- Email: [email protected]
Calving Interval and Survival Breeding Values as Measure of Cow Fertility in a Pasture-Based Production System with Seasonal Calving
J. Dairy Sci. 85:689–696 American Dairy Science Association, 2002.
Calving Interval and Survival Breeding Values as Measure of Cow Fertility in a Pasture-Based Production System with Seasonal Calving V. E. Olori,* T. H. E. Meuwissen,† and R. F. Veerkamp† *Irish Cattle Breeding Federation, Shinagh House, Bandon, Co. Cork, Ireland †Institute for Animal Science and Health, ID-Lelystad, P.O. Box 65, 8200 AB Lelystad, The Netherlands
ABSTRACT In a grass-based production system with seasonal calving, fertility is of major economic importance. A delay in conception due to poor fertility prolongs intercalving interval and causes a shift in calving pattern, which can lead to culling. Calving interval (CIV) information is readily available from milk records; analyzing it, however, presents a problem, as it is only available for cows that conceive and calve again. Calving interval should therefore be treated as a censored trait. In this study, survival to the next lactation (SUV) was analyzed jointly with CIV in a multivariate linear model to account for the selection in CIV data. Genetic parameters for first lactation calving interval were estimated with a sire model for Holstein Friesian cows in Ireland. SUV was preadjusted for production within herd-yearseason (HYS), while milk yield was included as a third trait in the analysis to account for the large effect it has on both traits. The residual covariance between CIV and SUV was fixed as 3 times the sire covariance within the model, as it was inestimable because of the structure of the data. Breeding values were estimated with various models to test the effect of culling and milk yield. Heritability was 0.04 ± 0.006 for CIV and 0.01 ± 0.003 for SUV, while the genetic correlation between them was −0.28 (±0.11). The genetic standard deviation was around 4% for SUV and 7 d for CIV. Sire predicted transmitting abilities for progeny tested bulls ranged between −5 and 3% for survival rate and between −4 and 8 d for calving interval. Differences between the best and worst bull varied with model. Including SUV and milk yield as traits in the model reduced the mean and variance of sire predicted transmitting abilities but increased the coefficient of variation by 30% compared with the univariate model. The current model is expected to account for most of the genetic variation in
Received September 25, 2001. Accepted November 5, 2001. Corresponding author: V. Olori; e-mail: [email protected].
fertility that is possible from calving dates and future extensions, such as the use of linear type trait or additional lactations for predicting survival, appear straightforward. These traits now form part of the national index for selecting dairy bulls in Ireland. (Key words: calving interval, survival score, fertility, seasonal calving) Abbreviation key: CIV = calving interval, HYS = herd-year-season, SUV = survival. INTRODUCTION Deterioration in cow fertility has become a major problem in dairy cattle production to the point that it is becoming increasingly apparent that high yielding cows are more difficult to get in calf. This negative association between production and fertility has been observed in several dairy breeds of cattle (Hermas et al., 1987; Hoekstra et al., 1994; Pryce and Veerkamp, 2001). The decline in reproductive efficiency of the Holstein Friesian cattle has been attributed to selection for increased production and or increase in the proportion of Holstein genes in the dairy cow population (Harris and Winkelman, 2000; Silvia, 1998). Results of recent studies in a grass-based production system indicated that cows of high genetic merit had a higher interval from calving to first service, poorer conception rate overall, poorer conception to first, second, and third services as well as higher number of services per conception compared to cows of medium genetic merit (Buckley et al., 2000a, 2000b; Snijders et al., 2001). There was no significant difference in milk yield, feed intake, and plasma glucose concentration between cows that conceive and those that did not conceive within the same genetic merit level (Snijders et al., 2001), which suggests that poor reproductive performance may not be a direct consequence of high milk production. An infertility rate of 21% for high genetic merit cows compared with 6% for the medium merit cows was reported (Snijders et al., 2001). The estimated margin after cost was about 0.02 US dollars per kilogram of milk higher
689
690
OLORI ET AL.
in medium-merit cows compared with the high genetic merit cows after considering loss due to poor fertility. Economic losses due to poor fertility are generally due to the cost of a prolonged calving interval, increased insemination cost, reduced returns from calves born, and forced replacement in the event of culling (Van Arendonk et al., 1989; Esselmont et al., 2001). Reduction in fertility and the attendant increase in the interval from calving to conception is a critical problem in seasonal calving herds, where calving interval has a high economic value because of the need for compact calving to maximize grass utilization during the grazing season. Increasing calving interval by 1 d costs the breeder about 1.8 US dollars in Ireland without accounting for the costs of higher culling due to poorer fertility (Veerkamp et al., 2001), and may be up to 4.5 US dollars, depending on the level of production and the extent of the delay (Esselmont et al., 2001). Measures of fertility generally have a low heritability, indicating a strong influence of environmental factors such as management decisions taken by the breeder. However, a review of several studies indicates that there is significant genetic variation between cows in the interval from calving to first breeding (Jansen, 1985), which is positively correlated with number of services per conception. The interval from calving to first breeding and number of services per conception both have positive correlation with number of days open, which is the variable component of calving interval. Calving interval is strongly correlated with days to first service (0.93) and moderately correlated with conception to first service in first lactation heifers (−0.56) as well as in cows, while measures of fertility in first lactation heifers are positively correlated with fertility in cows and have higher heritability (Pryce et al., 1997, 1998). Overall improvement in cow fertility can therefore be achieved by selecting on first-lactation heifer fertility based on calving interval. This will allow early selection but most importantly, the improvement of fertility by using calving dates available from milk records. Analyzing calving interval as a measure of fertility presents a problem because only animals that survive to the next lactation have a calving interval. Evaluation based on this trait alone will be biased because of preselection as culled animals with the worst fertility problems, will not be included in the analysis. Calving interval should therefore be treated as a censored trait and analyzed jointly with survival score to take into account the nonrandom scoring of calving interval. Furthermore, in a strict seasonal calving herd, poor fertility is expected to increase culling more than CIV, because a shift in calving pattern is not acceptable. Thus, the survival of a cow from one lactation to the next is also Journal of Dairy Science Vol. 85, No. 3, 2002
an important indicator of fertility. For these reasons, a combined analysis of calving interval and survival is expected to account for most of the genetic variation in fertility that is possible from calving dates. The objective of this study was to estimate genetic parameters and breeding values for first-lactation calving interval and survival rate for Holstein Friesian cattle in a grass-based and seasonal calving production system in the Republic of Ireland as a prelude to their inclusion in the national dairy breeding goal. MATERIALS AND METHODS Data The data used in this study were created from available milk records of about 400,000 Holstein Friesian cows commencing their first lactation between 1988 and 1998 in the Republic of Ireland. This included test-day and lactation milk yields as well as birth, calving, and lactation end dates. The file also contained information on lactation end reason, which indicated cows that were sold or died. It also indicated which cows were suckling, and those still in milk or dry. Records from about 200,000 lactations with no known end reason, which could not be conclusively identified as culled in the first lactation or survived to the second lactation were deleted. These include records from herds that stopped milk recording during the period or from cows that were sold to nonrecording herds. Usually, a cow sold to another milk recording herd will have a lactation end reason indicating that it was sold and its subsequent lactation will be found in another herd. A cow was assumed sold to a nonmilk recording herd when the end reason was ‘sold,’ but the next lactation record could not be traced. Such cows may also have been slaughtered thus it was safer to eliminate them from the analysis. Another 75,000 records from herds with less than 50 cows were also deleted to minimize the number of herd-year-season (HYS) classes. Cows were allocated to HYS subclasses based on the month of calving. Three seasons were defined as follows; January–April, May–August, and September–December. Each HYS class had a minimum of five records, while sire families were constrained to at least 10 cows. This led to a loss of another 35,000 records. The final data comprised of 89,101 cows sired by 1400 bulls in 21,139 HYS classes. About 71% of these cows survived to have a second lactation, while the remaining 29% were assumed culled in the first lactation. A standardized 305-d milk yield of each cow was calculated from available test-day records by the method of interpolation with standard lactation curves (Olori and Galesloot, 2000). For all cows with a subsequent calving date, calving interval (CIV in days) was calcu-
691
GENETIC EVALUATION OF COW FERTILITY Table 1. Mean and standard deviation of heifer lactation calving interval, milk, fat, and protein yields. Trait
Number of records
Mean
Standard deviation
Coefficient of variation
Calving interval, days Milk yield, kg Fat yield, kg Protein yield, kg
63,613 89,100 89,100 89,100
398 5475 204 180
75.9 1357 54 44
19 25 26 24
lated (Table 1). These cows were scored 1 for survival (SUV). CIV was set to null (missing) for cows that were culled in the first lactation, and these were scored 0 for survival. To account for voluntary culling, survival score was preadjusted for milk yield within HYS groups to take into account seasonal variation in the level of production within herd. This was achieved by obtaining the unadjusted residual score from a regression on milk yield deviated from the HYS mean. CIV was restricted to between 260 to 600 d (inclusive) to eliminate outliers that may be due to early abortion or cows kept open for repeated flushing to produce embryos for transfer. In a seasonal calving system, excessively long calving intervals may also result when a breeder opts to skip a year by leaving the cow open rather than change the calving pattern when a cow does not get in calf during the breeding season. About 2% of the records in this category had a survival score of 1, with CIV set to missing. No information was available in the current milk records to confirm these assumptions about the reason for the short or prolonged calving intervals observed in this data. Estimation of Variance Components Variance components were estimated from a joint analysis of calving interval with survival and milk yield using a multivariate sire model. Milk yield was included as a third trait in the analysis because of its positive association with calving interval and its effect on culling (not already accounted for by preadjustment of the data). It also accounts for a potential farmers’ attitude towards fertility and survival relative to production in
a seasonal production system. For instance, a breeder may not hesitate to cull a poor yielding cow that is difficult to get in calf. He may, however, decide to keep trying for a longer period to get a high yielding cow in calf rather than cull. There is also a higher probability of allowing a high yielding cow to survive in the herd albeit with a longer calving interval where poor cows will not. The inclusion of milk yield as a third trait accounts for the selection on yield, not accounted for by the preadjustment of SUV for milk yield, that is expected to account for the direct effects of yield on voluntary culling. The Statistical Model Variance components were estimated by REML with a sire model using ASREML (Gilmour et al., 1998). Relationships between sires (through their sires and dams) were included for two generations. Because only animals that survive (survival score = 1) have a calving interval value, the environmental covariance between survival rate and calving interval is undefined and inestimable (Janns and Bolder, 2000). However, the residual covariance is influenced by the genetic covariance between dams and mendelian sampling, and will normally be estimated as ³⁄₄*σa(i,j) + σe(i,j). Since the second term (σe(i,j)) cannot be estimated, the residual covariance becomes ³⁄₄*σa(i,j). The estimate of the sire covariance is ¹⁄₄*σa(i,j), hence, in this study, the residual covariance between survival score and calving interval was estimated as 3 times the sire covariance. Total variance was calculated as the sum of the residual and sire variances (σ2e + σ2s), while genetic variance was 4 times the
Table 2. Genetic and residual (co)variance components for yield, fertility and survival traits. Genetic (co)variances
Survival (%) × 10−02
Calving interval (days)
Milk yield (kg)
Survival Calving interval Milk yield Genetic standard deviation
0.002 −0.081 4.656 0.045
53.12 15,332 7.29
275,520 525
Residual (co)variances Survival Calving interval Milk yield
0.126 −0.061 −0.115
1418 2974
417,500 Journal of Dairy Science Vol. 85, No. 3, 2002
692
OLORI ET AL.
sire variance and heritability was calculated as 4σ2s/(σ2e + σ2s) (see Falconer and Mackay, 1997). Fixed effects in the model included management group (HYS), calving month, calving age, and the proportion of Holstein genes. The last two were fitted as covariate to the second degree. The linear model applied can thus be expressed as; 2 2 Yijkl = Ui + a1iX1i + a2iX1i b1iX2i + b2iX2i + Mij + HYSik + Sil + eijkl
where Yijkl = calving interval, survival between lactation 1 and 2 and/or milk yield, Ui = overall mean for the ith trait, a1i and a2i = the linear and quadratic regression coefficients (effects) of proportion of Holstein genes, b1i and b2i = the linear and quadratic regression coefficients (effects) of calving age, X1i and X2i = proportion of Holstein genes and calving age (linear and quadratic covariates), respectively, Mij = fixed effect of the jth month of calving on trait I, HYSik = fixed effect of kth herd year season class on trait I, Sil = the random genetic effect of the sire l of cows with records on trait i, and eijkl = random error term.
Figure 1. Distribution of first lactation calving interval for Holstein Friesian cows in Ireland.
for SUV, were expressed in percentage to correspond with the unit used in the derivation of the economic weight, while proofs for CIV were expressed in days. Further analysis, comparison, and conclusions were based on PTAs from about 650 bulls with at least five daughters in Ireland included in the breeding value estimation and having a minimum reliability of 30% for CIV in the single-trait analysis. The analysis of variance procedure (PROC ANOVA) in SAS (SAS, 1990) was used to analyze breeding values for each trait from the single, two- and three-trait analyses. Difference between models was based on an Fratio test from the one-way analysis of variance. RESULTS
Estimation of Breeding Values
Phenotypic Means
Breeding values were estimated by BLUP with a sire model using PEST (Groeneveld et al., 1990). The data used for this analysis was similar to the above, except that there was no limitation on the size of sire families. This was to allow as many bulls as possible to improve linkage in the relationship matrix, and to provide breeding values for all bulls in use. To see the effects of milk yield in the analysis as well as the effect of a combined analysis of SUV and CIV, single- versus multiple-trait estimates of breeding value were performed. Breeding values for SUV and CIV were first estimated separately, then jointly in a bivariate analysis and finally in a threetrait analysis with 305-d milk yield as the third trait. The genetic parameters used for breeding value estimation with the single-, two-, and three-trait models, were the same ones, estimated with a three-trait model as discussed in the parameters estimation section above. Predicted transmitting abilities were derived as half of the corresponding breeding value for all traits. Proofs
The average heifer yields and standard deviation for cows with records in the analysis were 5475 ± 1357, 204 ± 54, and 180 ± 44 kg for milk, fat, and protein yields, respectively. About 69% of these cows had calving intervals ranging from 260 to 600 d with a mean of 389 ± 55 d, while 66% of the cows had calving intervals ranging from 309 to 550 d. The modal calving interval was 363 d, which is close to the target of one calving per year desired in Ireland. First, second, and third quartile values were 352, 378, and 419 d, respectively. Figure 1 shows the frequency distribution of calving interval in the final data set before restriction to the range of 260 to 600 d.
Journal of Dairy Science Vol. 85, No. 3, 2002
Variance Components Table 2 shows the (co)variance components for survival rate, calving interval, and milk yield from the trivariate analysis. Genetic standard deviation was
693
GENETIC EVALUATION OF COW FERTILITY Table 3. Heritability (on diagonal), genetic (below), and phenotypic (above) correlations between survival, calving interval and milk yield with standard errors in parentheses.
Survival Calving interval Milk yield
Survival
Calving interval
Milk yield
0.01 (0.003) −0.28 (0.11) 0.22 (0.09)
−0.001 (0.003) 0.04 (0.006) 0.40 (0.07)
0.004 (0.003) 0.128 (0.004) 0.56 (0.03)
4.5% and 7.3 d for survival rate and calving interval, respectively. The corresponding heritabilities were 0.01 ± 0.003 and 0.04 ± 0.006, respectively (Table 3). The heritability of milk yield in the three-trait analyses was higher than usual at 0.56 ± 0.03. There was a negative genetic correlation between calving interval and survival (−0.28), while the correlation between milk yield and calving interval was positive (0.40). Milk yield was also positively correlated with survival (0.22). Predicted Transmitting Abilities Table 4 contains a summary of sire PTA for CIV and SUV from single- and multiple-trait analyses. The standard deviation of SUV PTA increased from 0.7% in the single-trait analysis to 1.1% in the three-trait analysis. There was no similar increase in the mean PTA, suggesting lack of systematic bias. The best bull had about 5% more daughters surviving, relative to the worst bull, between lactation one and two from the single-trait analysis. This increased to about 8% with the inclusion of CIV and milk yield in a three-trait analysis. The standard deviation of CIV sire PTA dropped from 1.6 d in the single-trait analysis to 1.38 d in the three-trait analysis; however, there was also a decrease in mean PTA, which resulted in an increase in coefficient of variation from 0.83 to 1.16. Figure 2 shows the distribution of sire PTA for CIV from the single, two- and three-trait analyses, while Figure 3 shows a similar distribution for SUV. The distribution of both CIV and SUV proofs from the single and two-trait models were relatively similar. There was a general shift of the curve to the left for CIV with
inclusion of milk yield suggesting that, analyzing calving interval alone, or along with survival rate without correcting for milk yield may result in the overestimation of breeding values for calving interval. This is because of the strong positive genetic and environmental effects of milk production on calving interval. The correlation between proofs from the single and two-trait analyses was 0.99 for CIV and 0.98 for SUV, and between the three-trait analysis and the single-trait analysis 0.91 and 0.92, respectively. Genetic trend for the CIV was similar for all the models. Figure 4 shows the trend of calving interval and survival by the birth year of the bulls based on sire PTA from the three-trait model. The trend indicates a steady increase in calving interval at the rate of 0.14 d per year, and a decline in survival rate at the rate of 0.05% per year since 1984. The %Holstein genes in the bull population was about 70%. Their daughters averaged 45%. Holstein gene percentage had a quadratic effect on SUV (R2 = 0.75), with significant drop in survival for bulls with 62% or more Holstein genes. For CIV, there was only a small linear (b = 0.049, R2 = 0.40) effect. The impact of %Holstein genes was more on CIV and less on SUV before inclusion of %Holstein genes in the model. DISCUSSION Cow fertility has become a major problem in dairy cattle production. Poor fertility delays conception, increases inter-calving interval, and causes a shift in calving pattern. In a seasonal calving production system, which relies on grass as a major source of feed, calving
Table 4. Summary statistics for calving interval and survival sire PTA from various models.1 Model
Mean
SD
Minimum
Maximum
Calving interval (days) Univariate (CIV) Bivariate (CIV-SUV) Trivariate (CIV-SUV-milk)
1.93 1.89 1.19
1.61 1.63 1.38
−3.36 −3.43 −3.71
8.10 8.15 6.55
Survival rate (%) Univariate (SUV) Bivariate (CIV-SUV) Trivariate (CIV-SUV-milk)
−0.27 −0.06 −0.14
0.73 0.78 1.11
−3.44 −3.34 −5.14
1.79 2.17 2.52
CIV = Calving interval, SUV = survival.
1
Journal of Dairy Science Vol. 85, No. 3, 2002
694
OLORI ET AL.
Figure 4. Genetic trend in calving interval and survival rate based on bulls with at least five daughters. Figure 2. Distribution of sire PTA for calving interval estimated from univariate (CIV1), bivariate CIV-SUV (CIV2), and trivariate CIV-SUV-Milk (CIV3) models for bulls with at least five daughters.
interval is an objective trait of economic importance because of the direct consequence of a shift in calving interval. Secondly, in the absence of data on direct measures of fertility, calving interval can be considered a good indicator of cow fertility because of the high correlation between CIV and several direct measures of fertility (Campos et al., 1994; Grosshans et. al, 1997; Pryce et al., 1997, 1998). The heritability of 4% and a genetic coefficient of variation of around 2% obtained in this study is within the range obtained in other studies (Janson, 1980, Hoekstra et al., 1994, Campos et al., 1994; Grosshans, et. al, 1997; Pryce et al., 1997, 1998). A genetic standard deviation of 7 d obtained for CIV in this study indicate significant genetic variation in the population suggesting that genetic improvement can be achieved through selection. Experience from the Scandinavian countries show that this is both feasible and beneficial (Phillipson, 1981; Lindhe and Phillipson, 2001).
Figure 3. Distribution of sire PTA for survival score estimated from univariate (SUV1), bivariate CIV-SUV (SUV2), and trivariate CIV-SUV-Milk (SUV3) models for bulls with at least five daughters. Journal of Dairy Science Vol. 85, No. 3, 2002
The main objective of this study was to investigate the feasibility of analyzing calving interval as a measure of cow fertility in routine genetic evaluation. The use in the past has not been contemplated perhaps because of the positive relationship with milk yield and the censored nature of calving interval. In this study, a multivariate linear model was used to analyze calving interval jointly with survival score and milk yield. This allows all cows (including those culled in the first lactation) to be included in the analysis thus avoiding the effect of preselection brought about by voluntary culling, which may be due among other things to poor fertility. The model was extended to include milk yield as a third trait because of the direct effect of production on voluntary culling. An increase in the coefficient of variation for sire PTA from the two- and three-trait models compared to the single-trait model suggests that a joint analysis of calving interval, survival score, and milk yield enable us to recover most of the variation in fertility that can be recovered from calving dates. Although only the interval and survival between first and second lactations have been studied here, the same principle is applicable to several lactations. Theoretically, nonlinear models seem to be more suitable for analyzing categorical data such as survival score and censored data such as calving interval. In one recent study, genetic parameters for fertility and health traits in cattle, estimated with a threshold model, where higher than estimates from a linear model (Kadarmideen et al., 2000). However, extension to an animal model for routine national genetic evaluation will be difficult to implement. Several studies have associated a decline in cow fertility with increase in milk yield (see Pryce and Veerkamp, 2001 for review) and percentage Holstein genes (Harris and Winkelman, 2000). Thus, a limitation in the analysis of calving interval and survival score, which in this study depends simply on reappearance in the sec-
695
GENETIC EVALUATION OF COW FERTILITY
ond lactation, is the potential for bias due to the influence of the breeder on the length of the intercalving interval. To avoid this, survival score was precorrected for yield within contemporary groups and 305-d yield was introduced as a third trait. Precorrection accounts for the phenotypic effect of milk yield on survival, hence, the phenotypic correlation between these traits was near zero (Table 3). However, there still exists, a significant genetic relationship between milk yield and survival, which explains the increase in the range of sire PTA for SUV from the three-trait model compared to the others. Our results indicate that these adjustments seem to eliminate potential bias, as indicated by the reduction in mean PTA, that may otherwise arise due to the farmers’ action dictated by the level of milk production. Examples of such action include delaying insemination of high yielding cows or production motivated selective culling of cows that fail to conceive during the breeding season. The correlation between sire PTA for CIV and milk yield was 0.51 in this study even with adjustment for percentage Holstein genes. This shows that the correction for Holstein percentage did not erode the positive genetic relationship between yield and calving interval. In this study, only the interval and survival between first and second lactation were analyzed. This keeps the model simple while still having the potential to improve overall cow fertility because of the strong correlation between first and later lactations (Hahn, 1969; Janson, 1980; Hansen et al., 1983; Pryce et al., 1998). Our heritability estimate was slightly higher than a recent estimate (0.02) from the UK (Kadarmideen et al., 2000). However, their estimate was obtained with a repeatability model including up to five lactation records per cow. Combining this with other direct measures of fertility and perhaps marker-assisted selection may speed up improvement. The genetic correlation between CIV and SUV indicate antagonistic relationship between calving interval and survival supporting the assertion that cows with prolonged calving interval are more likely to be culled in a seasonal calving herd scenario. We observed an unfavorable relationship between calving interval and milk yield similar to the findings in the UK (Kadarmideen et al., 2000). The difference in our correlation estimates and theirs (0.44 vs. 0.53) can also be attributed to the difference in data and model. The heritability of milk yield in this study was higher than those commonly reported in literature and from a recent UK study (Kadarmideen et al., 2000). However, this can be attributed to the data. Heritabilities from first-lactation records are usually higher than those estimated from multiple-lactation records. Also, in this study, it was based on a standardized 305-d
milk yield, even though the average lactation length in Ireland is about 280 d (ICBF, 2000). The high heritability may thus be due to a lower residual variance of standardized first-lactation records. One limitation of the current procedure is that cows can only be included in the analyses when they have a second lactation or when sufficient time has been given to ascertain that they will not have a second lactation. This delay means that young test bulls cannot be evaluated for these traits when they come off test, thus ruling out early selection decisions. A further delay is due to the time it takes to have sufficient daughters for a reliable proof because of the low heritability of the traits. A possible solution would be to include linear type traits that are significantly correlated with survival and calving interval in a joint analysis to facilitate earlier evaluation. The sire model approach also implies that proofs cannot be estimated directly for cows simultaneously with the bulls. Such will be possible with an animal model. The observed genetic variance may be currently limited because only first-lactation records are considered. A model that allows evaluation based on multiple lactations may increase variation in survival, as all animals that survive the first lactation may not have equal genetic potential to survive five lactations. It is suggested that further studies be put in place to investigate the possibility of an animal model with multiple lactation data. A multivariate analysis with linear predictors of calving interval should also be investigated as this will allow proofs to be calculated early for young test bulls as well as allow foreign bulls with no daughters in Ireland to be properly evaluated for these traits. It is suggested also that a good tracking system that promptly reports on cattle movement and lactation end reasons be put in place. This will allow us to distinguish between animals that were culled and those that have simply moved to nonmilk recording herds. CONCLUSION This study has shown how calving interval can be analyzed as a measure of fertility in dairy cattle without the bias due to culling and the effect of milk yield. This was achieved in a joint analysis of calving interval, precorrected survival score, and milk yield. This model allows the utilization of records from all milk-recorded cows, including those culled in the first lactation. It is possible that the use of genetic parameters obtained with a three-trait model could have affected our estimate of breeding values with a single- and two-trait model and hence our conclusions. However, the extent of this effect is not clear. Journal of Dairy Science Vol. 85, No. 3, 2002
696
OLORI ET AL.
ACKNOWLEDGMENTS We are grateful to B. Wickham and the Irish Cattle Breeding Federation (ICBF) for funding the ‘Dairy Breeding Objective Project’ of which this study was part. This study and V. Olori’s position in ICBF were partly funded with European Union structural funds. REFERENCES Buckley, F., P. Dillon, S. Crosse, F. Flynn, and M. Rath, 2000a. The performance of Holstein Friesian dairy cows of high and medium genetic merit for milk production on grass based feeding systems. Livest. Prod. Sci. 64:107–119. Buckley, F., P. Dillon, M. Rath, and R. F. Veerkamp. 2000b. The relationship between genetic merit for yield and live weight, condition score and energy balance of spring calving Holstein Friesian dairy cows on grass based systems of milk production. J. Dairy Sci. 83:1878–1886. Campos, M. S., C. J. Wilcox, C. M. Beceril, and A. Diz. 1994. Genetic parameters for yield and reproductive traits in Holstein and Jersey Cattle in Florida. J. Dairy Sci. 77:867–873. Darwash, A. O., G. E. Lamming, and J. A. Woolliams. 1997. The phenotypic association between the interval to postpartum ovulation and traditional measures of fertility in dairy cattle. Anim. Sci. 65:9–16. Esselmont, R. J., M. A. Kossaibati, and J. Allcock. 2001. Economics of fertility in dairy cows. Pages 19–29 in Fertility in the HighProducing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science, Occasional Publication No 26. Edinburgh, Scotland. Falconer, D. S., and T. F. C. Mackay, 1996. Introduction to Quantitative Genetics. 4th ed. Longman Group Ltd., Essex, England. Gilmour, A. R., B. R. Cullis, S. J. Welham, and R. Thompson. 2000. ASREML Reference Manual. NSW Agriculture, Orange Agric. Inst. Orange, Australia. Groeneveld, E., M. Kovac, and T. Wang. 1990. PEST, a general purpose BLUP package for multivariate prediction and estimation. Proc. 4th World Congr. Genet. Appl. Livest. Prod. Edinburgh, Scotland XIII:488–491. Grosshans, T., Z. Z. Xu, L. J. Burton, D. L. Johnson and K. L. Macmillan. 1997. Performance and genetic parameters of fertility in seasonal dairy cows in New Zealand. Livest. Prod. Sci. 51:41–51. Harris, B. L., and A. M. Winkelman. 2000. Influence of North American Holstein genetics on dairy cattle performance in New Zealand. Page 122 in Proc. 2000 Large Herds Australia Confr., Armidale, Australia. Hermas, S. A., C. W. Young, and J. W. Rust. 1987. Genetic relationships and additive genetic variation in productive and reproductive traits in Guernsey dairy cattle. J. Dairy Sci. 70:1252–1257. Hoekstra, J., A. W. van der Lugt, J. H. J. van der Werf, and W. Ouweltjes. 1994. Genetic and phenotypic parameters for milk production and reproductive performance traits in upgraded dairy cattle. Livest. Prod. Sci. 40:225–232.
Journal of Dairy Science Vol. 85, No. 3, 2002
Irish Cattle Breeding Federation (ICBF) 2000. Irish cattle breeding Statistics. 2000. Irish Cattle Breeding Society Limited, Shinagh House, Bandon. Co. Cork. Ireland. Jansen, J. 1985. Genetic aspects of fertility in dairy cattle based on analysis of AI data—a review with emphasis on areas for further research. Livest. Prod. Sci. 12:1–12. Janss, L. L., and N. M. Bolder. 2000. Heritabilities and genetic relationships between salmonella resistance traits in broilers. J. Anim. Sci. 78:2287–2291. Kadarmideen, H. N., R. Thompson, and G. Simm. 2001. Linear and threshold model genetic parameters for disease, fertility and milk production in dairy cattle. Anim. Sci. 71:411–419. Lindhe, B., and J. Philipsson. 2001. The Scandinavian experience of including reproductive traits in breeding programs. Pages 251– 261 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science Occasional Publication No 26. Edinburgh Scotland. Lucy, M. C., and B. A. Crooker. 2001. Physiological and genetic differences between low and high index dairy cows. Pages 223–236 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science occasional publication No 26. Edinburgh, Scotland. Olori, V. E., and P. Galesloot. 2000. Projection of partial lactations and calculation of 305-day yields in the republic of Ireland. Interbull Bull. No. 22:149–154. Philipsson, J. 1981. Genetic aspects of female fertility in dairy cattle. Livest. Prod. Sci. 8:307–319. Pryce, J. E., and R. F. Veerkamp. 2001. The incorporation of fertility indices in genetic improvement programmes. Pages 237–249 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science occasional publication No 26. Edinburgh, Scotland. Pryce, J. E., R. F. Veerkamp, R. Thompson, W. G. Hill, and G. Simm. 1997. Genetic aspects of common health disorders and measures of fertility in Holstein Friesian dairy cattle. Anim. Sci. 65:353– 360. Pryce, J. E., R. J. Esselmont, R. Thompson, R. F. Veerkamp, M. A. Kossaibati, and G. Simm. 1998. Estimation of genetic parameters using health, fertility and production data from a management recording system for dairy cattle. Anim. Sci. 66:577–584. SAS Institute. 1990. SAS/STAT User’s Guide, version 6, 4th ed. SAS Inst. Inc., Cary, NC. Silvia, W. J. 1998. Changes in reproductive performance of Holstein Dairy cows in Kentucky from 1972 to 1998. J. Dairy Sci. 81(Suppl. 1):244. (Abstr.) Snijders, S. E. M., P. G. Dillon, K. J. O’Farrell, M. Diskin, A. R. G. Wylie, D. O’Callaghan, M. Rath, and M. P. Boland. 2001. Genetic merit for milk production and reproductive success in dairy cows. Anim. Reprod. Sci. 65:17–31. Van Arendonk, J. A. M., R. Hovenier, and W. de Boer. 1989. Phenotypic and genetic association between fertility and production in dairy cows. Livest. Prod. Sci. 21:1–12. Veerkamp, R. F., P. Dillon, E. Kelly, A. R. Cromie, and A. F. Groen. 2001. Dairy cattle breeding objectives combining yield, survival and calving interval for pasture-based systems in Ireland. Livest. Prod. Sci. (in press).
Calving Interval and Survival Breeding Values as Measure of Cow Fertility in a Pasture-Based Production System with Seasonal Calving V. E. Olori,* T. H. E. Meuwissen,† and R. F. Veerkamp† *Irish Cattle Breeding Federation, Shinagh House, Bandon, Co. Cork, Ireland †Institute for Animal Science and Health, ID-Lelystad, P.O. Box 65, 8200 AB Lelystad, The Netherlands
ABSTRACT In a grass-based production system with seasonal calving, fertility is of major economic importance. A delay in conception due to poor fertility prolongs intercalving interval and causes a shift in calving pattern, which can lead to culling. Calving interval (CIV) information is readily available from milk records; analyzing it, however, presents a problem, as it is only available for cows that conceive and calve again. Calving interval should therefore be treated as a censored trait. In this study, survival to the next lactation (SUV) was analyzed jointly with CIV in a multivariate linear model to account for the selection in CIV data. Genetic parameters for first lactation calving interval were estimated with a sire model for Holstein Friesian cows in Ireland. SUV was preadjusted for production within herd-yearseason (HYS), while milk yield was included as a third trait in the analysis to account for the large effect it has on both traits. The residual covariance between CIV and SUV was fixed as 3 times the sire covariance within the model, as it was inestimable because of the structure of the data. Breeding values were estimated with various models to test the effect of culling and milk yield. Heritability was 0.04 ± 0.006 for CIV and 0.01 ± 0.003 for SUV, while the genetic correlation between them was −0.28 (±0.11). The genetic standard deviation was around 4% for SUV and 7 d for CIV. Sire predicted transmitting abilities for progeny tested bulls ranged between −5 and 3% for survival rate and between −4 and 8 d for calving interval. Differences between the best and worst bull varied with model. Including SUV and milk yield as traits in the model reduced the mean and variance of sire predicted transmitting abilities but increased the coefficient of variation by 30% compared with the univariate model. The current model is expected to account for most of the genetic variation in
Received September 25, 2001. Accepted November 5, 2001. Corresponding author: V. Olori; e-mail: [email protected].
fertility that is possible from calving dates and future extensions, such as the use of linear type trait or additional lactations for predicting survival, appear straightforward. These traits now form part of the national index for selecting dairy bulls in Ireland. (Key words: calving interval, survival score, fertility, seasonal calving) Abbreviation key: CIV = calving interval, HYS = herd-year-season, SUV = survival. INTRODUCTION Deterioration in cow fertility has become a major problem in dairy cattle production to the point that it is becoming increasingly apparent that high yielding cows are more difficult to get in calf. This negative association between production and fertility has been observed in several dairy breeds of cattle (Hermas et al., 1987; Hoekstra et al., 1994; Pryce and Veerkamp, 2001). The decline in reproductive efficiency of the Holstein Friesian cattle has been attributed to selection for increased production and or increase in the proportion of Holstein genes in the dairy cow population (Harris and Winkelman, 2000; Silvia, 1998). Results of recent studies in a grass-based production system indicated that cows of high genetic merit had a higher interval from calving to first service, poorer conception rate overall, poorer conception to first, second, and third services as well as higher number of services per conception compared to cows of medium genetic merit (Buckley et al., 2000a, 2000b; Snijders et al., 2001). There was no significant difference in milk yield, feed intake, and plasma glucose concentration between cows that conceive and those that did not conceive within the same genetic merit level (Snijders et al., 2001), which suggests that poor reproductive performance may not be a direct consequence of high milk production. An infertility rate of 21% for high genetic merit cows compared with 6% for the medium merit cows was reported (Snijders et al., 2001). The estimated margin after cost was about 0.02 US dollars per kilogram of milk higher
689
690
OLORI ET AL.
in medium-merit cows compared with the high genetic merit cows after considering loss due to poor fertility. Economic losses due to poor fertility are generally due to the cost of a prolonged calving interval, increased insemination cost, reduced returns from calves born, and forced replacement in the event of culling (Van Arendonk et al., 1989; Esselmont et al., 2001). Reduction in fertility and the attendant increase in the interval from calving to conception is a critical problem in seasonal calving herds, where calving interval has a high economic value because of the need for compact calving to maximize grass utilization during the grazing season. Increasing calving interval by 1 d costs the breeder about 1.8 US dollars in Ireland without accounting for the costs of higher culling due to poorer fertility (Veerkamp et al., 2001), and may be up to 4.5 US dollars, depending on the level of production and the extent of the delay (Esselmont et al., 2001). Measures of fertility generally have a low heritability, indicating a strong influence of environmental factors such as management decisions taken by the breeder. However, a review of several studies indicates that there is significant genetic variation between cows in the interval from calving to first breeding (Jansen, 1985), which is positively correlated with number of services per conception. The interval from calving to first breeding and number of services per conception both have positive correlation with number of days open, which is the variable component of calving interval. Calving interval is strongly correlated with days to first service (0.93) and moderately correlated with conception to first service in first lactation heifers (−0.56) as well as in cows, while measures of fertility in first lactation heifers are positively correlated with fertility in cows and have higher heritability (Pryce et al., 1997, 1998). Overall improvement in cow fertility can therefore be achieved by selecting on first-lactation heifer fertility based on calving interval. This will allow early selection but most importantly, the improvement of fertility by using calving dates available from milk records. Analyzing calving interval as a measure of fertility presents a problem because only animals that survive to the next lactation have a calving interval. Evaluation based on this trait alone will be biased because of preselection as culled animals with the worst fertility problems, will not be included in the analysis. Calving interval should therefore be treated as a censored trait and analyzed jointly with survival score to take into account the nonrandom scoring of calving interval. Furthermore, in a strict seasonal calving herd, poor fertility is expected to increase culling more than CIV, because a shift in calving pattern is not acceptable. Thus, the survival of a cow from one lactation to the next is also Journal of Dairy Science Vol. 85, No. 3, 2002
an important indicator of fertility. For these reasons, a combined analysis of calving interval and survival is expected to account for most of the genetic variation in fertility that is possible from calving dates. The objective of this study was to estimate genetic parameters and breeding values for first-lactation calving interval and survival rate for Holstein Friesian cattle in a grass-based and seasonal calving production system in the Republic of Ireland as a prelude to their inclusion in the national dairy breeding goal. MATERIALS AND METHODS Data The data used in this study were created from available milk records of about 400,000 Holstein Friesian cows commencing their first lactation between 1988 and 1998 in the Republic of Ireland. This included test-day and lactation milk yields as well as birth, calving, and lactation end dates. The file also contained information on lactation end reason, which indicated cows that were sold or died. It also indicated which cows were suckling, and those still in milk or dry. Records from about 200,000 lactations with no known end reason, which could not be conclusively identified as culled in the first lactation or survived to the second lactation were deleted. These include records from herds that stopped milk recording during the period or from cows that were sold to nonrecording herds. Usually, a cow sold to another milk recording herd will have a lactation end reason indicating that it was sold and its subsequent lactation will be found in another herd. A cow was assumed sold to a nonmilk recording herd when the end reason was ‘sold,’ but the next lactation record could not be traced. Such cows may also have been slaughtered thus it was safer to eliminate them from the analysis. Another 75,000 records from herds with less than 50 cows were also deleted to minimize the number of herd-year-season (HYS) classes. Cows were allocated to HYS subclasses based on the month of calving. Three seasons were defined as follows; January–April, May–August, and September–December. Each HYS class had a minimum of five records, while sire families were constrained to at least 10 cows. This led to a loss of another 35,000 records. The final data comprised of 89,101 cows sired by 1400 bulls in 21,139 HYS classes. About 71% of these cows survived to have a second lactation, while the remaining 29% were assumed culled in the first lactation. A standardized 305-d milk yield of each cow was calculated from available test-day records by the method of interpolation with standard lactation curves (Olori and Galesloot, 2000). For all cows with a subsequent calving date, calving interval (CIV in days) was calcu-
691
GENETIC EVALUATION OF COW FERTILITY Table 1. Mean and standard deviation of heifer lactation calving interval, milk, fat, and protein yields. Trait
Number of records
Mean
Standard deviation
Coefficient of variation
Calving interval, days Milk yield, kg Fat yield, kg Protein yield, kg
63,613 89,100 89,100 89,100
398 5475 204 180
75.9 1357 54 44
19 25 26 24
lated (Table 1). These cows were scored 1 for survival (SUV). CIV was set to null (missing) for cows that were culled in the first lactation, and these were scored 0 for survival. To account for voluntary culling, survival score was preadjusted for milk yield within HYS groups to take into account seasonal variation in the level of production within herd. This was achieved by obtaining the unadjusted residual score from a regression on milk yield deviated from the HYS mean. CIV was restricted to between 260 to 600 d (inclusive) to eliminate outliers that may be due to early abortion or cows kept open for repeated flushing to produce embryos for transfer. In a seasonal calving system, excessively long calving intervals may also result when a breeder opts to skip a year by leaving the cow open rather than change the calving pattern when a cow does not get in calf during the breeding season. About 2% of the records in this category had a survival score of 1, with CIV set to missing. No information was available in the current milk records to confirm these assumptions about the reason for the short or prolonged calving intervals observed in this data. Estimation of Variance Components Variance components were estimated from a joint analysis of calving interval with survival and milk yield using a multivariate sire model. Milk yield was included as a third trait in the analysis because of its positive association with calving interval and its effect on culling (not already accounted for by preadjustment of the data). It also accounts for a potential farmers’ attitude towards fertility and survival relative to production in
a seasonal production system. For instance, a breeder may not hesitate to cull a poor yielding cow that is difficult to get in calf. He may, however, decide to keep trying for a longer period to get a high yielding cow in calf rather than cull. There is also a higher probability of allowing a high yielding cow to survive in the herd albeit with a longer calving interval where poor cows will not. The inclusion of milk yield as a third trait accounts for the selection on yield, not accounted for by the preadjustment of SUV for milk yield, that is expected to account for the direct effects of yield on voluntary culling. The Statistical Model Variance components were estimated by REML with a sire model using ASREML (Gilmour et al., 1998). Relationships between sires (through their sires and dams) were included for two generations. Because only animals that survive (survival score = 1) have a calving interval value, the environmental covariance between survival rate and calving interval is undefined and inestimable (Janns and Bolder, 2000). However, the residual covariance is influenced by the genetic covariance between dams and mendelian sampling, and will normally be estimated as ³⁄₄*σa(i,j) + σe(i,j). Since the second term (σe(i,j)) cannot be estimated, the residual covariance becomes ³⁄₄*σa(i,j). The estimate of the sire covariance is ¹⁄₄*σa(i,j), hence, in this study, the residual covariance between survival score and calving interval was estimated as 3 times the sire covariance. Total variance was calculated as the sum of the residual and sire variances (σ2e + σ2s), while genetic variance was 4 times the
Table 2. Genetic and residual (co)variance components for yield, fertility and survival traits. Genetic (co)variances
Survival (%) × 10−02
Calving interval (days)
Milk yield (kg)
Survival Calving interval Milk yield Genetic standard deviation
0.002 −0.081 4.656 0.045
53.12 15,332 7.29
275,520 525
Residual (co)variances Survival Calving interval Milk yield
0.126 −0.061 −0.115
1418 2974
417,500 Journal of Dairy Science Vol. 85, No. 3, 2002
692
OLORI ET AL.
sire variance and heritability was calculated as 4σ2s/(σ2e + σ2s) (see Falconer and Mackay, 1997). Fixed effects in the model included management group (HYS), calving month, calving age, and the proportion of Holstein genes. The last two were fitted as covariate to the second degree. The linear model applied can thus be expressed as; 2 2 Yijkl = Ui + a1iX1i + a2iX1i b1iX2i + b2iX2i + Mij + HYSik + Sil + eijkl
where Yijkl = calving interval, survival between lactation 1 and 2 and/or milk yield, Ui = overall mean for the ith trait, a1i and a2i = the linear and quadratic regression coefficients (effects) of proportion of Holstein genes, b1i and b2i = the linear and quadratic regression coefficients (effects) of calving age, X1i and X2i = proportion of Holstein genes and calving age (linear and quadratic covariates), respectively, Mij = fixed effect of the jth month of calving on trait I, HYSik = fixed effect of kth herd year season class on trait I, Sil = the random genetic effect of the sire l of cows with records on trait i, and eijkl = random error term.
Figure 1. Distribution of first lactation calving interval for Holstein Friesian cows in Ireland.
for SUV, were expressed in percentage to correspond with the unit used in the derivation of the economic weight, while proofs for CIV were expressed in days. Further analysis, comparison, and conclusions were based on PTAs from about 650 bulls with at least five daughters in Ireland included in the breeding value estimation and having a minimum reliability of 30% for CIV in the single-trait analysis. The analysis of variance procedure (PROC ANOVA) in SAS (SAS, 1990) was used to analyze breeding values for each trait from the single, two- and three-trait analyses. Difference between models was based on an Fratio test from the one-way analysis of variance. RESULTS
Estimation of Breeding Values
Phenotypic Means
Breeding values were estimated by BLUP with a sire model using PEST (Groeneveld et al., 1990). The data used for this analysis was similar to the above, except that there was no limitation on the size of sire families. This was to allow as many bulls as possible to improve linkage in the relationship matrix, and to provide breeding values for all bulls in use. To see the effects of milk yield in the analysis as well as the effect of a combined analysis of SUV and CIV, single- versus multiple-trait estimates of breeding value were performed. Breeding values for SUV and CIV were first estimated separately, then jointly in a bivariate analysis and finally in a threetrait analysis with 305-d milk yield as the third trait. The genetic parameters used for breeding value estimation with the single-, two-, and three-trait models, were the same ones, estimated with a three-trait model as discussed in the parameters estimation section above. Predicted transmitting abilities were derived as half of the corresponding breeding value for all traits. Proofs
The average heifer yields and standard deviation for cows with records in the analysis were 5475 ± 1357, 204 ± 54, and 180 ± 44 kg for milk, fat, and protein yields, respectively. About 69% of these cows had calving intervals ranging from 260 to 600 d with a mean of 389 ± 55 d, while 66% of the cows had calving intervals ranging from 309 to 550 d. The modal calving interval was 363 d, which is close to the target of one calving per year desired in Ireland. First, second, and third quartile values were 352, 378, and 419 d, respectively. Figure 1 shows the frequency distribution of calving interval in the final data set before restriction to the range of 260 to 600 d.
Journal of Dairy Science Vol. 85, No. 3, 2002
Variance Components Table 2 shows the (co)variance components for survival rate, calving interval, and milk yield from the trivariate analysis. Genetic standard deviation was
693
GENETIC EVALUATION OF COW FERTILITY Table 3. Heritability (on diagonal), genetic (below), and phenotypic (above) correlations between survival, calving interval and milk yield with standard errors in parentheses.
Survival Calving interval Milk yield
Survival
Calving interval
Milk yield
0.01 (0.003) −0.28 (0.11) 0.22 (0.09)
−0.001 (0.003) 0.04 (0.006) 0.40 (0.07)
0.004 (0.003) 0.128 (0.004) 0.56 (0.03)
4.5% and 7.3 d for survival rate and calving interval, respectively. The corresponding heritabilities were 0.01 ± 0.003 and 0.04 ± 0.006, respectively (Table 3). The heritability of milk yield in the three-trait analyses was higher than usual at 0.56 ± 0.03. There was a negative genetic correlation between calving interval and survival (−0.28), while the correlation between milk yield and calving interval was positive (0.40). Milk yield was also positively correlated with survival (0.22). Predicted Transmitting Abilities Table 4 contains a summary of sire PTA for CIV and SUV from single- and multiple-trait analyses. The standard deviation of SUV PTA increased from 0.7% in the single-trait analysis to 1.1% in the three-trait analysis. There was no similar increase in the mean PTA, suggesting lack of systematic bias. The best bull had about 5% more daughters surviving, relative to the worst bull, between lactation one and two from the single-trait analysis. This increased to about 8% with the inclusion of CIV and milk yield in a three-trait analysis. The standard deviation of CIV sire PTA dropped from 1.6 d in the single-trait analysis to 1.38 d in the three-trait analysis; however, there was also a decrease in mean PTA, which resulted in an increase in coefficient of variation from 0.83 to 1.16. Figure 2 shows the distribution of sire PTA for CIV from the single, two- and three-trait analyses, while Figure 3 shows a similar distribution for SUV. The distribution of both CIV and SUV proofs from the single and two-trait models were relatively similar. There was a general shift of the curve to the left for CIV with
inclusion of milk yield suggesting that, analyzing calving interval alone, or along with survival rate without correcting for milk yield may result in the overestimation of breeding values for calving interval. This is because of the strong positive genetic and environmental effects of milk production on calving interval. The correlation between proofs from the single and two-trait analyses was 0.99 for CIV and 0.98 for SUV, and between the three-trait analysis and the single-trait analysis 0.91 and 0.92, respectively. Genetic trend for the CIV was similar for all the models. Figure 4 shows the trend of calving interval and survival by the birth year of the bulls based on sire PTA from the three-trait model. The trend indicates a steady increase in calving interval at the rate of 0.14 d per year, and a decline in survival rate at the rate of 0.05% per year since 1984. The %Holstein genes in the bull population was about 70%. Their daughters averaged 45%. Holstein gene percentage had a quadratic effect on SUV (R2 = 0.75), with significant drop in survival for bulls with 62% or more Holstein genes. For CIV, there was only a small linear (b = 0.049, R2 = 0.40) effect. The impact of %Holstein genes was more on CIV and less on SUV before inclusion of %Holstein genes in the model. DISCUSSION Cow fertility has become a major problem in dairy cattle production. Poor fertility delays conception, increases inter-calving interval, and causes a shift in calving pattern. In a seasonal calving production system, which relies on grass as a major source of feed, calving
Table 4. Summary statistics for calving interval and survival sire PTA from various models.1 Model
Mean
SD
Minimum
Maximum
Calving interval (days) Univariate (CIV) Bivariate (CIV-SUV) Trivariate (CIV-SUV-milk)
1.93 1.89 1.19
1.61 1.63 1.38
−3.36 −3.43 −3.71
8.10 8.15 6.55
Survival rate (%) Univariate (SUV) Bivariate (CIV-SUV) Trivariate (CIV-SUV-milk)
−0.27 −0.06 −0.14
0.73 0.78 1.11
−3.44 −3.34 −5.14
1.79 2.17 2.52
CIV = Calving interval, SUV = survival.
1
Journal of Dairy Science Vol. 85, No. 3, 2002
694
OLORI ET AL.
Figure 4. Genetic trend in calving interval and survival rate based on bulls with at least five daughters. Figure 2. Distribution of sire PTA for calving interval estimated from univariate (CIV1), bivariate CIV-SUV (CIV2), and trivariate CIV-SUV-Milk (CIV3) models for bulls with at least five daughters.
interval is an objective trait of economic importance because of the direct consequence of a shift in calving interval. Secondly, in the absence of data on direct measures of fertility, calving interval can be considered a good indicator of cow fertility because of the high correlation between CIV and several direct measures of fertility (Campos et al., 1994; Grosshans et. al, 1997; Pryce et al., 1997, 1998). The heritability of 4% and a genetic coefficient of variation of around 2% obtained in this study is within the range obtained in other studies (Janson, 1980, Hoekstra et al., 1994, Campos et al., 1994; Grosshans, et. al, 1997; Pryce et al., 1997, 1998). A genetic standard deviation of 7 d obtained for CIV in this study indicate significant genetic variation in the population suggesting that genetic improvement can be achieved through selection. Experience from the Scandinavian countries show that this is both feasible and beneficial (Phillipson, 1981; Lindhe and Phillipson, 2001).
Figure 3. Distribution of sire PTA for survival score estimated from univariate (SUV1), bivariate CIV-SUV (SUV2), and trivariate CIV-SUV-Milk (SUV3) models for bulls with at least five daughters. Journal of Dairy Science Vol. 85, No. 3, 2002
The main objective of this study was to investigate the feasibility of analyzing calving interval as a measure of cow fertility in routine genetic evaluation. The use in the past has not been contemplated perhaps because of the positive relationship with milk yield and the censored nature of calving interval. In this study, a multivariate linear model was used to analyze calving interval jointly with survival score and milk yield. This allows all cows (including those culled in the first lactation) to be included in the analysis thus avoiding the effect of preselection brought about by voluntary culling, which may be due among other things to poor fertility. The model was extended to include milk yield as a third trait because of the direct effect of production on voluntary culling. An increase in the coefficient of variation for sire PTA from the two- and three-trait models compared to the single-trait model suggests that a joint analysis of calving interval, survival score, and milk yield enable us to recover most of the variation in fertility that can be recovered from calving dates. Although only the interval and survival between first and second lactations have been studied here, the same principle is applicable to several lactations. Theoretically, nonlinear models seem to be more suitable for analyzing categorical data such as survival score and censored data such as calving interval. In one recent study, genetic parameters for fertility and health traits in cattle, estimated with a threshold model, where higher than estimates from a linear model (Kadarmideen et al., 2000). However, extension to an animal model for routine national genetic evaluation will be difficult to implement. Several studies have associated a decline in cow fertility with increase in milk yield (see Pryce and Veerkamp, 2001 for review) and percentage Holstein genes (Harris and Winkelman, 2000). Thus, a limitation in the analysis of calving interval and survival score, which in this study depends simply on reappearance in the sec-
695
GENETIC EVALUATION OF COW FERTILITY
ond lactation, is the potential for bias due to the influence of the breeder on the length of the intercalving interval. To avoid this, survival score was precorrected for yield within contemporary groups and 305-d yield was introduced as a third trait. Precorrection accounts for the phenotypic effect of milk yield on survival, hence, the phenotypic correlation between these traits was near zero (Table 3). However, there still exists, a significant genetic relationship between milk yield and survival, which explains the increase in the range of sire PTA for SUV from the three-trait model compared to the others. Our results indicate that these adjustments seem to eliminate potential bias, as indicated by the reduction in mean PTA, that may otherwise arise due to the farmers’ action dictated by the level of milk production. Examples of such action include delaying insemination of high yielding cows or production motivated selective culling of cows that fail to conceive during the breeding season. The correlation between sire PTA for CIV and milk yield was 0.51 in this study even with adjustment for percentage Holstein genes. This shows that the correction for Holstein percentage did not erode the positive genetic relationship between yield and calving interval. In this study, only the interval and survival between first and second lactation were analyzed. This keeps the model simple while still having the potential to improve overall cow fertility because of the strong correlation between first and later lactations (Hahn, 1969; Janson, 1980; Hansen et al., 1983; Pryce et al., 1998). Our heritability estimate was slightly higher than a recent estimate (0.02) from the UK (Kadarmideen et al., 2000). However, their estimate was obtained with a repeatability model including up to five lactation records per cow. Combining this with other direct measures of fertility and perhaps marker-assisted selection may speed up improvement. The genetic correlation between CIV and SUV indicate antagonistic relationship between calving interval and survival supporting the assertion that cows with prolonged calving interval are more likely to be culled in a seasonal calving herd scenario. We observed an unfavorable relationship between calving interval and milk yield similar to the findings in the UK (Kadarmideen et al., 2000). The difference in our correlation estimates and theirs (0.44 vs. 0.53) can also be attributed to the difference in data and model. The heritability of milk yield in this study was higher than those commonly reported in literature and from a recent UK study (Kadarmideen et al., 2000). However, this can be attributed to the data. Heritabilities from first-lactation records are usually higher than those estimated from multiple-lactation records. Also, in this study, it was based on a standardized 305-d
milk yield, even though the average lactation length in Ireland is about 280 d (ICBF, 2000). The high heritability may thus be due to a lower residual variance of standardized first-lactation records. One limitation of the current procedure is that cows can only be included in the analyses when they have a second lactation or when sufficient time has been given to ascertain that they will not have a second lactation. This delay means that young test bulls cannot be evaluated for these traits when they come off test, thus ruling out early selection decisions. A further delay is due to the time it takes to have sufficient daughters for a reliable proof because of the low heritability of the traits. A possible solution would be to include linear type traits that are significantly correlated with survival and calving interval in a joint analysis to facilitate earlier evaluation. The sire model approach also implies that proofs cannot be estimated directly for cows simultaneously with the bulls. Such will be possible with an animal model. The observed genetic variance may be currently limited because only first-lactation records are considered. A model that allows evaluation based on multiple lactations may increase variation in survival, as all animals that survive the first lactation may not have equal genetic potential to survive five lactations. It is suggested that further studies be put in place to investigate the possibility of an animal model with multiple lactation data. A multivariate analysis with linear predictors of calving interval should also be investigated as this will allow proofs to be calculated early for young test bulls as well as allow foreign bulls with no daughters in Ireland to be properly evaluated for these traits. It is suggested also that a good tracking system that promptly reports on cattle movement and lactation end reasons be put in place. This will allow us to distinguish between animals that were culled and those that have simply moved to nonmilk recording herds. CONCLUSION This study has shown how calving interval can be analyzed as a measure of fertility in dairy cattle without the bias due to culling and the effect of milk yield. This was achieved in a joint analysis of calving interval, precorrected survival score, and milk yield. This model allows the utilization of records from all milk-recorded cows, including those culled in the first lactation. It is possible that the use of genetic parameters obtained with a three-trait model could have affected our estimate of breeding values with a single- and two-trait model and hence our conclusions. However, the extent of this effect is not clear. Journal of Dairy Science Vol. 85, No. 3, 2002
696
OLORI ET AL.
ACKNOWLEDGMENTS We are grateful to B. Wickham and the Irish Cattle Breeding Federation (ICBF) for funding the ‘Dairy Breeding Objective Project’ of which this study was part. This study and V. Olori’s position in ICBF were partly funded with European Union structural funds. REFERENCES Buckley, F., P. Dillon, S. Crosse, F. Flynn, and M. Rath, 2000a. The performance of Holstein Friesian dairy cows of high and medium genetic merit for milk production on grass based feeding systems. Livest. Prod. Sci. 64:107–119. Buckley, F., P. Dillon, M. Rath, and R. F. Veerkamp. 2000b. The relationship between genetic merit for yield and live weight, condition score and energy balance of spring calving Holstein Friesian dairy cows on grass based systems of milk production. J. Dairy Sci. 83:1878–1886. Campos, M. S., C. J. Wilcox, C. M. Beceril, and A. Diz. 1994. Genetic parameters for yield and reproductive traits in Holstein and Jersey Cattle in Florida. J. Dairy Sci. 77:867–873. Darwash, A. O., G. E. Lamming, and J. A. Woolliams. 1997. The phenotypic association between the interval to postpartum ovulation and traditional measures of fertility in dairy cattle. Anim. Sci. 65:9–16. Esselmont, R. J., M. A. Kossaibati, and J. Allcock. 2001. Economics of fertility in dairy cows. Pages 19–29 in Fertility in the HighProducing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science, Occasional Publication No 26. Edinburgh, Scotland. Falconer, D. S., and T. F. C. Mackay, 1996. Introduction to Quantitative Genetics. 4th ed. Longman Group Ltd., Essex, England. Gilmour, A. R., B. R. Cullis, S. J. Welham, and R. Thompson. 2000. ASREML Reference Manual. NSW Agriculture, Orange Agric. Inst. Orange, Australia. Groeneveld, E., M. Kovac, and T. Wang. 1990. PEST, a general purpose BLUP package for multivariate prediction and estimation. Proc. 4th World Congr. Genet. Appl. Livest. Prod. Edinburgh, Scotland XIII:488–491. Grosshans, T., Z. Z. Xu, L. J. Burton, D. L. Johnson and K. L. Macmillan. 1997. Performance and genetic parameters of fertility in seasonal dairy cows in New Zealand. Livest. Prod. Sci. 51:41–51. Harris, B. L., and A. M. Winkelman. 2000. Influence of North American Holstein genetics on dairy cattle performance in New Zealand. Page 122 in Proc. 2000 Large Herds Australia Confr., Armidale, Australia. Hermas, S. A., C. W. Young, and J. W. Rust. 1987. Genetic relationships and additive genetic variation in productive and reproductive traits in Guernsey dairy cattle. J. Dairy Sci. 70:1252–1257. Hoekstra, J., A. W. van der Lugt, J. H. J. van der Werf, and W. Ouweltjes. 1994. Genetic and phenotypic parameters for milk production and reproductive performance traits in upgraded dairy cattle. Livest. Prod. Sci. 40:225–232.
Journal of Dairy Science Vol. 85, No. 3, 2002
Irish Cattle Breeding Federation (ICBF) 2000. Irish cattle breeding Statistics. 2000. Irish Cattle Breeding Society Limited, Shinagh House, Bandon. Co. Cork. Ireland. Jansen, J. 1985. Genetic aspects of fertility in dairy cattle based on analysis of AI data—a review with emphasis on areas for further research. Livest. Prod. Sci. 12:1–12. Janss, L. L., and N. M. Bolder. 2000. Heritabilities and genetic relationships between salmonella resistance traits in broilers. J. Anim. Sci. 78:2287–2291. Kadarmideen, H. N., R. Thompson, and G. Simm. 2001. Linear and threshold model genetic parameters for disease, fertility and milk production in dairy cattle. Anim. Sci. 71:411–419. Lindhe, B., and J. Philipsson. 2001. The Scandinavian experience of including reproductive traits in breeding programs. Pages 251– 261 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science Occasional Publication No 26. Edinburgh Scotland. Lucy, M. C., and B. A. Crooker. 2001. Physiological and genetic differences between low and high index dairy cows. Pages 223–236 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science occasional publication No 26. Edinburgh, Scotland. Olori, V. E., and P. Galesloot. 2000. Projection of partial lactations and calculation of 305-day yields in the republic of Ireland. Interbull Bull. No. 22:149–154. Philipsson, J. 1981. Genetic aspects of female fertility in dairy cattle. Livest. Prod. Sci. 8:307–319. Pryce, J. E., and R. F. Veerkamp. 2001. The incorporation of fertility indices in genetic improvement programmes. Pages 237–249 in Fertility in the High-Producing Dairy Cow. M. G. Diskin, ed. British Society of Animal Science occasional publication No 26. Edinburgh, Scotland. Pryce, J. E., R. F. Veerkamp, R. Thompson, W. G. Hill, and G. Simm. 1997. Genetic aspects of common health disorders and measures of fertility in Holstein Friesian dairy cattle. Anim. Sci. 65:353– 360. Pryce, J. E., R. J. Esselmont, R. Thompson, R. F. Veerkamp, M. A. Kossaibati, and G. Simm. 1998. Estimation of genetic parameters using health, fertility and production data from a management recording system for dairy cattle. Anim. Sci. 66:577–584. SAS Institute. 1990. SAS/STAT User’s Guide, version 6, 4th ed. SAS Inst. Inc., Cary, NC. Silvia, W. J. 1998. Changes in reproductive performance of Holstein Dairy cows in Kentucky from 1972 to 1998. J. Dairy Sci. 81(Suppl. 1):244. (Abstr.) Snijders, S. E. M., P. G. Dillon, K. J. O’Farrell, M. Diskin, A. R. G. Wylie, D. O’Callaghan, M. Rath, and M. P. Boland. 2001. Genetic merit for milk production and reproductive success in dairy cows. Anim. Reprod. Sci. 65:17–31. Van Arendonk, J. A. M., R. Hovenier, and W. de Boer. 1989. Phenotypic and genetic association between fertility and production in dairy cows. Livest. Prod. Sci. 21:1–12. Veerkamp, R. F., P. Dillon, E. Kelly, A. R. Cromie, and A. F. Groen. 2001. Dairy cattle breeding objectives combining yield, survival and calving interval for pasture-based systems in Ireland. Livest. Prod. Sci. (in press).