- Email: [email protected]
Phenotypic relationship between test results of Swedish Warmblood horses as 4-year-olds and longevity
Livestock Production Science 68 (2001) 97–105 www.elsevier.com / locate / livprodsci
Phenotypic relationship between test results of Swedish Warmblood horses as 4-year-olds and longevity Lena Wallin*, Erling Strandberg, Jan Philipsson Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden Received 8 May 2000; received in revised form 3 October 2000; accepted 3 October 2000
Abstract The relationship between longevity and different traits scored in the Swedish Riding Horse Quality Test (RHQT) was studied to evaluate their use as predictors of survival. Data comprised 1815 Warmblood horses born between 1969 and 1982 that had participated in the RHQT as 4-year-olds. Survival information was obtained via a questionnaire sent to owners of horses that had participated in the RHQT between 1973 and 1986. All phenotypic values of traits scored at 4 years of age were adjusted for the effect of place / year (event). Survival analysis was performed taking into account censoring. Traits having significant effects on longevity were: conformation, legs (included in conformation), orthopaedic status, jumping ability, and the horses’ combined classification score for dressage and jumping talents, respectively. Orthopaedic health had the greatest influence on longevity, and demonstrated the importance of judging health traits in young sports horses. The results of this study confirmed that there is a significant phenotypic relationship between many of the RHQT traits and longevity, and thus the possibility of using them as predictors of survival. 2001 Elsevier Science B.V. All rights reserved. Keywords: Horse; Longevity; Survival analysis; Test traits
1. Introduction From an economic point of view, longevity is a trait of significant importance in farm animals (Strandberg, 1997; Essl, 1998). Compared with most other farm animals, horses have a long productive life and considerable time and money are invested in the horses in order for them to express their performance potential. Riding horses in particular reach
*Corresponding author. Tel.: 146-18-671955; fax 146-18672648. E-mail address: [email protected] (L. Wallin).
their maximum sport potential as 10–15-year-olds, which shows the necessity of a long productive life. Thus, longevity should be an important trait in horsebreeding objectives. Unfortunately, records of culling and death are rarely available for analytical purposes in horses, which is why it is important to recognize appropriate traits associated with longevity. Ricard and Fournet-Hanocq (1997) analysed ‘length of competitive life’ in jumping horses as an indirect indicator of longevity. An Irish study (Foran et al., 1994) used ‘number of starts in show jumping’ as a measure of longevity. In a previous Swedish study (Wallin et al., 2000), where length of life was estimated in different horse populations, a direct
0301-6226 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0301-6226( 00 )00244-X
98
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
measure of longevity could be used, as records were accessible for culled and dead horses. Longevity studies focus on finding traits of value as early indicators of longevity. Some type traits scored in first-lactating cows have been found to be good predictors of longevity in dairy cattle (e.g. ¨ Solkner and Petschina, 1999; Larroque and Ducrocq, 1999; Schneider et al., 1999). An analogous early performance test for horses is the Swedish Riding Horse Quality Tests (RHQT), which is a 1-day field test for 4-year-old Warmblood horses, where conformation, gaits under rider, jumping ability, temperament, and health status are recorded. The question is: do these traits have any value as predictors of longevity? The ‘survival analysis’ method (Ducrocq and ¨ Solkner, 1999) seems to be the most appropriate method for analysing the phenotypic relationship between the RHQT traits and length of life, because it is important to be able to take into account horses that are censored, i.e. still alive at the end of the study. Without taking these horses into account, valuable information would be lost and the results would be biased. The aims of this study were: (i) to study whether the results of different scores awarded in field tests (RHQT) at an early age in a horse’s life are related to the risk of dying or being culled later in life; and (ii) to establish whether a thorough examination, including health tests, at about 1 year after a horse’s training has started, would be a good indicator of longevity for horses.
analysis. The distribution between females and males was fairly even (848 / 967). Of the males 70 were stallions at the time of the test. Traits scored in the RHQT were: Conformation traits: type head–neck and body correctness of legs locomotion walk showed at hand; locomotion trot showed at hand; Performance traits: jumping
temperament (J) walk trot canter temperament (G)
Health traits: orthopaedic status
medical status 2. Material Data on horses scored in the Swedish Riding Horse Quality Tests were supplied by the Swedish Horse Board. Data were combined with survival information obtained by a questionnaire sent to owners of Swedish Warmblood horses that had participated as 4-year-olds in the RHQT between 1973 and 1986. These horses were born between 1969 and 1982. For a more detailed description of the data, see Wallin et al. (2000). The response frequency to the questionnaire was 78%. Information on horses was available up to and including 1990. A total of 1815 horse records were available for
free jumping, or jumping under rider, or an average of both depending on year of test; temperament and general appearance in jumping; showed under rider; showed under rider; showed under rider; temperament and general appearance in gaits (walk, trot, and canter under rider); an orthopaedic health examination including flexion test. Orthopaedic status includes palpation of legs and a flexion test of all legs; a medical health examination. Medical status includes auscultation of heart and lungs and an examination of body and hooves. Mouth and mucous membranes are inspected and a cough test of each horse is made;
In addition, five secondary variables were devised: conformation
gaits under rider
a sum of individual scores for type; head–neck and body; legs; walk at hand; and trot at hand; based on an average of walk,
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
health score dressage quartile
jumping quartile
trot, and canter, showed under rider; based on: (medical status 1 2* orthopaedic status) / 3; based on: (health score 1 conformation 1 2*gaits 1 temperament gaits) / 5. All horses were divided into four groups depending on their results in the tests. Group 1 included the 25% best-scored horses at each place, and group 2 the next 25% best-scored horses, and so on. The purpose of classification into quartiles is to group the horses according to their talents for dressage as well as for jumping, based on their ranking within each test. based on (health score 1 conformation 1 2*jumping 1 temperament jumping) / 5. All horses were divided into four groups in the same way as for dressage quartile.
All traits were scored subjectively between one (very poor) and ten (excellent) and traits based on several traits also had values between one and ten. The conformation trait, which consisted of five subtraits with scores between one and ten, gave a maximum possible score of 50. All horses obtained
99
temperament scores when judged for gaits under rider and jumping, respectively. Both the medical and the orthopaedic health examination include clinical examination by qualified veterinary surgeons. Traits included in the survival analyses were conformation; correctness of legs; jumping; walk, trot, and canter under rider; gaits under rider; orthopaedic and medical status; dressage and jumping quartiles. Means and standard deviations for the traits scored in the RHQT are shown in Table 1. The differences in numbers of horses between different traits were due to some horses not completing the tests.
3. Methods Before analysing the relationship between the RHQT results and longevity, all phenotypic values, except the quartile traits, were adjusted for the effect of place / year (event) using PROC GLM in SAS ´ (1989) (Arnason, 1987). Information on judges was not available, but adjustment for place / year also includes the effects of different scoring between judges. Survival analysis was used in order to take account of censoring (Survival Kit V3.1 by Ducrocq ¨ and Solkner, 1999). The advantage of using this version of the package was the possibility of using discrete longevity data. Our longevity values were measured to the nearest year, which means that
Table 1 Number of horses, means, standard deviations, minimum and maximum values for traits scored in the RHQT, distributed by sex (males (M) and females (F)) Trait
Conformation Legs Jumping Walk Trot Canter Gaits Orthopaedic Medical Health score
Number
Mean
Standard deviation
Minimum
Max
M
F
M
F
M
F
All
All
961 961 950 950 950 950 950 967 958 958
846 846 818 832 832 832 832 848 837 837
37.1 7.0 6.7 6.5 6.4 6.7 6.6 7.9 8.8 8.2
36.9 7.0 6.7 6.3 6.2 6.4 6.3 7.8 8.8 8.1
3.0 0.8 1.5 1.2 1.2 1.1 1.0 1.4 1.0 1.0
3.0 0.8 1.6 1.2 1.2 1.1 1.0 1.4 1.0 1.0
25 4 1 2 2 2 2.7 3 4 4.3
47 10 10 10 10 10 9.3 10 10 10.0
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
100
several horses had exactly the same length of life (so-called ties). Longevity was measured as length of life defined as years between field test and death, and horses still alive at the end of the study period were right-censored. The proportion of censoring was 73%. A discrete proportional hazard model was used, fitting a Prentice and Gloeckler’s model (Prentice and Gloeckler, 1978; Ducrocq, 1999). The model used for the hazard at time t was: h(t 5 tk ; x) 5 e j k 1x9 b The effects included in the model were as follows:
jk
time-dependent covariate (time]unit) taking into account the discrete scale. The time-dependent covariate is a step function of time, with changes at t0 5 0, t1 5 1, t2 5 2, . . . . tk 5 k; k 5 17 years and in b
sex test year
RHQT trait
fixed time-independent effect (female and male); fixed time-independent effect of different year of testing (three test-year classes) and fixed time-independent classified effect (11 different traits). Depending on how the scale was used, some scores were grouped together in order to get reasonable number of horses in different groups.
Analyses were made, using one RHQT trait at a time. Test year was included in the model as the results of a previous study showed a positive trend in length of life over this period of time (Wallin et al., 2000). As in that study horses were divided into three groups according to year of testing (1973–1979; 1980–1983; 1984–1986). The proportional hazard assumption was verified for test year by plotting
log(2log S(t)) vs. log t using PROC LIFETEST in SAS (1989). All effects included in the analyses were tested for their significance, using a likelihood ratio test. Interactions between sex and RHQT traits were tested, but were not significant. Risk ratio (RR) measures the relative risk of a horse being culled due to its score for a certain RHQT trait, relative to a base class used for comparison. Risk ratios were calculated as the hazard for each level of an effect, divided by the hazard for the level of the effect that was chosen as a basis for comparison, e.g. year class 1980–1983.
4. Results The effects were tested in a likelihood ratio test and the results showed that sex had a highly significant effect (P,0.001) on a horse’s hazard, or the relative risk of being culled. Test-year groups were also highly significant (P,0.001), except when legs, jumping and medical status were included in the models (P,0.01). Time]unit had a significant effect (P,0.05) except for medical status. These results show that the discrete proportional hazard model was useful for analyses of these data for almost all traits. As risk ratios were estimated for the two covariates ‘sex’ and ‘test-year group’ in all models when all RHQT traits were modelled one at a time, an average RR was calculated for these effects (Table 2). There was a large difference in RR between sexes, showing that males were twice as likely to be culled as females. The RR for sex varied
Table 2 Average risk ratios (RR) and minimum (Min) and maximum (Max) values, for the covariates sex and test year group from all traits analysed Covariate
RR
Min
Max
Sex Male Female
2.02 1.00
2.00
2.05
Test year group 1973–1979 1980–1983 1984–1986
1.34 1.00 0.80
1.29
1.38
0.79
0.84
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
101
Table 3 Risk ratios (RR) for conformation and legs Conformation (P,0.05)
Legs (P,0.05)
Class
RR
Number of horses
Number of dead horses
Class
RR
Number of horses
Number of dead horses
#32 33–34 35–37 38–40 $41
1.20 a 1.04 a 1.00 a 0.95 a 0.64 b
88 204 696 653 166
38 62 173 174 46
#5 6 7 $8
1.32 a 1.00 a 0.79 b 0.73 b
67 357 918 465
29 113 228 113
a,b
Values within a column with different superscripts are significantly different (P,0.05).
little, depending on which RHQT trait was included in the model. A lower RR was observed when horses were tested between 1984 and 1986, compared with horses tested between 1973 and 1979. Again, only a small variation in RR for test-year group was revealed by the analyses including different RHQT traits. Table 3 shows the RR for conformation and its sub-trait correctness of legs. Both traits showed significant effects (P,0.05) on longevity. For conformation, RR decreased from 1.20 to 0.64 between the worst and best scored horses. This means that horses with the lowest scores were twice as likely to be culled as the best scored horses. A clear trend in RR was observed, from low to high conformation score. The highest best scored horses had a significantly lower risk of being culled than the average horse. The same trend could be seen for the sub-trait legs. Horses that scored seven or more had greater chance of surviving than horses scoring six or lower. Results for the health traits are shown in Table 4.
Orthopaedic status significantly (P,0.001) affected the RR, whereas the effects of medical status were non-significant. Horses with an orthopaedic score of nine or ten had about half the likelihood of being culled than horses scoring below 6. Jumping ability showed a curvilinear trend in RR. Horses with a score of five or less or a score above eight had less chance of surviving than horses with average scores (Table 5). Jumping ability had a significant effect (P,0.05) on longevity, whereas gaits under rider did not. However, results are presented because of their value for this particular study. Table 6 shows RR for walk, trot and canter, and for the gaits as an average of each horse. None of them, except canter, showed significant differences in RR explained by different scores. However, the tendency was that horses gaining higher average scores for their gaits survived somewhat longer than horses with lower average scores. In Table 7, the overall classification scores as talents for dressage and jumping are related to RR.
Table 4 Risk ratios (RR) for orthopaedic and medical status Orthopaedic status (P,0.001)
Medical status (n.s.)
Class
RR
Number of horses
Number of dead horses
#5 6 7 8 $9
1.49 a 1.22 ab 1.17 ab 1.00 b 0.76 c
81 151 342 590 651
29 50 104 157 148
a,b,c
Class
RR
Number of horses
Number of dead horses
#7 8 9 10
1.01 a 1.11 a 1.00 a 0.97 a
134 482 829 350
37 130 216 90
Values within a column with different superscripts are significantly different (P,0.05).
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
102 Table 5 Risk ratios (RR) for jumping
5. Discussion The results showed that sex had a significant effect on longevity. The rather large difference in risk ratio between females and males is explained by the fact that females were often used not only for riding and competition but also for reproduction, which prolonged their lives, as previously shown by Wallin et al. (2000). The same study showed that there was a time trend toward an increased length of life for both sexes among horses that participated in the RHQT. Length of life in that study used PROC LIFEREG in SAS (1989) for the analyses. The same time trend was seen when test-year group was included in the model in this study, using the Survival Kit V3.1 of ¨ Ducrocq and Solkner (1999). Both conformation and legs (included in conformation) showed lower risk ratios for horses with good and excellent scores. Horses with favourable
Jumping (P,0.05) Class
RR
Number of horses
Number of dead horses
#5 6 7 8 $9
1.36 a 0.91 b 1.00 b 0.97 b 1.16 ab
299 403 488 390 188
100 94 129 99 58
a,b Values within a column with different superscripts are significantly different (P,0.05).
Jumping quartile had quite a large effect (P,0.01) on longevity, whereas dressage quartile showed a smaller effect (P,0.05). Horses in group IV in both jumping and dressage quartiles were most likely to be culled.
Table 6 Risk ratios (RR) for walk, trot, canter and gaits (average) Class
Walk (n.s.)
#5 6 7 8 $9 a,b,c,d
Trot (n.s.)
Canter (n.s.)
Gaits (average) (n.s.)
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
0.95 a 0.95 a 1.00 a 0.74 a 1.13 a
369 578 527 251 57
101 147 149 58 20
1.08 a 1.05 a 1.00 a 0.84 a 1.09 a
376 593 491 252 70
101 155 135 63 21
1.19 ad 1.14 ad 1.00 a 1.33 bd 0.81 ac
233 587 592 281 89
62 152 154 86 21
1.17 a 1.04 a 1.00 a 0.98 a
256 632 629 265
73 158 173 71
Values within a column with different superscripts are significantly different (P,0.05).
Table 7 Risk ratios (RR) for dressage and jumping quartiles Quartile
I II III IV a,b
Dressage quartile (P,0.05)
Jumping quartile (P,0.01)
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
0.98 a 1.22 ab 1.00 a 1.35 b
495 462 427 404
124 130 109 118
0.97 a 0.99 a 1.00 a 1.43 b
492 453 445 398
126 110 115 130
Values within a column with different superscripts are significantly different (P,0.05).
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
conformation have been assumed to have a better chance of staying sound and be able to perform over a long period of time. An unfavourable conformation, especially incorrect extremities, does not make the horse unsound itself, but could predispose the horse to disease later in life. The results of this study support that hypothesis. Lately, in several studies, survival analysis was used when studying the relationships between several conformation traits and longevity. Both in dairy cattle and in pigs, significant relationships between some conformation traits and longevity have been ¨ shown (Solkner and Petschina, 1999; Larroque and Ducrocq, 1999; Schneider et al., 1999; Cederberg Ringmar, 1999). There are only a few reports on the relationship between conformation and soundness in horses. Nevertheless, Magnusson and Thavfelin (1985) found that 11% of the variation in soundness was explained by conformation traits for standard-bred trotters. However, several studies indicate that conformation does have an impact on performance. Langlois et al. (1978), Langlois (1979) and ¨ et al. (1990) found that conformation Holmstrom differed significantly between poor and good show ´ jumpers as well as between elite dressage horses / show jumpers and ‘ordinary’ horses. Both studies showed that successfully performing horses need a specific conformation to be able to stay in shape and be able to perform successfully. This study showed that there is a highly significant relationship between orthopaedic status scores and length of life. A score of six or less in orthopaedic status was usually awarded to horses with rather critical remarks on their legs and on joint soundness. Horses having reactions to flexion tests were generally scored six or less (L. Roepstorff pers. comm., 2000). The orthopaedic examination focuses on leg and joint problems, and one should bear in mind that leg problems are the reason for culling more than half of Warmblood horses (Wallin et al., 2000). These results are of great interest as they indicate a high predictive value for longevity already when orthopaedic status scores are awarded when the horses are 4 years old and have undergone about 1 year of training. Therefore, the possibility of using these orthopaedic examinations for regular health
103
recording in young horses for advisory purposes as a means of preventing horses from becoming diseased should be widely adopted. ¨ and Philipsson (1993) found that a Holmstrom few conformational measures were significantly correlated with orthopaedic status among 195 RHQT 4-year-old horses, which could explain how both conformation and orthopaedic status influenced longevity. Low scores (#5) as well as high scores ($9) in jumping gave higher risk ratios. A possible interpretation of that result is that low-scoring horses had a number of problems, including mental inability, leading to a higher culling rate than among average jumpers. On the other hand, the highly scored horses were used largely in sports at a high level (Ohlsson and Philipsson, 1992) and thus were used quite extensively. Therefore, such use may put greater strain on the horse and consequently lead to a shorter life. Ricard and Fournet-Hanocq (1997) used ‘number of years in competition’ for show jumpers as a measure of longevity. Their results showed that the relative culling rate becomes smaller for horses that started to compete at the age of four years, compared with horses starting their competition career at the age of 5 or 6 years. The classification into different quartiles for dressage and jumping influenced longevity, as horses in class IV of both disciplines were less likely to survive. These quartile scores include the health scores as previously shown. Ohlsson and Philipsson (1992) could also show that none of the horses in quartile IV had competition results at the most advanced level in either dressage or jumping. Medical health had no significant effect on longevity. This could be explained both by its lower variance and by its complexity, involving different independent health parameters. Neither gaits judged under rider nor overall gait scores influenced longevity significantly, but the overall gait score was favourable, though also in this case one may assume that highly scored horses are used extensively for sports purposes. Gaits under rider have rather high ´ genetic correlation with conformation (Arnason, 1993), as both walk and trot showed at hand are included in the conformation score. Therefore, it could be assumed that gaits as well as conformation
104
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
have an effect on longevity, but this could not be shown. This study emphasizes the value of judging different traits, especially health traits, early in horse’s life. The relationship between orthopaedic scores and longevity showed that orthopaedic scores could be used to predict longevity. When the RHQT was introduced in Sweden, it was emphasized that it should include assessments of health as well as of conformation and performance. Performance results from the RHQT have been used for a long time to estimate breeding values for Warmblood horses in ´ Sweden (Arnason, 1993). Health results from the RHQT have also been utilized for various purposes, e.g. for health examination in relation to sales and insurance of horses. For voluntary health recording of 4-year-old horses, the RHQT system has proved its worth. There is definitely a need for systematic recording of age of culling or death, for further investigations and monitoring of longevity. An extended study on the relationship between RHQT traits and longevity where the whole scoring scale could be used and hopefully reveal risk ratios for all scores would be desirable. Questionnaires are quite useful, but have the disadvantage of being time-consuming. The breeding associations would be appropriate places for registering survival data, i.e. age at culling or death and also cause of death. The practise in Sweden is to register all new-born foals, a procedure which has high acceptance among horse owners, and the question is whether registering deaths would be similarly accepted.
6. Conclusions The results of this study confirmed that there are significant phenotypic relationships between many of the RHQT traits judged early in horses’ lives and longevity on the one hand, and the possibility of using them as predictors of survival and for preventive medicine purposes on the other. Conformation, legs (included in conformation), orthopaedic status, jumping ability, and dressage and jumping quartiles were all traits associated with length of life. Orthopaedic health had the greatest influence on longevity, thus showing the importance of including health
recording early in horses’ lives. Conformation and performance, as well as health, must be of interest when selecting young horses for sports purposes.
7. Personal communication Vet. Med. Dr. Lars Roepstorff. Department of Equine Studies, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden.
Acknowledgements This work received financial support from the AGRIA Insurance Company, which is gratefully acknowledged. The Swedish Horse Board is gratefully acknowledged for providing materials for the study. We also thank individual horse owners who answered the questionnaires and made it possible for us to undertake this study.
References ´ Arnason, Th., 1987. Contribution of various factors to genetic evaluations of stallions. Livest. Prod. Sci. 16, 407–419. ´ Arnason, Th., 1993. Rapport om genetiska analyser av kvalitet¨ ˚ ¨ sbedomningar av svenska fyraariga ridhastar 1973–1993. (Report on genetic analysis of Riding Horse Quality Test data for ˚ Swedish 4-year-old horses). IHBC AB Knubbo, Morgongava, Sweden, 25 pp. Cederberg Ringmar, E., 1999. Environmental and Genetic Influence on Sow Longevity. Lic. Thesis, SLU, Uppsala. Ducrocq, V., 1999. Extension of survival analysis models to discrete measures of longevity. INTERBULL Bull. 21, 41–47. ¨ Ducrocq, V., Solkner, J., 1999. The Survival Kit-V3.1. http: / / www.boku.ac.at / nuwi / popgen Essl, A., 1998. Longevity in dairy cattle breeding: a review. Livest. Prod. Sci. 57, 79–89. Foran, M.K., Cromie, A.R., Reilly, M.P., Kelleher, D.L., Brophy, P.O., 1994. Analysis of show jumping data in the Irish sport horse population. 45th Ann. Meet. EAAP, Edinburgh, Scotland, 5–7 September, 4 pp. ¨ M., Magnusson, L.-E., Philipsson, J., 1990. Variation Holmstrom, in conformation of Swedish Warmblood horses and conforma´ tional characteristics of elite horses. Equine Vet. J. 22 (3), 186–193. ¨ Holmstrom, M., Philipsson, J., 1993. Relationship between conformation, performance and health of 4-year-old Swedish riding horses. Livest. Prod. Sci. 33, 293–312.
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105 Langlois, B., Froidevaux, J., Lamarche, L., Legault, C., Legault, P., Tassencourt, L.O., Theret, H., 1978. Analyse des liaisons entre la morphologie et l’aptitude au galop au trot au saut ´ ´ Sel. ´ Anim. 10, 443– d’obstacles chez le cheval. Ann. Genet. 474. Langlois, B., 1979. French results on the analysis of relationship between morphology and gallop, trot and jumping abilities in horses. 30th Ann. Meet. EAAP, Harrogate, UK, 23–26 July, 8 pp. Larroque, H., Ducrocq, V., 1999. Phenotypic relationship between type and longevity in the Holstein breed. INTERBULL Bull. 21, 96–103. Magnusson, L.-E., Thavfelin, B., 1985. Studies on conformation and related traits of standard-bred trotters in Sweden. IV. Relationship between the conformation and soundness of 4year-old Standard-bred trotters. Thesis, SLU, Skara. ISBN 91-576-2315-5. Ohlsson, L., Philipsson, J., 1992. Relationship between field test for Warmblood horses as 4-year-olds and later competition performance. 43rd Ann. Meet. EAAP, Madrid, Spain, 14–17 September, 7 pp.
105
Prentice, R., Gloeckler, L., 1978. Regression analysis of grouped survival data with application to breast cancer data. Biometrics 34, 57–67. Ricard, A., Fournet-Hanocq, F., 1997. Analysis of factors affecting length of competitive life of jumping horses. Genet. Sel. Evol. 29, 251–267. SAS Institute Inc., 1989. SAS Users Guide: Statistics, 1982 Edition. SAS Institute Inc., Cary, NC, p. 584. Schneider, M.D.P., Monardes, H.G., Cue, R.I., 1999. Effects of type traits on functional herd life in Holstein cows. INTERBULL Bull. 21, 111–116. ¨ Solkner, J., Petschina, R., 1999. Relationship between type traits and longevity in Austrian Simmental cattle. INTERBULL Bull. 21, 91–95. Strandberg, E., 1997. Breeding strategies to improve longevity. 48th Ann. Meet. EAAP, Vienna, Austria, 25–28 August, 12 pp. Wallin, L., Strandberg, E., Philipsson, J., Dalin, G., 2000. Estimates of longevity and causes of culling and death in Swedish warmblood and coldblood horses. Livest. Prod. Sci. 63 (3), 275–289.
Phenotypic relationship between test results of Swedish Warmblood horses as 4-year-olds and longevity Lena Wallin*, Erling Strandberg, Jan Philipsson Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden Received 8 May 2000; received in revised form 3 October 2000; accepted 3 October 2000
Abstract The relationship between longevity and different traits scored in the Swedish Riding Horse Quality Test (RHQT) was studied to evaluate their use as predictors of survival. Data comprised 1815 Warmblood horses born between 1969 and 1982 that had participated in the RHQT as 4-year-olds. Survival information was obtained via a questionnaire sent to owners of horses that had participated in the RHQT between 1973 and 1986. All phenotypic values of traits scored at 4 years of age were adjusted for the effect of place / year (event). Survival analysis was performed taking into account censoring. Traits having significant effects on longevity were: conformation, legs (included in conformation), orthopaedic status, jumping ability, and the horses’ combined classification score for dressage and jumping talents, respectively. Orthopaedic health had the greatest influence on longevity, and demonstrated the importance of judging health traits in young sports horses. The results of this study confirmed that there is a significant phenotypic relationship between many of the RHQT traits and longevity, and thus the possibility of using them as predictors of survival. 2001 Elsevier Science B.V. All rights reserved. Keywords: Horse; Longevity; Survival analysis; Test traits
1. Introduction From an economic point of view, longevity is a trait of significant importance in farm animals (Strandberg, 1997; Essl, 1998). Compared with most other farm animals, horses have a long productive life and considerable time and money are invested in the horses in order for them to express their performance potential. Riding horses in particular reach
*Corresponding author. Tel.: 146-18-671955; fax 146-18672648. E-mail address: [email protected] (L. Wallin).
their maximum sport potential as 10–15-year-olds, which shows the necessity of a long productive life. Thus, longevity should be an important trait in horsebreeding objectives. Unfortunately, records of culling and death are rarely available for analytical purposes in horses, which is why it is important to recognize appropriate traits associated with longevity. Ricard and Fournet-Hanocq (1997) analysed ‘length of competitive life’ in jumping horses as an indirect indicator of longevity. An Irish study (Foran et al., 1994) used ‘number of starts in show jumping’ as a measure of longevity. In a previous Swedish study (Wallin et al., 2000), where length of life was estimated in different horse populations, a direct
0301-6226 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0301-6226( 00 )00244-X
98
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
measure of longevity could be used, as records were accessible for culled and dead horses. Longevity studies focus on finding traits of value as early indicators of longevity. Some type traits scored in first-lactating cows have been found to be good predictors of longevity in dairy cattle (e.g. ¨ Solkner and Petschina, 1999; Larroque and Ducrocq, 1999; Schneider et al., 1999). An analogous early performance test for horses is the Swedish Riding Horse Quality Tests (RHQT), which is a 1-day field test for 4-year-old Warmblood horses, where conformation, gaits under rider, jumping ability, temperament, and health status are recorded. The question is: do these traits have any value as predictors of longevity? The ‘survival analysis’ method (Ducrocq and ¨ Solkner, 1999) seems to be the most appropriate method for analysing the phenotypic relationship between the RHQT traits and length of life, because it is important to be able to take into account horses that are censored, i.e. still alive at the end of the study. Without taking these horses into account, valuable information would be lost and the results would be biased. The aims of this study were: (i) to study whether the results of different scores awarded in field tests (RHQT) at an early age in a horse’s life are related to the risk of dying or being culled later in life; and (ii) to establish whether a thorough examination, including health tests, at about 1 year after a horse’s training has started, would be a good indicator of longevity for horses.
analysis. The distribution between females and males was fairly even (848 / 967). Of the males 70 were stallions at the time of the test. Traits scored in the RHQT were: Conformation traits: type head–neck and body correctness of legs locomotion walk showed at hand; locomotion trot showed at hand; Performance traits: jumping
temperament (J) walk trot canter temperament (G)
Health traits: orthopaedic status
medical status 2. Material Data on horses scored in the Swedish Riding Horse Quality Tests were supplied by the Swedish Horse Board. Data were combined with survival information obtained by a questionnaire sent to owners of Swedish Warmblood horses that had participated as 4-year-olds in the RHQT between 1973 and 1986. These horses were born between 1969 and 1982. For a more detailed description of the data, see Wallin et al. (2000). The response frequency to the questionnaire was 78%. Information on horses was available up to and including 1990. A total of 1815 horse records were available for
free jumping, or jumping under rider, or an average of both depending on year of test; temperament and general appearance in jumping; showed under rider; showed under rider; showed under rider; temperament and general appearance in gaits (walk, trot, and canter under rider); an orthopaedic health examination including flexion test. Orthopaedic status includes palpation of legs and a flexion test of all legs; a medical health examination. Medical status includes auscultation of heart and lungs and an examination of body and hooves. Mouth and mucous membranes are inspected and a cough test of each horse is made;
In addition, five secondary variables were devised: conformation
gaits under rider
a sum of individual scores for type; head–neck and body; legs; walk at hand; and trot at hand; based on an average of walk,
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
health score dressage quartile
jumping quartile
trot, and canter, showed under rider; based on: (medical status 1 2* orthopaedic status) / 3; based on: (health score 1 conformation 1 2*gaits 1 temperament gaits) / 5. All horses were divided into four groups depending on their results in the tests. Group 1 included the 25% best-scored horses at each place, and group 2 the next 25% best-scored horses, and so on. The purpose of classification into quartiles is to group the horses according to their talents for dressage as well as for jumping, based on their ranking within each test. based on (health score 1 conformation 1 2*jumping 1 temperament jumping) / 5. All horses were divided into four groups in the same way as for dressage quartile.
All traits were scored subjectively between one (very poor) and ten (excellent) and traits based on several traits also had values between one and ten. The conformation trait, which consisted of five subtraits with scores between one and ten, gave a maximum possible score of 50. All horses obtained
99
temperament scores when judged for gaits under rider and jumping, respectively. Both the medical and the orthopaedic health examination include clinical examination by qualified veterinary surgeons. Traits included in the survival analyses were conformation; correctness of legs; jumping; walk, trot, and canter under rider; gaits under rider; orthopaedic and medical status; dressage and jumping quartiles. Means and standard deviations for the traits scored in the RHQT are shown in Table 1. The differences in numbers of horses between different traits were due to some horses not completing the tests.
3. Methods Before analysing the relationship between the RHQT results and longevity, all phenotypic values, except the quartile traits, were adjusted for the effect of place / year (event) using PROC GLM in SAS ´ (1989) (Arnason, 1987). Information on judges was not available, but adjustment for place / year also includes the effects of different scoring between judges. Survival analysis was used in order to take account of censoring (Survival Kit V3.1 by Ducrocq ¨ and Solkner, 1999). The advantage of using this version of the package was the possibility of using discrete longevity data. Our longevity values were measured to the nearest year, which means that
Table 1 Number of horses, means, standard deviations, minimum and maximum values for traits scored in the RHQT, distributed by sex (males (M) and females (F)) Trait
Conformation Legs Jumping Walk Trot Canter Gaits Orthopaedic Medical Health score
Number
Mean
Standard deviation
Minimum
Max
M
F
M
F
M
F
All
All
961 961 950 950 950 950 950 967 958 958
846 846 818 832 832 832 832 848 837 837
37.1 7.0 6.7 6.5 6.4 6.7 6.6 7.9 8.8 8.2
36.9 7.0 6.7 6.3 6.2 6.4 6.3 7.8 8.8 8.1
3.0 0.8 1.5 1.2 1.2 1.1 1.0 1.4 1.0 1.0
3.0 0.8 1.6 1.2 1.2 1.1 1.0 1.4 1.0 1.0
25 4 1 2 2 2 2.7 3 4 4.3
47 10 10 10 10 10 9.3 10 10 10.0
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
100
several horses had exactly the same length of life (so-called ties). Longevity was measured as length of life defined as years between field test and death, and horses still alive at the end of the study period were right-censored. The proportion of censoring was 73%. A discrete proportional hazard model was used, fitting a Prentice and Gloeckler’s model (Prentice and Gloeckler, 1978; Ducrocq, 1999). The model used for the hazard at time t was: h(t 5 tk ; x) 5 e j k 1x9 b The effects included in the model were as follows:
jk
time-dependent covariate (time]unit) taking into account the discrete scale. The time-dependent covariate is a step function of time, with changes at t0 5 0, t1 5 1, t2 5 2, . . . . tk 5 k; k 5 17 years and in b
sex test year
RHQT trait
fixed time-independent effect (female and male); fixed time-independent effect of different year of testing (three test-year classes) and fixed time-independent classified effect (11 different traits). Depending on how the scale was used, some scores were grouped together in order to get reasonable number of horses in different groups.
Analyses were made, using one RHQT trait at a time. Test year was included in the model as the results of a previous study showed a positive trend in length of life over this period of time (Wallin et al., 2000). As in that study horses were divided into three groups according to year of testing (1973–1979; 1980–1983; 1984–1986). The proportional hazard assumption was verified for test year by plotting
log(2log S(t)) vs. log t using PROC LIFETEST in SAS (1989). All effects included in the analyses were tested for their significance, using a likelihood ratio test. Interactions between sex and RHQT traits were tested, but were not significant. Risk ratio (RR) measures the relative risk of a horse being culled due to its score for a certain RHQT trait, relative to a base class used for comparison. Risk ratios were calculated as the hazard for each level of an effect, divided by the hazard for the level of the effect that was chosen as a basis for comparison, e.g. year class 1980–1983.
4. Results The effects were tested in a likelihood ratio test and the results showed that sex had a highly significant effect (P,0.001) on a horse’s hazard, or the relative risk of being culled. Test-year groups were also highly significant (P,0.001), except when legs, jumping and medical status were included in the models (P,0.01). Time]unit had a significant effect (P,0.05) except for medical status. These results show that the discrete proportional hazard model was useful for analyses of these data for almost all traits. As risk ratios were estimated for the two covariates ‘sex’ and ‘test-year group’ in all models when all RHQT traits were modelled one at a time, an average RR was calculated for these effects (Table 2). There was a large difference in RR between sexes, showing that males were twice as likely to be culled as females. The RR for sex varied
Table 2 Average risk ratios (RR) and minimum (Min) and maximum (Max) values, for the covariates sex and test year group from all traits analysed Covariate
RR
Min
Max
Sex Male Female
2.02 1.00
2.00
2.05
Test year group 1973–1979 1980–1983 1984–1986
1.34 1.00 0.80
1.29
1.38
0.79
0.84
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
101
Table 3 Risk ratios (RR) for conformation and legs Conformation (P,0.05)
Legs (P,0.05)
Class
RR
Number of horses
Number of dead horses
Class
RR
Number of horses
Number of dead horses
#32 33–34 35–37 38–40 $41
1.20 a 1.04 a 1.00 a 0.95 a 0.64 b
88 204 696 653 166
38 62 173 174 46
#5 6 7 $8
1.32 a 1.00 a 0.79 b 0.73 b
67 357 918 465
29 113 228 113
a,b
Values within a column with different superscripts are significantly different (P,0.05).
little, depending on which RHQT trait was included in the model. A lower RR was observed when horses were tested between 1984 and 1986, compared with horses tested between 1973 and 1979. Again, only a small variation in RR for test-year group was revealed by the analyses including different RHQT traits. Table 3 shows the RR for conformation and its sub-trait correctness of legs. Both traits showed significant effects (P,0.05) on longevity. For conformation, RR decreased from 1.20 to 0.64 between the worst and best scored horses. This means that horses with the lowest scores were twice as likely to be culled as the best scored horses. A clear trend in RR was observed, from low to high conformation score. The highest best scored horses had a significantly lower risk of being culled than the average horse. The same trend could be seen for the sub-trait legs. Horses that scored seven or more had greater chance of surviving than horses scoring six or lower. Results for the health traits are shown in Table 4.
Orthopaedic status significantly (P,0.001) affected the RR, whereas the effects of medical status were non-significant. Horses with an orthopaedic score of nine or ten had about half the likelihood of being culled than horses scoring below 6. Jumping ability showed a curvilinear trend in RR. Horses with a score of five or less or a score above eight had less chance of surviving than horses with average scores (Table 5). Jumping ability had a significant effect (P,0.05) on longevity, whereas gaits under rider did not. However, results are presented because of their value for this particular study. Table 6 shows RR for walk, trot and canter, and for the gaits as an average of each horse. None of them, except canter, showed significant differences in RR explained by different scores. However, the tendency was that horses gaining higher average scores for their gaits survived somewhat longer than horses with lower average scores. In Table 7, the overall classification scores as talents for dressage and jumping are related to RR.
Table 4 Risk ratios (RR) for orthopaedic and medical status Orthopaedic status (P,0.001)
Medical status (n.s.)
Class
RR
Number of horses
Number of dead horses
#5 6 7 8 $9
1.49 a 1.22 ab 1.17 ab 1.00 b 0.76 c
81 151 342 590 651
29 50 104 157 148
a,b,c
Class
RR
Number of horses
Number of dead horses
#7 8 9 10
1.01 a 1.11 a 1.00 a 0.97 a
134 482 829 350
37 130 216 90
Values within a column with different superscripts are significantly different (P,0.05).
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
102 Table 5 Risk ratios (RR) for jumping
5. Discussion The results showed that sex had a significant effect on longevity. The rather large difference in risk ratio between females and males is explained by the fact that females were often used not only for riding and competition but also for reproduction, which prolonged their lives, as previously shown by Wallin et al. (2000). The same study showed that there was a time trend toward an increased length of life for both sexes among horses that participated in the RHQT. Length of life in that study used PROC LIFEREG in SAS (1989) for the analyses. The same time trend was seen when test-year group was included in the model in this study, using the Survival Kit V3.1 of ¨ Ducrocq and Solkner (1999). Both conformation and legs (included in conformation) showed lower risk ratios for horses with good and excellent scores. Horses with favourable
Jumping (P,0.05) Class
RR
Number of horses
Number of dead horses
#5 6 7 8 $9
1.36 a 0.91 b 1.00 b 0.97 b 1.16 ab
299 403 488 390 188
100 94 129 99 58
a,b Values within a column with different superscripts are significantly different (P,0.05).
Jumping quartile had quite a large effect (P,0.01) on longevity, whereas dressage quartile showed a smaller effect (P,0.05). Horses in group IV in both jumping and dressage quartiles were most likely to be culled.
Table 6 Risk ratios (RR) for walk, trot, canter and gaits (average) Class
Walk (n.s.)
#5 6 7 8 $9 a,b,c,d
Trot (n.s.)
Canter (n.s.)
Gaits (average) (n.s.)
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
0.95 a 0.95 a 1.00 a 0.74 a 1.13 a
369 578 527 251 57
101 147 149 58 20
1.08 a 1.05 a 1.00 a 0.84 a 1.09 a
376 593 491 252 70
101 155 135 63 21
1.19 ad 1.14 ad 1.00 a 1.33 bd 0.81 ac
233 587 592 281 89
62 152 154 86 21
1.17 a 1.04 a 1.00 a 0.98 a
256 632 629 265
73 158 173 71
Values within a column with different superscripts are significantly different (P,0.05).
Table 7 Risk ratios (RR) for dressage and jumping quartiles Quartile
I II III IV a,b
Dressage quartile (P,0.05)
Jumping quartile (P,0.01)
RR
Number of horses
Number of dead horses
RR
Number of horses
Number of dead horses
0.98 a 1.22 ab 1.00 a 1.35 b
495 462 427 404
124 130 109 118
0.97 a 0.99 a 1.00 a 1.43 b
492 453 445 398
126 110 115 130
Values within a column with different superscripts are significantly different (P,0.05).
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
conformation have been assumed to have a better chance of staying sound and be able to perform over a long period of time. An unfavourable conformation, especially incorrect extremities, does not make the horse unsound itself, but could predispose the horse to disease later in life. The results of this study support that hypothesis. Lately, in several studies, survival analysis was used when studying the relationships between several conformation traits and longevity. Both in dairy cattle and in pigs, significant relationships between some conformation traits and longevity have been ¨ shown (Solkner and Petschina, 1999; Larroque and Ducrocq, 1999; Schneider et al., 1999; Cederberg Ringmar, 1999). There are only a few reports on the relationship between conformation and soundness in horses. Nevertheless, Magnusson and Thavfelin (1985) found that 11% of the variation in soundness was explained by conformation traits for standard-bred trotters. However, several studies indicate that conformation does have an impact on performance. Langlois et al. (1978), Langlois (1979) and ¨ et al. (1990) found that conformation Holmstrom differed significantly between poor and good show ´ jumpers as well as between elite dressage horses / show jumpers and ‘ordinary’ horses. Both studies showed that successfully performing horses need a specific conformation to be able to stay in shape and be able to perform successfully. This study showed that there is a highly significant relationship between orthopaedic status scores and length of life. A score of six or less in orthopaedic status was usually awarded to horses with rather critical remarks on their legs and on joint soundness. Horses having reactions to flexion tests were generally scored six or less (L. Roepstorff pers. comm., 2000). The orthopaedic examination focuses on leg and joint problems, and one should bear in mind that leg problems are the reason for culling more than half of Warmblood horses (Wallin et al., 2000). These results are of great interest as they indicate a high predictive value for longevity already when orthopaedic status scores are awarded when the horses are 4 years old and have undergone about 1 year of training. Therefore, the possibility of using these orthopaedic examinations for regular health
103
recording in young horses for advisory purposes as a means of preventing horses from becoming diseased should be widely adopted. ¨ and Philipsson (1993) found that a Holmstrom few conformational measures were significantly correlated with orthopaedic status among 195 RHQT 4-year-old horses, which could explain how both conformation and orthopaedic status influenced longevity. Low scores (#5) as well as high scores ($9) in jumping gave higher risk ratios. A possible interpretation of that result is that low-scoring horses had a number of problems, including mental inability, leading to a higher culling rate than among average jumpers. On the other hand, the highly scored horses were used largely in sports at a high level (Ohlsson and Philipsson, 1992) and thus were used quite extensively. Therefore, such use may put greater strain on the horse and consequently lead to a shorter life. Ricard and Fournet-Hanocq (1997) used ‘number of years in competition’ for show jumpers as a measure of longevity. Their results showed that the relative culling rate becomes smaller for horses that started to compete at the age of four years, compared with horses starting their competition career at the age of 5 or 6 years. The classification into different quartiles for dressage and jumping influenced longevity, as horses in class IV of both disciplines were less likely to survive. These quartile scores include the health scores as previously shown. Ohlsson and Philipsson (1992) could also show that none of the horses in quartile IV had competition results at the most advanced level in either dressage or jumping. Medical health had no significant effect on longevity. This could be explained both by its lower variance and by its complexity, involving different independent health parameters. Neither gaits judged under rider nor overall gait scores influenced longevity significantly, but the overall gait score was favourable, though also in this case one may assume that highly scored horses are used extensively for sports purposes. Gaits under rider have rather high ´ genetic correlation with conformation (Arnason, 1993), as both walk and trot showed at hand are included in the conformation score. Therefore, it could be assumed that gaits as well as conformation
104
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105
have an effect on longevity, but this could not be shown. This study emphasizes the value of judging different traits, especially health traits, early in horse’s life. The relationship between orthopaedic scores and longevity showed that orthopaedic scores could be used to predict longevity. When the RHQT was introduced in Sweden, it was emphasized that it should include assessments of health as well as of conformation and performance. Performance results from the RHQT have been used for a long time to estimate breeding values for Warmblood horses in ´ Sweden (Arnason, 1993). Health results from the RHQT have also been utilized for various purposes, e.g. for health examination in relation to sales and insurance of horses. For voluntary health recording of 4-year-old horses, the RHQT system has proved its worth. There is definitely a need for systematic recording of age of culling or death, for further investigations and monitoring of longevity. An extended study on the relationship between RHQT traits and longevity where the whole scoring scale could be used and hopefully reveal risk ratios for all scores would be desirable. Questionnaires are quite useful, but have the disadvantage of being time-consuming. The breeding associations would be appropriate places for registering survival data, i.e. age at culling or death and also cause of death. The practise in Sweden is to register all new-born foals, a procedure which has high acceptance among horse owners, and the question is whether registering deaths would be similarly accepted.
6. Conclusions The results of this study confirmed that there are significant phenotypic relationships between many of the RHQT traits judged early in horses’ lives and longevity on the one hand, and the possibility of using them as predictors of survival and for preventive medicine purposes on the other. Conformation, legs (included in conformation), orthopaedic status, jumping ability, and dressage and jumping quartiles were all traits associated with length of life. Orthopaedic health had the greatest influence on longevity, thus showing the importance of including health
recording early in horses’ lives. Conformation and performance, as well as health, must be of interest when selecting young horses for sports purposes.
7. Personal communication Vet. Med. Dr. Lars Roepstorff. Department of Equine Studies, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden.
Acknowledgements This work received financial support from the AGRIA Insurance Company, which is gratefully acknowledged. The Swedish Horse Board is gratefully acknowledged for providing materials for the study. We also thank individual horse owners who answered the questionnaires and made it possible for us to undertake this study.
References ´ Arnason, Th., 1987. Contribution of various factors to genetic evaluations of stallions. Livest. Prod. Sci. 16, 407–419. ´ Arnason, Th., 1993. Rapport om genetiska analyser av kvalitet¨ ˚ ¨ sbedomningar av svenska fyraariga ridhastar 1973–1993. (Report on genetic analysis of Riding Horse Quality Test data for ˚ Swedish 4-year-old horses). IHBC AB Knubbo, Morgongava, Sweden, 25 pp. Cederberg Ringmar, E., 1999. Environmental and Genetic Influence on Sow Longevity. Lic. Thesis, SLU, Uppsala. Ducrocq, V., 1999. Extension of survival analysis models to discrete measures of longevity. INTERBULL Bull. 21, 41–47. ¨ Ducrocq, V., Solkner, J., 1999. The Survival Kit-V3.1. http: / / www.boku.ac.at / nuwi / popgen Essl, A., 1998. Longevity in dairy cattle breeding: a review. Livest. Prod. Sci. 57, 79–89. Foran, M.K., Cromie, A.R., Reilly, M.P., Kelleher, D.L., Brophy, P.O., 1994. Analysis of show jumping data in the Irish sport horse population. 45th Ann. Meet. EAAP, Edinburgh, Scotland, 5–7 September, 4 pp. ¨ M., Magnusson, L.-E., Philipsson, J., 1990. Variation Holmstrom, in conformation of Swedish Warmblood horses and conforma´ tional characteristics of elite horses. Equine Vet. J. 22 (3), 186–193. ¨ Holmstrom, M., Philipsson, J., 1993. Relationship between conformation, performance and health of 4-year-old Swedish riding horses. Livest. Prod. Sci. 33, 293–312.
L. Wallin et al. / Livestock Production Science 68 (2001) 97 – 105 Langlois, B., Froidevaux, J., Lamarche, L., Legault, C., Legault, P., Tassencourt, L.O., Theret, H., 1978. Analyse des liaisons entre la morphologie et l’aptitude au galop au trot au saut ´ ´ Sel. ´ Anim. 10, 443– d’obstacles chez le cheval. Ann. Genet. 474. Langlois, B., 1979. French results on the analysis of relationship between morphology and gallop, trot and jumping abilities in horses. 30th Ann. Meet. EAAP, Harrogate, UK, 23–26 July, 8 pp. Larroque, H., Ducrocq, V., 1999. Phenotypic relationship between type and longevity in the Holstein breed. INTERBULL Bull. 21, 96–103. Magnusson, L.-E., Thavfelin, B., 1985. Studies on conformation and related traits of standard-bred trotters in Sweden. IV. Relationship between the conformation and soundness of 4year-old Standard-bred trotters. Thesis, SLU, Skara. ISBN 91-576-2315-5. Ohlsson, L., Philipsson, J., 1992. Relationship between field test for Warmblood horses as 4-year-olds and later competition performance. 43rd Ann. Meet. EAAP, Madrid, Spain, 14–17 September, 7 pp.
105
Prentice, R., Gloeckler, L., 1978. Regression analysis of grouped survival data with application to breast cancer data. Biometrics 34, 57–67. Ricard, A., Fournet-Hanocq, F., 1997. Analysis of factors affecting length of competitive life of jumping horses. Genet. Sel. Evol. 29, 251–267. SAS Institute Inc., 1989. SAS Users Guide: Statistics, 1982 Edition. SAS Institute Inc., Cary, NC, p. 584. Schneider, M.D.P., Monardes, H.G., Cue, R.I., 1999. Effects of type traits on functional herd life in Holstein cows. INTERBULL Bull. 21, 111–116. ¨ Solkner, J., Petschina, R., 1999. Relationship between type traits and longevity in Austrian Simmental cattle. INTERBULL Bull. 21, 91–95. Strandberg, E., 1997. Breeding strategies to improve longevity. 48th Ann. Meet. EAAP, Vienna, Austria, 25–28 August, 12 pp. Wallin, L., Strandberg, E., Philipsson, J., Dalin, G., 2000. Estimates of longevity and causes of culling and death in Swedish warmblood and coldblood horses. Livest. Prod. Sci. 63 (3), 275–289.