This LaTeX document is available as postscript or asAdobe PDF.
Background Information
Reproductive performance of dairy cows includes many different aspects. Reproductive performance can change with age of the cow, often depending on previous performance. Many countries have genetic evaluation systems in place for one or more measures of female reproductive ability. Some of the traits that have been studied have been
Determining which trait to analyze is a problem, but usually the trait with the highest heritability is chosen. However, reproductive performance is really the entire array of traits given above as a complete picture. Also, these traits are observed around each pregnancy so that a cow may have repeated measurements as she ages. By considering all of these traits at one time, in a multiple trait model, then perhaps a better evaluation of reproductive ability may be achieved with an overall greater heritability. Also, with the data collection systems that currently exist in Canada, cows may be observed for some of the above traits, but not necessarily all of them. Insemination data is now becoming more readily available in Canada and is provided to CDN. An assessment of its quality and usefulness needs to be made.
The long term objectives, and their short term sub-objectives are as follows:
Literature Review
Work by Westell et al. (1982) on Canadian data showed that 25% of cow disposal reasons were attributed to reproductive failure. Canada has not had a national recording scheme aimed at accumulating information on reproductive traits of cows until the last couple of years with the new milk recording programs. Insemination data are now being accumulated and the time is appropriate for developing a genetic evaluation system for reproductive performance of cows. Insemination data generally has problems with accurate dates, identification of the cow or service sire, and a lack of information if the herd owner uses a bull rather than artificial insemination to impregnate problem cows. Thus, the last insemination record for a cow may not be the one that caused conception in the cow. Also, cows already pregnant from a previous insemination might be inseminated again. Number of inseminations, therefore, may not be a good measure of fertility for a cow.
Herd owners may deliberately delay breeding cows after calving and therefore, number of days from calving to first insemination may not be a good fertility measure. Herd owners may use hormones to synchronize cows for breeding on one particular day as a management labour-saving strategy. Thus, intervals between calving and first insemination or between first and last insemination may not be useful measures of reproductive ability. Measurements of ovarian activity through blood or milk tests for progesterone or other hormones are not very practical in large herds even though they might be very useful indicators of fertility.
Work in Guelph by Raheja et al. (1989a,b,c) looked at heifer fertility, and in particular age at first breeding, finding an estimate of heritability of 0.12, while the estimate for number of services was only 0.04. The genetic correlation of heifer fertility traits with production ranged from -0.10 to 0.06 indicating a slightly antagonistic relationship. Estimates of heritabilities for the same traits expressed on a cow basis were between 0.03 and 0.06, and fertility traits had low genetic correlations between lactations. Lastly, they found that bulls with high non-return rates had daughters that were younger at first breeding and that required fewer services to conceive. This work did not result in a genetic evaluation system for fertility traits. Ironically, in many countries information on heifer breedings is usually ignored, and this stage of life of a cow may have the highest heritability for fertility measures.
Canada has had a genetic evaluation system for calving ease for many years, but information about stillbirths and abortions have not been analyzed. These are also important traits associated with female reproduction. Gestation length is a good trait to have, but accurate determination of conception date is not always possible, and usually a standard 280 days is assumed.
Trait Definitions
Many of the early studies on cow reproduction had only calving dates from which calving intervals or days open could be computed assuming a standard gestation length (Jansen, 1986). Sewalem et al. (2002) have found that calving intervals in Canadian dairy breeds have been lengthening by about 16 days over the last ten years while the dry period has decreased about 4 days over the same period.
The availability of insemination data has allowed the calculation of intervals between calving and each insemination as well as counting the number of inseminations. Kadarmideen and Coffey (2001), however, questioned the quality of insemination data and discussed the biases that could exist. Age at first insemination, age at conception, and the interval from calving to first service in each lactation have been important traits in several studies.
Some experimental studies have looked at days to first observed heat (not usually available in field data), number of heat cycles, progesterone levels in milk (interval from calving to postpartum ovulation), and days from calving to the beginning of luteal activity. The heritabilities of some of these measures have been in the range from 0.10 to 0.20, but the collection of such information is not practical in commercial herds.
A stillbirth (or calf mortality) is defined as a calf that dies within 24 hours of birth after a normal gestation period. Abortions are the incomplete gestation of a foetus. These may be due to the sire of the calf or due to the cow, so that both factors must be considered as with calving ease. These traits have been found to differ between lactations. Bulls have ranged from 2 to 27% stillbirths (Philipsson et al., 1997) in Sweden. Some stillbirths may be due to a new chromosomal translocation in the Swedish Red and White breed (SRB).
Non-return rates of service sires have been evaluated for several years (Schaeffer, 1986), but these evaluations have been phenotypic rather than genetic. Dilution of semen before freezing, age of the sire when collected, and technician all have large effects on fertility. The number of sperm per dose is often altered for popular bulls in order to have more doses available and to maintain a certain fertility level. Thus, the fertility of a service sire can change over time due to the manner in which the semen was processed.
Factors Affecting Traits
The heritabilities of most reproductive traits were generally below 0.10, and many were below 0.05. Studies consulted were Jansen (1986), Raheja et al. (1989a,b,c), Pryce et al. (1997), Dematawewa and Berger (1998), Veerkamp et al. (2001), Hayes et al. (1992), and Weller (1989). Even though heritabilities were quite small the additive genetic variation for these traits was deemed to be sufficient such that selection for fertility could be effective.
Plaizier et al. (1998) found that larger herds tended to become more mechanized which resulted in less contact time with the cows and consequently reduced breeding efficiency for the herd. Months during the year, parity number, herd, service sire, sire of the cow, and age at calving were all factors that affected fertility traits in different studies. Williamson et al. (1980) found that short intervals between calving to first service (i.e. less than 40 days) had adverse effects on first service conception rates, number of services per conception, and calving interval. Pryce et al. (1997) found a genetic correlation between calving interval with days to first service of 0.50, and 0.61 between calving interval and conception rate. Body condition score has been found to be related to reproductive performance (Pryce et al., 2001; Pryce et al., 2000; and Veerkamp et al., 2001). Thin and more angular cows tended to have longer calving intervals. Dematawewa and Berger (1998) found that days open and number of services were negatively correlated with survival, and positively correlated with production, similar to Veerkamp et al.(2001). Miller et al. (2001) found no effect of somatic cell scores on fertility.
Boichard and Manfredi (1992) found male and female fertility to be quasi-independent in the French Holstein population. Raheja et al. (1989c), however, found that bulls with high non-return rates had daughters that were younger at first service and which needed fewer services to conceive.
For calving ease and stillbirths, Weller and Gianola (1989) included sex of calf, calving age, calving month, sire of cow, sire of calf, and herd-year-seasons in their threshold model. Heritability was higher for male calves than for female calves. Harbers et al. (2000) found a higher percentage of stillbirths in first parity cows (11.4%) compared to older cows (5.3%) in the Netherlands. Half of the stillborn calves were born unassisted (without any calving difficulty), so that dystocia and stillbirths are two separate traits. There were fewer stillbirths between July to September than in other months. Breed differences were important for stillbirths as well. Stillbirths could affect the number of services after calving and the number of days to reach peak yield. Lastly, Thaller (1997) suggested the possibility of strong non-additive genetic effects for fertility, but there have been no esimates in the literature for fertility traits.
Economic Values
Miesenberger et al. (1997) found the following economic values relative to conception rate to be 3 for longevity, 2 for mastitis resistance, 3.5 for fat yield, 4.0 for protein yield, 1.5 for daily gain and dressing percentage, and 0.5 for persistency and carcass merit in Austrian Simmental cattle. Thus, protein yields were deemed four times more valuable than conception rate. In a study of UK herds, Stott et al. (1999) found a cost of £7.44 for every day of delayed re-breeding. Further search of the literature for economic studies is needed, but the search will depend on the traits that will be analyzed in this project.
Methods and Proposed Approach
Concepts
Reproductive events occur during the lifetime of a dairy cow. After birth is the first estrus which occurs at a particular age. This is followed by the first insemination, the second, and so on until conception occurs. Some cows may abort the calf early, but most will have a normal gestation of around 280 days. At calving there could be dystocia or stillbirth. After calving the cow begins milk production and the level of this production could affect subsequent fertility. After calving the cow will begin to have estrus again, inseminations will occur, and subsequently conception occurs. This cycle is repeated after each calving. The literature indicated that heifer fertility traits had higher heritabilities than cows, and the importance of having a heifer fertility trait in the analyses seems vital, even though the relationship to later fertility measures may be weak.
A random regression model could be applied, in multiple trait fashion, to the fertility traits of cows. The time variable would be parity number. The traits, for each parity, would be age at conception, number of services (or number of estrus cycles) to conception, days open (or days from first service to conception), age at calving, calving ease score, stillbirth (0-1 trait), and abortion (0-1). Both direct and maternal genetic effects would be modeled, but with a zero correlation between direct and maternal effects. Thus, there would be 6 or 7 traits and the random regressions on parity number would allow the genetic ranking of animals to change with parity number (i.e. let each parity be a different trait). Genetic correlations between traits across parities would vary. The additive genetic relationship matrix would include inbreeding coefficients. The possibility exists to combine all breeds into one model with breed specific fixed curves within year- seasons of birth.
An alternative model would be a simple multiple trait model with traits nested within parities, and perhaps only 3 parities in the analysis. In either model, all cows would be required to have first lactation (i.e. heifer) breeding information to account for culling. With seven traits per parity, this would be a 21 trait model. The advantage of the random regression model would be that it is only a seven trait model and any number of parities could be included.
Genetic parameters would be estimated using the Bayesian approach and Gibbs sampling (Monte Carlo Markov Chains). Hierarchical modeling may accelerate the estimation of parameters. Random regression models and a multiple trait situation has not been reported previously. Two or three subsets of data will be used for the estimation process. All current data would be used to compute breeding values. Associations of trait EBVs with production and conformation traits, or with inbreeding coefficients is straightforward. Usual selection index procedures will be used to derive a fertility index, and to derive an economic value for the LPI.
This LaTeX document is available as postscript or asAdobe PDF.
Larry Schaeffer