Female Fertility Evaluations
The Animal Improvement Programs Laboratory (AIPL) will provide genetic evaluations for daughter pregnancy rate (DPR) beginning February 2003. Many producers have expressed concerns about the difficulty of achieving desired levels of reproductive performance in today's milking herds. Over the last year, AIPL staff examined several reproductive traits to determine if genetic selection for daughter fertility is possible and compared methods to evaluate and present this information. Calving interval and days open have been available from Dairy Herd Improvement Association (DHIA) data for many years, but were not evaluated routinely because fertility traits tend to have very low heritabilities (about .04). Predicted transmitting abilities (PTA) for cow fertility traits will have high reliabilities only after hundreds of daughters are recorded. For bulls with only first-crop daughters, reliabilities average about 60%, and parent averages still provide much of the information. Pregnancy rate and days open are almost the same trait genetically, and a 1% higher pregnancy rate equals 4 fewer days open. PTA DPR is not yet included in net merit, cheese merit, or fluid merit calculations, but will be added in future revisions of these indexes.
Pregnancy rate allows herd managers to measure how quickly their cows become pregnant again after having a calf and is defined as the percentage of nonpregnant cows that become pregnant during each 21-day period. In recent years, many reproductive specialists have recommended this measure of reproductive success over the more traditional measure, days open, because of several advantages. Pregnancy rate calculations are more current, cows that do not become pregnant are included in calculations more easily, and larger rather than smaller values are desirable. The pregnancy rate is 20% for a herd that averages 154 days open as compared to 25% for a herd with 133 days open. The nonlinear formula to convert from days open to pregnancy rate is
Pregnancy rate = 21/(days open - voluntary waiting period + 11),
where voluntary waiting period is the initial phase of lactation during which no inseminations occur (assumed to be 60 days). The factor of +11 adjusts to the middle day of the 21-day cycle so that cows that conceive during the first cycle receive 100% credit on average.
The formula for pregnancy rate is not very linear when graphed across the whole range of days open. However, the curve can be approximated by a straight line across the smaller changes in daughter means that result from sire genetic differences. Because of low heritability, PTA days open and PTA DPR both have small ranges and are nearly linear functions of each other. Each increase of 1% in PTA DPR equals a decrease of 4 days in PTA days open, and PTA days open can be approximated as PTA DPR multiplied by -4. Thus, a bull with a PTA of +2.0 for pregnancy rate would have a PTA of -8 for days open. The genetic correlation between days open and pregnancy rate is extremely high (.99) because the only way to reduce days open is for cows to become pregnant at a faster rate.
Pregnancy rates calculated by AIPL differ slightly from those reported by processing centers and reproductive specialists. AIPL calculations exclude additional cycles after day 250 of lactation and exclude lactations with no reported inseminations if the cow was sold during that lactation for reasons other than reproduction. Genetic evaluations are expressed as deviations from a base pregnancy rate within each breed. Several other countries also calculate and provide cow fertility evaluations, but trait definitions differ. Although a few countries have more complete fertility data than in the United States and evaluate more than one trait, most evaluate only first-insemination nonreturn rate, which has lower heritability and excludes much of the genetic variation in a cow's ability to cycle early in lactation and to express heat. Also, fertility evaluations in some countries may be adjusted for correlations with yield traits.
Bull and Cow Fertility
Reproductive performance is affected primarily by environment, management, and unexplained sources of variation, and to much smaller degrees by the genes of the cow, the fertility of her mate, and the interaction of the cow and bull. The fertility of AI bulls is monitored routinely by major organizations using microscopic examination of sperm count, motility, and abnormality, other laboratory tests, and technician data. Male fertility rankings expressed as estimated relative conception rate (ERCR) have been available from Dairy Records Management Systems (DRMS) since 1986. The ERCR rankings include data from three cooperating processing centers and measure a service bull's direct effect on fertility at the time of insemination. A bull's fertility is affected by changes in his environment over time and by his genes, but ERCR values have a fairly small range because numbers of live sperm per unit are equalized and infertile bulls are culled.
The PTA DPR is an additional measure of a sire's effect on fertility when his daughters are being bred. These two independent measures of bull and cow fertility (ERCR and DPR) are analogous to the service bull and daughter effects on calving ease introduced in August 2002. However, the units of expression for ERCR and DPR differ and the two cannot be compared directly or additively because mean conception rates are by definition higher than pregnancy rates. Bull and daughter effects on fertility and on calving ease may all be important economic traits but none are yet included in the calculation of net merit.
In addition to potential selection for pregnancy rate, mating programs can be valuable in reducing embryo losses that result from harmful gene interactions. Inbreeding should be avoided, and animals that carry the same lethal recessive gene should not be mated. Analogous advice has been provided for calving ease for many years. Many Holstein breeders have mated their heifers to bulls with favorable direct effects instead of selecting for calving ease and then mating cows and heifers at random to the selected bulls. Benefits of mating programs are larger for traits where interactions can be predicted.
Fertility Data Sources
Data used by AIPL to calculate DPR include 40 million lactation records from 16 million cows that calved since 1960. The national evaluations include up to five lactations for each cow (the same number as for yield traits). Date pregnant is determined from several information sources and is verified by the cow's calving interval when available. The best information on date pregnant is the date of last insemination verified by the next calf's birth date occurring within 15 days of the expected date. If no inseminations are reported through Dairy Herd Improvement records, the date pregnant is calculated by subtracting an average gestation length (280 days for Ayrshires, Guernseys, Holsteins, Jerseys, and Milking Shorthorns; 290 days for Brown Swiss) from the date of next calving. For lactations that do not have a date of next calving available (because the cow has been sold, the herd stopped testing, or the current date is less than the last breeding date plus the average gestation length), the date pregnant is assumed to be the date of last insemination but cannot be verified. The last reported breeding date is also used if the next lactation is initiated by abortion. The last insemination is assumed to have failed if no calving is reported within 295 days (305 days for Brown Swiss) and the cow is still alive at that time.
A final source of information for some lactations occurs when the owner reports that a cow was sold for beef because of reproductive problems. Such cows are assumed to be nonpregnant when sold, and the date of last insemination is disregarded. Records for pregnancy rate are considered to be complete at 250 days in milk (DIM), and pregnancy status after 250 DIM is not used. Date pregnant is set equal to 50 for any cows that become pregnant before 50 DIM. Some extremely early pregnancy dates obtained by calculation from date of next calving are inaccurate because of short gestation lengths or unreported abortions. The lower and upper limits of 50 and 250 are applied after adjusting days open for season effects and affect 5% and 14% of records, respectively. Further research may determine that different limits are needed.
Table 1 gives the percentages of recent cows with each type of fertility information or with no information. For Holsteins and Jerseys, only 6% of lactations had a next calf born with no previously reported breeding date and another 5% had breeding dates inconsistent with birth date of the next calf. Thus, accurate breeding dates are reported by most farms, but many of the reported final breeding dates (19% for Holstein) cannot be verified because the cow was sold for reasons other than reproductive problems. Differences among breeds in percentages of lactations terminated by reproductive culling seem consistent with the smaller decline in fertility experienced by Jerseys.
Records are used only if they have the opportunity to be complete so that selection bias can be avoided. Methods to use actual pregnancy diagnoses and records in progress are being developed so that pregnancy rate can be evaluated earlier in lactation in the future. Incomplete and unverified records should receive less weight in the evaluations but currently all records that are included get the same weight. Information on each insemination may be used in the future, but currently only the date of successful or last insemination is stored at AIPL. Also, AIPL does not yet collect data on pregnancy rate for heifers. Age at first calving could possibly be evaluated, but heifer fertility traits are somewhat different genetically from cow fertility traits.
Initial research at AIPL including heritability estimation, environmental adjustments, and preliminary genetic evaluations were based on days open. Genetic parameters were estimated using a sire model and restricted maximum likelihood with three large data sets. The first data set included calving interval records from 1,062,791 Holstein cows born from 1992 through 1994 so that complete productive life records were available for all cows. The heritability of days open (calculated by calving interval) in first lactation was 3.7%. Genetic correlation with productive life was -.59, which indicated that cow fertility plays a major role in longevity. Two more recent data sets were also examined and were based on actual insemination data for lactations initiated from 1998 through 2000 for 2,195,643 Holsteins and 145,976 Jerseys. Heritability of days to last breeding was 4.0% for Holsteins and 2.9% for Jerseys.
Parameters used for all breeds in routine animal model evaluations for cow pregnancy rate will be 4% for additive genetic effects, 1% for effect of interaction of sire and herd, and 6% for permanent environmental effect, resulting in a repeatability of 11% as obtained by Dematawewa and Berger (1998, J. Dairy Sci. 81:2700). Selection for high yield over several generations has contributed to longer calving intervals because of an unfavorable genetic correlation between yield and days open of about .3. Selection for productive life since 1994 apparently has slowed the decline in fertility, but direct selection for fertility could be more profitable.
Genetic trends for the different breeds show declining fertility for all breeds across time. Genetic bases were the means of progeny tested bulls born in 1995 rather than means of cows born in 1995 used for other traits. In general, the PTA for animals of different breeds cannot be compared directly, but increases or decreases across time can be compared. Milking Shorthorn, Jersey, and Ayrshire breeds had smaller losses of fertility across time, whereas Guernsey, Brown Swiss, and Holstein had larger losses. The smaller trend for Jerseys is not consistent with their higher rate of inbreeding but is consistent with the lower estimated heritability and smaller range of PTA The Holstein genetic trend has become nearly flat after 1994, perhaps because of selection for increased productive life. Phenotypic pregnancy rates, however, continue to decline. The genetic trends across four decades are consistent with correlated responses expected from selection for high yield, but explain only about 40% of the phenotypic fertility trends in days open.
Records are adjusted for season effects prior to analysis. Initial research indicated that for all breeds fertility is best with fall calvings and poorest with spring calvings because fewer cows express estrus or conceive during hot summer months. Some breeds had insufficient numbers of herds to obtain accurate adjustments in all five regions and five time periods. Thus, Jersey adjustments were applied to Guernsey data, and Holstein adjustments were applied to data from all other breeds. Season effects have increased over time such that adjustments are somewhat larger for current data and somewhat smaller for older data compared with the overall estimates in the graph. For recent Holstein data, regional differences show that spring calvings in the southeastern United States result in greatly reduced fertility.
Variance adjustments are applied to data after season adjustments using the same procedure introduced in 1991 for yield traits. Variation in days open has increased over time and is higher for herds with higher means. One difference from the yield trait adjustments is that heritability is assumed to remain constant regardless of variation within the herd. Fertility PTA's were calculated both with and without the variance adjustments and differences were small.
Genetic Evaluation Model
Cow fertility records are processed with the same animal model programs that AIPL uses for yield traits, productive life, and somatic cell score, which make interpretation of PTA DPR fairly simple. Cows in the same herd and management group are compared directly, and the definition of management groups is the same as for yield traits except for cows that change herds during a lactation. For yield traits, the herd providing the most information is the herd of evaluation. For pregnancy rate, the herd in which the cow became pregnant or from which the cow was sold because of reproductive problems is the herd of evaluation.
The animal model used for routine evaluation of cow fertility includes adjustments for parity that are defined separately within three regions of the United States and nine time periods. Records are not adjusted for age within parity because an older age at a given parity may indicate poorer fertility in the past and adjustment would remove part of the genetic effect. Records also are not adjusted for milk yield because the phenotypic correlation is lower than the genetic correlation, interpretation may be simpler without adjustment, and other traits such as productive life, somatic cell score, and udder depth are not adjusted for yield. However, a concern is that the correlation of milk yield with fertility may be caused not only by estrus expression or conception problems in high-producing cows but also by the owner voluntarily waiting more days to begin breeding the high producers. These causes and effects are difficult to separate. Another concern is that rankings could differ for herds that synchronize estrus versus herds that rely on tradition estrus detection.
Several possible cow fertility traits and evaluation methods were compared using simulated data, which allowed correlations with true transmitting abilities to be obtained. Transformation of days open to the logarithm of days open or to pregnancy rate using a nonlinear formula before evaluation did not increase accuracy. A weighted evaluation of pregnancy rate in which lactations that required more cycles for the cow to become pregnant received more weight was preferred slightly to the unweighted evaluation of pregnancy rate but was not more accurate than the initial evaluation of days open. Instead, the simpler linear approximation
pregnancy rate = .25(233 - days open)
is applied to adjusted data for days open before animal model analysis so that solutions are expressed as DPR. This linear formula appears as a straight green line in the graph shown earlier.
Bull and Cow Evaluations
Evaluations of DPR were calculated for all bulls and cows included in the November 2002 release. Table 2provides statistics for bulls of each breed with active AI status in November and for cows born in 1995. The PTA DPR are compared to the progeny tested bulls born in 1995 which formed the base for this trait. The bull base was chosen so that DPR evaluations of currently marketed bulls would be centered near 0. Means for active AI bulls may increase in February 2003 because individual evaluations will be available and can be considered when organizations decide which bulls will be marketed.
The PTA DPR for individual bulls are not yet official and are not being distributed at this time to avoid use in advertising before procedures are documented and understood by the public. However, unofficial DPR for older sires of sons may aid in the educational process. These PTA obtained from November 2002 data differ slightly from those obtained earlier from May 2002 data because of the change from a cow base population to a bull base and improved edits, enhancements, and additional records for fertility. PTA days open can be derived for educational use by multiplying PTA DPR by -4 but will not be provided routinely by AIPL.
PTA DPR was correlated by .46 with PTA productive life for recently progeny-tested Holstein bulls; the correlation was lower (.23) for recent Jersey bulls. Breeders should consider the sum of an animal's strengths and weaknesses instead of using independent culling levels for each trait. Many breeders have reduced the fertility of their herds by selecting for dairy form in addition to milk yield. Thin cows tend to have higher classification scores and lower pregnancy rates. Correlations of DPR with current net merit were near 0. Future indexes that include pregnancy rate will help to maintain or improve fertility but may reduce progress for less important traits.
The authors thank George Wiggans and Lillian Bacheller for improving the fertility database, John Clay for suggesting that the evaluations be expressed as pregnancy rate, and everyone in the DHIA system for providing data.