TY - JOUR
AU - Jørgensen, E.
AB - ABSTRACT Hypothermia is a major cause of mortality in neonatal piglets. However, there are considerable individual differences in the successful recovery from postnatal hypothermia in the common farrowing environment, and so far the causes and interactions of causes have not been studied in detail. Using 635 crossbred neonatal piglets, the aim of this study was to identify the links among different physiological and behavioral measures and their connections to the ability of piglets to overcome initial postnatal hypothermia, with rectal temperature at 2 h as the response variable. The data included birth weight, hypoxia at birth (viability score and lactate in umbilical cord blood), latency to first udder contact and first suckle, scans of individual piglet position during the first 2 h after birth, and rectal temperature at birth and 2 h postpartum. A graphical chain model was used to analyze data. The statistical variables were divided into blocks according to level (design and litter) and chronological order (prenatal, birth, perinatal, and thermoregulatory success at 2 h) before applying the graphical model to the data. Bayesian information criteria (BIC) was used for model selection. The BIC relates to maximum likelihood, but introduces an additional penalty term for the number of variables. The strength of an association between 2 variables is reported as the increase in BIC (BICinc) due to removing the link. Results indicate that at 2 h, 22.1% of the piglets had a rectal temperature below 37°C. Out of the 16 variables included in the model, only 3 had direct links to the response variable of rectal temperature at 2 h. There was a positive relationship between rectal temperature at 2 h and birth weight (BICinc = 26), and between being observed more often by the udder as opposed to alone during both the first (BICinc = 8) and second hours (BICinc = 19) after birth. Lighter piglets and piglets that had experienced hypoxia took longer to achieve first suckle, which in turn affected where the piglet positioned itself during the first and second hours after birth. Variables related to the birth process had no direct connection to thermoregulatory success, but were additive in the explanation of piglet behavior. The rectal temperature of individual piglets at 2 h depends largely on piglet birth weight and on piglet position in relation to sow and littermates. Birth weight is the most important single factor in successful recovery from postnatal hypothermia. INTRODUCTION Hypothermia is a major cause of neonatal piglet mortality and predisposes piglets to mortality by other causes including starvation, crushing, and disease. At birth, the piglet experiences a dramatic change in the ambient temperature and nature of its surroundings. From the homeostatic temperature in the sow uterus (38 to 40°C), the piglets are born into the much cooler environment of the farrowing crate or pen (20 to 22°C). At this time the lower critical ambient temperature of the piglet is 34 to 35°C (Mount, 1959; Berthon et al., 1993). This causes a well-documented decrease in body temperature shortly after birth, known as postnatal hypothermia (Berthon et al., 1993; Hoy et al., 1995; Tuchscherer et al., 2000; Malmkvist et al., 2006). The extent and duration of this postnatal hypothermia correlates negatively to the chances of survival of the piglet (Hoy et al., 1995; Tuchscherer et al., 2000). However, there are major individual differences both among and within litters in neonatal piglet success in recovering from initial postnatal hypothermia (Tuchscherer et al., 2000). To overcome postnatal hypothermia, neonatal piglets must rapidly adapt to postuterine life through thermogenesis and heat preservation. Rectal temperature at 2 h after birth is a strong indicator of thermoregulatory success (Baxter et al., 2008). For piglets up to 24 h after birth, a body temperature between 38 and 39°C would indicate thermal homeostasis (Berthon et al., 1993; Herpin et al., 1994). The aim of this study was to identify causes of differences in recovering from postnatal hypothermia measured as body temperature 2 h after birth. Physiological and behavioral data were collected on 635 live-born piglets from 45 litters born in farrowing crates or farrowing pens with air temperatures of 18 to 20°C, and a chain-graph model was applied to the data. MATERIALS AND METHODS All procedures involving animals were approved by the Danish Animal Experiments Inspectorate in accordance with the Danish Ministry of Justice Law No. 382 (June 10, 1987) and Acts 333 (May 19, 1990), 726 (September 9, 1993), and 1016 (December 12, 2001). The experimental subjects were 635 crossbred piglets from 45 litters born in 2 farrowing environments and 2 batches. The dams of the experimental subjects were 45 crossbred gilts (Danish Yorkshire × Danish Landrace), which were inseminated in their second estrus (approximately 210 d of age). All gilts within a batch were inseminated with semen from the same Duroc boar. The study was part of larger piglet mortality study, and therefore the gilts were pair-wise full siblings and distributed among the 2 different housing systems at farrowing (see farrowing housing and care). In batch A, 10 sows farrowed in pens and 11 sows farrowed in crates. In batch B, 11 sows farrowed in pens and 13 farrowed in crates. The 45 gilts gave birth to 679 piglets (15.1 ± 2.4 per litter), 641 of these were live born (14.3 ± 2.5 per litter), and 38 were stillborn (5.6% of total born), 3 of which were mummified and classified as prefarrowing deaths. Six piglets died during the first hour after birth, and an additional 3 piglets died during the second hour. Of the first 6, 2 suffocated in particularly thick amniotic sacks, 1 was weak and died a few minutes after birth, 2 were bitten by the sow, and 1 was crushed by the sow. Of the 3 piglets dying during the second hour after birth, 2 were crushed and 1 was bitten by the sow. Only the 635 piglets that survived as a minimum for the first hour after birth were included in data analysis. Farrowing Housing and Care The gilts farrowed in 2 different types of individual farrowing housing; a traditional farrowing crate (see Figure 1a) and loose housing in a modified version of the wide T-pen (see Figure 1b; Moustsen et al., 2007; Moustsen and Jensen, 2008). Both farrowing environments were thermostatically controlled with air temperatures between 18 and 20°C. In the T-pens, the sows were free to move around within the pen during the nest-building phase, farrowing, and throughout the lactation period. In the crates, the sows were confined to a specific area in the farrowing pen by horizontal bars, which did not allow them to turn around. In the farrowing units, sows were fed twice daily at 0730 and 1430 h with a standard sow meal containing 8.40 MJ of NE/kg as fed, and at the time of farrowing, the ration was 2.6 kg/d. From d 113 after insemination, the sows were given nesting materials in the form of approximately 2 kg/d of chopped wheat straw. Wet straw was removed from the pens every morning. After farrowing, the amount of straw given was reduced to 1 kg/d. The piglet creep areas differed in size and shape between the 2 farrowing environments. However, both were covered and had supplementary heat from a heat lamp fitted to a hole in the roof of the creep. Before farrowing, the floor of the piglet creep area was covered with an approximately 10-cm layer of chopped straw. Surface temperatures in the pens and crates were measured at farrowing and 24 h after farrowing. The surface temperature was assessed in 3 specific areas: creep area, by the udder of the sow (at the distance of the hoof, approximately 30 cm away from the udder of the sow), and on the slats, which are areas where the piglets could be expected to settle during the first 24 h of life. The measurements were taken using an infrared thermometer (Oakton TempTestr; IR Oakton Instruments, Vernon Hills, IL; accuracy ±2% or ±2°C, whichever was greater) from 50 cm above the floor surface, at least 20 cm away from the body of the sow. Distance-to-target-size ratio of this device was 1:6; thus, the area covered was a circle of approximately 8 cm in diameter as measures were taken vertically from 50 cm above the surface. Figure 1. View largeDownload slide Panel a. Traditional crate: the traditional farrowing crate with horizontal bars confining the sow; all measurements are given in centimeters. Vertical lines = cast-iron slatted floor. White space = solid concrete floor. Panel b. The modified wide T-pen: loose-housing farrowing pen; all measures are given in centimeters. Vertical lines = cast-iron slatted floor. White space = solid concrete floor. Small black circles = wall of vertical steel bars. Figure 1. View largeDownload slide Panel a. Traditional crate: the traditional farrowing crate with horizontal bars confining the sow; all measurements are given in centimeters. Vertical lines = cast-iron slatted floor. White space = solid concrete floor. Panel b. The modified wide T-pen: loose-housing farrowing pen; all measures are given in centimeters. Vertical lines = cast-iron slatted floor. White space = solid concrete floor. Small black circles = wall of vertical steel bars. The average surface temperature varied among the 3 measuring points (P < 0.0001; creep area: 30.1 ± 3.9°C; by sow: 24.7 ± 3.6°C; and on the slats: 19.8 ± 2.43°C). Surface temperatures also differed between batches (P < 0.0001), with the surface temperatures measured in the August batch being greater than the surface temperatures in the December batch (creep area: 31.1 vs. 27.7°C; by sow: 25.4 vs. 22.8°C; and on the slats: 21.0 vs. 17.1°C). There was no systematic difference in surface temperatures between farrowing environments (pen temperatures were not different from crate temperatures; P = 0.34). Procedures at Farrowing and 2 h All farrowings were overseen by the experimenters. Birth assistance was performed only when >3 h had elapsed without a piglet being expelled and there was no sign of the placenta being expelled or other signs that the farrowing had ended. Apart from the measures and procedures mentioned below, there was only minimal interference with the sows and piglets. However, any piglets that were found severely crushed or otherwise injured and would not survive were humanely euthanized by the experimenters. Time of birth and birth order of each individual piglet were noted at birth. From this information the interbirth interval (IBI_pig) for a given piglet was calculated as time elapsed from birth of the previous piglet. Consequently, the IBI_pig of the first piglet was treated as a missing value. Viability score of the piglet was determined based on the movement and breathing of the piglet during the first 15 s after expulsion (Table 1). Then blood samples for lactate analysis were collected (for procedures see blood sampling and lactate analysis). Immediately after collecting blood for lactate analysis, birth weight [using an OHAUS (Parsippany, NJ) digital scale of maximum 6 kg ± 2 g accuracy, calibrated by Dansk Vægt Industri A/S, Skanderborg, Denmark] and rectal temperature measurements (RT_birth; using a Kruuse, Digi-Temp Digital Thermometer, Langeskov, Denmark; of 32 to 42°C ± 0.10°C inserted approximately 1.5 cm into the rectum of the piglet) were taken. Time of determining RT_birth was noted, and initial data inspection showed a uniform distribution of temperatures taken within the first 5 min after birth. But there was also a general decline in temperatures taken later than 5 min after birth. Consequently, all temperatures determined later than 5 min after birth were discarded and renounced as missing values. Piglets were dried on the back with a paper towel only sufficiently to allow marking with piglet number (birth order) using a black permanent marker. This identification number made it possible to identify the individual piglets during direct observations and on video (see behavioral observations). Relative birth was calculated from birth order divided by litter size (Table 2). The sex of the piglet was also determined before carefully replacing the piglet in its original position in the pen. Two hours after birth, each piglet was picked up again to measure rectal temperature (RT_2h), crown-rump length (from top of the head to the base of the tail), and girth (right behind the front legs) before placing the piglet back in the pen at the place from where it had been removed. The crown-rump and girth data were used to calculate an index of piglet proportions (crown-rump/girth). Table 1. Viability score: viability as determined during the first 15 s after expulsion1 Viability score  Description  0  No movement, no breathing after 15 s, stillborn.  1  No movement after 15 s; piglet is breathing or attempting to breathe (coughing, spluttering, clearing its lungs).  2  Piglet shows some movement and attempts are made to right itself onto sternum within 15 s, breathing or attempting to breath.  3  Good movement and good breathing. Piglet rights itself onto sternum and attempts to stand within 15 s.  4  Good movement, good breathing, piglet stands within 15 s.  Viability score  Description  0  No movement, no breathing after 15 s, stillborn.  1  No movement after 15 s; piglet is breathing or attempting to breathe (coughing, spluttering, clearing its lungs).  2  Piglet shows some movement and attempts are made to right itself onto sternum within 15 s, breathing or attempting to breath.  3  Good movement and good breathing. Piglet rights itself onto sternum and attempts to stand within 15 s.  4  Good movement, good breathing, piglet stands within 15 s.  1Modified from Herpin et al. (1996) and Baxter et al. (2008). View Large Table 1. Viability score: viability as determined during the first 15 s after expulsion1 Viability score  Description  0  No movement, no breathing after 15 s, stillborn.  1  No movement after 15 s; piglet is breathing or attempting to breathe (coughing, spluttering, clearing its lungs).  2  Piglet shows some movement and attempts are made to right itself onto sternum within 15 s, breathing or attempting to breath.  3  Good movement and good breathing. Piglet rights itself onto sternum and attempts to stand within 15 s.  4  Good movement, good breathing, piglet stands within 15 s.  Viability score  Description  0  No movement, no breathing after 15 s, stillborn.  1  No movement after 15 s; piglet is breathing or attempting to breathe (coughing, spluttering, clearing its lungs).  2  Piglet shows some movement and attempts are made to right itself onto sternum within 15 s, breathing or attempting to breath.  3  Good movement and good breathing. Piglet rights itself onto sternum and attempts to stand within 15 s.  4  Good movement, good breathing, piglet stands within 15 s.  1Modified from Herpin et al. (1996) and Baxter et al. (2008). View Large Table 2. Variables included in the graphical block recursive model Level  Variable  Label  Identification tag  Transformation  Type  Description  n  Group level  Design  housing  ho  —  Factor/2 levels  Housing type pen/crate  635        batch  ba  —  Factor/2 levels  Batch August or December  635  Sow/litter level  Litter  littersize  li  —  Integer  Litter size  635        sowID  so  —  Factor/45 levels  ID of sow nested in housing  635  Piglet level  Prenatal  weight  w  —  Numeric  Birth weight in kilograms  635        sex  x  —  Factor/2 levels  Piglet sex  634        index  in     Numeric  Body length/girth  630        b_order  bo     Numeric  Birth order/litter size  627     Birth  viability  v  —  Integer  Viability score  606        lactate  la  Log  Numeric  ln(Umbilical cord lactate in mM)  555        IBI_pig  ib  Log  Numeric  ln(Interval from preceding piglet in seconds + 1)  635        RT_birth  rtb  —  Numeric  Rectal temperature at birth in °C  597     Perinatal  Time to suckle  su  Log  Numeric  ln(Latency to first successful suckle in hours + 0.001)  630        Touch to suckle  ts  Log  Numeric  ln(Interval from first touching the udder to first successful suckle in hours + 0.001)  625        pos_hour1  p1  Logit function1  Numeric  Logit function of piglet position scans during the first hour after birth.  635        pos_hour2  p2  Logit function1  Numeric  Logit function of piglet position scans during the second hour after birth.  630  Thermoregulatory success −  Response  RT_2h  rt2  —  Numeric  Response variable  614  Level  Variable  Label  Identification tag  Transformation  Type  Description  n  Group level  Design  housing  ho  —  Factor/2 levels  Housing type pen/crate  635        batch  ba  —  Factor/2 levels  Batch August or December  635  Sow/litter level  Litter  littersize  li  —  Integer  Litter size  635        sowID  so  —  Factor/45 levels  ID of sow nested in housing  635  Piglet level  Prenatal  weight  w  —  Numeric  Birth weight in kilograms  635        sex  x  —  Factor/2 levels  Piglet sex  634        index  in     Numeric  Body length/girth  630        b_order  bo     Numeric  Birth order/litter size  627     Birth  viability  v  —  Integer  Viability score  606        lactate  la  Log  Numeric  ln(Umbilical cord lactate in mM)  555        IBI_pig  ib  Log  Numeric  ln(Interval from preceding piglet in seconds + 1)  635        RT_birth  rtb  —  Numeric  Rectal temperature at birth in °C  597     Perinatal  Time to suckle  su  Log  Numeric  ln(Latency to first successful suckle in hours + 0.001)  630        Touch to suckle  ts  Log  Numeric  ln(Interval from first touching the udder to first successful suckle in hours + 0.001)  625        pos_hour1  p1  Logit function1  Numeric  Logit function of piglet position scans during the first hour after birth.  635        pos_hour2  p2  Logit function1  Numeric  Logit function of piglet position scans during the second hour after birth.  630  Thermoregulatory success −  Response  RT_2h  rt2  —  Numeric  Response variable  614  1ln[(number of observations out of 6 within the hour of the piglet being in “udder contact” + 0.5)/(number of observations out of 6 within the hour of the piglet being “alone on floor” + 0.5)]. View Large Table 2. Variables included in the graphical block recursive model Level  Variable  Label  Identification tag  Transformation  Type  Description  n  Group level  Design  housing  ho  —  Factor/2 levels  Housing type pen/crate  635        batch  ba  —  Factor/2 levels  Batch August or December  635  Sow/litter level  Litter  littersize  li  —  Integer  Litter size  635        sowID  so  —  Factor/45 levels  ID of sow nested in housing  635  Piglet level  Prenatal  weight  w  —  Numeric  Birth weight in kilograms  635        sex  x  —  Factor/2 levels  Piglet sex  634        index  in     Numeric  Body length/girth  630        b_order  bo     Numeric  Birth order/litter size  627     Birth  viability  v  —  Integer  Viability score  606        lactate  la  Log  Numeric  ln(Umbilical cord lactate in mM)  555        IBI_pig  ib  Log  Numeric  ln(Interval from preceding piglet in seconds + 1)  635        RT_birth  rtb  —  Numeric  Rectal temperature at birth in °C  597     Perinatal  Time to suckle  su  Log  Numeric  ln(Latency to first successful suckle in hours + 0.001)  630        Touch to suckle  ts  Log  Numeric  ln(Interval from first touching the udder to first successful suckle in hours + 0.001)  625        pos_hour1  p1  Logit function1  Numeric  Logit function of piglet position scans during the first hour after birth.  635        pos_hour2  p2  Logit function1  Numeric  Logit function of piglet position scans during the second hour after birth.  630  Thermoregulatory success −  Response  RT_2h  rt2  —  Numeric  Response variable  614  Level  Variable  Label  Identification tag  Transformation  Type  Description  n  Group level  Design  housing  ho  —  Factor/2 levels  Housing type pen/crate  635        batch  ba  —  Factor/2 levels  Batch August or December  635  Sow/litter level  Litter  littersize  li  —  Integer  Litter size  635        sowID  so  —  Factor/45 levels  ID of sow nested in housing  635  Piglet level  Prenatal  weight  w  —  Numeric  Birth weight in kilograms  635        sex  x  —  Factor/2 levels  Piglet sex  634        index  in     Numeric  Body length/girth  630        b_order  bo     Numeric  Birth order/litter size  627     Birth  viability  v  —  Integer  Viability score  606        lactate  la  Log  Numeric  ln(Umbilical cord lactate in mM)  555        IBI_pig  ib  Log  Numeric  ln(Interval from preceding piglet in seconds + 1)  635        RT_birth  rtb  —  Numeric  Rectal temperature at birth in °C  597     Perinatal  Time to suckle  su  Log  Numeric  ln(Latency to first successful suckle in hours + 0.001)  630        Touch to suckle  ts  Log  Numeric  ln(Interval from first touching the udder to first successful suckle in hours + 0.001)  625        pos_hour1  p1  Logit function1  Numeric  Logit function of piglet position scans during the first hour after birth.  635        pos_hour2  p2  Logit function1  Numeric  Logit function of piglet position scans during the second hour after birth.  630  Thermoregulatory success −  Response  RT_2h  rt2  —  Numeric  Response variable  614  1ln[(number of observations out of 6 within the hour of the piglet being in “udder contact” + 0.5)/(number of observations out of 6 within the hour of the piglet being “alone on floor” + 0.5)]. View Large Blood Sampling and Lactate Analysis Blood samples from the umbilical cord were collected immediately after determination of viability score (first 15 s after birth). The piglet was lifted out of the pen when collecting blood. It was noted if the umbilical cord had already broken and if it was blood-filled. If the cord was not already broken, it was gently ruptured at this time to disconnect the piglet from the sow (placenta) by slowly pulling the umbilical cord in the direction of the piglet, thus rupturing the cord as far away from the piglet as possible. Approximately 1 mL of blood was collected from the umbilical cord using a 2.5-mL plastic syringe with a 40-mm 21-ga needle, without further rupturing of the cord. Blood sampling was successful for 555 of 635 piglets included in this study (Table 2). After sampling, the umbilical cord was clamped by hand for the few minutes that it took for the bleeding to stop. The exact time of the sampling was noted. Blood was then carefully ejected into a 4.5-mL K3E K3 EDTA test tube. The blood-filled test tube was sealed and placed on ice for a maximum of 1 h before separating serum from blood cells by centrifugation at 3,500 × g for 10 min at 4°C. Serum was collected using a disposable pipette into a labeled 1.5-mL Eppendorf tube and stored at −20 to −30°C until analysis of lactate content. Blood plasma was analyzed for lactate according to standard procedures (Lactatoxidase, LAC, Bayer Clinical Methods), using an auto analyzer (ADVIA 1650 Chemistry System, Bayer Corporation, Tarrytown, NY). Intraassay CV was 1.19 and 0.69 for low and high control samples, respectively, and interassay CV was 1.23 and 0.94 for low and high control samples, respectively. Accuracy (% bias, n = 24) was −7.3% (1.49 mM) and −5.6% (5.68 mM), for low and high controls, respectively. Behavioral Observations Time to Touch Udder and Time to Suckle. The time for each individual piglet to first reach the udder and touch the udder with the snout tip was noted (time to touch udder). The time for each piglet to achieve first successful milk ingestion was also noted. This was defined as the first time the piglet suckled continuously on the same teat for duration of a min 5 s. From these data and time of birth, calculations for the variables time to suckle (su) and interval from touching the udder to first successful colostrum ingestion (ts) were calculated. Observation of Piglet Position and Use of the Creep Area. From the time of birth until 2 h after birth, the position in the pen or crate of the individual piglets were observed every 10 min (Table 3). See also the Statistical Analyses section on piglet position. Table 3. Ethogram of individual piglet position observed every 10 min during the first 2 h after expulsion Behavior  Description  Udder contact  The piglet is alive and in contact with the udder, or in contact with another piglet that touches the udder.  Alone on floor  Piglet is alive and not in contact with any other piglet, or in contact with the udder. It can be in contact with other parts of the body of the sow.  In huddle  The piglet is alive and in contact with at least one other piglet, but not in contact with the udder or another piglet touching the udder.  In creep area  The piglet is alive and inside the covered creep area.  Behavior  Description  Udder contact  The piglet is alive and in contact with the udder, or in contact with another piglet that touches the udder.  Alone on floor  Piglet is alive and not in contact with any other piglet, or in contact with the udder. It can be in contact with other parts of the body of the sow.  In huddle  The piglet is alive and in contact with at least one other piglet, but not in contact with the udder or another piglet touching the udder.  In creep area  The piglet is alive and inside the covered creep area.  View Large Table 3. Ethogram of individual piglet position observed every 10 min during the first 2 h after expulsion Behavior  Description  Udder contact  The piglet is alive and in contact with the udder, or in contact with another piglet that touches the udder.  Alone on floor  Piglet is alive and not in contact with any other piglet, or in contact with the udder. It can be in contact with other parts of the body of the sow.  In huddle  The piglet is alive and in contact with at least one other piglet, but not in contact with the udder or another piglet touching the udder.  In creep area  The piglet is alive and inside the covered creep area.  Behavior  Description  Udder contact  The piglet is alive and in contact with the udder, or in contact with another piglet that touches the udder.  Alone on floor  Piglet is alive and not in contact with any other piglet, or in contact with the udder. It can be in contact with other parts of the body of the sow.  In huddle  The piglet is alive and in contact with at least one other piglet, but not in contact with the udder or another piglet touching the udder.  In creep area  The piglet is alive and inside the covered creep area.  View Large Mortality and Postmortem Evaluation All seemingly stillborn piglets were recorded with birth order and weight, and the lungs were tested (float on water) to ensure that the piglet was stillborn and had not been breathing. If a live-born piglet died during the direct observation, the exact time of death was recorded. The piglet was weighed, and the cause of death was determined through observation and postmortem examination. Statistical Analyses The Graphical Chain Model. A graphical chain model (Lauritzen and Wermuth, 1989), also called a block-recursive model (Højsgaard and Thiesson, 1995), was used in the data analysis. A chain-graph model facilitates the search for the optimal statistical model of a multivariate data set. In recent years, standard software for estimation in these models has become available, such as MIM (Edwards, 2000). This has led to several applications especially within social science (e.g., Wermuth, 2003). Pedersen et al. (2006) used the model for study of early piglet mortality in loose-housed sows. The independence structure of the multivariate data set is shown as a so-called conditional independence graph. Each variable in the data set is shown with a node (or vertex) and edges (or links) between the nodes indicate whether the variables are conditionally dependent. In chain-graph models like the present block recursive model, the variables are partitioned into blocks. Within blocks, undirected links are assumed (e.g., cause and effect cannot be distinguished). Between blocks, directed links are assumed, which in some cases can be used to infer causal relationship (e.g., due to a temporal ordering of the blocks). We refer to Lauritzen and Richardson (2002) for a detailed discussion of this aspect. Piglet Position. Initially, there were 4 piglet positions in the scan data (see Table 3). To simplify the model and give more explanatory variables, the number of scan variables was reduced. Only “udder contact” and “alone on floor” were included from the scans of piglet position. “Udder contact” (79.9%) and “alone on floor” (11.1%) were the most frequent positions, accounting for 91.0% of all scans. The piglets were only rarely “in huddle” (4.6%) and “in creep area” (4.9%), and these 2 positions were subsequently eliminated from the model. Second, “alone on floor” and “udder contact” nearly cancelled out one another. Thermally, these 2 positions are very different or could even be considered as opposites. Thus, to have just 1 variable called pos_hour, which included both “alone on floor” and “udder contact,” the empirical logit function of the 2 was introduced using the equation ln[(No. of “udder contact” observations + 0.5)/(No. of “alone on floor” observations + 0.5)]. By using this approach, the thermoregulatory behavior of the piglet specific to the hour was categorized by one value, calculated based on the position scans of the piglet in that hour. This approach resulted in negative values for piglets observed more often alone than by the udder, 0 for piglets that were observed by udder as often as alone, and positive values for piglets that were found more often by the udder than alone on the floor. Transformation and Missing Values. Several of the data variables in the graphical model were transformed based on inspection of the standardized residual plots to comply with the assumption of constant variance. For an overview of the transformation methods applied, see Table 2. Table 2 also shows the number of piglets out of the 635 from which data were available for each of the variables; the remaining were treated as missing values. Furthermore, Table 2 includes a description of how the different variables were calculated based on raw data. The transformations were performed in the open-source statistical program R (R Development Core Team, 2008). Model Selection. The graphical model was fitted using the statistical program MIM 3.2 (Edwards, 2000), with the assistance of the mimR package (Højsgaard, 2007), which was used to call the MIM program (Edwards, 2000) from R and import the data from R into MIM. The Bayesian information criterion (BIC) was used to find the best model. The BIC relates to maximum likelihood but introduces an additional penalty term for the number of variables. The BIC criterion allows for comparison of the fit of different models. The best model is the model with the smallest collective BIC and therefore the best fit. The BIC of any given edge in the model signifies the contribution of this particular edge to improve the overall model fit. During model selection, negative BIC of an edge signifies that removing the edge will result in an equivalent decrease of the overall BIC, and therefore give a model with a better fit. Conversely, a positive BIC of a given edge signifies that removing this edge will result in an increase in the overall BIC, and a worse fit of the model. Consequently, the resulting block recursive model contains only positive BIC values for all edges. The larger the BIC for a given edge, the more important is the edge in the model. In graphical models, this is a way to represent the strength of relations between variables, in some ways comparable with the P-value used in more simple correlation and variance analysis. However, the BIC value of an edge in a block recursive model also considered all the other edges in the current and previous blocks, and therefore it tells us much more about the importance of the edge in the resulting model than the simple correlation of 2 variables. Before initiating the search for the graphical model with the best fit, the variables were ordered into 6 blocks. The ordering of these blocks was based on design criteria (blocks 1 and 2) and on chronological ordering of the piglet variables (blocks 3, 4, 5, and 6): block 1: design; block 2: litter; block 3: prenatal; block 4: birth; block 5: perinatal; and block 6: thermoregulatory success at 2 h. The chronological ordering in blocks 3 to 6 was based partly on the time that the measure was taken and partly on the order in which it was decided. For example, birth weight was placed in block 3: prenatal, because although it cannot be measured until after expulsion, it is decided before birth and is not influenced by expulsion. A thorough model selection was done using both forward and backward stepwise procedures in MIM. Furthermore, edges representing significant correlations (P-values < 0.05), according to initial data inspection and traditional correlation analysis, were also tested in the model one by one, followed by the backward stepwise procedure. The selected resulting block recursive model had the minimal BIC on each of the 6 blocks in the model, and could not be reduced any further without an increase in BIC. Other models came close to the final model in overall BIC, with similar structure, but the model presented was the model with the smallest collective BIC value, which means that adding or removing any edges resulted in a poorer model. The search for the best model was carried out in decomposable mode. This means that some of the edges could not be left out while still maintaining the decomposable structure. The BIC values were only calculated for edges that could be removed. Therefore, no BIC values are given for the edges including viability and umbilical cord lactate (v-la), birth interval and umbilical cord lactate (ib-la), weight and umbilical cord lactate (w-la), and birth order and birth interval (bo-ib). Decomposable mode, as compared with unrestricted mode, was the most suitable for the type of data analyzed in the current study. A detailed description of the search procedures mentioned can be found in Edwards (2000). RESULTS Rectal Temperatures at 2 h Figure 2 shows the distribution of the response variable of rectal temperature at 2 h. A total of 77.9% of piglets had rectal temperatures of 37.0°C or above at 2 h after birth. The thermometer could only measure temperatures within the range of 32 to 42°C; thus, the 6 piglets with rectal temperatures below 32°C are shown as <32.0°C. Figure 2. View largeDownload slide Distribution of piglet rectal temperatures 2 h after birth (n = 614). Figure 2. View largeDownload slide Distribution of piglet rectal temperatures 2 h after birth (n = 614). The Resulting Graphical Model The selected graphical model represents the main result in this study (Figure 3). The BIC values of each of the edges between variables are described in Figure 3. The greater the BIC value of a given edge, the more important the edge was to the collective model. Removing this edge would result in an equivalent increase in the collective BIC value of the model. The most important relations between the variables are also shown as scatter plots and a trend line based on simple linear regression. In the detailed presentation of the results (Figure 3), all variables are presented by full name supplemented by the letter/identification tag representing the variable. Furthermore, a detailed description of the selected block recursive model, block by block, and marginal difference in BIC values when removing each edge in the model are also presented and described below. Figure 3. View largeDownload slide Resulting block recursive graphical model. Variables are organized by block. Blocks 1 and 2 are the effects above the piglet level: experimental design and litter effects; these are organized by level. Blocks 3, 4, and 5 consist of variables on the piglet level. These are organized in chronological order: prenatal, birth, and perinatal. The 6th block contains only the response variable: thermoregulatory success at 2 h. Shaded notes are discreet or class variables, whereas open notes are continuous variables. The increase in collective Bayesian information criteria (BIC) if removing an edge is represented by the number signified next to the edge. The BIC values of each of the blocks in the selected block-recursive model are as follows: 1: Design (housing = ho, batch = ba); 2: litter/sow (littersize = li, sowID = so), df: 355, BIC: 8,793; 3: prenatal (weight = w, sex = x, index = in, b_order = bo), df: 5,203, BIC: 8,160; 4: birth (viability = v, lactate = la, IBI_pig = ib, RT_birth = rtb), df: 15,765, BIC: 11,751; 5: perinatal (Time to suckle = su, Touch to suckle = ts, pos_hour1 = p1, pos_hour2 = p2), df: 32,336, BIC: 17,062; 6: thermoregulatory success at 2 h (RT_2h = rt2), df: 37,345, BIC: 18,060. Figure 3. View largeDownload slide Resulting block recursive graphical model. Variables are organized by block. Blocks 1 and 2 are the effects above the piglet level: experimental design and litter effects; these are organized by level. Blocks 3, 4, and 5 consist of variables on the piglet level. These are organized in chronological order: prenatal, birth, and perinatal. The 6th block contains only the response variable: thermoregulatory success at 2 h. Shaded notes are discreet or class variables, whereas open notes are continuous variables. The increase in collective Bayesian information criteria (BIC) if removing an edge is represented by the number signified next to the edge. The BIC values of each of the blocks in the selected block-recursive model are as follows: 1: Design (housing = ho, batch = ba); 2: litter/sow (littersize = li, sowID = so), df: 355, BIC: 8,793; 3: prenatal (weight = w, sex = x, index = in, b_order = bo), df: 5,203, BIC: 8,160; 4: birth (viability = v, lactate = la, IBI_pig = ib, RT_birth = rtb), df: 15,765, BIC: 11,751; 5: perinatal (Time to suckle = su, Touch to suckle = ts, pos_hour1 = p1, pos_hour2 = p2), df: 32,336, BIC: 17,062; 6: thermoregulatory success at 2 h (RT_2h = rt2), df: 37,345, BIC: 18,060. Direct Effects on Thermoregulatory Success. Only 3 variables had direct edges with thermoregulatory success at 2 h after birth. These were birth weight (w), piglet position in the first hour after birth (p1), and piglet position in the second hour (p2). Heavier piglets had greater rectal temperatures at 2 h (Figure 4a), and piglets that spent more time by the udder as opposed to alone on the floor during the first (Figure 4b) and the second hour (Figure 4c) after birth had greater rectal temperatures 2 h after birth. Of these 3 affecting variables, the edge between birth weight and temperature at 2 h would result in the largest BIC value increase and hence was the most important of the 3 in the collective model (Figure 3). Figure 4. View largeDownload slide Panel a. Scatter plot of birth weight, kg (w), and rectal temperature at 2 h, °C (rt2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 25.66. Panel b. Scatter plot of piglet position in the first hour after birth, logit function of piglet position scans (p1), and rectal temperature at 2 h, °C (rt2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 7.65. Panel c. Scatter plot of piglet position in the second hour after birth, logit function of piglet position scans (p2), and rectal temperature at 2 h, °C (rt2). Bayesian information criteria = 18.68. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Figure 4. View largeDownload slide Panel a. Scatter plot of birth weight, kg (w), and rectal temperature at 2 h, °C (rt2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 25.66. Panel b. Scatter plot of piglet position in the first hour after birth, logit function of piglet position scans (p1), and rectal temperature at 2 h, °C (rt2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 7.65. Panel c. Scatter plot of piglet position in the second hour after birth, logit function of piglet position scans (p2), and rectal temperature at 2 h, °C (rt2). Bayesian information criteria = 18.68. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Birth Process and Thermoregulation. The variables related to the birth process (v, la, ib, rtb) constitute the fourth block (Figure 3). None of the variables in this block had direct connections to rectal temperature at 2 h. However, they did affect the behavioral variables related to rectal temperature at 2 h (p1 and p2) in different ways. Birth temperature had a weak connection to piglet position in the second hour (p2); piglets born with greater birth temperatures spent slightly less time by the udder (Figure 5a). Piglet interbirth interval was connected to piglet position in the first hour. The effect being that piglets born after longer intervals from the preceding piglet spend more time alone and less time by the udder during the first hour compared with piglets born after shorter intervals (Figure 5b). Umbilical cord lactate and viability score did not connect to piglet position in the first 2 h after birth (p1 and p2). However, increased lactate and decreased viability score, which are both indirect measures of hypoxia during delivery, delayed the time to achieve first successful suckle (Figure 5c, 5d). Time to achieve first suckle, in turn, had strong connections to piglet position during the first hour (Figure 5e) and the second hour (Figure 5f). Figure 5. View largeDownload slide Panel a. Scatter plot of rectal temperature at birth, °C (rtb) and piglet position in the second hour, logit function of piglet position scans (p2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 13.98. Panel b. Scatter plot of birth interval from the preceding piglet, seconds (ib), and piglet position in the first hour, logit function of piglet position scans (p1). Log scale on the x-axis. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 7.38. Panel c. Scatter plot of time to first suckle, log(h + 0.001) (su) and umbilical cord lactate, log(mM) (la). Log scale on both axes. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 15.85. Panel d. Boxplot of viability score (v) and time to first suckle, log(h + 0.001) (su). Log scale on the y-axis. Piglets are grouped according to viability score and plotted against their time to achieve first suckle. The broad edge in the middle of the box represents the average. Bayesian information criteria = 9. Panel e. Scatter plot of latency to first suckle, log(h + 0.001) (su) and piglet position in the first hour after birth, logit function of piglet position scans (p1). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 245.79. Panel f. Scatter plot of latency to first suckle, log(h + 0.001) (su) and piglet position in the second hour after birth, logit function of piglet position scans (p2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 25.03. Figure 5. View largeDownload slide Panel a. Scatter plot of rectal temperature at birth, °C (rtb) and piglet position in the second hour, logit function of piglet position scans (p2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 13.98. Panel b. Scatter plot of birth interval from the preceding piglet, seconds (ib), and piglet position in the first hour, logit function of piglet position scans (p1). Log scale on the x-axis. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 7.38. Panel c. Scatter plot of time to first suckle, log(h + 0.001) (su) and umbilical cord lactate, log(mM) (la). Log scale on both axes. Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 15.85. Panel d. Boxplot of viability score (v) and time to first suckle, log(h + 0.001) (su). Log scale on the y-axis. Piglets are grouped according to viability score and plotted against their time to achieve first suckle. The broad edge in the middle of the box represents the average. Bayesian information criteria = 9. Panel e. Scatter plot of latency to first suckle, log(h + 0.001) (su) and piglet position in the first hour after birth, logit function of piglet position scans (p1). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 245.79. Panel f. Scatter plot of latency to first suckle, log(h + 0.001) (su) and piglet position in the second hour after birth, logit function of piglet position scans (p2). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 25.03. Effects Above Piglet Level. The variables in block 1 and 2 did not directly affect thermoregulation at 2 h. Litter size affected birth weight, in that an increase in litter size decreased the birth weight of piglets in the litter. However, the edge between litter size and birth weight generated only a small increase in BIC (Figure 3) and hence had very limited importance in the model. Litter size also affected interbirth intervals, with larger litter size decreasing inter birth intervals between the individual piglets. There was a strong sow effect on piglet rectal temperature at birth. However, this early effect of sow was no longer present at 2 h. Rectal temperature at 2 h after birth was not related to rectal temperature at birth, nor was it related to sow or any other variables above the piglet level. Prenatal Effects. The third block of prenatal included the variables related to the prenatal phase (w, x, in, and bo). As already mentioned, birth weight had direct connections to thermoregulation and was furthermore connected to variables related to the birth process in block 4 and latency to first suckle. The connection was that heavier piglets took less time to achieve first suckle than lighter piglets (Figure 6), which in turn connected to all of the other behavioral variables (p1, p2, and ts) in the fifth block of perinatal. Birth weight and relative birth order were not connected. However, relative birth order affected the same variables in the fourth block (birth) as birth weight. Increased lactate was associated with lighter birth weights and high/later relative birth order. However, long interbirth intervals were associated with heavier piglets and early position in the relative birth order. Finally, greater rectal temperature at birth was related to heavier piglets and to late position in the birth order. Relative birth order further related to piglet position in the first hour, in that piglets born later in the birth order spent less time alone on the floor than piglets born earlier in the birth order. However, relative birth order did not relate to latency to first successful suckle. Figure 6. View largeDownload slide Scatter plot of birth weight, kg (w), and time to first suckle, log(h + 0.001) (su). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 19.05. Figure 6. View largeDownload slide Scatter plot of birth weight, kg (w), and time to first suckle, log(h + 0.001) (su). Each open dot represents the data of 1 piglet. The solid line represents a simple linear regression of the 2 variables. Bayesian information criteria = 19.05. DISCUSSION The results presented in this paper provide insight into the relationships between different piglet traits and behaviors associated with the recovery from early postnatal hypothermia. Identifying the indicators of neonatal survival and the role of postnatal hypothermia has been the subject of numerous studies (Tuchscherer et al., 2000; Malmkvist et al., 2006; Baxter et al., 2008, 2009). The application of the graphical model allows for a detailed overview of the structural relationships among different variables involved, which would not be possible through the use of ANOVA and other more traditional statistical methods. The ability of the individual piglet to overcome postnatal hypothermia and restore body temperature after the initial postpartum drop was directly linked to birth weight, and to piglet position in the pen during both the first and second hours after birth. Thus, low rectal temperature 2 h after birth was related to lighter birth weight and to the piglet being found more often alone on the floor as opposed to by the udder during the first and second hours after birth. The nature of these effects will be discussed in further detail below. The effect of birth weight on thermoregulatory ability is supported by the findings of others (Herpin et al., 2004; Baxter et al., 2008). Furthermore, Stanton et al. (1973) found that newborn piglet performance in a cold challenge was highly correlated with birth weight. Smaller piglets have a greater surface-area-to-body-volume ratio compared with larger piglets, and they are therefore more prone to heat loss in a cold environment. However, the simple body proportion index used in this study (crown-rump/girth) did not have the effect on thermoregulation, which might have been expected. Furthermore, it only had a weak edge with birth weight, which means that it is of little importance in the collective model. More advanced descriptions of body proportions might have a larger part of this effect than the simple index from the present study. Previous studies (Baxter et al., 2008, 2009) have used ponderal index [birth weight, kg/(crown – rump length, m)3] and body mass index [birth weight/(crown – rump length)2] to describe piglet shape in relation to BW. However, in both studies (Baxter et al., 2008, 2009) these indexes affected prenatal mortality more than postnatal mortality, and the effects on rectal temperature at 1 h were minor. Body weight-related differences in physiological maturity could be another possible explanation for the relationship between thermoregulatory success and birth weight. Herpin et al. (2004) related summit metabolic rate (SMR) to birth weight in neonatal European crossbred piglets, like the Landrace × Yorkshire used in the present study, and Chinese Meishan. The study showed impaired SMR in the European crossbred piglets of lighter birth weights. The SMR of the piglets <1,125 g could not be described by the same linear relationship with BW as for the piglets >1,125 g. The data on SMR and birth weight had to be described by a broken-line model (Herpin et al., 2004). This was in contrast to SMR in neonatal Meishan piglets from the same study (Herpin et al., 2004) that showed a uniform linear relationship with birth weight over the range. In a study on piglet mortality, Marchant et al. (2000) found a curvilinear relationship between birth weight and survival rate. In the same study (Marchant et al., 2000), only 28% of piglets <1,100 g were alive after 7 d. Relating the plot of birth weight and thermoregulatory success in the present study to the previous studies, there may be a minimum birth weight near 1,100 g below which piglets of modern European breeds have increasingly impaired thermogenic capacity (Herpin et al., 2004) and greater mortality rates (Marchant et al., 2000). The main focus of genetic selection in pigs has been to increase litter size. However, increased litter size is associated with reduced average birth weight (Roehe, 1999; Sorensen et al., 2000), which was also observed in the present study. Predictably, the percentage of piglets of lighter birth weight affects the prevalence of severe hypothermia. Based on the effect of birth weight in the present study and others, it might be beneficial to select against low birth weights in swine breeding programs in the future to reduce the occurrence of hypothermia. The other direct effect on piglet rectal temperature at 2 h apart from birth weight were piglet position in the pen during both the first and the second hours after birth. The result emphasizes the importance of appropriate behavioral adaptations to minimize heat loss and the need to increase heat production under less than favorable thermal conditions. Piglets that stayed by the udder rather than settling alone for longer periods gained a thermal advantage through heat conduction from contact with the udder of the sow, and by huddling with littermates, thereby reducing their overall surface area exposed to lower air temperature. Behavioral adaptations are therefore vital, not only to reduce heat loss, but huddling and keeping to the warm areas of the pen further enables heat influx from warmer objects. The edges between RT_2h and pos_1h and of RT_2h and pos_2h emphasize the significant importance of thermoregulatory behavior in achieving normothermic core temperature at 2 h after birth. The effect of settling in a thermally suitable position is shown by the direct edges of piglet position and thermoregulatory success. Early colostrum intake also has a strong influence on piglet position in the pen in both the first and the second hours after birth. However, there was no direct edge between time to suckle and rectal temperature at 2 h. This means that the effect of early colostrum intake on thermoregulatory success is mediated through its interaction with piglet position. This is supported by the findings of Le Dividich and Noblet (1981) who kept neonatal piglets away from the sow at 18 to 20°C (cool) and 30 to 32°C (warm) for the first 24 h after expulsion, while allowing them to suckle at intervals of 80 to 85 min. The development of piglet rectal temperature was strongly affected by the environmental treatment, whereas the effect of colostrum intake on rectal temperature was not significant before 15 h after birth and 11 sucklings. This supports the result in the present study that at this early stage, it is heat preservation rather than colostrum intake that determines thermoregulatory success. Although piglets are readily attracted to heat from birth (Mount, 1963), this is exceeded by their motivation to search for colostrum (Hartsock and Graves, 1976), and some of the piglets end up in the cold zones of the pen in their search for colostrum. Achieving the first successful ingestion of colostrum radically changes piglet behavior as described by Hartsock and Graves (1976). In the teat-seeking phase before the first successful ingestion of colostrum, any vertical surface with which the piglet comes in contact elicits a teat-seeking response. After the first successful suckle, teat recognition is facilitated and the piglet subsequently stays close to the udder. Piglets that have not achieved first suckle are more likely to settle or huddle in an area that is too cold (Hrupka et al., 2000) or to stay on a continuous search for colostrum in an area which is not warm enough, rather than to seek out the heated creep area. Therefore, although the idea of creating a heterothermic environment to accommodate the thermal comfort zones of both the piglets and the sow is compelling, the risk of fatal hypothermia in the piglets is great. Although the thermal environment in the present study was far from the comfort zone of the piglets, the piglets were only found in the heated creep area for 4.9% of the observations during the first 2 h. To reduce postnatal hypothermia and prevent the condition from becoming fatal, a heated creep area is insufficient, and further investigation on optimizing the thermal environment to better meet piglet needs at the birth site is required. Ingestion of colostrum has documented effects on metabolic capacity for heat production (Herpin et al., 1994). In this study, time to first suckle had no direct effect on thermoregulatory success, but very strong edges with piglet position, particularly during the first hour, and also during the second hour after birth. Colostrum intake was not measured in the present study, but it is highly likely that the piglets that were found more often by the udder were also the piglets that achieved the greatest colostrum intake. However, as Le Dividich and Noblet (1981) showed, the effect of colostrum intake on rectal temperature is not significant until much later. None of the variables related to the birth process had any direct effects on thermoregulatory success. The variables of the birth process did, however, connect to behavior during the first 2 h, and these in turn affect recovery from postnatal hypothermia. A relation between hypoxia and thermoregulation has been demonstrated earlier (Stanton et al., 1973; Herpin et al., 1996). However, the results of the present study suggest that in the thermal environment of the common production system, it is not a direct effect of hypoxia on heat production that causes poor recovery from postnatal hypothermia. In this environment, it is heat preservation rather than mere heat production that determines thermoregulatory success and recovery from postnatal hypothermia, although the birth process influences heat preservation behavior. Large variation in piglet mortality rate among litters has led to several studies investigating the effect of the sow on neonatal piglet mortality including sow reactivity to piglet scream (Wechsler and Hegglin, 1997), lying and rolling behavior (Damm et al., 2005), and other aspects of maternal behavior (Andersen et al., 2005; Pedersen et al., 2006). We therefore expected that there could be a sow effect on piglet recovery from hypothermia. In this study, we did not register sow behavior, but included sow identity as a means to address the issue of variation among litters caused by the sow. Using this approach, we were unable to confirm any direct or indirect effects of sow on piglet recovery from postnatal hypothermia. During the first 2 h of the life of the piglet, which has been the focus of this study, farrowing will in most cases still be in progress, during which time most of the sows lay practically immobile. The role of the sow in piglet thermoregulation at this early stage seems primarily to be to serve as a heat source. Second, the sow is a source of colostrum, and ingestion of colostrum will result in improved heat production capacity (Herpin et al., 1994). Maternal care at this stage is minimal, and the piglet is in control of utilizing the sow as a heat source. Some sows are more restless, and this could affect some of the piglets in their litter. However, any possible effect of the sow on individual piglets did not result in differences between litters; for all variables included in this model, the variation between litters was exceeded by the variation within litter. In this model of thermoregulatory success at 2 h, variables above the piglet level including sow identity and litter size had a very limited effect on other variables, and no direct or close connection to thermoregulatory success. The application of the block-recursive model on an extensive data set, as in the present study, provides new insights into the interactions and ideas of causal relationships, which were not possible through more traditional statistical methods such as variance analysis, correlations, and multivariate mixed models. Although correlations give the strength of the relationship between 2 variables, graphical block models take into account all of the links from the preceding blocks before relating one to the next. This makes it a great supplement to traditional statistics when addressing multifactorial challenges in animal husbandry and other fields of science. As with any other tool for analysis, the limitations should not be overlooked, and the resulting graphical model should be seen as a starting point specifying hypotheses that need to be investigated in confirmative experiments. In the block recursive model of thermoregulatory success at 2 h, birth weight and piglet position in the pen during both the first and the second hour were the ultimate affecting variables. Birth weight was the most important of the 3 affecting variables, and lighter birth weight piglets are particularly at risk of poor recovery from postnatal hypothermia. Piglets that were found more often by the udder as opposed to alone on the floor were also more successful in recovering from postnatal hypothermia. These results illustrate that in the common heterothermic environments in farrowing housing in the pig industry, being born with sufficient birth weight and keeping to the warm areas of the pen are absolutely necessary to reduce the extent and duration of postnatal hypothermia. Based on the results of the present study, it could be beneficial for reducing hypothermia to select against lighter birth weight in swine breeding programs. Improvement of the thermal farrowing environment is an area in need of further research. LITERATURE CITED Andersen I. L. Berg S. Boe K. E. 2005. Crushing of piglets by the mother sow (Sus scrofa)—Purely accidental or a poor mother? Appl. Anim. Behav. Sci.  93: 229– 243. Google Scholar CrossRef Search ADS   Baxter E. M. Jarvis S. D'Eath R. B. Ross D. W. Robson S. K. Farish M. Nevison I. M. Lawrence A. B. Edwards S. A. 2008. Investigating the behavioural and physiological indicators of neonatal survival in pigs. Theriogenology  69: 773– 783. https://doi.org/18242685 Google Scholar CrossRef Search ADS PubMed  Baxter E. M. Jarvis S. Sherwood L. Robson S. K. Ormandy E. Farish M. Smurthwaite K. M. Roehe R. Lawrence A. B. Edwards S. A. 2009. Indicators of piglet survival in an outdoor farrowing system. Livest. Sci.  124: 266– 276. Google Scholar CrossRef Search ADS   Berthon D. Herpin P. Duchamp C. Dauncey M. J. Le Dividich J. 1993. Modification of thermogenic capacity in neonatal pigs by changes in thyroid status during late-gestation. J. Dev. Physiol.  19: 253– 261. https://doi.org/7963301 Google Scholar PubMed  Damm B. I. Forkman B. Pedersen L. J. 2005. Lying down and rolling behaviour in sows in relation to piglet crushing. Appl. Anim. Behav. Sci.  90: 3– 20. Google Scholar CrossRef Search ADS   Edwards, D. 2000. Introduction to Graphical Modelling—Second Edition.  2nd ed. Springer-Verlag Inc., New York, NY. Google Scholar CrossRef Search ADS   Hartsock T. G. Graves H. B. 1976. Neonatal behavior and nutrition-related mortality in domestic swine. J. Anim. Sci.  42: 235– 241. https://doi.org/1248968 Google Scholar CrossRef Search ADS PubMed  Herpin P. Le Dividich J. Berthon D. Hulin J. C. 1994. Assessment of thermoregulatory and postprandial thermogenesis over the first 24 hours after birth in pigs. Exp. Physiol.  79: 1011– 1019. https://doi.org/7873158 Google Scholar CrossRef Search ADS PubMed  Herpin P. Le Dividich J. Hulin J. C. Fillaut M. DeMarco F. Bertin R. 1996. Effect of the level of asphyxia during delivery on viability at birth and early postnatal vitality of newborn pigs. J. Anim. Sci.  74: 2067– 2075. https://doi.org/8880407 Google Scholar CrossRef Search ADS PubMed  Herpin P. Vincent A. Damon M. 2004. Effect of breed and body weight on thermoregulatory abilities of European (Pietrain × (Landrace × Large White)) and Chinese (Meishan) piglets at birth. Livest. Prod. Sci.  88: 17– 26. Google Scholar CrossRef Search ADS   Højsgaard, S. 2007. The mimR Package for Graphical Modelling in R.  Accessed Jul. 7, 2008. http://gbi.agrsci.dk/~shd/public/mimR/. Højsgaard S. Thiesson B. 1995. Bifrost—Block recursive models induced from relevant knowledge, observations, and statistical techniques. Comput. Stat. Data Anal.  19: 155– 175. Google Scholar CrossRef Search ADS   Hoy S. Lutter C. Puppe B. Wahner M. 1995. Influence of early postnatal vitality of newborn piglets on live weight development and mortality up to 28th day of life. Arch. Tierzucht  38: 319– 330. Hrupka B. J. Leibbrandt V. D. Crenshaw T. D. Benevenga N. J. 2000. The effect of thermal environment and age on neonatal pig behavior. J. Anim. Sci.  78: 583– 591. https://doi.org/10764064 Google Scholar CrossRef Search ADS PubMed  Lauritzen S. L. Richardson T. S. 2002. Chain graph models and their causal interpretations. J. Royal Stat. Soc. Series B Stat. Method  64: 321– 348. Google Scholar CrossRef Search ADS   Lauritzen S. L. Wermuth N. 1989. Graphical models for associations between variables, some of which are qualitative and some quantitative. Ann. Stat.  17: 31– 57. Google Scholar CrossRef Search ADS   Le Dividich J. Noblet J. 1981. Colostrum intake and thermoregulation in the neonatal pig in relation to environmental-temperature. Biol. Neonate  40: 167– 174. https://doi.org/7284502 Google Scholar CrossRef Search ADS PubMed  Malmkvist J. Pedersen L. J. Damgaard B. M. Thodberg K. Jorgensen E. Labouriau R. 2006. Does floor heating around parturition affect the vitality of piglets born to loose housed sows? Appl. Anim. Behav. Sci.  99: 88– 105. Google Scholar CrossRef Search ADS   Marchant J. N. Rudd A. R. Mendl M. T. Broom D. M. Meredith M. J. Corning S. Simmins P. H. 2000. Timing and causes of piglet mortality in alternative and conventional farrowing systems. Vet. Rec.  147: 209– 214. https://doi.org/10994922 Google Scholar CrossRef Search ADS PubMed  Mount L. E. 1959. The metabolic rate of the new-born pig in relation to environmental temperature and to age. J. Physiol.  147: 333– 345. https://doi.org/14424736 Google Scholar CrossRef Search ADS PubMed  Mount L. E. 1963. Environmental temperature preferred by young pig. Nature  199: 1212– 1213. https://doi.org/14072066 Google Scholar CrossRef Search ADS PubMed  Moustsen, V. A., and T. Jensen 2008. Inventar til forbedring af hygiejne i stier til løsgående farende og diegivende søer.  Accessed Apr. 14, 2011. http://vsp.lf.dk/Publikationer/Kilder/notater/2008/0809.aspx. Moustsen, V. A., L. J. Pedersen, and T. Jensen 2007. Afprøvning af stikoncepter til løse farende og diegivende søer.  Accessed Apr. 14, 2011. http://vsp.lf.dk/Publikationer/Kilder/lu_medd/2007/805.aspx. Pedersen L. J. Jorgensen E. Heiskanen T. Damm B. I. 2006. Early piglet mortality in loose-housed sows related to sow and piglet behaviour and to the progress of parturition. Appl. Anim. Behav. Sci.  96: 215– 232. Google Scholar CrossRef Search ADS   R Development Core Team 2008. R: A Language and Environment for Statistical Computing.  R Foundation for Statistical Computing, Vienna, Austria. Roehe R. 1999. Genetic determination of individual birth weight and its association with sow productivity using Bayesian analyses. J. Anim. Sci.  77: 330– 343. https://doi.org/10100660 Google Scholar CrossRef Search ADS PubMed  Sorensen D. Vernersen A. Andersen S. 2000. Bayesian analysis of response to selection: A case study using litter size in Danish Yorkshire pigs. Genetics  156: 283– 295. https://doi.org/10978292 Google Scholar PubMed  Stanton H. C. Brown L. J. Mueller R. L. 1973. Interrelationships between maternal and neonatal factors and thermoregulation in fasted neonatal swine (Sus domesticus). Comp. Biochem. Physiol.  44: 97– 105. https://doi.org/4404872 Google Scholar CrossRef Search ADS   Tuchscherer M. Puppe B. Tuchscherer A. Tiemann U. 2000. Early identification of neonates at risk: Traits of newborn piglets with respect to survival. Theriogenology  54: 371– 388. https://doi.org/11051321 Google Scholar CrossRef Search ADS PubMed  Wechsler B. Hegglin D. 1997. Individual differences in the behaviour of sows at the nest-site and the crushing of piglets. Appl. Anim. Behav. Sci.  51: 39– 49. Google Scholar CrossRef Search ADS   Wermuth, N. 2003. Analysing social science data with graphical Markov models. Pages 33–39 in Highly Structured Stochastic Systems.  P. Green, N. Hjort, and S. Richardson ed. Oxford University Press, Oxford, UK. American Society of Animal Science
TI - Hypothermia in neonatal piglets: Interactions and causes of individual differences
JF - Journal of Animal Science
DO - 10.2527/jas.2010-3022
DA - 2011-07-01
UR - https://www.deepdyve.com/lp/oxford-university-press/hypothermia-in-neonatal-piglets-interactions-and-causes-of-individual-y1vHPU883A
SP - 2073
EP - 2085
VL - 89
IS - 7
DP - DeepDyve
ER -