Within-litter variation of birth weight in hyperprolific Czech Large White sows and its relation to litter size traits, stillborn piglets and losses until weaning

Available online at www.sciencedirect.com

Livestock Science 115 (2008) 195 – 205 www.elsevier.com/locate/livsci

Within-litter variation of birth weight in hyperprolific Czech Large White sows and its relation to litter size traits, stillborn piglets and losses until weaning J. Wolf a,⁎, E. Žáková a , E. Groeneveld b b

a Institute of Animal Science, P.O. Box 1, CZ 10401 Prague Uhříněves, Czech Republic Department of Breeding and Genetic Resources, Institute for Animal Science Mariensee, Federal Agricultural Research Center (FAL), DE 31535 Neustadt, Germany

Received 8 August 2006; received in revised form 28 June 2007; accepted 8 July 2007

Abstract Data from about 2900 litters (approximately 40,000 piglets) originating from 1063 Czech Large White hyperprolific sows were analyzed. The phenotypic and genetic relations between litter size traits, piglet mortality during farrowing and from birth to weaning and several statistics referring to the distribution of the birth weight within litter were analyzed. All genetic parameters were estimated from multi-trait animal models including the following factors: mating type (natural service or insemination), parity, linear and quadratic regression on age at first farrowing (1st litter) or farrowing interval (2nd and subsequent litters), herd-year-season effect and additivegenetic effect of the sow. The phenotypic correlations of the mean birth weight with the total number of piglets born and piglets born alive were −0.30. Traits describing the variability of the birth weight within litter (range, variance, standard deviation, coefficient of variation) were mostly positively correlated with litter size. A statistically significant phenotypic correlation (−0.09 to −0.15) between mean birth weight and losses at birth and from birth to weaning was found. The heritability for the number of piglets born, piglets born alive and piglets weaned was around 0.15. The number of stillborn piglets had only a very low heritability less than 0.05, whereas the heritability for losses from birth to weaning was 0.13. The heritabilities of the mean, minimal and maximal birth weight were 0.16, 0.10 and 0.10, respectively. The heritability for all statistics and measures referring to the variability of the birth weight within litter was very low and did never exceed the value of 0.05. An increase in litter size was shown to be genetically connected with a decrease in the mean piglet birth weight and an increase in the within-litter variability of birth weight. Selection on litter size should be accompanied by selection on mortality traits and/or birth-weight traits. Losses from birth to weaning and the minimal birth weight in the litter were proposed as potential traits for a selection against piglet mortality. © 2007 Elsevier B.V. All rights reserved. Keywords: Pig; Large White; Hyperprolific line; Birth weight; Litter traits; Piglet mortality

1. Introduction

⁎ Corresponding author. Tel.: +420 267009573; fax: +420 267710779. E-mail addresses: [email protected], [email protected] (J. Wolf). 1871-1413/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.livsci.2007.07.009

Birth weight is considered to be one of the most important factors influencing pig survival (Leenhouwers et al., 2001). The low-birth-weight piglet is particularly at risk for preweaning morbidity and mortality. It is physiologically compromised in terms of energy stores

196

J. Wolf et al. / Livestock Science 115 (2008) 195–205

and susceptibility to cold and is at a disadvantage in competing with larger littermates at the udder (Lay et al., 2002). Having in mind the whole litter of the sow, the distribution of the birth weight within the litter (mean birth weight and variability within the litter) is of importance for the overall productivity of the sow. Selection for the sow's ability to give birth to a higher number of piglets has led to an increased within-litter variation in piglet birth weight (Tribout et al., 2003). Large litters result in a longer farrowing duration and thus may be critical to survival for piglets born toward the end of farrowing. Litter size can also influence piglet survival after birth as piglet losses tend to be greater in larger litters which may be attributed to the within-litter variation in piglet body weight (Marchant et al., 2000; Lay et al., 2002). Genetic analyses of the birth weight and its variation within litter have been presented in several papers (Högberg and Rydhmer, 2000; Hermesch et al., 2001; Damgaard et al., 2003; Huby et al., 2003; Tribout et al., 2003; Täubert and Henne, 2003). In these studies, the standard deviation and/or the coefficient of variation of the birth weight were investigated in data sets from 15,000 to 30,000 piglets. Using a large data set with nearly 40,000 weighed piglets, the aim of the present study is to investigate the question to what extend measures of the distribution of the birth weight within litter are heritable. Besides the standard deviation and coefficient of variation which have been currently used in the literature, a number of further statistics will be considered. Next, the phenotypic and genetic relations between the statistics for birth weight and litter size traits including farrowing losses and losses from birth to weaning will be analyzed. Finally, the possible use of the statistics for birth weight in selection programmes will be discussed. 2. Materials and methods 2.1. Animals and structure of data A hyperprolific line has been formed as a subpopulation within the Czech Large White breed starting apTable 1 Numbers of piglets, litters, sows and boars Characteristics

Number or value

Total number of piglets born alive Number of litters Number of sows Number of boars Mean number of litters/sow Mean number of litters/boar

40,102 2933 1063 393 2.76 7.46

Table 2 Relative frequencies (%) of individual parities Parity

Relative frequency

1 2 3 4 5 or 6 ≥7

22.3 21.2 18.6 14.9 15.7 7.3

proximately in the year 2000. To be included into a hyperprolific line, a sow must have a breeding value for litter size (number of piglets born alive in the second and subsequent litters) among the top 15%. The top 15% is calculated from the breeding values of all sows with a minimal age of two years. The candidate sow may be on her first to third litter and must have an average of 12 or more live-born piglets per litter. The number of functional nipples must be at least 7 on both sides and the sow must be MHS negative. For testing the MHS status, the ryanodine receptor gene (RYR1) was used and the test was carried out according to Brenig and Brem (1992). Data of about 2900 litters originating from 1063 Czech Large White sows from the hyperprolific line were analyzed. The farrowing years were in the range from 2000 to 2005. The sows were distributed over 24 farms. Table 1 gives a survey on the number of litters, number of sows, number of boars and the mean number of litters per sow and boar. The relative frequencies of individual parities are listed in Table 2. Data were available on age at first farrowing, the farrowing interval, total number of piglets born, number of piglets born alive and number of piglets weaned. The number of piglets born and born alive were recorded immediately after birth (not later than 24 h after birth). The number of piglets born included piglets born alive and stillborn piglets, but no mummified piglets. The number of weaned piglets was stated between 18 and 24 days after birth. Crossfostering was not used. From these data the number of stillborn piglets and the losses from birth to weaning, both as absolute (number of piglets) and relative (per cent of the total number of piglets born or per cent of the number of piglets born alive, respectively) numbers were calculated. Furthermore, the individual birth weight for each live-born piglet was given. 2.2. Parameters describing the distribution of individual birth weight within litter The distribution of the birth weight within the litter was described by several quantities. The arithmetic and

J. Wolf et al. / Livestock Science 115 (2008) 195–205

harmonic mean represented the average birth weight. The harmonic mean x˜ was defined as n x˜ ¼ P : n i¼1

1 xi

Variance, standard deviation and range were measures of the variability of the birth weight. The coefficient of variation related the variability to the mean. Minimal and maximal birth weight showed the extremes in the distribution. Skewness described the deviation of the distribution of birth weight from the (symmetric) normal distribution.

197

the herd-year-season class, the season was prolonged as long as the number of 10 was reached. The aggregation of sows from different herds into one herd-year-season class was not allowed. The pedigree was traced back until the year 1985. In matrix notation, Eq. (1) can be written as y ¼ Xb þ Zh h þ Zu u þ e where the vectors y, h, u and e are assumed to have the following joint distribution: 02 31 2 3 3 2 y Xb V R Zh Gh Zu GA B6 C 6e7 7 6 R 0 0 7 7C 6 7fN B 6 0 7; 6 R @ 4 0 5 4 Gh Z hV 0 4h5 0 0 5A GA Z uV 0 0 0 u 0

2.3. Phenotypic and genetic analyses The basic statistic analyses of the litter size traits and of the statistical parameters referring to the distribution of birth weight within litter were carried out using SAS® 8.2 software. All genetic parameters were estimated from multitrait animal models with the following structure: ytijkl ¼ mti þ ptj þ b1tj ðxjl  ¯ xj: Þ þ b2tj ðxjl  ¯ xj: Þ2 þ hysk þ atl þ etijkl

ð1Þ

where ytijkl is the value of trait t in parity j of sow l which was measured in the herd-year-season class k, mti, is the effect of mating type i on trait t, ptj is the effect of parity j on trait t, xjl is a covariable (see below) observed on sow l in parity j, b1tj(xjl − ¯x j.) + b2tj(xjl − ¯x j.)2 is the quadratic regression within parity j for trait t on the covariable xjl (b1tj,b2tj are regression coefficients specific for parity j and trait t), ¯x j is the mean of the covariable in parity j, hysk is the herd-year-season effect, atl is the additivegenetic effect of trait t in sow l and etijkl is the residual effect. In the first parity, age at farrowing was used as covariable whereas the farrowing interval was used in the subsequent parities. The herd-year-season effect, the residual effect and the additive-genetic effect were treated as random; all remaining effects were considered as fixed. There were two classes for the mating type: natural mating and artificial insemination. The parity classes were formed as shown in Table 2. A flexible allocation of records to herd-year-season classes was applied. For this purpose, a FORTRAN program was written. Herdyear-season classes were preferably formed according to natural seasons and had normally a length of three months: March to May, June to August, September to November and December to February of the following year. If the number of observations was less than 10 in

with V = R + ZhGhZ′h + ZuGAZ′u, R = ⊕ jn= 1R0ij, Gh = Ih ⊗ Gh0, and GA = A ⊗ Gu0 where y is the vector of observations, X is the incidence matrix for the fixed effects, Zh and Zu are the incidence matrices of the random herdyear-season and additive-genetic effects, β, h, u and e are the vectors of fixed, herd-year-season, additive-genetic and residual effects, V is the covariance matrix for y, R0ij, Gh0 and Gu0 are the residual, the herd-year-season and the additive-genetic covariance matrices of traits, respectively, Ih is the identity matrix of the order of the number of herdyear-season effects, A is the additive-genetic relationship matrix between animals, ⊗ is the Kronecker product and ⊕ is the direct sum. The covariance matrices and further genetic parameters were estimated by restricted maximum likelihood (REML) and optimization by a quasi Newton algorithm with analytical gradients (Neumaier and Groeneveld, 1998) as implemented in VCE 5 program (Kovač et al., 2002). Approximate standard errors of the covariance components representing the lower bound of the real standard errors were calculated from the Hessian matrix. As there were many functional dependencies between traits (i.e. farrowing loss was calculated as total number of piglets born minus piglets born alive), it was impossible to calculate a complete covariance matrix in one run. Therefore, four series of VCE runs were designed in such a way that each omitted functional dependencies between traits and jointly covered all combinations of traits of interest. If there was more than one estimate of a genetic parameter, the median of the estimates of the parameter and the median of the estimated standard errors were presented in the Results section. The number of simultaneously estimated covariance components was between 18 and 63 in the individual runs. An estimate was considered to be significantly different from zero when it was greater than twice its approximate standard error.

198

J. Wolf et al. / Livestock Science 115 (2008) 195–205 Table 4 Statistical parameters of the distribution of birth weight within litter (number of litters n = 2933)

3. Results 3.1. Phenotypic analyses Some basic statistics of litter traits are summarized in Table 3. Table 4 lists several statistical parameters and measures which refer to the distribution of the individual birth weight within litter and which were defined in Section 2.2. The mean birth weight was approximately 1500 g and the range, that means the difference between the maximal and the minimal birth weight, was approximately 550 g on average. Considerable differences between litters in the distribution of the individual birth weights within litter were reflected e.g. by the coefficient of variation which ranged from 2.1% to 38%. The distribution of the birth weight within the individual litters was not symmetric, both negative and positive values for skewness were observed. However on average, the skewness differed not significantly from zero. The estimated phenotypic correlations between traits related to individual birth weight and litter traits (total number of piglets born, number of piglets born alive and number of piglets weaned) were low to moderate (see Table 5). They were mostly negative for minimal and maximal birth weight and for the arithmetic and harmonic mean of the birth weight as well as for skewness. The correlation of the (arithmetic or harmonic) mean birth weight with the total number of piglets born and piglets born alive was around −0.30 whereas the correlation of the same trait with the number of piglets weaned was −0.20. Traits describing the variability of the birth weight within litter (range, variance, standard deviation, coefficient of variation) were mostly positively correlated Table 3 Basic statistics of litter traits Trait

na

Age at 1st farrowing (d) Farrowing interval (d) Total number of piglets born Number of piglets born alive Number of piglets weaned Litter weight at birth (kg) Number of stillborn piglets Losses from birth to weaning (piglets) Stillborn piglets as percentage of the total number of piglets born Losses from birth to weaning as percentage of the number of piglets born alive

655 371 363 2278 164 155 2933 13.7 14.0 2933 13.1 13.0 2912 11.3 11.0 2933 19.3 19.3 2933 0.6 0.0 2912 1.8 2.0

a

Number of litters.

Mean Median Standard deviation 38.4 26.1 2.32 2.14 1.83 3.62 0.92 1.60

2933

3.7

0.0

6.03

2912

13.0

12.5

10.90

Statistical parameter

Min.

Mean

Median Max.

0.50 1.18 1.18 2.40 Minimal birth weight in litter (kg) Maximal birth weight 1.05 1.74 1.70 2.90 in litter (kg) Range (maximum– 0.10 0.56 0.50 1.80 minimum, kg) Arithmetic mean (kg) 0.86 1.48 1.47 2.79 Harmonic mean (kg) 0.77 1.45 1.45 2.78 Variance (kg2) 0.001 0.037 0.023 0.292 Standard deviation (kg) 0.034 0.171 0.152 0.540 Coefficient of 2.1 11.8 10.7 38.8 variation (%) Skewness − 3.37 −0.20 − 0.16 2.44

Standard deviation 0.272 0.268 0.299 0.224 0.227 0.039 0.090 6.29 0.613

with the three litter size traits. Though these correlations were significant, they did never exceed the value of 0.20. The estimated phenotypic correlation between the mean individual birth weight and the litter weight at birth was relatively high (0.50). Whereas the minimal birth weight in the litter had a correlation of 0.30 with the litter weight at birth, the correlation of the maximal birth weight with the litter weight was greater than 0.50. Among the variability measures within the litter, the range showed the highest correlation with the litter weight. The coefficient of variation was not correlated with the litter weight at birth. The estimated phenotypic correlations between the statistical parameters related to individual birth weight and the losses during farrowing and from birth to weaning (both as absolute values in piglets and as percentages) are summarized in Table 6. A statistically significant relation between the mean birth weight and losses at birth and from birth to weaning was found, though the relation was relatively weak (correlations in the range from −0.09 to −0.15). That means, on average the losses increased with lower mean birth weight. The correlations between the variability measures and losses were near zero. 3.2. Heritabilities and proportion of variance caused by the herd-year-season effect for the investigated traits The genetic parameters of litter size traits were estimated from 12 VCE runs (16 runs for litter weight at birth) with different combinations of traits including three to six traits in each run and producing very similar results. The numbers presented in Table 7 are the medians of the estimates and of their approximate standard errors. The heritability for the three litter size traits total number of

J. Wolf et al. / Livestock Science 115 (2008) 195–205

199

Table 5 Phenotypic correlations between litter size traits and litter weight at birth on the one hand and statistical parameters related to individual birth weight within litter on the other hand (n = 2912 for correlations with number of piglets weaned, n = 2933 for all other correlations, n is the number of litters) Statistical parameter

Total number of piglets born

Number of piglets born alive

Number of piglets weaned

Litter weight at birth

Minimal birth weight in litter (kg) Maximal birth weight in litter (kg) Range (maximum–minimum, kg) Arithmetic mean (kg) Harmonic mean (kg) Variance (kg2) Standard deviation (kg) Coefficient of variation (%) Skewness

−0.37+ −0.17+ 0.18+ −0.32+ −0.33+ 0.11+ 0.12+ 0.21+ −0.06+

− 0.35+ − 0.14+ 0.20+ − 0.28+ − 0.30+ 0.11+ 0.13+ 0.21+ − 0.06+

− 0.30+ − 0.09+ 0.20+ − 0.22+ − 0.24+ 0.12+ 0.14+ 0.20+ − 0.06+

0.30+ 0.53+ 0.21+ 0.54+ 0.52+ 0.17+ 0.17+ 0.02 − 0.11+

+

Different from zero for P = 0.05.

piglets born, piglets born alive and number of piglets weaned was around 0.15. The heritability of litter weight at birth was in the same order of magnitude. The number of stillborn piglets as well as the percentage of stillborn piglets had only a very low heritability less than 0.05. The losses from birth to weaning expressed as number of piglets or as percentage of the number of piglets born alive had a similar heritability as the litter size traits. The effect of the herd-year-season effect expressed as proportion of the total variance was highest for litter weight at birth (about 25%). In all other traits listed in Table 7, the proportion of variance caused by the herd-year-season effect was less than 10%. Herd-year-season had only a very small effect especially on the total number of piglets born and the number of piglets born alive. Also among all factors acting on the losses from birth to weaning, herdyear-season caused only 5% of the total variability. The genetic parameters of different statistics and measures connected with the distribution of the birth weight within litter were estimated from 3 VCE runs mostly. The medians of the estimates and of their ap-

proximate standard errors are presented in Table 8. The arithmetic and the harmonic mean showed the highest heritability (0.16). The minimal and the maximal birth weight had a heritability of approximately 0.10. The heritability for all statistics and measures referring to the variability of the birth weight within litter was very low and did never exceed the value of 0.05. The factor herd-year-season had a large impact on all traits listed in Table 8. Whereas the proportion of variance caused by this factor was around 30% for the arithmetic and the harmonic mean, this proportion was between 40 and more than 50% for the variability measures as variance, standard deviation, range and coefficient of variation. There was only a minor impact on the skewness. 3.3. Genetic correlations between birth weight and litter size traits including stillborn piglets and losses from birth to weaning The genetic correlation between the litter size traits and the mean, minimal and maximal birth weight

Table 6 Phenotypic correlations between stillborn piglets and losses from birth to weaning on the one hand and statistical parameters related to individual birth weight within litter on the other hand (n = 2912 for correlations with losses from birth to weaning, n = 2933 for correlations with stillborn piglets, n is the number of litters) Statistical parameter

Number of stillborn piglets

Losses from birth to weaning (piglets)

Percentage of stillborn piglets

Losses from birth to weaning (%)

Minimal birth weight in litter (kg) Maximal birth weight in litter (kg) Range (maximum–minimum, kg) Arithmetic mean (kg) Harmonic mean (kg) Variance (kg2) Standard deviation (kg) Coefficient of variation (%) Skewness

− 0.12+ − 0.12+ 0.00 − 0.14+ − 0.15+ 0.01 0.00 0.04+ 0.00

−0.13+ −0.08+ 0.04+ −0.13+ −0.13+ 0.01 0.01 0.05+ −0.02

− 0.08+ − 0.10+ − 0.01 − 0.11+ − 0.11+ 0.00 − 0.01 0.02 0.01

− 0.08+ − 0.07+ 0.01 − 0.09+ − 0.09+ 0.00 − 0.01 0.02 − 0.01

+

Different from zero for P = 0.05.

200

J. Wolf et al. / Livestock Science 115 (2008) 195–205

Table 7 Heritability and proportion of variance caused by herd-year-season for litter size traits including losses and litter weight at birth Trait

n a Heritability

Proportion of variance caused by herd-year-season

Total number of piglets born Number of piglets born alive Number of piglets weaned Litter weight at birth (kg) Number of stillborn piglets Losses from birth to weaning (piglets) Stillborn piglets as percentage of the total number of piglets born Losses from birth to weaning as percentage of the number of piglets born alive

12 12 12 16 12 12

0.02 ± 0.006 0.03 ± 0.008 0.07 ± 0.011 0.24 ± 0.020 0.07 ± 0.013 0.04 ± 0.012

0.13 ± 0.018 0.14 ± 0.020 0.16 ± 0.018 0.13 ± 0.018 0.04 ± 0.012 0.13 ± 0.020

12 0.04 ± 0.013 0.07 ± 0.013

12 0.13 ± 0.019 0.06 ± 0.013

Both the estimates and their standard errors are calculated as medians from the given number of estimates. a Number of estimates.

were approximately in the range from −0.30 to −0.50 (Table 9). Thus a higher litter size was connected with lower birth weights on average. The range of the birth weight within litter similarly as the coefficient of variation had a relatively high positive genetic correlation (around 0.40) with the total number of piglets born and the number of piglets born alive whereas the genetic correlation of these two statistics with the number of piglets weaned was zero or close to zero. The genetic correlations of the variance and the standard deviation of the birth weight within litter with the three litter size traits showed a similar behaviour, though the estimates of the correlations with the total number of piglets born and the number of piglets born alive were lower (approximately 0.20). Litter weight at birth showed high genetic correlations (between 0.45 and 0.60) with the arithmetic and harmonic mean of birth weight and with the minimal and maximal birth weight. The standard deviation, variance and the range of birth weight showed also positive, but somewhat lower correlations (between 0.25 and 0.40) with the litter weight at birth. This resulted in a low and negative genetic correlation between the coefficient of variation and litter weight at birth. The skewness of the birth-weight distribution within litter showed low negative correlations with the three litter size traits and with litter weight at birth. Whereas negative correlations due to herd-year-season were calculated between the three litter size traits and the minimal birth weight, the estimates of the correlations of these traits with the maximal birth weight were positive and not significantly different from zero (total number of

piglets born). All estimates of correlations for the herdyear-season effect among the litter size traits and all variability measures (range, variance, standard deviation and coefficient of variation) were positive. The highest values of around 0.40 were calculated for the correlations of the variability measures with the number of piglets weaned. Skewness was not or only very weakly correlated with the litter size traits. The herd-year-season correlation between litter weight at birth and the mean birth weight was near one. Also the minimal and the maximal birth weight showed high correlations with the litter weight at birth. All measures referring to the variability of birth weight within litter showed only low positive correlations with litter weight at birth. There was a negative herd-year-season correlation between litter weight at birth and skewness of the birth weight within litter. The additive-genetic correlations and the correlations caused by the herd-year-season effect between stillborn piglets or losses from birth to weaning and several statistics related to the distribution of individual birth weight within litter are listed in Table 10. The genetic correlation between the mean birth weight and the number of stillborn piglets or losses from birth to weaning (in piglets) was approximately −0.40 or − 0.20, respectively. The genetic correlation of the mean birth weight with the relative values (stillborn piglets and losses from birth to weaning in per cent) were somewhat lower (− 0.30 and −0.13, respectively). The minimal birth weight showed the highest genetic correlation (− 0.50) with the number of stillborn piglets whereas the Table 8 Heritability and proportion of variance caused by herd-year-season for statistical parameters of the distribution of birth weights within litter Statistical parameter

n a Heritability Proportion of variance caused by herd-year-season

Minimal birth weight in litter (kg) Maximal birth weight in litter (kg) Range (maximum– minimum, kg) Arithmetic mean (kg) Harmonic mean (kg) Variance (kg2) Standard deviation (kg) Coefficient of variation (%) Skewness

3

0.10 ± 0.012 0.37 ± 0.022

3

0.11 ± 0.011 0.37 ± 0.019

3

0.02 ± 0.009 0.51 ± 0.028

6 3 3 3

0.16 ± 0.015 0.16 ± 0.020 0.04 ± 0.010 0.03 ± 0.009

3

0.05 ± 0.011 0.49 ± 0.027

3

0.04 ± 0.008 0.04 ± 0.008

0.32 ± 0.014 0.32 ± 0.026 0.40 ± 0.023 0.53 ± 0.026

Both the estimates and their standard errors are calculated as medians from the given number of estimates. a Number of estimates.

J. Wolf et al. / Livestock Science 115 (2008) 195–205

201

Table 9 Additive-genetic correlations (1st line) and correlations caused by herd-year-season (2nd line) between statistical parameters related to individual birth weight within litter on the one hand and litter size traits and litter weight at birth on the other hand Statistical parameter

Total number of piglets born Number of piglets born alive Number of piglets weaned Litter weight at birth

Minimal birth weight in litter (kg) − 0.53 ± 0.090 − 0.33 ± 0.126 Maximal birth weight in litter (kg) − 0.31 ± 0.100 0.11 ± 0.144 Range (maximum–minimum, kg) 0.40 ± 0.134 0.32 ± 0.164 Arithmetic mean (kg) − 0.43 ± 0.082 − 0.08 ± 0.135 Harmonic mean (kg) − 0.46 ± 0.090 − 0.09 ± 0.166 Variance (kg2) 0.26 ± 0.143 0.18 ± 0.162 Standard deviation (kg) 0.21 ± 0.130 0.30 ± 0.110 Coefficient of variation (%) 0.45 ± 0.101 0.32 ± 0.162 Skewness − 0.21 ± 0.085 0.09 ± 0.127

− 0.44 ± 0.100 − 0.20 ± 0.135 − 0.25 ± 0.105 0.20 ± 0.137 0.36 ± 0.131 0.29 ± 0.151 − 0.37 ± 0.085 0.10 ± 0.123 − 0.39 ± 0.091 0.09 ± 0.155 0.20 ± 0.148 0.16 ± 0.145 0.17 ± 0.135 0.28 ± 0.120 0.38 ± 0.106 0.29 ± 0.152 − 0.17 ± 0.086 − 0.15 ± 0.120

genetic correlation of the maximal birth weight with the number of stillborn piglets was in the same order of magnitude as the correlation of the mean birth weight with stillborn piglets. The range, variance, standard deviation and coefficient of variation were positively correlated with the number of stillborn piglets (values between 0.25 and 0.30). The losses from birth to weaning were genetically highly correlated with the range in birth weight. The

−0.30 ± 0.097 −0.42 ± 0.091 −0.30 ± 0.096 0.20 ± 0.102 0.00 ± 0.152 0.44 ± 0.097 −0.33 ± 0.088 −0.09 ± 0.110 −0.35 ± 0.094 −0.14 ± 0.119 0.02 ± 0.147 0.34 ± 0.103 −0.09 ± 0.158 0.42 ± 0.097 0.12 ± 0.127 0.43 ± 0.099 −0.18 ± 0.081 0.11 ± 0.111

0.46 ± 0.099 0.60 ± 0.056 0.61 ± 0.076 0.78 ± 0.039 0.41 ± 0.138 0.13 ± 0.072 0.55 ± 0.079 0.95 ± 0.015 0.51 ± 0.082 0.95 ± 0.013 0.31 ± 0.137 0.13 ± 0.064 0.27 ± 0.138 0.17 ± 0.042 − 0.17 ± 0.106 0.05 ± 0.066 − 0.13 ± 0.080 − 0.38 ± 0.097

genetic correlation between the coefficient of variation and the losses from birth to weaning was about 0.45. Whereas the minimal birth weight had a correlation of about − 0.30 with the losses from birth to weaning, there was no significant genetic correlation between the maximal birth weight and the losses from birth to weaning. The herd-year-season correlation between the mean birth weight and the number of stillborn piglets was

Table 10 Additive-genetic correlations (1st line) and correlations caused by herd-year-season (2nd line) between statistical parameters related to individual birth weight within litter on the one hand and stillborn piglets and losses from birth to weaning on the other hand Statistical parameter

Number of stillborn piglets

Losses from birth to weaning (piglets)

Percentage of stillborn piglets

Losses from birth to weaning (%)

Minimal birth weight in litter (kg)

− 0.50 ± 0.086 − 0.18 ± 0.072 − 0.34 ± 0.077 − 0.14 ± 0.067 0.24 ± 0.306 0.03 ± 0.101 − 0.39 ± 0.065 − 0.26 ± 0.080 − 0.42 ± 0.068 − 0.25 ± 0.080 0.31 ± 0.106 0.02 ± 0.050 0.25 ± 0.183 0.01 ± 0.032 0.30 ± 0.189 0.06 ± 0.116 − 0.22 ± 0.235 0.33 ± 0.124

−0.33 ± 0.111 0.45 ± 0.108 −0.04 ± 0.108 −0.04 ± 0.131 0.55 ± 0.135 −0.38 ± 0.154 −0.17 ± 0.088 0.28 ± 0.135 −0.19 ± 0.094 0.34 ± 0.132 0.24 ± 0.159 −0.30 ± 0.140 0.36 ± 0.174 −0.30 ± 0.140 0.46 ± 0.145 −0.41 ± 0.145 −0.04 ± 0.116 −0.28 ± 0.176

− 0.36 ± 0.123 − 0.16 ± 0.052 − 0.22 ± 0.092 − 0.16 ± 0.056 0.22 ± 0.310 0.00 ± 0.116 − 0.28 ± 0.087 − 0.25 ± 0.075 − 0.31 ± 0.104 − 0.23 ± 0.076 0.31 ± 0.129 0.00 ± 0.048 0.25 ± 0.161 − 0.02 ± 0.048 0.22 ± 0.215 0.03 ± 0.076 − 0.11 ± 0.250 0.27 ± 0.128

− 0.26 ± 0.090 0.41 ± 0.099 0.01 ± 0.100 − 0.05 ± 0.086 0.53 ± 0.191 − 0.36 ± 0.129 − 0.12 ± 0.095 0.24 ± 0.115 − 0.14 ± 0.100 0.29 ± 0.118 0.26 ± 0.164 − 0.30 ± 0.115 0.37 ± 0.182 − 0.29 ± 0.117 0.44 ± 0.141 − 0.39 ± 0.120 − 0.05 ± 0.151 − 0.10 ± 0.166

Maximal birth weight in litter (kg) Range (maximum–minimum, kg) Arithmetic mean (kg) Harmonic mean (kg) Variance (kg2) Standard deviation (kg) Coefficient of variation (%) Skewness

202

J. Wolf et al. / Livestock Science 115 (2008) 195–205

negative, whereas the correlation between the mean birth weight and the losses from birth to weaning was positive. There were no significant herd-year-season correlations between the number of stillborn piglets and the variability measures for birth weight within litter (range, variance, standard deviation, coefficient of variation). The appropriate correlations between the losses from birth to weaning and the mentioned variability measures were negative and in the range from −0.30 to −0.40. A relatively high positive herd-year-season correlation between the minimal birth weight and the losses from birth to weaning was observed.

(2001) reported that their data provided little support for the hypothesis that high birth-weight variation results in decreased survival. Leenhouwers et al. (1999) did not find any relation on the phenotypic scale between the withinlitter standard deviation of birth weight and the proportion of stillborn piglets. Low-birth weight was found to increase the probability for stillbirth (Le Cozler et al., 2002) and preweaning mortality (Milligan et al., 2001). All these results are in good agreement with our findings on the phenotypic relations between the mean birth weight or its within-litter variability and farrowing losses or losses from birth to weaning.

4. Discussion

4.2. Model for the estimation of genetic parameters

The present study analyzed a wide range of statistics referring to the distribution of individual birth weight within litter and the relation of these statistics to litter size traits including farrowing losses and losses from birth to weaning on a large data set of nearly 40,000 individually weighed piglets. This number was exceeded only by Knol et al. (2002) who analyzed a data set with records on approximately 63,000 piglets. In other investigations (Roehe, 1999; Högberg and Rydhmer, 2000; Kaufmann et al., 2000; Hermesch et al., 2001; Damgaard et al., 2003; Huby et al., 2003; Täubert and Henne, 2003; Tribout et al., 2003), the number of weighed piglets was in the range from 15,000 to 30,000. All literature sources analyzed only a low number of statistics referring to the individual birth weight whereas in our investigation nine statistics of the distribution of individual birth weight within litter were included. Though these statistics were partially related with each other, they differed in the heritabilities and in genetic correlations. Therefore, there was a good opportunity to select the statistics which will ensure the most effective selection against piglet mortality. Furthermore, here, the herd-year-season effect was considered to be random and covariance estimates for the herd-year-season effect were calculated for the first time in the present study.

The model for the estimation of the genetic parameters had to be kept simple and robust because of the (relatively) low number of observations. When using more sophisticated models the final status of convergence of the optimization algorithm raised some doubts, whether an optimum had indeed been reached, especially in multitrait models. These problems occurred mainly when the permanent effect of the sow was added to the model. In this case, numerical problems occurred, the estimates of the standard errors of all parameters increased considerably and often correlations of 1 were estimated for the permanent effect. Therefore, the results from the model with the permanent effect did not seem to be trustworthy. The same problem was reported in two studies. Kremer et al. (1999) found that when estimating genetic correlations, including the permanent environmental effect resulted either in estimates with large standard errors or in poor convergence (most of the situations) indicating that the data structure and size was insufficient to disentangle all effects. This problem was solved through simplification of the model by excluding the permanent environmental effect. Also Grandinson et al. (2005) had the intention to use full models for the genetic analyses of the piglet traits, including both the direct genetic effect of the piglet, the maternal genetic effect of the sow, a random permanent effect of the sow and a random litter effect. However, this model could not be used in bi-variate analyses because of convergence problems and the permanent effect of the sow was omitted. As many traits in pig breeding programs exhibit a substantial amount of variation due to the permanent environmental effect, not including this effect in the model will overestimate the additive-genetic component. This was also the case here, for those traits that optimized cleanly (gave a unique solution) with the permanent effect included. However, the additive-genetic correlations did

4.1. Phenotypic relations between birth weight and litter size traits The observed relation that mean birth weight decreases as litter size (total number of piglets born and number of piglets born alive) increases was confirmed in numerous investigations (e.g. Johnson et al., 1999; Roehe, 1999; Knol, 2001; Lund et al., 2002). Within-litter variation in birth weight has been shown to be positively related to preweaning mortality in several papers (Roehe and Kalm, 2000; Milligan et al., 2002a,b), whereas Milligan et al.

J. Wolf et al. / Livestock Science 115 (2008) 195–205

not change too much. As they are of prime interest in this investigation, the authors feel confident, that the generally more appropriate model that includes the permanent environment effect will not lead to different conclusions. 4.3. Estimates of genetic parameters High estimates between 0.30 and nearly 0.50 for the heritability of mean birth weight were published in a number of papers (Högberg and Rydhmer, 2000; RuízFlores and Johnson, 2001; Hermesch et al., 2001; Kim et al., 2002; Täubert and Henne, 2003; Damgaard et al., 2003; Huby et al., 2003). A litter size trait as total number born or number born alive was used as covariable in the statistical model which explains the high values of the estimates. As it was our intention to estimate also correlations between litter size traits and birth weight, litter size was not used as covariable in the present investigation similar as in Kerr and Cameron (1995) and in Crump et al. (1997). This explains the lower heritability estimates in the present study (0.16) and in the two papers mentioned above (0.16 and 0.21). The higher estimates (around 0.10) for heritability of the standard deviation of the birth weight in the literature (Högberg and Rydhmer, 2000; Damgaard et al., 2003; Huby et al., 2003; Täubert and Henne, 2003) compared to 0.03 estimated in the present investigation may be also explained by the use of different models (litter size trait as covariate) in most cases. For the heritability of the coefficient of variation of birth weight, Hermesch et al. (2001) published a value of 0.11 being also higher than our estimate of 0.05. No literature values were found for the heritabilities of the remaining statistics connected with the distribution of birth weight within litter. In a complex selection programme for reproductive traits, probably both piglet weight and litter size traits will be included. In a real multiple-trait model, litter size traits will influence piglet weight via correlations. An additional inclusion of litter size traits as a covariable in the model for birth weight will be far from a mathematically clean solution because the linear relationship between birth weight and litter size would be included twice. The high proportions of variance caused by the herdyear-season effect estimated for all birth-weight statistics show that there were considerable differences among herd-year-season classes. This variance may be partly caused by environmental changes between years and seasons. Roehe and Kalm (2000) analyzed preweaning mortality in piglets with a deviance analysis. They found that year-season was the major contributory factor to preweaning mortality of all fixed effects used in the model. They suggested that this may be due to

203

several factors such as pathogenic and immunological factors, temperature and thermoregulation. A further part of the herd-year-season variance may be induced by the persons who recorded the data in the individual herds. But the high variability among herds indicates mainly that there is a great potential for improving birthweight traits by management. Our finding of a moderate positive genetic correlation between within-litter standard deviation of birth weight and losses from birth to weaning was confirmed by Damgaard et al. (2003). On the other hand, these authors did find only a negligible positive genetic correlation (0.06 vs. 0.25 in the present study) between within-litter standard deviation of birth weight and the proportion of stillborn piglets. The herd-year-season correlations between the birthweight statistics and relative losses (both at birth and from birth to weaning) had often a different sign than the corresponding genetic correlations (Table 10). This may be interpreted in such a way that the management of the farms tries to counteract to the unfavourable genetic correlations. In other words, a preferential treatment of large litters has to be assumed. 4.4. Potential use of birth-weight traits in selection programmes A main question of interest in the present study was if traits referring to individual birth weight may be used in the selection against losses during farrowing and from birth to weaning. The losses from birth to weaning had a relatively high heritability (0.13) and therefore the direct selection on this trait should be the best way of decreasing these losses. The number of stillborn piglets showed a heritability of only 0.04. The minimal birth weight in the litter was the trait which showed the highest genetic correlation (− 0.50) with the number of stillborn piglets and had a heritability of 0.10. It could therefore be used instead of the number of stillborn piglets in selection. Although the effect of selection on farrowing losses should be expected to be similar for direct and indirect selection, there are two further arguments in favour of indirect selection using minimal birth weight. Minimal birth weight has, from the mathematical point of view, a more suitable distribution and it is also moderately correlated with the losses from birth to weaning, so that selection on minimal birth weight should have a positive effect on more than one trait. Gondret et al. (2005), for example, found that light birth-weight pigs were 12 days older at slaughter (at constant weight) than heavy birth-weight litter mates. Dunshea et al. (2003) reported that pigs heavier at

204

J. Wolf et al. / Livestock Science 115 (2008) 195–205

weaning were also heavier at every subsequent age. Unlike cattle and sheep, pigs have no ability for compensating live weight differences between litter mates during growth. Therefore, any reduction in its weight gains from birth onwards will have an amplified detrimental effect on growth from weaning to slaughter (Cole and Varley, 2000). These are further arguments to select for an increased minimal birth weight. It could be argued that selection on the mean birth weight might have a similar effect than selection on the minimal birth weight. The mean birth weight (both arithmetic and harmonic mean) had in our investigation a similar correlation as minimal birth weight with the mortality traits. However, the results of Damgaard et al. (2003) differed partly from our findings. In accordance with our results, they found a moderate negative, and therefore favourable, genetic correlation between mean birth weight and the proportion of dead piglets during suckling. In contrast, they observed a moderate positive genetic correlation between mean birth weight and the proportion of stillborn piglets. They concluded therefore in agreement with Knol (2001) that the net result of an attempt to improve piglet mortality by selection for higher mean birth weight might prove to be very limited. According to our opinion the mean birth weight is not the trait to be changed but the piglets with low-birth weight will cause problems and increase piglet mortality. Though there is a high genetic correlation between the minimal birth weight and the mean birth weight (a value of 0.94 was calculated from our data), the selection on the minimal birth weight is the direct and natural way to reduce the number of piglets with low birth weight. Furthermore, a genetic correlation of − 0.19 was calculated between the minimal birth weight and the within-litter variance of birth weight; therefore it should be expected a positive side effect on the within-litter variability when selecting on minimal birth weight. 5. Conclusion Selection on high litter size was shown to decrease the mean piglet birth weight and to increase the within-litter variability of birth weight. These changes in the birthweight traits cause an increase in the number of stillborn piglets and in the losses from birth to weaning. Therefore, selection on litter size should be accompanied by selection on mortality traits and/or birth-weight traits. On the basis of the estimated heritabilities and genetic correlations, preweaning mortality and the minimal birth weight in the litter are proposed as potential traits for selection against piglet mortality.

Acknowledgements Thanks are due to V. Jelínková from the Association of Pig Breeders in the Czech Republic for making available the data. The technical assistance of P. Mrázková is acknowledged. The research was supported by the Ministry of Agriculture of the Czech Republic (Project MZE 0002701401) and by Project No. 2/04 of GermanCzech cooperation in the field of agricultural research. References Brenig, B., Brem, G., 1992. Molecular cloning and analysis of the porcine “halothane” gene. Arch. Tierz. 35, 129–135. Cole, M., Varley, M., 2000. Weight watchers from birth. New data emphasise the importance of maximising piglet weight from birth onwards. Pig Int. 30, 13–16. Crump, R.E., Haley, C.S., Thompson, R., Mercer, J., 1997. Individual animal model estimates of genetic parameters for reproduction traits of Landrace pigs performance tested in a commercial nucleus herd. Anim. Sci. 65, 285–290. Damgaard, L.H., Rydhmer, L., Løvendahl, P., Grandinson, K., 2003. Genetic parameters for within-litter variation in piglet birth weight and change in within-litter variation during suckling. J. Anim. Sci. 81, 604–610. Dunshea, F.R., Kerton, D.K., Cranwell, P.D., Campbell, R.G., Mullan, B.P., King, R.H., Power, G.N., Pluske, J.R., 2003. Lifetime and post-weaning determinants of performance indices of pigs. Aust. J. Agric. Resour. Econ. 54, 363–370. Gondret, F., Lefaucheur, L., Louveau, L., Lebret, B., Pichodo, X., Le Cozler, Y., 2005. Influence of piglet birth weight on postnatal growth performance, tissue lipogenic capacity and muscle histological traits at market weight. Livest. Prod. Sci. 93, 137–146. Grandinson, K., Rydhmer, L., Strandberg, E., Solanes, F.X., 2005. Genetic analysis of body condition in the sow during lactation, and its relation to piglet survival and growth. Anim. Sci. 80, 33–40. Hermesch, S., Luxford, B.G., Graser, H.U., 2001. Genetic Parameters for Piglet Mortality, within Litter Variation of Birth Weight, Litter Size and Litter Birth Weight. Proc. Assoc. Advmt. Anim. Breed. Genet., vol. 14, pp. 211–214. Högberg, A., Rydhmer, L., 2000. A genetic study of piglet growth and survival. Acta Agric. Scand., A Anim. Sci. 50, 300–303. Huby, M., Gogué, J., Maignel, L., Bidanel, J.P., 2003. Corrélations génétiques entre les caractéristiques numériques et pondérales de la portée, la variabilité du poids des porcelets et leur survie entre la naissance et le sevrage. Journ. Rech. Porc. 35, 293–300. Johnson, R.K., Nielsen, M.K., Casey, D.S., 1999. Responses in ovulation rate, embryonal survival, and litter traits in swine to 14 generations of selection to increase litter size. J. Anim. Sci. 77, 541–557. Kaufmann, D., Hofer, A., Bidanel, J.P., Künzi, N., 2000. Genetic parameters for individual birth and weaning weight and for litter size of Large White pigs. J. Anim. Breed. Genet. 117, 121–128. Kerr, J.C., Cameron, N.D., 1995. Reproductive performance of pigs selected for components of efficient lean growth. Anim. Sci. 60, 281–290. Kim, B.W., Kim, S.D., Lee, I.J., Chung, K.H., Kwon, O.S., Ha, J.K., Lee, J.G., 2002. Estimation of direct and service sire genetic parameters for reproductive traits in Yorkshire. Asian Austral. J. Anim. Sci. 15, 1232–1236.

J. Wolf et al. / Livestock Science 115 (2008) 195–205 Knol, E.F., 2001. Genetic aspects of piglet survival. Ph.D. Thesis, Institute for Pig Genetics and Animal Breeding and Genetics Group, Wageningen Universitet, The Netherlands. Knol, E.F., Ducro, B.J., van Arendonk, J.A.M., van der Lende, T., 2002. Direct, maternal and nurse sow genetic effects on farrowing-, pre-weaning- and total piglet survival. Livest. Prod. Sci. 73, 153–164. Kovač, M., Groeneveld, E., García-Cortés, L.A., 2002. VCE-5 package for the estimation of dispersion parameters. Proceedings of the 7th World Congr. Genet. Appl. Livest. Prod., 33, pp. 741–742. Kremer, V.D., van der Lende, T., de Rouw, O.L.A.M., van Arendonk, J.A.M., 1999. Genetic parameters, inbreeding depression and genetic trend for litter traits in a closed population of Meishan pigs. 50th Annual Meeting of the EAAP, paper Ga 4.5. Lay Jr., D.C., Matteri, R.L., Carroll, J.A., Fangman, T.J., Safranski, T.J., 2002. Preweaning survival in swine. J. Anim. Sci. 80, E74–E86. Le Cozler, Y., Guyomarch, C., Pichodo, X., Quinio, P.Y., Pellois, H., 2002. Factors associated with stillborn and mummified piglets in high-prolific sows. Anim. Res. 51, 261–268. Leenhouwers, J.I., van der Lende, T., Knol, E.F., 1999. Analysis of stillbirth in different lines of pig. Livest. Prod. Sci. 57, 243–253. Leenhouwers, J.I., de Almeida, C.A., Knol, E.F., van der Lende, T., 2001. Progress of farrowing and early postnatal pig behavior in relation to genetic merit for pig survival. J. Anim. Sci. 79, 1416–1422. Lund, M.S., Puonti, M., Rydhmer, L., Jensen, J., 2002. Relationship between litter size and perinatal and pre-weaning survival in pigs. Anim. Sci. 74, 217–222. Marchant, J.N., Rudd, A.R., Mendl, M.T., Broom, D.M., Meridith, M.J., Corning, S., Simmons, P.H., 2000. Timing and causes of piglet mortality in alternative and conventional farrowing systems. Vet. Rec. 147, 209–214.

205

Milligan, B.N., Fraser, D., Kramer, D.L., 2001. Birth weight variation in the domestic pig: effects on offspring survival, weight gain and suckling behaviour. Appl. Anim. Behav. Sci. 73, 179–191. Milligan, B.N., Dewey, C.E., De Grau, A.F., 2002a. Neonatal-piglet weight variation and its relation to pre-weaning mortality and weight gain on commercial farms. Prev. Vet. Med. 56, 119–127. Milligan, B.N., Fraser, D., Kramer, D.L., 2002b. Within-litter birth weight variation in the domestic pig and its relation to pre-weaning survival, weight gain, and variation in weaning weights. Livest. Prod. Sci. 76, 181–191. Neumaier, A., Groeneveld, E., 1998. Restricted maximum likelihood estimation of covariances in sparse linear models. Genet. Sel. Evol. 30, 3–26. Roehe, R., 1999. Genetic determination of individual birth weight and its association with sow productivity using Bayesian analyses. J. Anim. Sci. 77, 330–343. Roehe, R., Kalm, E., 2000. Estimation of genetic and environmental risk factors associated with pre-weaning mortality in piglets using generalized linear mixed models. Anim. Sci. 70, 227–240. Ruíz-Flores, A., Johnson, R.K., 2001. Direct and correlated responses to two-stage selection for ovulation rate and number of fully formed pigs at birth in swine. J. Anim. Sci. 79, 2286–2297. Täubert, H., Henne, H., 2003. Groβe Würfe und wenig Ferkelverlusteein erreichbares Zuchtziel beim Schwein? Züchtungskunde 75, 442–451. Tribout, T., Caritez, J.C., Gogué, J., Gruand, J., Billon, Y., Bouffaud, M., Lagant, H., Le Dividich, J., Thomas, F., Quesnel, H., Guéblez, R., Bidanel, J.P., 2003. Estimation, par utilisation de semence congelée, du progrés genétique réalisé en France entre 1977 et 1998 dans la race porcine Large White: résultats pour quelques caractéres de reproduction femelle. Journ. Rech. Porc. 35, 285–292.