Multiple advanced resting ECG (A-ECG) techniques have improved the diagnostic or prognostic value of ECG in detecting human cardiac diseases even before onset of clinical signs or changes in conventional ECG.
Multiple advanced resting ECG (A-ECG) techniques have improved the diagnostic or prognostic value of ECG in detecting human cardiac diseases even before onset of clinical signs or changes in conventional ECG.
To determine which A-ECG parameters, derived from 12-lead A-ECG recordings, change with severity of mitral regurgitation (MR) caused by myxomatous mitral valve disease (MMVD) in Cavalier King Charles Spaniels (CKCSs) in sinus rhythm.
Seventy-six privately owned CKCSs.
Dogs were prospectively divided into 5 groups according to the degree of MR (estimated by color Doppler mapping as the percentage of the left atrial area affected by the MR jet) and presence of clinical signs. High fidelity approximately 5-minute 12-lead ECG recordings were evaluated using custom software to calculate multiple conventional and A-ECG parameters.
Nineteen of 76 ECG parameters were significantly different (P < .05) across the 5 dog groups. A 4-parameter model that incorporated results from 1 parameter of heart rate variability, 2 parameters of QT variability, and 1 parameter of QRS amplitude was identified that explained 82.4% of the variance with a correlation coefficient (R) of 0.60 (P < .01). When age or murmur grade was included in the statistical model the prediction value further increased the R to 0.74 and 0.85 (P < .01), respectively.
In CKCSs with sinus rhythm, 4 selected A-ECG parameters further improve prediction of MR jet severity beyond age and murmur grade, although the predictive increment in this study probably is not sufficient to warrant utilization in clinical veterinary practice.
analysis of covariance
analysis of variance
congestive heart failure
Cavalier King Charles Spaniel
dipolar voltage of the signal-averaged QRS-waveform, calculated as the square-root of the sum of the 1st through 3rd squared singular values of the QRS, normalized with the square-root of the number of samples in QRS
heart rate variability
instantaneous and time independent component of change in QT, that is dependent only on the change in RR intervals in lead II
left ventricular end diastolic diameter
ratio between the left atrial (LA) and aortic root (Ao) diameters
myxomatous mitral valve disease
principal component analysis
QT interval variability
QT interval variability index
R-wave to R-wave variability
standard deviation of normal-to-normal RR intervals
Poincare-related first standard deviation
singular value decomposition
the total spectral power of RR interval variability by autoregression analysis
index of variability for unexplained part of QT interval obtained from lead II
vasovagal tonus index
Mitral regurgitation (MR) caused by myxomatous mitral valve disease (MMVD) is the most common cause of congestive heart failure and cardiac-related mortality in dogs. Standard ECG is of limited use in the diagnosis or management of canine MMVD other than to document and classify certain arrhythmias, such as premature atrial depolarizations and atrial fibrillation. Therefore, any improvements in the resting ECG that might assist with better identification of the severity of MR before onset of complications could be clinically useful.
Over the past 2 decades, several advanced techniques for high-resolution ECG (A-ECG) implemented using software have improved the diagnostic or predictive value of resting ECG in humans. When the results of these techniques are combined, they are more accurate than conventional ECG for identifying underlying coronary artery disease, myocardial ischemia, and dilated and hypertrophic cardiomyopathy, even before onset of clinical signs or changes in the standard ECG.[3-7]
A-ECG techniques include beat-to-beat QT variability (QTV)[8, 9] and R-wave to R-wave variability (RRV), “3-dimensional (3D)” (spatial and spatiotemporal) ECG,[11, 12] high-frequency (HF) QRS ECG, and detailed studies of waveform complexity by singular value decomposition.[13, 14]
Considering the promising results of A-ECG in humans, we investigated the utility of A-ECG in Cavalier King Charles Spaniels (CKCSs), a breed that suffers both earlier development and higher prevalence of MMVD than other breeds. We hypothesized that in CKCSs in sinus rhythm, several A-ECG parameters derived from 12-lead ECG recordings would change with the severity of MR related to MMVD.
In this cross-sectional study, 99 privately owned CKCSs were examined. All dogs underwent clinical examination, cardiac auscultation, echocardiography, and high fidelity approximately 5-minute 12-lead ECG. A-ECG was performed after approximately 5 minutes of acclimation, 20 minutes of echocardiography, and 5 minute of additional rest in a quiet environment at all study sites. None of the dogs were sedated for examination. All of the dogs were echocardiographically free of other cardiac diseases and with no clinical signs of systemic disease. Twenty-three CKCSs were excluded because of insufficient quality of ECG recording (n = 10), premature atrial depolarizations (n = 8), 2nd degree AV block (n = 3), 2nd degree sinoatrial block (n = 1), or euthanasia before completion of all study measures (n = 1). All dogs enrolled in the study as well as the reasons for exclusion are presented in the flow diagram (Fig 1).
Seventy-six CKCSs were categorized according to the degree of mitral regurgitation (MR) estimated as the percentage of the left atrial area affected by the MR jet using color Doppler mapping performed in the left apical 4-chamber view. Left ventricular end diastolic diameter (LVIDd) and the ratio between the left atrial (LA) and aortic root (Ao) diameters (LA/Ao) also were assessed from the echocardiographic recordings, as described previously. Echocardiographic examinations were recorded for later evaluation by observers who were blinded to the identity of the dog and results of ECG analysis.
Dogs were divided into 5 groups according to MR severity and presence of clinical signs: (i) 18 CKCSs with no MR and no clinical signs; (ii) 20 CKCSs with minimal MR (0 ≤ 15%) and no clinical signs; (iii) 13 CKCSs with mild MR (15% < MR ≤ 50%) and no clinical signs; (iv) 17 CKCSs with moderate-to-severe MR (>50%) and no clinical signs; and (v) 8 CKCSs in congestive heart failure (CHF) with moderate-to-severe MR (>50%) and clinical signs of CHF responsive to furosemide treatment. At the time of examination, all 8 CKCSs in CHF were undergoing medical treatment (Table 1).
|Group 1||Group 2||Group 3||Group 4||Group 5|
|MR (%)||0||0 ≤ 15||15 < MR ≤ 50||>50||>50|
|No. of dogs||18||20||13||17||8|
|Clinical signs of CHF||none||none||none||none||present†|
|Age (years)||3.6 ± 1.2#3,4,5||4.5 ± 1.4#3,4,5||6.7 ± 1.9#1,2,5||7.6 ± 1.9#1,2||9.5 ± 2.4#1,2,3|
|Body weight (kg)||8.8 ± 2.1||9.7 ± 2.0||9.1 ± 1.4||9.5 ± 2.0||10.8 ± 3.0|
|LVIDd (mm)||27.5 ± 3.6#5||29.7 ± 3.8#5||29.8 ± 4.7#5||33.2 ± 4.2#1||41.3 ± 5.0#1,2,3,4|
|LA/Ao (ratio)||1.4 ± 0.2#5||1.3 ± 0.2#5||1.5 ± 0.3#5||1.6 ± 0.2||1.8 ± 0.4#1,2,3|
|Therapy*||none||none||none||none||7 (F+A+P), 1(F+A+P+S)|
The study was performed between March 2009 and January 2010 at the Department of Basic Animal and Veterinary Sciences, Faculty of Life Sciences, University of Copenhagen, Denmark; at the Clinic for Surgery and Small Animal Medicine, Veterinary Faculty, University of Ljubljana, Slovenia; and at the small animal hospital Din Veterinar, Helsingborg, Sweden. The study was approved by the Slovenian Veterinary Administration, the Danish Animal Welfare Division and the Local Ethical Committee in Gothenburg, Sweden. All owners entered their dogs voluntarily and signed an informed consent form.
Dogs were positioned in right lateral recumbency. Short duration (approximately 5 minutes) 12-lead ECG was recorded as described previously. At all sites, a high fidelity (1,000 samples/s) computerized 12-lead ECG systema was used to acquire approximately 256 beat waveforms acceptable for both signal averaging and variability analyses. The recordings were stored on portable PC hard disk and analyzed using custom programs[9, 19] adapted for use in dogs.
Signals from conventional ECG were analyzed automatically using software for RR, QRS, and QTc (Fridericia-corrected) intervals, and waveform amplitudes.
Signal averaging was performed over the entire recording to acquire 256 beats and generate results for parameters of (i) 12-lead HF (150–250 Hz) QRS ECG; (ii) derived 3D ECG, using the regression-related Frank lead reconstruction of Kors et al to generate several vectorcardiographic parameters, including for example the spatial mean QRS-T angle, ventricular gradient and its components, and Z integral; and (iii) waveform complexity derived from singular value decomposition (SVD) plus signal averaging to derive dipolar (DPV) and nondipolar voltage (NDPV) equivalents[13, 14] of the P, QRS, and T waveforms, including for example dipolar voltage of the signal-averaged QRS-waveform, calculated as the square-root of the sum of the 1st through 3rd squared singular values of the QRS, normalized using the square root of the number of samples in QRS (DPV QRS).
Time series (256 beats) for the RR and QT intervals were analyzed. Specific variability analyses included: the standard deviation of normal-to-normal RR and QT intervals (SDNN RR and SDNN QT, respectively), the root mean square of the successive interval difference of normal-to-normal RR and QT intervals (RMSSD RR and RMSSD QT, respectively), Poincare-related 1st standard deviation (SD1), and, by autoregression analysis (AR), the very low (0.0–0.04 Hz), low (0.04–0.15 Hz), high (0.15–0.40 Hz) and total (TP) (0.0–0.40 Hz, TP AR) frequency powers of RRV. QT signals were analyzed using a dynamic model that considers the influence of RR intervals, QRS-T angles and QRS amplitudes on QTV, including calculation of the RR interval-corrected QTc interval using Fridericia's formula and the hysteresis properties of the QT interval. The unexplained part of QTVI for lead II (UnexQTVI (II)), the instantaneous and time independent component of the change in QT that is dependent only on the change in RR intervals in lead II (inst dQT/dRR (II)), and other parameters also were calculated as described previously.[9, 19, 23]
Statistical analyses were performed using commercially available software.,2,3 Data shown are mean ± SD, unless otherwise stated. One-way analysis of variance (ANOVA) was used to find differences among 5 groups of dogs for the ECG parameters studied. The normality of data and homoscedasticity assumptions were tested using Shapiro-Wilk, the Bartlett's Chi-Square, and Levene tests, respectively.
When ANOVA showed significance (P < .05), Bonferroni post-hoc tests also were performed. An analysis of covariance (ANCOVA) controlled by covariates sex and age was used to examine the influence of LVIDd and LA/Ao as dichotomous variables on the ECG parameters studied, with significance accepted at P < .05. LVIDd was considered normal when it had a normal value according to allometric scaling and LA/Ao was considered normal when <1.5.
Principal component factor analysis (PCA) was performed to decrease the number of relevant ECG parameters. Multiple linear regression (MLR) was performed to predict the severity of MR using relevant ECG parameters. ANCOVA again was used to test the influence of age, murmur grade, weight and sex on a MLR model incorporating 4 selected ECG parameters and a correlation model also was used to determine a percentage influence of age and murmur grade on each incorporated parameter individually.
Seventy-six ECG recordings of 76 CKCSs (5.9 ± 2.6 years; 29 males and 47 females) were studied.
At the 1st data reduction step, individual ECG parameters that had no significant discrimination value in the ANOVA (MR jet) and ANCOVA (LVIDd and LA/Ao) were eliminated from further consideration. After Bonferroni post-hoc tests and tests for the similarity of meaningful information for the remaining individual ECG parameters, 19 parameters remained for further statistical analysis (see Supporting Information).
Principal component factor analysis then was used to identify representative ECG parameters from the subset of 19. Eigenvalues of the correlation matrix were computed and then scree test criteria were applied to decide on the number of factors to be retained (Fig 2).
Parameters TP (AR), DPV QRS, Inst dQT/dRR (II), and UnexQTVI (II) were the best surrogate variables for 4 dimensional factor space, with 82.4% of variance explained (Table 2).
|Parameters||Factor 1||Factor 2||Factor 3||Factor 4|
|TP (AR) (ms2/Hz)||−0.91|
|DPV QRS (μV)||0.97|
|Inst dQT/dRR (II)||0.86|
|Explained variance (%)||29.9||23.8||15.4||13.3|
|Total explained variance variance (%)||82.4|
The predictive value of the 4 selected ECG parameters was examined using them as an independent variable in a MLR wherein MR jet was the dependent variable. The correlation coefficient (R = 0.60) indicated significant predictive power (P < .01).
ANCOVA for the 4 selected parameters showed a significant influence of both age and murmur grade on the overall model (P < .05). Although the individual parameters were not significantly influenced by age or murmur grade (Table 3), the inclusion of age or murmur grade as a 5th independent variable in the MLR analysis increased the R to 0.74 or to 0.85 (P < .01), respectively (Fig 3). Relative contributions of each variable in the prediction of MR jet for all 3 MLR models are shown in Table 4. Age (49%) and murmur grade (81%) made significant contributions (P < .05) in the prediction of MR jet when added as a 5th independent variable in the MLR analysis.
|ECG parameters||Covariates (P-value)|
|Inst dQT/dRR (II)||.110||.580||.895||.123|
|Variable||β Coefficients of MLR Model|
|4 A-ECG parameters||5 A-ECG parameters with age||5 A-ECG parameters with murmur grade|
|Inst dQT/dRR (II)||0.21*||0.14||0.05|
The influence of age (murmur grade) on the individual ECG parameters, calculated from correlation model was 12% (23%) for TP (AR), 11% for both DPV QRS (22%) and Inst dQT/dRR (II) (32%) and 7% (14%) for UnexQTVI (II).
Weight (P = .084) and sex (P = .070) did not significantly influence the overall model. However, sex had a significant influence on DPV QRS (P = .006), and weight had a significant influence on TP (AR) (P = .001) (Table 3).
MLR analysis showed that the 4 selected A-ECG parameters could predict the severity of MR with standard error of estimate ± 16% of the MR jet value. With inclusion of age or murmur grade the standard error of estimate was decreased to ± 13% or ± 10% of the MR jet value, respectively.
Our results suggest that several A-ECG parameters obtained from short-duration 12-lead resting high fidelity ECGs change with severity of MR caused by MMVD in CKCSs in sinus rhythm. Furthermore, improved detection of severity of MR relative to conventional ECG can be accomplished by the use of 4 A-ECG parameters: 1 parameter of heart rate variability (TP [AR]) that decreases significantly with MR severity, and 2 parameters of QT variability (Inst dQT/dRR [II] and UnexQTVI [II], respectively) and 1 parameter of QRS amplitude (DPV QRS) that increase significantly with MR severity. Detection of severity was furthermore improved with inclusion of age, a well-known risk factor for MMVD, and especially with inclusion of murmur grade.
Previous clinical studies of cardiac vagal tone have indicated that in conscious dogs with spontaneous heart disease, the cardiac index of parasympathetic activity (CIPA), an electronically generated measurement of cardiac vagal tone, is a sensitive, simple, and inexpensive measure of the severity of heart failure.[26, 27] In another study that investigated HRV in relation to severity of MR in CKCSs, both HRV, defined as the vasovagal tonus index (VVTI), and left atrial diameter were found to be most efficient in separating decompensated from compensated dogs.
The present study used a computerized ECG system in dogs that allowed multiple A-ECG techniques to be performed using software during a single digital 12-lead ECG recording.
Although several individual HRV parameters had utility for identifying the severity of MR in our study, the HRV parameter with the greatest utility was TP (AR). At the same time, only 1 conventional ECG parameter, the RR mean, changed (decreased) significantly with severity of MR, and all of the most useful A-ECG parameters with the exception of DPV QRS could be derived from lead II alone.
There is paucity of literature regarding QT and QTV-related parameters in dogs, possibly because of lack of agreement regarding formulas for rate correction of QT interval. Although the severity of MR in this study was not reflected by changes in the conventional QTc (Fridericia-corrected) interval, it was reflected by changes in 2 advanced parameters of QT interval variability (UnexQTVI [II] and Inst dQT/dRR[II], respectively) that characterize temporal (beat-to-beat) lability of ventricular repolarization.
In humans, age has a well-known effect on several ECG parameters. For example QRS and T wave amplitudes decrease with age and their axes also change directionally. Several HRV parameters also decrease with age in humans and 1 study of dogs revealed similar diminutions in HRV with age in clinically healthy Dachshunds. On the other hand, in a recent study of Holter monitoring in clinically healthy CKCSs, Wire-haired Dachshunds, and Cairn Terriers, HRV parameters were not influenced by age.
In our study, there was a significant influence of age on the overall ANCOVA model of 4 ECG parameters. However, age had only minor, nonsignificant influences on the individual ECG parameters and the model was able to accurately identify MR severity without inclusion of age. Moreover, DPV QRS, a global measure of QRS amplitude, increased in CKCSs with MR severity, whereas in humans QRS amplitude decreases with age. Differences in DPV QRS among dog groups therefore were much more likely related to disease than age.
Although the ECG equipment used in this study was “high fidelity”, the best-performing ECG parameters probably did not require such equipment and thus should also be derivable from many “standard fidelity” 12-lead ECG devices. Although 3 of the 4 most useful ECG parameters were derived from lead II alone, the 4th (DPV QRS) was derived with SVD which in our method required the presence of all 8 independent channels that constitute the 12-lead ECG. Therefore, for ideal results, 12-lead digital ECG must be recorded. From our experience, 12-lead ECG can be adequately recorded with use of ECG gel, self-adhesive disposable electrodes connected with alligator clips, and self-adherent wraps or tapes for additional fixation of electrodes.
Because 6-lead rather than 12-lead ECG currently is the standard of practice in veterinary medicine, it remains to be determined whether a DPV QRS result obtained from SVD of the fewer independent channels constituting a 6-lead recording could substitute for the 12-lead ECG-based DPV QRS parameter described here.
In conclusion, our results suggest that several A-ECG parameters change with severity of MR before changes in conventional ECG in CKCSs. Inclusion of age and murmur grade, when added to A-ECG, further improves the prediction of MR jet severity. However, although A-ECG itself further improves prediction of MR jet severity beyond age and murmur grade, its additive predictive value in this study probably is not enough to warrant use in clinical veterinary practice. More extensive clinical studies, including in other breeds, therefore are needed to better clarify the independent diagnostic utility of A-ECG in larger groups of dogs and to determine whether A-ECG parameters carry any prognostic information when MMVD is present.
This study had several limitations. First, the 3D-ECG parameters we studied rely on transformation coefficients that depend in part on the morphology of the human torso for deriving the Frank XYZ leads from the standard 12 leads. Important interspecies differences exist between dogs and humans regarding thorax morphology and only 1 study has described an exact placement of the precordial leads, adapted from human medicine, for 12-lead ECG in dogs. Moreover, the T wave may be concordant or discordant with the QRS in normal dogs whereas in healthy humans, the QRS complex and T wave usually are concordant. These facts probably complicate the analysis of 3D-ECG parameters in dogs when using software designed more specifically for humans and may help explain why none of the 3D ECG parameters were incorporated into our final model.
Second, comparison of the size of MR jet with the size of the left atrium is semiquantitative in dogs with MR. However, the severity of MR jet correlated well with several A-ECG parameters in this study. With evaluation of ECG parameters independently from blinded MR jet evaluations (ie, with blinded measurements of LVIDd and LA/Ao) we also showed that the changes in the 4 ECG parameters that had the greatest diagnostic utility correlated with changes in LVIDd and LA/Ao and with the severity of MR.
Third, this study was limited to CKCSs and results may not be applied to other breeds. It is known that VVTI, a measure of HRV, differs significantly among different breeds of dog, in particular being significantly greater in brachycephalic dogs than nonbrachycephalic dogs.
Because the CKCS is a brachycephalic breed, we speculate that differences in other ECG parameters also might exist. Additional studies are needed to identify such possible interbreed variations.
Fourth, measurements were not done in all dogs at the same time of the day, so we cannot exclude an influence of circadian rhythm on ECG parameters.[31, 35]
Fifth, CBC, clinical biochemistry, blood pressure measurements, and thoracic radiography only were performed in 7 dogs with CHF to determine whether the dogs had other systemic disease. None of these 7 dogs had clinically relevant renal disease.
Finally, although we handled the dogs gently, and examinations were performed under calm conditions, some of the dogs still may have experienced stress as suggested by their higher heart rates. Therefore, the state of mind of the dog also is important, and some of our results may have been influenced by stress reactions at the time of examination.
Research was supported by the Slovenian Research Agency (P3-0019) and the Slovene Human Resources Development and Scholarship Fund.
Cardiax, IMED, Budapest, Hungary
Statsoft Statistica 7.0, Statsoft, Tulsa, OK
Microsoft Office Proffesional 2003 Analysis Tools, Microsoft Corporation, Redmond, WA