References
1 Xu KM, Li KS, Smith GJ et al. Evolution and molecular epidemiology
of H9N2 influenza A viruses from quail in southern China, 2000 to
2005. J Virol 2007; 81:2635–2645.
2 Xu KM, Smith GJ, Bahl J et al. The genesis and evolution of H9N2
influenza viruses in poultry from southern China, 2000 to 2005. J
Virol 2007; 81:10389–10401.
3 Peiris M, Yuen KY, Leung CW et al. Human infection with influenza
H9N2. Lancet 1999; 354:916–917.
4 Butt KM, Smith GJ, Chen H et al. Human infection with an avian
H9N2 influenza A virus in Hong Kong in 2003. J Clin Microbiol
2005; 43:5760–5767.
5 Cong YL, Pu J, Liu QF et al. Antigenic and genetic characterization of
H9N2 swine influenza in China. J Gen Virol 2007; 88:2035–2041.
6 Peiris JS, Guan Y, Markwell D et al. Cocirculation of avian H9N2
and contemporary ‘‘human’’ H3N2 influenza A viruses in pigs in
southeastern China: potential for genetic reassortment? J Virol
2001; 75:9679–9686.
7 Matrosovich MN, Krauss S, Webster RG. H9N2 influenza A viruses
from poultry in Asia have human virus-like receptor specificity.
Virology 2001; 281:156–162.
8 Li KS, Xu KM, Peiris JS et al. Characterization of H9 subtype influ-
enza viruses from the ducks of southern China: a candidate for the
next influenza pandemic in humans? J Virol 2003; 77:6988–6994.
9 Hall TA. BioEdit: a user-friendly biological sequence alignment editor
and analysis program for Window 95 ⁄ 98 ⁄ NT. Nucleic Acids Symp
Ser 1999; 41:95–98.
10 Zwickl DL. Genetic algorithm approaches for the phylogenetic anal-
ysis of large biological sequence datasets under the maximum likeli-
hood criterion. Ph.D. dissertation. Austin, TX: The Univeristy of
Texas, 2006.
11 Swofford DL. PAUP*: Phylogenetic Analysis Using Parsimony (and
other methods) 4.0 Beta. Sunderland, MA: Sinauer Associates, 2001.
12 Zhang P, Tang Y, Liu X et al. A novel genotype H9N2 influenza virus
possessing human H5N1 internal genomes has been circulating in
poultry in eastern China since 1998. J Virol 2009; 83:8428–8438.
13 Zhang P, Tang Y, Liu X et al. Characterization of H9N2 influenza
viruses isolated from vaccinated flocks in an integrated broiler
chicken operation in eastern China during a 5 year period (1998–
2002). J Gen Virol 2008; 89:3102–3112.
14 Aamir UB, Wernery U, Ilyushina N et al. Characterization of avian
H9N2 influenza viruses from United Arab Emirates 2000 to 2003.
Virology 2007; 361:45–55.
15 Banks J, Speidel EC, Harris PA et al. Phylogenetic analysis of influ-
enza A viruses of H9 haemagglutinin subtype. Avian Pathol 2000;
29:353–359.
16 Cameron KR, Gregory V, Banks J et al. H9N2 subtype influenza A
viruses in poultry in Pakistan are closely related to the H9N2 viruses
responsible for human infection in Hong Kong. Virology 2000;
278:36–41.
17 Huang Y, Hu B, Wen X et al. Diversified reassortants H9N2 avian
influenza viruses in chicken flocks in northern and eastern China.
Virus Res 2010; 151:26–32.
18 Sun Y, Pu J, Jiang Z et al. Genotypic evolution and antigenic drift of
H9N2 influenza viruses in China from 1994 to 2008. Vet Mircobiol
2010; 146:215–225.
Swine influenza in Vietnam: preliminary results of
epidemiological studies
Karen Tre
´
vennec,
a,b
Fre
´
deric Mortier,
a
Faouzi Lyazrhi,
b
Ho Thu Huong,
c
Ve
´
ronique Chevalier,
a
Franc¸ois Roger
a
a
CIRAD, AGIRs Research Unit, Montpellier, France.
b
National Veterinary School of Toulouse, Toulouse, France.
c
National Institute of Veterinary
Research (NIVR), Hanoi, Vietnam.
Keywords cross-species transmission, epidemiology, swine influenza, Vietnam.
Please cite this paper as: Tre
´
vennec et al. (2011) Swine influenza in Vietnam: preliminary results of epidemiological studies. Influenza and other Respiratory
viruses 5 (Suppl. 1), 60–78.
Introduction
In Vietnam, the modelling of the pandemic H1N1 progres-
sion estimates that 460 000 (260 000–740 000) pigs might
be exposed to the virus on the basis of 410 000 cases
among swine owners (220 000–670 000).
1
A poor level of
biosecurity, high animal densities, and a mix of species
could increase the risk of influenza virus flow, persistence,
and emergence on swine and poultry farms. This study was
set up in the Red River Delta, where a third of the national
pig husbandry is produced.
2
The aims are to give prelimin-
ary information of the epidemiological state of swine influ-
enza and in order to further assess the risk of infection of
SwIV, through cross-species transmissions from poultry to
pigs. This paper will present the preliminary results on
SwIV and the risk factors of pig seropositivity in Vietnam.
Materials and methods
A cross-sectional study was conducted in two provinces of
the Red River Delta in April 2009. Pig farms were randomly
selected from nine communes representative of at risk area
of avian H5N1. In each farm, pig and poultry were sampled
and collected to virological and serological analyses. Inter-
views were conducted in all farms by trained interviewees.
Questionnaires included closed and open questions on
ª 2011 Blackwell Publishing Ltd, Influenza and Other Respiratory Viruses, 5 (Suppl. 1), 60–78 71
Options for the control of influenza VII
livestock husbandry ⁄ management and household character-
istics, such as herd size and structure, health history and
vaccination, pig housing, watering and feeding system,
reproduction, purchasing of animals, biosecurity measures,
pig contact with poultry, and environmental factors.
The virological detection assay was performed on pools
of nasal swab specimens from pigs. We investigated
whether real-time RT-PCR assay could detect gene M on
pools of nasal swab specimens before attempting virus iso-
lation from individual nasal swab specimens. The poultry
and pig sera were tested against influenza type A with an
Enzyme-like immunosorbant assay (ELISA) competition
test IDVETª. This commercial kit is designed to specifi-
cally detect antibodies directed against the NP protein anti-
gen of influenza type A viruses. The positive serum
samples were examined in hemagglutination inhibition
(HI) to determine antibody titers and subtypes. The HI test
was tailored for H1, H3, and H9 subtypes in pigs and H6
and H9 subtypes in poultry. Seroneutralization tests by
pseudo particles were used to test the presence of antibod-
ies directed against H5 subtype.
We analysed the data for relationships between Influenza
A serological status (the outcome variable) and possible
risk factors using R version 2Æ11Æ1 (R Development Core
team). The statistical unit was the individual. Initially, the
quantitative variables were encoded into categorical vari-
ables according to the quartiles or median. Descriptive sta-
tistics (e.g., means or medians, proportions, standard
deviations) were calculated for all herd-level and commune
level predictors to assist in the subsequent modeling pro-
cess. We also performed the independence test among all
variables to determine if variables were dependant. Then,
univariate analysis of potential risk factors for the pigs
being positive for SwIV and estimation of odds ratios were
performed using generalised linear mixed models with bin-
ary outcome and logit link function for each herd-level and
commune-level variable to determine which variables were
individually associated with influenza A seropositivity at a
significance level of P <0Æ30. Herd and commune of resi-
dence were included as a random effect to account for the
correlation of observations at the herd level.
The third stage of the analyses included the four herd-
level variables found to be significantly (P <0Æ30) associated
with Influenza A seropositivity. An automatic process using
all possible associations between the selected variables was
computed into a mixed logistic regression models, with ran-
dom effects. When two variables were collinear, as deter-
mined before, only one variable was likely to enter the
multivariable model, and therefore, the selection of which
collinear variable to enter the model was guided by biologi-
cal plausibility and statistical significance.
Results
All of the 146 pools of nasal swabs were RT-PCR negative.
The maximal possible prevalence considering perfect diag-
nostic tests would be of 2Æ03% at a confidence level of
95%, in an infinite population within these regions (Win-
Episcope 2Æ0).
Six hundred-and-nine pig sera were tested in 76 non-
vaccinating farms. The herd seroprevalence of swine influ-
enza in the commune previously infected by the avian
H5N1 in the Red River Delta raised by 17Æ1% [8Æ7; 25Æ6] in
April 2009. But among 13 seropositive farms, only four
had at least two seropositive pigs. The within-herd seropre-
valence is very low, and no seropositivity was detected in
the majority of farms. Estimates had large confidence inter-
vals due to small sample sizes. The individual seropreva-
lence raised 3Æ62% [1Æ98; 5Æ27]. The subtyping of
seropositive sera is still in process.
Descriptive statistical analyses on five major risk factors
of SwIV: farm size, breeding vs. fattening, purchasing, per-
centage of family income, and poultry production, were
conducted. Based on this analysis, three types of farming
systems were identified and included in mixed models
(
Table 1). Percentage of family income by pig production
and poultry production were not differentiating factors for
this typology. Whereas types 1 and 2 seem to be specialized
in fattening, the type 3 produces and might sell piglets on
the farm site.
The exploration of the different variance components
indicated that the random effect variances were mainly asso-
ciated with the herd, while the commune did not seem to
have any effect. Therefore we included in all models only
the herd as a random effect. The random effect term for
herd was modelled, assuming a normal distribution with a
Table 1. Typology of farming system
Type 1: Large fattening farms Largest scale production, with more than 40 pigs per year
Specialized in fattening, and purchase more than 20 pigs per year
Type 2: Small fattening farms Small scale of production, with less than 20 pigs per year
Specialized in fattening, and purchase less than 20 pigs per year
Type 3: Medium breeding-fattening farms Medium scale of production, with less than 40 pigs per year
Breeding and fattening piglets, with rare purchase
72 ª 2011 Blackwell Publishing Ltd, Influenza and Other Respiratory Viruses, 5 (Suppl. 1), 60–78
Tre
´
vennec et al.
common variance [N(0,r2herd)].
3
The univariate analyses
were conducted on 22 variables and typology variables, with
herd as random effect. Some coefficient or confidence inter-
vals were inconsistent because of small effectives, especially
for the percentage of self-product culture or the pig free-
grazing because of the lack of positive results in the dataset.
The only one significant (P value < 0Æ1) parameter was the
percentage of pig sales in the familial annual income. Sur-
prisingly, common risk factors of swine influenza infection,
such as farm size, animal movements, and sanitary parame-
ters got low odds ratio individually (without being signifi-
cant); the typology provides the hypothesis of complex
interactions effects that increase the risk of infection. As
shown in Table 2, the farming system type 3 got a higher se-
roprevalence of 6Æ47% [3Æ00–11Æ94] and a higher risk indica-
tor, with OR = 5Æ26 (P-value = 0Æ18) in comparison with
type 1. This finding was not significant. In the multivariate
mixed model, the percentage of familial income provided
by pig production was the only one significant variable,
with OR = 0Æ22 [0Æ04–1Æ25].
Discussion
The focus on diseased animals in the winter-time is usually
required in order to increase the likelihood to isolate the
virus, although the isolation rate on healthy or clinical
samples never exceed 6%.
4
The season and the lack of dis-
ease reports might explain the difficulties to detect influ-
enza viruses. Additionally, the pooling method tends to
decrease the isolation rate because of a dilution effect,
potential presence of PCR assay inhibitors, or uneven dis-
tribution of virus in the sample.
5
Our seroprevalence results must be confirmed and the
subtypes identified, especially because we found only one
positive animal in a few farms that could be attributed to
false positive results of the ELISA test (performances are
not known). These preliminary results are in favor of a
virus circulation at low level in the spring, but must be
completed by further surveys in the winter and before the
New Year (Te
ˆ
t celebration) when pig production, trade,
and movement increase at their maximum.
No clear prior information on the expected prevalence
of swine influenza in Vietnam, tests sensitivity, and speci-
ficity could be obtained from literature or reliable sources.
Bayesian methods will be carried out in the future in order
to compute prevalence and ⁄ or to estimate the probabilities
of freedom.
The risk factors analysis was limited by the lack of positive
results. Further studies are necessary to identify the at-risk
season and type of farming systems at risk of swine influenza
infection. However, this investigation of risk factors leads to
the hypothesis that medium size breeding-fattening farms
had a higher risk than large or small size fattening farms.
Further investigation are needed to precise this typology.
The risk of SwIV infection increases with a combination of
three major factors. Poultry production does not seem to
play any role on swine infection. The generalized linear
mixed model afforded to take into account all the non
investigated parameters at the herd level. Although we inves-
tigated the most common risk factors of swine influenza
infection covering different kind of fields, the herd random
effect might explain risk variations. Mixed models have
become a frequently used tool in epidemiology. Due to soft-
ware limitations, random effects are often assumed to be
normally distributed. Since random effects are not observed,
the accuracy of this assumption is difficult to check.
6
Further studies, such as case-control or cohort studies
could help to identify more precisely risk factors of swine
influenza seropositivity, as these study designs are more
adapted than cross-sectional studies.
Acknowledgements
We thank all French and Vietnamese field staff involved in
the data collection in Viet Nam for their enthusiasm and
support and we are grateful to the pig farmers participating
in the study for their cooperation and patience. This study
was a part of the GRIPAVI project and was funded by the
French Ministry of Foreign Affairs.
References
1 Boni MF et al. Modelling the progression of pandemic influenza A
(H1N1) in Vietnam and the opportunities for reassortment with
other influenza viruses. BMC Med 2009; 7:43.
2 GSO. Number of pigs by province, in General Statistics Office of
Vietnam. Hanoi: General Statistics Office Of Vietnam, 2008.
3 Osterstock JB et al. Familial and herd-level associations with paratu-
berculosis enzyme-linked immunosorbent assay status in beef cattle.
J Anim Sci 2008; 86:1977–1983.
4 Li H et al. Serological and virologic surveillance of swine influenza
in China from 2000 to 2003. Int Congr Ser 2004; 1263:754–757.
5 Landolt GA et al. Use of real-time reverse transcriptase polymerase
chain reaction assay and cell culture methods for detection of swine
influenza A viruses. Am J Vet Res 2005; 66:119–124.
6 Litiere S, Alonso A, Molenberghs G. Type I and Type II error under
random-effects misspecification in generalized linear mixed models.
Biometrics 2007; 63:1038–1044.
Table 2. Seroprevalence of SwIV and univariate analysis
with typology as fixed effect and herd as random effect
Seroprevalence (%) IC95% OR P-value
Type 1 1Æ93 0Æ53–4Æ87 1 –
Type 2 4Æ76 1Æ77–10Æ08 3Æ11 0Æ39
Type 3 6Æ47 3Æ00–11Æ94 5Æ26 0Æ18
ª 2011 Blackwell Publishing Ltd, Influenza and Other Respiratory Viruses, 5 (Suppl. 1), 60–78 73
Options for the control of influenza VII