1. Introduction
In recent decades, selection for improved litter size in pigs has led to a significant increase in the total number of piglets born and the number of piglets born alive, mostly from hyperprolific sow lines [
1]. However, in addition to positive effects of increased litter size, the development of highly prolific hybrid lines resulted in changes that required the adaptation of existing approaches in farm management, with a focus on nursing piglets of low viability, usually those with low birth weights [
2]. The variability of birth weight and the lower survival rate of piglets with low birth weights are now recognised as one of the key problems in the management of hyperprolific sows [
3,
4]. In addition, common beliefs regarding the high productivity of hyperprolific sows can also be questioned in terms of the number of teats and longer farrowing durations [
5].
Piglet growth is influenced by litter size while the piglets are still in the uterus: larger litters result in intrauterine crowding and embryos that are implanted first may physically prevent the development of additional embryos [
5] resulting in a higher number of light piglets. Impaired intrauterine growth is strongly correlated with litter size [
6]. Since growth rate strongly determines the economic efficiency of pig production, it is important to assess numerous factors that influence this trait. Since growth is a longitudinal trait that can be measured several times during the lifespan, a traditional approach with single trait linear models may not be sufficient to describe the nature of the trait. The reason for this is that the variability between measurements on the same experimental unit might change through time and the measurements may therefore be characterised by heterogeneous variances [
7]. Because of their properties, random regression models can be used to analyse such traits. The main advantages of random regression models over conventional models are the improved accuracy of estimation, the avoidance of adjustment of phenotypic data, a smaller number of parameters to describe longitudinal measurements, smoother (co)variance estimates, and the possibility to estimate covariance components and to predict breeding values at any point along the trajectory [
8,
9,
10].
Random regression models have typically been applied to pigs in estimating breeding values for growth [
11], feed intake [
12], and fertility traits [
13]; however, the application of a random regression coefficient might also be useful in the estimation of non-genetic factors [
14] for traits with a longitudinal structure.
The aim of the study was to examine the possibility of using a random regression coefficient model to determine the main effects on body weight of piglets from hyperprolific sows during the growth period from birth to 85 days of age.
4. Discussion
The covariance structure between measurements justified the application of a random regression coefficient model on the collected data. One of the most important features of random regression coefficient models is the ability to use it as an unstructured model and take different variances into account for each period and different covariances between periods [
7]. This might be particularly important when considering the litter size of highly prolific sows since there is a big variability in the number of liveborn piglets between parities. Moreover, the large variability in their birth weights collectively makes the growth traits of the piglets one of the most important factors in efficient fattening. Thus, the use of a method that can describe the individual growth of piglets is a method of choice for the analysis of the growth data. In large litters, within-litter variation in birth weight might significantly affect future growth. Thus, the application of a random regression coefficient model allows for the analysis of the growth of every individual, which can be useful for the piglets with variable birth weights. Due to the heterogeneous covariance structure between measurements, classical approaches, such as classical fixed models and random linear models, are not sufficient to describe the growth pattern of piglets since the data structure does not fulfil the assumptions of the homogeneous variance and balanced data [
20]. The literature describing the use of models that account for heterogeneous variance between measurements in a random regression coefficient model in the growth of piglets, especially for the analysis of non-genetic factors, is scarce.
The application of the random regression coefficient model confirmed a slower growth in piglets with lower birth weights. The ability of a random regression coefficient model to describe individual growth patterns of piglets can be helpful for breeders in forming groups and adopting feeding and regimes and in planning staying capacities in the farm. The results correspond to earlier studies by Berard et al. [
21], who found the significant effect of birth weight groups on myogenesis in piglets. Gondred et al. [
22] found a significant association between low birth weight and reduced average daily gain during suckling and the postweaning period. A similar conclusion was brought by Vaclakova et al. [
23], who also found an association between low birth weights and impaired growth rates during postweaning and the fattening period. It is commonly recognised that low birth weight in piglets correlates with decreased survival and lower postnatal growth rates. In the majority of low birth weight piglets, low numbers of muscle fibres differentiate during prenatal myogenesis, for genetic or maternal reasons, and those low birth weight piglets with reduced fibre numbers are unable to exhibit postnatal catch-up growth [
24]. The strong effect of birth weight on the growth of piglets is relevant to discussions regarding upper limits in the selection for litter sizes and its final effect in managing sow reproduction. Ocepek et al. [
25] discussed the sustainability of further artificial selection for litter size in pigs, suggesting that further genetic improvement for litter size might be unsustainable because increments in the number of piglets weaned have increasing costs, such as sibling competition, mortality, and uneven growth, which compromises piglet welfare and fitness.
Riddersholm et al. [
2] found that the main critical risk factor for low birth weights observed was the litter size. According to Ocepek et al. [
25], piglets from large litters had significantly lower and more variable body weights at weaning. Such a variation might lead to non-uniform piglets in the nursing and fattening period. The negative effects of high litter sizes could be connected to a lower uterine blood flow per foetus when litter size increases [
26] and a lack of space in the uterus as a consequence of overcrowded uterus horns, where embryos that were implanted first also prevent the development of additional embryos [
5]. Moreover, larger piglets in the litter will have more access to teats or milk replacer during the suckling period, resulting with better growth than in smaller piglets in the litter [
27]. On the other hand, Božičković et al. [
28] found no statistically significant effect of the litter size on the final weight of piglets at the end of the fattening period, although the birth weight of piglets was higher in animals from small litters.
Milligan et al. [
3] found a significant effect of parity on weights at weaning and they reported a higher birth weight of piglets farrowed in the first parity compared to subsequent parities. On the contrary, Carney et al. [
29] found that growth performance in the nursery may be affected by dam parity, where the results of the study suggested that the progeny born in higher parities have increased body weight and growth performance during the nursery phase of production compared to piglets born in the first parity. The significant effect of parity on the birth weight of piglets and consequently on the growth of piglets was reported by Amatucci et al. [
30], who found that a litter's average daily gain and final weights were higher in multiparous sows than in gilts, probably due to the differences in the colostrum composition within different parities. No effect of parity on the weaning weight was found in the study of Akdag et al. [
31]. These results suggest that the piglets in the analysis were not affected by differences in sex when they were in the same environmental conditions such as microclimate and rearing density. Additionally, the reduction of sex difference on the growth rate is considered in the new paternal line because results by Cisneros et al. [
32] showed that the sex difference could vary with the genotype. This is in accordance with studies by Škorjanc et al. [
33], Bocian et al. [
34], and Lee et al. [
35]. However, straightforward comparison between different studies might be affected by different experimental conditions and methodologies applied throughout the experiments. According to Kielly et al. [
36] castration at 3 days of age can temporarily reduce weight gain. Since castration occurred at the third day of life, the impact of early castration on growth has been accounted for.