| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Animal Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705-2350
Corresponding author: Melvin Kuhn; e-mail: mkuhn{at}aipl.arsusda.gov.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
45 d and 2 herds with mean DD of 30 d. Some herds (125) had unusually long mean DD (exceeding 71 d). Variation across regions of the United States was minimal, although southwestern herds did show somewhat longer DD than other regions. Days open (DO) accounted for the most variation in DD with longer DO leading to longer dry periods. Heritability of DD was 7% and repeatability 12%. Error correlations indicated that longer dry periods benefit both milk yield and somatic cell score in the subsequent lactation. Genetic and permanent environmental correlations were large, which indicated that 1) bias would result if genetic evaluations for milk yield included an adjustment for DD, and 2) correction for cow effects is warranted in estimation of DD effects on performance. The relationship between DD and DO is due primarily to permanent cow effects. Improved fertility will enhance the feasibility of shortened dry periods.
Key Words: days dry • days open
Abbreviation key: DD = days dry, DO = days open, PE = permanent environment.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
Characterization of factors affecting DD can be useful to extension and consulting personnel, researchers, and dairy producers. To formulate management recommendations, extension scientists and other consultants need to first be familiar with current practices. For research aimed at delineating the dry period length to maximize performance, knowledge of factors affecting DD can be useful in determining appropriate models for field data or factors to control in designed studies. Finally, although dairy producers are no doubt aware of what factors affect their individual decisions to dry off cows, the extent that various factors play overall may be largely unknown. More insight might lead to improved management practices.
Given within-herd variation for DD and evidence for its effect on subsequent lactation milk yield, several researchers have advocated adjustment for DD in genetic evaluation for production (Funk et al., 1987; Makuza and McDaniel, 1996). Estimates of genetic correlations, however, are critical in evaluating the propriety of this recommendation. If the genetic correlation between 2 traits differs from zero, then adjusting one trait for the other in genetic evaluation redefines the trait being evaluated. The genetic evaluations for the redefined trait are biased with respect to the original, unadjusted trait. Estimates of genetic, permanent environment (PE), and error correlations of DD with other economically important traits can also be useful in research aimed at determining the effects of DD on performance. For example, if genetic and PE correlations with performance are not zero, this would indicate that adjustment for cow effects is warranted in estimation of DD effects on performance.
The purpose of this research was to characterize DD for modern US Holsteins. This included summarization of variation in herd means as well as within herd variability for DD, determination of factors affecting DD, and estimation of parameters related to DD including heritability and genetic and environmental correlations with other economically important traits.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
There are numerous DHI testing plans for US herds. Some plans involve herd owners taking the measurements themselves, rather than having an official DHI tester take milk weights and samples. Such records generally contain more error than supervised testing (Animal Improvement Programs Laboratory, 2004). Animal Improvement Programs Laboratory considers any record where more than 50% of the test-days were taken by the herd owner to be owner-sampled and these records must pass additional edits to be included in the national genetic evaluations (Animal Improvement Programs Laboratory, 2004a). Thus, to ensure quality of data, a herd’s average owner-sampler percentage had to be less than 50% across the period specified to be included.
Because not all centers began sending complete lactation records at the same time, the date at which dry period length was available varied somewhat across dairy records processing centers. Thus, the January 1997 inclusion date had to be moved forward for some herds. This edit change applied primarily to herds processed by the 2 centers located in the western United States.
Records initiated by abortion were deleted. Cows known to be embryo transfer donors were also excluded because embryo transfer could lead to abnormally long dry periods or lactations. Days dry was required to be between 0 and 120 d; dry periods longer than 120 d were deleted. Another important edit was that the expected calving date based on last reported days open (DO) and the actual calving date had to agree within 10 d. Some researchers (Bachman and Schairer, 2003; Grummer, 2004) have argued that analyses using DHI data are biased because most short dry periods included are those that are unplanned, due to cows calving earlier than expected. They indicated that if the producers had known when the cows were going to calve, the cows would have been managed differently to deal with the short DD. Hence, this edit ensured that producers knew when cows were going to calve because, in effect, they reported it to DHI.
Descriptive Statistics
Within-herd standard deviations of DD were calculated and summarized by an overall mean and range. Within-herd standard deviations were also averaged by region. A frequency distribution of individual DD records, in contrast to herd mean DD, was also calculated using dry periods up to the beginning of fifth lactation. For this distribution, no upper limit was placed on DD. This was the only analysis without the DD 120 d requirement.
Factors Affecting DD
In this study, previous lactation was defined as the lactation preceding a given dry period and subsequent lactation as the lactation following a given dry period. For example, for the dry period between first and second lactations, the previous lactation is parity 1, and the subsequent lactation is parity 2.
In studies aimed at determining the effect of DD, the dependent variable would be the trait of interest (e.g., milk yield) in the subsequent lactation, and DD would be included in the model as an independent variable for that trait. For such studies, the previous lactation is generally of little interest. However, to determine what factors affect DD, the opposite is true. For example, the first DD would be influenced by what occurred in the first, not second lactation (which had not occurred yet when the cow went dry). Thus, to determine to what extent various factors affect DD, the dependent variable is DD, and factors in the previous lactation are the independent variables of interest. The equation used for modeling DD was
 | ([1]) |
where DD is DD between previous and subsequent parities, MO is calving month, YR is calving year, Parity is lactation number (1, 2, ..., 6), Age, Milk, and Last_SCS were categorical variables for age at calving, actual lactational milk yield, and last test-day SCS, respectively, and e was a random error. All effects were fit as fixed effects and were from the previous lactation.
Calving age categories were defined as 1) <24 mo, 2) 24 to <30 mo, 3) 30 to <36 mo, 4) 36 to <42 mo, 5) 42 to <48 mo, and 6) 
48 mo. Age categories were largely nested within parity. Thus, the effect of age investigated here is an age effect, beyond parity or within parity. Categories for milk, SCS, and DO were defined as 1) lower third of data sorted by each trait, 2) middle third, and 3) top third. Actual milk yields were calculated from test-day yields using the test-interval method (Sargent et al., 1968) and lactation-shape adjustment factors (Shook et al., 1980). A total of 459,075 records from 295,067 cows in 3527 herds were included in this analysis. The mean of each category for milk, SCS, and DO is given in Table 1
.
|
.
|
Another alternative considered was lactational vs. last test-day SCS. Although SCS increases slightly at the end of lactation, it is much less sensitive to stage of lactation than milk yield (Miller et al., 2004). Preliminary analyses showed that last test-day SCS accounts for more variation than lactational SCS. Thus, last test-day SCS, rather than lactational SCS, was used in the final model for analysis.
Days in milk was also considered for inclusion in model [1], but DIM was largely confounded with DO. Days open was chosen as the variable to include in the final analysis for several reasons. First, DIM was substituted for DO in a preliminary analysis and resulted in a lower coefficient of determination than did the model including DO. Days open is more of a causative agent toward dry period length than is DIM. Cows have long DIM in part because they had long DO. The days corresponding to the long open period are split between a longer lactation and a longer dry period. Thus, it is expected that DO would account for more variation in DD than DIM. Finally, DO, as a measure of fertility, is probably of more interest as a variable affecting dry period length. Least squares means for DIM from the preliminary analysis in which it was substituted for DO in model [1] are given in the results.
Estimation of Heritability and Correlation
The 2-trait animal model used for estimation of variances and covariances of DD with each of DO, 305-d, twice daily milking, mature-equivalent milk yield, SCS, fat percentage, and protein percentage was
 | ([2]) |
where HY is herd-year of calving, YR_ST_MO is year-state-month of calving; Age is calving age fit as linear and quadratic covariates with regression coefficients b1 and b2; a is the random animal effect, pe is a random PE effect, and e is random error. Herd-yr was used, instead of HY-season, to avoid small group sizes. Month was added to the model to account for season effects. Month effects were allowed to differ by state as well year within state, hence YR-ST-MO. Co(variance) matrices for a, pe, and e were
 |
where A is a relationship matrix, I is an identity matrix, and 
2 is a scalar variance.
Variances and covariances were estimated using REML. Only a random sample of the data, used to study factors affecting DD, was used for parameter estimation because the complete data set was more than could be handled computationally. Herds were randomly selected for inclusion. Each data set, used for parameter estimation, contained between 30,000 and 35,000 cows with records.
Days dry is related to previous lactation milk yield as well as milk yield in the subsequent lactation, but in different ways. Milk yield in the lactation before the dry period (e.g., lactation 1) partly determines the length of the dry period, but then the dry period length, in turn, has an effect on the subsequent lactation (e.g., lactation 2) yield. The correlations of DD with other traits could conceivably be different depending on whether the previous or subsequent lactation was used. When DD is associated with the previous lactation, milk yield would be a causative agent toward DD, whereas with subsequent lactation, the roles of cause and effect would be reversed. Therefore, correlations of DD with other traits were estimated for both previous and subsequent lactations. Because of the computing time involved, only 2 traits at a time were fit. Thus, correlations of DD with previous and subsequent lactations were estimated separately rather than simultaneously.
Correlation of DD with subsequent lactation traits implies that DD was missing for first lactation. To avoid selection bias, however, first lactations were still included, and DD was treated as missing for first lactation. For correlations of DD with previous lactation traits, DD was missing for the last lactation.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
Table 2 has the frequency distribution of individual DD records. Less than 3% of all dry periods included were less than 35 d. About 65% of all dry periods were between 46 and 65 d and about 1.5% of dry periods were greater than 150 d.
|
presents mean squares for model [1] computed from SAS type III sums of squares (SAS Institute, 2004). Table 4 shows the least squares means for each effect except herd. Of the effects included in model [1], DO, by far, accounted for the most variation in DD. Longer DO result in longer dry periods. The difference (3.4 d) in DD between cows with shortest and medium DO was small, but the difference in mean DD between cows with shortest and longest DO was 11.3 d. When cows have delayed conception, their milk yield likely declines during the extended lactation to a point where keeping them in milk is no longer profitable and they are dried off, which results in long open as well as long dry periods. Excessively long dry periods associated with long DO is additional motivation to improve fertility through both genetics and management.
|
|
Milk yield and SCS accounted for the second and third most variation in DD, respectively. As expected, high milk cows had the shortest DD (56 d), whereas the poorest producers had the longest DD (65 d). This simply indicates that cows are dried off according to production level. Cows with the highest SCS on last test-day had 2.4 more DD than those with the lowest SCS. This probably indicates that some cows that get mastitis near the end of lactation are dried off earlier than they would have been otherwise.
Cows had shorter (4 to 5 d) dry periods after first lactation than after second and later lactations. First-parity cows are more persistent than later parity cows (Stanton et al., 1992), which could influence managers to keep them milking longer. Differential culling may also account for part of the differences between parities. Only the more productive cows make it to second and later lactations and longer dry periods may be tolerated more for those cows than for first-lactation cows; a first-parity cow may instead get culled.
Although significant (P <0.0001), month, year, and age at calving accounted for only a minimal amount of variation in DD. Cows calving in warmer months (May through August) had slightly shorter dry periods than cows calving in cooler months (especially November through February). The means in Table 4
indicate slightly shorter dry periods for 2002, but this probably reflects a slight bias caused by the final date used for inclusion of records. Cows with long dry periods that calved late in 2002 may not have had the opportunity to calve again by the cutoff date. Age at calving had little influence on DD beyond the parity effect, although younger cows did have slightly fewer DD than did older cows.
In contrast to many traits of economic importance in dairy cattle, herd actually accounted for the smallest amount of variation in DD. The overall mean DD was 60.5 d, and the standard deviation among herd means was 5.8 d. The lowest herd mean was 28 d, whereas the highest herd averaged 82 d. The distribution of herd means is in Table 5. About 65% of all herds maintained a mean DD between 56 and 65 d. Although variation among herd means was low, 12 herds had a mean of 40 d, and 2 herds had a mean DD of 30 d. Thus, although most herd managers appear to be maintaining, on average, the traditional 60-d dry period, some are utilizing a shorter dry period. Some herds (125) had unusually long mean DD, exceeding 71 d.
|
) was also minimal, although southwestern herds did show slightly longer mean DD than did herds in other regions. This may simply reflect differences in management, perhaps related to larger herd sizes that are more common in the western United States.
|
Correlations of DD with previous and subsequent mature-equivalent milk, fat and protein percentage, SCS, and DO are presented in Table 7. The error correlations of –0.17 and 0.14 between DD and previous and subsequent lactation milk yield, respectively, indicate that high milk cows receive fewer DD, but cows afforded a longer dry period produce more milk in the subsequent lactation. The error correlations between DD and component percentages had signs opposite to those for milk yield, which is not surprising given the negative correlation between percentages and milk yield (Welper and Freeman, 1992). Cows with high fat and protein percentages tended to have longer dry periods, but cows with longer dry periods have lower percentages in the subsequent lactation. These results indicate that long dry periods are detrimental to fat and protein percentages and that protein is more affected than fat. Cows with high SCS have longer DD than cows with lower SCS as indicated by the error correlation of 0.07 with previous lactation. However, long dry periods favor somewhat improved SCS in the subsequent lactation, as indicated by the error correlation of –0.05 with subsequent lactation.
|
What this reflects, however, is a partitioning of the DD effect, across model terms, an effect which is present in the subsequent lactation record but not in the previous. Consider the following simplified model for first-lactation yield (y1):
 |
where Am is breeding value for milk yield, PEm is a permanent environmental effect for yield, and em1 is a first-lactation error. Given the correlation of DD with previous yield, an equation for DD between first and second lactations (DD1) can be written as:
 |
where ßa, ßpe, and ße are regressions of DD breeding value, PE, and error on milk yield breeding value, PE, and error, respectively, and ed is an independent error term. An equation for second-lactation yield would be:
 |
where DD_effect1 is the effect of DD associated with a dry period of length DD1. The covariances of y1 and y2 with DD1 are:
 |
Considering DD_effect1 as a random variable, the probability of occurrence of DD_effect1i is the same as the probability of occurrence of DD1i. Thus, DD_effect1 and DD1 are clearly correlated and if DD1 is correlated with Am, PEm, and em, the DD effect will be as well. Hence, the DD effect will contribute to genetic and environmental correlations when DD is associated with subsequent lactation.
With the exception of milk yield, genetic correlations between DD and other traits (Table 7) were generally the same for previous and subsequent lactations. The negative genetic correlations between mature-equivalent milk yield and DD were consistent with the findings of Schaeffer and Henderson (1972), who reported genetic correlations ranging from –0.18 to –0.41 across parities. One implication of the nonzero genetic correlations between DD and milk yield is that caution and further research are warranted before implementing a correction for DD in genetic evaluation for milk yield, in spite of relatively low heritability for DD. Such research, perhaps using simulation, should determine the amount of the breeding value that would be partitioned out of the animal effect and the possible long-term consequences of such a partitioning. Furthermore, with an error correlation between DD and milk yield in subsequent lactation of only 0.14, the benefit of adjusting for DD (correction for a known environmental effect) may be less than the cost (redefinition of the trait or bias relative to total breeding value for the original trait).
As with the error correlations, the genetic correlations between DD and component percentages were opposite in sign compared with those for milk yield and generally of lower magnitude as well. The negative genetic correlation between DD and SCS indicates that cows that are genetically prone to high SCS have lower breeding values for DD. This finding may seem somewhat peculiar because high SCS would be expected to be associated with longer DD (a positive correlation), rather than shorter DD, as indicated in Table 4. The genetic relationship between DD and SCS may result from their relationships with milk yield. Breeding value for high milk yield is associated with breeding value for poorer SCS (Welper and Freeman, 1992). High breeding value for milk yield, then, may lead to breeding values for higher SCS but also for shorter dry period.
The PE correlation of 0.19 between DD and previous lactation milk yield indicates that PE effects for high production tend to predispose the cow toward longer dry periods. However, the –0.89 correlation for subsequent lactation milk yield shows that cows with PE effects for high production tend to receive low DD effects. The genetic and PE correlations for subsequent lactation milk yield clearly indicate that cows with inherent high-producing ability receive the shortest dry periods.
The genetic and PE correlations of DD with DO converged to 1. Table 3 shows that DO had the most pronounced effect on DD. Table 7 indicates that the magnitude of that effect resulted mainly from additive genetic and PE components and those components are not subject to quick and easy change. Milk yield or SCS in the subsequent lactation, for example, could be improved simply by deciding to give the cow a longer dry period as indicated by the subsequent lactation error correlations for those traits. However, manipulation of DD is not an answer for cows with long DO because such cows are almost certainly dried off and receive longer dry periods because the producer believes that keeping them in milk is no longer economically prudent. Genetic improvement of female fertility, along with identification of and improvement in environmental factors that have permanent effects on female fertility, will be a useful, if not necessary, companion for viable shortened dry periods.
The error correlations of DD with DO were low. The –0.06 error correlation with subsequent DO does imply some reproductive benefit of a shortened dry period, but that benefit may result from the reduced milk yield associated with fewer DD as implied by the 0.14 error correlation with milk yield.
The error correlations in this study indicate direction of difference but not magnitude of difference. The positive error correlation (0.14) between DD and subsequent lactation yield, for example, implies that production increases as DD increases. However, the magnitude of differences, or what particular DD maximizes performance in the subsequent lactation, cannot be gleaned from these results. Further research should reexamine the effect of DD on subsequent lactation using recent data not only for production but for components, SCS, and fertility as well. One implication of the large genetic and PE correlations between DD and the other traits in subsequent lactation is that correction for cow effects is warranted in studies aimed at discerning the effect of DD on subsequent lactation performance.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
Heritability and repeatability estimates for DD were 7% and 12%, respectively, and were generally consistent with previously published research. Sizeable genetic correlations warrant caution and further study before following recommendations to correct for DD in genetic evaluation. The relationship between DD and DO results largely from permanent cow effects that are not quickly or easily manipulated. Genetic and environmental improvement in factors affecting reproduction will enhance the viability of shortened dry periods.
Further research should reexamine the effect of DD on subsequent lactation using recent data. Most genetic and PE correlations were of a sizeable magnitude, which indicates that correction for cow effects in such studies would be warranted.
|
APPENDIX |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
for DD, DO, and mature-equivalent milk yield, fat percentage, protein percentage, and SCS.
|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication October 18, 2004. Accepted for publication November 9, 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSAPPENDIXACKNOWLEDGEMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. T. Kuhn, J. L. Hutchison, and H. D. Norman Dry Period Length in US Jerseys: Characterization and Effects on Performance J Dairy Sci, April 1, 2007; 90(4): 2069 - 2081. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. T. Kuhn, J. L. Hutchison, and H. D. Norman Dry Period Length to Maximize Production Across Adjacent Lactations and Lifetime Production J Dairy Sci, May 1, 2006; 89(5): 1713 - 1722. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. T. Kuhn and J. L. Hutchison Methodology for Estimation of Days Dry Effects J Dairy Sci, April 1, 2005; 88(4): 1499 - 1508. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
