Research in Veterinary Science 93 (2012) 1231–1240

Contents lists available at SciVerse ScienceDirect

Research in Veterinary Science

journal homepage: www.elsevier .com/locate / rvsc

Infectious risk factors for individual postweaning multisystemic wastingsyndrome (PMWS) development in pigs from affected farms in Spain and Denmark

Llorenç Grau-Roma a,b,1, Anders Stockmarr c,d,1, Charlotte S. Kristensen e, Claes Enøe d,Sergio López-Soria b, Miquel Nofrarías b, Vivi Bille-Hansen d,f, Charlotte K. Hjulsager d, Marina Sibila b,Sven E. Jorsal e, Lorenzo Fraile b,g, Poul Baekbo e, Hakan Vigre d, Joaquim Segalés a,b,⇑, Lars E. Larsen d

a Departament de Sanitat i Anatomia Animals, Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spainb Centre de Recerca en Sanitat Animal (CReSA), UAB-IRTA, Campus de la Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spainc Department of Informatics and Mathematical Modelling, Technical University of Denmark, Asmussens Alle, Building 305/126, DK-2800 Lyngby, Denmarkd National Veterinary Institute, Technical University of Denmark, Bülowsvej 27, DK-1790 Copenhagen V, Denmarke Pig Research Centre, Vinkelvej 11, DK-8620 Kjellerup, Denmarkf Dana Lab ApS, Agern Alle 3, 2970 Hørsholm, Denmarkg Institut de Recerca i Tecnologia Agroalimentàries (IRTA), Barcelona, Spain

a r t i c l e i n f o a b s t r a c t

Article history:Received 19 September 2011Accepted 2 July 2012

Keywords:Porcine circovirus type 2 (PCV2)Postweaning multisystemic wastingsyndrome (PMWS)Infectious risk factorsSurvival analysis

0034-5288/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.rvsc.2012.07.001

⇑ Corresponding author at: Departament de Sanitatsitat Autònoma de Barcelona, 08193 Bellaterra, Barce

E-mail addresses: llorenc.grau@uab.cat (L. Grau-Rouab.cat (J. Segalés).

1 These authors contributed equally to the authorsh

Two prospective longitudinal studies in 13 postweaning multisystemic wasting syndrome (PMWS)-affected farms from Spain (n = 3) and Denmark (n = 10) were performed. Blood samples from pigs werelongitudinally collected from 1st week until the occurrence of the PMWS outbreak. Wasted and healthyage-matched pigs were euthanized, necropsied and histopathologically characterised. PMWS diagnosiswas confirmed by means of lymphoid lesions and detection of porcine circovirus type 2 (PCV2) in thesetissues by in situ hybridization or immunohistochemistry. Serological analyses were performed in longi-tudinally collected serum samples to detect antibodies against, PCV2, porcine reproductive and respira-tory syndrome virus (PRRSV), porcine parvovirus (PPV), swine influenza virus (SIV) and Lawsoniaintracellularis (law), Mycoplasma hyopneumoniae, Aujeszky’s disease virus (ADV) and Salmonella spp. ACox proportional hazards model was used to investigate the simultaneous effects of seroconversionand maternal immunity against the studied pathogens. Results showed that high levels of maternalimmunity against PCV2 had a protecting effect in farms from both countries. Moreover, for the Danishdataset, seroconversion against law had an overall protecting effect, but for animals with very low levelsof maternal antibody levels against this pathogen, the effect appeared neutral or aggravating. Otherwise,for the Spanish dataset, maternal immunity against PPV and PRRSV gave protective and aggravatingeffects, respectively. In conclusion, the present study reflects the complex interaction among differentpathogens and their effects in order to trigger PMWS in PCV2 infected pigs.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction cepted criteria to diagnose PMWS on individual basis include the

Postweaning multisystemic wasting syndrome (PMWS) is aworldwide spread disease that affects pigs in nursery and/orfattening units (Grau-Roma et al., 2011; Madec et al., 2008) and isconsidered to have a severe economic impact on swine production(Armstrong and Bishop, 2004). The main clinical sign of PMWS iswasting, but can also include pallor of the skin, icterus, respiratorydistress and diarrhoea (Harding, 1998). The internationally ac-

ll rights reserved.

i Anatomia Animals, Univer-lona, Spain.ma), joaquim.segales@cresa.

ip of this work.

presence of compatible clinical signs, moderate to severe lympho-cyte depletion with granulomatous inflammation in lymphoidtissues, and detection of moderate to high amount of porcinecircovirus type 2 (PCV2) within these lesions (Segalés et al., 2005;Sorden, 2000).

PCV2 is considered the essential infectious agent for PMWSdevelopment. However, the experimental reproduction of the se-vere clinical expression of the disease observed under field condi-tions has been difficult to achieve and only few studies have beenable to reproduce PMWS by inoculating PCV2 alone (Bolin et al.,2001; Lager et al., 2007; Okuda et al., 2003). Consequently, it hasbeen suggested that other infectious or non-infectious factors mustbe concomitantly influencing the development of PMWS in thefield. A meta-analysis on reported experimental infections con-cluded that one out of the most successful approaches to reproduce

1232 L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240

the disease was achieved when using a co-infection with anotherswine pathogen as a triggering factor (Tomás et al., 2008). Specifi-cally, the apparently most successful models comprise the infec-tion with PCV2 together with other swine pathogens, such asporcine parvovirus (PPV) (Allan et al., 1999; Krakowka et al.,2000), porcine reproductive and respiratory syndrome virus(PRRSV) (Allan et al., 2000; Harms et al., 2001; Rovira et al.,2002) or Mycoplasma hyopneumoniae (Opriessnig et al., 2004), orthe use of non-infectious immunomodulators like keyhole limpethemocyanin in incomplete Freund’s adjuvant (KLH-ICFA) (Graslandet al., 2005; Krakowka et al., 2001; Ladekjaer-Mikkelsen et al.,2002).

Accordingly to experimental data, several field studies haveshown a wide spectrum of infectious agents present concomitantlywith PCV2 infection in PMWS affected farms (Grau-Roma et al.,2011). However, no single other co-pathogen has been possibleto identify as the unique responsible of enhancing the severityand the incidence of PMWS. Most of epidemiological studies haveanalysed the disease at farm level (Alarcon et al., 2011; López-Soriaet al., 2005; Rose et al., 2003) and few data at individual level isavailable (Rose et al., 2009; Woodbine et al., 2011). Therefore,the aim of the present study was to gain further insight on infec-tious agents as potential risk factors for PMWS triggering at indi-vidual level. This work was carried out taking into accountmaternal immunity humoral levels and longitudinal antibody pro-files to different pathogens on a set of pigs that developed PMWSversus healthy age-matched animals in several PMWS affectedfarms from Denmark and Spain. The hypothesis behind this studywas that infection with and/or following seroconversion againstthe studied pathogens may influence the risk of developing PMWSon individual basis. Moreover, it was also hypothesized that the le-vel of maternal immunity against studied pathogens may also alterthe risk, either upwards or downwards, of PMWS occurrence atindividual level.

2. Materials and methods

2.1. Study design

Two prospective longitudinal studies in PMWS affected farms,one in Spain and one in Denmark, were performed as previouslydescribed (Grau-Roma et al., 2009). Briefly, both studies were car-ried out using similar designs, and were performed on 16 farmbatches (six in Spain and 10 in Denmark) of 100 to 154 animalsper batch, coming from 13 different farms (three from Spain and10 from Denmark). Ten to 21 sows were selected in order to reachat least 100 piglets per batch. Piglets showing any clinical defects(i.e., herniation, lameness, deformities) were not included in thestudy. The diagnosis of PMWS at the farm level was confirmedaccording to established criteria (Grau-Roma et al., 2012) beforethe start of the study. Diagnostic procedures included a prevalenceof pigs with wasting and mortality in nurseries plus fattening/fin-ishing areas higher than 10%, as well as the individual diagnosticcase definition fulfilment of PMWS (Segalés et al., 2005) in at leastone out of five necropsied pigs. Studied piglets were ear-tagged atone week of age and monitored until the occurrence of the PMWSoutbreak. Blood samples (Vacutainer�, Becton–Dickinson, MeylanCedex, France) from all pigs were serially collected from the cavacranial vein at established weeks of age (1, 3, 7 and 11 in Spain,and 1, 4, 6 and 9 in Denmark) and at the time of PMWS outbreakoccurrence (necropsy). Blood samples were allowed to clot at4 �C and then centrifuged at 1500g for 10 min at 4 �C. All serumsamples were frozen at �80 �C until testing. When the PMWS com-patible clinical picture (Segales and Domingo, 2002) appeared atthe studied farms, clinically healthy animals and pigs displaying

PMWS-like signs were selected, euthanized and necropsied. Theselection of the clinically healthy animals was not based on expo-sure to pathogens but their clinical status; therefore, it is assumedthat selected animals were representative for all clinically healthyanimals at the given time with respect to pathogen exposure. Eachbatch was followed until no more animals suffering from clinicalsigns compatible to PMWS were observed among the tagged ani-mals. At necropsy, sections of lymphoid tissues were collectedand fixed by immersion in neutral-buffered 10% formalin to assessthe pathological status of both clinically healthy and diseasedanimals.

All sows in the Spanish herds were vaccinated against Erysipelo-thrix rhusiopathiae and porcine parvovirus (PPV). In addition, pig-lets were vaccinated against Aujeszky disease virus (ADV) atearly fattening ages (11 and 14 weeks of age). Moreover, at thesame time, farm 3 from Spain (Sp-3, Table 1) vaccinated alsoagainst swine influenza virus (SIV). All sows in the Danish herdswere vaccinated against Erysipelothrix rhusiopathiae, PPV and Clos-tridium perfringens. Additionally, sows in herd Dk-4, DK-5, DK-7and DK-10 were vaccinated against Escherichia coli. Pigs in farmsDK-3, DK-5 and DK-6 were vaccinated against M. hyopneumoniaeat weaning. None of the farms included in this study vaccinatedagainst PCV2 in sows and piglets.

All treatments, housing, husbandry and slaughtering conditionsfollowed the European Union Guidelines and Good ClinicalPractices.

2.2. Histopathology

Formalin fixed tissues were dehydrated and embedded in paraf-fin blocks. Two consecutive 4 lm thick sections containing col-lected lymphoid tissues from each pig were cut from each block.One section was processed for haematoxylin and eosin stain, whilethe other was used for PCV2 nucleic acid detection by in situhybridization (ISH) (Rosell et al., 1999) for Spanish cases, and forimmunohistochemistry (IHC) for PCV2 antigen detection (Jensenet al., 2006) for Danish cases. Pathological evaluation and PMWSdiagnosis was carried out using a previously described scoring sys-tem evaluating the PCV2 amount and the intensity of lymphoiddepletion and granulomatous infiltration present in lymphoid tis-sues (Grau-Roma et al., 2009). Three different categories of pigswere established: (i) PMWS cases: corresponding to pigs showingclinical wasting and with moderate to severe lymphoid lesionsand moderate to high amount of PCV2 antigen/nucleic acid; (ii)wasted non-PMWS cases: corresponding to pigs showing clinicalwasting but without or with slight PMWS characteristic histopa-thological lesions and no or low amount of PCV2 antigen/nucleicacid within lymphoid tissues; (iii) healthy pigs: corresponding topigs showing good clinical condition, without or slight PMWS char-acteristic histopathological lesions and no or low amount of PCV2antigen/nucleic acid within lymphoid tissues.

2.3. Serology

Serological analyses were performed in all longitudinally col-lected serum samples from necropsied pigs belonging to one ofthe three categories described above. In Spain, several ELISAs wereused to detect antibodies against PCV2 (by immunoperoxidasemonolayer assay, IPMA) (Rodríguez-Arrioja et al., 2000), PRRSV(HerdCheck PRRS virus antibody test, IDEXX, Inc., USA), ADV (Herd-Check anti-PRV gpI, IDEXX, Inc., USA), PPV (Ingezim PPV, Ingenasa,Spain), SIV (Civtest Suis Influenza, Laboratorios Hipra, Spain),Lawsonia intracellularis (law) (Boesen et al., 2005), M. hyopneumo-niae (Civtest Suis Mycoplasma hyopneumoniae, Laboratorios Hipra,Spain), and Salmonella spp. (Salmonella covalent mix-ELISA, Svano-vir™, Svanova, Sweden). In Denmark, assays used were previously

Table 1Number of studied farms, batches and pigs, ages at pig necropsies (when PMWSoutbreak was observed) and pathological category of studied pigs.

Farm Batch Weeks of age at necropsy Number of necropsied pigs duringPMWS outbreak

PMWS Wasted non-PMWS Healthy

SP-1 SP-1 18–21 3 8 5

SP-2 SP-2a 15–17 6 5 4SP-2b 13–15 3 7 5SP-2c 11–15 6 5 5

SP-3 SP-3a 12–15 7 9 5SP-3b 12–15 10 9 5

Total number of studied animals inSpanish study

35 43 29

DK-1 DK-1 9–10 2 2 2DK-2 DK-2 10–13 5 3 3DK-3 DK-3 10–13 21 25 5DK-4 DK-4 10–13 9 3 3DK-5 DK-5 10–11 2 2 2DK-6 DK-6 10–13 7 8 4DK-7 DK-7 8–10 3 1 3DK-8 DK-8 9 4 2 3Dk-9 DK-9 11 1 0 1Dk-10 Dk-10 7–14 1 1 5

Total number of studied animals inDanish study

55 47 31

L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240 1233

described IPMAs to detect antibodies against European and Amer-ican PRRSV strains (Sorensen et al., 1998), ELISAs to detect antibod-ies against PCV2 (Kristensen et al., 2009), PPV (Madsen et al., 1997)and law (Boesen et al., 2005), and haemagglutination inhibition(HI) test to detect antibodies against H1N1 and H3N2 subtypesof SIV, which was essentially carried out according to the OIE man-ual (OIE manual of Diagnostic Tests and Vaccines for TerrestrialAnimals 2008).

2.4. Statistical analysis

2.4.1. Data descriptionDue to the fact that two different sets of techniques were used

to make the serological analysis in both countries and that some ofthe pathogens were studied only in one of them, Danish and Span-ish data sets were analysed separately. All serological results ob-tained in Denmark and PCV2 and law determinations in Spainwere expressed as antibody titres, which were log transformedprior to statistical analysis. On the other hand, antibodies to SIV,PRRSV, PPV and M. hyopneumoniae in Spain were expressed as opti-cal density (OD) values, while results from Salmonella spp. com-prised only positive or negative values. In Denmark, antibodiesagainst both American and European PRRSV strains were assessed.However, due to the potential cross-reactivity of the serologicaltests used (Sorensen et al., 1998), an animal infected with PRRSVwas considered to have either the European (PRRSVe) or American(PRRSVu) variant throughout its life, not both. Therefore, a herdwas considered to be infected only by the first detected antibodiesto PRRSVe or PRRSVu. The Danish covariate SIV was comprised ofH1N1 and H3N2, while it was a single variable in the Spanish data-set, since the test used in Spain detected antibodies against bothH1N1 and H3N2. The exact individual sampling dates were re-corded and used in the analysis.

Since PMWS-affected animals were intermingled with the restof pigs, it was assumed that the pattern of exposures in healthypigs tested was the same than those in the not tested healthy pigs.

Pigs classified as wasted non-PMWS constituted a heteroge-neous group of animals which was not easy to interpret

collectively (Grau-Roma et al., 2009). There is a wide range ofcauses, apart from PMWS, that can produce wasting in pigs (Har-ding, 1998). Thus, animals classified as wasted non-PMWS wereinitially removed from the analysis in order to focus on compari-sons between healthy and PMWS animals. Afterwards, a controlanalysis was run where wasting non-PMWS animals were includedas healthy animals (constituting a global group of non-PMWS af-fected pigs) and comparisons to the original analysis (includingexclusively PMWS affected and clinically healthy pigs) wereperformed.

All analyses were carried out using the Splus� software pack-age, version 6.1.

2.4.2. Model buildingA statistical model was built to investigate the effects of sero-

conversion and maternal immunity against the above mentionedanalysed pathogens in each country. The effects were representedby whether and when seroconversion against a pathogen did oc-cur, and maternal immunity against the pathogen. However, theexact moment of seroconversion was not observable and maternalantibody levels in the animals within the first 24 h of life were notpart of the data. As a consequence, the time for seroconversionoccurrence and the level of maternal immunity from the availabledata were estimated as described below.

2.4.3. Estimation of seroconversion time pointsOccurrence and age-time for seroconversion was estimated tak-

ing into account that the amount of maternal antibodies, if present,declined from first days of life onwards, and an increase (measuredby an antibody titre/OD value higher than the one of the precedingsampling) indicated that an infection and subsequent seroconver-sion had taken place. In this sense, for each ‘infected’ animal witha given pathogen, an estimate for the time of seroconversion (hereequalled to the time where antibody titre was assumed to switchfrom decrease to increase) was constructed. Any measurementsafter an increase that showed an antibody titre/OD value decliningagain were not considered. Moreover, if there were only two mea-surements to estimate the time of seroconversion, the midpointbetween the ages at observation for these two measurementswas selected as the seroconversion time. If three or more measure-ments were available, a second order polynomial was fitted to themeasurements as a function of age-time. If the fitted polynomialdid not attain a minimum, the midpoint between the first mea-surement that showed increase and the preceding observationwas selected. If the fitted polynomial did attain a minimum andthis minimum was positive, the age for which the minimum wasattained was selected. If the fitted polynomial did attain a mini-mum and this minimum was negative, the first time point afterthe time where the minimum was attained, and where the polyno-mial attained a positive value, was selected. The reason for the lastprocedure was to mimic a situation in which a minimum wasreached and the antibody titre/OD value was close to constantfor a period of time, for then to rise again due to seroconversion.The fitted polynomial constituted an estimated development inantibody titre/OD value, and the time-point where the minimumwas attained was exactly the point where this estimated develop-ment switches from decrease to increase.

The polynomial method was chosen for descriptive reasonsafter visual inspection of the longitudinal sequences of antibody ti-tres/OD values. For the Danish data, the polynomial method wasnot applied to the covariate PCV2, where the midpoint betweenthe first observation showing an increase of antibody titre andthe preceding was used as an estimate for the time of seroconver-sion; this was because increase in antibody titres seemed to hap-pen much faster than decrease, and the form of a 2nd orderpolynomial was therefore not suitable to fit the data. For the

Fig. 1. Example of a pig that seroconverted towards porcine parvovirus (PPV) (A) and did not seroconvert towards PCV2 (B). The vertical dotted line in (A) indicates theestimated time for seroconversion towards PPV.

1234 L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240

Spanish data, a method similar to the Danish PCV2 was applied tothe covariates SIV and Salmonella spp. For SIV it was due to thesame reason as above, and for Salmonella spp. due to the lack ofnumerical values for levels of antibodies. Examples of applicationof the polynomial method are shown in Fig. 1.

2.4.4. Estimation of levels of maternal immunityMaternal immunity was defined as the maximum value of anti-

body titre/OD values observed within the first three weeks of age.These data were not available for Salmonella spp. The possibility ofestimating maternal immunity through measurements on sowswas considered. However, despite antibody data being presentfor sows in the Danish data, cross-fostering practices invalidatedthe approach. In the Spanish data, sow data were available forPCV2 only. Thus, sow data could not properly represent maternalimmunity in any of the datasets.

2.4.5. Survival analysisThe data were analysed within a survival analysis framework,

where development of PMWS was considered as death/failure inthe survival context (case status as PMWS), while the lack of

20 40 0

PRRSV.eBirth

A

20 40 0

0.0

0.1

0.2

0.3

0.4

0.5

B

Haza

rd ra

tio

Fig. 2. (A) Age-times for birth, death and seroconversions against porcine respiratory anand Lawsonia intracellularis are marked for animal 208 from Denmark. (B) Longitudinalanimal with average maternal immunity.

PMWS diagnosis at necropsy (wasting-non-PMWS and healthypigs) was considered as censoring. The age of the animal was usedas the time variable, starting at day 0 and ending with the age atautopsy. A Cox proportional hazards model (Andersen et al.,1993) was used to estimate and test the effects of the covariates.Herd effects were included as nuisance parameters in the survivalmodels, as fixed effects. Many of the ten Danish herds did only con-tribute with few animals, while three herds contributed with morethan half of the animals. Due to this situation, it was not possible toestimate a random herd effect in the models applied. An investiga-tion of seroconversions as defined in Section 2.4.3 did not revealany of the major contributing herds as standing out compared tothe remainders. For the Spanish data, structural inhomogeneities(pathogens apparently present in more or less all animals or nonewithin a given herd, and some herds being included more thanonce) ruled out a randomised effect. The remaining covariates fallinto three categories: (a) maternal immunity, with an effect thatwas considered independent of time (age); (b) seroconversion sta-tus, which was considered as dependent of time; and (c) interac-tions within and between these two groups (seroconversion andmaternal immunity). No interaction between PRRSVe and PRRSVu

6 80 0

PPV Lawsonia Death/censoring

6 80 0

d reproductive syndrome virus European strain (PRRSVe), porcine parvovirus (PPV)development of hazard ratio in piglet 208 from Denmark, relative to a theoretical

L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240 1235

was modelled, as these were diagnosed on herd basis so that noanimal was modelled to possibly be infected with both pathogens.

The covariates corresponding to seroconversion were piece-wise constant in time. It took the value 0 from age-time 0 (birth),and until the animal had seroconverted for the given combinationof pathogens that comprised the covariate. From there, it changedvalue to 1, which it continued until autopsy. An exemplified se-quence of events is depicted in Fig. 2.

Separate analyses were carried out for the Danish and Spanishdatasets, based on the same Cox proportional hazards model. Thehazard function ki for the Cox model for the i’th individual is givenby the formula:

kiðtÞ ¼ k0ðtÞ expX

j

bjXijðtÞ !

; t > 0 ð1Þ

where k0 is a baseline hazard for a non-infected animal with nomaternal immunity, which is left unspecified, and where Xij iscovariate j for animal i, constructed as described in Sections 2.4.3and 2.4.4. The bj’s are the effect parameters. The effect parametersdo not depends on the level of maternal immunity that was as-sumed for the baseline hazard k0. As the baseline hazard k0 is leftunspecified, the effect parameters models the relative risk of thecovariate and not the absolute risk through the hazard function.For example, the hazard ratio exp(bPPV) is the risk that an animalwho has seroconverted towards PPV will develop PMWS duringan infinitesimal time interval, relative to an animal that has notseroconverted. However, the full risk cannot be modelled with thistechnique because k0 remains unspecified. If bPPV < 0, the hazard ra-tio is less than one (seroconverted animals had less risk than non-seroconverted ones). A value of bPPV > 0 represents the oppositesituation.

The proportional hazard assumption was checked with a good-ness of fit test following Grambsch and Therneau (1994) formula(8), applied to Kaplan–Meier transformed event times. Statisticalinference in the model corresponding to the intensities in formula1 was conducted with the Likelihood Ratio method. The test se-quence was found through the Akaike Information Criteria, andthe test statistics were evaluated through Chi-square-distributions.

Due to the relatively small number of animals in the analyseswhich made a simultaneous analysis of all covariates impossible,the analyses were performed in steps. First, all seroconversionsand interactions were included as covariates, and the model wasreduced to the statistical significant ones. Then, maternal immuni-ties and interactions with these were included through forwardselection. Last, all covariates that had been removed as insignifi-cant prior to or during inclusion of maternal immunity were in-cluded again through forward selection, and the model wasreduced. Finally, this three-step procedure was then iterativelyapplied until no changes to the model were observed. For eachpathogen, a single-term survival analysis was also carried out forboth seroconversion and maternal immunity, including only thissingle covariate and herd effect.

2.4.6. Sensitivity analysisIn order to contemplate the impact of the built-in impreciseness

of the estimates of seroconversions, a sensitivity analysis wascarried out. Gaussian noise was added to the seroconversion timesconsidering that 95% of the new seroconversion times should bewithin one week of the original estimates. Model reduction corre-sponding to the procedure described in Section 2.4.5 was thencarried out, giving a new, possibly different, model. This procedurewas repeated 20 times. In order to pass the sensitivity test, a covar-iate should appear in at least half of the resultant models. Other-wise, the significance of the covariate would be ascribed as anartefact. The ±1 week interval for the Gaussian noise was chosen

after considering the distance in time between longitudinal mea-surements, and the potential changes in seroconversion timesdue to this.

3. Results

3.1. Histopathology

The number of studied farms, batches, pigs, their pathologicalclassification and the range of age when PMWS appeared areshown in Table 1. One animal from Spain and two from Denmarkwithout clinical signs were not classified as healthy, because theyhad moderate lesions and amounts of PCV2 in lymphoid tissuescompatible to PMWS (Grau-Roma et al., 2009), and they werenot represented in the table.

3.2. Serology

The number and the percentage of seroconverted pigs accordingto the criteria described above are displayed in Table 2. Consider-ing all the animals included in the survival analysis (PMWS andhealthy), only PRRSVu in Denmark, SIV in Spain and PCV2 in bothcountries had percentages of seroconversion higher than 40%. Onthe other hand, the percentage of seroconverted pigs to PPV andPRRSV in Spain were lower than 10%. The percentage of serocon-verted pigs to PRRSVe, law, SIV and PPV in Denmark and to Salmo-nella spp. and law in Spain was between 10 and 40%. Finally, noneof the studied animals showed signs of seroconversion to ADV andM. hyopneumoniae.

Mean and standard deviation (SD) of estimated level of mater-nal immunity towards the analyzed pathogens are displayed in Ta-ble 3.

3.3. Statistical analysis

One PMWS pig from the Danish dataset was excluded from theanalysis due to inconsistent serological determinations. Three pigswith moderate lesions and moderate amounts of PCV2 but withoutapparent clinical signs (two from Denmark and one from Spain)were also excluded from the statistical analysis. Therefore, survivalanalysis was performed comparing 55 PMWS versus 31 healthypigs in Denmark, and 35 PMWS versus 29 healthy pigs in Spain.M. hyopneumoniae and ADV were not included, because no evi-dence of seroconversions was observed against those pathogens.The potential effect of vaccinations in piglets from Spain and Den-mark was considered, but was found to be not significant. On theother hand, the potential effect of ADV vaccination in Spanish pig-lets could not be assessed, since all pigs received such vaccine.

The statistically significant factors in the survival models forDanish data were seroconversion against law and PRRSVe, andmaternal immunity against PCV2 and law. However, seroconver-sion against PRRSVe did not pass the sensitivity analysis and wasconsequently not included in the final model. For the Spanish data,the statistically significant factors were maternal immunity againstlaw, PCV2, PPV, PRRSV and SIV. Because no seroconversions werestatistically significant, no sensitivity analysis was performed.The effect parameters in the final models for Spain and Denmark,corrected for herd effects, are shown in Table 4.

For those factors that interacts with maternal immunities (lawin the Danish model, and all maternal immunities in the Spanishmodel), the effect was not obvious from the effect parameters, be-cause the effect can vary with the specific maternal immunity levelfor the animal. To attain an overall classification of such a covari-ate, an infinitesimal linear effect coefficient for the covariate atits mean (describing the effect of perturbations around the mean

Table 2Serology results against studied pathogens in Denmark (A) and Spain (B). Results are indicated as the number of seroconverted pigs at euthanasia during the PMWS outbreak withrespect to the total of studied pigs. Percentage is given between brackets.

PCV2 PPV PRRSVe PRRSVu SIV_H1N1 SIV_H3N2 SIV* Law

(A)PMWS 27/55 15/55 5/55 27/55 6/55 3/55 8/55 7/55

(49) (27) (9) (49) (11) (6) (15) (13)Healthy 11/31 15/31 5/31 15/31 7/31 1/31 8/31 7/31

(36) (48) (16) (48) (23) (3) (26) (23)Total (PMWS + Healthy) 38/86 30/86 10/86 42/86 13/86 4/86 16/86 19/133

(44) (35) (12) (49) (15) (5) (19) (14)Wasted non-PMWS 22/47 14/47 6/47 21/47 1/47 1/47 2/47 5/47

(47) (30) (13) (45) (2) (2) (5) (11)

law PCV2 PPV PRRSV SIV ADV Salm Myco

(B)PMWS 8/35 28/35 2/35 1/35 20/35 0/35 8/35 0/35

(23) (80) (6) (3) (57) (0) (23) (0)Healthy 7/29 26/29 3/29 2/29 22/29 0/29 9/29 0/29

(24) (90) (10) (9) (76) (0) (31) (0)

Total (PMWS + Healthy) 15/64 54/64 5/64 3/64 42/64 0/64 17/64 0/64(23) (87) (8) (5) (66) (0) (27) (0)

Wasted non-PMWS 8/43 42/43 2/43 1/43 33/43 0/43 19/43 0/43(19) (98) (5) (2) (77) (0) (44) (0)

Porcine parvovirus (PPV), porcine reproductive and respiratory virus (PRRSV), Aujeszky disease virus (ADV), swine influenza virus (SIV), Salmonella spp. (salm), Mycoplasmahyopneumoniae (Myc), porcine circovirus type 2 (PCV2), SIV subtype H1N1 and H3N2 (SIV_H1N1, SIV_H3N2) European and American PRRSV strains (PRRSVe, PRRSVu),Lawsonia intracellularis (law).* Number of pigs that seroconverted to SIV H1N1 and/or SIV H3N2.

Table 3Estimated level of maternal immunity towards the studied pathogens in Denmark (A) and Spain (B). Antibody titres are presented as log(antibody titre) or optical density (OD)values.

log(PCV2) log(PPV) log(PRRSVe) log(PRRSVu) log(SIV) log(Law)

(A)PMWS 6.31 ± 2.09 9.19 ± 2.16 5.33 ± 2.58 5.24 ± 2.37 4.20 ± 2.87 3.16 ± 2.06Healthy 7.24 ± 3.47 9.30 ± 2.45 4.95 ± 1.82 5.53 ± 2.03 4.07 ± 2.15 2.92 ± 2.04

log(PCV2) log(PPV) PRRSV SIV log(law)

(B)PMWS 6.64 ± 2.32 7.00 ± 1.01 0.64 ± 1.71 1.08 ± 0.44 3.18 ± 2.19Healthy 7.28 ± 2.60 7.16 ± 1.01 0.73 ± 1.50 1.07 ± 0.55 3.45 ± 1.63

Porcine parvovirus (PPV), porcine reproductive and respiratory virus (PRRSV), swine influenza virus (SIV), porcine circovirus type 2 (PCV2), Lawsonia intracellularis (law).

Table 4Parameters estimated for the Danish (A) and Spanish (B) models, corrected for herdeffects. mat.law refers to antibody levels for maternal immunity against lawsonia,with similar notation for other pathogens. Interaction factors are represented by theproduct of the factors interacting, and a factor with a superscript of 2 represents effectof the square of the original factor.

Covariate Estimated b ± 1.96SE p*

(A)Law 10.322 ± 7.10 0.002log(mat.pcv2) �0.561 ± 0.26 <0.0001log(mat.law) �4.02 ± 2.73 0.0005**

Covariate Estimated b ± 1.96SE p*

(B)log(mat.law) �11.46 ± 6.49 0.002log(mat.pcv2) 7.26 ± 7.22 0.007log(mat.pcv2)2 �0.72 ± 0.60 0.008log(mat.ppv) 11.29 ± 6.49 <0.0001mat.prrsv 11.08 ± 5.97 0.0001mat.siv 64.87 ± 47.39 <0.0001log(mat.law): log(mat.pcv2) 0.64 ± 0.60 0.03log(mat.law): mat.prrsv �2.64 ± 1.46 0.0003log(mat.law): mat.siv 7.94 ± 4.58 0.0008log(mat.ppv): mat.siv �13.46 ± 6.87 <0.0001

* Tests of main effects includes removal of interaction terms.** The effect of mat.law extends to lawsonia sero-converted animals only.

1236 L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240

of when all other covariates are kept at their respective means)was calculated and labelled as the index I. The index was calcu-lated as the derivative of the log of the hazard function (1) with re-spect to the given covariate and subsequent empirical expectation,as follows:

ðaÞ IðlawjDanishÞ ¼ blaw þ blaw:mat:law�meanðlogðmat:lawÞÞ ð2Þ

ðbÞ Iðmat:lawjSpanishÞ ¼ bmat:law

þ bmat:law:mat:pcv2�meanðlogðmat:pcv2ÞÞþ bmat:law:mat:prrsv�meanðmat:prrsvÞþ bmat:law:mat:siv�meanðmat:sivÞ

ðcÞ Iðmat:pcv2jSpanishÞ ¼ bmat:pcv2

þ 2bmat:pcv2:2�meanðlogðmat:pcv2ÞÞþ bmat:pcv2:mat:siv�meanðmat:sivÞ

ðdÞ Iðmat:ppvjSpanishÞ ¼ bmat:ppv þ bmat:ppv:mat:siv�meanðmat:sivÞ

ðeÞ Iðmat:prrsvjSpanishÞ ¼ bmat:prrsv þ bmat:prrsv:mat:siv�meanðmat:sivÞ

ðfÞ Iðmat:sivjSpanishÞ ¼ bmat:siv

þ bmat:pcv2:mat:siv�meanðlogðmat:pcv2ÞÞ;

Table 5Effect parameters/index adjusted for interaction effects, for animals with average maternal immunity against the involved pathogens (see Index in methods section).

Pathogen type Covariate type Calculated index ± 1.96SE Hazard ratio* (CI) p (Chisq)

I (law|Danish) Seroconversion �1.45 ± 1.44 0.23 (0.06;0.99) 0.05I (mat.law|Spanish) Maternal immunity �0.29 ± 0.64 0.75 (0.39;1.42) 0.63I (mat.PCV2|Spanish) Maternal immunity �2.75 ± 1.05 0.06 (0.02;0.18) <0.0001I (mat.PPV|Spanish) Maternal immunity �3.35 ± 1.80 0.04 (0.01;0.21) <0.0001I (mat.PRRSV|Spanish) Maternal immunity 2.32 ± 1.23 10.18 (2.97;34.81) 0.0002I (mat.SIV|Spanish) Maternal immunity �4.15 ± 4.14 0.02 (0.00;0.99) 0.05

* For continuous covariates, the hazard ratio is per increase of 1.

L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240 1237

where the mean in (a) extents to data for animals seroconvertingagainst law only, while the means in (b–f) extends to all animals.The symbol mat.pcv2.2 refers to the squared effects of log(mat.pcv2) (Table 4). For factors not interacting with maternal immuni-ties, the index I was equalled to the effect parameter, shown inTable 4. The results of the indexes in (a–f) are shown in Table 5.Positive and negative index indicate aggravating and protectingeffects on ‘average animals’, respectively (formula (1)), for thedevelopment of PMWS.

While the index I for maternal immunities was dependent onthe chosen scale for the covariate just as the coefficients in Table4, the signs were similarly not, and the sign of the index I indicatedpositive/negative average effects in the same way as the coeffi-cients b in Table 4. The final results of the survival analysis andthe single-term survival analyses are presented in Table 6, wherean indication of significance or not and type of effect of each covar-iate are listed for both analyses.

For the Danish analysis, Table 6 shows that seroconversionagainst law had a protecting effect. For animals seroconvertedagainst law, the protective effect increased with the level of mater-nal immunity against law. High maternal immunity against PCV2had a protecting effect. For the Spanish analysis, Table 6 shows thathigh maternal immunity against PCV2, maternal immunity againstPPV and maternal immunity against SIV had a protective effect.Maternal immunity against PRRSV had an aggravating effect. No

Table 6Statistical significances and protecting/aggravating effects in the survival model andthe marginal models in Denmark (A) and Spain (B).

Covariate Survival analysis Single-term analysis

(A)Seroconversion law Protecting Not significantSeroconversion PCV2 Not significant Not significantSeroconversion PPV Not significant Not significantSeroconversion SIV Not significant Not significantSeroconversion PRRSVe Not significant Not significantSeroconversion PRRSVu Not significant Not significantMaternal law Protecting* Not significantMaternal PCV2 Protecting ProtectingMaternal PPV Not significant Not significantMaternal SIV Not significant AggravatingMaternal PRRSVe Not significant Not significantMaternal PRRSVu Not significant Not significant

Covariate Survival analysis Marginal model

(B)Seroconversion PCV2 Not significant Not significantSeroconversion PPV Not significant Not significantSeroconversion SIV Not significant Not significantSeroconversion PRRSV Not significant Not significantSeroconversion Salmonella Not significant Not significantMaternal law Not significant Not significantMaternal PCV2 Protecting Not significantMaternal PPV Protecting Not significantMaternal PRRSV Aggravating AggravatingMaternal SIV Protecting Not significant

* Applied to Lawsonia intracellularis seroconverted animals only.

significant average effect was found for maternal immunity againstlaw.

Fig. 2 shows an example of how the statistical model worked.Seroconversions for the Danish animal 208 (healthy) are listed asage of the animal progresses (Fig. 2A). The corresponding longitu-dinal development of the hazard ratio for this animal, relative to ananimal with average maternal immunity, is shown in Fig. 2B. Inthis example, the hazard ratio did not vary after seroconversionagainst PRRSVe and PPV (non-significant covariates) but decreasedafter seroconversion against law.

Control analyses including wasted non-PMWS pigs as healthyanimals showed results similar to Table 6, with two exceptions:the effect for of maternal immunity against PRRSV (Spanish data)could not be found, and the maternal immunity against law couldno not be found (Danish data). The sensitivity analysis for the Dan-ish model rejected the effect of seroconversion against PRRSVe, butconfirmed the effects of seroconversion against law, and maternalimmunity towards PCV2 and law. The results were consistent inthe sense that already after the first 10 repeats of the sensitivityprocedure, the results were as described and did not change inthe last 10 out of the 20 repeats.

4. Discussion

PMWS is defined as a multifactorial disease (Grau-Roma et al.,2011; Madec et al., 2008). Since a number of years, the assessmentof the risk factors linked to disease occurrence has been a matter ofresearch focus (López-Soria et al., 2005; Rose et al., 2009, 2003;Woodbine et al., 2007). Moreover, different methodologies can beapplied to investigate risk factors for a given disease. In the presentstudy, survival analysis was applied on two sets of longitudinaldata on PMWS occurrence in Denmark and Spain, including allcovariates. At the same time, a single-term analysis for each path-ogen was also carried out for both seroconversion and maternalimmunity, including the nuisance herd effect and only a singlecovariate. The comparison between both survival and single-termanalysis results showed that, in several cases, the single-term anal-yses did not detect the effects found in the survival analysis, illus-trating that the complexity of the problem demands a multivariateapproach. However, when factors were significant in both types ofanalyses, the effects were similar. The Cox proportional hazardsmodel was chosen partly because it was adequate for the longitu-dinal results of the data collection, but also because it is a widelyused model in clinical studies in human and veterinary medicine(Ahmed et al., 2007; Andersen et al., 1993). It has the advantageof not being based on any assumptions concerning the nature orshape of the underlying survival distribution, and it appropriatelycorrects for the fact that animals not showing signs of PMWS atthe end of the study are censored here rather than consideredhealthy. Very few assumptions are made about the hazard func-tion, and this makes it likely to describe a large number of scenar-ios. There is only one previous study using survival analysis on thestudy of PMWS (Rose et al., 2009). In this work, PMWS diagnosiswas established at herd level and individual status was assigned

1238 L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240

based on clinical data (Rose et al., 2009). In the present work, allpigs were individually diagnosed as PMWS, wasted non-PMWSand healthy cases, as previously described (Grau-Roma et al.,2009). However, the fact that we only recommend the sign of aneffect as the proper outcome of the study (Table 6) is also a conse-quence of the study design; only a selection of the clinicallyhealthy animals followed were included in the basis for the statis-tical analysis due to financial considerations. This would poten-tially bias effect parameters, but would not expectedly reversethe direction of these, since the selected animals were representa-tive of the study population clinically healthy at time of censoring.

One of the limitations of the present study is that it was basedon the presence of antibodies against the studied pathogens, andnot on the detection of the pathogens directly. Therefore, if a givenseroconversion did not occur, the infection could not be ruled out,since pigs may have been euthanized before seroconversion wouldhave been detected. This is the reason because results derived fromthe present study must be considered carefully. This situation mayhave been circumvented if PCR-longitudinal data would have beenused instead of serology, but it was not affordable from economicaland technical points of view.

Maternal immunity against PCV2 in both Danish and Spanishdata showed a protecting effect against PMWS development,which is in agreement with previous works (Calsamiglia et al.,2007; Grau-Roma et al., 2009; López-Soria et al., 2005; Rodrí-guez-Arrioja et al., 2002; Rose et al., 2009). Therefore, it is expectedthat the higher the PCV2 antibody level in colostrum, the lower thelikelihood of developing PMWS in a context of non-vaccinated pigs.This situation would support the interest of PCV2 vaccination insows in order to transfer a higher load of PCV2 antibodies to theoffspring (Kekarainen et al., 2010).

For the Danish data, the sensitivity analysis confirmed all signif-icant factors identified by the survival analysis except seroconver-sion against PRRSVe. However, if different criteria for passing thesensitivity analysis were chosen, such as appearance in 90% ofthe models in the sensitivity repeats instead of 50%, only serocon-version against PCV2 and maternal immunity against PCV2 wouldpass. This shows that the plus/minus one week window for theGaussian noise in the sensitivity analysis has an impact on themodel. The fact that the effect of seroconversion against PRRSVedid not show up in more than half of the models when the serocon-version time point was perturbed within plus/minus one week,leads us to describe the statistical significance as an artefact, the ef-fect being contradicted by the sensitivity of the model to the esti-mated seroconversion time. On the other hand, the fact that all ofthe remaining factors did appear in more than half of the models inthe sensitivity repeats, consistently in both the first and the secondhalf of the repeats, supports the conclusion of significant impacts.

The obtained Danish model showed a significant interaction be-tween seroconversion against law and law maternal immunity.This interaction complicates the interpretation of the effect oflaw, because the sign (aggravating or protecting) of seroconversionagainst law depended on the level of maternal immunity. By com-bining the values for law from Tables 3A and 4A, it would appearthat pigs with moderate to high levels of maternal antibodiesagainst law, seroconversion would have a protective effect, whileanimals with a low levels of maternal immunity against law, lawseroconversion would have detrimental effects. The formal turningpoint from the tables is a (log) maternal immunity of 10.322/4.02 = 2.57, but this value should be considered in the light ofthe parameter uncertainty from Table 4A and the general reserva-tions on the effect of the study design. From a biological point ofview, as mentioned above, it should be emphasized that the effectdetected in this analysis does not refer directly to infection withlaw, but merely to the fact that seroconversion took place. At the

same time, it is also widely reported that PMWS affected animalsare severely immunocompromised (Darwich et al., 2004; Segaléset al., 2004). Thus, seroconversion may only happen in animalswith an intact immune system, which in turn will make the animalless likely to develop PMWS. Thus, only healthy pigs would becapable of mounting a significant immune response against law.However, for animals with very low maternal immunity againstlaw, a seroconversion may not be so beneficial according to themodel. Therefore, the time of law infection seemed to be importantfor the impact on PMWS development in that early law infections(which would be more probably present in animals with lowmaternal immunity) would act as a triggering factor for PMWS,while later law infection (which would be more probably presentin animals with average or higher maternal immunity) does notrepresent a risk for PMWS development. This idea is supportedby the fact that low maternal immunity was associated with earlyseroconversion against law and high frequency of PMWS cases.These data supports the previous reported experimental evidencesindicating that co-infections by other infectious agents differentfrom PCV2 may be an important factor for PMWS development (Al-lan et al., 2004; Tomás et al., 2008). This is the first report that sug-gests law as a potential trigger of PMWS, even it has not beenproved experimentally (Opriessnig et al., 2011).

For the Spanish data, maternal PPV resulted also as a protectingeffect against PMWS development. Therefore, maternal PPV antibod-ies would protect against PPV infection, which has been suggested asa triggering factor for PMWS development (Allan et al., 1999). Thissituation would be expectable since antibodies against PPV in sowsare mainly elicited by vaccination and, therefore, the likelihood ofnatural infection at early ages by this virus would be limited.

On the other hand, PRRSV is a known trigger of PMWS (Allanet al., 2000; Harms et al., 2001; Rovira et al., 2002), thus the findingthat maternal PRRSV had an aggravating effect in the Spanish datawould not be surprising since high level of maternal antibodies insows may be indicative of circulating PRRSV in the herd. However,obtained data may be difficult to support by the fact that only fewpigs were infected with this virus during the experimental periodin the Spanish farms.

For SIV, maternal antibodies were found to have an aggravatingeffect in the single-term model for the Danish data and a protect-ing effect in the survival analysis for the Spanish data. These find-ings are not interpretative from a biological point of view; howeverit is reasonable to expect that SIV may have an aggravating effecton PMWS development in line with other viruses such as PPVand PRRSV (Tomás et al., 2008).

It should be noted that the Spanish data showed a distinct inho-mogeneity regarding seroconversions against the studied patho-gens, since the major part of the animals either seroconverted ornot against a given pathogen. This scenario made statistical signif-icances highly fragile to individual results. For this reason, care wastaken when generalizing the results of the Spanish analysis. How-ever, the Spanish animals can be sorted after the hazard ratio withrespect to an animal with average immunity, because no timedependent covariates entered into the final model. Sorting theSpanish animals in three groups after hazard ratio values indicatedthat the full survival analysis was in agreement with the observeddata. Specifically, the group with the lowest hazard ratios had aPMWS frequency of 27.7%, while the middle group had a frequencyof 50.0%, and the group with the highest hazard ratios had a PMWSfrequency of 78.9%. Thus, the higher the risk, the higher the fre-quency of PMWS cases.

The lack of seroconversion against M. hyopneumoniae and ADVruled out the potential relation of those pathogens with PMWSdevelopment in the Spanish analysed pigs. However, this resultshow only that these pathogens were not involved in PMWS

L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240 1239

development in the studied farms, but does not rule out their rela-tion with PMWS occurrence. In fact, infection with these pathogenshas been experimentally (Opriessnig et al., 2004) or naturally (Rod-ríguez-Arrioja et al., 1999) linked to PMWS development.

In summary, maternal PCV2 antibodies in both Danish andSpanish data showed a protecting effect against PMWS develop-ment. The protective effect of maternal immunity against PPVand the aggravating effect of maternal PRRSV are in agreementwith previous works suggesting these viruses as a co-infectioustrigger for PMWS (Allan et al., 1999; Harms et al., 2001; Roviraet al., 2002). Besides, the fact that seroconversion against law re-sulted protective for PMWS development may reflect a well-devel-oped immune system, which in turn would make the animal lesslikely to develop PMWS. In any case, as stated before, the fact thatthis work used indirect determinations of pathogens (serology)makes that this later results must be considered with precaution.Similar studies using PCR-longitudinal data would be useful to fur-ther confirm present results.

Acknowledgement

This work was funded by the Project No. 513928 from the SixthFramework Programme of the European Commission.

References

Ahmed, F.E., Vos, P.W., Holbert, D., 2007. Modeling survival in colon cancer: amethodological review. Molecular Cancer 6, 15.

Alarcon, P., Velasova, M., Mastin, A., Nevel, A., Stark, K.D., Wieland, B., 2011. Farmlevel risk factors associated with severity of post-weaning multi-systemicwasting syndrome. Preventive Veterinary Medicine 101, 182–191.

Allan, G.M., Kennedy, S., McNeilly, F., Foster, J.C., Ellis, J.A., Krakowka, S.J., Meehan,B.M., Adair, B.M., 1999. Experimental reproduction of severe wasting disease byco-infection of pigs with porcine circovirus and porcine parvovirus. Journal ofComparative Pathology 121, 1–11.

Allan, G.M., McNeilly, F., Ellis, J., Krakowka, S., Botner, A., McCullough, K., Nauwynck,H., Kennedy, S., Meehan, B., Charreyre, C., 2004. PMWS: experimental modeland co-infections. Veterinary Microbiology 98, 165–168.

Allan, G.M., McNeilly, F., Ellis, J., Krakowka, S., Meehan, B., McNair, I., Walker, I.,Kennedy, S., 2000. Experimental infection of colostrum deprived piglets withporcine circovirus 2 (PCV2) and porcine reproductive and respiratory syndromevirus (PRRSV) potentiates PCV2 replication. Archives of Virology 145, 2421–2429.

Andersen, P.K., Borgan, Ø., Gill, R., Keiding, N., 1993, Statistical models based oncounting processes. Springer-New York, 477,539 pp.

Armstrong, D., Bishop, S.C., 2004. Does genetics or litter effect influence mortatily inPMWS. In: Armstrong, D. (Ed.), Proceeding of International Pig VeterinarySociety Congress, pp. 62.

Boesen, H.T., Jensen, T.K., Jungersen, G., Riber, U., Boye, M., Moller, K., 2005.Development, characterization and diagnostic application of a monoclonalantibody specific for a proteinase K resistant Lawsonia intracellularis antigen.Veterinary Microbiology 105, 199–206.

Bolin, S.R., Stoffregen, W.C., Nayar, G.P., Hamel, A.L., 2001. Postweaningmultisystemic wasting syndrome induced after experimental inoculation ofcesarean-derived, colostrum-deprived piglets with type 2 porcine circovirus.Journal of Veterinary Diagnostic Investigation 13, 185–194.

Calsamiglia, M., Fraile, L., Espinal, A., Cuxart, A., Seminati, C., Martin, M., Mateu, E.,Domingo, M., Segalés, J., 2007. Sow porcine circovirus type 2 (PCV2) statuseffect on litter mortality in postweaning multisystemic wasting syndrome(PMWS). Research in Veterinary Science 82, 299–304.

Darwich, L., Segalés, J., Mateu, E., 2004. Pathogenesis of postweaning multisystemicwasting syndrome caused by Porcine circovirus 2: an immune riddle. Archivesof Virology 149, 857–874.

Grambsch, P., Therneau, T., 1994. Proportional hazards tests and diagnostics basedon weighted residuals. Biometrika 81, 515–526.

Grasland, B., Loizel, C., Blanchard, P., Oger, A., Nignol, A.C., Bigarre, L., Morvan, H.,Cariolet, R., Jestin, A., 2005. Reproduction of PMWS in immunostimulated SPFpiglets transfected with infectious cloned genomic DNA of type 2 porcinecircovirus. Veterinary Research 36, 685–697.

Grau-Roma, L., Baekbo, P., Rose, N., Wallgren, P., Fraile, L., Larsen, L.E., Segalés, J.,2012. Clinical and laboratory studies on postweaning multisystemic wastingsyndrome affected herds in Denmark, France, Spain, and Sweden: diseaseprogression and a proposal of herd case definition. Journal of Swine Health &Production 20 (3), 129–136.

Grau-Roma, L., Fraile, L., Segalés, J., 2011. Recent advances in the epidemiology,diagnosis and control of diseases caused by porcine circovirus type 2. TheVeterinary Journal 187, 23–32.

Grau-Roma, L., Hjulsager, C.K., Sibila, M., Kristensen, C.S., López-Soria, S., Enoe, C.,Casal, J., Botner, A., Nofrarias, M., Bille-Hansen, V., Fraile, L., Baekbo, P., Segalés,J., Larsen, L.E., 2009. Infection, excretion and seroconversion dynamics ofporcine circovirus type 2 (PCV2) in pigs from post-weaning multisystemicwasting syndrome (PMWS) affected farms in Spain and Denmark. VeterinaryMicrobiology 135, 272–282.

Harding, J., 1998. Postweaning multisystemic wasting syndrome: epidemiology andclinical presentation. Journal of Swine Health & Production 6, 249–254.

Harms, P.A., Sorden, S.D., Halbur, P.G., Bolin, S.R., Lager, K.M., Morozov, I., Paul, P.S.,2001. Experimental reproduction of severe disease in CD/CD pigs concurrentlyinfected with type 2 porcine circovirus and porcine reproductive andrespiratory syndrome virus. Veterinary Pathology 38, 528–539.

Jensen, T.K., Vigre, H., Svensmark, B., Bille-Hansen, V., 2006. Distinction betweenporcine circovirus type 2 enteritis and porcine proliferative enteropathy causedby Lawsonia intracellularis. Journal of Comparative Pathology 135, 176–182.

Kekarainen, T., McCullough, K., Fort, M., Fossum, C., Segalés, J., Allan, G.M., 2010.Immune responses and vaccine-induced immunity against Porcine circovirustype 2. Veterinary Immunology and Immunopathology 136, 185–193.

Krakowka, S., Ellis, J.A., McNeilly, F., Ringler, S., Rings, D.M., Allan, G., 2001.Activation of the immune system is the pivotal event in the production ofwasting disease in pigs infected with porcine circovirus-2 (PCV-2). VeterinaryPathology 38, 31–42.

Krakowka, S., Ellis, J.A., Meehan, B., Kennedy, S., McNeilly, F., Allan, G., 2000. Viralwasting syndrome of swine: experimental reproduction of postweaningmultisystemic wasting syndrome in gnotobiotic swine by coinfection withporcine circovirus 2 and porcine parvovirus. Veterinary Pathology 37, 254–263.

Kristensen, C.S., Baekbo, P., Bille-Hansen, V., Botner, A., Vigre, H., Enoe, C., Larsen,L.E., 2009. Induction of porcine post-weaning multisystemic wasting syndrome(PMWS) in pigs from PMWS unaffected herds following mingling with pigsfrom PMWS-affected herds. Veterinary Microbiology 138, 244–250.

Ladekjaer-Mikkelsen, A.S., Nielsen, J., Stadejek, T., Storgaard, T., Krakowka, S., Ellis, J.,McNeilly, F., Allan, G., Botner, A., 2002. Reproduction of postweaningmultisystemic wasting syndrome (PMWS) in immunostimulated and non-immunostimulated 3-week-old piglets experimentally infected with porcinecircovirus type 2 (PCV2). Veterinary Microbiology 89, 97–114.

Lager, K.M., Gauger, P.C., Vincent, A.L., Opriessnig, T., Kehrli Jr., M.E., Cheung, A.K.,2007. Mortality in pigs given porcine circovirus type 2 subgroup 1 and 2 virusesderived from DNA clones. Veterinary Record 161, 428–429.

López-Soria, S., Segalés, J., Rose, N., Vinas, M.J., Blanchard, P., Madec, F., Jestin, A.,Casal, J., Domingo, M., 2005. An exploratory study on risk factors forpostweaning multisystemic wasting syndrome (PMWS) in Spain. PreventiveVeterinary Medicine 69, 97–107.

Madec, F., Rose, N., Grasland, B., Cariolet, R., Jestin, A., 2008. Post-weaningmultisystemic wasting syndrome and other PCV2-related problems in pigs: a12-year experience. Transboundary and Emerging Diseases 55, 273–283.

Madsen, E.S., Madsen, K.G., Nielsen, J., Jensen, M.H., Lei, J.C., Have, P., 1997.Detection of antibodies against porcine parvovirus nonstructural protein NS1may distinguish between vaccinated and infected pigs. Veterinary Microbiology54, 1–16.

Okuda, Y., Ono, M., Yazawa, S., Shibata, I., 2003. Experimental reproduction ofpostweaning multisystemic wasting syndrome in cesarean-derived, colostrum-deprived piglets inoculated with porcine circovirus type 2 (PCV2): investigationof quantitative PCV2 distribution and antibody responses. Journal of VeterinaryDiagnostic Investtigation 15, 107–114.

Opriessnig, T., Madson, D.M., Roof, M., Layton, S.M., Ramamoorthy, S., Meng, X.J.,Halbur, P.G., 2011. Experimental reproduction of porcine circovirus type 2(PCV2)-associated enteritis in pigs infected with PCV2 alone or concurrentlywith Lawsonia intracellularis or Salmonella typhimurium. Journal of ComparativePathology 145 (2–3), 261–270.

Opriessnig, T., Thacker, E.L., Yu, S., Fenaux, M., Meng, X.J., Halbur, P.G., 2004.Experimental reproduction of postweaning multisystemic wasting syndrome inpigs by dual infection with Mycoplasma hyopneumoniae and porcine circovirustype 2. Veterinary Pathology 41, 624–640.

Rodríguez-Arrioja, G.M., Segalés, J., Balasch, M., Rosell, C., Quintana, J., Folch, J.M.,Plana-Duran, J., Mankertz, A., Domingo, M., 2000. Serum antibodies to porcinecircovirus type 1 and type 2 in pigs with and without PMWS. Veterinary Record146, 762–764.

Rodríguez-Arrioja, G.M., Segalés, J., Calsamiglia, M., Resendes, A.R., Balasch, M.,Plana-Duran, J., Casal, J., Domingo, M., 2002. Dynamics of porcine circovirustype 2 infection in a herd of pigs with postweaning multisystemic wastingsyndrome. American Journal of Veterinary Research 63, 354–357.

Rodríguez-Arrioja, G.M., Segalés, J., Rosell, C., Quintana, J., Ayllon, S., Camprodon, A.,Domingo, M., 1999. Aujeszky’s disease virus infection concurrent withpostweaning multisystemic wasting syndrome in pigs. Veterinary Record 144,152–153.

Rose, N., Eveno, E., Grasland, B., Nignol, A.C., Oger, A., Jestin, A., Madec, F., 2009.Individual risk factors for Post-weaning Multisystemic Wasting Syndrome(PMWS) in pigs: a hierarchical Bayesian survival analysis. Preventive VeterinaryMedicine 90, 168–179.

Rose, N., Larour, G., Le Diguerher, G., Eveno, E., Jolly, J.P., Blanchard, P., Oger, A., LeDimna, M., Jestin, A., Madec, F., 2003. Risk factors for porcine post-weaningmultisystemic wasting syndrome (PMWS) in 149 French farrow-to-finish herds.Preventive Veterinary Medicine 61, 209–225.

Rosell, C., Segalés, J., Plana-Duran, J., Balasch, M., Rodríguez-Arrioja, G.M., Kennedy,S., Allan, G.M., McNeilly, F., Latimer, K.S., Domingo, M., 1999. Pathological,immunohistochemical, and in-situ hybridization studies of natural cases of

1240 L. Grau-Roma et al. / Research in Veterinary Science 93 (2012) 1231–1240

postweaning multisystemic wasting syndrome (PMWS) in pigs. Journal ofComparative Pathology 120, 59–78.

Rovira, A., Balasch, M., Segalés, J., García, L., Plana-Duran, J., Rosell, C., Ellerbrok, H.,Mankertz, A., Domingo, M., 2002. Experimental inoculation of conventional pigswith porcine reproductive and respiratory syndrome virus and porcinecircovirus 2. Journal of Virology 76, 3232–3239.

Segalés, J., Allan, G.M., Domingo, M., 2005. Porcine circovirus diseases. AnimalHealth Research Reviews 6, 119–142.

Segalés, J., Domingo, M., 2002. Postweaning multisystemic wasting syndrome(PMWS) in pigs. A review. Veterinary Quarterly 24, 109–124.

Segalés, J., Domingo, M., Chianini, F., Majo, N., Domínguez, J., Darwich, L., Mateu, E.,2004. Immunosuppression in postweaning multisystemic wasting syndromeaffected pigs. Veterinary Microbiology 98, 151–158.

Sorden, S., 2000. Update on porcine circovirus and postweaning multisystemicwating syndrome. Journal of Swine Health & Production 8, 136.

Sorensen, K.J., Strandbygaard, B., Botner, A., Madsen, E.S., Nielsen, J., Have, P., 1998.Blocking ELISA’s for the distinction between antibodies against European andAmerican strains of porcine reproductive and respiratory syndrome virus.Veterinary Microbiology 60, 169–177.

Tomás, A., Fernandes, L.T., Valero, O., Segalés, J., 2008. A meta-analysis onexperimental infections with porcine circovirus type 2 (PCV2). VeterinaryMicrobiology 132, 260–273.

Woodbine, K.A., Medley, G.F., Slevin, J., Kilbride, A.L., Novell, E.J., Turner, M.J.,Keeling, M.J., Green, L.E., 2007. Spatiotemporal patterns and risks of herdbreakdowns in pigs with postweaning multisystemic wasting syndrome.Veterinary Record 160, 751–762.

Woodbine, K.A., Turner, M.J., Medley, G.F., Scott, P.D., Easton, A.J., Slevin, J., Brown,J.C., Francis, L., Green, L.E., 2011. A cohort study of post-weaning multisystemicwasting syndrome and PCV2 in 178 pigs from birth to 14 weeks on a single farmin England. Preventive Veterinary Medicine 97, 100–106.