Bio-economic modeling to support insemination decisions in dairy cows
Chaidate Inchaisri 2011
Bio-economic modeling to support insemination decisions in dairy cows Chaidate Inchaisri Dissertation, Faculty of Veterinary Medicine, Utrecht University, the Netherlands ISBN:978-90-393-5610-4 Copyright © C. Inchaisri, E-mail:
[email protected] Cover, pictures and lay-out: Chaidate Inchaisri Printed by: B.V.S Service
Bio-economic modeling to support insemination decisions in dairy cows Bio-economische modellen ter ondersteuning van inseminatiebeslissingen bij melkkoeien (met een samenvatting in het Nederlands)
(พร้ อมบทสรุปภาษาไทย) Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr.G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op vrijdag 7 oktober 2011 des middags te 4.15 uur
door
Chaidate Inchaisri geboren op 28 april 1970, te Loei, Thailand
Promotor:
Prof. dr. G.C. van der Weijden
Co-promotoren:
Dr. ir. H. Hogeveen Dr. R. Jorritsma Dr. P.L.A.M. Vos
The research described in this thesis was financially supported by the Commission on Higher Education, Ministry of Education, Royal Thai Government
Contents Chapter 1
General introduction
1
Chapter 2
Economic consequences of reproductive
9
performance in dairy cattle Chapter 3
Effect of milk yield characteristics, breed and
33
parity on success of the first insemination in Dutch dairy cows Chapter 4
Analysis of the economically optimal voluntary
53
waiting for first insemination Chapter 5
Cow effects and estimation of success of first
77
and following inseminations in Dutch dairy cows Chapter 6
Improved knowledge about conception rates
95
influences the decision to stop insemination in dairy cows Chapter 7
Summarizing discussion
111
Summary
121
Nederlandse samenvatting (Dutch summary)
125
บทสรุป (Thai summary)
129
Acknowledgements
133
Curriculum Vitae
137
List of Publications
141
CHAPTER 1
General introduction
Chapter 1 Worldwide, a high reproductive efficiency guarantees an optimal economic profit on dairy farms. The achievement of a calving interval (CI) of 12 to 13 months is generally regarded as economically optimal. As a consequence of an increased milk production, it appears that this objective is difficult to achieve as conception rates are reported to decline during the last decade in most European countries as well as in the United States (e.g., CRV, 2009, De Vries and Risco, 2005, Gonzalez-Recio et al., 2004, Lopez-Gatius, 2003, Lucy, 2001). To avoid efforts to compensate for the lower conception rates in early lactation, both farmers and advisors suggest that the general objective of an economically optimal CI of 12 to 13 months may not be able to apply to all herds or to all cows in a specific herd (Allore and Erb, 2000, van Amburgh et al., 1997). In fact, several studies on this subject have pointed at the possibility to postpone the first insemination and extend the calving interval for a cow with a high and persistent milk production (Arbel et al., 2001, Sørensen and Østergaard, 2003, van Amburgh et al., 1997). However, the complexity of interactions and dynamics of factors influencing reproductive performance do influence seriously the determination of an economically optimal time to start and, likely more important, to terminate inseminating an individual cow within a herd. Thus, although there is a strong demand from dairy practice to provide this information for supporting fundamental management decisions, so far useful answers to these questions are not available on cow level for Dutch dairy farms. For the development of an insemination decision support system, an economic simulation model may serve as a tool for modeling studies to determine the optimal breeding and replacement decisions (De Vries, 2006, Groenendaal et al., 2004, Houben et al., 1994, Nielsen et al., 2010, Sørensen and Østergaard, 2003, van Arendonk and Dijkhuizen, 1985). Although such economic simulation models are available for decision support, interactions between the milk yield, insemination numbers and reproductive performances of an individual cow are many times lacking in those models. To build a simulation model for making economically relevant decisions which may be used for an optimal insemination strategy, significant epidemiological inputs like information on probabilities of a successful first insemination and hence following inseminations are needed. Unfortunately, these data are not available for the Dutch situation. Therefore, the objective of this thesis was to develop bio-economic simulation models which may be used for decision support with regard the insemination strategy of individual cows within a herd including all relevant interactions between important cow factors estimated from epidemiological data derived under Dutch situations. Economic simulation models for insemination and replacement decisions Because of the dynamic and complex interactions of cow factors on reproductive performance, economic simulation models have been developed to provide information to assist management decisions. Several studies have reported on the economics effects of various scenarios of reproductive performance (Boichard, 3
General introduction 1990, Meadows et al., 2005, Oltenacu et al., 1980, Plaizier et al., 1998). Optimization models for insemination and replacement have also been developed as decision support tools to assist dairy farmers with insemination and replacement (De Vries, 2006, Groenendaal et al., 2004, Houben et al., 1994, Nielsen et al., 2010, Sørensen and Østergaard, 2003, van Arendonk and Dijkhuizen, 1985). These optimization models were based on marginal net revenues and dynamic programming. When using dynamic programming, sequential decisions can be modeled using a Markov decision process (Houben et al., 1994, Nielsen et al., 2010, van Arendonk and Dijkhuizen, 1985). The hierarchical Markov decision process was developed to build a series of Markov chain processes into one (smaller) Markov chain process (Kristensen, 1992) to increase computational efficiency. This method was extended to a multilevel hierarchical Markov decision process, enabling the use of smaller and transition matrices. Multilevel hierarchical Markov decision processes were recently used to evaluate the economic consequences of clinical mastitis (Bar et al., 2008) and lameness (Cha et al., 2010). Recently, a multilevel hierarchical Markov decision model has been incorporated with a Bayesian reevaluation based on previously information to predict the performance of each individual cow to evaluate the optimal replacement policy for an individual cow on a daily basis (Nielsen et al., 2010). However, even with the use of hierarchical Markov decision processes, with dynamic programming it is difficult to model complex interactions, for instance between the relation of probability of conception and cow factors such as lactation number, calendar months of insemination, level of milk yield and months in milk. Moreover, because of the consequences for the transition matrices, many dynamic programming models use for the length of a stage a relatively long period of one month or, in some models even one year. To model the biological processes of reproduction, a shorter stage length is necessary. Using marginal net revenue in a spreadsheet model (Groenendaal et al., 2004, Meadows et al., 2005) is a relatively easy modeling method, but it lacks of the possibility to take the variation expected performance of individual cows into account. Furthermore, the method of marginal net revenues lack the possibility to model the dynamics of reproduction Dynamic stochastic Monte Carlo model have been introduced to evaluate the economic effects of several diseases in dairy cattle such as mastitis (Steeneveld et al., 2011), lameness (Bruijnis et al., 2010), and infertility (Hockey and Morton, 2010). The dynamic stochastic Monte Carlo model can take into account the variation and dynamics of biological and economical parameters. Moreover, because no strict transition matrix has to be made, it becomes easier to model interaction between cow factors. However, Monte Carlo models are not very well capable of working with sequential decisions, which means that it is difficult to evaluate long term effects of decisions, such as the cost of genetic improvement and the economic optimization of culling. 4
Chapter 1 Developing a new bio-economic model to support insemination decisions In general after calving, the ovarian cyclicity is resumed and as a consequence the first ovulation results in the first estrus cycle. Due to the voluntary waiting period, cows will be inseminated at a predetermined time after calving and likely may become pregnant. Failure of pregnancy, leads to the new start of an estrus cycle and hence cows may either be re-inseminated, culled immediately after a diagnosed nonpregnant status or culled at the end of lactation. After a successful pregnancy, the first or next calving will occur after a complete gestational period. The decision on the optimal voluntary waiting period associates with the successful first insemination whereas the decision for the optimal time to stop insemination relates to the success of the following inseminations. Many intrinsic and extrinsic cow factors do have interactive effects on the occurrence of a successful pregnancy. In this respect, not only the stage of lactation but also other cow factors, e.g., milk yield, parity, breed and season of insemination time, affect a successful pregnancy (Kuhn et al., 2006, Loeffler et al., 1999, Veerkamp et al., 2001, Windig et al., 2005). Several milk parameters do affect the successful pregnancy in dairy cows such as cumulative milk yield, 305–d milk yield, peak yield and milk persistency (Kadarmideen et al., 2000, Kinsel and Etherington, 1998, Muir et al., 2004, Veerkamp et al., 2001). However, using those milk parameters to predict reproductive performances have to be improved because they are partly dependent on the reproductive performance. From a practical point of view, farmers do use only those parameters which are measurable such as daily milk yield at insemination date and days in milk, before making the decision to start or to delay an insemination. Moreover, recent studies have shown that a serial number of estrous cycles and hence matings do influence the probability of pregnancy (Friggens and Labouriau, 2010, Kuhn and Hutchison, 2008). Thus, to improve the accuracy of the estimation for a successful insemination, one has to incorporate the insemination number postpartum with daily milk yield and days in milk at the date of insemination in a new designed model which is limited for Dutch circumstances. The optimal decision on insemination depends on many uncertainty factors not only the interactions between cow factors, e.g., lactation number, milk yield, conception rate and days in milk, but also the market circumstances such as milk price and replacement cost (De Vries, 2006, Groenendaal et al., 2004). A dynamic modeling approach including variation may be considered to capture the complex interactions between cow factors and reproductive performances over the lactation. A Monte Carlo stochastic simulation model can take the dynamics of reproduction and interactions between reproductive performance and cow factors into account for a number of time steps throughout the lactation, including the variation in conception and estrous detection rates. By developing a dynamic stochastic simulation model to calculate the economic consequences of change in the time to start inseminating, insight can be gained in the optimal time to start insemination for a specific cow. However, the 5
General introduction decision to stop inseminating is in fact a decision to cull a cow (immediately or later in lactation). Therefore, to support decisions with regard to the time to stop inseminating, the effect of sequential insemination and culling decisions become more important. The comparison between the present net revenue and the expected future net revenue representing by the retention pay-off value in a dynamic programming model should be used to support decisions on stopping or continuing inseminations. Therefore, for the development of an inseminating decision support tool, a dynamic stochastic Monte Carlo model is a possibly and hence suitable instrument to calculate the economic consequences when making a decision for the optimal time to start inseminating whereas the dynamic programming is a suitable tool to calculate the economic consequences when making the decision to stop inseminating. OBJECTIVES The main objective of this thesis was to develop a bio-economic model to support insemination decisions of an individual cow within a herd, based on epidemiological studies performed in Dutch dairy cows. To reach this objective, several studies were conducted with the following goals: 1. Development of an economic simulation model to evaluate the economic impact of individual reproduction parameters for individual cows under varying circumstances using inputs from literatures and expertise (Chapter 2). 2. Epidemiological evaluation of the relationship between cow factors, such as daily milk production and stage of lactation, parity and breed on the probability of a successful first insemination in Dutch dairy cows (Chapter 3). 3. Development of an economic tool to determine the optimal time to start inseminating after calving for a specific dairy cow in Dutch circumstances and evaluation of the differences in optimal time to start inseminating between cows (Chapter 4). 4. Epidemiological evaluation of the probability of successful insemination of subsequent inseminations of Dutch dairy cows. The influences of cow factors and the failure of the previous inseminations on the next successive inseminations within the same lactation were estimated (Chapter 5). 5. Evaluation for the optimal moment to stop inseminating and the economic consequences in a specific cow using different epidemiological inputs of successful inseminations for Dutch dairy cows (Chapter 6). REFERENCES Allore, H. G. and H. N. Erb. 2000. Simulated effects on dairy cattle health of extending the voluntary waiting period with recombinant bovine somatotropin. Prev. Vet. Med. 46(1):29-50.
6
Chapter 1 Arbel, R., Y. Bigun, E. Ezra, H. Sturman, and D. Hojman. 2001. The effect of extended calving intervals in high-yielding lactating cows on milk production and profitability. J. Dairy Sci. 84(3):600-608. Bar, D., L. W. Tauer, G. Bennett, R. N. González, J. A. Hertl, H. F. Schulte, Y. H. Schukken, F. L. Welcome, and Y. T. Gröhn. 2008. Use of a dynamic programming model to estimate the value of clinical mastitis treatment and prevention options utilized by dairy producers. Agri. Sys. 99(1):6-12. Boichard, D. 1990. Estimation of the economic value of conception rate in dairy cattle. Livest. Prod. Sci. 24(3):187-204. Bruijnis, M. R. N., H. Hogeveen, and E. N. Stassen. 2010. Assessing economic consequences of foot disorders in dairy cattle using a dynamic stochastic simulation model. J. Dairy Sci. 93(6):2419-2432. Cha, E., J. A. Hertl, D. Bar, and Y. T. Gröhn. 2010. The cost of different types of lameness in dairy cows calculated by dynamic programming. Prev. Vet. Med. 97(1):1-8. CRV. 2009. CRV(Coöperatie Rundvee Vebetering) Annual Reports 2009. https://www.cr-delta.nl/servlets/dbupload?id=6161 Accessed May 1, 2010. De Vries, A. 2006. Economic value of pregnancy in dairy cattle. J. Dairy Sci. 89(10):38763885. De Vries, A. and C. A. Risco. 2005. Trends and seasonality of reproductive performance in Florida and Georgia dairy herds from 1976 to 2002. J. Dairy Sci. 88(9):3155-3165. Friggens, N. C. and R. Labouriau. 2010. Probability of pregnancy as affected by oestrus number and days to first oestrus in dairy cows of three breeds and parities. Anim. Reprod. Sci. 118(2-4):155-162. Gonzalez-Recio, O., M. A. Perez-Cabal, and R. Alenda. 2004. Economic value of female fertility and its relationship with profit in Spanish Dairy Cattle. J. Dairy Sci. 87(9):3053-3061. Groenendaal, H., D. T. Galligan, and H. A. Mulder. 2004. An economic spreadsheet model to determine optimal breeding and replacement decisions for dairy cattle. J. Dairy Sci. 87(7):2146-2157. Hockey, C. D. and J. M. Morton. 2010. Use of a stochastic simulation model to assess effects of diagnostic specificity of systems for detecting ovulating cows on herd reproductive performance in year-round calving dairy herds. Anim. Reprod. Sci. Houben, E. H., R. B. Huirne, A. A. Dijkhuizen, and A. R. Kristensen. 1994. Optimal replacement of mastitic cows determined by a hierarchic Markov process. J. Dairy Sci. 77(10):2975-2993. Kadarmideen, H. N., R. Thompson, and G. Simm. 2000. Linear and threshold model genetic parameters for disease, fertility and milk production in dairy cattle. Anim. Sci. 71(3):411-419. Kinsel, M. L. and W. G. Etherington. 1998. Factors affecting reproductive performance in Ontario dairy herds. Theriogenology 50(8):1221-1238. Kristensen, A. R. 1992. Optimal replacement in the dairy herd: A multi-component system. Agr. Syst. 39(1):1-24.
7
General introduction Kuhn, M. T. and J. L. Hutchison. 2008. Prediction of dairy bull fertility from field data: Use of multiple services and identification and utilization of factors affecting bull fertility. J. Dairy Sci. 91(6):2481-2492. Kuhn, M. T., J. L. Hutchison, and G. R. Wiggans. 2006. Characterization of Holstein heifer fertility in the United States. J. Dairy Sci. 89(12):4907-4920. Loeffler, S. H., M. J. de Vries, Y. H. Schukken, A. C. de Zeeuw, A. A. Dijkhuizen, F. M. de Graaf, and A. Brand. 1999. Use of AI technician scores for body condition, uterine tone and uterine discharge in a model with disease and milk production parameters to predict pregnancy risk at first AI in Holstein dairy cows. Theriogenology 51(7):12671284. López-Gatius, F. 2003. Is fertility declining in dairy cattle? A retrospective study in northeastern Spain. Theriogenology 60(1):89-99. Lucy, M. C. 2001. Reproductive loss in high-producing dairy cattle: where will it end? J. Dairy Sci. 84(6):1277-1293. Meadows, C., P. J. Rajala-Schultz, and G. S. Frazer. 2005. A spreadsheet-based model demonstrating the nonuniform economic effects of varying reproductive performance in Ohio dairy herds. J. Dairy Sci. 88(3):1244-1254. Muir, B. L., J. Fatehi, and L. R. Schaeffer. 2004. Genetic relationships between persistency and reproductive performance in first-lactation Canadian Holsteins. J. Dairy Sci. 87(9):3029-3037. Nielsen, L. R., E. Jørgensen, A. R. Kristensen, and S. Østergaard. 2010. Optimal replacement policies for dairy cows based on daily yield measurements. J. Dairy Sci. 93(1):75-92. Oltenacu, P. A., R. A. Milligan, T. R. Rounsaville, and R. H. Foote. 1980. Modelling reproduction in a herd of dairy cattle. Agric. Sys. 5(3):193-205. Plaizier, J. C. B., G. J. King, J. C. M. Dekkers, and K. Lissemore. 1998. Modeling the relationship between reproductive performance and net-revenue in dairy herds. Agric. Sys. 56(3):305-322. Sørensen, J. T. and S. Østergaard. 2003. Economic consequences of postponed first insemination of cows in a dairy cattle herd. Livest. Prod. Sci. 79(2-3):145-153. Steeneveld, W., T. van Werven, H. W. Barkema, and H. Hogeveen. 2011. Cow-specific treatment of clinical mastitis: An economic approach. J. Dairy Sci. 94(1):174-188. van Amburgh, M. E., D. M. Galton, D. E. Bauman, and R. W. Everett. 1997. Management and economics of extended calving intervals with use of bovine somatotropin. Livest. Prod. Sci. 50(1-2):15-28. van Arendonk, J. A. M. and A. A. Dijkhuizen. 1985. Studies on the replacement policies in dairy cattle. III. Influence of variation in reproduction and production. Livest. Prod. Sci. 13(4):333-349. Veerkamp, R. F., E. P. Koenen, and G. De Jong. 2001. Genetic correlations among body condition score, yield, and fertility in first-parity cows estimated by random regression models. J. Dairy Sci. 84(10):2327-2335. Windig, J. J., M. P. Calus, and R. F. Veerkamp. 2005. Influence of herd environment on health and fertility and their relationship with milk production. J. Dairy Sci. 88(1):335-347.
8
CHAPTER 2 Economic consequences of reproductive performance in dairy cattle C. Inchaisri1,2, R. Jorritsma2, P.L.A.M. Vos2, G.C. van der Weijden2, and H. Hogeveen2,3
1
Department of Veterinary Medicine, Faculty of Veterinary Science, Chulalongkorn University, Bangkok, 10330, Thailand 2 Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, P.O. Box 80152, 3508 TD, The Netherlands 3 Business Economics, Wageningen University, P.O. Box 8130, 6706 KN Wageningen, The Netherlands
Published in Theriogenology (2010) 74: 835-846
Chapter 2 ABSTRACT The net economic value of reproductive efficiency in dairy cattle was estimated using a stochastic dynamic simulation model. The objective was to compare the economic consequences of reproductive performance scenarios (“average” and “poor”) of a cow having a good reproductive performance and to explore which reproductive factors have an important impact on economic efficiency. A “good” reproductive performance scenario was defined with 1 ovulation rate (POVUi), 0.7 estrus detection rate (PEst), 0.7 conception rate (PCon), 0.03 incidence rate of postpartum disorders prolonging the ovarian cyclicity (CO), 0.2 incidence rate of postpartum disorders reducing conception (ME), 0.05 embryonic death rate (ED) and voluntary waiting period (VWP) of 9 wks pp (post partum). In the current situation of dairy cows in the Netherlands, an “average” reproductive scenario (0.95 POVUi, 0.5 PEst, 0.5 Pcon, 0.07 CO, 0.27 ME, 0.07 ED and VWP of 12 wks pp) and a “poor” reproductive scenario (0.90 POVUi, 0.3 PEst, 0.3 Pcon, 0.11 CO, 0.33 ME, 0.09 ED and VWP of 15 wks pp) were identified. A sensitivity analysis was performed by comparing changes of single effect of factors in a good and poor scenario with the average scenario. The mean net economic loss (NELi) compared with the good scenario was €34 and €231 per cow per year for the average and poor reproductive performance scenario, respectively. Increasing the calving interval resulted in greater economic loss. The important factors on the cost of reproductive efficiency were the involuntary culling cost and the return of milk production. Variation in PCon, PEst , ME , ED and VWP had large impacts on economic benefits. (Keywords: Dairy cow; reproductive performance; simulation model; economics)
11
Economic consequences of reproductive performance INTRODUCTION In the dairy industry, the genetic composition of the herd, in combination with proper housing conditions and balanced management conditions, is important to obtain acceptable economic results. Many cow and management related factors that may affect the reproductive performance of the herd have been described (Pryce et al., 2004, Windig et al., 2005). For example, early onset of ovarian follicular development post partum, and hence regular cyclicity, adequate estrus detection, and insemination at the correct moment in relation to ovulation are crucial steps for successful conception and pregnancy. Several studies have shown that the continuous increase in milk production coincides with a decrease in cow fertility, measured in terms of prolongation of the calving interval (CI) (De Vries, 2004, Gonzalez-Recio et al., 2004, Hare et al., 2006, Lucy, 2001, Royal et al., 2000). In addition to a more biological approach to this situation, e.g., by exploring the (hormonal) mechanisms between negative energy balance and reproductive performance, it is evident that commercial dairy farmers need practical instruments and information for their management decisions. Less interested in the underlying mechanisms, these farmers weigh the costs of any intervention required to improve reproduction performance with its benefits. They may, for example, choose not to shorten the interval from parturition to first insemination when the economic benefits of a higher milk production per year, associated with a longer lactation period, are higher (Dhaliwal et al., 1996, López et al., 2004, Mayne et al., 2002, Opsomer et al., 1998). Moreover, farmers usually use estimated economic outcomes to prioritize their options, e.g., to focus on improving the estrus detection rates or on preventing postpartum uterine infections. As a result of more fundamental research findings in this field, extensive data are available about biological effects and environmental interactions (Jorritsma et al., 2000, Opsomer et al., 1998, Windig et al., 2005). However, the economic consequences of these biological effects are rarely reported, even for the average cow (Boichard, 1990, Plaizier et al., 1997, Stott et al., 1999). Plaizier et al. (1997) showed the effect of an adjusted calving interval for pregnant cows that were not culled for reproductive failure on the net revenue and Meadows et al.(2005) reported the economic loss for an increase in days open. Boichard (1990) reported the cost of a change in conception rate and Marsh et al. (Marsh et al., 1987) showed the effect of a change in conception rate and estrus detection rate under various breeding and culling policies on the gross margin of a farm. Significant economic returns of extending the first insemination time after calving for high milk producing cows have been reported (Arbel et al., 2001, Tenhagen et al., 2003). Although commonly described, the models used in these valuable reports usually lack interactions between factors such as (relative) milk production level, lactation stage, age, or disease and fertility parameters. The objective of the present work was to calculate the economic consequences of different reproductive performances for individual cows under varying circumstances. To this 12
Chapter 2 end, we evaluated reproductive performance scenarios with average and poor reproductive performance and compared them with cows with a good reproductive performance. Unlike studies that compared the economic consequences of different reproductive performances, we included also interactions between milk production level, lactation stage, age, postpartum disorders and pregnancy rate in our stochastic dynamic simulation model. Finally, the developed simulation model was used to evaluate the economic impact of individual reproduction parameters such as estrus detection and conception rate. MATERIALS AND METHODS A Monte Carlo stochastic-simulation model was developed to estimate the economic loss due to decreased reproductive performance. The model was created in Excel (Microsoft Corporation, Redmond, WA, USA) with @Risk add-in software (Palisade Corporation, Ithaca, NY, USA). The model took the individual cow as a basis. First, the parity and the level of milk production of a cow in a herd were randomly selected from relevant distributions. Then the model simulated fertility events, based on discrete distributions, using weekly time steps. The probability of an event occurring in a particular time step was dependent on a basic probability distribution of the events and on cow information from the previous time step. For every cow, the results of a “good”, “average” and “poor” reproductive performance scenario were calculated simultaneously. The relevant parameters, e.g. calving interval, number of inseminations and milk production, were collected for each simulation. In a next step, these parameters were compared and used for an economic evaluation. The iterative process stopped when convergence (a steady state) was reached. Convergence was reached when the standard deviation of each output variable did not change by more than 1.5% after a new iteration. The model is described in more detail in the following paragraphs. Default values (Tables 1 and 2) for cow information, reproductive dynamics and economics were based on the scientific literature and the authors’ expertise and were chosen in such a way that they resembled the dairy situation in the Netherlands. Milk production characteristics of a specific cow In each iteration, a specific cow in a herd with an average rolling herd milk production (HMP) of 8008 kg per cow per year (CRV, 2009) was simulated. For each cow i (i = 1, ..., n), the parity number (Pi), with the possible values 1 to ≥ 5, was based on a discrete distribution with a proportion identified for every parity number (P1, P2, P3, P4, P5+).
Pi = discrete [(1, 2, 3, 4, 5+)]; [(P1, P2, P3, P4, P5+)]
13
[1]
Economic consequences of reproductive performance
To simulate the milk production level (kg/305 days) of the individual cows, a relative level of production, lactation value (LVi), was drawn from a normal distribution with mean = 1 and standard deviation = 0.1. Then, the milk production level (kg/305 days) of each individual cow i in parity p (MPip) was calculated as follows: LVi × HMP × MPPp MPip = i = 1, ..., n
[2]
where MPPp is a milk production factor of 0.9, 1.02, 1.07 and 1.08 for, respectively, parity 1, 2, 3, and ≥ 4 (CRV, 2009) and n represents the number of runs of individual cows to reach convergence. In the next step, the daily milk yield of cow (i) in wk (j) adjusting for parity (p) effect (MYijp) was calculated using Wood’s function (Wood, 1967). (A × jB × exp(-Cp j))i × MPip 44 (A × j B × exp(-Cp j))i × 7 j=1 [3] MYijp = i = 1, ..., n j = 1, ..., week at drying off In this function, factors A, B, and C modeled the lactation curve characteristics. A represents the milk production (kg/day) at the peak of the lactation curve, B represents the timing of the peak in milk production (set constant at 6 wks pp) and C p represents the parity dependent milk persistency. Cp has a value of 1, 1.59 and 1.77 for parity 1, 2 and ≥ 3, respectively (CRV, 2009) Reproductive cycle The dynamics of the reproductive cycle were simulated from calving to 60 wks pp. Because of the difference in time of the first ovulation (OVU1) between parities (Opsomer et al., 1999), the number of weeks between parturition and first ovulation of a cow (i) was determined by a lognormal distribution (Table 2). After the first ovulation, the following ovulations (OVUij) were modeled to occur, based on a probability (POVUi) (Table 1). The duration of the ovulation interval was assumed to be at least 3 wks. Disorders (CO) that occurred in lactation were assumed to interfere with normal cyclicity and were included in the model. These disorders were assumed to occur not more than once per lactation and to follow a uniform distribution (Table 2). After the occurrence of these postpartum disorders (COij) at wk j, a fixed delay of 4 wks in ovulation was modeled (Table 2). 14
Chapter 2 if COij = 1, COij-1 = 1, COij-2 = 1, and COij-3 = 1 then 0 if j < OVU1 then 0 if OVU ij-1 = 1, OVU ij-2 = 1 then 0 OVUij = else discrete[(1, 0), (POVU i , 1-POVU i )] i = 1, ..., n j = 1, ..., 60
[4]
The estrus detection rate of an average cow in the herd (Pest) was given (Table 1) and adjusted for LVi and MYij (PestAdjij). In our model, the estrus detection rate that was included was somewhat higher for cows with a lower LV i (using a factor FLV) and for cows with a lower actual milk yield at ovulation. PestAdjij at the first ovulation was 50% of Pest. After the peak of milk yield (6 wks pp), PestAdj ij increased with a decrease of MYij, (Harrison et al., 1990, López et al., 2004). For each cow i (i=1, …, n) in wks pp j (j=1, …, 60) the PestAdjij was calculated as follows: if OVUij = 0 then 0 if j = OVU1 then 0.50 × PEst × FLV if OVU1 < j < 6 weeks then PEst × FLV [5] PEst × FLV × MYi6 PEstAdjij = else MYij i = 1, ..., n j = 1, ..., 60 MYi6 denotes the peak milk yield of the lactation curve in Wood’s function. Cows were detected in estrus (EDij) with a certain probability (PestAdjij) at wk j pp. The FLV was 1.1, 1, and 0.9 for cows (i) having LVi < 0.9, 0.9 to1.1 and > 1.1, respectively. Cows were assumed to be inseminated (INSij) at the first detected estrus after the voluntary waiting period. if EDij = 0 then 0 if EDij = 1 and j < VWP then 0 [6] INSij = else EDij = 1 and j > VWP then 1 i = 1, ..., n j = 1, ..., 60 Conception rate of an average cow in a herd (Pcon) was given (Table 1). Conception of individual cows (Cij) was dependent on wks pp (Pconij), parity (FP) 15
Economic consequences of reproductive performance (Windig et al., 2005), LV (FLV)) (Dhaliwal et al., 1996), and postpartum disorders such as metritis (ME). if INSij = 0 then 0 if j < 5 then 0 PCon × FP × FLV × FME [7] PCon ij = else j > 5 then (41.5 - 345 × 0.6 j ) × 0.37 i = 1, ..., n j = 1, ..., 60 The FP is, respectively, 1.05 and 1 for parity 1 and parity ≥ 2. The values for FLV were the same as for the calculation of PestAdjij. FME is the adjustment factor of metritis on conception given the value of 0.81 in the occurrence of ME and 1 without the occurrence of ME (Table 2) (Loeffler et al., 1999). After conception, some cows might experience embryonic death (ED) (Table 1) (Silke et al., 2002) which prolonged the reproductive cycle for 6 - 8 wks before returning to normal cyclicity. After pregnancy, the subsequent calving interval (CIi) was calculated using a normal distribution function for the length of gestation (G i) (Table 2). If the cow was still not pregnant at 60 wks pp, the simulation process was stopped and the cow was kept open. Assuming a fixed dry period of 8 wks, the daily milk yield of cow (i) in wk (j) (MYij) was estimated up to 8 wks before the expected calving date. MY ij was adjusted (Myadjij) when the cow became pregnant (Olori et al., 1997). MYij - MYlossik MYlossik = 0.18 [exp(0.03 × k)] MYadjij = j = 1, ..., week at drying off k = 5, ..., 42 i = 1, ..., n
[8]
Mylossik represents the milk production loss of the cow (kg) at wk k of pregnancy from the 5th wk of pregnancy onwards. The milk production per cow per year (MP/cow/yr(i)) was calculated as well as the number of calves per cow per year (Noc/cow/yr(i)) and the number of inseminations (Noai(i)) needed.
16
POVUi Pest Pcon CO ME ED VWP
Ovulation rate after the 1st ovulation Estrus detection rate Conception rate
Incidence rate of disorders prolonging ovarian cyclicity Incidence rate of disorders reducing conception rate Incidence rate of embryonic death Voluntary waiting period (wks)
Abbreviation
0.03 0.20 0.05 9
Good 1.00 0.70 0.70
17
0.07 0.27 0.07 12
Scenarios Average 0.95 0.50 0.50 0.11 0.33 0.09 15
Poor 0.90 0.30 0.30
Kyle et al. (1992); Villa-Godoy et al. (1990) Grimard et al. (2005); Villa-Godoy et al. (1990) Dhaliwal et al. (1996); Melendez and Pinedo (2007); Windig et al. (2005) Hooijer et al. (1999); Laporte et al. (1994) Loeffler et al. (1999) Silke et al. (2002) Arbel et al. (2001); Bertilsson et al. (1997); Schindler et al. (1991)
Source
Table 1. Default input values for the variables representing reproduction dynamics for “good”, “average” and “poor” reproductive performance scenario.
Chapter 2
Opsomer et al. (1999) Opsomer et al. (1999) De Vries et al. (2006) Hooijer et al. (1999); Laporte et al. (1994) Authors Authors
1 0.126 0.021 0.18(exp(0.093* wks of gestation) LogNorm(5.36, 5.04), Truncate(1, 10) LogNorm(4.53, 3.08), Truncate(1, 10) Norm(40.00, 0.86), Truncate( 38.29, 41.71) Uniform(6, 14) 4 Uniform(6, 8)
A B C Mylossik
Gi j
Cp Cai Ccm
-1265.41+1730*LV -1511.93+2133*LV -1453.09+2080*LV -1384.01+1983*LV -1309.77+1864*LV 100 20 152
0.12
Olori et al. (1997)
0.32 0.25 0.18 0.11 0.14 Norm(1, 0.1)
P1 P2 P3 P4 P5 LVi
Cmp RPOi
CRV (2009) CRV (2009) CRV (2009)
8310 8008
HMP
Authors Authors Authors
Authors Houben et al. (1994)
Authors CRV (2009)
Source
Default value
Abbreviation
b
18
Time of the occurrence and delay for the next ovulation caused by cystic ovary. The calving management cost was calculated on average per cow by summing the labor cost during calving, the cost of periparturient and postpartum disorders, the cost of drug delivery and dry off treatment cost
a
Parameters Cow factors Milk production potential (kg/cow/305 days) Average rolling herd milk production per year (kg) Proportion of cows in herd Parity 1 Parity 2 Parity 3 Parity 4 Parity 5+ Lactation value Wood’s curve Factor A Factor B Factor C Milk production loss per week (kg) st The 1 ovulation time (wks) Parity 1 Parity 2+ Gestation period (wks) Week of lactation stage to occur disordera Recovery period (wks)a Delay of normal cyclicity by the effect of embryonic death (wks) Economic values Cost of lower milk production (€/kg) Retention pay-off (€/cow) Parity1 Parity 2 Parity 3 Parity 4 Parity 5+ Calf price (€/head) Insemination cost (€/service) Calving management cost (€/calving)b
Table 2. Default input values for cow factors and their relation with reproductive performance.
Economic consequences of reproductive performance
Chapter 2 Simulation and economic calculations The reference situation for the simulation in this study was a good situation with an ovulation rate, an estrus detection rate, a conception rate, CO, ME, ED and VWP as provided in Table 1. The voluntary waiting period in this good reproductive performance scenario was set at 9 wks pp. As stated before, for every cow, a good reproductive performance scenario was calculated together with the other two reproductive performance scenarios, based on user input. The performance of a cow, represented by MP/cow/yr(i)), Noc/cow/yr(i), and Noai(i) in “average” and “poor” situation, was compared with the performance of the same cow in “good” situation. The economic effect of reproductive performances was generated using the partial budget approach similar to the stochastic model of Steeneveld, et al.(2007) who evaluated the economic effect on treatment of chronic subclinical mastitis comparing the situation with and without treatment (Steeneveld et al., 2007). Change in the sum of additional revenues or revenues forgone (milk production, culling, calves), and reduced costs or extra costs (AI, calving management) were calculated as the basis for net economic loss (NELi) by comparing the “average” and “poor” reproductive performance scenarios with the good reproductive performance scenario in the same cow.
[9] ∆MPi, ∆Ci, ∆AIi represent the loss of MP/cow/yr(i), Noc/cow/yr(i) and Noai(i) of the “average” and “poor” reproductive performance scenario compared to the good reproductive performance scenario, while Cmp, Cp, Ccm, and Cai indicate costs of lower milk production, calf price, costs of calving management and costs of insemination, respectively. The costs of calving management depended on Noc/cow/yr(i) and the costs of treatment, time and labor costs during the calving period. The treatment costs were, on average per cow, treatment of peripartum and postpartum disorders in the Netherlands situation, plus drying off treatment, labor costs during calving and the costs of drug delivery. Because cows that are not pregnant after 60 wks pp also give economic losses, these losses were approached by considering these cows to be culled. Consequently, the economic damage of culling was taken as the retention pay-off. Retention pay-off (RPOi) was defined as the difference in expected future net revenue between a culled cow and its replacing heifer, including the cost of buying and raising this replacing heifer. RPOi was dependent on LVi and Pi, and was based on a study of Houben, et al. (1994), adjusting for the current situation (Table 2) (Houben et al., 1994).
19
Economic consequences of reproductive performance Input data Input data concerning milk production, fertility and economics were based on reports and reviewed literature for circumstances in the Netherlands. When necessary, authors’ expertise was used (Table 2). Three realistic reproductive performance scenarios, representing poor, average and good reproductive performance were modeled. These scenarios were based on reported fertility data in the literature (Table 1). In general, the input data included parameters on milk production (HMP, LVi, Pi,), probabilities necessary to simulate the dynamics of a reproductive cycle (OVU1, POVUi, Pest, Pcon, CO, ME, ED; Table 1) and economics (Cmp, Cp, Ccm, Cai, RPOi; Table 2). Costs of a lower milk production were estimated to be €0.12 per kg (Huijps and Hogeveen, 2007). These costs represent the marginal costs of keeping extra cows to fill the milk quota. The model used to calculate RPO figures was recently updated to represent current market circumstances. The incidence (CO) and the timing of the occurrence of postpartum disorder prolonging cyclicity was based on data of cystic ovarian disease (Hooijer et al., 1999, Laporte et al., 1994). The incidence of postpartum disorders reducing conception was based on metritis (Loeffler et al., 1999). Owing to the increase of milk yield during the first weeks of lactation, the negative energy balance might have a negative effect on fertility (Butler and Smith, 1989). The association between Pest and milk production level (Harrison et al., 1990, López et al., 2004, Silvia, 2003) and the association between Pcon, the milk production level and lactation stage (Grimard et al., 2006, Mayne et al., 2002) are reported in the literature and they were included in the model, adjusting and incorporating data in the light of authors’ expertise if the existing literature data available were insufficient to run the dynamics of a reproductive cycle. The Mylossik described in [8] was a conversion of data reported in the literature (Olori et al., 1997). Economic data were deterministic. Validation Because of a lack of suitable data for validation, external validation was not possible and several other methods of verification were used, e.g. the rationalism method and face validity (Sørensen, 1990). With the rationalism method, several scenarios of inputs were used and checked for consistency and credibility of the model output. Experts in the field of fertility gave feedback on the assumptions and the credibility of the relationship between input, processing and output of the model in face validity. Our results were compared with the performance of Dutch dairy cows (Berends et al., 2008, Windig et al., 2006) to guarantee the credibility of the simulation model.
20
Chapter 2 Sensitivity analysis In the sensitivity analysis, outputs resulting from the “average” scenario were compared to the outputs based on alternative values (good and poor). To assess the importance of individual factors on the economic outcome, only one factor was changed at a time and run in the simulation. The minimum and maximum value of this factor, e.g. conception rate, estrous detection rate, postpartum disorders rate (CO and ME), embryonic death rate or VWP, was compared with its average value in “average” scenario to calculate the NELi. RESULTS Technical and economic results in various scenarios Average annual performances of cows in the three defined scenarios (good, average and poor) are given in Table 3. The average calving interval in the good situation was 362 days, achieved after the first AI at 10.5 wks pp on average. All parameters affecting the reproductive performance of the cows were worse in the “average” and “poor” scenarios compared to the “good” reproductive performance scenario. For example, the average calving interval deteriorated to 407 and 507 days for the “average” and “poor” reproductive performance scenario, respectively (Table 3). Within scenarios there was a large variation between cows. For example, the calving interval of individual cows in the average scenario varied from 365 days (5% percentile) to 490 days (95 % percentile). Because of the longer calving intervals, the milk production decreased from 8068 kg/year in the good reproductive performance scenario to 7031 kg/year in the poor reproductive performance scenario. The dynamics of conception in our model, related to days in milk, resulted in a conception rate in the second insemination of 32, 63, and 89% for the “poor”, “average” and “good” reproductive performance scenarios, respectively. The annual net economic losses (€ per cow) of the “average” and “poor” reproductive performance scenarios related to the “good” scenario are shown in Table 4. The mean net economic losses per cow per year were €34 and €231 for the “average” and “poor” reproductive performance scenario, respectively. For an “average” scenario, the economic loss caused by decreased milk production was the most important. The economic loss caused by increase a number of non-pregnant cows became the most important in a “poor” scenario. The cost of decreased milk production accounted on average for 100% and 52% of the total net economic losses for the “average” and “poor” reproductive performance scenario, respectively. The cost for increase of non-pregnant cows accounted on average for 55% for the “poor” reproductive performance scenario. Compared to the “good” reproductive performance scenario, the net economic loss per day of increase in calving interval was calculated and included the economic losses and benefits influenced by the increase in calving interval (the economic losses of milk production, selling calves and 21
Economic consequences of reproductive performance the cost of calving management), given the average of economic losses of €0.57, and €0.70 per cow per day in “average” and “poor” reproductive performance scenarios, respectively (Table 4). For each of the reproductive performance scenarios there was a large variation in losses as is illustrated by the 5% and 95% percentiles presented in Table 4 and in Fig. 1. In comparison with other parities, the cows in parity 1 had, in general, a lower milk production, a higher milk persistency and a better conception rate (Table 5). Although the difference in calving intervals between parities did not differ by more than 2 days, the net economic losses, increased with increasing parity for all scenarios. Sensitivity analysis An increase of the VWP extended the calving interval and reduced milk production per year. Varying the input value of CO compared to the “average” scenario had limited effect on the fertility parameters (Fig. 2). An increase of ME and ED reduced conception rate, extended open days, and increased the number of nonpregnant cows. Lower Pest and Pcon resulted in an extended calving interval, a reduction in milk production per year, and an increase of non-pregnancy rate (up to 0.21). The most important variable influencing the annual net economic loss was the conception rate, followed by the estrus detection rate, the length of the VWP, the incidence rate of ME, the incidence rate of ED, the ovulation rate after the first ovulation and the incidence rate of CO (Fig. 2). The net annual economic loss increased with decreasing value of fertility parameters. An improvement of the conception rate from 0.30 to 0.50 resulted in a reduction of the net annual economic loss of €75.54 per cow per year. An improvement of the conception rate from 0.50 to 0.70 resulted in a reduction of the net annual economic loss of €16.69 per cow per year. An improvement in the estrus detection rate from 0.30 to 0.50 and from 0.50 to 0.70 resulted in a reduction of the net economic loss of, respectively, €53.29 and €11.20 per cow per year. DISCUSSION Several economic models have reported the economics of fertility (Boichard, 1990, Meadows et al., 2005, Oltenacu et al., 1980, Plaizier et al., 1998) but they failed to evaluate the dynamics of interactions between reproductive performance and milk production within the individual cow. The model presented in this study is novel because the stochastic dynamic reproductive cycle was dependent on various cow factors and milk production levels. This provides insight into the variation of reproductive performance in an individual cow during the complete lactation. On a farm, reproductive performances of cows are not independent as, e.g., estrus detection will increase when more animals are in heat around the same day. However, the exact interactions between cows on a farm are largely unknown and, therefore, its modeling 22
Chapter 2 will be highly speculative. We therefore decided to perform calculations at cow level only. Simulation studies that nevertheless did model at the herd level also kept the reproductive performance of single cows independent from each other. Because we took the individual cow characteristics, such as milk production level and parity from distributions representing the average Dutch dairy farm, the average model outcomes do represent the average farm. An estimate of the average net economic losses on a farm can be calculated by multiplying the net economic losses per cow in a specific scenario with the number of cows on a farm. No data were available for external validation of our model but internal validation techniques, such as sensitivity analysis, rationalism and face validation, are assumed to be suitable alternatives to check a simulation model for credibility (Sørensen, 1990). In this model, the effects of changes of input parameters such as the conception rate, were compared with the original situation and expected changes and found adequate. By this means, the process and its outcomes were checked. The most recent study using data from the Royal Dutch Cattle Syndicate (Coöperatie Rundvee Verbetering, CRV) reported an average interval of calving to first AI, days open, calving interval and numbers of inseminations per conception of 87 to 93, 125 to 128, 405 to 412, and 1.6 to 1.7, respectively (Berends et al., 2008). These results are in the predicted range by our model (Table 3). Moreover, the average milk production per cow per year (Table 3) fitted closely with the findings of Windig et al.(2006). Table 3. Reproductive performance of dairy cows with a “good”, “average” and “poor” reproductive performance scenario (5% to 95% percentiles in parentheses). Technical performances
Good
Average
Poor
Week to start AI
10.5 (9-13) 83 (67-130) 362 (342-408) 1.16 (1-2) 8068 (6329-9831) 36.32 (26.55-45.66) -0.095 (-0.129-(-0.056)) 1.02 (0.90-1.08) 0.00a
14.5 (12-20) 127 (88-207) 407 (365-490) 1.61 (1-4) 7775 (6188-9438) 36.32 (26.55-45.66) -0.091 (-0.125-(-0.051)) 0.91 (0.75-1.01) 0.00a
21.7 (15-35) 228 (116-389) 507 (394-670) 2.76 (1-6) 7031 (5441-8790) 36.16 (26.39-45.46) -0.086 (-0.120-(-0.047)) 0.74 (0.55-0.93) 0.21a
Days open Calving interval No. AI/cow MP/cow/yr (kg) Peak milk yield (kg/day) Milk persistency (kg/day) No.Calves/cow/yr Non-pregnancy ratea a
Percentile was not calculated because it is irrelevant for rate of non-pregnancy.
23
Economic consequences of reproductive performance Table 4. Annual economic consequence (€/cow) of a poor and average reproductive performance scenario related to a “good” reproductive performance scenario (5% to 95% percentiles in parentheses).
a
Economic losses of
Average
Poor
Milk production
35 (0-119)
120 (15-292)
Selling calves
11 (0-28)
27 (7-48)
Non-pregnant cows
0 (0-0)
128(0-749)
AI costs
5 (-19-32)
20 (-12-80)
Calving management cost
-16 (-43-0)
-42(-73-(-11))
Total Net
34 (-19-131)
231 (7-806)
Net cost per day of increase in calving intervala
0.57 (0-1.04)
0.70 (0.21-1.08)
Excluding the economic losses of culling cost and AI cost in calculation.
Figure 1. Distribution of annual net economic losses (€/cow) over cows within a “poor”, and “average” reproductive performance scenario related to the “good” reproductive performance scenario.
24
Chapter 2 Table 5. Reproductive performance and net economic loss due to reproduction (€/cow) in various parities of “poor” and “average” reproductive performance scenario related to a “good” reproductive performance scenario (5% to 95% percentiles in parentheses). Performances Calving interval
No. AI/cow
MP/cow/yr (kg)
Peak milk yield (kg/day) Milk persistency (kg/day) Total Net (`€/cow)
Parity 1 2 3+ 1 2 3+ 1 2 3+ 1 2 3+ 1 2 3+ 1 2 3+
Good 363 (342-412) 362 (341-403) 362 (342-406) 1.16 (1-2) 1.16 (1-2) 1.16 (1-2) 7125 (5939-8313) 8206 (6807-9625) 8682 (7164-10138) 29.54 (24.79-34.35) 37.40 (31.26-43.60) 40.68 (33.73-47.31) -0.061 (-0.069-(-0.052)) -0.101 (-0.116-(-0.086)) -0.116 (-0.134-(-0.099)) _ _ _
Average 407 (365-490) 407 (365-494) 406 (365-488) 1.57 (1-4) 1.64 (1-4) 1.62 (1-4) 6987 (5803-8168) 7873 (6477-9288) 8297 (6780-9778) 29.54 (24.79-34.35) 37.40 (31.26-43.60) 40.68 (33.73-47.31) -0.057 (-0.066-(-0.048)) -0.097 (-0.112-(-0.082)) -0.112 (-0.130-(-0.094)) 15 (-19-73) 39 (-19-141) 46 (-19-153)
Poor 508 (394-674) 506 (396-666) 507 (394-669) 2.73 (1-6) 2.80 (1-6) 2.77 (1-6) 6523 (5250-7799) 7077 (5471-8723) 7374 (5698-9096) 29.40 (24.64-34.11) 37.20 (31.18-43.47) 40.49 (33.59-47.06) -0.052 (-0.062-(-0.043)) -0.092 (-0.108-(-0.077)) -0.108 (-0.125-(-0.090)) 167 (3-667) 260 (20-886) 262 (26-837)
Figure 2. Changes in annual net economic losses (€/cow) when the value of the single input parameters (ovulation rate after the first ovulation, estrus detection rate, conception rate, incidence rate of postpartum disorders prolonging normal cyclicity (CO; cystic ovary), incidence rate of postpartum disorders reducing conception rate (ME; metritis), embryonic death (ED; embryonic death) and voluntary waiting period (VWP)) were changed in the “average” scenario (5% to 95% percentiles in parentheses).
25
Economic consequences of reproductive performance A large variation in the average number of inseminations is reported for the situation in the Netherlands, as Berends et al. (2008) found 1.6 to 1.7 inseminations per conception while Windig et al. (2006) found 1.8 to 2.5 inseminations per cow per lactation. The outcome of our model was in between the results of the studies mentioned above, while we also found large variation in technical performances and economic losses (Table 3 and 4). Differences in number of inseminations per conception as well as the average calving interval can be caused by deciding when to cull a cow with fertility problems. In our model, because the model runs for 60 wks pp (which is a long time), we could not calculate the economic consequences of cows that were not pregnant at that time. To approach the economic consequences, cows not pregnant at 60 wks pp were considered to be culled and the economic damage of culling was taken as costs. In a normal farming situation, cows failing to conceive within a reasonable amount of time are being culled. The moment that cows are culled depend on the specific farm situation, such as the availability and price of pregnant heifers. Modeling this process would require complex systems that take away the attention of physiological effects, such as the calving interval, of not optimal fertility. By using Monte Carlo simulation instead of optimization modeling methods, such as dynamic programming, we were allowed to model the complex interrelationship between cow factors and fertility. The calf productivity in our model might differ from the real situation, because it accounted for only non-pregnant cows as being culled. In reality, cows are also culled for diseases, longer periods of days open and the influx of available pregnant heifers (Dijkhuizen et al., 1985). These culling reasons were not included in our model. When cows are culled due to the influx of available pregnant heifers, it will not be different from the calf productivity in our model. The fact that we did not included culling of diseased cows or cows with a too many days open most likely results in an over-estimation of calf productivity in our study. Because the losses due to calf productivity are a small proportion of the total costs, this will not affect the total costs of not optimal fertility very much. In this study the “good” reproductive performance scenario was used as a base line or reference to calculate the economic loss of an “average” and “poor” reproductive performance and was, therefore, a hypothetical situation. In some specific circumstances, the “average” reproductive scenario was economically better than a “good” scenario due to the effect of ED given the differences of reproductive cycles. The high annual net economic loss in a cow showing poor reproductive performance were mainly caused by high costs of non-pregnant cows. When interpreting these types of loss, it is important to note that the absolute amount of the economic loss (loss of a reproductive performance in relation to a good reproductive performance) are not of much practical value but were needed in our partial budgeting approach. The annual net economic effects found reflect the total economic losses. This makes the losses due to not “good” reproductive performance comparable with 26
Chapter 2 net economic losses due to production diseases. For instance, the average estimated economic loss due to subclinical mastitis was estimated to be €89 per average cow per year (Huijps et al., 2009). This is higher than the average net economic loss due to reproduction of €34 per average cow per year found in this study. Comparisons between two scenarios are useful to estimate avoidable losses. For example, by improving the fertility of a farm from “average” to “good”, a farmer may save €34 per cow per year, which is €2210 per year for an average farm of 65 lactating dairy cows in the Netherlands. Economic losses due to non-optimal fertility are often expressed in costs per day of increased calving interval. In our model, the net cost of a one day increase in calving interval depended on the reproductive performance scenario, on average, from €0.57 per cow per day for the “average” scenario to €0.70 per cow per day for the “poor” scenario. These figures include costs for lower milk production per day, lower number of calvings and lower costs for calving management. However, they do not include costs for non-pregnant cows because these only concern cows that do conceive. Moreover, costs for additional inseminations are not taken into account. This means that the cost per day of increased calving interval is lower than the cost for decreased fertility. Because the net economic loss increases relatively more than proportionally with an increase in calving interval, the increase in cost per day can become much higher than that mentioned (€0.57 to €0.70), ranging from €0.76 to €1.95 per cow per day. Other studies have reported costs of one day extra for days open and calving interval as follows: €0.06 to1.03 (Groenendaal et al., 2004), €0.28 to 1.10 (Meadows et al., 2005), €0.47 to €0.79 (De Vries and Conlin, 2003), €2.07 (Veerkamp et al., 2002) and €2.95 (Plaizier et al., 1997). Differences between our results and the results in other publications may be caused by a difference in factors taken in to account. For instance, by extending the calving interval, there will be fewer problems caused by calving. In fact this is a benefit of an increased calving interval. Moreover, price levels may differ between countries. For instance, the loss due to milk production decrease (kg/cow/year) was evaluated at €0.12 per kg, under Dutch quota circumstances. In a situation without a milk quota, milk production losses have a much higher economic impact, depending on the milk price. Thus, the methodological differences, the differences in the economic context between areas that are also time dependent are considered to explain the differences with these studies (Seegers et al., 1994). The sensitivity analysis showed clearly that changes in estrus detection and conception rate have a considerable impact on the mean number of days open and the calving interval. Consequently, the conception rate, estrus detection rate, voluntary waiting period, incidence rate of ME, and incidence rate of ED do considerably affect the annual net economic losses (Fig. 2). Clearly, metritis and embryonic death reduce pregnancy rate (Loeffler et al., 1999, Santos et al., 2009, Silke et al., 2002). Several cow risk factors cause ME and ED (Bruun et al., 2002, Santos et al., 2009). Efforts to 27
Economic consequences of reproductive performance decrease those risk factors and farmer’s management skills can result in an improvement in the incidence rate. The economic value of a 1% increase of conception was estimated to be €1.22 per cow per year (van Arendonk and Dijkhuizen, 1985). Similarly, the economic effect of an increase in the estrus detection of 1% ranged from €1.81 to €14.10, depending on the re-breeding program (Plaizier et al., 1998), and from €0.50 to €1.74, depending on the level of estrus detection rate (De Vries and Conlin, 2003). The economic importance of estrus detection and pregnancy rates are commonly described by others, and are still of practical value for farmers (De Vries and Conlin, 2003, Oltenacu et al., 1981, Plaizier et al., 1998) The observed economic consequences of extending the VWP was also reported by others (Sørensen and Østergaard, 2003). The VWP is an interesting variable because it relates to a direct management decision of the dairy farmer. This is in contrast with variables such as conception rate and estrus detection which are influenced by the dairy farmer’s skills. Because of the effects of milk production and persistency of the lactation curve on the costs of an increased calving interval, the net economic effect of an extended VWP may differ from cow to cow. To be more precise, for specific cows an extended VWP might even be beneficial under some circumstances (Bertilsson et al., 1997, López et al., 2004, Mayne et al., 2002). The differences between cows shown in our study as well as in other studies (Allore and Erb, 2000, Arbel et al., 2001, Bertilsson et al., 1997), demonstrated that the outcomes of a study cannot be generalized. Thus, the length of the VWP should be tailor made and depends on, e.g. age, production level and persistency.
CONCLUSIONS Our stochastic dynamic simulation model was suited for the study of the economic consequences of non-optimal fertility of a dairy cow. Non-optimal fertility, which was defined in our study as an “average”, or “poor” reproductive performance, resulted in average net economic losses compared to a “good” situation of €34 and €231 per cow per year, respectively. These losses are mainly caused by decreased milk production and increased non pregnant cows, especially in the situation of poor fertility. Net economic losses due to reproduction become higher with increasing parity. Finally, conception rates and estrus detection rates do have the largest effect on the loss of open days postpartum and hence on the calving interval. ACKNOWLEDGEMENTS The authors grateful acknowledge Edward Hopkin for his editorial comments.
28
Chapter 2 REFERENCES Allore, H. G. and H. N. Erb. 2000. Simulated effects on dairy cattle health of extending the voluntary waiting period with recombinant bovine somatotropin. Prev. Vet. Med. 46(1):29-50. Arbel, R., Y. Bigun, E. Ezra, H. Sturman, and D. Hojman. 2001. The effect of extended calving intervals in high lactating cows on milk production and profitability. J. Dairy Sci. 84(3):600-608. Berends, I. M. G. A., W. A. J. M. Swart, K. Frankena, J. Muskens, T. J. G. M. Lam, and G. van Schaik. 2008. The effect of becoming BVDV-free on fertility and udder health in Dutch dairy herds. Prev. Vet. Med. 84(1-2):48-60. Bertilsson, J., B. Berglund, G. Ratnayake, K. Svennersten-Sjaunja, and H. Wiktorsson. 1997. Optimising lactation cycles for the high-yielding dairy cow. A European perspective. Livest. Prod. Sci. 50(1-2):5-13. Boichard, D. 1990. Estimation of the economic value of conception rate in dairy cattle. Livest. Prod. Sci. 24(3):187-204. Bruun, J., A. K. Ersbøll, and L. Alban. 2002. Risk factors for metritis in Danish dairy cows. Prev. Vet. Med. 54(2):179-190. Butler, W. R. and R. D. Smith. 1989. Interrelationships between energy balance and postpartum reproductive function in dairy cattle. J. Dairy Sci. 72(3):767-783. CRV. 2009. CRV(Coöperatie Rundvee Verbetering) Annual Reports 2009. https://www.cr-delta.nl/servlets/dbupload?id=6161 Accessed May 1, 2010. De Vries, A. 2004. Trends in reproductive performance in dairy cows: What do the numbers tell us? Pages 1-8 in Proc. of 2004 Florida Dairy Reproduction Road Show, Florida. De Vries, A. and B. J. Conlin. 2003. Economic value of timely determination of unexpected decreases in detection of estrus using control charts. J. Dairy Sci. 86(11):3516-3526. De Vries, A., M. B. Crane, J. A. Bartolome, P. Melendez, C. A. Risco, and L. F. Archbald. 2006. Economic comparison of timed artificial insemination and exogenous progesterone as treatments for ovarian cysts. J. Dairy Sci. 89(8):3028-3037. Dhaliwal, G. S., R. D. Murray, and H. Dobson. 1996. Effects of milk yield, and calving to first service interval, in determining herd fertility in dairy cows. Anim. Reprod. Sci. 41:109-117. Dijkhuizen, A. A., J. Stelwagen, and J. A. Renkema. 1985. Economic aspects of reproductive failure in dairy cattle. I. Financial loss at farm level. Prev. Vet. Med. 3(3):251-263. Gonzalez-Recio, O., M. A. Perez-Cabal, and R. Alenda. 2004. Economic value of female fertility and its relationship with profit in Spanish Dairy Cattle. J. Dairy Sci. 87(9):3053-3061. Grimard, B., S. Freret, A. Chevallier, A. Pinto, C. Ponsart, and P. Humblot. 2006. Genetic and environmental factors influencing first service conception rate and late embryonic/foetal mortality in low fertility dairy herds. Anim. Reprod. Sci. 91(1-2):3144. Grimard, B., S. Freret, H. Seegers, C. Ponsart, A. Chevallier, and P. Humblot. 2005. Sensitivity and specificity of visual oestrus detection 3 weeks after first AI in low fertility dairy herds in France. Reprod. Domest. Anim. 40:358.
29
Economic consequences of reproductive performance Groenendaal, H., D. T. Galligan, and H. A. Mulder. 2004. An economic spreadsheet model to determine optimal breeding and replacement decisions for dairy cattle. J. Dairy Sci. 87(7):2146-2157. Hare, E., H. D. Norman, and J. R. Wright. 2006. Trends in calving ages and calving intervals for dairy cattle breeds in the United States. J. Dairy Sci. 89(1):365-370. Harrison, R. O., S. P. Ford, J. W. Young, A. J. Conley, and A. E. Freeman. 1990. Increased milk production versus reproductive and energy status of high producing dairy cows. J. Dairy Sci. 73(10):2749-2758. Hooijer, G. A., K. Frankena, M. M. H. Valks, and M. Schuring. 1999. Treatment of cystic ovarian disease in dairy cows with gonadotrophin-releasing hormone: A field study. Vet. Q. 21(1):33-37. Houben, E. H., R. B. Huirne, A. A. Dijkhuizen, and A. R. Kristensen. 1994. Optimal replacement of mastitic cows determined by a hierarchic Markov process. J. Dairy Sci. 77(10):2975-2993. Huijps, K., S. De Vliegher, T. Lam, and H. Hogeveen. 2009. Cost estimation of heifer mastitis in early lactation by stochastic modelling. Vet. Microbiol. 134(1-2):121-127. Huijps, K. and H. Hogeveen. 2007. Stochastic modeling to determine the economic effects of blanket, selective, and no dry cow therapy. J. Dairy Sci. 90(3):1225-1234. Jorritsma, R., H. Jorritsma, Y. H. Schukken, and G. H. Wentink. 2000. Relationships between fatty liver and fertility and some periparturient diseases in commercial Dutch dairy herds. Theriogenology 54(7):1065-1074. Kyle, S. D., C. J. Callahan, and R. D. Allrich. 1992. Effect of progesterone on the expression of estrus at the first postpartum ovulation in dairy cattle. J. Dairy Sci. 75(6):1456-1460. Laporte, H. M., H. Hogeveen, Y. H. Schukken, and J. P. T. M. Noordhuizen. 1994. Cystic ovarian disease in Dutch dairy cattle, I. Incidence, risk factors and consequences. Livest. Prod. Sci. 38(3):191-197. Loeffler, S. H., M. J. de Vries, Y. H. Schukken, A. C. de Zeeuw, A. A. Dijkhuizen, F. M. de Graaf, and A. Brand. 1999. Use of AI technician scores for body condition, uterine tone and uterine discharge in a model with disease and milk production parameters to predict pregnancy risk at first AI in Holstein dairy cows. Theriogenology 51(7):12671284. López, H., L. D. Satter, and M. C. Wiltbank. 2004. Relationship between level of milk production and estrous behavior of lactating dairy cows. Anim.Reprod. Sci. 81(34):209-223. Lucy, M. C. 2001. Reproductive loss in high-producing dairy cattle: where will it end? J. Dairy Sci. 84(6):1277-1293. Marsh, W. E., A. A. Dijkhuizen, and R. S. Morris. 1987. An economic comparison of four culling decision rules for reproductive failure in United States dairy herds using DairyORACLE. J. Dairy Sci. 70(6):1274-1280. Mayne, C. S., M. A. McCoy, S. D. Lennox, D. R. Mackey, M. Verner, D. C. Catney, W. J. McCaughey, A. R. G. Wylie, B. W. Kennedy, and F. J. Gordon. 2002. Fertility of dairy cows in Northern Ireland. Vet. Rec. 150(23):707-713. Meadows, C., P. J. Rajala-Schultz, and G. S. Frazer. 2005. A spreadsheet-based model demonstrating the nonuniform economic effects of varying reproductive performance in Ohio dairy herds. J. Dairy Sci. 88(3):1244-1254.
30
Chapter 2 Melendez, P. and P. Pinedo. 2007. The association between reproductive performance and milk yield in Chilean Holstein cattle. J. Dairy Sci. 90(1):184-192. Olori, V. E., S. Brotherstone, W. G. Hill, and B. J. McGuirk. 1997. Effect of gestation stage on milk yield and composition in Holstein Friesian dairy cattle. Livest.Prod.Sci. 52(2):167-176. Oltenacu, P. A., R. A. Milligan, T. R. Rounsaville, and R. H. Foote. 1980. Modelling reproduction in a herd of dairy cattle. Agri. Sys. 5(3):193-205. Oltenacu, P. A., T. R. Rounsaville, R. A. Milligan, and R. H. Foote. 1981. Systems analysis for designing reproductive management programs to increase production and profit in dairy herds. J. Dairy Sci. 64(10):2096-2104. Opsomer, G., M. Coryn, and A. De Kruif. 1999. Measurement of ovarian cyclicity in the postpartum dairy cow by progesterone analysis. Reprod. Domest. Anim. 34(3-4):297300. Opsomer, G., M. Coryn, H. Deluyker, and A. De Kruif. 1998. An analysis of ovarian dysfunction in high yielding dairy cows after calving based on progesterone profiles. Reprod. Domest. Anim. 33(3-4):193-204. Plaizier, J. C. B., G. J. King, J. C. M. Dekkers, and K. Lissemore. 1997. Estimation of economic values of indices for reproductive performance in dairy herds using computer simulation. J. Dairy Sci. 80(11):2775-2783. Plaizier, J. C. B., G. J. King, J. C. M. Dekkers, and K. Lissemore. 1998. Modeling the relationship between reproductive performance and net-revenue in dairy herds. Agri. Sys. 56(3):305-322. Pryce, J. E., M. D. Royal, P. C. Garnsworthy, and I. L. Mao. 2004. Fertility in the highproducing dairy cow. Livest. Prod. Sci. 86:125-135. Royal, M. D., A. O. Darwash, A. P. F. Flint, R. Webb, J. A. Woolliams, and G. E. Lamming. 2000. Declining fertility in dairy cattle: Changes in traditional and endocrine parameters of fertility. Anim. Sci. 70(3):487-501. Santos, J. E. P., H. M. Rutigliano, and M. F. S. Filho. 2009. Risk factors for resumption of postpartum estrous cycles and embryonic survival in lactating dairy cows. Anim. Reprod. Sci. 110(3-4):207-221. Schindler, H., S. Eger, M. Davidson, D. Ochowski, E. C. Schermerhorn, and R. H. Foote. 1991. Factors affecting response of groups of dairy cows managed for different calving-conception intervals. Theriogenology 36(3):495-503. Seegers, H., C. Fourichon, X. Malher, and M. L'Hostis. 1994. A framework for animal health management. Vet. Res. 25(2-3):165-173. Silke, V., M. G. Diskin, D. A. Kenny, M. P. Boland, P. Dillon, J. F. Mee, and J. M. Sreenan. 2002. Extent, pattern and factors associated with late embryonic loss in dairy cows. Anim. Reprod. Sci. 71(1-2):1-12. Silvia, W. J. 2003. Addressing the decline in reproductive performance of lactating dairy cows: a researcher's perspective. Vet. Sci. Tomorrow 3:1-5. Sørensen, J. T. 1990. Validation of livestock herd simulation models: a review. Livest. Prod.Sci. 26(2):79-90. Sørensen, J. T. and S. Østergaard. 2003. Economic consequences of postponed first insemination of cows in a dairy cattle herd. Livest. Prod. Sci. 79(2-3):145-153.
31
Economic consequences of reproductive performance Steeneveld, W., J. Swinkels, and H. Hogeveen. 2007. Stochastic modelling to assess economic effects of treatment of chronic subclinical mastitis caused by Streptococcus uberis. J. Dairy Res. 74(04):459-467. Stott, A. W., R. F. Veerkamp, and T. R. Wassell. 1999. The economics of fertility in the dairy herd. Anim. Sci. 68(1):49-57. Tenhagen, B. A., C. Vogel, M. Drillich, G. Thiele, and W. Heuwieser. 2003. Influence of stage of lactation and milk production on conception rates after timed artificial insemination following Ovsynch. Theriogenology 60(8):1527-1537. van Arendonk, J. A. M. and A. A. Dijkhuizen. 1985. Studies on the replacement policies in dairy cattle. III. Influence of variation in reproduction and production. Livest. Prod. Sci. 13(4):333-349. Veerkamp, R. F., P. Dillon, E. Kelly, A. R. Cromie, and A. F. Groen. 2002. Dairy cattle breeding objectives combining yield, survival and calving interval for pasture-based systems in Ireland under different milk quota scenarios. Livest. Prod. Sci. 76(12):137-151. Villa-Godoy, A., T. L. Hughes, R. S. Emery, E. P. Stanisiewski, and R. L. Fogwell. 1990. Influence of energy balance and body condition on estrus and estrous cycles in Holstein heifers. J. Dairy Sci. 73(10):2759-2765. Windig, J. J., M. P. Calus, and R. F. Veerkamp. 2005. Influence of herd environment on health and fertility and their relationship with milk production. J. Dairy Sci. 88(1):335-347. Windig, J. J., M. P. L. Calus, B. Beerda, and R. F. Veerkamp. 2006. Genetic correlations between milk production and health and fertility depending on herd environment. J. Dairy Sci. 89(5):1765-1775. Wood, P. D. P. 1967. Algebraic model of the lactation curve in cattle. Nature 216:164-165.
32
CHAPTER 3 Effect of milk yield characteristics, breed and parity on success of the first insemination in Dutch dairy cows C. Inchaisri1,2, H. Hogeveen2,3P.L.A.M. Vos2, G.C. van der Weijden2, and R. Jorritsma2
1
Department of Veterinary Medicine, Faculty of Veterinary Science, Chulalongkorn University, Bangkok, 10330, Thailand 2 Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, P.O. Box 80152, 3508 TD, The Netherlands 3 Business Economics, Wageningen University, P.O. Box 8130, 6706 KN Wageningen, The Netherlands
Published in Journal of Dairy Science (2010) 93: 5179-5187
Chapter 3 ABSTRACT The objective of this study was to determine the contribution of cow factors to the probability of a successful first insemination (SFI). The investigation was performed with 51,791 lactations from 1,396 herds obtained from the Dutch dairy cow database of the Cattle Improvement Co-operative (CRV). Cows that had the first insemination (AI) between 40 and 150 d postpartum were selected. The first AI was classified as successful when cows were not re-inseminated and either calved between 267 and 295 d later or were culled within 135 to 295 d after first AI. The lactation curve characteristics of individual lactations were estimated by Wilmink’s curve using the test-day milk records from CRV. The lactation curve characteristics (peak milk yield, milk yield at the first-AI date, time of peak yield (PT), and milk persistency) were calculated. Breed, parity, interval from calving to first AI (CFI), lactation curve characteristics, milk production traits, moment of AI related to PT (before or after PT), calf status, month of AI, and month of calving were selected as independent variables for a model with SFI as a dependent variable. A multivariable logistic regression model was used with farm as a random effect. Overall SFI was 44%. The effect of parity on SFI depended on CFI. The first-parity cows had the greatest SFI (0.43) compared with other parities (0.32–0.39) at the same period of CFI before 60 d in milk (DIM), and cows in parity ≥5 had the least SFI (0.38–0.40) when AI was after 60 DIM. After 60 DIM, extending CFI did not improve SFI in the first-parity cows, but SFI was improved in multiparous cows. Holstein-Friesian cows had lesser SFI (0.37) compared with cross-breed cows (0.39–0.46). Twin and stillbirth calving reduced SFI (0.39) compared with a single female calf (0.45) or a male calf (0.43) calving. The SFI in different months of AI varied and depended on CFI. Cows that received AI before 60 DIM had a lesser SFI, especially in March, June, and July (0.18, 0.35, and 0.34, respectively). Artificial insemination before PT reduced SFI (0.39) in comparison with AI after PT (0.44). The effect of milk yield at the first-AI date on SFI varied depending on CFI. After 60 DIM at the same period of CFI, a high level of milk yield at the first-AI date reduced SFI. In conclusion, knowledge of the contribution of cow factors on SFI can be applied to support decision making on the moment of insemination of an individual cow in estrus. (Key words: milk production, first insemination, lactation curve, reproductive)
35
Success of the first insemination in Dutch dairy INTRODUCTION A high reproductive efficiency is an important prerequisite to guarantee a profitable dairy production. Several factors, including cow factors as well as management and environmental factors are generally known to affect the economical return of a dairy farm. In general, factors like milk production, BCS, postpartum disorders, breed, parity, DIM, feeding regimes, and seasonality have substantial effects on the reproductive efficiency in dairy cows and were applicable in the Netherlands (Berry et al., 2003, Pryce et al., 1997, Tsuruta et al., 2009). Moreover, where milk production has been increased, reproductive performance of dairy cattle has decreased (Huang et al., 2009, Norman et al., 2009). A successful first AI (SFI) is a key in reproductive efficiency, but the probability of conception after the first AI differed from cow to cow (Windig et al., 2005). At least part of the differences between cows is explained by the 305-d milk yield (MY305) or the cumulative milk yield in the early stage of lactation (Kadarmideen et al., 2000, Pryce et al., 1997, Veerkamp et al., 2001). This is observation has been confirmed by other studies, in which characteristics of the lactation curve of an individual cow, such as milk persistency (Muir et al., 2004) and peak milk yield (Kinsel and Etherington, 1998, Lean et al., 1989) affect the probability of SFI. In general, the high-producing cows within herds have a low conception rate at the first AI (Windig et al., 2005). Moreover, conception rate improved with an increase of DIM (Loeffler et al., 1999, Tenhagen et al., 2003), but decreased when the milk yield produced on the test-day closest to the AI date on conception is higher (Loeffler et al., 1999). Therefore, one may predict that the milk yield around the time of AI is an important factor determining the success of this AI. Due to the effect of gestation, cumulative milk yield, MY305 or milk persistency is partly dependent on the reproductive performance (Olori et al., 1997). Thus, milk yield at the day of first AI (MYFI) and the interval from calving to first AI (CFI) are more suitable as predictor for SFI. Moreover, from a practical perspective knowledge about the influence of MYFI and CFI on SFI may assist farmers with their decisions whether to start or to delay the first AI. Hence, it is important to know which factors contribute to SFI and to which extent SFI is influenced by these factors. Although many studies reported on the possible relationship between cow factors and conception rates, no studies showed a modeled SFI using cow factors only that are measurable before first AI. Therefore, the objective was to determine the effects of cow factors, especially those before the first AI postpartum on the probability of SFI in Dutch dairy cows.
36
Chapter 3 MATERIALS AND METHODS Data The original dataset, consisting of AI and production records of 104,969 lactations following calving in 3,674 herds in 2006, was selected randomly from the database of the Cattle Improvement Co-operative. In general, Dutch farmers use a dry period length of about 2 mo and house these cows separately. Also, their feeding is usually adjusted to meet the requirements in the dry period. Many farmers have a close look at individual parturient cows, e.g., by using a separate calving pen. Herds with fewer than 25 lactations in the dataset were removed (n = 7,942 lactations). To standardize AI techniques, herds in which not all cows were AI by a technician from the Cattle Improvement Co-operative were removed (n = 9,438 lactations). Lactations with incomplete data with respect to cow identification, birth date, calving date, AI date, or parity were removed (n = 18,290 lactations). Herds that had at least 1 cow that was naturally serviced or inseminated for embryo transfer, were deleted from the dataset as insemination records were supposedly less reliable (n = 1,093 lactations). Lactations from cows that were AI before 40 or after 150 DIM were assumed not representative for normal healthy cows and removed (n = 9,028 lactations). When the interval between 2 AI was less than 10 d, the first AI of that pair was considered as missing. Lactations with only 1 AI without information on culling, death or a next calving date were removed (n = 135 lactations). Lactations with an aberrant fitted curve (see Equation 1; n = 489 lactations) were removed, as described in the next paragraph. Also, lactations with less than 5 test-day milk records (n = 815 lactations), lactations with a test-day milk record interval < 14 d or > 70 d, or a first-test day milk records after 70 DIM (n = 2120 lactations) were removed because this is rarely seen in normal healthy cows. Moreover, lactations with a high fluctuation in test-day milk records between test-day interval were removed (n = 826 cows). Lactations of culled cows were removed when culling took place within < 50 d after the first AI (n = 537). The obtained cow fertility records were matched with milk production data and unmatched data were removed (n = 110 lactations). Other lactations were removed after the estimation of lactation curve when it gave unsatisfied characteristics, e.g., too low milk production and too early of peak time (n = 2,599 lactations). Finally, a dataset of 51,791 lactations from 1,396 herds with 59.2 ± 0.7 (mean ± se) lactations per herd was used. Of all lactations, 32% was parity 1, 25% parity 2, 17% parity 3, 12% parity 4, and 15% parity 5 or higher. The breeds of the cows were 100% Holstein Friesian (HF; 63% of lactations), 50 - < 100% HF (32% of lactations), 51% - 100% Dutch Red-and-White (MRY; 4% of lactations) and other breeds (1% of lactations).
37
Success of the first insemination in Dutch dairy Lactation Curve A lactation curve for each lactation was modeled using the test-day milk recording data of the cow. A combined exponential and linear model was chosen to depict the shape of the lactation curve (Wilmink, 1987):
MYti = a i + bi × t i + ci × exp(-0.05 t i ) ,
[1]
Where MYti is milk (kg) on day t postpartum of cow i, a is the scaling factor estimating production at time zero, b is the rate of descent after the peak, and c is the rate of ascent to the peak. The initial values in different parities of a, b, and c given the age at calving (months) were obtained from an earlier study (Wilmink, 1987). Parameter estimates including a, b and c for individual lactation were obtained using the Gauss-Newton method in the nonlinear procedure (PROC NLIN) of SAS (version 9.2, SAS Inst. Inc., Cary, NC). A lactation curve was classified as non-aberrant and remained in the dataset when the estimated values for b and c were negative, resulting in the well-known convex parabola milk production curve. Also, lactation curves with a decreasing production curve due to the lack of a somewhat lower test-day milk record early in lactation were classified as non-aberrant as long as the first AI was after the first test-day milk record. These curves had a negative value for b and a positive for c in the modeled lactation curve. Based on the estimated lactation curve parameters, the 4 lactation curve characteristics: peak time (PTi), peak yield (PYi), MYFIi and persistency (MPERi) were calculated for individual lactations of cow i as follows:
PTi = -20log 20bi / ci
[2]
,
PYi = a i + bi × PTi + ci × exp(-0.05 PTi ) ,
[3]
MYFIi = a i + bi × CFIi + ci × exp(-0.05 × CFIi ) ,
[4]
MPER i = MY280i - MY90i 190 ,
[5]
Where PTi is DIM at PYi and CFIi is interval between calving and first AI date of cow i. MY280i and MY90i are the milk yield of cow i at 280 and 90 d postpartum, respectively. For cows with a decreasing lactation curve, the first test-day milk record was used as PTi, and PYi.
38
Chapter 3 Definition of a Successful First AI Cows that delivered a calf with a gestation length between 267 and 295 d after a single first AI were classified as having a SFI. Cows that were culled during lactation had a successful first AI when they were not re-inseminated after the first AI and when they were culled between at least 135 d (95th percentile of interval from the first to the second AI calculated from the dataset) and 295 d after first AI (the maximum length of gestation). All other cows with at least 1 AI were classified as having a non-successful first AI. Statistical Analyses The following measures were calculated to obtain descriptive statistics and Pearson’s correlations coefficients for continuous variables or Spearman rank correlations coefficients for categorical variables (r): parity, breed, milk production measures (305-d milk yield (MY305), 305-d protein yield (PROT305), and 305-d fat yield (FAT305)), lactation curve characteristics (PY, PT, MPER, and MYFI), and reproductive performances (CFI, calving date, the first AI date, time of AI related to PT, and calf status). Statistical analyses were carried out to determine the association between SFI and independent variables, using SAS (PROC GLIMMIX) with binary distribution and logit link (version 9.2, SAS Institute Inc., Cary, NC). Farm was added in the model as a random effect: [6]
,
,
[7]
Where pij is the comparison between the probability of the success and non-success of the first AI of cow i on farm j; β0 is the estimated intercept, and the regression coefficients of log odds ratio (β1 - βk) were estimated for each independent factor (X1i - Xki). The Xki are the predictor values of an independent variable k for cow i. λj is the random effect of farm j, assumed λj ~ N(0, σ2), and the relationship between the probability pij and the binary outcome is success of the first AI (p(SFIij = 1) = pij. SFIij is success of the first AI of cow i on farm j. Combining equations 6 and 7, the estimation of SFIi adding farm as a random effect, and its link function (E[SFIijk| λj]) were analyzed as follows:
, 39
[8]
Success of the first insemination in Dutch dairy All independent variables were modeled, including breed, parity, calf status after birth, CFI, lactation curve characteristics (PY, PT, MPER, and MYFI), milk production measures (MY305, PROT305, and FAT305), month of calving and month of first AI (Jan to Dec), and time of AI related to PT. Due to non-normal distribution of CFI, it was categorized by 3 wk as estrous interval by group of ≤ 60, 61 to 81, 82 to 102, 103 to 123, and ≥ 124 DIM for the subsequence analysis. Cow breeds were categorized as 100% of HF, 50 - < 100% HF, 51 - 100% MRY, and others. Parities were categorized as 1, 2, 3, 4, and ≥ 5. Registered calves from included cows were categorized as female, male, twin, and stillbirth. The AI times related to PT were categorized for AI before or after PT. For categorized variables, all dummy variables were entered. The factors that affected on SFI in the univariable analysis (P < 0.15) and not highly correlated (r > 0.5) with other predictors were included in multivariable analysis by stepwise forward regression. Yet, MYFI and AI date were included in the multivariable analysis, as they were preferred above other measures for milk production and timing. Other independent variables were entered freely or were removed, using twice the negative of the residual log likelihood (model deviance), type 3 sums of squares, and the coefficients and error terms of the main effects. Confounding was assessed if the exclusion variables changed the remaining parameter estimates > 10%. Interactions between the remaining parameters (MYFI and CFI, parity and MYFI, parity and CFI, breed and MYFI, breed and CFI, calf status and MYFI, calf status and CFI, month of AI and MYFI, month of AI and CFI, time of AI related to PT and MYFI) were analyzed. A variable remained in the model when its significance level was lower than 0.1 and the linearity assumption between log-odds and variables were met. The final model was acceptable when the ratio of the generalized chi-square was close to 1 (Brown and Prescott, 1999). The model diagnostics were performed by plotting the residuals against log-odds and the fitted values judged for peculiarities for the final multivariable model. When categorized data were statistically significant, the estimates from the final model were expressed as means of the probabilities of SFI and pairwise means comparison between categorized data were carried out with Turkey adjustment at P < 0.05. RESULTS Data Description The overall SFI was 43.7%. Descriptive data and an average of the variables studied are in Table 1. Culling percentage of all cows was 14.3% and SFI of culled cows was 18.1%. Of all lactations, 0.04% had a shorter CI than 267 d and 9.5% had a second AI more than 3 reproductive cycles (63 d) after the first AI, whereas 7.8% of calvings were stillbirths, and 2.3% were twins. In the univariable analysis, the associations between SFI with PY, MPER, MYFI, MY305, PROT305, FAT305, CFI, breed, parity, calf status, AI time related to 40
Chapter 3 PT, month of calving, and month of first AI were statistically significant for each individual variable (Table 2). Multivariable Logistic Regression Model Results of the final multivariable logistic regression model are given in Table 3 and the associations between SFI and independent categorized factors and their interactions are plotted in Figures 1, 2, 3, and 4. The random effect of farms was small, given λ being 0.06 (SE = 0.007). No confounding was found and the inspection of residual plots showed no irregularities. The MYFI had an effect on SFI depending on CFI (Table 3). When AI before 60 DIM, the increase of MYFI had less of an effect on SFI, but a high level of MYFI reduced SFI for the AI after 60 DIM. Moreover, cows that were AI before PT had a statistically significant (OR = 0.8; P = < 0.01) lower SFI (0.39) compared with AI after PT (SFI = 0.44). After adjusting for other factors, SFI was higher for 50 - < 100% HF cows, 51 - 100% MRY, and other breeds compared with 100% HF cows (Table 3 and Figure 1). Calf status had an effect on SFI (Table 3 and Figure 2). With interaction of CFI, extending CFI did not improve SFI in the first parity cows, but SFI was improved significantly in other parities when the first AI was after 60 DIM (Figure 3). Comparison with other parities at the same period of CFI after 60 DIM, cows in parity ≥ 5 had the lowest SFI and the first parity cows had the highest SFI compared with other parities when AI before 60 DIM. Table 1. Descriptive statistics (mean, 5th and 95th percentile) of the studied variables. Variable1 n Mean 5th - 95th percentiles CFI (d) 51,791 87 (51 - 136) CI (d) 44,269 409 (336 - 554) PY (kg) 51,791 35 (24 - 48) PT (d) 51,791 46 (17 - 73) MPER (kg/d) 51,791 -0.07 (-0.13 - (-0.02)) MYFI (kg) 51,791 33 (22 - 45) MY305 (kg) 51,791 8,542 (5,929 - 11,345) PROT305 (kg) 51,791 296 (209 - 385) FAT305 (kg) 51,791 370 (261 - 486) 1 SFI = success of first AI; CFI = interval from calving to first AI; CI = calving interval; PY = peak milk yield; PT = time of peak yield; MPER = milk persistency; MYFI = milk yield at the first-AI date; MY305 = 305-d milk yield; PROT305 = 305-d protein yield; FAT305 = 305-d fat yield
41
Success of the first insemination in Dutch dairy Table 2. Results of the analysis of risk factors for successful first AI from the univariable logistic regression analysis using farm as a random effect in the model1 Variable2
β
SE
P-value < 0.01
Breed HF 100%
Ref.
HF 50 - < 100%
3
-
-
0.10
0.02
< 0.01
MRY 51 - 100%
0.30
0.05
< 0.01
Others
0.42
0.08
< 0.01
Parity
< 0.01
1
0.21
0.03
< 0.01
2
0.14
0.03
< 0.01
3
0.15
0.03
< 0.01
4
0.08
0.03
0.01
.
.
≥5
Ref.
3
< 0.01
Calf status Female
Ref.3
.
.
Male
-0.06
0.02
< 0.01
Twin
-0.23
0.06
< 0.01
Still birth
-0.25
0.03
< 0.01
Month of calving4
-
-
< 0.01
Month of AI4
-
-
< 0.01
CFI (days postpartum)
< 0.01
≤ 60
-0.22
0.03
< 0.01
61 to 81
-0.05
0.03
NS5
82 to 102
-0.007
0.03
NS5
103 to 123
-0.006
0.03
NS5
≥ 124
Ref.3
.
.
42
Chapter 3 Table 2 continued Variable2
β
SE
P-value
PY (kg/d)
-0.01
0.001
< 0.01
PT (d)
0.0002
0.0005
NS5
MPER (kg/d)
-5.13
0.26
< 0.01
MYFI (kg/d)
-0.013
0.001
< 0.01
MY305 (kg)
-0.0001
0.000005
< 0.01
FAT305 (kg)
-0.003
0.0001
< 0.01
PROT305 (kg)
-0.003
0.0002
< 0.01 < 0.01
Time of AI After PT
Ref.3
-
-
Before PT
-0.26
0.03
< 0.01
1
Estimated coefficient (β), SE for the coefficient, and significance level are given for each cow-specific risk factor 2 HF = Holstein Friesian; MRY = Dutch Red-and-White; CFI = interval from calving to first AI; PY = peak milk yield; PT = time of peak yield; MPER = milk persistency; MYFI = milk yield at the first-AI date; MY305 = 305-d milk yield; PROT305 = 305-d protein yield; FAT305 = 305-d fat yield 3 Ref. = reference category. 4 Categorized data not shown 5 NS = P > 0.1
The interaction between months of first AI with CFI had an effect on the SFI after correction for other factors. The association between SFI and the interaction between months of first AI with CFI is shown in Figure 4. In all months, the SFI was lower when cows were AI before 60 DIM compared with after 60 DIM. This was particular for AI in March before 60 DIM, when SFI was very low and SFI was improved after 81 DIM. In every period of CFI, the AI in summer especially in July had the lowest SFI compared with other seasons.
43
Success of the first insemination in Dutch dairy Table 3. Results of the analysis of risk factors for successful first AI from the multivariable logistic regression analysis using farm as a random effect in the model (λ = 0.06, SE = 0.007)1 Variable2
β
Intercept
0.23
OR3
95% CI for OR P-value < 0.01
Breed 4
HF 100%
Ref.
1
.
HF 50 - < 100% MRY 51 - 100% Others
0.09 0.27 0.38
1.10 1.31 1.46
Female
Ref.4
1
Male Twin Still birth
-0.06 -0.24 -0.27
0.94 0.79 0.76
0.90 - 0.97 0.70 - 0.90 0.71 - 0.81
Time of AI Before PT
-0.23
0.80
0.74 - 0.86
After PT
Ref.4
1
.
MYFI (kg/d)
-0.02
MYFI*CFI MYFI*CFI ≤ 60 d MYFI*CFI 61 to 81d
0.02 0.01
1.02 1.01
1.00 - 1.05 1.00 - 1.03
< 0.01 0.05 0.02 0.1
MYFI*CFI 82 to 102 d
0.003
1.00
0.99 - 1.01
NS6
MYFI*CFI 103 to 123 d
0.002
1.00
0.99 - 1.01
NS6
MYFI*CFI ≥ 124 d
Ref.4
1
1.05 - 1.14 1.17 - 1.47 1.24 - 1.73
Calf status
CFI*Parity
5
< 0.01 < 0.01 < 0.01 < 0.01 . < 0.01 < 0.01 < 0.01 < 0.01 < 0.01
. 0.02
5
< 0.01 CFI*month of AI The ratio of the generalized chi-square statistic and its degrees of freedom equals 0.99 and estimated coefficients (β), odds ratio (OR), 95% confidence interval (CI) for OR, and significance level are given for each cow-specific risk factor. 2 HF = Holstein Friesian; MRY = Dutch Red-and-White; CFI = interval from calving to first AI; MYFI = milk yield at the first-AI date; PT = DIM at peak milk yield 3 Odds ratio for successful first AI vs. failure of the first AI. 4 Ref. = reference category. 5 Categorized data not shown. 6 NS = P > 0.1 1
44
Chapter 3
Figure 1. The estimated the probabilities of successful first AI in different breeds correcting for other factors in the final multivariable logistic regression model. a, b, and c indicate differences (P < 0.05) of estimated probability of successful first AI between breeds (HF = Holstein Friesian; MRY = Dutch Red-and-White).
Figure 2. The estimated probabilities of successful first AI in different calf status correcting for other factors in the final multivariable logistic regression model. a, b, and c indicate differences (P < 0.05) of estimated probability of successful first AI between calf status.
45
Success of the first insemination in Dutch dairy
Estimated probability of sucessful first AI
Figure 3. The effect of interaction of parity and days in milk at first AI on successful first AI in the final multivariable model. The differences (P < 0.05) of estimated probability of successful first AI were shown with a and b indicating the differences between DIM at first AI (≤ 60 d (white bars), 61 to 81 d (white gray bars), 82 to 102 d (gray bars), 103 to 123 d (dark gray bars), and ≥ 124 d (black bars)) within the same parity and 1, 2 and 3 indicate the differences between parities within the same category of DIM. 0.55 0.50
bc
0.45
bc
b ab
b
b
a
0.40
a
a
c
b b ab ab
ab ab
b ab ab ab
a a b
ab ab ab
b ab ab a
a
a
a
0.35 0.30 0.25 a
0.20 0.15 Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Months in different days in milk at first AI
Figure 4. The effect of interaction of months of AI and DIM at first AI on successful first AI in the final multivariable model. The differences (P < 0.05) of estimated probability of successful first AI were shown with a, b, and c indicating the differences between DIM at first AI (≤ 60 d (white bars), 61 to 81 d (white gray bars), 82 to 102 d (gray bars), 103 to 123 d (dark gray bars), and ≥ 124 d (black bars)) within the same month of AI.
46
Chapter 3 DISCUSSION The percentage of pregnancy after first AI in Dutch dairy cows varied between 40 and 45% and was dependent on fat and protein-corrected milk yield, parity, postpartum disorders, and BCS (Loeffler et al., 1999). Here, SFI varied depending on breed, parity, calf status, and time to AI. Before 60 DIM, the SFI in multiparous cows was lower in all months and the SFI increased when CFI was > 60 DIM (Figure 3 and 4). The higher SFI at larger CFI may be caused by an improved energy balance (Butler and Smith, 1989, Loeffler et al., 1999). Although other studies did not look at the interaction between MYFI and CFI, other studies suggest amelioration of SFI when the first AI is postponed (Loeffler et al., 1999, Tenhagen et al., 2003). Given the interaction between MYFI and CFI, the results of our study suggest that the negative effect of CFI on SFI is greater for higher producing cows compared with lower producing cows. Insemination after PT resulted in a higher SFI than AI before PT. Cows having a lower MYFI in combination with a longer CFI at the same period of CFI showed a higher SFI than cows with high MYFI whenever the first AI was performed after 60 DIM and especially when these cows were AI after PT. The average SFI in first parity cows was 41 to 44%, supporting the success of first AI that was 46% in the first parity and 44% in higher parity cows (Windig et al., 2005). Supporting other studies, older cows AI before 60 DIM were less likely to become pregnant after first AI than younger cows (Darwash et al., 1997, Eicker et al., 1996, Tenhagen et al., 2003). In contrast to these results, other studies have shown lower conception rates at the first AI in first parity cows (Lean et al., 1989, Melendez and Pinedo, 2007, Rocha et al., 2001) and explained this by a more severe negative energy balance due to higher energy requirement for growth (Lucy, 2001). Yet, when the effects of environment and feeding were controlled, the first parity cows had a lower negative energy balance; hence, an improved fertility outcome than multiparous cows (Friggens et al., 2007). Results of this study showed that the SFI of the first parity cows when AI during early lactation was not different from AI performed in later stages of lactation (Figure 3). In general, differences between studies are difficult to explain, but may be due to different social and nutritional management strategies of the included cows. Season is an important factor affecting the fertility, both in the season of calving (Eicker et al., 1996, Grohn and Rajala-Schultz, 2000, Melendez and Pinedo, 2007) and the season of AI (Benzaquen et al., 2007, Huang et al., 2009, Melendez et al., 2008). The effect of DIM on conception rate varied between seasons in our study as well as in others (Eicker et al., 1996, Huang et al., 2008). The reason for a somewhat lower conception rate in summer was explained by the effects of high temperature. Moreover, a lower SFI at the start and finish of the grazing season may be due to difficulties with regard to the switch between feeding and housing systems. 47
Success of the first insemination in Dutch dairy In winter, dairy cows in the Netherlands are housed inside the barn and, in most herds, cows are pastured from April to October. Our results showed that 100% HF cows had lower SFI compared to cross-bred cows and cows of other breeds. Previous studies have found that a high milk yield cow has a greater BCS change than a low milk yield cow (Berry et al., 2003, Buckley et al., 2003, Loeffler et al., 1999) as a result of the mobilization of more body energy in early lactation of the Holstein compared with the Danish red and Jersey breed (Friggens et al., 2007). Moreover, a negative correlation between BCS loss and reproductive performance in different breeds is well established (Pryce et al., 2001). It has been suggested that differences between breeds in their ability to partition energy between milk production and body reserves could be responsible for the differences in fertility (Friggens et al., 2007, Walsh et al., 2008, Yan et al., 2006). Stillbirth and twin calving were associated with the reduction of SFI supporting other previous studies (Berry et al., 2007, Nielen et al., 1989). Stillbirth may be caused by dystocia (Chassagne et al., 1999) and dystocia itself may be caused by twin calving or larger sized calves, hence a heavier weight. Cows with dystocia negatively affect postpartum reproductive performance (Chassagne et al., 1999). Moreover, cows which experienced stillbirth showed a slightly more severe negative energy balance (Berry et al., 2007) that may contribute to a lower fertility. The analysis was carried out on an individual cow level, using farm as a random effect, like a recent study (Friggens and Labouriau, 2010) and included culled cows in the statistical calculations. Therefore, it is not likely that SFI was overestimated. A potential lower SFI for cows having postpartum disorders was not studied because data on postpartum disorders was not available. Using the existing databases, the variation between farms might have biased the results. Farms were included as random effects in our model, which has the advantage of correcting for variation of herd management such as differences in nutritional management and environment. The results showed that the variation in SFI between farms was relatively small. Only one lactation per cow was included in the analysis to prevent a bias from repeated individual cows. To study accurately the relationship between milk yield and fertility, sick and aborted cows should be deleted from the analysis. Because the dataset did not contain information about illness and abortions, the situation allowed selection of only cows that had a non-aberrant fitted milk production curve. The modeled cow factors were all factors that are known at the moment that cows are eligible for AI. At that time, the decision on whether a cow will be inseminated or not and the probability of a successful first AI are important variables. Therefore, these results can be applied in models that support AI decisions. Instead of using a generic voluntary waiting period for all cows in a herd, the voluntary waiting period may be adjusted for individual cows, taking into account cow factors such as the production level, persistency of milk production and parity. 48
Chapter 3 CONCLUSIONS The results show that SFI was influenced by MYFI, CFI, time of AI related to PT, parity, breed, calf status, and season. For multiparous cows, AI in an early lactation stage (before 60 DIM) reduced SFI moderately, but AI in a later stage of lactation and a low MYFI improved SFI. Insemination before peak milk yield reduced SFI. Knowledge of the influence of cow factors that contribute to SFI can be applied to assist in deciding on the moment of AI of individual cows that are in estrus. ACKNOWLEDGMENTS We thank Mathijs van Pelt for his support in preparation of the dataset and the Cattle Improvement Co-operative (Arnhem, the Netherlands) for providing the data used in the present study. We thank Jan van Den Broek (Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, the Netherlands) for his suggestions on the statistics. REFERENCES Benzaquen, M. E., C. A. Risco, L. F. Archbald, P. Melendez, M. J. Thatcher, and W. W. Thatcher. 2007. Rectal temperature, calving-related factors, and the incidence of puerperal metritis in postpartum dairy cows. J. Dairy Sci. 90(6):2804-2814. Berry, D. P., F. Buckley, P. Dillon, R. D. Evans, M. Rath, and R. F. Veerkamp. 2003. Genetic relationships among body condition score, body weight, milk yield, and fertility in dairy cows. J. Dairy Sci. 86(6):2193-2204. Berry, D. P., J. M. Lee, K. A. Macdonald, and J. R. Roche. 2007. Body condition score and body weight effects on dystocia and stillbirths and consequent effects on postcalving performance. J. Dairy Sci. 90(9):4201-4211. Brown, H. and R. Prescott. 1999. Applied mixed models in medicine. John Wiley and Sons, Ltd., Chichester, UK. Buckley, F., K. O'Sullivan, J. F. Mee, R. D. Evans, and P. Dillon. 2003. Relationships among milk yield, body condition, cow weight, and reproduction in spring-calved HolsteinFriesians. J. Dairy Sci. 86(7):2308-2319. Butler, W. R. and R. D. Smith. 1989. Interrelationships between energy balance and postpartum reproductive function in dairy cattle. J. Dairy Sci. 72(3):767-783. Chassagne, M., J. Barnouin, and J. P. Chacornac. 1999. Risk factors for stillbirth in holstein heifers under field conditions in France: A prospective survey. Theriogenology 51(8):1477-1488. Darwash, A. O., G. E. Lamming, and J. A. Woolliams. 1997. The phenotypic association between the interval to post-partum ovulation and traditional measures of fertility in dairy cattle. Anim. Sci. 65(1):9-16. Eicker, S. W., Y. T. Grohn, and J. A. Hertl. 1996. The association between cumulative milk yield, days open, and days to first breeding in New York Holstein cows. J. Dairy Sci. 79(2):235-241.
49
Success of the first insemination in Dutch dairy Friggens, N. C., P. Berg, P. Theilgaard, I. R. Korsgaard, K. L. Ingvartsen, P. LØvendahl, and J. Jensen. 2007. Breed and parity effects on energy balance profiles through lactation: Evidence of genetically driven body energy change. J. Dairy Sci. 90(11):5291-5305. Friggens, N. C. and R. Labouriau. 2010. Probability of pregnancy as affected by oestrus number and days to first oestrus in dairy cows of three breeds and parities. Anim. Reprod. Sci. 118(2-4):155-162. Grohn, Y. T. and P. J. Rajala-Schultz. 2000. Epidemiology of reproductive performance in dairy cows. Anim. Reprod. Sci. 60-61:605-614. Huang, C., S. Tsuruta, J. K. Bertrand, I. Misztal, T. J. Lawlor, and J. S. Clay. 2008. Environmental effects on conception rates of Holsteins in New York and Georgia. J. Dairy Sci. 91(2):818-825. Huang, C., S. Tsuruta, J. K. Bertrand, I. Misztal, T. J. Lawlor, and J. S. Clay. 2009. Trends for conception rate of Holsteins over time in the southeastern United States. J. Dairy Sci. 92(9):4641-4647. Kadarmideen, H. N., R. Thompson, and G. Simm. 2000. Linear and threshold model genetic parameters for disease, fertility and milk production in dairy cattle. Anim. Sci. 71(3):411-419. Kinsel, M. L. and W. G. Etherington. 1998. Factors affecting reproductive performance in Ontario dairy herds. Theriogenology 50(8):1221-1238. Lean, I. J., J. C. Galland, and J. L. Scott. 1989. Relationships between fertility, peak milk yields and lactational persistency in dairy cows. Theriogenology 31(5):1093-1103. Loeffler, S. H., M. J. de Vries, Y. H. Schukken, A. C. de Zeeuw, A. A. Dijkhuizen, F. M. de Graaf, and A. Brand. 1999. Use of AI technician scores for body condition, uterine tone and uterine discharge in a model with disease and milk production parameters to predict pregnancy risk at first AI in Holstein dairy cows. Theriogenology 51(7):12671284. Lucy, M. C. 2001. Reproductive loss in high-producing dairy cattle: Where will it end? J. Dairy Sci. 84(6):1277-1293. Melendez, P., M. Duchens, A. Perez, L. Moraga, and L. Archbald. 2008. Characterization of estrus detection, conception and pregnancy risk of Holstein cattle from the central area of Chile. Theriogenology 70(4):631-637. Melendez, P. and P. Pinedo. 2007. The association between reproductive performance and milk yield in Chilean Holstein cattle. J. Dairy Sci. 90(1):184-192. Muir, B. L., J. Fatehi, and L. R. Schaeffer. 2004. Genetic relationships between persistency and reproductive performance in first-lactation Canadian Holsteins. J. Dairy Sci. 87(9):3029-3037. Nielen, M., Y. H. Schukken, D. T. Scholl, H. J. Wilbrink, and A. Brand. 1989. Twinning in dairy cattle: A study of risk factors and effects. Theriogenology 32(5):845-862. Norman, H. D., J. R. Wright, S. M. Hubbard, R. H. Miller, and J. L. Hutchison. 2009. Reproductive status of Holstein and Jersey cows in the United States. J. Dairy Sci. 92(7):3517-3528. Olori, V. E., S. Brotherstone, W. G. Hill, and B. J. McGuirk. 1997. Effect of gestation stage on milk yield and composition in Holstein Friesian dairy cattle. Livest. Prod. Sci. 52(2):167-176. Pryce, J. E., M. P. Coffey, and G. Simm. 2001. The relationship between body condition score and reproductive performance. J. Dairy Sci. 84(6):1508-1515.
50
Chapter 3 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(3):353-360. Rocha, A., S. Rocha, and J. Carvalheira. 2001. Reproductive parameters and efficiency of inseminators in dairy farms in Portugal. Reprod. Domest. Anim. 36(6):319-324. Tenhagen, B. A., C. Vogel, M. Drillich, G. Thiele, and W. Heuwieser. 2003. Influence of stage of lactation and milk production on conception rates after timed artificial insemination following Ovsynch. Theriogenology 60(8):1527-1537. Tsuruta, S., I. Misztal, C. Huang, and T. J. Lawlor. 2009. Bivariate analysis of conception rates and test-day milk yields in Holsteins using a threshold-linear model with random regressions. J. Dairy Sci. 92(6):2922-2930. Veerkamp, R. F., E. P. Koenen, and G. De Jong. 2001. Genetic correlations among body condition score, yield, and fertility in first-parity cows estimated by random regression models. J. Dairy Sci. 84(10):2327-2335. Walsh, S., F. Buckley, K. Pierce, N. Byrne, J. Patton, and P. Dillon. 2008. Effects of breed and feeding system on milk production, body weight, body condition score, reproductive performance, and postpartum ovarian function. J. Dairy Sci. 91(11):4401-4413. Wilmink, J. B. M. 1987. Adjustment of test-day milk, fat and protein yield for age, season and stage of lactation. Livest. Prod. Sci. 16(4):335-348. Windig, J. J., M. P. Calus, and R. F. Veerkamp. 2005. Influence of herd environment on health and fertility and their relationship with milk production. J. Dairy Sci. 88(1):335-347. Yan, T., C. S. Mayne, T. W. J. Keady, and R. E. Agnew. 2006. Effects of dairy cow genotype with two planes of nutrition on energy partitioning between milk and body tissue. J. Dairy Sci. 89(3):1031-1042.
51
CHAPTER 4 Analysis of the economically optimal voluntary waiting for first insemination C. Inchaisri1,2, R. Jorritsma2, P.L.A.M. Vos2, G.C. van der Weijden2, and H. Hogeveen2,3
1
Department of Veterinary Medicine, Faculty of Veterinary Science, Chulalongkorn University, Bangkok, 10330, Thailand 2 Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, P.O. Box 80152, 3508 TD, The Netherlands 3 Business Economics, Wageningen University, P.O. Box 8130, 6706 KN Wageningen, The Netherlands
Published in Journal of Dairy Science (2011) 94: 3811-3823
Chapter 4 ABSTRACT The voluntary waiting period (VWP) is defined as the time between parturition and the time at which the cow is first eligible for insemination. Determining the optimal VWP from field data is difficult and unlikely to happen. Therefore, a Monte-Carlo dynamic-stochastic simulation model was created to calculate the economic effects of different VWPs. The model is dynamic and uses time steps of one week to simulate the reproductive cycle (ovulation, estrous detection, and conception), the occurrence of postpartum disorders, and the lactation curve. Inputs of the model were chosen to reflect the situation of Dutch dairy cows. In the model, we initially created a cow of a randomly-selected breed, parity, month of calving, calf status of last calving, and expected 305-d milk yield. The randomly varied variables were based upon relevant distributions and adjusted for cow statuses. The lactation curve was modeled by Wood’s function. The economic input values in the analysis included: cost of milk production (€0.07 to €0.20 per kg), calf price (€35 to €150 per calf), AI cost (€7 to €24 per AI), calving management cost (€137 to €167 per calving), and culling cost, expressed as the retention pay-off (€118 to €1,117). A partial budget approach was used to calculate the economic effect of varying the VWP from 7 to 15 wk postpartum, using a VWP of 6 wk as reference. Per iteration, the VWP with either the lowest economic loss or the maximum profit was determined as the optimal VWP. The optimal VWP of most cows (90%) was less than 10 wk. On average, every VWP longer than 6 wk gave economic losses. Longer VWPs were in particular optimal for the first parity of breeds other than HF, cows calving in winter with a low milk production, a high milk persistency, a delayed peak milk yield time, a delayed time of first ovulation, occurrence of a postpartum disorder, and low cost of milk production and high cost for AI. (Keywords: dairy cow, optimal voluntary waiting period, milk production, reproductive performance)
55
Analysis of the economically optimal voluntary waiting INTRODUCTION In the Netherlands milk is produced under a quota system, which requires farmers to produce milk as efficiently as possible for economic viability (Halasa et al., 2007). The average Dutch farm milks 60 to 80 lactating cows (CRV, 2009), housing them indoors during the winter and allowing at least partial access to pasture during the rest of the seasons. Suboptimal reproductive performance can have considerable economic consequences, largely due to decreased milk production per cow per yr and a reduction of the number of calves per yr (Inchaisri et al., 2010b, Sørensen and Østergaard, 2003). In general, a calving interval (CI) of 12 to 13 mo is regarded as being economically optimal (Stevenson, 2007). However if attempting to increasing milk production per lactation by shifting the shape of the lactation curve towards a more persistent milk production, a calving interval of 12 to 13 mo may not apply to all cows in an individual herd (Allore and Erb, 2000, van Amburgh et al., 1997). In the Netherlands, the time to start insemination varies among farms and cows. On average, the interval of calving to first insemination is 87 d (Inchaisri et al., 2010a). Estrous synchronization programs combined with timed AI are scarcely used and typical length of the dry period is approximately 2 mo. The average milk production increased from 7,671 kg per cow per yr in 2002 to 8,218 kg per cow per yr in 2009 (CRV, 2009). At the same time, the average CI increased from 402 d in 2002 to 417 d in 2009 (CRV, 2009). Increased CI can be caused by a lower conception rate. The conception rate of a cow is influenced by a number of factors such as the stage of lactation and milk yield (Inchaisri et al., 2010a). Additionally an increased CI may be caused by an extended interval from parturition to first AI. For example, the observed delayed onset of resumption of ovarian follicular activity postpartum (pp) in high producing dairy cows facing a severe negative energy balance is well described, causing the event of the first ovulation pp, and consequently the event of the first AI, to also be delayed (Dhaliwal et al., 1996, Leroy et al., 2008, Lucy, 2001). Moreover, cows show a less prominent expression of estrous behavior due to a negative energy balance (Harrison et al., 1990, López et al., 2004, Roelofs et al., 2010). This may become more important when resumption of ovarian follicular activity is delayed, as in general estrus expression is observed not before the second and following estruses (Dhaliwal et al., 1996, Kinsel and Etherington, 1998, Kyle et al., 1992). Finally, the farmer may also influence the interval of calving to first AI by altering the voluntary waiting period (VWP) based on management decisions. The VWP is defined as the time period (d or wk pp) at which the cow is first eligible for insemination. Because of the increase of milk productions and persistency, farmers may in some instances deliberately choose for an extension of the VWP to allow for longer lactations. In general, a minimal VWP of 45 to 60 d pp is recommended, allowing for complete uterine involution and resumption of normal ovarian cyclicity to improve the 56
Chapter 4 rate of successful conception after AI (Fetrow et al., 2007). Although a shorter CI is regarded as economically optimal, it is suggested that a small extension of the CI may be advantageous in a cow showing a high milk persistency in combination with a high milk production level (Allore and Erb, 2000, Arbel et al., 2001). Since individual cows differ largely in milk production (both in level and shape of the lactation curve), the economically optimal VWP may differ between cows. The interactions and dynamics of factors influencing the optimal VWP of a specific cow make the determination of an optimal VWP for individual cows difficult using field data for analysis. Therefore, simulation modeling may support decisions on an economically optimal VWP and do provide insight in the background of cows for which an extended VWP is profitable. The objective of this study was to explore the optimal length of the VWP for individual cows, taking into account individual cow factors. The described dynamic stochastic simulation model is used to explore the relation between the individual cow factors, economic variables, and the optimal VWP. MATERIALS AND METHODS Model Description A Monte-Carlo stochastic simulation model was developed to calculate the economic effects of a non-optimal fertility status in dairy cows in previous study (Inchaisri et al., 2010b). The model was built in Excel (Microsoft Cooperation, Redmond, WA, USA) with @Risk add-in software (Palisade Corporation, Ithaca, NY, USA). The model is dynamic and uses time periods of one week. For this study, the simulation model was modified to determine the effect of an extended VWP for a specific lactation from calving to the next calving. Every iteration simulates the reproductive cycle (ovulations, estrous detections, and conceptions) of a single cow during one lactation. The maximum length of a lactation was 82 wk, consisting of a maximum interval calving to insemination for 40 wk and a gestation length of 42 wk. Assume there is a number (n) of cows i (i = 1, ….n). For each simulated lactation of cow i, the model simultaneously simulates the reproductive cycle of this lactation of cow i using a VWP of 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 wk pp. The dynamics of the reproductive cycle for the lactation of cow i are based on cow-specific probabilities for the first ovulation, following ovulations, estrous detections, conceptions, and occurrence of postpartum disorders (Inchaisri et al., 2010b). For each time step in the reproductive cycle, the expected reproductive status of the cow, given the input, was simulated using binomial distributions for each specific state of the reproductive cycle. The cow-specific probabilities were based on milk production curves, parity, and DIM simulated for each individual lactation. To simulate the milk production curve, a Wood’s curve (Wood, 1967) was fitted for each lactation, using a simulated 305-d milk yield (MP305i), a time of peak milk production (PTi), and a milk persistency.
57
Analysis of the economically optimal voluntary waiting The 305-d milk yield was based on an average herd 305-d milk yield (HMP) and a cow-specific relative production level, the lactation value (LVi). Outcome variables of the simulation model were for each lactation of cow i with VWPj, the milk production per yr (MPij), calves per yr (CAij), the number of AI per lactation (AIij), the number of non-pregnant cows per yr (CUij), and milk yield at drying off (kg/d). These outcomes were used for further economic analysis. The milk yield at drying off was calculated by taking the simulated milk production level of 224 d (gestation length minus dry period) after the day of conception. Economic Calculations A partial budget approach was used to calculate the economic outcomes for the single lactation of each individual cow using a VWP of 6 wk as the basic situation and VWPs longer than 6 wk as alternatives. This average net economic losses (ANELij) represent the economic effects of the differences in the reproductive outcomes of VWPj compared to a VWP of 6 wk and were calculated within the same lactation of each cow i (VWPj = 7 to 15 wk): ANELij = (MPij - MPi6 ) × Cmp + (CAij - CAi6 ) × Cp + (CAij - CAi6 ) × Ccmij + (AIij -AIi6 ) × Cai + (CUij - CUi6 ) × RPOi [1]
Where Cmp, Cp, Ccmij, Cai, and RPOi indicate respectively the costs of a lower milk production (€/kg), calf price (€), cost of calving management (€), cost of AI (€), and retention pay-off (€). Ccmij is calculated by adjusting a basic Ccm (input) for the estimated milk yield at dry-off milk yield of cow i at VWPj, representing, the increased risk of mastitis in the next lactation following a higher milk yield at drying off (Dingwell et al., 2004) with 10% of incidence rate of postpartum clinical mastitis. RPOi is based on a function of parity and LV of cow i and is defined as the difference in expected future net revenue between a culled cow and its replacing heifer, including the cost of buying and raising this replacing heifer as well as the returns of selling the culled cow. Input Data For each iteration, values of stochastic variables were drawn from relevant distributions (Table 1 and 2). Normal distributions were used for cow variables and uniform distributions were used for economic variables. Input concerning milk production, fertility and economics was based on results of a recent epidemiological study on the probability of conception in Dutch dairy herds (Inchaisri et al., 2010a), reports, and peer-reviewed literature. When necessary, authors’ expertise was used. Input values were based as much as possible on the Dutch dairy farming system. The input variables included factors on cow status (parity, breed, season of calving, the calf status of last calving, HMP, and LV; Table 1), probabilities necessary to simulate the dynamics of a reproductive cycle (Table 1), and economics (Cmp, Cp, Ccm, Cai, 58
Chapter 4 RPO; Table 2). To model the lactation curve, PT and milk persistency were given (Inchaisri et al., 2010a). The lactation curve was adjusted for parity and breed (Inchaisri et al., 2010a). The incidence and the timing of the occurrence of postpartum disorders during lactation were derived from data on cystic ovarian disease (Hooijer et al., 1999, Laporte et al., 1994). These data were used to represent all diseases resulting in a delay of ovulation in addition to the variation in resumption of ovarian follicular activity in early lactation and assuming no effect of postpartum disorders on milk production. The cow-specific probability of estrous was made dependent on 305-d milk yield and daily milk yield (López et al., 2004). The cow-specific probability of conception depended on the breed, parity, milk yield at the date of AI, calf status of last calving, the DIM at the date of AI, the calendar month of AI, and time of AI related to the time of the peak yield (before or after) (Inchaisri et al., 2010a). Breed, parity, calf status of the last calving, and calendar month of calving were generated by discrete distributions with appropriate proportions (Inchaisri et al., 2010a).When a cow was inseminated, the calendar month of AI was determined by summing the calving month and DIM. Moreover, when a cow was pregnant, the milk yield was negatively adjusted due to a gestation effect (Olori et al., 1997). The cost of a lower milk production per cow was calculated under Dutch quota circumstance by taking the marginal costs of keeping extra cows to fill the milk quota. Costs of calving management were based on the costs of drying off, consisting of labor costs and dry cow treatment, costs for veterinary assistance and labor during calving, and costs of peripartum and postpartum disorders. The costs of peripartum and postpartum disorders, such as dystocia, retained placenta, metritis, mastitis, and clinical ketosis depended on the incidence rate of peripartum and postpartum disorders, and the costs per case (Luttikholt et al., 2009). Cows that do not become pregnant do also give economic losses. The RPO values were based on a study of Houben et al. (1994) with price levels adjusted for the current situation (Table 2). The Optimal Week of VWP The model was run for 100,000 iterations, each representing one lactation with all VWP scenarios (VWP of 6 wk to VWP of 15 wk). By using 10 scenarios, the combinations of 1,000,000 events were evaluated. For every iteration, the optimal VWP (wk) was defined as the week in which the ANEL reached the minimum value. Because ANEL is calculated as the difference between the economic losses of the used VWP in relation to a VWP of 6 wk, 6 wk was seen as optimal when ANEL of all other VWP were greater or equal to zero. .
59
60
40 0.07
Norm2(40, 0.86), Truncate2(38.29, 41.71) Norm2(0.07, 0.012), Truncate2(0.03, 0.11)
0.18(exp(0.093 x weeks of gestation))
0.18(exp(0.093 x weeks of gestation)) x Norm3(1, 0.038)
Gestation period (wk) Incidence rate of cystic ovary
7
0.25 0.5
-0.025
Norm2(-0.025, 0.003), Truncate2(-0.035, -0.015) Norm2(6.6, 2.5), Truncate2(2, 10) 0.49, 0.84, 0.96, 0.99, and 1.0 0.96, 1.0, 0.95, and 0.85
50% of estrous detection rate in later stage Norm2(0.5, 0.05), Truncate2(0.30, 0.70)
8422 100
Norm2(8422, 968) Norm2(100, 10) 0.89, 1.04, 1.09, 1.10, and 1.08 1.04, 0.99, 0.82, and 0.85
Estrous detection rate of the 1st ovulation Estrous detection rate in later stage
Winter Female
0.23, 0.19, 0.25, and 0.33 0.44, 0.46, 0.08, and 0.02
5 5 0.95
parity 2 100% HF
0.32, 0.25, 0.17, 0.12, and 0.15 0.63, 0.32, 0.04, and 0.01
LogNorm2(5.36, 5.04), Truncate2(1, 10) LogNorm2(4.53, 3.08), Truncate2(1, 10) 0.95
Default value
Variation value
Reproductive cycle The 1st ovulation time (wk) Parity 1 Parity 2+ Probabilities of the ovulations after the 1st ovulation
Week of peak yield (PTi) Adjusting factors of parity 1, 2, 3, 4, and ≥ 5 on milk persistency Adjusting factors of breed 100% HF1, 50 -
