Preventive Veterinary Medicine 107 (2012) 253 259

Contents lists available at SciVerse ScienceDirect

Preventive Veterinary Medicine

j ourna l ho me pag e: ww w.elsev i er .com/ locate /prev etmed

Using egg production data to quantify withinpathog erc

J.L. Gonz . BonJ.A. Stegea Department of Epidemiology, Crisis organization and Diagnostics, Central Veterinary Institute (CVI) part of Wageningen UR, P.O. Box 65, 8200 AB Lelystad,The Netherlandsb Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University P.O. Box 80, 151, 3508 TD Utrecht, The Netherlandsc Avian Virology group, Department of Virology, Central Veterinary Institute (CVI), part of Wageningen UR, P.O. Box 65, 8200 AB Lelystad, The Netherlandsd Animal Healt

a r t i c l

Article history:Received 13 OReceived in reAccepted 28 Ju

Keywords:Avian inuenzLPAITransmissionTransmission Reproduction Transmission

1. Introdu

Low patease of vari

Corresponnization and Wageningen UTel.: +31 3202

E-mail add(J.L. Gonzales)

0167-5877/$ http://dx.doi.oh Service, P.O. Box 9, 7400 AA Deventer, The Netherlands

e i n f o

ctober 2011vised form 15 June 2012ne 2012

a

experimentsratioparameters

a b s t r a c t

Even though low pathogenic avian inuenza viruses (LPAIv) affect the poultry industryof several countries in the world, information about their transmission characteristics inpoultry is sparse. Outbreak reports of LPAIv in layer chickens have described drops in eggproduction that appear to be correlated with the virus transmission dynamics. The objectiveof this study was to use egg production data from LPAIv infected layer ocks to quantify thewithin-ock transmission parameters of the virus. Egg production data from two commer-cial layer chicken ocks which were infected with an H7N3 LPAIv were used for this study.In addition, an isolate of the H7N3 LPAIv causing these outbreaks was used in a transmis-sion experiment. The eld and experimental estimates showed that this is a virus with hightransmission characteristics. Furthermore, with the eld method, the day of introductionof the virus into the ock was estimated. The method here presented uses compartmentalmodels that assume homogeneous mixing. This method is, therefore, best suited to studytransmission in commercial ocks with a litter (oor-reared) housing system. It wouldalso perform better, when used to study transmission retrospectively, after the outbreakhas nished and there is egg production data from recovered chickens. This method can-not be used when a ock was affected with a LPAIv with low transmission characteristics(R0 < 2), since the drop in egg production would be low and likely to be confounded withthe expected decrease in production due to aging of the ock. Because only two ocks wereused for this analysis, this study is a preliminary basis for a proof of principle that trans-mission parameters of LPAIv infections in layer chicken ocks could be quantied using theegg production data from affected ocks.

2012 Elsevier B.V. All rights reserved.

ction

hogenic avian inuenza (LPAI) is a mild dis-ous avian species, which is caused by Inuenza

ding author at: Department of Epidemiology, Crisis orga-Diagnostics, Central Veterinary Institute (CVI) part ofR, P.O. Box 65, 8200 AB Lelystad, The Netherlands.38852; fax: +31 320238961.resses: pepevet2@hotmail.com, jose.gonzales@wur.nl.

A viruses belonging to one of 16 Hemagglutinin (H) and9 Neuraminidase (N) subtypes (Fouchier et al., 2005;Alexander, 2007). LPAI virus (LPAIv) infections in poultrywith H5 or H7 virus subtypes are of major importancedue to their ability to mutate to a highly pathogenic avianinuenza virus (HPAIv) (Alexander, 2007). In addition, H9and H6 LPAIv subtypes in particular, have been affecting thepoultry industry of different countries in Asia (Cheung et al.,2007; Xu et al., 2007; Hadipour, 2011; Park et al., 2011).

LPAIv surveillance programmes have been imple-mented in many countries (Gonzales et al., 2010). Although

see front matter 2012 Elsevier B.V. All rights reserved.rg/10.1016/j.prevetmed.2012.06.010enic avian inuenza virus in comm

alesa,b,, A.R.W. Elbersa, J.A. van der Goota, Dmanb-ock transmission of lowial layer chickens

tjea, G. Kochc, J.J. de Witd,

254 J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259

these programmes may be useful to determine whether theprevalence of infected birds is below a pre-set level, theirusefulness in early detection is still unknown. To establishthe latter (Graat et al., 2001; Fischer et al., 2005), quantita-tive knowle

The trandeterminedet al., 2003(Stegeman experimentters in a conappears to characterissame H sub2011, 2012mentally wthe quantiter would htransmissiotransmissiotransmissiosion experiavailable wperspective

LPAIv inpresent unproductioninvolving cZanella, 20production(Zanella, 20increase (frSuch dropsin experim2006; Gonzworthwhiletion can be would be a

In 2003formed in high prevaldetected initives out ofarms (30 sewas later isVelkers et amate the trin a transmdata of the

2. Method

2.1. Experimtransmissio

The chicthe H7N3 et al., 2006rst quantiimental prodescribed e

et al., 2011). Briey, two experimental trials were car-ried out. Each trial consisted of 10 specied pathogen free(spf) White Leghorn chickens (6 weeks old). Five chick-ens were inoculated and the remaining ve were kept as

cts. ChtracheID50 (5monito

samp (viruswere t.i. 10 xperi

paramntact ay, the) thatecoverf animated uing a

estimaull dis

and asic ref annknowe Carlnd Wnalysi

Devesed foe tranmmitrimentainmengeninether

Estimaegg pr

g prod chickereportsis. Egily egg

week th oceek (w

Wit ata po

the inministptiblehich wing anted asumber

populmpliesdge of transmission of LPAIv is necessary.smission characteristics of a pathogen can be

in transmission experiments (van der Goot; Velthuis et al., 2007) or in eld outbreakset al., 1999; Bos et al., 2009). Transmissions allow for quantifying transmission parame-trolled environment, but, in case of LPAIv, therebe considerable variation in the transmissiontics of different virus strains even within thetype (van der Goot et al., 2003; Gonzales et al.,). To quantify the existing variability experi-ould be very costly. An alternative would becation of transmission from eld data. The lat-ave the following benets: (i) the quantiedn parameters would be a direct indicator of then characteristics of the virus in the eld, (ii)n could be studied faster than with transmis-ments, and (iii) the use of indicators alreadyould be cheaper and more desirable from the

of animal welfare.fections in poultry are often subclinical orspecic clinical signs. However, drops in egg

have been often reported during outbreakshicken layers ocks (Henzler et al., 2003;03; de Wit et al., 2004); with sudden drops in

ranging from 10% (de Wit et al., 2004) to 40%03) in a couple of weeks followed by a slightom the biggest drop level) some weeks later.

in egg production have been also observedentally infected layer chickens (Trampel et al.,ales et al., 2012). Consequently, it would be

to examine whether the drop in egg produc-used to estimate transmission parameters. Thischeap alternative to transmission experiments., a cross-sectional serological survey was per-The Netherlands (de Wit et al., 2004) and aence of seropositive animals to H7N3 LPAIv was

a cluster of three farms: one turkey (10 seropos-f 10 samples) and two free-range layer chickenropositives out of 30 samples). The H7N3 virus

olated from the turkey farm (de Wit et al., 2004;l., 2006). The objective of this study was to esti-ansmission characteristics of this H7N3 LPAIvission experiment and from the egg productiontwo infected layer ocks.

s

ental estimation of within-ockn parameters

ken-to-chicken transmission characteristics ofLPAIv (cleavage side: PEIPKGR*GLF (Velkers)) causing the outbreaks here analysed wereed in a transmission experiment. The exper-cedure and data analysis was similar to thatlsewhere (van der Goot et al., 2003; Gonzales

containtra106 Ewas swabvirusples to d.pthis erate of coper d(daysthe rber oestimassumwas Weibdata)The buct owas uMontmal athe aR (Rwas u

Thcal coexpeContof WaThe N

2.2. from

Eglayer4 as analyof dafromin bolast wby de25 dlateddetersuscein w(Keelmulathe ntotaltion iickens were inoculated both intranasally andally with 0.1 ml/route of inoculum containing0% egg infectious dose)/ml. Virus transmissionred by regularly collecting cloaca and tracheales, which were examined for the presence of

isolation in embryonated chicken eggs). Sam-aken daily from day post inoculation (d.p.i.) 1and later at d.p.i 14, 17 and 21. The data fromment were used to estimate the transmissioneter (day1), which is the expected numberinfections caused by an infectious individual

infectious period T, which is the average time an infected individual remains infectious, andy rate (day1), which is the expected num-als recovering from infection per day. wassing a generalised lineal model (GLM) method,

latent period 1 day. The mean length of Tted using a parametric survival model with atribution (the distribution that best tted the

was estimated as the inverse of T ( = 1/T).production ratio R0 was estimated as the prod-d T. Because the correlation between and Tn, condence intervals for R0 were derived by

o (MC) simulations assigning to and T Lognor-eibull distributions, respectively (Table 2). Alls were performed using the statistical packagelopment Core Team, 2005). The library Survivalr the survival analysis.smission experiment was approved by an ethi-tee and complied with the Dutch law on Animals. The experiment was carried out in the Hight Unit at the Central Veterinary Institute part

gen University and Research centre, in Lelystad,lands.

tion of within-ock transmission parametersoduction data

uction data from the two infected free-rangen ocks, here referred to as Farm-3 and Farm-ed by de Wit et al. (2004), were used for theg production data consisted of weekly averages

production. For both ocks, we selected data38 (calendar week) of 2002 when productionks appeared to be maximal and stable to theeek 10 of 2003) that production was reported

et al. (2004). This period resulted in a total ofints (Fig. 1). To analyse these data, we simu-fection dynamics in these ocks constructingic susceptible-infectious-recovered (SIR) and-exposed-infectious-recovered (SEIR) models,e assumed a homogeneous contact structured Rohani, 2008). The transmission term was for-

S(t)I(t)/N(t), with S(t), I(t) and N(t) denoting of susceptible S and infectious I chickens in theation of size N at time t (days). This formula-

that the transmission pressure is independent

J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259 255

Fig. 1. Simularecovered (SIRvalues as presweek 38 of the

of populati(Bouma et data sets, frexperiment

Egg proassumed tois denoted function ofsusceptibleinfected) (Etime t in th

pe(t) = [S(tFor the s

for the tranwere derivvalues for derived froage egg proinfection wduction wawere nishted LPAI outbreaks based on egg production data of affected ocks. Egg produ) infection dynamics (solid line, showing only infectious chickens) and simulatedented in Table 3 for Farm-3 and Farm-4. The arrows indicate the estimated day o

calendar year 2002.

on size, which appears to be appropriate hereal., 1995). By applying this model to bothom the simulations and from the transmission, the results could be compared.duction in latently infected chickens was

be equal to that of susceptible chickens, andas ps. Egg production (pe) was simulated as a

the expected production (ps, pi and pr) of the chickens (S) plus that of the exposed (latently), infectious (I) and recovered (R) chickens ate epidemic.

) + E(t)]ps + I(t)pi + R(t)pr (1)imulations, the starting values (initial guesses)smission parameters ( = 0.49 and = 1/7.7)

ed from Gonzales et al. (2011). The startingthe production parameters ps and pr werem the production data by calculating the aver-duction from the rst 7 weeks (we assumed thatas not yet introduced) and the last 6 weeks (pro-s stable and we assumed that both outbreaksed) of the study period (Farm-3: ps = 0.944,

pr = 0.827 avalue for pi imental es2012).

The predcompared wof the prediwas calcula(SSQ) was uparameter best t. Themission and

The SIR/using an in-out in Exceland selectinthe covarianrequest). Th1/40 days (tion and acwas iterateMarquardt ction data (diamonds), simulated susceptible-infectious- egg production (dashed line) based on the tted parameterf start of the outbreak. Day zero represents the rst day of

nd Farm 4: ps = 0.852, pr = 0.766). The starting= 0.56 (for both farms) was derived from exper-timates reported elsewhere (Gonzales et al.,

icted egg production was averaged weekly andith the observed data. The squared deviation

cted production from the observed productionted and the total sum of squared deviationssed as a measure for goodness of t. The set ofvalues with the lowest SSQ was selected as the

simulation models were specied by the trans- production parameters described in Table 1.

SEIR simulations and parameter optimization built routine to minimize the SSQ were carried using the add-in tool PopTools (Hood, 2010)g the Marquardt method for the estimation ofce matrix (the Excel le can be provided upone time step (t) used in the simulations wasthis small time step was optimised for dura-curacy of the simulations). The tting routined until the SSQ reached a constant value. Thetting method nds local minima for the SSQ.

256 J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259

Table 1Parameters used for the susceptible-infectious-recovered (SIR) andsusceptible-exposed-infectious-recovered (SEIR) models simulationsintended to t the dynamics in egg production.

Parameter Description Units

TL

I(0)

psa

pia

pra

a The valuemodels.

It might wedifferent loues, are fouto the initiand , we ater values estudy (Sectiminima. Codone with t

3. Results

3.1. Experimtransmissio

All inocbecame intact chickeobserved anisolation atperiod mighinfectious pbetween inmean estimwhile the mlatter givesmate of of contact-ioverview odence inter

3.2. Estimafrom egg pr

Optimizfrom Farm-from Farm-

after the peak of the outbreak (here referred to as recov-ery phase), was more variable (Fig. 1), and optimizationsresulted in unexpected estimates for pi and (e.g. pi = 10%and = 0.46 day1). To improve the optimisation proce-

for thi optimr xedates osults data f

3. Thent stame m

simila estim

close he SIR e 3). T-3, whs. The rffecterst inferm-3 wtudy pe 3).

iscussi

e tranens, evted omissioo othe, 2012with td of ind of twf the v

both ean ereportTransmission rate parameter day1

Recovery rate day1

Infectious period, equal to 1/ dayLatent period, this parameter was used inthe SEIR model

day1

Starting value for the Infectious (SIRmodel) or the Exposed (SEIR model)compartment. This parameter allows themodel to identify the time the epidemicstarted in each ock.

day

Level of egg production of Susceptible andExposed chickens. Since Exposed animalsare latently infected, their production isexpected to be the same as that ofSusceptible chickensLevel of egg production of InfectiouschickensLevel of egg production of Recoveredchickens

s of these parameters are included as proportions in the

ll be that with different initial values guessed,cal minima, and thus different parameter val-nd. To test the robustness of the tted resultsal conditions, in particular the estimates of lso initiated tting iterations with the parame-stimated in the transmission experiment in thison 3.1). Both approaches yielded the same localmparison between SIR and SEIR model ts werehe Akaikes Information Criterion (AIC).

ental estimation of within-ockn parameters

ulated chickens in both trials (ve per trial)fected and transmitted virus to their con-

dureto beand pestim

Retion Tabledifferthe sweremeanwereels, t(TablFarmfarmwas athe in Fathe s(Tabl

4. D

Thchickaffectranstion t2011tion, perioperiotion owithest meld ns (Table 2). No apparent clinical signs wered all inoculated chickens were positive in virus

d.p.i. 1, the latter indicating that the latentt be less than one day. The mean length of theeriod (T) was signicantly different (P < 0.05)oculated and contact-infected chickens. Theate of T in inoculated chickens was 13.32 days,ean T in contact chickens was 10.03 days. The

a recovery rate = 0.10 day1. The mean esti-was 0.91 day1. The MC result for R0 using Tnfected chickens was 9.1. Table 2 provides anf all parameter estimates and their 95% con-vals (CI).

tion of within-ock transmission parametersoduction data

ation of parameter estimates using the data4 did not converge as easily as that using data3. This was because egg production in Farm-4,

2004). Infoof a LPAIv aimplement(backward whenever ait would beproduction

The meathe reproduto be highelow numbeing to statiswith the when comptransmissioGoot et al.,et al., 2009)better suitenation on tto provide s farm, we reduced the number of parametersised, by keeping the production parameters ps

to the starting values. This resulted in robustf the optimised parameters (Table 3).of the parameter estimation using egg produc-rom Farm-3 and Farm-4 are summarised inese estimates were robust and insensitive toarting conditions. All simulations converged toinimum (Table 3, Fig. 1). Parameter estimatesr when using either a SIR or a SEIR model. Theates of the latent period (L), in the SEIR model,to zero. Based on the SSQ and AIC of these mod-model showed a better t than the SEIR modelhe estimated was lower in Farm-4 than inile the estimates of were similar for bothesults of the simulations also show that Farm-3d between 2 and 3 weeks before Farm-4. The dayctious chicken was present in the ock (t(I = 1))as around day 54 (counting from the start of

eriod in week 38), and around day 72 in Farm-4

on

smission characteristics of the H7N3 LPAIv inaluated using either egg production data of the

cks (here referred to as the eld method) or then experiment, showed that this virus in rela-r LPAIv (van der Goot et al., 2003; Gonzales et al.,) was highly transmissible in chickens. In addi-he eld method, we were able to estimate thetroduction of the virus in each ock. There was ao to three weeks between the day of introduc-irus in Farm-3 and Farm-4. This is in accordancethe difference in the time at which the low-gg production was observed in each ock ands that described these outbreaks (de Wit et al.,rmation about the transmission characteristicsnd its introduction into a ock is relevant for theation, design or evaluation of control measuresand forward tracing, surveillance, etc.). Hence,

seropositive ock is detected by surveillance, advisable to obtain retrospectively daily egg

data from that ock.n estimates of the transmission rate () andction ratio (R0) obtained experimentally appearr (because of low statistical power due to ther of observations in this study, it is not interest-tically test this difference) than those obtainedeld method. A similar trend is also observedaring the outcomes of two separate studies onn of an H7N7 HPAIv: one experimental (van der

2005) and the other based on eld data (Bos. Group transmission experiments appear to bed to compare treatments, e.g. effect on vacci-ransmission (De Jong and Kimman, 1994), thanprecise estimates of transmission parameters

J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259 257

Table 2Experimental estimates of transmission parameters and basic reproduction ratio of the H7N3 LPAIva.

Latent period (days) Infectious period Tb (days) Recovery rate c (day1) Transmission rate (day1) Reproduction ratio R0(95%CId) (95%CI) (95%CI) (95% CI)e

Contacts:1 10.03 (8.5011.56) 0.10 (0.090.12) 0.91 (0.451.62) 9.1 (3.619.5)

Inoculated:13.32 (11.2815.35) 0.07 (0.060.09)

a Low pathogenic avian inuenza virus.b The estimated infectious period (T) of contact infected chickens was signicantly different from the estimated T of inoculated infected animals.c The recovery rate was estimated from = 1/T.d CI = condence interval.e The limits of this interval are the 2.5% and 97.5% quantiles of the Monte Carlo procedure. R0 = T, where was assigned a Lognormal distribution

(mean = 0.094, standard deviation = 0.322), and T was assigned a Weibull distribution (shape = 4.616; scale = 10.977). The Weibull parameters are thoseestimated for T of contact-infected chickens.

(Velthuis et al., 2007). Hence, extrapolation of experimen-tal results, where conditions (age, breed, management, etc.)are different from the eld, should be carried out care-fully. However, in the absence of eld data, experimentallyderived estimates offer a useful insight into the transmis-sion of a pathogen. In this study, the experimental resultsshowed that this LPAIv is highly transmissible in chick-ens, which is in agreement with the results obtained withthe eld method. In addition, the experimental estimates from this and other studies (Gonzales et al., 2011, 2012) provided information for the initial parameter values forthe optimization process in the eld method.

It has been shown that the level and duration ofvirus shedding is directly proportional to the inocu-lation/infectious dose (Stoyanov and Vladimirov, 2008;Chaves et al., 2011). This experiment used the sameinoculationiments usinGonzales etiments, the

inoculated-infected chickens was longer than that of thecontact-infected chickens. We hypothesize that, for theH7N3 LPAIv, this difference is a consequence of a pos-sible difference in the infectious doses received by theinoculated- and contact- infected animals, with the for-mer receiving a higher dose, to a level that resulted in theobserved difference in T. However, we expect that the inoc-ulation dose had no signicant effect on the infectiousnessof the inoculated chickens, and therefore, the estimates ofthe transmission parameters (Spekreijse et al., 2011).

Both the LPAIv and the characteristics of the infectedock have inuence in the transmission dynamics. Hence,variation in the within-ock transmission characteristics ofa virus between ocks can be expected (Comin et al., 2011).This variation could be related to different managementconditions, breed of the chicken, presence of concomitant

ses, ag differee transin Farm

Table 3Estimated (op ta. Valubrackets, then

Parametersb F

S

(day1) 0 (day1) 0T (day)c 9L (day) pi (%) 6pr (%) 7ps (%) 8I(0) 5

(Sum of Squa 4AICd

Compound pR0t(I = 1) (day)

a SIR = susceb Transmiss

value of I whec The infectd AIC = Akaike Condenc

quantiles.f t(I = 1) den

describing the dose as that used in other transmission exper-g LPAIv in chickens (van der Goot et al., 2003;

al., 2011). However, in contrast to those exper- length of the infectious period (T) in the

diseamainin ththan

timized) transmission parameters by tting models to egg production da values are xed.

Farm-3

SIRa SEIRa

0.72 (0.680.77) 0.73 (0.690.77) 0.13 (0.090.17) 0.13 (0.100.16) 7.69 (5.8811.11) 7.69 (6.2510.00)

0.02 (0.010.04) 47.4 (38.356.4) 47.1 (38.655.7) 82.8 (82.383.3) 82.8 (82.383.3) 94.4 (94.094.9) 94.4 (94.094.9) 8.2 1015(5.7 10163.9 1014)

7.5 1015(1.4 10152.5 1014)

res 1.98 1.98

29.13 31.10 1

arameterse

5.6 (4.37.7) 5.6 (4.47.8) 4f 54 (4962) 54 (5061) 7

ptible-infectious-recovered. SEIR = susceptible-expoused-infectious-recovered.ion rate , recovery rate , infectious period T, latent period L, egg production of inn time is zero I(0), Reproduction ratio R0.ious period T was calculated as T = 1/ .es information criterion.

e intervals for the compound parameters were estimated by Monte Carlo samp

otes the time I = 1 meaning the day that the rst infectious animal was present in growth of an epidemic I(t) = I(0)e()t (Keeling and Rohani, 2008), where I(t) = 1e of production and others. In this study, thence between Farm-3 and Farm-4 was observedmission rate (), which was higher in Farm-3-4. This implies that the virus spread faster in

es between brackets are the 95% condence intervals. If no

arm-4

IR SEIR

.50 (0.450.55) 0.50 (0.420.59)

.11 (0.050.16) 0.11 (0.040.17)

.09 (6.2520.00) 9.09 (5.8825.00)0.03 (0.010.04)

6.4 (60.272.5) 66.3 (56.376.3)6.6 76.65.2 85.2.0 10132.9 10137.9 1013)

4.4 1013(1.8 10139.2 1013)

9.17 49.17

05.38 107.38

.7 (3.08.6) 4.7 (2.211.0)2 (6189) 72 (5897)

fectious pi, recovered pr and susceptible ps chickens, initial

ling. The limits of these intervals are the 2.5% and 97.5%

the ock. This was estimated by solving t from the formula and I(0), and are the above estimated values.

258 J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259

Farm-3. Unfortunately, some characteristics of the ocks,such as breed, were not provided in the paper by de Witet al. (2004). A possible explanation for the differencein could be that Farm-3 also suffered from a nema-tode infect2004). The lsusceptibilitransmissiotion, a lowe

Egg prodlast phase eters for srespectivelymation prooutbreaks uproductiondata to be uof Farm-4, we found thestimating subsequentthe estimatresults withsimulated dadvisable towhich will process tha

Becausethis methodthe outbreacase, whena low frequIn these cirtransmissiomercial layeintroductioinvestigatesurveillancediscover onbreaks in lais most likeegg producfor the estithe time ofting an assuobtained froas this stud

The SIR tion dynamlatent (expzero. This iswith other2003; Gonzinoculated isolation onsidering thewould be panalysing thinfections iHowever, tstrain or in

et al., 2009; Spekreijse et al., 2011), with some studiesreporting latent periods longer than one day (Spekreijseet al., 2011). Therefore, the decision to use a SIR or a SEIRmodel might depend on the virus to be analysed.

me LP 1

.5) (vahe pree wolenceroducxpecte. By ueld meter vw trad not bon shothere aetho

data ay/backous cood to cr (oo

utbreaased mof the; Beltrality inrametlems; a

was b these ), and

on thestimatshould

for an for a pAIv infsing ths the o

of difnexpee a drock. Thpriate

owled

is stuc Strurograer (Ceersity s to th

ences

der, D.Jccine 25ion at the time of the outbreak (de Wit et al.,atter could have contributed to: (i) increase thety of the ock, which consequently enhancedn, and (ii) induce, together with the viral infec-r egg production of infectious chickens (pi).uction data before the outbreak and from the

of the outbreak dene the production param-usceptible (ps) and recovered (pr) chickens,. These parameters are inuential in the esti-cess of the transmission parameters. For thesed for this study, only weekly averaged egg

data were available, which resulted in fewersed to optimise ps and pr. As a result, in the casefor the reason already explained (Section 3.2),at better convergence and t was obtained by

these parameters separately from the data andly keeping them xed. Having had daily data,es would surely have been more robust than our

weekly data points. We conrmed this withata (results not shown). Therefore, it would be

obtain daily egg production data from the ock,provide more information for the optimizationn weekly averages.

pr is inuential for the optimization process, is best suited to be applied in situations wherek has already died out. This is likely to be the

surveillance programmes are performed withency (e.g. once a year) (Gonzales et al., 2010).cumstances, this method would help to studyn of LPAIv, that have been circulating in com-r ocks. In addition, the estimates of the day of

n could be used to reconstruct an epidemic to between farm spread. Occasionally, serological

(Elbers et al., 2007) or early detection systems-going outbreaks, e.g. in 2011, two LPAIv out-yer chickens in The Netherlands (OIE). Detectionly to happen at the time of the biggest drop intion. In such a situation, with no data availablemation of pr, the transmission parameters and

introduction could be still optimised, by set-med value for pr. This assumed value could bem previous outbreaks or reported studies suchy.models appeared to explain the egg produc-ics better than the SEIR models. The estimatedosed) periods in the SEIR models were close to

in agreement with transmission experiments H5 or H7 LPAIv strains (van der Goot et al.,ales et al., 2011, 2012), which showed that allchickens were already positive for PCR or viruse d.p.i. The results of this study suggest con-

law of parsimony that the use of a SIR modelreferable above a SEIR model for simulating ore within-ock transmission dynamics of LPAIv

n chickens, since it requires fewer parameters.he latent period varies depending of the virusoculation dose (Spickler et al., 2008; Bouma

SoR0 < 1and tin timprevaegg pthe eockthe paramfor lowoulmissithis, this mtion hobbgenemetha litte

Oincretime 2004mortof paprobtalityafter2004encethe e

It usedbasisof LPed uopenisticsand iinduccial appro

Ackn

ThnomiFES pKoeijUnivment

Refer

AlexanVaAIv spread slowly within a ock ( < 0.22 day ;n der Goot et al., 2003; Gonzales et al., 2012)valence of infectious chickens at any momentuld be low (in the peak of the outbreak, the

would be lower than 10%). Therefore, drops intion might be unnoticed or confounded withd decrease in production due to aging of the

sing simulated data, we saw that with R0 < 2,ethod was not able to reproduce the originalalues consistently (data not shown). Therefore,nsmitting viruses, the method proposed heree applicable, and other methods to study trans-uld be applied (Comin et al., 2011). Other thanre also other limitations to this method. First,

d performs better with consistent egg produc-s is the case in commercial layer ocks unlikeyard ocks. Secondly, the assumption of homo-ntact structure limits the application of thisommercial ocks (free-range and indoors) withr reared) housing system.ks of LPAI have also been associated withortality, with the peak of mortality around the

biggest drop in egg production (de Wit et al.,n-Alcrudo et al., 2009). We did not include

the model because: (i) the increased numberers to be estimated could lead to convergencend (ii) mortality was very low (the peak mor-elow 0.5% per week and mortality before andpeak was below 0.25% per week) (de Wit et al.,was therefore assumed not to have a big inu-

population dynamics of the infection and oned transmission parameters.

be noted that data from only two ocks werealysis, and therefore, this study is a preliminaryroof of principle that transmission parametersections in layer chicken ocks could be quanti-e egg production data from affected ocks. This

pportunity to study the transmission character-ferent LPAIv affecting chickens with a practicalnsive method, provided that these viruses doop in egg production in the affected commer-is information will be of importance to develop

control measures against LPAI epidemics.

gements

dy was supported by the Foundation for Eco-cture Strengthening (FES), in The Netherlands:mme on Avian Inuenza. We thank Aline dentral Veterinary Institute part of Wageningenand Research centre) for her constructive com-is manuscript.

., 2007. An overview of the epidemiology of avian inuenza., 56375644.

J.L. Gonzales et al. / Preventive Veterinary Medicine 107 (2012) 253 259 259

Beltrn-Alcrudo, D., Carpenter, T.E., Cardona, C., 2009. A ock-tailoredearly warning system for low pathogenic avian inuenza (LPAI) incommercial egg laying ocks. Prev. Vet. Med. 92, 324332.

Bos, M.E.H., Nielen, M., Koch, G., Bouma, A., De Jong, M.C.M., Stegeman, A.,2009. Back-calculation method shows that within-ock transmissionof highly pathogenic avian inuenza (H7N7) virus in the Netherlandsis not inuenced by housing risk factors. Prev. Vet. Med. 88, 278285.

Bouma, A., Claassen, I., Natih, K., Klinkenberg, D., Donnelly, C.A., Koch, G.,van Boven, M., 2009. Estimation of transmission parameters of H5N1avian inuenza virus in chickens. PLoS Pathog. 5, e1000281.

Bouma, A., de Jong, M.C.M., Kimman, T.G., 1995. Transmission of pseu-dorabies virus within pig populations is independent of the size of thepopulation. Prev. Vet. Med. 23, 163172.

Chaves, A.J., Busquets, N., Campos, N., Ramis, A., Dolz, R., Rivas, R., Valle, R.,Abad, F.X., Darji, A., Maj, N., 2011. Pathogenesis of highly pathogenicavian inuenza A virus (H7N1) infection in chickens inoculated withthree different doses. Avian Pathol. 40, 163172.

Cheung, C.L., Vijaykrishna, D., Smith, G.J.D., Fan, X.H., Zhang, J.X., Bahl, J.,Duan, L., Huang, K., Tai, H., Wang, J., Poon, L.L.M., Peiris, J.S.M., Chen,H., Guan, Y., 2007. Establishment of inuenza A virus (H6N1) in minorpoultry species in Southern China. J. Virol. 81, 1040210412.

Comin, A., Klinkenberg, D., Marangon, S., Toffan, A., Stegeman, A., 2011.Transmission dynamics of low pathogenicity avian inuenza infec-tions in turkey ocks. PLoS ONE 6, e26935.

De Jong, M.C.M., Kimman, T.G., 1994. Experimental quantication ofvaccine induced reduction in virus transmission. Vaccine 12,761766.

de Wit, J.J., Koserologicathe presen33, 56557

Elbers, A.R.WFabri, T., in the Ne15th Worlp. 138.

Fischer, E.A.J., vde Jong, Mtuberculosdemiologi

Fouchier, R.A.S., Smith,2005. Chatinin subt2814282

Gonzales, J.L., E2012. Tranvirus of H155, 2072

Gonzales, J.L., E2010. Lowthe risk of active survViruses 4,

Gonzales, J.L., 2011. Tranavian inuVet. Micro

Graat, E.A.M., de Jong, M.C.M., Frankena, K., Franken, P., 2001. Modellingthe effect of surveillance programmes on spread of bovine herpesvirus1 between certied cattle herds. Vet. Microbiol. 79, 193208.

Hadipour, M.H., 2011. Serological evidence of inter-species transmissionof H9N2 avian inuenza virus in poultry, Iran. Int. J. Anim. Vet. Adv. 3,2932.

Henzler, D.J., Kradel, D.C., Davison, S., Ziegler, A.F., Singletary, D., DeBok,P., Castro, A.E., Lu, H., Eckroade, R., Swayne, D., Lagoda, W., Schmucker,B., Nesselrodt, A., 2003. Epidemiology, production losses, and con-trol measures associated with an outbreak of avian inuenza subtypeH7N2 in Pennsylvania (199698). Avian Dis. 47, 10221036.

Hood, G.M., 2010. PopTools version 3.2.3.Keeling, M.J., Rohani, P., 2008. Modeling Infectious Diseases in Humans

and Animals. Princeton University Press, Princeton, p.366.OIE, WAHID-Interface, Summary of Immediate notications and Follow-

ups: 20062011. World Organization of Animal Health.Park, K.J., Kwon, H.-i., Song, M.-S., Pascua, P.N.Q., Baek, Y.H., Lee, J.H., Jang,

H.-L., Lim, J.-Y., Mo, I.-P., Moon, H.-J., Kim, C.-J., Choi, Y.K., 2011. Rapidevolution of low-pathogenic H9N2 avian inuenza viruses followingpoultry vaccination programmes. J. Gen. Virol. 92, 3650.

R Development Core Team, 2005. R: A Language and Evironment forStatistical Computing. Foundation for Statistical Computing, Vienna,Austria.

Spekreijse, D., Bouma, A., Stegeman, J.A., Koch, G., de Jong, M.C.M., 2011.The effect of inoculation dose of a highly pathogenic avian inuenzavirus strain H5N1 on the infectiousness of chickens. Vet. Microbiol.147, 5966.

er, A.R., d clinic-pathoan, A., E

ansmiss97199ov, S.I.,ian H6Nsponse on. Revuel, D.W

serum aian inur Goot, Jn of theian inur Goot, Jtion of t7N7) ins, F.C., Ban, J.A., 2e Netheris, A.G.J07. Desth animM., Li, Kiris, J.S.Mmiology00 to 20a, A., 20ers in Itch, G., Fabri, T.H.F., Elbers, A.R.W., 2004. A cross-sectionall survey of the Dutch commercial poultry population force of low pathogenic avian virus infections. Avian Pathol.0.

., De Wit, J.J., Hulsbergen, H.B.A., van der Spek, A.N.,Koch, G., 2007. Avian inuenza surveillance in poultrytherlands between 20042006. In: Proceedings of thed Veterinary Poultry Association Congress, Beijing, China,

an Roermund, H.J.W., Hemerik, L., van Asseldonk, M.A.P.M.,.C.M., 2005. Evaluation of surveillance strategies for bovineis (Mycobacterium bovis) using an individual based epi-cal model. Prev. Vet. Med. 67, 283301.M., Munster, V., Wallensten, A., Bestebroer, T.M., Herfst,

D., Rimmelzwaan, G.F., Olsen, B., Osterhaus, A.D.M.E.,racterization of a novel inuenza A virus hemagglu-ype (H16) obtained from black-headed Gulls. J. Virol. 79,2.lbers, A.R.W., Bouma, A., Koch, G., de Wit, J.J., Stegeman, J.A.,smission characteristics of low pathogenic avian inuenza7N7 and H5N7 subtypes in layer chickens. Vet. Microbiol.13.lbers, A.R.W., Bouma, A., Koch, G., Wit, J.J.d., Stegeman, J.A.,

-pathogenic notiable avian inuenza serosurveillance andinfection in poultry a critical review of the European Unioneillance programme (20052007). Inuenza Other Respir.9199.van der Goot, J.A., Stegeman, J.A., Elbers, A. R.W., Koch, G.,smission between chickens of an H7N1 low pathogenicenza virus isolated during the epidemic of 1999 in Italy.biol. 152, 187190.

Spicklanlow

StegemTr19

Stoyanavretio

Trampinav

van deisoav

van deca(H

Velkermth

Velthu20wi

Xu, K.Pede20

ZanelllayTrampel, D.W., Roth, J.A., 2008. The onset of virus sheddingal signs in chickens infected with high-pathogenicity andgenicity avian inuenza viruses. Avian Pathol. 37, 555577.lbers, A.R.W., Bouma, A., de Smit, H., de Jong, M.C.M., 1999.

ion of classical swine fever virus within herds during the8 epidemic in The Netherlands. Prev. Vet. Med. 42, 201218.

Vladimirov, B.I., 2008. Inuence of inoculation dose of2 inuenza A virus on virus shedding and humoral immunef chickens after articial experimental intravenous infec-

e Med. Vet. 159, 489493.., Zhou, E.K.J.Y., Koehler, K.J., 2006. Detection of antibodiesnd egg yolk following infection of chickens with an H6N2enza virus. J. Vet. Diagn. Invest. 18, 437442..A., De Jong, M.C.M., Koch, G., van Boven, M., 2003. Compar-

transmission characteristics of low and high pathogenicityenza A virus (H5N2). Epidemiol. Infect. 131, 10031013..A., Koch, G., de Jong, M.C.M., van Boven, M., 2005. Quanti-he effect of vaccination on transmission of avian inuenza

chickens. PNAS 102, 1814118146.ouma, A., Matthijs, M.G.R., Koch, G., Westendorp, S.T., Stege-006. Outbreak of avian inuenza H7N3 on a turkey farm inlands. Vet. Rec. 159, 403405.., Bouma, A., Katsma, W.E.A., Nodelijk, G., De Jong, M.C.M.,ign and analysis of small-scale transmission experimentsals. Epidemiol. Infect. 135, 202217..S., Smith, G.J.D., Li, J.W., Tai, H., Zhang, J.X., Webster, R.G.,., Chen, H., Guan, Y., 2007. Evolution and molecular epi-

of H9N2 inuenza A viruses from Quail in Southern China,05. J. Virol. 81, 26352645.03. Avian inuenza attributable to serovar H7N1 in lightaly. Avian Dis. 47, 11771180.