Materials and methods to reproductive problems and lameness Dagorn and

´ M . Lopez-Serrano et al. Livestock Production Science 64 2000 121 –131 123 interest because of their heritability and, moreover, 2710 for Landrace sows, and the total number of they have a relationship with the leg weakness animals in the pedigree was 24 444 for Large White syndrome Grøndalen, 1974a. This relationship and 17 255 for Landrace. Selection of gilts was could indirectly be the cause for culling and con- carried out in the five nucleus farms examining sequent reduction of longevity. The heritabilities of performance traits and exterior condition at approxi- linearly scored exterior traits, such as length of back mately 105 kg live weight. The positively selected and width of hams, representing muscle, were esti- sows were distributed to 65 and 54 multiplier herds mated and resulted in higher estimates of 28 and for Large White and Landrace, respectively. 36 Van Steenbergen, 1990, respectively. Other The performance traits were daily gain and backfat objective measurements are available in the literature thickness, registered as weight at selection day showing higher heritability values due to linearity divided by age in days g day, and as the average and normality of the chosen traits. Lundeheim value of three points measured at the back mm. All 1987 estimated heritabilities of 0.59 for carcass animals were recorded for these traits. Five exterior length. Other objective measurements of slaughter traits describing leg status, length, muscle, height carcass traits, estimated with animal models in boars, and overall type were also scored at the selection day ranged from 0.35 to 0.7 for lean percentage Johan- using a scale from 1 worst to 9 best. Table 1 nson, 1987; Ducos et al., 1993; De Vries et al., 1994; shows the number of records per exterior trait. Schmutz, 1995; Engellandt et al., 1997, and from Two categorical longevity traits were defined for 0.34 to 0.44 for carcass length Schmutz, 1995; those sows that were distributed to multiplier farms Engellandt et al., 1997 in different breeds. and had at least one litter. Survival data from nucleus The genetic relationship of these exterior traits to sows were not considered because their culling rates the leg weakness syndrome has been reviewed by are considerably higher due to genetic selection ¨ different authors and may show an indirect relation Muller, 1997. The number of animals with longevi- to longevity. Lameness as a consequence of the ty records was 8879 for Large White with 34 668 selected exterior is a reason for involuntary culling. litters and 4881 for Landrace with 20 023 litters. The There is a relationship between back length and joint farrowing dates were between July 1987 and April shapes and leg weakness because of joint lesions 1994. The first trait was stayability from first to Schilling, 1963; Grøndalen, 1974a; Lodde et al., second litter STAY12 and the second stayability 1985. Some authors have estimated genetic correla- from first to third litter STAY123, as used previ- tions and concluded that a long body predisposes to ously in dairy cattle Everett et al., 1976a. A value leg weakness Webb et al., 1983; Lundeheim, 1987; of 0 was assigned to a culled sow and 1 for a sow Van Steenbergen, 1990. Using subjective visual surviving up to second or to third litter. All multi- measurements, lower unfavourable genetic correla- plier sows had an observation for both stayability tions were found between ham score and leg score values. End of follow-up of the sows for both traits traits r 5 2 0.28 in crossbred sows Von Brevern, was culling or end of data collection before the g 1996. Grøndalen 1974b also found a relationship respective second or third parity event. They get a between back length and possible skeleton problems. zero in all cases if the survival status is unknown. Stayability records were taken in the first three parities because they have a higher culling rate due

3. Materials and methods to reproductive problems and lameness Dagorn and

Aumaitre, 1979; Dijkhuizen et al., 1989; Dourmad et 3.1. Data al., 1994. Data from 21 870 Large White and 14 944 Land- 3.2. Test of risk race sows, born in five nucleus herds between September 1986 and April 1993, were available for Hazard curves were calculated for both lines from this investigation. Sires and dams of the total popula- raw data using the life-table estimate of the hazard tion were 728 and 4731 for Large White and 567 and function to show the instantaneous risk of disposal ´ 124 M . Lopez-Serrano et al. Livestock Production Science 64 2000 121 –131 for a sow surviving to that instantaneous time after the other one for stayability traits, are as follows: the first litter along the productive life. The life-table herd-year-season at birth of the sows in the nucleus estimate of the hazard function in the jth time farms, with 416 levels for Large White and 418 for 9 interval is: ht 5 d n 2 d 2 t , where d is the Landrace, to consider the influence of the environ- j j j j j 9 number of disposals in this jth time interval, n is the ment at birth and of rearing on the performance and j average number of animals at risk of disposal in that exterior of the sows season is month of birth. interval, i.e., number of sows alive minus the number The fixed effects for the model of stayability are of censored survival times in the interval, t is the herd-quarter-year 144 levels for Large White and j length of the time interval, and the denominator is 141 levels for Landrace, also taking into account the the average time survived in that interval Collet, birth conditions on the longevity of the sows and 1994. The analysis was carried out using the lifetest herd-year-season at first litter 233 levels for Large procedure of SAS 1990 for every line in survival White and 331 for Landrace, considering environ- intervals of 7 days. mental conditions at first farrowing. Age at first litter in six classes corrects for the influence of reproductive maturity on longevity. An uncorrelated 3.3. Statistical analysis random common litter-environment effect was also included. A linear regression coefficient on the final All performance and exterior data from the nu- weight at the selection day in the nucleus farm was cleus and multiplier herds were considered using the considered in both models to correct for weight. following mixed model: Linear and quadratic regression coefficients for the Model 1: Y 5 m 1 HYS 1 bw 1 a 1 e ij i j ij number of piglets born alive in the last litter were taken into account to consider the decision for and for all sows with at least one litter in a multiplier selection of the farmer and to correct the data for farm: production. Model 2: Y 5 m 1 FH 1 LH 1 FA 1 bw ijklm i j k Bayesian multitrait covariance components esti- 2 mation was carried out using the 1997 version of the 1 d x 1 d x 1 c 1 a 1 e 1 2 l m ijklm MTGSAM1.11 program Van Tassel and Van Vleck, where Y , exterior or performance record of animal 1995. Flat prior distributions for fixed effects and ij ij; Y , stayability record 0 or 1 of animal ijklm; variance components were chosen to calculate the ijklm m, general mean; HYS , fixed effect of herd-year- variance components in bivariate analyses of a i season at birth i i 51–416 LW; i 51–418 LR; FH , performance or an exterior trait with one of the i fixed effect of herd-quarter-year at birth i i 51–144 stayability traits in each line. Seven bivariate runs LW; i 51–141 LR; LH , fixed effect herd-year- were carried out between performance or exterior j season at first litter j j 51–233 LW; j 51–331 LR; traits with the first model and a stayability trait with FA , fixed effect of first litter age k k 56; b, the second model in each line; in total 14 runs for k regression coefficient; w, weight at the selection day each line with both stayability traits. To estimate the of the sow; d ,d , regression coefficients; x, number bivariate genetic correlations between performance 1 2 of piglets born alive in the last litter; a , random or leg score and stayability, Gibbs chains consisting j additive genetic effect of animal j j 51–21 870 LW; of 15 000 iterations were used, and between the j 51–14 944 LR; c , uncorrelated random effect of other exterior traits and stayability Gibbs chains of l litter-environment l l 51–5116 LW; l 51–3060 LR; 25 000 cycles were used in both lines. The cycles of a , random additive genetic effect of animal m the chains were analysed visually. After a burn-in m m 51–8879 LW; m 51–4881 LR; e , residual period of 6000 rounds for performance and leg traits ij random error associated with the ijth observation; bivariate runs, a sample size of 8000 rounds was and e , residual random error associated with the taken, and for the bivariate estimate of the other four ijklm ijklmth observation. exterior and stayability traits the burn-in period was The fixed effects in two mixed models, one for 12 000 cycles and the sample size considered was growth and backfat thickness and exterior traits and 13 000. Heritabilities were calculated as mean value ´ M . Lopez-Serrano et al. Livestock Production Science 64 2000 121 –131 125 of every cycle in the chosen sample as the chain al. 1983 found a significantly poorer overall leg achieved a stationary trend, one value for every condition in Landrace than in Large White boars. performance and exterior trait and a mean value from Other frequencies of culling due to lameness were seven bivariate runs for STAY12 and STAY123 in 10.5 of all sows in Dutch herds, with an average each line. The standard deviations of the mean productive lifespan of 2.9 parities Dijkhuizen et al., values of the samples were taken as standard errors 1989, and 12.6 and 10.2 for Landrace and of the estimates, one for every trait, and the mean of crossbreds at early parities Sehested and Scherve, seven estimations for both stayability traits in each 1996, respectively. Culling reasons were not com- line. pletely recorded in our data. 4.2. Pattern of hazard curves

4. Results and discussion