Activity Monitoring and an Enzyme Immunoassay for Milk Progesterone to Aid in the Detection of Estrus A. S. MOORE' and S. L. SPAHR University of Illinois Department of Animal Sciences 1207 West Gregory Drive Urbana 61801
ABSTRACT
INTRODUCTION
The accuracy and efficiency of estrus detection using an electronic activity monitor tag in conjunction with an enzyme immunoassay for milk progesterone were studied during 55 observed estrous periods in 37 cows. At approximately 30 d postpartum, cows were equipped with an activity tag and visually observed for estrus, an activity tag with a flashing light-emitting diode, or both twice daily. Milk progesterone concentrations were determined from cows observed in estrus or with an activated tag. Mean daily activity was greater on the day of estrus than during any of the 3 d preceding or 3 d following estrus. Functioning activity tags correctly identified 55% of all visually observed estrous periods that coincided with low milk progesterone levels. The overall accuracy of a flag by the tag for identifying true estrus was 21%. The enzyme immunoassay for milk progesterone was in agreement with 98% of all visually observed estrous periods and false flags committed by the tags. Although the activity tags detected some cows in estrus, a more durable and reliable tag must be developed before it is of practical value to dairy producers. (Key words: estrus, activity, milk progesterone)
Perhaps the single most important factor associated with a successful reproductive program is efficient and accurate detection of estrus. Without a good estrus detection regimen, AI can become impractical (23). Visual observance for changes in sexual behavior is the most frequently used method to detect cows in estrus (19). However, visual checks often result in relatively low on-farm estrus detection rates (5, 7, 8, 12, 19). This has created an interest in developing aids designed to improve the efficiency and accuracy of estrus detection on dairy farms. One approach that has shown promise as an estrus detection aid is the monitoring of cow activity. It is well documented that COWS exhibit an increase in physical activity around the time of estrus (9, 11, 12, 25). Mechanical pedometers (10, 11, 12, 25) and electronic activity monitors (1, 4, 13, 17, 18) have been used with limited success to monitor cow activity in an attempt to improve estrus detection rates. One drawback to the use of activity monitoring is the low accuracy of detection due to a large number of false positive identifications. Studies reporting the accuracy of detection have found accuracies to range from 83% (25) to 39% (10). On-farm enzyme immunoassays (EIA) have proven to be rapid and reliable tests for milk progesterone (2, 14) and are useful for identifying errors in estrus detection (3, 15). One use for the EL4 could be to serve as a second screen to check the accuracy of an activity monitoring device. Determination of the progesterone concentration in milk collected on the day when an increase in activity is noted could be used to confirm whether or not the cow in question is truly in estrus. This could greatly improve the accuracy of estrus detection and make the use of activity monitoring more practical.
Abbreviation key: EIA = enzyme immunoassay, LED = light-emitting diode.
Received January 7, 1991. Accepted July 1, 1991. 'Present address: Illinois Stap University, Department of Agriculture, Normal. 1991 J Dairy Sci 74:3857-3862
3857
3858
MOORE AND SPAHR
The objective of this research was to determine the accuracy and efficiency of estrus detection using an electronic activity monitor tag in conjunction with an EIA for milk progesterone. MATERIALS AND METHODS
Approximately 1 wk after freshening, COWS were housed on a grooved concrete lot with an adjacent free-stall barn accomodating up to 40 animals. Cows were confined to the lot at all times except during milking. Thirty-seven Holstein and Brown Swiss cows were equipped with electronic activity monitor tags (Dairy Equipment Company, Madison, WI) approximately 30 d postpartum as they freshened. The tag was strapped around the left rear leg just below the hock. Cows were visually observed and recorded for estrus (n = 55), given a tag with a flashing light-emitting diode (LED), or both twice daily at 0700 and 2200 h for 30 min. Cows that were designated to be bred (41 of the 55 observed estrous periods) were placed in stanchions until serviced, a period that lasted between 30 min and 6 h. Tags remained on cows until 25 d postbreeding without any subsequent return to estrus. The tag contained an electronic microprocessor powered by a nonreplaceable lithium battery. A motion sensor (mercury switch) located within the tag sensed cow movement and incremented a binary counter with each sigdicant movement. The microprocessor stored the activity counts every 2 h up to 7.5 d. The microprocessor then compared the most recent 12 h of activity with a baseline activity level from the same 12-h period during the previous 3 d. The continuous blinking of LED 1. 2, or 3 within the tag represented a two-, three-, or fourfold and greater increase in activity, respectively. Activity data were collected once per week following the manufacturer's instructions (20). This involved placing a transceiver over the tag while the tag was on the cow's leg. The transceiver received and amplified the data transmitted from the tag via the three LED within the tag. Activity data were recorded on a cassette tape through a direct connection of the transceiver to a tape deck and later transferred to floppy disk A 20-ml whole milk sample was obtained from weigh jars after the completion of milkJ o d of Dairy Science Vol. 74. No. 11, 1991
ing, following the observance of a cow in estrus or with an activated tag. Milk was preserved using tablets of potassium dichromate and stored at 4°C until analysis for progesterone was conducted, usually within 1 w k Progesterone concentration was determined on duplicate 10-pl samples of whole milk. A strip assay (Noctech Ltd., Dublin, Ireland) was used to determine milk progesterone concentrations (16). A 5 ng/ml standard, supplied as part of the kit, served as the reference point for the determination of estrus. Cows were classified as in estrus if the optical density of their milk sample was equal to or greater than the optical density of the standard, indicative of milk progesterone concentrations S5 ng/ml. If a cow's milk sample had an optical density below that of the standard (milk progesterone concentrations >5 mg/ml), the cow was classified as not in estrus. No cows were classified as questionable. The optical density of each sample was read at 490 nm using a BIO-TEK EL4 autoreader model EL 310 (BIO-TEX, Burlington, VT). True estrus, defmed as all visually observed estrous periods that coincided with low milk progesterone concentrations, was used as the basis for all comparisons. Manual calculations were performed on activity data to assess the effects of algorithm and threshold ratio on the efficiency and accuracy of estrus detection. The algorithms and threshold ratios used in these comparisons were independent of those utilized by the activity tag. Activity data were obtained from all cows with valid data for d -6 through d +3 relative to estrus (d 0). Four different algorithms were used to calculate activity rates manually at 2-h intervals from identical data based upon 12-, lo-, 8-, and 6-h comparisons. The general equation for the calculation of activity rates was activity rate = sum of the x most recent hours of activity data mean activity count during the same x-hour period for the previous 3 d
where x = 12, 10, 8. or 6.
Three manually calculated threshold ratios, 22.0, 22.5, and 23.0, signifying 22.0-,22.5,
3859
DETECTION OF ESTRUS VIA ACTIVITY MONITORING
and 23.0-fold increases in activity, respectively, were used to identlfy significant increases in cow activity. Each threshold ratio was tested in conjunction with each of the four algorithms. Efficiency of the tag for detecting and flagging cows in estrus was calculated by dividing the number of correct identifications (Le.. true estrous periods correctly flagged) by the total number of true estrous periods observed and multiplying by 100. Flag accuracy was expressed as the number of correct identifications divided by the sum of false-positive flags and correct identifications and then multiplied by 100. Peak 2-h activity count during estrus was compared with the peak 2-h activity count for d 2 and 3 preceding and d 2 and 3 following estrus to determine the percentage of increase in physical activity associated with estrus. Withincow comparisons of increased activity during estrus were compared using the Zstatistic (22). RESULTS AND DISCUSSION
During the course of the 8-mo trial, 44% of all tags had to be replaced because of malfunc-
tions. Vasquez (24) reported a 24% failure rate over a 3-mo period using a similar tag. Others have expressed wncem over the ease by which mechanical pedometers and electronic pedometers were damaged (10). Activity data were analyzed from 37 cows representing 55 visually observed estrous periods that coincided with low milk progesterone concentrations (true estrus). An additional 43 visually observed estrous periods (also supported by EIA) failed to be detected because of tag malfunction and were not included in the analyses. A summary of the efficiencies and accuracies of the tag is presented in Table 1. Functional tags correctly identified slightly more than one-half (55%) of all true estrous periods. The remaining 45% of true estrous periods failed to be detected by the tags. When malfunctioning tags were taken into account, only 31% (30 out of 98) of all estrous periods were detected by the tags. Tags were in use for an average of 48 d per cow. During this time, there were 113 false positives recorded by the tags. Approximately 4 out of every 5 flags registered by the tags were false identifications. The overall accuracy of a flag emitted by a functional tag for identifying true estrus was 21%.
TABLE 1. Efficiencies and accuracies of electronic activity monitor tags for detecting estrus.
nagged activity level' 1
2
3
Total2
15 87 40 27 15
11 22 44 20 33
4 4 51 7 50
30 113
Correct identificatiOllS
False positives False negatives % Efficiency7 % Flag accuracy4
25 55
21
lLight-emitting diodes activated at levels 1, 2, or 3 signify a two-, three-,or fourfold and greater haease in baseline activity, respectively. 2 ~ ~upon t d 55 me estrous periods in 37 cows. 3C0rrect identifications/true estrous periods. 4Correct identifications/(correct identifications + false positives).
Efficiency of estrus detection in this study (55%) is much lower than previous reports using an earlier version of the tag. Previous studies found the tag to be over 90% efficient for detecting estrus (1, 24) and for predicting second and third postpartum ovulations (18). None of the earlier studies reported on the accuracy of the tag. Research conducted with mechanical pedometers has yielded values similar to those reported in this study. Electronic pedometers utilized by Holdsworth and Markillie (10) detected 59% of estrous periods with 39% accuracy. Williams et al. (25) correctly identified 74% of estrous periods with 58% accuracy at 1 SD and 83% accuracy at 2 SD. Lewis and Newman (12) reported that 73% of maximum recorded activity coincided with the day of estrus. The efficiency of the activity tags during this trial was affected by the performance of the microprocessor in the tag. Manual calculation of activity ratios for each estrous period was conducted according to the algorithm (12h) followed by the microprocessor in the tag. According to these calculations, 73% of all estrous periods should have been identified by the activity tags instead of the 55% that were actually flagged. Data transmission was accurate, suggesting that the low efficiency may be due in part to the failure of some microprocessors to perform calculations as expected. This may also account for some of the false positives recorded. Journal of Dairy Science Vol. 74, No. 11, 1991
3860
MOORE AND SPAHR
Mean daily activity was significantly greater (P e .05) on the day of observed estrus than during any of the 3 d preceding or 3 d following estrus. Activity increased 163% on the day of estrus. This increase is comparable with the 186% increase cited by Vasquez (24). Lewis and Newman (12) found the daily increase to be around 225%. Pennington et al. (17) reported that 95% of cows had at least a 100% increase in activity during estrus. Overall mean increase in 2-h peak estrus activity compared with the mean 2-h nonestrus activity was 417%. The individual values in our study ranged from 159 to 953%. Cows showed significant (P c .05) increases in 2-h peak estrus activity during 48 of the 55 (87%) true estrous periods. The mean increase of 417% in this study is similar to increases reported earlier (4, 11, 24), which ranged from 390 to 421%. However, a direct comparison of the values from the present study to those from the literature is impossible because most of the cows in our study were held in stanchions up to 6 h prior to breeding and because each reference used a different time interval to compare increases in activity during estrus. Data concerning the duration of increased activity at estrus were analyzed from 14 COWS not restrained for breeding purposes. The duration of increased activity at estrus (defined as an increase of at least 1.5 SD above the nonestrus mean) for individual cows ranged from 2 to 18 h. The mean duration of increased activity for all cows was 9.6 h. Activity data were only available at 2-h intervals, which limited the precision with which data could be utilized for this analysis. Average duration of increased activity reported in this study is similar to the 10-h duration found by Vasquez (24), whose activity data were available at hourly intervals. The results concerning the effect of algorithm and threshold ratio on activity tag efficiency and accuracy are summarized in Table 2. There was a decrease in efficiency and an increase in flag accuracy as the threshold ratio was increased from 22.0 to 23.0. This trend was apparent regardless of the algorithm used. A second trend was a slight increase in efficiency with a more evident decrease in flag accuracy as the hours used in the calculation of the algorithm were decreased. Taking into account both the efficiency and accuracy of the tag, it appears that the J o d of Dairy Science Vol. 74, No. 11, 1991
TABLE 2. Effect of algorithm and threshold ratio on the efficiency and accuracy of estrus detection by electronic activity tags. Threshold' A l ~ o r i t h m% ~ Efficieac?
% Accuracy4
(h)
22.0
22.5
23.0
12 10 8
73
6
81 62 62 69
77
n
12 10 8 6 12
73 46
10
52
8
60
6
64
100 89 80 59 100 94 97 81 100 100 100 85
lThrcsholds 22.0,22.5, and 23.0 sigaify 22.@, ;?2.5-, and 23.Gfold increases in baseline activity, respectively. ~ H O U Sused to calculate dgorithim. 3Correct ideatifkations/true eseous periods. 4Correct ideatifications/(corct identifications + false positives).
12-h algorithm in conjunction with a threshold ratio of 22.0 offers the best system to detect estrus. It is possible to increase the efficiency of the tag at the expense of accuracy; however, this only detracts from the reliability and practicality of the activity tag as a reproductive aid. This conclusion is in agreement with Vasquez (24), who concluded that a 12-h algorithm was the most effective for detecting estrus. Several specific possibilities for improving the performance of the activity tags were noted. The most obvious need was to increase the reliability of the tags to operate over an extended period of time. Possibilities for failure include poor battery performance, microcracks in the potting material that allow moisture to seep into the electronics, and failure of electronic components because of mechanical shock during use or because of spontaneous failure. Possibilities for increasing the efficiency of the tag while decreasing the occurrence of false positives include changing the sensitivity or dampening of the mercury switch to record a count, changing the angle of the mercury switch, improving the reliability of the microprocessor to start the flashing LED at the appropriate time, and adjusting the algorithm used to switch on the flashing LED.
DETECTION OF ESTRUS VIA ACTIVITY MONITORING
Application of the system to a commercial basis also would be made more attractive if the requirement for visual observation of the LED were replaced by a capability for automatic acquisition of data from the tag, especially if the data could be examined for increased activity at several adjacent 2-h periods. The EIA agreed with the incidence of observed estrus in 98% of all cases. These results are similar to the findings of others who report that EIA for milk progesterone accurately confirmed 96 to 100% of all estrous periods (6, 21). There were five separate cases in which a cycling cow had an activated tag and milk progesterone concentrations below 5 n g / d 18 to 27 d after the last recorded estrus but was never observed in estrus. It is very likely, considering the efficiency of visual observations, that these 5 cows were correctly identified as in estrus by the tag and by EL4 but were either missed by the observers or exhibited no signs of standing estrus. These results indicate that the EIA is a very useful tool for validating flags made by activity tags. In addition, there is a possibility that activity tags used in conjunction with the EIA for milk progesterone can assist in the detection of estrus among problem cows, such as those having short duration and silent estrous periods. CONCLUSIONS
The efficacy of electronic activity monitor tags for estrus detection when used in conjunction with an EIA for mi& progesterone was examined over an 8-mo period. During this trial, 44% of the tags malfunctioned. Functioning activity tags successfully identified 55% of visually observed estrous periods that coincided with low milk progesterone concentrations. There were 113 false positives during the period of use, resulting in an overall flag accuracy of 21%. Activity increased an average of 163% on the day of estrus. The average duration of increased physical activity associated with estrus was 9.6 h. The mean increase in peak 2-h activity during estrus compared with mean 2-h nonestrus activity was 417%. Eighty-seven percent of all estrous periods were characterized by a significant increase (P e .05) in peak 2-h activity, Analysis of four different algorithms for calcdating activity ra-
3861
tios confirmed that the 12-h algorithm is more effective than lo-, 8-, or 6-h algorithms for the efficient and accurate detection of estrus. The EIA agreed with 98% of all visually observed estrous periods and correctly identified 111 out of 113 (98%) of the false flags committed by the tags. The EIA proved to be an invaluable tool for screening cows flagged by tags. In conclusion, electronic activity tags can detect some cows in estrus. However, a more durable and reliable tag must be developed before the tags are of practical value to dairy farmers. The electronic performance should be improved to increase the efficiency of the tag while decreasing the occurrence of false positives. ACKNOWLEDGMENTS
This study was a contribution of Project Number 35-0343 of the AgricUltural Experiment Station, University of Illinois at UrbanaChampaign, as part of the work of NC 119, Dairy Herd Management Strategies for Improved Decision Making and Profitability. It was supported in part by Grant US-439-82 from the US-Israel Binational Agriculture Research and Development Program and by a gift from the Commodity Credit Corporation, USDA, Washington, DC. Appreciation is expressed to Dairy Quipment Company, Madison, WI for supplying electronic activity tags. The authors gratefully acknowledge the technical assistance of D. E. Dill, R. J. Favero, T. A. Gipson, and R. S. Serber. REFERENCES 1 Armstrong, D. V., F. Wiersmao, C. P. Hammond, and D. S. Ammom 1984. Evaluation of an activity monitoring device to detect estrus in dairy cattle. J. Dairy Sci. 67(Supp1. 1):157.(Abstr.) 2 Amstadt, K. I., and W. F. Cleere.. 1981. Enzymeimmunoassay for determination of progesterone in milk from cows. J. Reprod. Fed. 62173. 3 Eddy, R. G., and P. J. Clark. 1987. Oestrus prediction in diury cows using an ELISA progesterone test. Vet. Rec. 12031. 4 Favero, R J.. S. L. Spahr, H. B. Puckett, and H. L. Whitmore. 1984. Rear leg, front leg, and neck for measufemcnt of increased activity at estrus. J. Dairy Sci. 67(Suppl. 1):156.(Abstr.) 5 Fon~e~a, F.A., J. H.Britt, B. T. McDaniel, I. C. Wik, and A. H. Rakes. 1983. Reproductive kaits of Holsteins and Jerseys. Effects of age, milk yield, and clinical abnormalities on involution of cervix and uterus, owlation, estrous cycles, detection of estrus,
Journal of Dairy Science Vol. 74, No. 11, 1991
3862
MOORE AND SPAHR
conception rate, and days open. J. Dairy Sci. 66:1128. 6 Foulkes, J. A., A. D. Cookson, and M. J. Sauer. 1982. AI in cattle based on daily microtiter plate enzymeimmunoassay of progesterone in whole milk. Br. Vet. J. 138515. 7 FulkQsoq W. J., G. J. Sawyer, and I. Crothtrs. 1983. The accuracy of several aids in detecting oestrus in dairy cattle. Appl. Anim. Ethol. 1O:lW. 8Gwazdauskas, F? C., J. A. Lineweaver, and M. L. McGilliard. 1983. Winmental and management factors affecting estrous activity in dairy cattle. J. Dairy Sci. 66:1510. 9 Hammond, J. 1927. The physiology of reproduction in the cow. Cambridge Univ. Ress, New York NY. 10 Holdsworth, R J., and N.A.R. Markillie. 1982. Evaluation of pedometers for oestrus detection in dairy cows. Vet. Rec. 111:16. 11 Eddy, C. A. 1977. Variation in physical activity as an indication of estnw: in dairy cows. J. Dairy Sci. 60: 235. 12Lewis, G. S., and S. K. Newman. 1984. Changes throughout estrous cycles of variables that might indicate estrus in dairy cows. J. Dairy Sci. 67:146. 13Mmre, A. S., and S. L. Spahr. 1986. Electronic activity tags and milk progesterone enzyme immunoassays as aids for reproductive management. J. Dairy Sci. 69(Suppl. 1):92.(Abstr.) 14Morris, D.. and J. M. Srecnan. 1983. New enzyme tests to improve animal health and production: routine practical applications. Western Research Centre, Belcare, Ireland. 15 Nebel, R. L., W. D. Whittier, B. G. Cassell, and J. H. Britt. 1987. Comparison of on-farm and laboratory milk progesterone assays for identifying errors in detection of estrus and diagnosis of pregnancy. J.
Journal of Dairy Science Vol. 74, No. 11, 1991
Dairy Sci. 701471. 16Noctech LTD. 1984. Enzyme immunoassay for the determination of progesterone in bovine milk.Noctech Ltd., Dublin, Ireland. 17Pennington, J. A., J. L. Albright, and C. J. callahaa 1986. Relationships of sexual activities in estrous cows to different frequencies of observation and pcdomeler meawwnents. J. Dairy Sci. 69:2925. 18Peter. A. T., and W.TX. Bosu. 1986. Postpartum ovarian activity in dairy cows: correlation between behavioral estrus, pedometer measuranents and o w lations. Thenogenology 26:ll l. 19Reimm. T. J.. R. D. Smith, and S. K. Newman. 1985. Management factors affecting reproductive paformance of dairy cows in the northeastern United States. J. Dairy Sci. 68:%3. 20Rodrian. J. A. 1984. Research support packageactivity monitor. EL SYN Inc., Grafton, WI. 21 Romagnolo, D., L. Buttazmni, R. Aleandri. and R. L. Nebel. 1989. Estrous validation and early pregnancy diagnosis by two on-farm milk progesterone kits. J. Dairy Sci. 72(Suppl. 1):454.(Abstr.) 22Steel RG.D.. and J. H. Torre. 1980. Rinciples and procedures of statistics. 2nd ed. McGraw-Hill, New Yolk, NY. 23 Ihompson, P. D., and J. A. Rodriaa 1983. Transducers for capture of activity data Page 115 in Proc Symp. Automation Dairying. Wageningen, Neth. 24Vasquez. L. H. 1985. Relationship between physical activity, stage of estrous cycle, sexual behavior and plasma luteinizing hormone in dairy cattle. Ph.D. ~ i s s . ,univ. nlinois, Urbana. 25 Williams, W. F.. D. R. Yver, and T. S. Gross. 1981. Comparison of estNS detection tecbmques in dairy heifers. J. Dairy Sci. W1738.
ABSTRACT
INTRODUCTION
The accuracy and efficiency of estrus detection using an electronic activity monitor tag in conjunction with an enzyme immunoassay for milk progesterone were studied during 55 observed estrous periods in 37 cows. At approximately 30 d postpartum, cows were equipped with an activity tag and visually observed for estrus, an activity tag with a flashing light-emitting diode, or both twice daily. Milk progesterone concentrations were determined from cows observed in estrus or with an activated tag. Mean daily activity was greater on the day of estrus than during any of the 3 d preceding or 3 d following estrus. Functioning activity tags correctly identified 55% of all visually observed estrous periods that coincided with low milk progesterone levels. The overall accuracy of a flag by the tag for identifying true estrus was 21%. The enzyme immunoassay for milk progesterone was in agreement with 98% of all visually observed estrous periods and false flags committed by the tags. Although the activity tags detected some cows in estrus, a more durable and reliable tag must be developed before it is of practical value to dairy producers. (Key words: estrus, activity, milk progesterone)
Perhaps the single most important factor associated with a successful reproductive program is efficient and accurate detection of estrus. Without a good estrus detection regimen, AI can become impractical (23). Visual observance for changes in sexual behavior is the most frequently used method to detect cows in estrus (19). However, visual checks often result in relatively low on-farm estrus detection rates (5, 7, 8, 12, 19). This has created an interest in developing aids designed to improve the efficiency and accuracy of estrus detection on dairy farms. One approach that has shown promise as an estrus detection aid is the monitoring of cow activity. It is well documented that COWS exhibit an increase in physical activity around the time of estrus (9, 11, 12, 25). Mechanical pedometers (10, 11, 12, 25) and electronic activity monitors (1, 4, 13, 17, 18) have been used with limited success to monitor cow activity in an attempt to improve estrus detection rates. One drawback to the use of activity monitoring is the low accuracy of detection due to a large number of false positive identifications. Studies reporting the accuracy of detection have found accuracies to range from 83% (25) to 39% (10). On-farm enzyme immunoassays (EIA) have proven to be rapid and reliable tests for milk progesterone (2, 14) and are useful for identifying errors in estrus detection (3, 15). One use for the EL4 could be to serve as a second screen to check the accuracy of an activity monitoring device. Determination of the progesterone concentration in milk collected on the day when an increase in activity is noted could be used to confirm whether or not the cow in question is truly in estrus. This could greatly improve the accuracy of estrus detection and make the use of activity monitoring more practical.
Abbreviation key: EIA = enzyme immunoassay, LED = light-emitting diode.
Received January 7, 1991. Accepted July 1, 1991. 'Present address: Illinois Stap University, Department of Agriculture, Normal. 1991 J Dairy Sci 74:3857-3862
3857
3858
MOORE AND SPAHR
The objective of this research was to determine the accuracy and efficiency of estrus detection using an electronic activity monitor tag in conjunction with an EIA for milk progesterone. MATERIALS AND METHODS
Approximately 1 wk after freshening, COWS were housed on a grooved concrete lot with an adjacent free-stall barn accomodating up to 40 animals. Cows were confined to the lot at all times except during milking. Thirty-seven Holstein and Brown Swiss cows were equipped with electronic activity monitor tags (Dairy Equipment Company, Madison, WI) approximately 30 d postpartum as they freshened. The tag was strapped around the left rear leg just below the hock. Cows were visually observed and recorded for estrus (n = 55), given a tag with a flashing light-emitting diode (LED), or both twice daily at 0700 and 2200 h for 30 min. Cows that were designated to be bred (41 of the 55 observed estrous periods) were placed in stanchions until serviced, a period that lasted between 30 min and 6 h. Tags remained on cows until 25 d postbreeding without any subsequent return to estrus. The tag contained an electronic microprocessor powered by a nonreplaceable lithium battery. A motion sensor (mercury switch) located within the tag sensed cow movement and incremented a binary counter with each sigdicant movement. The microprocessor stored the activity counts every 2 h up to 7.5 d. The microprocessor then compared the most recent 12 h of activity with a baseline activity level from the same 12-h period during the previous 3 d. The continuous blinking of LED 1. 2, or 3 within the tag represented a two-, three-, or fourfold and greater increase in activity, respectively. Activity data were collected once per week following the manufacturer's instructions (20). This involved placing a transceiver over the tag while the tag was on the cow's leg. The transceiver received and amplified the data transmitted from the tag via the three LED within the tag. Activity data were recorded on a cassette tape through a direct connection of the transceiver to a tape deck and later transferred to floppy disk A 20-ml whole milk sample was obtained from weigh jars after the completion of milkJ o d of Dairy Science Vol. 74. No. 11, 1991
ing, following the observance of a cow in estrus or with an activated tag. Milk was preserved using tablets of potassium dichromate and stored at 4°C until analysis for progesterone was conducted, usually within 1 w k Progesterone concentration was determined on duplicate 10-pl samples of whole milk. A strip assay (Noctech Ltd., Dublin, Ireland) was used to determine milk progesterone concentrations (16). A 5 ng/ml standard, supplied as part of the kit, served as the reference point for the determination of estrus. Cows were classified as in estrus if the optical density of their milk sample was equal to or greater than the optical density of the standard, indicative of milk progesterone concentrations S5 ng/ml. If a cow's milk sample had an optical density below that of the standard (milk progesterone concentrations >5 mg/ml), the cow was classified as not in estrus. No cows were classified as questionable. The optical density of each sample was read at 490 nm using a BIO-TEK EL4 autoreader model EL 310 (BIO-TEX, Burlington, VT). True estrus, defmed as all visually observed estrous periods that coincided with low milk progesterone concentrations, was used as the basis for all comparisons. Manual calculations were performed on activity data to assess the effects of algorithm and threshold ratio on the efficiency and accuracy of estrus detection. The algorithms and threshold ratios used in these comparisons were independent of those utilized by the activity tag. Activity data were obtained from all cows with valid data for d -6 through d +3 relative to estrus (d 0). Four different algorithms were used to calculate activity rates manually at 2-h intervals from identical data based upon 12-, lo-, 8-, and 6-h comparisons. The general equation for the calculation of activity rates was activity rate = sum of the x most recent hours of activity data mean activity count during the same x-hour period for the previous 3 d
where x = 12, 10, 8. or 6.
Three manually calculated threshold ratios, 22.0, 22.5, and 23.0, signifying 22.0-,22.5,
3859
DETECTION OF ESTRUS VIA ACTIVITY MONITORING
and 23.0-fold increases in activity, respectively, were used to identlfy significant increases in cow activity. Each threshold ratio was tested in conjunction with each of the four algorithms. Efficiency of the tag for detecting and flagging cows in estrus was calculated by dividing the number of correct identifications (Le.. true estrous periods correctly flagged) by the total number of true estrous periods observed and multiplying by 100. Flag accuracy was expressed as the number of correct identifications divided by the sum of false-positive flags and correct identifications and then multiplied by 100. Peak 2-h activity count during estrus was compared with the peak 2-h activity count for d 2 and 3 preceding and d 2 and 3 following estrus to determine the percentage of increase in physical activity associated with estrus. Withincow comparisons of increased activity during estrus were compared using the Zstatistic (22). RESULTS AND DISCUSSION
During the course of the 8-mo trial, 44% of all tags had to be replaced because of malfunc-
tions. Vasquez (24) reported a 24% failure rate over a 3-mo period using a similar tag. Others have expressed wncem over the ease by which mechanical pedometers and electronic pedometers were damaged (10). Activity data were analyzed from 37 cows representing 55 visually observed estrous periods that coincided with low milk progesterone concentrations (true estrus). An additional 43 visually observed estrous periods (also supported by EIA) failed to be detected because of tag malfunction and were not included in the analyses. A summary of the efficiencies and accuracies of the tag is presented in Table 1. Functional tags correctly identified slightly more than one-half (55%) of all true estrous periods. The remaining 45% of true estrous periods failed to be detected by the tags. When malfunctioning tags were taken into account, only 31% (30 out of 98) of all estrous periods were detected by the tags. Tags were in use for an average of 48 d per cow. During this time, there were 113 false positives recorded by the tags. Approximately 4 out of every 5 flags registered by the tags were false identifications. The overall accuracy of a flag emitted by a functional tag for identifying true estrus was 21%.
TABLE 1. Efficiencies and accuracies of electronic activity monitor tags for detecting estrus.
nagged activity level' 1
2
3
Total2
15 87 40 27 15
11 22 44 20 33
4 4 51 7 50
30 113
Correct identificatiOllS
False positives False negatives % Efficiency7 % Flag accuracy4
25 55
21
lLight-emitting diodes activated at levels 1, 2, or 3 signify a two-, three-,or fourfold and greater haease in baseline activity, respectively. 2 ~ ~upon t d 55 me estrous periods in 37 cows. 3C0rrect identifications/true estrous periods. 4Correct identifications/(correct identifications + false positives).
Efficiency of estrus detection in this study (55%) is much lower than previous reports using an earlier version of the tag. Previous studies found the tag to be over 90% efficient for detecting estrus (1, 24) and for predicting second and third postpartum ovulations (18). None of the earlier studies reported on the accuracy of the tag. Research conducted with mechanical pedometers has yielded values similar to those reported in this study. Electronic pedometers utilized by Holdsworth and Markillie (10) detected 59% of estrous periods with 39% accuracy. Williams et al. (25) correctly identified 74% of estrous periods with 58% accuracy at 1 SD and 83% accuracy at 2 SD. Lewis and Newman (12) reported that 73% of maximum recorded activity coincided with the day of estrus. The efficiency of the activity tags during this trial was affected by the performance of the microprocessor in the tag. Manual calculation of activity ratios for each estrous period was conducted according to the algorithm (12h) followed by the microprocessor in the tag. According to these calculations, 73% of all estrous periods should have been identified by the activity tags instead of the 55% that were actually flagged. Data transmission was accurate, suggesting that the low efficiency may be due in part to the failure of some microprocessors to perform calculations as expected. This may also account for some of the false positives recorded. Journal of Dairy Science Vol. 74, No. 11, 1991
3860
MOORE AND SPAHR
Mean daily activity was significantly greater (P e .05) on the day of observed estrus than during any of the 3 d preceding or 3 d following estrus. Activity increased 163% on the day of estrus. This increase is comparable with the 186% increase cited by Vasquez (24). Lewis and Newman (12) found the daily increase to be around 225%. Pennington et al. (17) reported that 95% of cows had at least a 100% increase in activity during estrus. Overall mean increase in 2-h peak estrus activity compared with the mean 2-h nonestrus activity was 417%. The individual values in our study ranged from 159 to 953%. Cows showed significant (P c .05) increases in 2-h peak estrus activity during 48 of the 55 (87%) true estrous periods. The mean increase of 417% in this study is similar to increases reported earlier (4, 11, 24), which ranged from 390 to 421%. However, a direct comparison of the values from the present study to those from the literature is impossible because most of the cows in our study were held in stanchions up to 6 h prior to breeding and because each reference used a different time interval to compare increases in activity during estrus. Data concerning the duration of increased activity at estrus were analyzed from 14 COWS not restrained for breeding purposes. The duration of increased activity at estrus (defined as an increase of at least 1.5 SD above the nonestrus mean) for individual cows ranged from 2 to 18 h. The mean duration of increased activity for all cows was 9.6 h. Activity data were only available at 2-h intervals, which limited the precision with which data could be utilized for this analysis. Average duration of increased activity reported in this study is similar to the 10-h duration found by Vasquez (24), whose activity data were available at hourly intervals. The results concerning the effect of algorithm and threshold ratio on activity tag efficiency and accuracy are summarized in Table 2. There was a decrease in efficiency and an increase in flag accuracy as the threshold ratio was increased from 22.0 to 23.0. This trend was apparent regardless of the algorithm used. A second trend was a slight increase in efficiency with a more evident decrease in flag accuracy as the hours used in the calculation of the algorithm were decreased. Taking into account both the efficiency and accuracy of the tag, it appears that the J o d of Dairy Science Vol. 74, No. 11, 1991
TABLE 2. Effect of algorithm and threshold ratio on the efficiency and accuracy of estrus detection by electronic activity tags. Threshold' A l ~ o r i t h m% ~ Efficieac?
% Accuracy4
(h)
22.0
22.5
23.0
12 10 8
73
6
81 62 62 69
77
n
12 10 8 6 12
73 46
10
52
8
60
6
64
100 89 80 59 100 94 97 81 100 100 100 85
lThrcsholds 22.0,22.5, and 23.0 sigaify 22.@, ;?2.5-, and 23.Gfold increases in baseline activity, respectively. ~ H O U Sused to calculate dgorithim. 3Correct ideatifkations/true eseous periods. 4Correct ideatifications/(corct identifications + false positives).
12-h algorithm in conjunction with a threshold ratio of 22.0 offers the best system to detect estrus. It is possible to increase the efficiency of the tag at the expense of accuracy; however, this only detracts from the reliability and practicality of the activity tag as a reproductive aid. This conclusion is in agreement with Vasquez (24), who concluded that a 12-h algorithm was the most effective for detecting estrus. Several specific possibilities for improving the performance of the activity tags were noted. The most obvious need was to increase the reliability of the tags to operate over an extended period of time. Possibilities for failure include poor battery performance, microcracks in the potting material that allow moisture to seep into the electronics, and failure of electronic components because of mechanical shock during use or because of spontaneous failure. Possibilities for increasing the efficiency of the tag while decreasing the occurrence of false positives include changing the sensitivity or dampening of the mercury switch to record a count, changing the angle of the mercury switch, improving the reliability of the microprocessor to start the flashing LED at the appropriate time, and adjusting the algorithm used to switch on the flashing LED.
DETECTION OF ESTRUS VIA ACTIVITY MONITORING
Application of the system to a commercial basis also would be made more attractive if the requirement for visual observation of the LED were replaced by a capability for automatic acquisition of data from the tag, especially if the data could be examined for increased activity at several adjacent 2-h periods. The EIA agreed with the incidence of observed estrus in 98% of all cases. These results are similar to the findings of others who report that EIA for milk progesterone accurately confirmed 96 to 100% of all estrous periods (6, 21). There were five separate cases in which a cycling cow had an activated tag and milk progesterone concentrations below 5 n g / d 18 to 27 d after the last recorded estrus but was never observed in estrus. It is very likely, considering the efficiency of visual observations, that these 5 cows were correctly identified as in estrus by the tag and by EL4 but were either missed by the observers or exhibited no signs of standing estrus. These results indicate that the EIA is a very useful tool for validating flags made by activity tags. In addition, there is a possibility that activity tags used in conjunction with the EIA for milk progesterone can assist in the detection of estrus among problem cows, such as those having short duration and silent estrous periods. CONCLUSIONS
The efficacy of electronic activity monitor tags for estrus detection when used in conjunction with an EIA for mi& progesterone was examined over an 8-mo period. During this trial, 44% of the tags malfunctioned. Functioning activity tags successfully identified 55% of visually observed estrous periods that coincided with low milk progesterone concentrations. There were 113 false positives during the period of use, resulting in an overall flag accuracy of 21%. Activity increased an average of 163% on the day of estrus. The average duration of increased physical activity associated with estrus was 9.6 h. The mean increase in peak 2-h activity during estrus compared with mean 2-h nonestrus activity was 417%. Eighty-seven percent of all estrous periods were characterized by a significant increase (P e .05) in peak 2-h activity, Analysis of four different algorithms for calcdating activity ra-
3861
tios confirmed that the 12-h algorithm is more effective than lo-, 8-, or 6-h algorithms for the efficient and accurate detection of estrus. The EIA agreed with 98% of all visually observed estrous periods and correctly identified 111 out of 113 (98%) of the false flags committed by the tags. The EIA proved to be an invaluable tool for screening cows flagged by tags. In conclusion, electronic activity tags can detect some cows in estrus. However, a more durable and reliable tag must be developed before the tags are of practical value to dairy farmers. The electronic performance should be improved to increase the efficiency of the tag while decreasing the occurrence of false positives. ACKNOWLEDGMENTS
This study was a contribution of Project Number 35-0343 of the AgricUltural Experiment Station, University of Illinois at UrbanaChampaign, as part of the work of NC 119, Dairy Herd Management Strategies for Improved Decision Making and Profitability. It was supported in part by Grant US-439-82 from the US-Israel Binational Agriculture Research and Development Program and by a gift from the Commodity Credit Corporation, USDA, Washington, DC. Appreciation is expressed to Dairy Quipment Company, Madison, WI for supplying electronic activity tags. The authors gratefully acknowledge the technical assistance of D. E. Dill, R. J. Favero, T. A. Gipson, and R. S. Serber. REFERENCES 1 Armstrong, D. V., F. Wiersmao, C. P. Hammond, and D. S. Ammom 1984. Evaluation of an activity monitoring device to detect estrus in dairy cattle. J. Dairy Sci. 67(Supp1. 1):157.(Abstr.) 2 Amstadt, K. I., and W. F. Cleere.. 1981. Enzymeimmunoassay for determination of progesterone in milk from cows. J. Reprod. Fed. 62173. 3 Eddy, R. G., and P. J. Clark. 1987. Oestrus prediction in diury cows using an ELISA progesterone test. Vet. Rec. 12031. 4 Favero, R J.. S. L. Spahr, H. B. Puckett, and H. L. Whitmore. 1984. Rear leg, front leg, and neck for measufemcnt of increased activity at estrus. J. Dairy Sci. 67(Suppl. 1):156.(Abstr.) 5 Fon~e~a, F.A., J. H.Britt, B. T. McDaniel, I. C. Wik, and A. H. Rakes. 1983. Reproductive kaits of Holsteins and Jerseys. Effects of age, milk yield, and clinical abnormalities on involution of cervix and uterus, owlation, estrous cycles, detection of estrus,
Journal of Dairy Science Vol. 74, No. 11, 1991
3862
MOORE AND SPAHR
conception rate, and days open. J. Dairy Sci. 66:1128. 6 Foulkes, J. A., A. D. Cookson, and M. J. Sauer. 1982. AI in cattle based on daily microtiter plate enzymeimmunoassay of progesterone in whole milk. Br. Vet. J. 138515. 7 FulkQsoq W. J., G. J. Sawyer, and I. Crothtrs. 1983. The accuracy of several aids in detecting oestrus in dairy cattle. Appl. Anim. Ethol. 1O:lW. 8Gwazdauskas, F? C., J. A. Lineweaver, and M. L. McGilliard. 1983. Winmental and management factors affecting estrous activity in dairy cattle. J. Dairy Sci. 66:1510. 9 Hammond, J. 1927. The physiology of reproduction in the cow. Cambridge Univ. Ress, New York NY. 10 Holdsworth, R J., and N.A.R. Markillie. 1982. Evaluation of pedometers for oestrus detection in dairy cows. Vet. Rec. 111:16. 11 Eddy, C. A. 1977. Variation in physical activity as an indication of estnw: in dairy cows. J. Dairy Sci. 60: 235. 12Lewis, G. S., and S. K. Newman. 1984. Changes throughout estrous cycles of variables that might indicate estrus in dairy cows. J. Dairy Sci. 67:146. 13Mmre, A. S., and S. L. Spahr. 1986. Electronic activity tags and milk progesterone enzyme immunoassays as aids for reproductive management. J. Dairy Sci. 69(Suppl. 1):92.(Abstr.) 14Morris, D.. and J. M. Srecnan. 1983. New enzyme tests to improve animal health and production: routine practical applications. Western Research Centre, Belcare, Ireland. 15 Nebel, R. L., W. D. Whittier, B. G. Cassell, and J. H. Britt. 1987. Comparison of on-farm and laboratory milk progesterone assays for identifying errors in detection of estrus and diagnosis of pregnancy. J.
Journal of Dairy Science Vol. 74, No. 11, 1991
Dairy Sci. 701471. 16Noctech LTD. 1984. Enzyme immunoassay for the determination of progesterone in bovine milk.Noctech Ltd., Dublin, Ireland. 17Pennington, J. A., J. L. Albright, and C. J. callahaa 1986. Relationships of sexual activities in estrous cows to different frequencies of observation and pcdomeler meawwnents. J. Dairy Sci. 69:2925. 18Peter. A. T., and W.TX. Bosu. 1986. Postpartum ovarian activity in dairy cows: correlation between behavioral estrus, pedometer measuranents and o w lations. Thenogenology 26:ll l. 19Reimm. T. J.. R. D. Smith, and S. K. Newman. 1985. Management factors affecting reproductive paformance of dairy cows in the northeastern United States. J. Dairy Sci. 68:%3. 20Rodrian. J. A. 1984. Research support packageactivity monitor. EL SYN Inc., Grafton, WI. 21 Romagnolo, D., L. Buttazmni, R. Aleandri. and R. L. Nebel. 1989. Estrous validation and early pregnancy diagnosis by two on-farm milk progesterone kits. J. Dairy Sci. 72(Suppl. 1):454.(Abstr.) 22Steel RG.D.. and J. H. Torre. 1980. Rinciples and procedures of statistics. 2nd ed. McGraw-Hill, New Yolk, NY. 23 Ihompson, P. D., and J. A. Rodriaa 1983. Transducers for capture of activity data Page 115 in Proc Symp. Automation Dairying. Wageningen, Neth. 24Vasquez. L. H. 1985. Relationship between physical activity, stage of estrous cycle, sexual behavior and plasma luteinizing hormone in dairy cattle. Ph.D. ~ i s s . ,univ. nlinois, Urbana. 25 Williams, W. F.. D. R. Yver, and T. S. Gross. 1981. Comparison of estNS detection tecbmques in dairy heifers. J. Dairy Sci. W1738.
