| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853
Department of Population Medicine, Ontario Veterinary College, University of Guelph, Ontario, Canada
1 Corresponding author: rcb28{at}cornell.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
3 were hoof trimmed and the presence or absence of a painful lesion (PL), defined as a reaction to digital pressure, was recorded. A strongly increasing pattern in the proportion of cows with PL was detected as VLS increased. The proportions of cows with painful lesions were 5.6% (n = 53), 20.1% (n = 78), 55.5% (n = 164), 79.9% (n = 159), and 100% (n = 5) for VLS 1 to 5, respectively. Study 2 was conducted on a different farm. The entire farm was visually locomotion scored by 3 veterinarians on the same day, and the cows were Stepmetrix locomotion scored by walking through the Stepmetrix system. Every cow was trimmed during the following 2 d by 1 of 8 professional hoof trimmers. The 3 veterinarians identified, scored, and recorded any PL. Interobserver agreement for the 3 veterinarians had a kappa coefficient of between 0.45 and 0.48 ± 0.05. In total, 518 cows were used in the analysis, from which 11.2% were identified with a PL. Of the cows diagnosed with a PL, 32.8% were detected with a sole ulcer, 25.9% with white line disease, 13.8% with white line abscess, and 27.5% with other diseases. A receiver operating characteristic analysis was performed; the area under the curve was larger for VLS (0.80; 95% confidence interval, 0.76 to 0.83) than SLS (0.62; 95% confidence interval, 0.57 to 0.66). When performed by trained veterinarians, VLS performed better than SLS in detecting PL.
Key Words: lameness • Stepmetrix • dairy cow
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Claw disorders account for approximately 90% of lameness in dairy cattle (Murray et al., 1996). Early detection and treatment of lameness reduces economic losses and the prevalence of lameness (Clarkson et al., 1996; Espejo et al., 2006). Lameness is usually detected by visual observation of abnormal locomotion by the farmer or farm personnel. As farms grow larger, the ratio of cows to farm personnel increases, and this reduces the likelihood that a lame cow will be detected and treated in a timely fashion. This has led to interest in automated methods of lameness detection (Neveux et al., 2006). Whay et al. (2003) reported that dairy producers are aware of only 25 to 50% of the cows that are lame, and Wells et al. (1993) reported that producers correctly identified only 40 to 45% of lame cows. Scoring herds for locomotion is important for observing the prevalence of lameness in the herd and detecting cows that require treatment.
Biomechanical technology for lameness detection in cattle has been researched, such as the video motion analysis by Flower et al. (2005) and ground reaction force detection by Rajkondawar et al. (2006). Although studies have shown encouraging results, the use of automated lameness detection is not widely used. One automated lameness detection system, Stepmetrix (BouMatic, Madison, WI), which uses ground reaction force measurements, is commercially available.
Several visual locomotion scoring (VLS) systems have been described (Sprecher et al., 1997; Engel et al., 2003); however, the lack of a gold standard for the presence of lameness makes it difficult to evaluate the accuracy and optimal cutoff point for intervention. The presence of a painful lesion in the foot, such as a sole ulcer or an abscess, which is diagnosed at hoof trimming, should be associated with an abnormal gait.
The objectives of this study were to evaluate the association between VLS and an automated locomotion scoring system (Stepmetrix locomotion scoring, SLS) in the presence or absence of painful digit lesions to compare the accuracy of VLS vs. SLS for detecting painful foot lesions. Our hypotheses were that a significant association would be found between VLS or SLS and painful foot lesions, and that VLS would have a higher accuracy in detecting painful foot lesions than SLS.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Design and Data Collection.
A prospective observational study design was used. Three pens of lactating dairy cows, 1 pen with 310 cows and 2 pens with 150 cows each, were enrolled in the study. The criteria for selection of the pens were based on pen location, farm milking schedule, and ability of the veterinarian present to score cows at a fixed time every week. Enrolled cows were visual locomotion scored as the cows exited the milking parlor. Scoring was done by the same veterinarian and the system used was a 5-point scoring scale, where 1 = normal, 2 = presence of a slightly asymmetric gait, 3 = cow clearly favors one or more limbs (moderately lame), 4 = severely lame, or 5 = extremely lame (non-weight-bearing lame). The veterinarian recorded the cow identification number, VLS, and suspected affected limb for VLS 3 into a digital voice recorder while the cows were walking by. There were 7 categories of affected limbs by visual assessment: left front, left hind, right front, right hind, both hind, both front, and all limbs. On the same day, the data were entered into a database (Access, Microsoft Corporation, Redmond, WA) created for the study.
A parallel clinical study was done to evaluate the effect of treatment on lame cows using 2 different VLS thresholds. At the beginning of the study, every cow on the farm was randomly assigned a code of 1 or 2. Code 1 cows would be considered lame on a VLS of 4 or 5 (LVLS4), and code 2 cows were considered lame if VLS was 3, 4, or 5 (LVLS3). Therefore, only a portion of the cows with a VLS of 3 were treated.
Criteria for treatment were a VLS of 3 (for LVLS3 cows), 4, or 5, and cows scheduled for routine hoof trimming. In total, 246 cows were allocated to trimming because they were detected with a VLS 3, and 213 cows were allocated to trimming because they were due for a routine trim. Selected cows were trimmed the day after VLS or the following day if too many cows were scheduled for treatment. The treatment lists were generated at the end of data entry each week using the study database in Microsoft Access according to historical treatment data and VLS. Treated cows would not return to the trimming table for the following 2 wk unless an increase in locomotion score was observed for that period. All hoof trimmings were done by the same veterinarian. Cows were trimmed in a standing hoof-trimming chute and all 4 feet were trimmed and examined. Lesions were treated, recorded, and graded according to severity. A list of 15 diseases was created to categorize the lesions. Cows were classified as having either the presence or absence of a painful lesion (PL). Pain was defined by reaction to gentle pressure applied to the lesions by hand. In the instance of a clearly PL, such as a sole ulcer with substantial exposure of corium tissue, the cow was spared from having pressure applied to the lesion but was classified as having a PL. All PL were scored for severity. Severity 1 was used for mild lesions and severity 2 was used to classify advanced lesions. Affected limb or limbs identified at trimming were recorded using the same categories as for the visual assessment of affected limbs.
Only the first treatment of each cow was used for analyses. Although data were collected after the first treatment, no data for that cow were used after the first treatment date.
Statistical Analysis.
For study 1, accuracy analyses (sensitivity, specificity, positive predictive value, and negative predictive value) were not performed, because the research sample was not representative of the farm population; animals assigned for hoof trimming were not randomly selected from the population.
An increasing or decreasing trend on the frequency of PL by VLS, and disease severity by VLS was evaluated using the Cochran-Armitage trend test using PROC FREQ of SAS (SAS Institute, Cary, NC).
The variable VLS was dichotomized twice, which generated 2 new variables, LVLS3 and LVLS4. For LVLS3, the cows classified as 1 or 2 at VLS were considered sound and were therefore classified as LVLS3 = 0; the cows classified from 3 to 5 at VLS were considered lame and were therefore classified as LVLS3 = 1. For LVLS4, the cows classified as 1 to 3 at VLS were considered sound and were therefore classified as LVLS4 = 0; the cows classified as 4 or 5 at VLS were considered lame and were therefore classified as LVLS4 = 1. To analyze the association between VLS and the presence or absence of a PL, a Spearman correlation coefficient was calculated using PROC FREQ of SAS, with the PLCORR option.
To facilitate analysis and interpretation of the association between affected limb at visual scoring and affected limb at treatment, the variables were dichotomized into front limb and hind limb lameness. The association between affected limb at visual scoring and affected limb at treatment was analyzed using the Spearman correlation coefficient.
Study 2
Farm and Management.
Data were collected from one dairy farm located in Lancaster County, Pennsylvania, from March 15 to 18, 2006. The farm milked 600 Holstein cows 3 times daily in a double-12 milking parlor. Lactating cows were fed a TMR. The cows were housed in free-stall barns with slatted floors. The concrete stalls were covered with mattresses, and the tops of the mattresses were bedded daily with wood shavings. Footbaths were located in the exit lanes of the milking parlor. Every cow was scheduled to receive routine hoof trimming at least twice yearly. Lame cows were identified and treated by trained farm employees and professional hoof trimmers. Moreover, cows enrolled in this study were not trimmed by the farm employees.
Design and Data Collection.
A prospective observational study design was used. All cows were scored using VLS once as they were exiting the milking parlor. Scoring was done by 3 veterinarians who were previously trained to perform VLS. The VLS system used was identical to that used in study 1. The veterinarians recorded the cow identification number, VLS, and suspected affected limb for VLS 3 into digital voice recorders while the cows were walking by. There were 7 categories of affected limb, as in study 1. The data were entered into a database (Microsoft Access) created for the study on the same day.
The Stepmetrix machine was located in a single-lane alley returning to the free-stall barn. The research cows walked through the Stepmetrix machine 3 times daily when returning from the milking parlor. A report, consisting of an SLS for each lactating cow, was generated by the Stepmetrix software and recorded on the same day the VLS were performed. The SLS was an average of all measurements captured for each cow for a period of 7 d. The SLS ranged from 1 to 100. The manufacturer’s suggested cutoff point for SLS was 39; cows with SLS of 39 were considered lame.
For the 2 d following the day of VLS, 8 professional hoof trimmers working as a group and using 4 professional hoof-trimming chutes performed maintenance hoof trimming and, whenever necessary, therapeutic trimming on all lactating cows. The same group of veterinarians that performed the VLS inspected the cows’ feet while the cows were being trimmed; diseases were recorded based on the same list of diseases previously described for study 1. Cows were classified as having either the presence or absence of a PL.
Statistical Analysis.
Three veterinarians performed VLS. Interobserver agreement for the VLS performed by the different veterinarians was assessed using the weighted kappa (K) statistic (Cohen’s kappa), calculated using MedCalc (www.medcalc.be). Furthermore, a receiver operating characteristic (ROC) analysis was performed for each veterinarian, assuming detection of PL at hoof trimming as the reference test. For the accuracy analysis, the mean of the available scores for the 3 veterinarians was calculated and rounded to a decimal point. If a cow had been scored by only 2 veterinarians, but not by the third veterinarian, the mean was calculated by adding the 2 scores and dividing by 2.
For study 2, sensitivity, specificity, positive predictive value, and negative predictive value analyses were conducted, assuming that the diagnosis of PL during hoof trimming was the gold standard method (reference test) for detection of painful digit lesions. This accuracy analysis was done using MedCalc. Differences between SLS and VLS in the accuracy of detecting PL were evaluated by ROC curves, by estimating differences of the area under the ROC curves and by using the statistical software package MedCalc.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Presence of a PL by VLS.
A total of 289 cows had PL at first trimming. An increasing pattern in the proportion of cows with PL was detected as the VLS increased (P < 0.001; Figure 1
). The proportions of cows with PL were 5.6, 20.1, 55.5, 79.9, and 100% for VLS 1 to 5, respectively.
|
3 as being lame and then considering VLS 4 as being lame, and a correlation analysis was done to measure the degree of association of the 2 dichotomized VLS and the presence or absence of a PL. When cows with VLS 3 were considered lame and the cows with VLS <3 were considered sound, the Spearman correlation coefficient was 0.48 ± 0.04. When cows were considered lame only if their VLS were 4 and cows with VLS <4 were considered sound, the Spearman coefficient was 0.41 ± 0.04 (Table 1).
|
).
|
Interobserver VLS Agreement.
Agreement beyond chance was performed among the observers that performed VLS, according to the kappa statistic analysis. The kappa coefficient for observers 1 and 2 was 0.46 ± 0.05 (n = 459). A complete summary of the kappa statistic for all comparisons is shown in Table 3.
|
). The area under the curve was 0.77 [confidence interval (CI) = 0.72 to 0.81], 0.74 (CI = 0.69 to 0.78), and 0.76 (CI = 0.71 to 0.80) for veterinarians 1, 2, and 3, respectively].
|
After excluding the cows with missing data, a total of 518 cows had complete data and were used for the final analysis. The overall prevalence of PL was 11.2% (n = 518). The most prevalent disease was sole ulcer (32.8%, n = 58), followed by white line disease (25.9%, n = 58), white line abscesses (13.8%, n = 58), digital dermatitis (8.7%, n = 58), and others (18.8%, n = 58). A strongly increasing pattern in the proportion of cows with PL was detected as VLS increased (P < 0.001). The proportions of cows with PL were 2.2, 5.1, 29.6, 61.9, and 100% for VLS of 1 to 5, respectively (Figure 1
).
Accuracy Analyses for VLS by a PL.
An accuracy analysis for VLS by PL was performed using the ROC curve. When cows classified as VLS 3, 4, and 5 were considered lame, the sensitivity of VLS to detect PL was 67.5% (95% CI, 54.6 to 78.1%). The specificity for this same cutoff was 84.6% (95% CI, 81.0 to 87.8%). A complete summary of the accuracy analysis for VLS is provided in Table 4.
|
.
|
).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The frequency of disease diagnosed in study 1 represents the disease that caused the first episode of lameness during the study period, whereas the frequency of disease in study 2 represents the prevalence of disease at the time of whole-herd trimming. In study 1, the cows that had VLS of 1 and 2 but were trimmed were the cows scheduled for routine hoof trimming. These cows would not have been trimmed for 6 mo or more and were therefore at a higher risk of hoof overgrowth and infectious claw diseases (Fjeldaas et al., 2006) than if they had been chosen at random within all the cows with VLS of 1 or 2. This might have caused a bias by increasing the proportion of PL detected in cows with VLS of 1 and 2.
There are many limitations to visual observations as a method for diagnosing lameness, such as the subjective nature of locomotion scoring systems (Neveux et al., 2006), the reliance on the skill of the observer to detect gait abnormalities (Flower et al., 2005), and the variation between and within observers (O’Callaghan et al., 2003). Despite these shortcomings, only 15.5% of cows with VLS of 1 or 2 in study 1 had a PL, 80.0% of cows with VLS of 4 or 5 had a PL, and 54.9% of cows with VLS of 3 had a PL. Visual locomotion score 3 represented 36.7% of all trimmings. Hoof trimming is time-consuming and expensive; therefore, the benefits of trimming cows with VLS of 3 needs to be studied further to determine if intervention in these cows is warranted.
In study 1, only one person was performing locomotion scoring; therefore, analyses of interobserver agreement were not possible. Cows in study 1 were visual locomotion scored every week, and some cows might not have had time for the lesions to develop into PL detectable by trimming. Cows were scored every week, and only the first trimming per cow was included in the study; therefore, the majority of cows trimmed were in the VLS 3 or 4 groups. Only 1.6% (n = 428) of cows detected lame by VLS had a score of 5. Septic arthritis, which can develop from complications of common digit diseases and can result in significant economic losses (Bicalho et al., 2006), was not diagnosed. This suggests that most lameness conditions take longer than a week to progress to this stage, and early intervention can reduce economic losses (Espejo et al., 2006).
Visual detection of lameness was more accurate for hind feet compared with front feet. Prevalence of lameness was higher in hind feet independently of the diagnosing technique. The increased accuracy of VLS might be due to an increased probability of finding problems on the hind legs. It should be noted that hoof overgrowth was not considered a PL and upper limb lameness could account for cows with visual gait abnormalities not having a PL on hoof trimming. As expected, a lower threshold for lameness in LVLS3 resulted in a higher association with a detected PL; the Spearman correlation coefficient for LVLS3 was 0.48 when compared with LVLS4, which had a Spearman correlation coefficient of 0.41.
For study 2, interobserver agreement among the 3 observers was moderate; the kappa coefficient ranged from 0.45 to 0.48. O’Callaghan et al. (2003), using percentage agreements, reported interobserver agreement on a 5-point VLS scale of 37%; however, the near agreement defined by the difference of one score was 81%. Engel et al. (2003), using a 9-point classification of gait assessment, found 47% agreement among 9 observers and an expert after training, which could be improved to 80%, if a difference of one class was allowed. Herein, the ROC analysis was performed for each veterinarian that performed VLS separately, and the areas under the curves did not differ significantly among the veterinarians.
In study 2, a ROC curve analysis for VLS and PL was performed, and the optimal sensitivity-specificity relationship was determined when a cutoff point of VLS 3 was used to detect a PL. An ROC curve analysis was performed for the SLS and a PL, and the optimal sensitivity-specificity relationship was determined when a cutoff point of 33 was used to detect a PL. The sensitivities to detect a PL were 67.5 and 33.3%, whereas the specificities were 84.6 and 89.5%, respectively, for VLS and SLS at the optimal cutoff point. The manufacturer’s recommended cutoff point for Stepmetrix was 39, and at this cutoff, the sensitivity was 22.2% and the specificity was 93.8%. When the ROC curve for VLS was compared with that for SLS to detect a PL, the area under the curve for VLS was significantly larger than that of SLS. Visual locomotion scoring was performed by trained veterinarians in this study and appears to perform better than SLS for detecting lameness. The application of ground reaction force measurements to detect lameness is promising, but may require further development to outperform VLS and become more attractive to dairy producers.
Received for publication February 5, 2007. Accepted for publication March 26, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. Chapinal, A. M. de Passille, D. M. Weary, M. A. G. von Keyserlingk, and J. Rushen Using gait score, walking speed, and lying behavior to detect hoof lesions in dairy cows J Dairy Sci, September 1, 2009; 92(9): 4365 - 4374. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. C. Bicalho, V. S. Machado, and L. S. Caixeta Lameness in dairy cattle: A debilitating disease or a disease of debilitated cattle? A cross-sectional study of lameness prevalence and thickness of the digital cushion. J Dairy Sci, July 1, 2009; 92(7): 3175 - 3184. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Liu, N. K. Neerchal, U. Tasch, R. M. Dyer, and P. G. Rajkondawar Enhancing the prediction accuracy of bovine lameness models through transformations of limb movement variables J Dairy Sci, June 1, 2009; 92(6): 2539 - 2550. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. J. Channon, A. M. Walker, T. Pfau, I. M. Sheldon, and A. M. Wilson Variability of Manson and Leaver locomotion scores assigned to dairy cows by different observers Vet Rec., March 28, 2009; 164(13): 388 - 392. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Chapinal, A. M. de Passille, and J. Rushen Weight distribution and gait in dairy cattle are affected by milking and late pregnancy J Dairy Sci, February 1, 2009; 92(2): 581 - 588. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. C. Bicalho, L. D. Warnick, and C. L. Guard Strategies to Analyze Milk Losses Caused by Diseases with Potential Incidence Throughout the Lactation: A Lameness Example J Dairy Sci, July 1, 2008; 91(7): 2653 - 2661. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. C. Bicalho, F. Vokey, H. N. Erb, and C. L. Guard Visual Locomotion Scoring in the First Seventy Days in Milk: Impact on Pregnancy and Survival J Dairy Sci, October 1, 2007; 90(10): 4586 - 4591. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
