The experiment followed the guidelines documented in The Ethical Principles and Guidelines for the Use of Animals for Scientific Purposes edited by the National Research Council of Thailand, and was approved by the Institutional Animal Care and Use Committee (IACUC) in accordance with the university regulations and policies governing the care and use of experimental animals (approval no. 1431063).

The present study was carried out in a 3,500-sows commercial swine herd in the western part of Thailand between June and August 2013. The average ambient temperature during the experimental period ranged from 25.8°C to 30.0°C. The minimum and maximum temperature ranged from 21.1°C to 26.3°C and from 28.1°C to 37.6°C, respectively. The average relative humidity varied from 72.0% to 96.0%. Sows were kept in individual crates (1.2 m²) during gestation in a conventional open-housing system and were provided with fans and individual water sprinklers to reduce the impact of high ambient temperature. During gestation, sows were fed a commercial gestation diet that met or exceed their nutritional requirement estimates (19.7% crude protein, 5.8% fat, 3.8% fibre, and 14.0 metabolizable energy, MJ/kg) [13]. Feed was provided twice a day following a standardised feeding pattern, resulting in an average of 2.5 kg per sow per day. During lactation, sows were fed twice a day with a commercial lactation diet, increasing the daily amount of feed offered, until ad libitum feed was reached after one week of lactation. The animals were received water ad libitum in a continuous water channel. Pregnant sows were moved to the farrowing house about one week before their expected farrowing date. Sows were kept in individual farrowing crates (1.2 m²) placed at the centre of the pens with a space allowance of 4.2 m². The pens were fully slatted with concrete at the centre for sows and with steel slats at both sides of the farrowing crate for piglets. Each pen was provided with a creep area for piglets (0.60 m²) placed on the floor on one side, covered by a plastic plate and heating lamp during the first week after farrowing. The heating lamp was usually turned on during the night or when the environment temperature fell below 30°C. The temperature in the creep area was between 30°C to 36°C.

The parturition process was carefully supervised by the first author (M. Nuntapaitoon). Briefly, sows were interfered with as little as possible during parturition, and birth intervention was performed only when dystocia was clearly identified. Dystocia was considered when an interval of >30 min elapsed from the birth of the last piglet, and when the sow showed intermittent straining accompanied by paddling of the legs or when the sow expelled small quantities of foetal fluid together with marked tail switching for >30 min without any piglet being born. Routine procedures performed on piglets included weighing, tail docking, tooth clipping and 1 mL (200 mg) iron supplement administered intramuscularly (Gleptosil, Alstoe Ltd. Animal Health, Leicestershire, England) on the day of birth. Piglets were orally administered a coccidiocide of 20 mg/kg body weight (Baycox [Toltrazuril 5.0% oral suspension], Bayer Inc., Mississauga, ON, Canada) to control neonatal coccidiosis at 3 days of age. In total, 805 piglets born from 57 Landrace×Yorkshire crossbred sows were included. The mean parity was 4.0±1.6 (ranged 2 to 7). Lactation length was on an average of 23.0±2.0 days.

The following reproductive variables of the sows were recorded: farrowing duration (i.e., time between the first and last born piglets), total number of piglets born per litter (TB), number of piglets born alive per litter (BA), and mummified foetuses, stillbirths and number of piglets at weaning per litter. The piglets’ heart rate and blood oxygen saturation (SatO₂) were monitored within 5 min after birth by using a veterinary pulse oximetry (EDAN VE-H100B Pulse Oximeter, Edan Instrument Inc., San Diego, CA, USA). Thereafter, blood samples for glucose analysis were collected from piglets within 5 to 10 min after birth (i.e., before first suckling). A small amount of blood sample (a mixture of venous and arterial blood) was obtained from the cut umbilical cord. Few drops of blood sample were used to determined blood glucose concentration by using a portable test kit of human glucometer (Accu-Chek Performa, Roche, Mannheim, Germany). Birth order, birth interval (the time elapsed between each piglet born), sex, and time elapsed from birth until first attempts to stand (3 groups: <1 min, 1 to 5 min, and >5 min) were recorded for each piglet. Also, birth assistance was recorded (yes or no) if required by piglets. Skin color of the piglets was recorded at birth and was classified into 2 groups (normal or pale). Integrity of the umbilical cord of each piglet was examined and classified into two groups (intact or broken). Rectal temperature was measured at 24 h after birth (RT24h) with a digital thermometer (Microlife, Microlife AG Swiss Corporation, Widnau, Switzerland, with a display resolution of 0.01°C and ±0.1°C accuracy). All piglets were individually identified by an ear tattoo. Body weight of the piglets was measured immediately at birth. Litters were equalized to 13 or 14 piglets per litter after 24 h and within 48 h after birth. The LS was defined as the number of piglets after cross-fostering. Piglets were weighed again at day 7 and 21 after birth. Mortality rate of the piglets was also determined at day 7 and 21 of lactation. Creep feeding and drinking water were provided to the piglets from 7 days of age until weaning.

All statistical analyses were performed using SAS 9.0 (SAS Inst. Inc., Cary, NC, USA) [14]. Descriptive analysis (mean±standard deviation [SD] median and range) and frequency analysis were obtained for all reproductive parameters. To identify the potential indicators for piglet mortality and piglet average daily weight gain (ADG) at days 7 and 21 of lactation, each recorded factor was individually tested using univariate analyses. Continuous variables indicating the neonatal piglet characters (i.e., TB, BA, LS, birth order, BW_B, birth interval, heart rate, SatO₂ and blood glucose concentration) were compared between piglets dying and surviving at days 7 and 21 of lactation by using Student’s t test (PROC TTEST). The association between categorical variables (i.e., sex, skin color, attempts to stand, umbilical cord integrity and birth assistance) and the piglet mortality (dead or alive) were analyzed using Chi-square test. The effect of these categorical variables on ADG at days 7 and 21 of lactation was analyzed using the PROC GLM of SAS. Pearson’s correlation was performed to study collinearity among the continuous variables in the univariate models.

For multivariate analyses, generalised linear mixed models (GLMMIX macro) with the dependent variable (mortality) modeled as a binary outcome (dead or alive) was conducted. Factors with significant levels of p<0.10 were included in the final models. Highly correlated variables were not included in the same multivariate analyses models. Then, the final multivariate GLMMIX models for piglet mortality at days 7 and 21 of lactation included BW_B of the piglet classes (<1.30, 1.30 to 1.79, and ≥1.80 kg) and RT24h classes (<37.0°C, 37.0°C to 38.5°C, and >38.5°C). The classification of the independent variables was made according to frequency distribution with some minor adjustment based on biological reliability. Similarly, general linear mixed models (PROC MIXED) were conducted to analyze piglet ADG at days 7 and 21 of lactation. The final multivariate MIXED models for piglet ADG at day 7 and 21 included BW_B classes (<1.30, 1.30 to 1.79, and ≥1.80 kg), blood glucose concentration classes (≤24 and >24 mg/dL), BA classes (<12, 12 to 14, and ≥15 piglets), sex and attempts to stand (<1, 1 to 5, and >5 min). In all models, sow or litter were introduced as a random effect. Least squares means were obtained from each class of the variables and were compared using Tukey–Kramer adjustment for multiple comparisons. p< 0.05 was regarded to be statistically significance and 0.05<p<0.10 was considered as a tendency for statistically significance. Piglets were defined as experimental units.