SEARCH

SEARCH BY CITATION

Keywords:

  • horse;
  • treadmill exercise test;
  • electrocardiogram;
  • exercise-induced arrhythmias

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

Reasons for performing study: Frequent supraventricular or ventricular arrhythmias during and after exercise are considered pathological in horses. Prevalence of arrhythmias seen in apparently healthy horses is still a matter of debate and may depend on breed, athletic condition and exercise intensity.

Objectives: To determine intra- and interobserver agreement for detection of arrhythmias at rest, during and after exercise using a telemetric electrocardiography device.

Materials and methods: The electrocardiogram (ECG) recordings of 10 healthy Warmblood horses (5 of which had an intracardiac catheter in place) undergoing a standardised treadmill exercise test were analysed at rest (R), during warm-up (W), during exercise (E), as well as during 0–5 min (PE0–5) and 6–45 min (PE6–45) recovery after exercise. The number and time of occurrence of physiological and pathological ‘rhythm events’ were recorded. Events were classified according to origin and mode of conduction. The agreement of 3 independent, blinded observers with different experience in ECG reading was estimated considering time of occurrence and classification of events.

Results: For correct timing and classification, intraobserver agreement for observer 1 was 97% (R), 100% (W), 20% (E), 82% (PE0–5) and 100% (PE6–45). Interobserver agreement between observer 1 vs. observer 2 and between observer 1 vs. 3, respectively, was 96 and 92.6% (R), 83 and 31% (W), 0 and 13% (E), 23 and 18% (PE0–5), and 67 and 55% (PE6–45). When including the events with correct timing but disagreement for classification, the intraobserver agreement increased to 94% during PE0–5 and the interobserver agreement reached 83 and 50% (W), 20 and 50% (E), 41 and 47% (PE0–5), and 83.5 and 65% (PE6–45). The interobserver agreement increased with observer experience.

Conclusions: Intra- and interobserver agreement for recognition and classification of events was good at R, but poor during E and poor-moderate during recovery periods. These results highlight the limitations of stress ECG in horses and the need for high-quality recordings and adequate observer training.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

In human athletes, cardiac rhythm disturbances during exercise occur with a prevalence of 4.9%, with supraventricular premature depolarisations (SVPD; prevalence 1.6%) or ventricular premature depolarisations (VPD; prevalence 2.7%) being the most frequent encountered anomalies (Sofi et al. 2008). One of the main indications to perform stress electrocardiograms (ECG) in man is prevention of sudden cardiac death during exercise (Urhausen et al. 2007; Sofi et al. 2008). However, in the absence of structural heart disease and abnormal clinical findings, most arrhythmias are not considered relevant and do not require exclusion from competition (König et al. 2007; Sofi et al. 2008).

In apparently healthy Thoroughbred racehorses during peak exercise, prevalence of arrhythmias was recently reported to be 3% (Ryan et al. 2005). The aetiological relationship between cardiac arrhythmias and sudden death in racehorses is poorly characterised to date (Kiryu et al. 1987, 1999; Brown et al. 1988; Johnson et al. 1994; Boden et al. 2005). Nonetheless, detection of exercise induced cardiac arrhythmias is certainly relevant, particularly in horses with a history of exercise intolerance, poor performance or collapse during exercise. Frequent occurrences of supraventricular or ventricular arrhythmias during and after exercise are considered pathological in horses (Reef 1999; Martin et al. 2000) and it has been proposed that >2 isolated premature depolarisations during peak exercise, or multiple (i.e. >5) or pairs or paroxysms of premature depolarisations in the immediate recovery period are abnormal (Martin et al. 2000). However, these cut-offs have been questioned, especially in apparently healthy (Ryan et al. 2005) and poorly performing (Jose-Cunilleras et al. 2006) racehorses. Hence, the number of arrhythmic events occurring during and after exercise that can still be considered ‘normal’ remains a matter of debate and may depend on breed, athletic condition and exercise intensity.

Electrocardiography is the gold standard for detection and characterisation of arrhythmias in horses. For collection of ECG recordings during exercise, telemetric ECG devices or ambulatory Holter ECG recorders are used. Electrode placement and fixation greatly influence the quality of the recordings. Correct interpretation of ECG tracings further relies on the careful reading by an experienced equine clinician or cardiologist. However, evenwith strict adherence to recording guidelines (User's Guide for Televet-1001 and Reef 1999) and adequate observer training, ECG tracings obtained during exercise are subject to motion artefacts and are often difficult to interpret.

In human medicine it has been shown that a cardiologist reading the same ECG on separate occasions (intraobserver agreement) as well as different cardiologist reading the same ECG (interobserver agreement) may come up with different interpretations. Intra- and interobserver agreement may vary between moderate and very good (Holmvang et al. 1998; Massel 2003). Both agreement and accuracy in ECG reading depend on the ECG features identified (Pinkerton et al. 1981; Gillespie et al. 1996; Holmvang et al. 1998; Eslava et al. 2009). They are generally low in inexperienced observers and increase with practice (Pinkerton et al. 1981; Gillespie et al. 1996; Massel 2003; Eslava et al. 2009).

The purpose of this study was to determine the intra- and interobserver agreement for assessment of ECG recordings obtained in healthy Warmblood horses at rest, during treadmill exercise and immediately after exercise testing using a digital telemetric ECG device. We also aimed to determine to what extent observer experience influences interobserver agreement.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

Study population

The study population consisted of 2 groups of Warmblood sports horses. All horses were healthy based on history, physical examination (including lameness examination), haematology, serum biochemistry, resting ECG and routine echocardiography. Five horses were competing in 3-day events and another 5 horses were performing at an intermediate level in dressage or showjumping events. All horses were accustomed to the treadmill (Mustang 2200)2 over 2 days with at least 3 training sessions lasting 40 min. The horses of the second group were enrolled in a cardiovascular study, which included intracardiac pressure measurements. Therefore, they performed the standardised exercise test with an intracardiac catheter placed through the jugular vein, right atrium and right ventricle into the pulmonary artery. The intracardiac catheters were expected to provoke a certain number of premature ectopic depolarisations, thereby slightly increasing the number of abnormal findings to be analysed by the observers.

ECG recordings

The ECG was recorded with a digital telemetric ECG device (Televet-100)1. Surface electrodes (modified Polar Equine T51H/T52H)3 were placed without clipping at the following locations: ventrally in the median, approximately 5 cm caudal to the olecranon (i.e. at the level of girth placement, green electrode); left (red electrode) and right (yellow electrode) on the thorax, about 30 cm below the withers in the sixth or seventh intercostal space; and right on the thorax, about 50 cm below the withers in the sixth or seventh intercostal space (black grounding electrode). The electrodes were tightly fixed with a lungeing girth, following manufacturer's instructions (User's Guide for Televet-100)1. ECGs were recorded and digitally stored onto a laptop computer prior to the exercise test at rest (R), during the warm-up period (W), during the exercise test (E) and during the post exercise recovery period (PE). The ECG tracings were subsequently analysed using the ECG analysis software of the telemetry system.1

Treadmill exercise test

All horses performed a submaximal incremental exercise test on a 6% inclined treadmill. Before starting the exercise test, the horses were warmed-up for approximately 30 min at all 3 gaits. The exercise test consisted of one or 2 steps at trot (3.5–4.0 m/s) and 4 or 5 steps at canter and gallop (6.0–10.0 m/s). Speed increments were 1 m/s. The first step lasted 2 min, all subsequent steps 90 s. The horses were not run to exhaustion; instead, the exercise test was terminated when the horses reached a speed at which the blood lactate concentration was >4 mmol/l. The time from peak exercise to complete stop of the treadmill was 30 s. Then the treadmill was stopped for 5 min of passive recovery (post exercise period PE0–5) before the horses walked for active cool-down for at least 15 min and maximally 1 h (post exercise period PE6–45).

ECG analysis and agreement

Three veterinarians at different levels of their postgraduate training analysed the 10 ECG recordings. The level of experience in ECG reading increased from Observer 3 to Observer 1. Observer 1 (DST) was a veterinarian specialised in equine medicine with particular interest in equine cardiology and 9 years' experience in ECG reading. Observer 2 (NW) was a veterinarian with 3 years' experience in equine medicine, but no specific training in cardiovascular medicine. Observer 3 (CB) was a veterinarian in the second year after graduation from veterinary school. All 3 observers underwent a short training session to agree on the specific definitions of the different arrhythmias to be detected. All observers were independent and blinded to the results of the ECG readings of the other observers.

Assessment of ECGs for recognition and classification of cardiac arrhythmias was performed on ECG tracings showing simultaneous derivation from 3 bipolar leads (I: yellow to red electrode, II: red to green electrode, III: yellow to green electrode). Thereby, QRS complexes were automotically detected and R-R intervals were measured by the ECG analysis software1. R-R intervals deviating >5% from the preceding R-R interval where highlighted automatically, but the software did not further characterise the highlighted R-R intervals and associated QRS complexes. Detection of 5% deviation was chosen in order to detect R-R inconsistencies at high HR. The marked R-R intervals were defined as ‘rhythm events’, unless they were identified as artefacts by the observers. Each observer noted the time when rhythm events occurred and classified the events in one of the following categories: sino-atrial blocks (SAB, RR-interval lasting approximately twice as long as the preceding RR-intervals with no visible P wave during the interval), second degree atrio-ventricular blocks (AVB), supraventricular premature depolarisations (SVPD; including single depolarisations, runs of 4 or more supraventricular premature depolarisations or supraventricular tachycardia [SVT] lasting >30 s), atrial fibrillation (AF), ventricular premature depolarisations (VPD; including single depolarisations, couplets, triplets, or runs of 4 or more ventricular premature depolarisations or sustained ventricular tachycardia [VT] lasting >30 s). If the rhythm event could not be classified in one of these categories, it was reported as ‘not classified’. For AVBs, only the time of first occurrence and the number of following blocks were recorded. Occurrence of sinus arrhythmias (SA) was not reported unless the observer felt that it was severe. The differentiation between SA and SVPD was based upon the pattern of occurrence of the premature beats and shape of the P waves. Specifically, suddenly occurring, obviously premature beats, occasionally even superimposed on the preceding T waves and associated with P waves of markedly altered shape, were considered SVPDs. Conversely, mild to marked variation or waxing and waning of P-P intervals occurring over a prolonged period of time, associated with no or only slight alterations in the shape of the P waves, was considered SA. Occurrence of artefacts was not reported.

Data analysis

Observer agreement: Observer 1 reviewed the results of all readings performed by the 3 observers to determine the type of agreement/disagreement, calculate observer agreement and determine which categories of rhythm events caused the highest degree of disagreement.

Intra- and interobserver agreement, respectively, were estimated as percentage of full agreement in assessment of rhythm events (= number of events with agreement in time of occurrence and classification/total number of events counted by the respective observers × 100), classification disagreement (= number of events with agreement in time of occurrence but disagreement in regards to classification/total of number of events counted by the respective observers) and time disagreement (=number of events that were not recognised on one occasion or by one of the observers/total number of events counted by the respective observer × 100).

The rhythm events were then grouped into physiological events (i.e. SAB, SA if noted, AVB) or pathological events (i.e. SVPD, SVT, VPD, VT) and the agreement calculated for the detection of ‘pathological rhythm events’ (without further specification) in the same manner as described above.

To determine the categories of rhythm events that caused the highest degree of disagreement, the number of occurrences of specific disagreement (e.g. disagreement between SA and SVP, or time disagreement for a AVB) among all observers was counted, divided by total number of occurrences of disagreement noted for all observers during the observation period, and multiplied by 100 to obtain % values.

Descriptive statistics: For descriptive purposes, data are listed as raw data or expressed as medians and ranges.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

The duration of the different observation periods analysed was 40 (5–82) min (median [range]) for R, 50 (39–73) min for W, 10 (9–15) min for E, 5 min for PE0–5, and 44 (15–58) min for PE6–45.

The number of physiological and pathological rhythm events noted by the 3 observers is shown in Table 1. The number of detected events varied greatly between the observation periods and was highest during R; in this period the majority of events were physiological AVBs. No AF, SVT and VT were detected in any of the horses.

Table 1. Number of physiological and pathological rhythm events observed by 3 independent, blinded observers on electrocardiograms obtained from 10 healthy, athletic Warmblood horses at different time points before, during and after a standardised treadmill exercise test. Note that 5 of the horses underwent an invasive cardiovascular study and therefore had an intracardiac catheter in place. Therefore, these data do not reflect the true prevalence of arrhythmias in a healthy Warmblood horse population
 RestWarming-upExercise test0–5 min after the exercise test6–45 min after the exercise test
  1. SAB: sino-atrial block; AVB: second degree atrio-ventricular block; SVPD: supraventricular premature depolarisation; VPD: ventricular premature depolarisation.

Total number of rhythm events     
 Observer 1 (1st reading)2961141724
 Observer 1 (2nd reading)2951121524
 Observer 2301112619
 Observer 3310135724
Physiological rhythm events     
 SAB     
  Observer 1 (1st reading)00010
  Observer 1 (2nd reading)00000
  Observer 250000
  Observer 310000
 AVB     
  Observer 1 (1st reading)29140014
  Observer 1 (2nd reading)28640014
  Observer 228640013
  Observer 328640014
Pathological rhythm events     
 SVPD     
  Observer 1 (1st reading)20010
  Observer 1 (2nd reading)50000
  Observer 270203
  Observer 3193440
 VPD     
  Observer 1 (1st reading)37267
  Observer 1 (2nd reading)47277
  Observer 236053
  Observer 321121
 Not classified     
  Observer 1 (1st reading)00293
  Observer 1 (2nd reading)00083
  Observer 201020
  Observer 325019

Table 2 summarises the results for observer agreement for detection of all (i.e. physiological and pathological) rhythm events. At rest, intraobserver full agreement was 97% and interobserver full agreement was 96 and 92.6%, for observer 1 vs. observer 2 and for observer 1 vs. observer 3, respectively. With exercise, full agreement decreased markedly reaching values as low as 20% for intraobserver agreement and 0 and 13% for interobserver agreement. During the recovery period, full agreement improved to 82% (PE0–5) and 100% (PE6–45) for intraobserver comparisons, to 23 and 18% (PE0–5) and to 67 and 55% (PE6–45), for interobserver comparisons. When combining full agreement and classification disagreement (i.e. recognised but differently classified events), the intraobserver agreement did not change substantially at R, E and PE6–45, but increased to 94% during PE0–5. Similarly, the interobserver agreement increased to 99 and 93% (R), to 20 and 50% (E), to 41 and 47% (PE0–5), and to 83.5 and 65% (PE6–45), for observer 1 vs. observer 2 and for observer 1 vs. observer 3, respectively.

Table 2. Observer agreement expressed as percent of agreement for 3 independent, blinded observers including diagnosis of all (i.e., physiological and pathological) rhythm events on electrocardiograms obtained from 10 healthy, athletic Warmblood horses. Note that the level of experience increased from Observer 3 to Observer 1
 RestWarmingupExercise Test0–5 min after the exercise test6–45 min after the exercise test
  1. Full agreement: agreement in time of occurrence and classification of a rhythm event. Classification disagreement: agreement in time of occurrence but disagreement in regards to classification of a rhythm event. Time disagreement: lack of recognition of a rhythm event on one occasion or by one of the observers.

Intraobserver agreement     
 Total number of rhythm event counted by the respective observers (Observer 1 first vs. second reading)3001151724
 Full agreement291 (97%)11 (100%)1 (20%)14 (82%)24 (100%)
 Classification disagreement0 (0%)0 (0%)0 (0%)2 (12%)0 (0%)
 Time disagreement9 (3%)0 (0%)4 (80%)1 (6%)0 (0%)
Interobserver agreement     
 Total number of rhythm events counted by the respective observers (Observer 1 vs. Observer 2)3011251724
 Full agreement290 (96%)10 (83%)0 (0%)4 (23%)16 (67%)
 Classification disagreement8 (3%)0 (0%)1 (20%)3 (18%)4 (16.5%)
 Time disagreement3 (1%)2 (17%)4 (80%)10 (59%)4 (16.5%)
 Total number of rhythm events counted by the respective observers (Observer 1 vs. Observer 3)3131681729
 Full agreement290 (92.6%)5 (31%)1 (13%)3 (18%)16 (55%)
 Classification disagreement1 (0.4%)3 (19%)3 (37%)5 (29%)3 (10%)
 Time disagreement22 (7%)8 (50%)4 (50%)9 (53%)10 (35%)

Table 3 summarises the results for the observer agreement for detection of pathological rhythm events. At rest, intraobserver full agreement was 56% and interobserver full agreement was 50 and 18%, respectively. Agreement was lowest during exercise testing and increased again during the recovery period. When combining full agreement and classification disagreement, the intraobserver agreement again increased in PE0–5 to 94%, while interobserver agreement increased to 80 and 18% (R), to 33 and 56% (E), to 54 and 44% (PE0–5), and to 100 and 33% (PE6–45).

Table 3. Observer agreement expressed as percent of agreement for 3 independent, blinded observers including diagnosis of pathological rhythm events on electrocardiograms obtained from 10 healthy, athletic Warmblood horses. Note that the level of experience increased from Observer 3 to Observer 1
 RestWarmingupExercise test0–5 min after the exercise test6–45 min after the exercise test
  1. Full agreement: agreement in time of occurrence and classification of a rhythm event. Classification disagreement: agreement in time of occurrence but disagreement in regards to classification of a rhythm event. Time disagreement: lack of recognition of a rhythm event on one occasion or by one of the observers.

Intraobserver agreement     
 Total number of rhythm event counted by the respective observers (Observer 1 first vs. second reading)9751610
 Full agreement5 (56%)7 (100%)1 (20%)14 (88%)10 (100%)
 Classification disagreement0 (0%)0 (0%)0 (0%)1 (6%)0 (0%)
 Time disagreement4 (44%)0 (0%)4 (80%)1 (6%)0 (0%)
Interobserver agreement     
 Total number of rhythm events counted by the respective observers (Observer 1 vs. Observer 2)1073117
 Full agreement5 (50%)6 (86%)0 (0%)4 (36%)3 (43%)
 Classification disagreement3 (30%)0 (0%)1 (33%)2 (18%)4 (57%)
 Time disagreement2 (20%)1 (14%)2 (67%)5 (46%)0 (0%)
 Total number of rhythm events counted by the respective observers (Observer 1 vs. Observer 3)221281615
 Full agreement4 (18%)1 (8%)1 (13%)3 (19%)2 (13%)
 Classification disagreement0 (0%)3 (25%)3 (37%)4 (25%)3 (20%)
 Time disagreement18 (82%)8 (67%)4 (50%)9 (56%)10 (67%)

Generally, intraobserver agreement was higher than interobserver agreement and interobserver agreement increased with observer experience.

Time disagreement (i.e. rhythm event only detected by one observer or on one occasion during repeated assessment by Observer 1) was the most common problem at all observed periods and accounted for 34/43 (79%) of disagreement during R, for 10/13 (76%) during W, for 12/16 (75%) during E, for 20/30 (66%) during PE0–5 and for 14/21 (66%) during PE6–45. At R, the main problem causing time disagreement was the high number of AVB that had to be counted. For classification disagreement, the most frequent cause was an uncertainty to distinguish a pronounced SA from SVPD (5/9, 55%; Figs 1a–c), and a SAB from an AVB (3/9, 33%). With motion (i.e. during W, E and PE6–45), the main difficulty leading to classification disagreement consisted in distinguishing between SVPD and VPD (8/14, 57%; Fig 2a) and in differentiating rhythm events from artefacts (5/14, 36%; Fig 2b). In the early post exercise period (PE0–5), when the horse was standing but the heart rate was still high, distinction between SVPD and VPD was the most common problem (7/10, 70%; Figs 3a,c).

image

Figure 1. ECG tracings with example of rhythm events occurring at rest (R). The vertical marks indicate 1 s intervals, tracings show Lead II (Televet-1001). a) Example of SA. Note the irregularity in the R-R intervals but normal-shaped P wave (in comparison to the preceding P waves), QRS complex and T wave; b) Example of rhythm event leading to classification disagreement (SA vs. SVPD). Note the pronounced irregularity in the R-R intervals and the slightly variable shape of the P wave in the presence of normal QRS complexes and T waves; c) Example of 2 succeeding SVPDs. Note the pronounced irregularity in the R-R intervals associated with altered shape of the P waves in comparison to preceding P wave (arrowheads) but normal-shaped QRS complex and T wave.

Download figure to PowerPoint

image

Figure 2. ECG tracings with examples of rhythm events occurring during the exercise test (E). The vertical marks indicate 1 s intervals, the tracings show Leads I, II and III (Televet-1001). a) Example of a rhythm event (arrowheads) occurring while the horse was galloping. This event led to classification disagreement (VPD vs. SVPD). Note the irregular R-R interval, altered shape of QRS complex and T wave, and the pause after the complex seen in Leads II and III. Also note that P waves are difficult to identify at high heart rates during exercise; b) Example of rhythm event occurring while the horse was galloping. This event led to disagreement regarding classification as a rhythm event. Note the absence of a distinguishable premature QRS complex and the regularly occurring T waves (arrowheads) seen in Leads II and III.

Download figure to PowerPoint

image

Figure 3. ECG tracings with examples of rhythm events occurring immediately after the exercise test (PE0–5). The vertical marks indicate 1 s intervals, the tracings show Leads I, II and III (Televet-1001). a) Example of a rhythm event occurring during the first 60 s after the end of the exercise test and leading to classification disagreement (SVPD vs. VPD). Note the irregular R-R interval and the normal shaped QRS complex seen in Leads II and III. A P wave, superimposed to the preceding T wave, is barely distinguishable preceding the QRS complex (arrowheads); b) Example of a couplet of VPDs occurring in the first 30 s after the end of the exercise test. Note the irregular R-R intervals, absence of a distinguishable P wave, the altered shape of QRS complex and T wave and the pause after the premature complexes; c) Example of rhythm event occurring in the first 120 s after the end of the exercise test, leading to classification disagreement (SA vs. SVPD vs. VPD). Note the irregular R-R interval, the absence of a distinguishable P wave and the altered shape of QRS complex and T wave.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

The number of rhythm events in this study varied greatly between observation periods. The intra- and interobserver agreement for recognition and correct classification of arrhythmias was good at rest, but markedly lower during and immediately after exercise. Interobserver agreement generally increased with observer experience.

Studies in human medicine showed that the accuracy for recognising normal ECGs at rest varies between 92.7 and 97.6% (Willems et al. 1991; Gillespie et al. 1996), but is markedly lower and ranges from 72.6–81% (Willems et al. 1991) and from 29–97% (Gillespie et al. 1996) for correctly classifying cardiovascular diseases based on ECG. Furthermore, observer agreement also depends on the ECG feature to be recognised (Holmvang et al. 1998). Because of the differences in indications for ECG examinations, study design (i.e. number of observers and number of features to be interpreted), recording conditions and assessment of ECGs (i.e. diagnosing specific disease vs. characterising specific rhythm events), direct comparison of these results with those of the present study should be made with caution. In the present study, we calculated agreement for recognising and classifying ‘rhythm events’, but not accuracy of the observer for reaching a correct diagnosis.

As expected, and in concordance with human studies (Weston et al. 1976; Holmvang et al. 1998; Massel 2003), intraobserver agreement was greater than interobserver agreement. Furthermore, interobserver agreement seemed to be influenced by the level of observer experience. This finding is also concordant to findings in human studies, in which poor agreement (Massel 2003) or poor accuracy in ECG reading (Pinkerton et al. 1981; Gillespie et al. 1996; Eslava et al. 2009) was related to lack of experience. These results underline the importance of adequate observer training and experience in ECG reading. Obtaining a second opinion from a more experienced observer may allow correct classification of ambiguous ECG findings. This has also been shown in a human study in which the accuracy of ECG diagnoses could be increased from 67 to 91% by having the ECG reviewed by a cardiologist (Woolley et al. 1992).

A substantial number of disagreement occurred by miscounting events when a high number of events were present. This type of error could possibly be reduced by automated ECG analysis. However, computerised ECG interpretation programs have a high rate of misclassification, especially in nonsinus rhythm events (Shah and Rubin 2007). The high number of physiological AVB and high amplitude and variable morphology of the T wave make the use of such fully automated algorithms very difficult in horses. Automated R-R interval analysis, as conducted in this study, may help identify inconsistencies in R-R intervals, but still requires considerable observer input and does not aid in classifying abnormal rhythm events.

Distinction between pronounced SA and SVPD (Figs 1a–c) may be difficult and in the present study led to a great number of classification disagreements. Applying different lead positions might have facilitated recognition of abnormal P waves or QRS complexes and could have provided further information to distinguish between ambiguous findings. However, in this study we used the standard lead position as recommended by the manufacturer of the telemetric ECG device. Investigating different lead positions was beyond the scope of this study. Agreement was generally poor during the exercise period, probably because of the high numbers of motion artefacts during this observation period impairing the ability to correctly classify events. Also, at high heart rates the P wave is very close to, or superimposed on, the T wave, making the differentiation between SVPD and VPD difficult (Fig 3a), unless the QRS morphology is markedly altered (Figs 3b,c). With motion, the percentage of full agreement generally dropped more in observers with less experience.

Some limitations of this study need to be discussed. The study population was relatively small and consisted of healthy horses in athletic condition. Therefore, the number of pathological rhythm events was generally low. This may certainly have influenced the results, and it may be argued that observer agreement may have improved with a greater number of pathological rhythm events in a population of horses with heart disease. Furthermore, the degree of observer agreement reported in this study might be overestimated as our data do not account for agreement by chance. However, assessment of observer agreement by a kappa statistic (Dohoo et al. 2010) addressing this point was not applicable to our data because of the fact that during certain observation periods, rhythm events occurred infrequently or were absent, leading to erroneous values or large confidence intervals for kappa values. Finally, because of the low number of horses and because one group of horses had intracardiac catheters in place, no conclusions can be made regarding the prevalence of pathological arrhythmias in healthy, athletic Warmblood horses at rest and during and after exercise.

In conclusion, our results show that intra- and interobserver agreement for recognition and correct classification of cardiac arrhythmias is generally good at rest but markedly lower during and immediately after exercise. The results further highlight the need for high-quality recordings and adequate observer training, particularly for assessment of ECGs recorded during exercise. Studies including a larger number of healthy horses and horses with cardiac disease will be necessary to determine the prevalence of pathological arrhythmias in athletic Warmblood horses at rest, during exercise and during recovery, to define cut-off values for the number of rhythm events that should be considered pathological and to define more precisely the reliability of diagnosing pathological arrhythmias by resting ECG and stress ECG.

Acknowledgement

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

We would like to thank Dr S. Hartnack for her advice concerning the statistical analyses.

Conflicts of interest

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

The study has been supported by a grant from the Research Fund of the University of Zurich and the Foundation ‘Forschung für das Pferd’.

Manufacturers' addresses

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References

1 Roesch&Associates Information Engineering GmbH, Frankfurt, Germany.

2 Graber AG, Fahrwangen, Switzerland.

3 Polar Electr. Oy, Kempele, Finland.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. Conflicts of interest
  9. Manufacturers' addresses
  10. References
  • Boden, L.A., Charles, J.A., Slocombe, R.F., Sandy, J.R., Finnin, P.J., Morton, J.M. and Clarke, A.F. (2005) Sudden death in racing Thoroughbreds in Victoria, Australia. Equine vet. J. 37, 269-271.
  • Brown, C.M., Kaneene, J.B. and Taylor, R.F. (1988) Sudden and unexpected death in horses and ponies: an analysis of 200 cases. Equine vet. J. 20, 99-103.
  • Dohoo, I., Wayne, M. and Stryhn, H. (2010) Screening and diagnostic tests. In: Veterinary Epidemiologic Research, 2nd edn., VER Inc, Charlottetown. pp 97-98.
  • Eslava, D., Dhillon, S., Berger, J., Homel, P. and Bergmann, S. (2009) Interpretation of electrocardiograms by first-year residents: The need for change. J. Electrocardiol. 42, 693-697.
  • Gillespie, N.D., Brett, C.T., Morrison, W.G. and Pringle, S.D. (1996) Interpretation of the emergency electrocardiogram by junior hospital doctors. J. Accid. Emerg. Med. 13, 395-397.
  • Holmvang, L., Hasbak, P., Clemmensen, P., Wagner, G. and Grande, P. (1998) Differences between local investigator and core laboratory interpretation of the admission electrocardiogram in patients with unstable angina pectoris or non-Q-wave myocardial infarction (a Thrombin Inhibition in Myocardial Ischemia [TRIM] substudy). Am. J. Cardiol. 82, 54-60.
  • Johnson, B.J., Stover, S.M., Daft, B.M., Kinde, H., Read, D.H., Barr, B.C., Anderson, M., Moore, J., Woods, L., Stoltz, J. and Blanchard, P. (1994) Causes of death in racehorses over a 2 year period. Equine vet. J. 26, 327-330.
  • Jose-Cunilleras, E., Young, L.E., Newton, J.R. and Marlin, D.J. (2006) Cardiac arrhythmias during and after treadmill exercise in poorly performing Thoroughbred racehorses. Equine vet. J., Suppl. 36, 163-170.
  • Kiryu, K., Machida, N., Kashida, Y., Yoshihara, T., Amada, A. and Yamamoto, T. (1999) Pathologic and electrocardiographic findings in sudden cardiac death in racehorses. J. vet. med. Sci. 61, 921-928.
  • Kiryu, K., Nakamura, T., Kaneko, M., Oikawa, M. and Yoshihara, T. (1987) Cardiopathology of sudden cardiac death in the race horse. Heart Vessels Suppl. 2, 40-46.
  • König, D., Zehender, M., Mewis, C. and Dickhuth, H.-H. (2007) Herzrhythmusstörungen. In: Sportkardiologie: Körperliche Aktivität b Herzerkrankungen, 2. Auflage Edn., Eds: W.Kindermann, H.-H.Dickhuth, A.Niess, K.Röcker and A.Urhausen, Steinkopff Verlag, Darmstadt. pp 241-256.
  • Martin, B.B., Jr, Reef, V.B., Parente, E.J. and Sage, A.D. (2000) Causes of poor performance of horses during training, racing, or showing: 348 cases (1992-1996). J. Am. vet. med. Ass. 216, 554-558.
  • Massel, D. (2003) Observer variability in ECG interpretation for thrombolysis eligibility: experience and context matter. J. Thromb. Thrombolysis 15, 131-140.
  • Pinkerton, R.E., Francis, C.K., Ljungquist, K.A. and Howe, G.W. (1981) Electrocardiographic training in primary care residency programs. J. Am. med. Ass. 246, 148-150.
  • Reef, V.B. (1999) Ambulatory and exercise electrocardiography and post-exercise echocardiography. In: Cardiology of the Horse, Ed: C.M.Marr, W.B. Saunders, London. pp 150-160.
  • Ryan, N., Marr, C.M. and McGladdery, A.J. (2005) Survey of cardiac arrhythmias during submaximal and maximal exercise in Thoroughbred racehorses. Equine vet. J. 37, 265-268.
  • Shah, A.P. and Rubin, S.A. (2007) Errors in the computerized electrocardiogram interpretation of cardiac rhythm. J. Electrocardiol. 40, 385-390.
  • Sofi, F., Capalbo, A., Pucci, N., Giuliattini, J., Condino, F., Alessandri, F., Abbate, R., Gensini, G.F. and Califano, S. (2008) Cardiovascular evaluation, including resting and exercise electrocardiography, before participation in competitive sports: Cross sectional study. Br. Med. J. 337, a346.
  • Urhausen, A., Scharhag, J. and Kindermann, W. (2007) Plötzlicher Herztod beim Sport und kardiovaskuläres Screening. In: Sportkardiologie: Körperliche Aktivität b Herzerkrankungen, 2. Auflage, Eds: W.Kindermann, H.-H.Dickhuth, A.Niess, K.Röcker and A.Urhausen, Steinkopff Verlag, Darmstadt. pp 21-38.
  • Weston, M.J., Bett, J.H. and Over, R. (1976) Consensus opinion and observer accuracy in electrocardiography with reference to coronary arteriographic information. Aust. N. Z. J. Med. 6, 429-432.
  • Willems, J.L., Abreu-Lima, C., Arnaud, P., Van Bemmel, J.H., Brohet, C., Degani, R., Denis, B., Gehring, J., Graham, I., Van Herpen, G., Machado, H., Macfarlane, P.W., Michaelis, J., Moulopoulos, S.D., Rubel, P. and Zywietz, C. (1991) The diagnostic performance of computer programs for the interpretation of electrocardiograms. N. Engl. J. Med. 325, 1767-1773.
  • Woolley, D., Henck, M. and Luck, J. (1992) Comparison of electrocardiogram interpretations by family physicians, a computer, and a cardiology service. J. Fam. Pract. 34, 428-432.