crows (n � 163; Taylor et al.1956). However, our results corroborate some results of pre-vious studies: the high point prevalence observed in theColumbidae family was reported in Israel (Nir et al. 1969,1972); Woodchat Strike (Lanius senator) was identified as can-didate species for virus introduction from sub-SaharanAfrica to Europe (Jourdain 2006).
More information is needed on the biology of the sampledspecies. As a matter of fact, seroprevalence is an indirect in-dicator of the exposure since other factors may influence itsvalue. The high prevalence rate in Streptopelia species maybe explained by the fact that they live in, or close to, urbanareas, which are preferential habitats for Culex mosquitoes,but also by a greater longevity compared with birds of otherfamilies. Bird age may be a confounding factor to study theexposure level using serological data. Juvenile and adultbirds can be easily distinguished. However, this distinctionis not accurate enough to take into account the age as a con-founding factor. Due to a lack of field method to precisely
CHEVALIER ET AL.594
Table 3. Results of the Logistic Model Selected by the Stepwise Regression Procedure
Variable Reference Value OR (95% CI) p valuea
Migration status Migrating Resident 1.4 (0.2–29.6) 0.8Nest type: migrating birds Bulky or soil nest Medium nest 1.9 (0.3–37.9) 0.6Nest type: resident birds Bulky or soil nest Medium nest 11.0 (3.6–49.8) 0.0003Herd instinct Low High 0.3 (0.1–08) 0.01Intercept 0.006 (0.0003–0.03) �0.0001
OR, odds ratio; CI, confidence interval.aWald test.
estimate bird ages, this factor could not be considered in thisstudy.
The most important risk factor observed in this study wasthe interaction between migration status and nest type. Aspreviously suggested, nesting could be a predominant pe-riod for WNV infection: the limited mobility of adults andfledglings, and the bare skin of the latter, presumably makethem easy prey for mosquitoes (Marra et al. 2004). The nest-ing type would then be an important factor to explain birdexposure to mosquito bites. In resident birds, high preva-lence was associated with medium nests that include plat-form, cupped, and cavity nest. Platform nests are rudimen-tary, flat structures located on the ground, in a tree, or onthe tops of rooted vegetation or debris in shallow water.Cupped nests are an arrangement of hard and soft materialfrom floor and walls shaped in a cup. Cavity nest is a hol-lowed-up opening in the trunk of a tree. These three kindsof nests are located in the midstory and provide a large ac-cess to mosquitoes. Ground nests are usually just scrapes onthe ground forming a depression. Many mosquito’ speciesexhibit vertical height specialization for host seeking(Balenghien et al. 2006): birds building nests on the groundmay be less exposed than birds building medium nests. Thestructure of bulky nests is complex and compact. They aremade of twigs or grass weave or not with a grass-lined cham-ber inside. Access to outside is rather reduced; thus, the ex-posure to mosquito bites of birds building these kinds of nestis probably weak.
The increased point prevalence in nongregarious birdsmight be explained by a dilution effect previously suggestedin horses (Durand et al. 2002).
Results of model internal validation were satisfactory at theindividual and species level. Considering two prevalenceclasses (below or above 10%), the model adequately fitted thestatus of each species with �5 trapped birds, thus giving someconfidence in model predictions. However, an external vali-dation remains to be performed using independent data. Thehigher predicted prevalence in migratory birds was 7%, inparticular for White Wagtail (Motacilla alba), the Common Red-start (Phoenicurus phoenicurus), and the Rufous Scrub-Robin(Cercotrichas galactotes). Previous studies in 1965–1967 in Israelsuggested that the White Wagtail was exposed to the WNVand might be a relevant indicator of WNV circulation for sur-veillance (Nir et al. 1969, 1972). The Common Redstart wasreported as having WNV antibodies in another study in south-ern France in 2004 (Jourdain 2006). To our knowledge, no in-fection of Rufous Scrub-Robin by WNV was reported in theliterature. However, our analysis provides evidence that thesemigratory species are possibly exposed to WNV and mightplay a role in the dissemination of WNV from Africa to Eu-rope. Finally, high prevalence rates were predicted for threeresident species: the Black Shrub-Robin (Cercotrichas podobe),the Common Gonolek (Laniarius barbarus), and a woodpeckerspecies (Mesopicos goertae). Nothing is known about the sus-ceptibility to the WNV and the reservoir competence of thesetwo species. More field and biological data are needed to con-firm these preliminary results.
Acknowledgments
This publication is catalogued by the EDEN Steering Com-mittee as EDEN00116 (www.edenfp6project. net). The con-
tents of this publication are the sole responsibility of the au-thors and do not necessarily reflect the views of the Euro-pean Commission.
Disclosure Statement
This publication was partially funded by the CORUS proj-ect (French Ministry of Foreign Affairs) and GOCE-2003-010284 EDEN grants.
References
Balenghien, T, Fouque, F, Sabatier, P, Bicout, DJ. Horse, bird andhuman-seeking behaviour and seasonal abundance of mos-quitoes in a West Nile focus of southern France. J Med Ento-mol 2006; 43: 936–946.
Blitvich, BJ, Marlenee, NL, Hall, RA, Calisher, CH, et al. Epi-tope-blocking enzyme-linked immunosorbent assays for thedetection of serum antibodies to West Nile virus in multipleavian species. J Clin Microbiol 2003; 41:1041–1047.
Burnham, KP and Anderson, DR. Model selection and multi-model inference: a practical information -theoretic approach.New-York: Sringer-Verlag; 2002.
Chevalier, V, Lancelot, R, Diaité, A, Mondet, B, et al. Serologi-cal assessment of West Nile fever virus activity in the pastoralsystem of Ferlo (Senegal). Ann N Y Acad Sci 2006;10801:216–225.
Chevalier, V, Lancelot, R, Diaïté, A, Mondet, B, et al. Use of sen-tinel chickens to study the
transmission dynamics of West Nile virus in a Sahelian ecosys-tem. Epidemiol Infect 2007; 136:1–4.
Dupuis, AP, Marra, PP, Kramer, LD. Serologic evidence of WestNile virus transmission, Jamaica, West Indies. Emerg InfectDis 2003; 9:860–863.
Durand, B, Chevalier, V, Pouillot, R, Labie, J, et al. West NileVirus outbreak in horses, Southern France, 2000: results of aserosurvey. Emerg Infect Dis 2002; 8:777–7782.
Hamer, GL, Walker, ED, Brawn, JD, Loss, SR, et al. Rapid am-plification of West Nile virus: the role of hatch-year birds. Vec-tor Borne Zoonotic Dis 2008; 8:57–67.
Jourdain, E. Oiseaux sauvages et virus West Nile: étude éco-épidémiologique en Camargue.Grenoble I. Grenoble: Univer-sité Joseph Fourier; 2006.
Jupp, PG. The ecology of West Nile virus in South Africa andthe occurrence of outbreaks in humans. Ann N Y Acad Sci2001; 951:143–152.
Kilpatrick, AM, Kramer, LD, Jones, MJ, Marra, PP, et al. WestNile virus epidemics in North America are driven by shifts inmosquito feeding behaviour. PLoS Biol 2006; 4:606–610.
Komar, N, Langevin, SA, Hinten, SR, Nemeth, N, et al. Experi-mental infection of North American birds with the New-York1999 strain of West Nile virus. Emerg Infect Dis 2003;9:311–322.
Komar, O, Robbins, MB, Klenk, K, Blitvitch, BJ, et al. West Nilevirus transmission in resident birds, Dominican Republic.Emerg Infect Dis 2003; 9:1299–1302.
Malkinson, M and Banet, C. The role of birds in the ecology ofWest Nile virus in Europe and Africa. Curr Top Microbiol Im-munol 2002; 267:309–322.
Marra, PP, Griffing, S, Caffrey, C, Kilpatrick, M, et al. West Nilevirus and wild life. Bioscience 2004; 54:393–402.
Morel, GJ, Morel, MY. Recherches écologiques sur une savanesahélienne du Ferlo septentrionnal, Sénégal. Etude d’une com-munauté avienne. Cahiers de l’ORSTOM. Série Biologie 1978;13:3–34.
WEST NILE INFECTION ON WILD BIRDS IN SENEGAL 595
Murgue, B, Zeller, H, Deubel, V. The ecology and epidemiologyof West Nile Virus in Africa, Europe and Asia. Curr Top Mi-crobiol Immunol 2002; 267:195–221.
Nir, Y, Avivi, A, Lasowski, Y, Margalit, J, et al. Arbovirus ac-tivity in Israel. Isr J Med Sci 1972; 8:1695–1701.
Nir, Y, Lasowski, Y, Avivi, A, Goldwasser, R. Survey for anti-bodies to arboviruses in the serum of various animals in Is-rael during 1965–1966. Am J Trop Med Hyg 1969; 18:416–422.
Scott, TW, Edman, JD. Effects of avian host age and arbovirusinfection on mosquito attraction and blood-feeding success.In Loye, J, Zuk, M, eds. Bird-parasite interactions: Ecology,Evolution, and Behaviour. New York: Oxford UniversityPress; 1991:179–204.
Stone, M. Cross-validation choice and assessment of statisticalpredictions (with Discussion). Journal of the Royal StatisticalSociety B 1974; 36:111–147.
Taylor, RM, Work, TH, Hurlbut, HS, Rizk, F. A study of the ecol-ogy of West Nile virus in Egypt. Am J Trop Med Hyg 1956;5:867–869.
Traore-Lamizana, M, Zeller, HG, Mondo, M, Hervy, JP, et al.Isolations of West Nile and Bagaza viruses from mosquitoes(Diptera: Culicidae) in central Senegal (Ferlo). J Med Entomol1994; 31:934–938.
Address correspondence to:Dr. Veronique Chevalier
CIRADCampus International de Baillarguet
Montpellier 34398France
E-mail: [email protected]
CHEVALIER ET AL.596