Pig housing and management
This study was completed on a 1,500 Large White × Landrace sow farrow-to-finish Irish commercial farm, with weekly farrowing batches of approximately 80 sows. The farm was seropositive for App, Mhyo and SIV but seronegative for porcine respiratory and reproductive syndrome virus. The farm vaccinated piglets against Mhyo between 10 to 12 days of age and at weaning (i.e. 28 days of age), while a blanket SIV vaccination was used for sows every 6 months with a trivalent vaccine. This commercial farm purported that it implemented an AIAO production system, with batches of pigs post-weaning, progressing to the first nursery stage (4 weeks), then advancing to the second nursery stage (4 weeks), moving to the grower stage (4 weeks) and finally to the finisher stage (8 weeks). This was an observational study, whereby pigs were managed as per routine farm practice and the weekly movement of animals was tracked. Information regarding animal management was previously described by Calderón Díaz et al.  and Diana et al. .
In brief, a total of 1,016 pigs, born within one weekly farrowing batch, were followed through the different production stages. All pigs were individually tagged at birth and information on sow parity, number of piglets born alive per litter (NBA), sex and cross-fostering status (i.e. cross-fostered or not cross-fostered) was recorded. All pigs were weaned at approximately 28 days of age. At weaning, entire litters were transferred to the first nursery stage and housed in groups of 55 pigs, (minimum of 0.30 m2 pen space per pig) comprised of 4 to 5 litters. During the second nursery and grower stage, pigs were split and regrouped according to size and/or body weight (BW) and housed in groups of 36 animals, with a minimum of 0.55 m2 pen space per pig. Upon transfer to the finisher period, pigs were housed in groups of 35 pigs, with a minimum of 0.65 m2 pen space per pig. Housing facilities were uniform (i.e. pens, floor surface and ventilation system) within each of the production stages. Nursery and growing facilities had an automatic temperature control system with ceiling fans, while finisher facilities had natural ventilation. In all stages, animals were housed on fully slatted floors; plastic floors for nursery and concrete floors for the grower and finisher stages. Pigs were provided with wet-feed ad libitum during nursery; [18.3% crude protein (CP) and 10.5 MJ/DE per kg of feed]; grower (18.1% CP and 10.0 MJ/DE per kg of feed) and finisher diets (16.9% CP and 9.9 MJ/DE per kg of feed). Pigs had ad libitum access to water via at least one nipple drinker in each pen.
Mortality was recorded during the study. A total of 145 pigs died and 47 pigs were euthanised during the various phases of production representing 18.9% of all pigs in the study. Regarding mortality, 104 pigs died during lactation (54.2% out of 192 pigs), 24 pigs died during the nursery stages (12.5% out of 192 pigs), 3 pigs died during the grower stage (1.5% out of 192 pigs) and 14 pigs died during the finisher stage (7.3% out of 192 pigs). The remaining 47 pigs (24.5% out of 192 pigs) were selected for euthanasia due to the presence of abnormalities such as external lesions, hernias, tail loss, severe lameness, external abscesses and emaciation. Eight-hundred-and-twenty-four pigs reached slaughter age and they were slaughtered within 1 week, regardless of body weight at 24 weeks of age. Pigs were retrospectively classified into three production flows, depending on the time spent in each production stage: flow 1 = pigs that advanced through the normal production stages ‘in a timely manner’; n = 620; flow 2 = pigs which were delayed by one week from advancing forward to the next production stage; n = 111; and flow 3 = pigs delayed by more than one week from advancing to the next production stage; n = 93.
Blood sampling and serological analysis
At slaughter, individual blood samples were obtained at exsanguination using labelled red-stopper sterile BD Vacutainer® blood collection tubes (Becton, Dickinson U.K. Ltd., Berkshire, U.K.). All blood samples were individually labelled on collection, with corresponding sample delivery documents. Samples were transported to the Irish Department of Agriculture, Food and the Marine’s, Blood Testing Laboratory, Cork, for analysis. Blood samples were processed following clot formation, with serum aliquoted into individually labelled, anonymised cryovials (STARSTEDT®, Nümbrecht, Germany) and stored at -80oC until required for analysis. All serum samples were analysed by ELISA using commercial pathogen-specific ELISA kits (IDEXX Europe B.V., Hoofddorp, The Netherlands) for the three respiratory pathogens of interest [App - ApxIV (Apx IV toxin is produced during an episode of infection) Ab Test (97.8% sensitivity, 99.6% specificity); Mhyo – HerdChek® Mycoplasma hyopneumoniae Antibody Test (100% sensitivity, 99.67% specificity); SIV – Influenza A Ab Test (95.3% sensitivity, 99.6% specificity) for swine sera]. Manufacturer’s instructions were strictly adhered to during analysis, with positive in-house and also positive and negative test-kit controls incorporated during the serodiagnostic testing.
For each sample, the serostatus was determined for each pathogen of interest using the immunodiagnostic assay. Quantification of the antibody response was determined by colorimetric detection by spectrophotometry using TECAN Sunrise™ microplate reader (Tecan Group Ltd., Männedorf, Switzerland) in conjunction with TECAN Magellan™ data analysis software v7.1 (Tecan Group Ltd., Männedorf, Switzerland). Sample-to-positive (S/P) ratio values for App and Mhyo and sample-to-negative (S/N) ratio values for SIV were extrapolated from the optical density values obtained as per manufacturer’s instructions.
At slaughter, lesions of the lungs and heart were visually scored by a single trained observer. The macroscopic EP-like lesions were scored according to severity using BPEX Pig Health Scheme  on a scale from 0 to 55, where 0 indicates no lesion and 55 denotes the extensive presence of the EP-like lesions. Pleurisy was scored using the Slaughterhouse Pleurisy Evaluation System grid, developed by Dottori et al.  on a 5 point scale, dependent on lesion location and severity where; 0 = absence of chronic pleuritis lesions; 1 = ventrocranial lesion; 2 = dorsocaudal monolateral focal lesion; 3 = bilateral lesion or extended monolateral lesion (minimum of 1/3 of the diaphragmatic lobe); 4 = severely extended bilateral lesion (minimum of 1/3 of both diaphragmatic lobes. The presence of pericarditis (i.e. purulent inflammation of the pericardium resulting in adhesion of the pericardium to the epicardium and the pericardium with lungs and/or pleura) and heart condemnations were also recorded following the decision of the on-site veterinary inspector.
Data management and statistical analysis
All data were analysed in R v3.5.2 . Initially, ANOVA tests were performed for sow parity, birth weight and NBA, including data from all 824 animals in the batch that reached slaughter, to check for differences between production flows. Statistical differences were detected for the three variables between each flow. While litter size and NBA were similar between production flows, mean sow parity (2.9 ± 1.50) and mean body weight at birth (1.19 ± 0.30 kg) were lower in flow 3, compared with flow 1 (mean parity = 3.4 ± 1.43 and 1.44 ± 0.28 kg of body weight) and flow 2 (mean parity = 3.3 ± 1.49 and 1.26 ± 0.29 kg of body weight). Additionally, 29% of pigs in flow 3 originated from first parity sows (versus 13.4% of pigs in flow 1 and 19.4% of pigs in flow 2). Therefore, a nested case-control design was applied, whereby pigs originating from each flow were matched by sow parity, birth weight and NBA, resulting in a final data set of 120 in flow 1, 60 pigs in flow 2 and 60 pigs in flow 3. Pleurisy and EP-like lesions were reclassified as present or absent, due to the low number of higher scores recorded. Likewise, due to the low number of sows with parity ≥ 5, these were re-classified into a single group (i.e. 5+).
Residuals of predicted variables were tested for normality using the Shapiro-Wilk test and by examining the normal plot. Predicted variables were not normally distributed and thus, were analysed using generalised linear mixed models using the stats package. A Gamma distribution was fitted to each predicted variable. Univariable linear models were used with S/P or S/N ratio values as predicted variables with early-life indicators (sow parity, NBA, birth body weight, litter size, weaning body weight and cross-fostering status), production flow and pluck lesions (pleurisy, EP-like lesions, pericarditis and heart condemnations) as predictor variables. For all analyses, alpha level for determination of significance and trends was 0.05 and 0.10, respectively. Results for fixed effects are reported as least square means (LSM) ± standard error (SE), and results for continuous variables are reported as the regression coefficient ± SE.