Evaluating the Predictive Value of Estimated Breeding Values of Piétrain Sires for Uniformity and Survival in Their Crossbred Progeny

ABSTRACT

High litter sizes in pigs are associated with lower birth weights and increased within‐litter variation, which poses challenges to pig farming. Improving piglet uniformity and survival via genetic selection of terminal boars could be a strategy to enhance these traits. This study evaluated how well estimated breeding values for uniformity and survival of terminal Piétrain sires predict within‐litter uniformity and mortality rates in their crossbred offspring from birth until slaughter. We selected six Piétrain boars with contrasting estimated breeding values for uniformity and survival (3 high vs. 3 low) and mated these with 93 hybrid sows, which produced 1421 liveborn piglets. Moderate correlations were observed between crossbred piglets' within‐litter uniformity and the paternal estimated breeding values (r = 0.17 to 0.28), indicating that the used estimated breeding values for uniformity and survival of terminal sires have a low predictive value for within‐litter uniformity. Nonetheless, significant associations were found between these estimated breeding values and specific survival‐ and uniformity‐related traits, such as pre‐weaning survival and uniformity of growth, supporting their potential for selection. When analysing piglets' weights over time, low correlations were observed between the coefficient of variation at birth and later in life (r = 0.05 to 0.24). On the other hand, moderate to high correlations were found between coefficients of variation at weaning and later in life (r = 0.34 to 0.88), indicating stability of the within‐litter body weight uniformity after weaning. These findings suggest that pre‐ and post‐weaning within‐litter body weight uniformity should be considered as two distinct traits.

1. Introduction

The selective breeding for increased litter sizes has led to an unintended rise in the number of lightweight piglets and greater within‐litter body weight variation (Quesnel et al. 2008; Zindove et al. 2014; Marandu et al. 2015; Nam and Sukon 2022). Until the early 2000s, genetic selection in pigs has been mainly focused on improving production efficiency traits, such as growth rate, litter size and feed efficiency. Since then, most breeding organizations have adopted more balanced breeding goals that also emphasize resilience, survival and robustness, although selection for increased litter size continues to be a key driver (Knap 2020; EFFAB 2023).

The higher prevalence of lightweight piglets and increased birth weight variation negatively impact both piglet survival and the economic efficiency of pig production (Quiniou et al. 2002; Baxter and Edwards 2018). Knol et al. (2022) indicated that birth weight is crucial for piglet survival with little chance of survival for piglets weighing less than 0.5 kg at birth. The probability of survival rises to 63% for 0.8 kg piglets, 84% for 1.0 kg piglets and reaches 95% for piglets weighing 1.5 kg at birth. Increased variation in within‐litter body weights further lowers the survival chances of lightweight piglets due to competition with heavier littermates (Milligan, Dewey, and de Grau 2002; Milligan, Fraser, and Kramer 2002; Deen and Bilkei 2004). Piglets with lower birth weights have reduced growth rates throughout the entire production cycle, requiring more days to reach market weight (Gondret et al. 2005; Beaulieu et al. 2010; Douglas et al. 2014). These underperforming lightweight pigs typically remain longer on the farm, thereby increasing feed and labour costs and complicating farm logistics. Moreover, the practice of sorting pigs by body weight to manage such variation is a labour and time‐consuming task. Therefore, minimizing within‐litter variation improves not only piglet survival and growth performance but also contributes towards a more efficient and manageable production system.

Body weight uniformity is a multifactorial and complex trait, influenced by a wide range of factors, such as litter size, parity and other maternal characteristics (Quesnel et al. 2008; Wientjes et al. 2013; Zindove et al. 2014; Lavery et al. 2019; Buthelezi et al. 2024). In the short term, targeted management strategies such as sow nutrition, split‐suckling and weight‐based sorting of piglets may reduce variability by supporting development and the growth of lightweight piglets and minimize competition (Donovan and Drits 2000; Campos et al. 2012; Faccin et al. 2020; Hooyberghs, Goethals, de Mey, Gorssen, et al. 2025).

However, genetic selection could be a way forward for sustainable and long‐term improvements. Damgaard et al. (2003) reported low heritabilities of 0.08 and 0.06 for within‐litter weight uniformity at birth and at 3 weeks of age. The genetic correlations (r _g) between SD of birth weight and pre‐weaning mortality (r _g = 0.25) and mean growth of piglets during lactation (r _g = −0.31) indicate that increased variation in birth weight is associated with increased pre‐weaning mortality and lower growth rates (Damgaard et al. 2003). Furthermore, Camargo et al. (2020) noted genetic correlations between the CV of birth weight and average birth weight (r _g = −0.93) and litter size (r _g = 0.64), suggesting that selection for lower within‐litter birth weight variation might increase average birth weight but lower litter sizes. Yang et al. (2023) found a positive genetic correlation between litter size and number of weak piglets born (r _g = 0.52), indicating that selecting for increased litter sizes could unintentionally increase the number of piglets born weak. Moreover, Gorssen et al. (2025) showed that a Piétrain breeding program for improved piglet vitality and less congenital defects led to substantially higher vitality and survival of the crossbred progeny (+3%), but the effects on body weight uniformity were not studied.

Selection based on sire estimated breeding values (EBVs) for uniformity, vitality and survival of terminal sires might be an approach to improve the performance of their crossbred offspring. However, no study has yet tested whether sire EBVs for uniformity and survival in Piétrain sires translate into measurable improvements in crossbred offspring. Therefore, the aim of this study is to evaluate whether sire EBVs for uniformity (CV and SD; see below) and survival of terminal Piétrain sires improve within‐litter body weight uniformity and reduce mortality in crossbred offspring from birth to slaughter.

2. Materials and Methods

According to Belgian and EU regulations (Council Directive 86/609/EEC), approval from the local ethics committee was not required for this study.

2.1. Animals and Management

This study was conducted at two experimental farms of ILVO (Merelbeke‐Melle, Belgium). First, six Piétrain boars from the Flemish Piétrain breeding organization (Vlaamse Piétrain Fokkerij vzw) were selected based on a selection index specifically designed for this study (see below). The index was based on the available progeny testing information in the breeding program, which is explained in detail in Gorssen et al. (2024) and Gorssen et al. (2025). The sire EBVs for the post‐weaning traits are routinely calculated based on a subset of the offspring of each boar, likely excluding the lightweight piglets. In short, EBVs were calculated using the blupf90 software (Misztal et al. 2023) with a single‐step GBLUP sire‐dam model using genotypes and pedigree of Piétrain sires and pedigree only of crossbred dams. For the litter traits (vitality, number of lightweight piglets, preweaning mortality, congenital defects), factors in the model were sow parity, farm, gestation length, number of piglets born alive, weaning age (fixed effects) and common sow group (random effect), as explained in Gorssen et al. (2025). For the production traits (SD of daily gain, daily gain, feed conversion ratio…), fixed factors in the models were sex‐ratio of the pen (linear, quadratic), number of pigs at start in the pen, space per pig (m²/pig), parity of the litter, breed of the dam and batch (random effect).

The EBVs of all sires of the breeding company were rescaled to a range between −1 and +1, with −1 representing an unfavourable outcome (e.g., low vitality and high mortality) and + 1 indicating a favourable outcome (e.g., high vitality and low mortality). This rescaling was performed to correct for the difference in variance across the sire EBVs and to directly compare EBVs. Next to sire EBVs, two extra uniformity traits were collected per boar based on the phenotypes of their previously tested offspring: the CV of pre‐testing average daily gain (ADG; average daily gain from birth to the start of the finishing phase) and CV of total ADG of previous offspring.

For this study, an ad‐hoc sire selection index for survival and uniformity was designed using the following formula:

$$\begin{array}{c}\text{sire survival}-\text{uniformity index}\hfill \\ =0.20*\text{piglet vitality sire index}\hfill \\ +0.10*\text{sire}\mathrm{EBV}\text{lightweight piglets}\hfill \\ +0.15*\mathrm{EBV}\text{production trait sire index}\hfill \\ +0.20*\mathrm{CV}\text{of}\mathrm{pre}-\text{testing daily gain of previous offspring}\hfill \\ +0.20*\text{sire}\mathrm{EBV}\mathrm{SD}\text{of daily gain}\hfill \\ +0.15*\text{sire}\mathrm{EBV}\text{mortality finishing phase}\hfill \end{array}$$

Three boars with a low sire survival‐uniformity index (−0.15 to −0.01) and three boars with a high sire survival‐uniformity index (0.33 to 0.51) were selected (Table S1 includes the data per sire). Boars with a high sire survival‐uniformity index were expected to produce more vital and uniform offspring compared to boars with a lower sire survival‐uniformity index. Litters were obtained by mating each of these six Piétrain boars with 12–18 hybrid sows in three batches at each farm (93 sows in total). Per boar, sows were balanced by parity and batch within farms. At farm 1, 42 sows (Topigs Norsvin; TN70) produced a total of 678 liveborn piglets. At farm 2, a total of 743 piglets were born alive from 51 sows (RA‐SE genetics). Since sow line and farm effect were confounded, they were treated as two independent datasets.

2.2. Phenotypic Data Collection and Trait Definitions

The body weight and backfat thickness of the sows were recorded upon entering the farrowing pen (±1 week before farrowing) and at weaning. Backfat thickness was measured using a real‐time ultrasound instrument (Renco lean indicator, Golden Valley, Minnesota, United states). Piglets were individually weighed within 24 h after birth (N = 1421), at weaning (27.2 ± 1.5 days of age; N = 1146), 1 week post‐weaning (34.2 ± 1.5 days of age; N = 1104), 2 weeks post‐weaning (41.2 ± 1.5 days of age; N = 1099) and at the end of the nursery (62.2 ± 1.5 days of age; N = 1095). Due to limited housing capacity at the farms, it was not possible to retain all piglets until slaughter. As part of the standard farm practice, a proportion of piglets was sold at the end of the nursery phase. At weaning, five piglets per litter from farm 1 were selected for follow‐up to slaughter. Selection was performed in such a way that the CV of the selected subsets reflected the CV of the original weaned litters, thereby maintaining the within‐litter weight distribution. These five piglets per litter were transferred together to the finishing barn. This procedure was applied in batches 1 and 3. In batch 2, all weaned piglets were retained and moved to the finishing barn, no selection or sale occurred in that round. Only a subset of the pigs was monitored during the finishing phase (358 pigs selected at farm 1 from a total of 1095 pigs at the end of the nursery phase). During the finishing phase, body weight was collected at 16 weeks of age (109.3 ± 1.6 days; N = 357) and at the day of slaughter (167.4 ± 6.7 days; N = 354). The pigs were fasted the day before slaughter. In the slaughterhouse, after a transport and lairage time of approximately two hours, pigs were stunned using carbon dioxide and then exsanguinated. At the slaughter line, carcass quality traits including lean meat content, carcass weight and center of gravity (COG) were assessed using the ’AutoFOM III’ system (Fully Automatic Ultrasonic Carcass Grading AutoFom IIITM 2019). Carcass data could only be retrieved for 331 of the 354 slaughtered pigs, due to the loss of electronic ear tags on the slaughter line (93.5%). Daily carcass growth was determined by subtracting birth weight from the hot carcass weight and dividing the result by the age at slaughter. Furthermore, individual ADG was calculated for five periods: pre‐weaning (ADG_lactation), the nursery phase (ADG_nursery), the finishing phase (ADG_testing), from birth to the start of the finishing phase (ADG_pre‐testing) and from birth to slaughter (ADG_total). Within‐litter CV and SD were calculated for body weight at each weighing moment to quantify within‐litter uniformity. For each litter, the following information was recorded: sow and boar identity, sow's parity, farrowing and weaning dates, number of piglets born alive (NBA) and the number of male and female liveborn piglets. Parities were regrouped into seven categories representing parity 1 (N _sows = 10), parity 2 (N _sows = 19), parity 3 (N _sows = 11), parity 4 (N _sows = 14), parity 5 (N _sows = 10), parity 6 (N _sows = 12) and older sows (parity 7–11; N _sows = 17). The within‐litter sex ratio was calculated and expressed as the percentage of male piglets in a litter. Also, pre‐weaning mortality was calculated as the number of liveborn piglets that died before weaning in a litter. Furthermore, z‐scores (standardized weight) were calculated for the body weight of each pig ($x_i$) as $z_t=\left(x_i-{\mu }_t\right)/{\sigma }_t$, where ${\mu }_t$ is the average body weight of all piglets at age t and ${\sigma }_t$ the SD of body weight at that age. The z‐score indicates how much a pig's body weight deviates from the population mean, measured in SDs. Furthermore, the natural logarithm of the variance in z‐scores (lnvarZ) was calculated per pig from weaning to the start of the finishing phase, as a resilience trait. This measure of growth consistency, based on longitudinal deviations in body weight, indicates the stability of growth compared to the pigs in the dataset (Berghof et al. 2019; Gorssen et al. 2024). A larger lnvarZ indicates more weight deviations compared to the population. Also, an average lnvarZ was calculated per litter (lnvarZ_litter).

2.3. Data Handling and Quality Control of Genotypes

Data quality was validated using descriptive statistics on the variables included in this study (Table 1). Three litters were excluded from the analysis: two because they contained only four liveborn piglets and another because all piglets (N = 23) died before weaning for unknown reasons. Hair root samples were collected from the sows using pliers approximately 14 days into lactation and stored dry on hair cards. These samples were used for genotyping on a medium‐density single nucleotide polymorphism (SNP) chip (Illumina 60 K porcine bead chip array) which includes 67,611 SNPs. Genotype quality control was conducted using PLINK v1.9 software (Chang et al. 2015) (PLINK commands in brackets). First, individual quality control was performed, focusing on outliers in missing (−missing) and heterozygosity (−het) values. Individuals were included when they had a minimum individual call‐rate of 90% and when their average SNP heterozygosity was within three SDs of the mean (genotypes from five sows were excluded). Next, SNP‐level quality control was applied by retaining only autosomal SNPs (−autosome) and ensuring a minimum SNP call rate of 95% (−geno 0.05). Potential duplication errors were checked using genomic relatedness estimates (−genome). The VanRaden genetic relationship matrix was estimated between the sows using the R package AGHmatrix (Amadeu et al. 2023). After quality control, phenotypes of 90 litters/sows (96.8%) with 1394 liveborn piglets and genotypes of 85 sows (91.4%) were available. For a separate project, a selection of piglets (N = 651) was also genotyped and parentage analysis was performed as well based on these data. The PCA plots of piglets, sows and sires were added to the additional files (Figure S1).

TABLE 1.

Descriptive statistics of sow and litter traits (N = 1421 liveborn piglets, N = 93 litters).

Trait	Average	SD	Min; Max
Parity	4.4	2.6	1; 11
Piglets born alive per litter (N = 1421 liveborn piglets)	15.3	3.5	4; 22
Percentage of males per litter (%) (N = 721 male piglets/N = 1421 liveborn piglets)	51.0	13.5	25.0; 81.8
Weight of sow at entering the farrowing house (kg)	305.6	45.1	196.0; 396.0
Weight of sow at weaning (kg)	263.4	44.0	169.0; 381.0
Backfat thickness of sow at entering the farrowing house (mm)	16.5	3.6	5.5; 26.0
Backfat thickness of sow at weaning (mm)	13.1	3.3	5.0; 21.5
Weaning age (d)	27.2	1.5	24; 32
Birth weight (kg; N = 1421 liveborn piglets)	1.45	0.38	0.36; 2.82
Weaning weight (kg; N = 1146 piglets)	7.6	1.8	1.8; 13.9
Weight 1 week post‐weaning (kg; N = 1104 piglets)	8.3	1.8	3.0; 12.9
Weight 2 weeks post‐weaning (kg; N = 1099 piglets)	10.0	2.2	3.1; 16.1
Weight start finishing phase (kg; N = 1095 piglets)	20.4	4.2	6.0; 31.2
Weight at 16 weeks of age (kg; N = 357 piglets)	60.2	10.5	29.0; 94.0
Slaughter weight (kg; N = 354 piglets)	114.7	14.1	70.5; 150.0

2.4. Statistical Analyses

Statistical analyses were performed in R (R Core Team 2024) using the stats, lme4 (Bates et al. 2015), AGHmatrix (Amadeu et al. 2023), and lme4GS (Caamal‐Pat et al. 2021) packages. Normality and linearity were checked visually based on histograms and pairwise correlation plots. First, Pearson correlations were calculated at the litter level between the CVs at seven time points (birth, weaning, 1 week post‐weaning, 2 weeks post‐weaning, start of finishing phase, 16 weeks of age and slaughter) throughout the piglets' life. The data from the five sows that failed genotype quality control were still included in this correlation analysis as genotypic information was not required for these analyses and their exclusion would have unnecessarily reduced the sample size. Second, relationships between paternal EBVs and crossbred piglet phenotypes were investigated via Pearson correlation analyses at the litter level for CV, SD, mortality and lnvarZ_litter and at the individual piglet level for the piglet‐specific traits (body weight, ADG, lnvarZ, percentage of lean meat, COG, or carcass growth), followed by linear (mixed) modelling. Final models were obtained via stepwise regression (Kutner et al. 2005) starting from full models that included all potential associated variables. Non‐significant factors (p > 0.10) were removed using backward selection, starting with the factor showing the highest p‐value.

The final models for the litter‐specific traits (CV, SD, mortality and lnvarZ_litter) were constructed as follows:

$$\mathbf{y}\mathbf{=}\mathbf{Xb}\mathbf{+}\mathbf{e}$$

where $\mathbf{y}$ represents the vector with one of the litter‐specific traits, $\mathbf{b}$ is a vector with the fixed effects boar group (2 levels), batch (3 levels), farm (2 levels) and parity (7 levels) and the covariables sire EBV (when a sire EBV was added, the fixed effect of boar group was removed), NBA, condition of the sow (body weight and backfat thickness before farrowing), sex ratio and weaning age. Weaning age was only included for phenotypes at or after weaning. Next, $\mathbf{e}$ is the vector of residual effects $~N\left(0,{\mathbf{I}\mathrm{\sigma }}_{\mathrm{e}}^2\right)$; $\mathbf{X}$ is an incidence matrix for fixed effects.

The final mixed models for the piglet‐specific traits (body weight, ADG, lnvarZ, percentage of lean meat, COG or carcass growth) were constructed as follows:

$$\mathbf{y}\mathbf{=}\mathbf{Xb}\mathbf{+}\mathbf{Zs}\mathbf{+}\mathbf{e}$$

where $\mathbf{y}$ represents the vector with one of the piglet‐specific traits, $\mathbf{b}$ is a vector with the fixed effects boar group (2 levels), batch (3 levels), farm (2 levels) and sex (2 levels) and covariables sire EBV (when a sire EBV was added, the fixed effect of boar group was removed), NBA, condition of the sow (body weight and backfat thickness before farrowing), sex ratio and weaning age. Next, $\mathbf{s}$ is a vector with the random sow effect (85 levels), assuming that $\mathbf{s}~\mathrm{N}\left(0,{\mathbf{G}\mathrm{\sigma }}_{\mathrm{s}}^2\right)$ and where $\mathbf{G}$ is the genomic relationship matrix; ${\mathbf{\sigma }}_{\mathbf{s}}^{\mathbf{2}}$ represents the variance component associated with the sow effect; $\mathbf{e}$ is the vector of residual effects $~N\left(0,{\mathbf{I}\mathrm{\sigma }}_{\mathrm{e}}^2\right)$; $\mathbf{X}$ and $\mathbf{Z}$ are incidence matrices for fixed effects and random sow effects, respectively.

To assess the effects of the boar group and sire EBVs on the survival of piglets (a binary outcome), generalized linear mixed models with a binomial distribution and logit link function were used. The initial models included the following fixed effects: boar group (2 levels), batch (3 levels), farm (2 levels), parity (7 levels) and sex (2 levels) and the following covariables: sire EBV (when a sire EBV was added, the fixed effect of boar group was removed), NBA, condition of the sow (body weight and backfat thickness before farrowing) and sex ratio (percentage of male piglets). Stepwise regression (removal criterion p > 0.10) was used to obtain the final models (Kutner et al. 2005). For each variable in the models, odds ratios were calculated along with their corresponding 95% confidence intervals (CI).

3. Results

3.1. Relationships Between Within‐Litter Body Weight Uniformity Over Time

No significant correlation coefficients were observed between body weight uniformity at birth and later in life (r = 0.05 to 0.24; p > 0.10; Figure 1). In contrast, the CVs from weaning up to slaughter were moderately to strongly correlated (r = 0.34 to 0.88; p < 0.05). The number and percentage of stillborn piglets were negatively, but not significantly, correlated with the number and percentage of dead piglets in the nursery (r = −0.20 and r = −0.19, respectively; p > 0.10; Figures S2 and S3).

FIGURE 1.

Pairwise Pearson correlation plot between within‐litter body weight coefficients of variation (CVs) from birth to slaughter. p < 0.001***, p < 0.01**, p < 0.05*. [Colour figure can be viewed at wileyonlinelibrary.com]

3.2. Relationships Between Sire EBVs and Within‐Litter Body Weight Uniformity

The ‘high sire survival‐uniformity index’ group had crossbred offspring with lower within‐litter body weight SD and CV at birth compared to the ‘low sire survival‐uniformity index’ group (SD: estimate (ES) = −0.04, p = 0.02; CV: ES = −2.52, p = 0.07; Figure 2, Figure 3, Table 2). Similar results were found for the regression on individual index values per boar for SD and CV (SD: r = −0.17, p = 0.01; CV: r = −0.12, p = 0.08; Figures S4 and S5). The sire survival‐uniformity index had no significant effect on the SD or CV at other timepoints. Piglets of boars from the ‘high sire survival‐uniformity index’ group tended to have less weight variability over time (natural logarithm of the variance in z‐scores; lnvarZ), thus a more stable individual growth (ES = −0.29; p = 0.08; Figure 4 (per boar group) and Figure S6 (per boar)). No significant effect on lnvarZ_litter was observed.

FIGURE 2.

Association between boar groups based on ad‐hoc sire selection index for survival and uniformity and within‐litter body weight uniformity (SD) in pigs. Boxplots and correlation analysis for the individual estimated breeding values are given in Figure S4. [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 3.

Association between boar groups based on ad‐hoc sire selection index for survival and uniformity and within‐litter body weight uniformity (CV) in pigs. Boxplots and correlation analysis for the individual estimated breeding values are given in Figure S5. [Colour figure can be viewed at wileyonlinelibrary.com]

TABLE 2.

Associations between boar group/sire EBVs of sire and offspring phenotypes (uniformity).

Offspring phenotype	Variable of sire	Estimate	p	Correlation (r)
Within‐litter SD at birth (kg)	Boar group (high versus low)	−0.04	0.02	—
	Sire survival‐uniformity index	−0.09	0.01	−0.17
	Sire EBV SD ADG_total	−0.08	0.01	−0.19
	CV ADG_pre‐testing of tested offspring	−0.05	0.001	−0.28
	CV ADG_total of tested offspring	−0.10	0.01	−0.18
Within‐litter CV at birth (%)	Boar group (high versus low)	−2.52	0.07	—
	Sire survival‐uniformity index	−5.08	0.08	−0.12
	CV ADG_pre‐testing of tested offspring	−3.00	0.01	−0.22
	CV ADG_total of tested offspring	−5.82	0.07	−0.17
SD at the start of the finishing phase (kg)	Sire EBV pre‐weaning mortality	−0.84	0.01	−0.23
	Sire EBV stillborn piglets	−1.47	0.005	−0.26
	Sire EBV piglet vitality	−1.24	0.009	−0.20
	Sire EBV ADG_pre‐testing	0.98	0.01	0.28
	Sire EBV daily feed intake	0.51	0.08	0.19
	Sire EBV feed conversion ratio	−1.09	0.02	−0.20
Individual lnvarZ	Boar group (high versus low)	−0.29	0.08	—
	Sire EBV stillborn piglets	0.73	0.08	0.11
	Sire EBV ADG_pre‐testing	−0.81	0.004	−0.17
	Sire EBV ADG_total	−0.46	0.08	−0.12
	Sire EBV SD ADG_total	−0.61	0.03	−0.15

Note: Correlations were not calculated with the variable boar group (N = 2).

Abbreviations: ADG, average daily gain; EBV, estimated breeding value.

FIGURE 4.

Association between boar groups based on ad‐hoc sire selection index for survival and uniformity and uniformity of growth: Individual (lnvarZ) and within‐litter (lnvarZ_litter) in pigs. Boxplots and correlation analysis for the individual estimated breeding values are given in Figure S6. Abbreviation: LnvarZ = natural logarithm of the variance in z‐scores. [Colour figure can be viewed at wileyonlinelibrary.com]

Focusing on the individual sire EBVs, we found that the boars' EBV for SD of ADG_total was negatively associated with within‐litter SD at birth (r = −0.19; p = 0.01; Figure 5 and Table 2). Moreover, the boars' previously tested offspring's CV of ADG_pre‐testing and ADG_total were associated with both within‐litter SD at birth (r = −0.28; p = 0.001 and r = −0.18; p = 0.01) and within‐litter CV at birth in this study (r = −0.22; p = 0.01 and r = −0.17; p = 0.07).

FIGURE 5.

Associations between sire EBVs and within‐litter uniformity at birth. ADG, average daily gain; EBV, estimated breeding value. [Colour figure can be viewed at wileyonlinelibrary.com]

The within‐litter SD at the start of the finishing phase was associated with the sire EBVs for pre‐weaning mortality (r = −0.23; p = 0.01; Figure 6; Table 2), stillborn piglets (r = −0.26; p = 0.005), piglet vitality (r = −0.20; p = 0.009), ADG_pre‐testing (r = 0.28; p = 0.01), daily feed intake (r = 0.19; p = 0.08), and feed conversion ratio (r = −0.20; p = 0.02).

FIGURE 6.

Associations between sire EBVs and within‐litter uniformity at the start of the finishing phase. ADG, average daily gain; EBV, estimated breeding value. [Colour figure can be viewed at wileyonlinelibrary.com]

Furthermore, the individual lnvarZ was associated with the scaled sire EBVs ADG_pre‐testing (r = −0.17; p = 0.004) and SD of ADG_total (r = −0.15; p = 0.03; Figure 7; Table 2).

FIGURE 7.

Associations between sire EBVs and individual growth stability (lnvarZ individual). Abbreviations: LnvarZ = the natural logarithm of the variance in z‐scores; ADG, average daily gain; EBV, estimated breeding value. [Colour figure can be viewed at wileyonlinelibrary.com]

3.3. Associations Between Sire EBVs and Mortality

The ‘high sire survival‐uniformity index’ group had crossbred offspring with a higher pre‐weaning mortality rate (estimate = 0.91; p = 0.02; Figure 8 and Table 3). No other significant associations between the boar group and mortality were observed (Figures S7–S9). Furthermore, pre‐weaning mortality was associated with the scaled sire EBVs for ADG_pre‐testing (r = 0.22; p = 0.01; Figure 9) and SD of ADG_total (r = 0.23; p = 0.02). In our phenotypes, the average ADG_pre‐testing per litter was also lowly correlated to both pre‐weaning mortality (r = 0.10) and the percentage of pre‐weaning mortality (r = 0.16) within litters, although these associations were not significant.

FIGURE 8.

Association between boar groups based on ad‐hoc sire selection index for survival and uniformity and dead offspring in pigs. Association with percentage of dead offspring is given in Figure S7. Boxplots and correlation analysis for the individual estimated breeding values are given in Figures S8 and S9. [Colour figure can be viewed at wileyonlinelibrary.com]

TABLE 3.

Associations between boar group/sire EBVs of sire and offspring phenotypes (mortality).

Offspring phenotype	Variable of sire	Estimate	p	Correlation (r)	Odds ratio (OR)	95% CI
Pre‐weaning mortality	Boar group (high versus low)	0.91	0.02
	Sire EBV stillborn piglet	−1.94	0.08	−0.10
	Sire EBV ADG_pre‐testing	2.02	0.01	0.22
	Sire EBV SD_total	1.73	0.02	0.23
Pre‐weaning mortality odds	Boar group (high versus low)	−0.27	0.10		0.76	0.55 to 1.06
	Sire EBV stillborn piglets	−0.81	0.08		0.45	0.18 to 1.09
	Sire EBV ADG_pre‐testing	0.84	0.01		2.31	1.21 to 4.38
	Sire EBV SD ADG_total	0.60	0.04		1.83	1.04 to 3.21

Abbreviations: ADG, average daily gain; CI, confidence interval; EBV, estimated breeding value.

FIGURE 9.

Associations between sire EBVs and mortality rate of the litters. ADG, average daily gain; EBV, estimated breeding value. [Colour figure can be viewed at wileyonlinelibrary.com]

Similarly, the binomial analysis showed that the ‘low sire survival‐uniformity index’ group (estimate = −0.27; OR = 0.76; CI = 0.55 to 1.06; p = 0.10) tended to be linked to lower mortality odds of piglets. Higher sire EBVs for ADG_pre‐testing (estimate = 0.84; OR = 2.31; CI = 1.21 to 4.38; p = 0.01) and SD of ADG_total (estimate = 0.60; OR = 1.83; CI = 1.04 to 3.21; p = 0.04) were linked to higher odds of pre‐weaning mortality.

3.4. Relationships Between Sire EBVs and Production Traits

We investigated associations between production traits (BW, ADG, COG, lean meat percentage and carcass growth) and the ad‐hoc sire selection index for survival and uniformity, corrected for relevant covariates. This analysis indicated that pigs sired by the ‘high sire survival‐uniformity index’ group had lower ADG_testing (estimate = −29.40; p = 0.07; Figure 10 (per boar group), Figure S10 (per boar) and Table 4).

FIGURE 10.

Association between boar groups based on ad‐hoc sire selection index for survival and uniformity and growth performance in pigs. Boxplots and correlation analysis for the individual estimated breeding values are given in Figure S10. ADG, average daily gain. [Colour figure can be viewed at wileyonlinelibrary.com]

TABLE 4.

Associations between boar group/sire EBVs of sire and offspring phenotypes (production traits).

Offspring phenotype	Variable of sire	Estimate	SE	p	Correlation (r)
ADG_pre‐testing (g/d)	Sire EBV ADG_pre‐testing	49.10	12.49	< 0.001	0.17
	Sire EBV stillborn piglets	−54.28	16.59	0.001	−0.17
	Sire EBV SD of ADG_total	28.24	12.24	0.02	0.06
	Sire EBV pre‐weaning mortality	−35.13	10.90	< 0.001	−0.26
ADG_nursery (g/d)	Sire EBV stillborn piglets	−104.98	25.52	< 0.001	−0.21
	Sire EBV SD of ADG_total	51.22	19.03	0.007	0.11
	Sire EBV pre‐weaning mortality	−64.66	16.91	< 0.001	−0.18
ADG_testing (g/d)	Boar group (high versus low)	−29.40			0.07
	Sire EBV pre‐weaning mortality	−50.50	22.93	0.03	−0.23
Body weight 1 week post‐weaning (kg)	CV ADG_pre‐testing of tested offspring	−0.33	0.16	0.04	−0.13
	Sire EBV mortality finishing phase (kg)	1.17	0.61	0.06	0.15
Body weight at the start of the finishing phase (kg)	Sire EBV pre‐weaning mortality	−2.33	0.70	< 0.001	−0.12
	Sire EBV stillborn piglets	−4.04	1.06	< 0.001	−0.14
Body weight at 16 weeks of age (kg)	Sire EBV pre‐weaning mortality	−6.51	2.82	0.02	−0.25
	Sire EBV stillborn piglets	−10.91	4.33	0.01	−0.26
Carcass growth (g/d)	Sire EBV pre‐weaning mortality	−35.62	18.77	0.06	−0.24
COG	Sire EBV stillborn piglets	−2.74	1.48	0.06	−0.16
Lean meat percentage	Sire EBV piglet vitality	1.23	0.66	0.06	0.17
	Sire EBV pre‐weaning mortality	1.31	0.45	0.004	0.22
	Sire EBV stillborn piglets	1.86	0.72	0.009	0.22

Abbreviations: ADG, average daily gain; COG, center‐of‐gravity; EBV, estimated breeding value; SE, standard error.

The ADG_pre‐testing of the crossbred offspring was positively associated with the sire EBV for ADG_pre‐testing (r = 0.18; p < 0.001; Figure 11), but the ADG_total was not significantly linked to the sire EBV for ADG_total (r = 0.11; p = 0.25) and the percentage lean meat was not significantly associated with the sire EBV for carcass quality (r = 0.14; p = 0.18).

FIGURE 11.

Associations between sire EBVs and production traits. ADG, average daily gain; EBV, estimated breeding value. [Colour figure can be viewed at wileyonlinelibrary.com]

The ADG_nursery and ADG_pre‐testing showed associations with the scaled sire EBVs for stillborn piglets (r = −0.21; p < 0.001 and r = −0.17; p = 0.001) and SD of ADG_total (r = 0.11; p = 0.007 and r = 0.06; p = 0.02). The ADG_total, ADG_nursery, and ADG_pre‐testing were associated with the scaled sire EBV for pre‐weaning mortality (r = −0.23; p = 0.03; r = −0.18; p < 0.001; and r = −0.26; p < 0.001).

The body weight 1 week post‐weaning showed associations with the scaled CV of ADG_pre‐testing of the previously tested offspring of the sires (r = −0.13; p = 0.04) and the scaled sire EBV for mortality in the finishing phase in terms of kilograms (r = 0.15; p = 0.06). The body weight at the start of the finishing phase and at 16 weeks was associated with the scaled sire EBVs for pre‐weaning mortality (r = −0.12; estimate = −2.33; SE = 0.70; p < 0.001 and r = −0.25; estimate = −6.51; SE = 2.82; p = 0.02) and stillborn piglets (r = −0.14; estimate = −4.04; SE = 1.06; p < 0.001 and r = −0.26; estimate = −10.91; SE = 4.33; p = 0.01).

Lean meat percentage showed associations with the scaled sire EBVs for pre‐weaning mortality (r = 0.22; estimate = 1.31; SE = 0.45; p = 0.004) and stillborn piglets (r = 0.22; estimate = 1.86; SE = 0.72; p = 0.009).

4. Discussion

4.1. Relationships Between Sire EBVs and Within‐Litter Uniformity

We selected six Piétrain boars based on an ad‐hoc sire survival‐uniformity index and divided these into a high and a low group to generate crossbred offspring. The high sire survival‐uniformity index group had a lower within‐litter birth weight variation (Figures 2 and 3), but did not significantly affect body weight uniformity at later timepoints. The small effect can be attributed to several factors. First, the heritability of birth weight uniformity is low, limiting the accuracy and predictive value of EBVs (Damgaard et al. 2003; Zhang et al. 2016; Camargo et al. 2020). Second, we combined several sire EBVs related to uniformity and survival into one sire survival‐uniformity index. This method might reduce the predictive value for specific traits. Third, the contrast in sire index values between our low and high group was not very high (Table S1), as the worst performing sires were already removed from the breeding population and were no longer available for use in our experiment. As a result, the low group in this study was only just around the population average. A greater contrast in sire index values between low and high groups could possibly lead to more significant results. Fourth, the sire EBVs of the post‐weaning traits were estimated based on the phenotypes of a selection of offspring of the sires, possibly introducing selection bias.

Furthermore, uniformity at birth does not predict later uniformity, but post‐weaning measures are consistent over time. We found low correlations between within‐litter body weight uniformity at birth and later in life (r = 0.05 to 0.24; Figure 1), which aligns with previous research that birth weight uniformity has limited predictive value for post‐weaning within‐litter body weight uniformity (Hooyberghs, Goethals, Gorssen, Meyermans, et al. 2025). This suggests the existence of two distinct traits: within‐litter body weight uniformity at birth and within‐litter body weight uniformity post‐weaning. Therefore, selection programs aiming to improve uniformity throughout the production cycle should target both traits separately.

Our sire selection index also proved effective in increasing growth stability, with the piglets of the boars of the ‘high sire survival‐uniformity index’ group having a lower lnvarZ (estimate = −0.29; Figure 4 and Figure S6), indicating less deviations in growth. This suggests that selection for greater uniformity and survival might enhance growth stability and thus increase resilience. Similarly, a mice line selected on more uniform birth weights demonstrated increased survival and robustness (Formoso‐Rafferty et al. 2016b). In pigs, Putz et al. (2019) found that more stable feed intake was genetically correlated with lower mortality rates (r _g = 0.37). Furthermore, Casto‐Rebollo et al. (2025) reported that the offspring of resilient sires had an 80% chance of having a survival rate at least 2.5% higher.

Additionally, the negative association between the SD of birth weight and the sire EBV for SD of ADG_total (estimate = −0.08 kg per unit increase in the scale sire EBV; Figure 5) suggests that selection for more uniform growth can simultaneously decrease the within‐litter birth weight variation. Similarly, increased CV of ADG_pre‐testing and ADG_total of previously tested offspring of the selected boars is associated with higher SD of birth weight of the litters in this study. Moreover, Gorssen et al. (2024) reported a positive correlation between individual growth stability and uniformity of body weight at pen level (r = 0.14 to 0.29), supporting the idea that breeding for increased resilience and growth stability could also increase the uniformity of body weight.

The SD of body weight at the start of the finishing phase is linked to the sire EBVs for ADG_pre‐testing and daily feed intake, indicating that selection for increased ADG_pre‐testing and feed intake might unintentionally increase variation in body weight of pigs at the start of the finishing phase. Furthermore, selection for lower pre‐weaning mortality, less stillbirths, and increased piglet vitality might reduce the SD of body weight at the start of the finishing phase. Damgaard et al. (2003) found a moderately positive genetic correlation between SD of birth weight and pre‐weaning mortality (r _g = 0.25), and Zhang et al. (2016) between CV at birth and number of stillborn piglets (r _g = 0.26). Also, lower SD in body weight at the start of the finishing phase is found in the offspring of sires with a high sire EBV for feed conversion ratio. Similarly, Bruininx et al. (2001) showed that more uniform weaning groups had a more efficient feed conversion from weaning until 34 days later.

The paternal EBVs of ADG_pre‐testing, ADG_total and lower SD of ADG_total are associated with a lower lnvarZ, indicating enhanced growth stability of the offspring of the boars. Similarly, a previous study has shown that pigs with higher ADG during the first week post‐weaning exhibit a lower lnvarZ. On the other hand, selection aimed at lowering stillbirth rates is associated with higher lnvarZ, potentially affecting growth stability. A possible explanation is that reducing the number of stillborn piglets, which typically have birth weights comparable to their liveborn littermates (Langendijk and Plush 2019), may increase competition for nutrients not only in the womb but also during lactation, because the litter size increases. In other research, resilience traits such as deviations in growth, feed intake and feeding behaviour are low to moderately heritable (0.03 to 0.3) and genetically correlated with feed conversion (r _g = 0.39 to 0.49), tail biting (r _g = 0.22 to 0.30), lameness (r _g = 0.15 to 0.31) and mortality (r _g = 0.19 to 0.33) (Gorssen et al. 2023, 2024).

4.2. Relationships Between Sire EBVs and Mortality

Contrary to expectations, boars with higher survival–uniformity EBVs produced litters with greater pre‐weaning mortality compared to those sired by the other three boars (estimate = 0.91; Figure 6). This effect may, in part, be explained by the inverse relationship between ADG and pre‐weaning mortality. In our study, ADG_pre‐testing and lower SD of ADG_total increased pre‐weaning mortality rates, with estimates of 2.02 and 1.73, respectively. Gorssen et al. (2025) found an unfavourable correlation between sire EBVs for ADG and pre‐weaning mortality in Piétrain sires of the same population as we used (r = −0.16 for ADG_pre‐testing, r = −0.21 for ADG_total). Within our selection experiment, the average sire EBVs for ADG were higher for the boar group selected on high survival‐uniformity compared to the low group (0.28 vs. 0.05 for ADG_pre‐testing and 0.47 vs. 0.05 for ADG_total; Table S1). As this is an association, it is hard to make assumptions about causality. Breeding for lower pre‐weaning mortality might increase the survival rates of lightweight piglets and these pigs with lower birthweights tend to grow slower (Beaulieu et al. 2010; Douglas et al. 2014). Larger litter sizes are often linked to lower birth weights and more within‐litter variation, intensifying competition and reducing growth rates (Milligan, Dewey, and de Grau 2002; Milligan, Fraser, and Kramer 2002; Quiniou et al. 2002, Deen and Bilkei 2004, Quesnel et al. 2008, Zindove et al. 2014). Therefore, maternal traits such as number of functional teats, sow condition and appropriate litter size relative to maternal capacity are important to support uniform growth and reduce competition among piglets. Another hypothesis is that animals with a genetic predisposition for a higher ADG are more vulnerable in early life.

4.3. Relationships Between Sire EBVs and Production Traits

Selection based on sire survival–uniformity index and EBVs influenced several production traits, including growth rate and carcass composition. Piglets originating from the ‘high sire survival‐uniformity index’ group showed lower ADG_testing, with an estimate of −29.40 g/day. No significant association was observed between the sire EBV for ADG_total and its corresponding phenotype (ADG from birth to slaughter), although the correlation was positive (r = 0.12). These results were unexpected as ADG generally has a moderate to high heritability, with an average of 29% over multiple studies (Rothschild and Ruvinsky 2011). In the finishing phase, data were collected from only 358 of the 1095 pigs, resulting in a relatively small dataset with possible selection bias. Also, the sire EBV for ADG_total is routinely calculated based on a subset of the offspring of each boar, likely excluding the lightweight piglets. In contrast, our data also included the lightweight pigs which were followed up until slaughter.

The sire EBVs for SD of ADG_total (estimate = 28.24 to 51.22 g/day per unit increase in sire EBV) are associated with increased ADG_pre‐testing and ADG_nursery. This suggests that selecting for animals with higher growth stability post‐weaning may lead to higher ADG. Animals with more stable growth patterns tend to have lower mortality rates (r _g = 0.19) (Putz et al. 2019; Gorssen et al. 2024). However, in previous research, a negative correlation was observed between the ADG the first week post‐weaning and growth stability (r = −0.18), suggesting that pigs that had a higher ADG during the first week post‐weaning had less deviations in growth later in life.

The sire EBVs for stillbirths (estimate = +1.86%), pre‐weaning mortality (estimate = +1.31%), and vitality at birth (estimate = +1.23%) were associated with increased lean meat percentage. This suggests that improving early‐life survival and vitality may impact carcass composition. However, as mentioned above, the sire EBVs for pre‐weaning mortality and vitality were also linked to reduced ADG, and ADG and lean meat tissue growth rate are known to have a negative genetic correlation (r _g = −0.09) (Rothschild and Ruvinsky 2011).

4.4. Advice on the Use of a Boar Index to Improve Piglet Survival and Uniformity

Our current sire survival‐uniformity index was effective for lowering within‐litter birth weight variation and increasing post‐weaning growth stability. However, the impact on within‐litter body weight uniformity further in life was limited. Because birth and post‐weaning body weight uniformity seem to be two distinct uncorrelated uniformity traits, effective selection programs should address both traits to improve uniformity throughout the production cycle. Our study found no clear associations between within‐litter body weight uniformity or survival and the sire EBV for lightweight piglets at the boar level. As the percentage of lightweight piglets in a litter has been found to be moderately correlated to birth weight SD and CV (r = 0.38 to 0.70) (Hooyberghs, Goethals, Gorssen, Chapard, et al. 2025), the sire EBV can be further improved. Furthermore, we observed negative effects of the sire index on survival, leading us to conclude that the sire index, in its current form, is suboptimal to be used for selection decisions at the commercial farm level. We would like to stress that this is different from long‐term genetic selection using such a sire index by breeding organizations. Our study only focused on a single generation of terminal boars with variability in sire EBVs for uniformity and survival, and the practical use of selecting these boars to improve uniformity and survival on commercial pig farms. Long‐term selection could yield more significant results between the current and future generations of terminal boars. In a selection experiment for mice to increase birth weight uniformity, for example, significant differences between lines were only observed after 2 generations of selection (Formoso‐Rafferty et al. 2016a).

4.5. Limitations of the Study

Mainly small effect sizes were found, which might be due to a few limitations during this study. First, the number of boars included in this study was small (n = 6), which limits the contrast in EBVs and reduces statistical power. Second, as different sow lines were housed at the two farms, the sow line and farm effects are confounded, making it impossible to disentangle their respective contributions to litter and piglet performance. Third, some litters were removed due to limited litter size (N = 2) or unexplained pre‐weaning mortality (N = 1); this might have introduced some bias; however, still a wide range of litter sizes were kept (from 9 to 22 liveborn piglets). Also, removal of piglets reduced sample size for the traits during the finishing phase and may have introduced selection bias.

5. Conclusion

Our aim to improve survival and uniformity through sire EBVs was partially achieved, as the observed only effects in our study were limited. Nonetheless, significant associations were found between sire EBVs and specific traits related to survival and uniformity, supporting their potential for selection. At the same time, the observed increase in pre‐weaning mortality in offspring of sires with a higher sire survival–uniformity index highlights important limitations associated with the current sire index. Pigs' within‐litter body weight uniformity can be divided in two distinct traits: pre‐weaning uniformity, which is crucial for piglet survival, and post‐weaning uniformity, which stabilizes from weaning to the finishing phase. These traits should be targeted separately to improve overall uniformity. From a practical perspective, the current survival–uniformity index appears unsuitable for short‐term, farm‐level decision‐making, given the small effect sizes. Future research should focus on evaluating larger contrasts between sires, refining selection indices to better disentangle growth and survival, and assessing responses across multiple generations to determine the long‐term effectiveness and robustness of similar selection for uniformity and survival.

Author Contributions

K.H. performed the analysis of the data and wrote the manuscript. The methodology of the data analysis was designed by K.H., W.G., R.M., L.C., J.D.K., I.P. and S.J. Moreover, N.B., S.J. and S.M. acquired the funding and provided the scientific supervision. K.H., S.G., W.G., L.C., R.M., J.D.K., I.P., S.M., S.J. and N.B. discussed the results. All authors read and approved the manuscript.

Funding

We acknowledge the Agency for Innovation and Entrepreneurship in Flanders (VLAIO) and the co‐financers from the industry for funding the UNIPIG project (HBC.2019.2866). WG was funded through an FR PhD fellowship (1104320N) of the Research Foundation Flanders (FWO). Moreover, LC was funded by a KU Leuven C2 project (C24/18/036). RM was partially funded by Internal Funds KU Leuven (PDMT2/23/035). R.M., J.D.K. and I.P. were funded by the Genobreed project (S020024N) of the Research Foundation Flanders (FWO). The funding bodies played no role in the design of the study, collection analysis, interpretation of data and in writing the manuscript.

Ethics Statement

National (Belgian) legislation for the use of animals was respected. No procedures requiring approval from the local ethics committee according to Belgian and EU legislation (Council Directive 86/609/EEG) were performed.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Overview of data per sire. The EBVs were rescaled to a range between −1 and + 1, with −1 representing an unfavourable outcome (e.g., low vitality and high mortality) and + 1 indicating a favourable outcome (e.g., high vitality and low mortality). Next to EBVs, two extra uniformity traits were collected per boar based on the phenotypes of their previously tested offspring: the CV of ADG_pre‐testing and CV of ADG_toal of previous offspring. EBV for mortality in the finishing barn in terms of kilograms, is based on the body weight of the died tested offspring of the sires and the EBV for mortality in the finishing barn in terms of percentage of the percentage of tested offspring that died in the finishing barn per sire.

Figure S1: Principal component analysis (PCA) showing clustering of samples from sires, piglets and sows from two farms. Each point represents an individual, coloured according to group: piglets farm 1, piglets farm 2, sires, sows farm 1 and sows farm 2. The percentage values on the axes indicate the proportion of variance explained by each principal component.

Figure S2: Pairwise Pearson correlation plot between number of stillborn piglets, pre‐weaning mortality and mortality in the nursery. p < 0.001 ***.

Figure S3: Pairwise Pearson correlation plot between percentage stillborn piglets, pre‐weaning mortality and mortality in the nursery. p < 0.001.

Figure S4: Association between ad‐hoc selection index for survival and uniformity and within‐litter body weight uniformity (SD) in pigs.

Figure S5: Association between ad‐hoc selection index for survival and uniformity and within‐litter body weight uniformity (CV) in pigs.

Figure S6: Association between ad‐hoc selection index for survival and uniformity and individual uniformity of growth (lnvarZ) in pigs. lnvarZ = natural logarithm of the variance in z‐scores.

Figure S7: Association between boar groups based on ad‐hoc selection index for survival and percentage of dead offspring.

Figure S8: Association between ad‐hoc selection index for survival and uniformity and number of dead offspring pigs.

Figure S9: Association between ad‐hoc selection index for survival and uniformity and percentage of dead offspring in pigs.

Figure S10: Association between ad‐hoc selection index for survival and uniformity and growth performance in pigs. Abbreviation: ADG = average daily gain.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.