• Open Access

Detection of Congestive Heart Failure in Dogs by Doppler Echocardiography


  • Presented in part at the Annual Forum of the American College of Veterinary Internal Medicine, Montreal, Canada, June 3–6, 2009. Dr Hart is presently affiliated with University of Minnesota Veterinary Medical Center, 1365 Gortner Avenue, St Paul, MN 55108. Dr Stern is presently affiliated with Department of Veterinary Clinical Sciences, Washington State University, 100 Grimes Way, Pullman, WA 99164. This work was completed at The Ohio State University, Columbus, OH.

Corresponding author: Karsten E. Schober, DVM, PhD, Department of Veterinary Clinical Sciences, College of Veterinary Medicine, The Ohio State University, 601 Vernon L. Tharp Street, Columbus, OH 43210; e-mail: schober.4@osu.edu.


Background: Echocardiographic prediction of congestive heart failure (CHF) in dogs has not been prospectively evaluated.

Hypothesis: CHF can be predicted by Doppler echocardiographic (DE) variables of left ventricular (LV) filling in dogs with degenerative mitral valve disease (MVD) and dilated cardiomyopathy (DCM).

Animals: Sixty-three client-owned dogs.

Methods: Prospective clinical cohort study. Physical examination, thoracic radiography, analysis of natriuretic peptides, and transthoracic echocardiography were performed. Diagnosis of CHF was based upon clinical and radiographic findings. Presence or absence of CHF was predicted using receiver-operating characteristic (ROC) curve, multivariate logistic and stepwise regression, and best subsets analyses.

Results: Presence of CHF secondary to MVD or DCM could best be predicted by E : isovolumic relaxation time (IVRT) (area under the ROC curve [AUC]=0.97, P < .001), respiration rate (AUC=0.94, P < .001), Diastolic Functional Class (AUC=0.93, P < .001), and a combination of Diastolic Functional Class, IVRT, and respiration rate (R2=0.80, P < .001) or Diastolic Functional Class (AUC=1.00, P < .001), respiration rate (AUC=1.00, P < .001), and E : IVRT (AUC=0.99, P < .001), and a combination of Diastolic Functional Class and E : IVRT (R2=0.94, P < .001), respectively, whereas other variables including N-terminal pro-brain natriuretic peptide, E : Ea, and E : Vp were less useful.

Conclusion and Clinical Importance: Various DE variables can be used to predict CHF in dogs with MVD and DCM. Determination of the clinical benefit of such variables in initiating, modulating, and assessing success of treatments for CHF needs further study.


duration of the late diastolic transmitral flow wave


aortic annular dimension


duration of the late diastolic pulmonary vein atrial reversal flow wave


area under the ROC curve


congestive heart failure


coefficient of variation


dilated cardiomyopathy


Doppler echocardiography


deceleration time of the early diastolic transmitral flow


fractional area change


isovolumic relaxation time


left atrial


maximum left atrial area


minimum left atrial area


maximum left atrial dimension


minimum left atrial dimension




left ventricular


left ventricular internal dimension in diastole


left ventricular internal dimension in systole


left ventricular filling pressure


degenerative mitral valve disease


N-terminal pro-atrial natriuretic peptide


N-terminal pro-brain natriuretic peptide

Peak A

peak velocity of late diastolic transmitral flow

Peak E

peak velocity of early diastolic transmitral flow

Peak Ea

peak velocity of early diastolic mitral annular motion of the mitral annulus

Peak TR

peak velocity of tricuspid regurgitation

Peak Vp

peak velocity of propagation of early transmitral flow


receiver-operating characteristic




shortening fraction

Congestive heart failure (CHF) is a common and often fatal clinical syndrome in dogs characterized by cardiac dysfunction, neurohormonal activation, sodium and water retention, and increase in left ventricular (LV) filling pressures (LVFP).1,2 It occurs most often secondary to degenerative mitral valve disease (MVD)3–5 and dilated cardiomyopathy (DCM).4 Early recognition of CHF is of clinical importance.5 CHF can be suspected by clinical signs although reliability of such findings may be limited. Thoracic radiography is the most commonly applied method for the diagnosis of CHF and is considered the clinical “gold standard.”6 However, radiography is of unspecified sensitivity and specificity, especially in the setting of combined heart and lung disease, and can suffer from considerable observer variation.6,7 Plasma concentration of N-terminal pro-brain natriuretic peptide (NT-proBNP) is increased in patients with advanced MVD and DCM and may be useful in the diagnosis of CHF.4,5,8–11 However, a wide overlap of circulating NT-proBNP concentrations in dogs with and without CHF has been reported and generally accepted discrimination limits have not been determined.5,8,12 In addition, effects of renal function, day-to-day variability, and considerable turn-around time make this variable poorly suited for situations where bedside decisions are immediately required.

The development of cardiogenic pulmonary edema is predicted largely by the magnitude of volume overload and resulting increase in LVFP.2,13,14 A simple but quantifiable noninvasive method for estimation of volume status and filling pressures could not only refine the diagnosis, but also promote the early recognition of CHF, advance optimal medical management, and facilitate therapeutic monitoring. The recent introduction of novel Doppler echocardiographic (DE) techniques has sparked considerable interest in the noninvasive prediction of CHF by DE.15–19 One variable, the ratio between peak velocity of early diastolic transmitral flow (Peak E) to peak early tissue Doppler mitral annulus velocity (Peak Ea; E : Ea), has gained the most attention in the prediction of LVFP in dogs15–17,20,21 and people.18,22 Previous validation studies15,16 in experimental dogs reported on the use of isovolumic relaxation time (IVRT) and the ratio between Peak E to IVRT (E : IVRT) in the diagnosis of increased LVFP. These variables, however, have not been validated in dogs with naturally acquired heart disease.

Therefore, we undertook a study to test the hypothesis that DE indices of LV filling would predict CHF in dogs with spontaneous heart failure with clinically acceptable accuracy. More specifically, we hypothesized that E : Ea, E : IVRT, and IVRT would be most predictive of high CHF scores in dogs with MVD and DCM.

Materials and Methods

The study protocol was reviewed and approved by the Institutional Animal Care and Use Committee (protocol #2004A0196) and the Review Board of the Department of Veterinary Clinical Sciences, College of Veterinary Medicine, The Ohio State University, Columbus, OH.

Dogs, Clinical Examinations, and Group Assignment

Sixty-three client-owned dogs were prospectively studied. Dogs were consecutively selected over a time period of 2 years (2007 and 2008) based upon the echocardiographic diagnosis of MVD3,23 and DCM.24 All dogs underwent a thorough physical examination, a noninvasive measurement of systolic blood pressure,a thoracic radiography,b,c,d,e blood biochemical analyses, and a 2-dimensional (2D), M-mode, and DE study.f Heart rate and respiration rate were taken during initial physical examination without consideration of ambient temperature and determined as the number of beats or respirations per minute, respectively. If respiration rate could not be obtained due to panting, dogs were reassessed within 1 hour of arrival in order to obtain a definitive rate. Dogs with atrial flutter and fibrillation, arrhythmogenic right ventricular cardiomyopathy, systemic hypertension (systolic blood pressure >170 mmHg), and evidence of concomitant diseases or conditions such as hypothyroidism, renal failure, primary tracheal or pulmonary disease, anemia, or cancer were excluded. Using clinical, radiographic, and echocardiographic data, dogs were divided into 4 groups for statistical analyses: MVD or DCM with or without evidence of CHF (Groups 1–4), respectively. In preclinical (asymptomatic) dogs with MVD (Group-1), no treatments other than an ACE inhibitor were permitted. In dogs with asymptomatic DCM (Group-3), no treatments other than an ACE inhibitor, spironolactone, carvedilol, and pimobendan were permitted. In dogs with CHF (MVD, Group-2; DCM, Group-4), no treatments other than an ACE inhibitor, spironolactone, carvedilol, pimobendan, and furosemide (given within 12 hours of thoracic radiographs and echocardiography) were permitted. Dogs with signs of CHF were sedated upon arrival at the hospital with acepromazineg (0.025–0.050 mg/kg, IM; n = 2), butorphanolh (0.15–0.25 mg/kg, IM; n = 21), or a combination of both drugs (n = 7).

Thoracic Radiography

Thoracic radiographs were taken in 3 different imaging planes (right lateral, left lateral, and ventral dorsal projections) before echocardiography. Radiographs were assessed by the attending clinician only after the echocardiographic exam was performed. If noncardiac disease or cardiac disease other than MVD and DCM were observed, dogs were disqualified from the study. For final radiographic diagnosis and definitive group assignment (CHF absent or CHF present), radiographic images were reassessed by 2 independent board-certified radiologists (V.F.S. and L.J.Z.) en bloc at the end of the case recruitment period. All 63 studies were ordered randomly and coded by the principal investigator. The radiologists were aware of the 2 principal diagnoses (MVD and DCM); however, they were blinded to the animal's identification, the date of the examination, the initial interpretation of the images, the results of natriuretic peptide analysis, echocardiographic findings, the clinical diagnosis, and each other's assessment. A radiographic composite CHF score including criteria for left atrial (LA) enlargement, pulmonary venous congestion, pulmonary infiltrates compatible with cardiogenic edema, and pleural effusion was used (Appendix 1).i,25 The final radiographic assessment revealed a numerical radiographic CHF score and a qualitative, binary variable (CHF absent or CHF present). For group assignment, only the latter variable was used. To assure proper application of suggested diagnostic criteria and thus consistency of data, investigators met for a 1-hour training session before final analysis by mock radiographic images. After interpretation of films, results of the 2 readers were compared. If there was disagreement with regard to the principal diagnosis (CHF absent or CHF present), the investigators collectively reassessed the images in question to come up with a final definitive diagnosis that satisfied both raters. Intraobserver reproducibility of the radiographic diagnosis of CHF was determined by 15 studies (7 from dogs with compensated disease and 8 from dogs with CHF) interpreted 3 times by 1 blinded observer (V.F.S.) using a random order list.

Collection of Blood and Analysis of Natriuretic Peptides

In each dog, 5 mL of blood were collected from a jugular vein directly into serum tubes. After 20 minutes of storage at room temperature (22–24°C) to allow for stable clot formation, samples were centrifuged at 3,000 ×g for 10 minutes at 5°C and further processed within 15 minutes. The supernatant (serum) was harvested, divided into four 0.5 mL aliquots, transferred into plastic cryotubes, and stored at −80°C for a maximum of 4 weeks until analysis. Two aliquots of each sample were shipped on dry ice to the reference laboratoryj where batched analysis was done once a month. Assays were run in duplicate, and the average of the 2 values was used for final data analysis.

Serum N-terminal pro-atrial natriuretic peptide (NT-proANP) and serum NT-proBNP were analyzed with commercial test kits (sandwich enzyme immunoassays) with antibodies raised against human NT-proANP31-67k,26 and canine NT-proBNP25-41 (capture antibody) and canine NT-proBNP1-22 (detection antibody).l The detection limits were 50 pmol/L for the NT-proANP assay and 42 pmol/L for the NT-proBNP assay. The intra-assay coefficients of variation (CV) were below 10% for the NT-proANP assay and below 15% for the NT-proBNP assay as reported by the manufacturer.


Transthoracic 2D, M-mode, and DE examinations were performed by a single operator (K.E.S.; n = 53) or under direct supervision of the principal investigator (n = 10) with the dogs in right and left lateral recumbency with a digital ultrasound systemf with a transducer array of 3.5 MHz nominal frequency as recently described.15 Right parasternal standard views optimized for the LA, the LV outflow tract (long axis), or the LV (short axis) and left apical parasternal standard views optimized for the LV inflow tract, longitudinal motion of the lateral and septal mitral annulus, or the LV outflow tract were used for data acquisition. 2D cine loops and Doppler tracings were recorded and stored on the internal hard drive of the echocardiograph or on magneto-optical disc and analyzed off-line. A simultaneous 1-lead ECG was recorded. Heart rate was determined from the preceding R-R interval on the ECG and represents the mean of 5–12 calculations. Measurements were obtained from digital still images as an average of 5–12 consecutive cardiac cycles, irrespective of respiratory phase. Only high-quality images were used for data analysis. All measurements were done off-line by 1 trained investigator (T.M.H.) blinded to the dogs' identification, clinical signs, thoracic radiographs, and hemodynamic status and verified by the principal investigator (K.E.S.).

Echocardiographic Data Analysis

Fourteen variables were measured and 10 variables were calculated as recently described in dogs.15,16,27 In brief, from right parasternal long-axis recordings, the maximum (LADmax) and minimum (LADmin) septal-to-posterior dimension of the LA,27 the maximum (LAAmax) and minimum (LAAmin) area of the LA, and the end-diastolic dimension of the aortic valve (Ao) were measured using imaging views optimized for the LA or the LV outflow tract and inner edge projections. From a LV short-axis M-mode recording at the level of the chordae tendineae, LV dimensions in systole (LVDs) and diastole (LVDd) were measured. From the left apical 3-chamber view, IVRT was measured as the time period from the Doppler signal of aortic valve closure to the beginning of the transmitral early diastolic flow wave (E wave) with a pulsed wave sample volume of 6–10 mm placed in an intermediate position between the LV inflow and outflow tracts. Transmitral flow was recorded from the left apical 2-chamber view with a pulsed wave sample volume (2 mm in width) placed between the tips of the opened mitral valve leaflets. Peak E and the peak velocity of the late diastolic transmitral flow wave (Peak A) were measured. Deceleration time of the early diastolic transmitral flow (DTE) was measured as the time from Peak E velocity to the end of E at the baseline. Summated E and A waves were not measured. Duration of the A wave (Aduration) was measured from the beginning to the end of the A wave. Pulmonary venous flow was recorded from the same left apical 2-chamber view, with minimized baseline filter, optimized velocity scale, and with a pulsed wave sample volume of 4–6 mm placed 5–10 mm within the pulmonary vein of the left caudal lung lobe.28 Only the duration of the late diastolic reversal wave (ARduration) was measured. Pulsed-wave Doppler-derived velocities of myocardial motion were recorded from an apical 2-chamber view, with a sample volume of 5–6 mm placed in the septal or lateral aspects of the mitral annulus. Frame rate during tissue Doppler studies was optimized (>160 frames per second) by narrowing the imaging sector. Peak velocity of early diastolic septal (Peak Ea Sept) and lateral (Peak Ea Lat) mitral annulus motion was measured. Color M-mode recordings of early diastolic LV inflow were used to determine LV flow propagation velocity (Peak Vp). Color Doppler transmitral flow recordings were obtained, the Nyquist limit was reduced to approximately 50% of Peak E, and the slope of the 1st aliasing velocity line from the tip of the mitral valve to 3 cm distally into the LV cavity was used for determination of Peak Vp.29 From a right parasternal tilted short-axis view or a left apical cranial view optimized for the right ventricular inflow tract, peak tricuspid regurgitation (Peak TR) velocity was recorded. A Peak TR velocity of ≥3.0 m/s was considered an echocardiographic finding suggestive of mild to severe pulmonary hypertension.30,31 Variables calculated included LA shortening fraction (SF; LA-SF = [LADmax− LADmin]/LADmax× 100%), LA fractional area change (LA-FAC = [LAAmax− LAAmin]/LAAmax× 100%), the LA-to-Ao ratio (LADmax : Ao), LV shortening fraction (LV-SF = [LVDd− LVDs]/LVDd× 100%), the ratio between Peak E (cm/s) to IVRT (ms; E : IVRT); the ratio between Peak E to Peak A (E : A); the ratio between the duration of A to the duration of AR (Aduration : ARduration), the ratios between Peak E to Peak Ea Sept (E : Ea Sept), Peak E to Peak Ea Lat (E : Ea Lat), and Peak E to Peak Vp (E : Vp), and [(E : Ea Lat) + (E : Ea Sept)]/2. LV diastolic function was classified22 based on E : A and qualified based on E : Ea Lat − Class-1: Normal pattern (1.0 ≤ E : A ≤ 2.0), Class-2: Relaxation delay pattern (E : A < 1.0), Class-3: Pseudonormal pattern (1.0 ≤ E : A ≤ 2.0), and Class-4: Restrictive pattern (E : A > 2.0). Class-1 and Class-3 were discriminated using cut-offs of E : Ea Lat of 11.0 for dogs with MVD and 9.0 for dogs with DCM. These discrimination values were obtained from a previous pilot study of 25 dogs with DCM and 52 dogs with MVD.m,32 In dogs older than 10 years, any observed E : A between 1.0 and 2.0 was considered pseudonormal transmitral flow because of the fact that a physiologic age-related relaxation delay pattern would be expected in such dogs.33

Measurement reliability was determined for selected echocardiographic variables. Four echocardiograms each from dogs with MVD and DCM were randomly selected from the pool of studies to undergo 5 repeated analyses by 1 observer (T.M.H.) to determine intraobserver measurement variability. The same 8 echocardiograms underwent repeated analyses by a 2nd independent observer (K.E.S.) to determine interobserver measurement variability. Both investigators were blinded to the results of prior studies.

Statistical Analyses

Statistical analyses were done with commercially available software.n,o,p All continuous outcome variables were visually inspected and tested for normality by the Kolmogorov-Smirnov test. Descriptive statistics were determined: frequencies for categorical variables and median and 5 and 95 percentiles for continuous variables. Selected data were graphically depicted as median and scatter plot of individual data points. Groups of dogs with MVD (Group-1 versus Group-2) and DCM (Group-3 versus Group-4) were compared using an unpaired t-test if standard assumptions were fulfilled or the Mann-Whitney rank-sum test if not. Proportions were compared by Fisher's exact test. Receiver-operating characteristic (ROC) curve analysis was used to determine the diagnostic ability of heart rate (taken at physical examination), respiratory rate, NT-proANP, NT-proBNP, and various DE variables in the diagnosis of CHF and to define optimal discrimination limits (diagnostic cut-off values) for such prediction. The area under the ROC curve (AUC) was used as a summary measure for diagnostic accuracy and to quantify the ability of variables to predict CHF and is reported with 95% confidence intervals. Cut-off values were chosen based on the highest Youden index (Y = (Se + Sp) − 1) of various combinations of Se and Sp.34 Multivariate logistic regression was used to predict the qualitative dichotomous variable “CHF absent or CHF present” (based on radiographic diagnosis) from observations of 1 or more independent variables by fitting a logistic function to the data. In addition to all tabulated variables, variables such as prior treatment with furosemide and presence of azotemia were also considered in the regression. Once variables with significant (P < .05) associations were identified, forward selection and best subsets regression analyses were performed to identify the combination of variables that predicted presence of CHF best and to determine the contribution of individual variables to the final model. For model fitting, a 1 : 7 approach was used (ie, addition of no more than 1 independent variable to the model for every 7 observations).

Conformity among observers with regard to radiographic interpretation of recordings after 1st assessment were determined by the nonparametric McNemar's test,35 Bowker's test of symmetry,36 and by calculating Cohen's Kappa coefficients (κ).37 Intraobserver reproducibility of the radiographic diagnosis of CHF was determined by 1-way random effects models for calculation of the intraclass correlation coefficient (ICC).38 Observer variability of echocardiographic measurements was calculated by coefficients of variation (CV = [standard deviation / average of measurements] × 100) and expressed as CV in percent and also as absolute value of the respective variable.39 For all analyses, P-values ≤.05 were considered significant.


Demographic data, historical findings, and results of physical examination and blood pressure measurement are summarized in Table 1. At presentation, 11 (52%) dogs of Group-1 and 10 (42%) dogs of Group-2 were on long-term treatment with enalapril or benazepril. In Group-2, 12 (50%) had received furosemide within the last 12 hours before the study; however, they were still diagnosed with CHF at the time of thoracic radiography and echocardiography. At presentation, 1 (9%) dog of Group-3 and 2 (28%) dogs of Group-4 were on long-term treatment with enalapril or benazepril. In Group-4, 2 (28%) dogs had received furosemide within the last 12 hours before the study; however, they were still diagnosed with CHF at the time of thoracic radiography and echocardiography. Azotemia was detected in 1/10 dogs of Group-1, 5/24 dogs of Group-2, and no dogs in Group-3 or Group-4. In all dogs, BUN was <50 mg/dL (reference range, 5–20 mg/dL) and creatinine was <2.7 mg/dL (reference range, 0.6–1.6 mg/dL).

Table 1.   Demographic data and results of history, physical examination, and blood pressure measurement in 63 study dogs.
  • Group-1, degenerative mitral valve disease (MVD) without congestive heart failure (CHF); Group-2, MVD with CHF; Group-3, dilated cardiomyopathy (DCM) without CHF; Group-4, DCM with CHF. Median (5 and 95 percentiles) for continuous data and number (n) or percent (%) for frequency data.

  • *

    Within a row, values between Group-3 and Group-4 differ significantly (P≤ .05).

  • Within a row, values between Group-1 and Group-2 differ significantly (P≤ .05).

Number of different breeds101455
Most common breed (n)CKCS (9)CKCS (4)Doberman (5)Doberman (4)
Age (years)10 (6–12)10 (7–13)6 (1–13)7 (5–12)
Body weight (kg)11 (5–37)10 (4–36)32 (11–58)37 (30–54)
Sex (female : male)1.100.851.200.75
Exercise intolerance (%)1010018100*
Cough (%)4384086*
Tachypnea and dyspnea (%)075971*
Syncope and near-syncope (%)029914
Respiration rate (per minute)28 (19–57)54 (38–85)26 (20–32)56 (36–60)*
Heart rate (per minute)132 (91–169)140 (100–206)114 (70–198)160 (120–210)*
Systolic heart murmur (%)10010091100
Murmur grade (0/6–6/6)3 (2–6)4 (3–6)2 (0–3)2 (1–3)
Systolic blood pressure (mmHg)130 (97–160)125 (90–156)131 (112–170)113 (90–128)*

Results on radiographic composite score, NT-proANP, and selected echocardiographic variables are presented in Table 2. One dog of Group-1 and 2 dogs of Group-2 had a radiographic composite score of 4.0. All other dogs were either below 4.0 (Group-1) or above 4.0 (Group-2). Two dogs of Group-3 had a radiographic composite score of 4.0. All other dogs were either below 4.0 (Group-3) or above 4.0 (Group-4). The minority of dogs with compensated heart disease (15% in Group-1 and 27% in Group-3) and the majority of dogs with CHF (71% in Group-2 and 86% in Group-4) had echocardiographic evidence of mild to severe pulmonary hypertension using a diagnostic cut-off of 3.0 m/s Peak TR velocity for estimation.30,31 Summated E and A waves seen in 2 dogs with MVD were discarded from further data analysis.

Table 2.   Radiographic score, serum concentrations of NT-proANP, and selected echocardiographic variables in 63 study dogs.
  1. NT-proANP, N-terminal pro-atrial natriuretic peptide; LADmax, maximum left atrial dimension; Ao, aortic annular dimension; FAC, fractional area change; SF, shortening fraction; Peak TR, peak velocity of tricuspid regurgitation; Peak E, peak velocity of early transmitral flow; Peak Vp, peak velocity of propagation of early transmitral flow; Peak Ea, peak velocity of early diastolic mitral annular motion; Lat, measured at the lateral mitral annulus; septal, measured at the septal mitral annulus.

  2. See Table 1 for key.

  3. Median (5 and 95 percentiles) for continuous data and number (n) or percent (%) for frequency data.

Radiographic composite score1 (0–4)7 (4–9)1 (1–4)7 (5–8)*
NT-proANP (pmol/L)362 (151–980)941 (325–3,035)565 (120–1,644)542 (234–990)
LADmax : Ao2.23 (1.80–3.12)3.19 (2.31–4.23)2.18 (1.53–2.66)2.81 (2.51–3.60)*
LA-FAC (%)37 (17–49)24 (10–43)26 (18–43)11 (5–34)*
LA-SF (%)20 (7–28)11 (5–23)15 (8–27)6 (3–19)*
LV-SF (%)39 (24–55)43 (24–59)21 (10–25)16 (10–23)
TR (%)86929186
Peak TR (m/s)2.81 (2.19–3.15)3.43 (2.08–4.61)2.77 (2.18–3.46)3.18 (2.28–3.83)
TR > 3 m/s (%)1571‡2786*
Peak E (m/s)0.92 (0.60–1.45)1.41 (1.04–1.74)0.80 (0.54–1.02)1.39 (0.70–1.74)*
DTE (ms)88 (65–112)86 (62–120)100 (64–128)62 (57–104)*
Peak Vp (cm/s)51 (30–95)52 (38–87)33 (23–48)33 (28–43)
E : Vp1.66 (0.88–3.94)2.63 (1.36–3.77)2.17 (1.41–3.56)3.63 (2.47–4.56)*
Peak Ea Lat (cm/s)10.1 (6.3–16.1)8.7 (6.2–17.4)10.3 (6.7–17.7)8.54 (6.7–17.1)
Peak Ea Sept (cm/s)8.82 (5.16–14.16)9.80 (5.49–15.10)7.37 (5.57–16.54)7.08 (5.55–12.52)
E : Ea Sept10.6 (6.7–16.6)14.7 (7.6–22.8)9.4 (5.9–13.2)14.9 (8.7–24.6)*
[(E : Ea Sept) + (E : Ea Lat)]:210.0 (6.5–15.1)14.4 (6.1–22.3)8.4 (4.2–11.0)15.2 (8.9–22.5)*

Figures 1 and 2 illustrate median and scatter plots of respiration rate, NT-proBNP, E : A, Diastolic Functional Class, IVRT, E : IVRT, E : Ea Lat, and Aduration : ARduration in dogs with MVD (Fig 1) and DCM (Fig 2) and differences between dogs with compensated disease (Group-1 and Group-3) and dogs with CHF (Group-2 and Group-4).

Figure 1.

 Medians and scatter plots of respiration rate, NT-proBNP, and Doppler echocardiographic variables of left ventricular (LV) filling in 21 dogs with compensated degenerative mitral valve disease (MVD; Group-1) and 24 dogs with MVD and congestive heart failure (Group-2). Respiration rate (median, 28 versus 54 bpm, P < .001); NT-proBNP (median, 848 versus 2,750 pmol/L, P < .001); E : A ratio (median, 1.34 versus 2.21, P < .001); Class of LV Diastolic Function (median, 1 versus 4, P < .001); isovolumic relaxation time (IVRT) (median, 57 versus 36 ms, P < .001); E : IVRT ratio (median, 1.74 versus 3.71, P < .001); E : Ea (median, 8.9 versus 15.2, P < .001); and Aduration : ARduration ratio (median, 1.30 versus 1.26, P= .79).

Figure 2.

 Medians and scatter plots of respiration rate, NT-proBNP, and Doppler echocardiographic variables of left ventricular (LV) filling in 11 dogs with compensated dilated cardiomyopathy (DCM; Group-3) and 7 dogs with DCM and congestive heart failure (Group-4). Respiration rate (median, 26 versus 56 bpm, P < .001); NT-proBNP (median, 1,314 versus 3,830 pmol/L, P < .001); E : A ratio (median, 1.10 versus 2.73, P < .001); Class of LV Diastolic Function (median, 1 versus 4, P < .001); isovolumic relaxation time (IVRT) (median, 66 versus 32 ms, P < .001); E : IVRT ratio (median, 1.11 versus 4.41, P < .001); E : Ea (median, 7.9 versus 13.3, P= .003); and Aduration : ARduration ratio (median, 1.47 versus 0.73, P < .001).

In dogs with MVD, DTE (P= .59), heart rate taken at physical exam (P= .39), Aduration : ARduration (P= .79), Peak Vp (P= .98), and SF (P= .69) did not predict presence of CHF (Tables 3 and 4). In dogs with DCM, NT-proANP (P= .89) and Peak Vp (P= .87) did not predict presence of CHF.

Table 3.   Areas under the receiver-operating characteristic curve (AUC) and optimal diagnostic cut-offs of clinical and echocardiographic variables in the prediction of congestive heart failure in 45 dogs with degenerative mitral valve disease.
 AUC95% CICut-offSeSpP
  1. CI, confidence interval; Se, sensitivity; Sp, specificity.

  2. See Table 2 for key.

E : IVRT0.970.92–1.022.500.920.96< .001
Respiration rate (per minute)0.940.84–1.04410.920.94< .001
Diastolic Functional Class0.930.85–1.0120.921.00< .001
LA : Ao0.900.81–0.992.520.920.81< .001
E : A0.890.78–0.991.580.870.86< .001
Peak E (m/s)0.870.77–0.981.080.960.71< .001
IVRT (ms)0.860.75–0.98460.880.76< .001
E : Ea Lat0.830.70–0.9511.50.750.91< .001
NT-proBNP (pmol/L)0.830.75–0.981,9510.750.86< .001
NT-proANP (pmol/L)0.830.71–0.955840.780.71< .001
[(E : Ea Lat) + (E : Ea Sept)]/20.790.64–0.9312.40.740.86.001
E : Vp0.750.61–0.901.870.790.62.004
E : Ea Sept0.740.59–0.9014.70.550.95.006
Table 4.   Areas under the receiver-operating characteristic curve (AUC) and optimal diagnostic cut-offs of clinical and echocardiographic variables in the prediction of congestive heart failure in 18 dogs with dilated cardiomyopathy.
 AUC95% CICut-offSeSpP
  1. See Tables 2 and 4 for key.

Diastolic Functional Class1.001.00–1.0021.001.00< .001
Respiration rate (per minute)1.001.00–1.00341.001.00< .001
E : IVRT0.990.95–1.031.881.000.91< .001
IVRT (ms)0.980.93–1.03431.000.91< .001
Aduration : ARduration0.980.93–< .001
E : A0.970.89–
E : Vp0.950.85–1.053.560.860.90.002
E : Ea Lat0.940.82–
NT-proBNP (pmol/L)0.940.83–1.052,4921.000.82.002
[(E : Ea Lat) + (E : Ea Sept)]/20.940.82–1.0611.00.861.00.002
DTE (ms)0.910.76–1.06720.860.90.005
Peak E (m/s)0.890.69–
E : Ea Sept0.870.68–1.0713.90.711.00.013
LA : Ao0.860.69–1.032.461.000.73< .001
HR (per minute)0.830.60–1.061350.711.00.021
SF (%)0.790.57–1.00190.710.82.046

Regression analyses revealed that presence of CHF in dogs with MVD can be predicted from a combination of Diastolic Functional Class (step 1, R2= 0.58, P < .001), IVRT (step 2, cumulative R2= 0.72, P < .001), and respiration rate (step 3, cumulative R2= 0.80, P < .001) leading to the general prediction model: CHF = X + (0.190 Diastolic Functional Class) − (0.0104 IVRT [ms]) + (0.00935 Respiration Rate [per minute]). With the same methods in dogs with DCM, presence of CHF could be predicted from a combination of Diastolic Functional Class (step 1, R2= 0.91, P < .001) and E : IVRT (step 2, cumulative R2= 0.94, P= .021) leading to the general prediction model: CHF = X + (14.273 Diastolic Functional Class) + (0.843 E : IVRT).

Kappa (κ) for assessment of conformity between the 2 radiographic observers was 0.78 (95% CI, 0.70–0.86) for LA enlargement; 0.44 (95% CI, 0.32–0.55) for pulmonary infiltrates compatible with cardiogenic edema; 0.26 (95% CI, −0.05 to 0.56) for pleural effusion; 0.37 (95% CI, 0.21–0.52) for pulmonary venous congestion; and 0.65 (95% CI, 0.52–0.78) for the final diagnosis of CHF. Consistency of 1 radiographic observer (V.F.S.) in the diagnosis of CHF based on ICC was 0.92.

Results of studies on measurement variability of echocardiographic indices are summarized in Table 5. Coefficients of variation for intra- and interobserver measurement variability were <10 % for all but 3 variables.

Table 5.   Intra- and interobserver measurement variability of 18 echocardiographic variables assessed in 8 randomly selected dogs.
CV (%)AbsoluteCV (%)Absolute
  1. CV, coefficient of variation.

  2. See Table 2 for key.

LADs (cm)1.410.071.410.06
Ao (cm)2.500.044.320.08
LADs : Ao3.530.104.000.11
IVRT (ms)4.282.337.204.00
Peak E (m/s)0.480.013.440.03
DTE (ms)5.215.1911.0410.75
Peak A (m/s)1.360.011.840.01
E : A1.930.043.920.07
E : IVRT4.640.1010.450.22
Aduration (ms)3.413.165.615.25
ARduration (ms)3.473.056.245.50
Aduration : ARduration5.
Vp (cm/s)5.602.153.781.41
E : Vp5.
Ea Sept (cm/s)
E : Ea Sept3.360.403.180.38
Ea Lat (cm/s)
E : Ea Lat1.590.182.260.26


Increase in filling pressure is a unifying feature of CHF regardless of underlying cause.2,13 Because filling pressure cannot be directly measured in clinical situations; an easily applicable, reliable, noninvasive method is needed. Results of this study support the contention that CHF secondary to MVD and DCM can be predicted by DE. The major finding is that E : IVRT, Diastolic Functional Class, and IVRT allow for a rapid and feasible estimation of whether or not CHF is present. In addition, respiration rate taken at physical exam was equally useful in the prediction of CHF. Moreover, disease-specific differences between dogs with MVD or DCM regarding the diagnostic accuracy of individual DE variables and clinical decision thresholds were identified and thus need to be considered clinically.

Diagnosis of CHF by Single DE Variables

Our results are in agreement with previous findings from studies in anesthetized, volume loaded dogs16 and dogs with rapid pacing-induced CHF.15 In the former16 in which mean LA pressure was measured directly, E : IVRT outperformed 8 commonly used DE variables of filling pressure and predicted increased LA pressure (≥15 mmHg) with high accuracy. In the latter,15 a decrease of LV end-diastolic pressure after furosemide was best predicted by a decrease of E : IVRT. In both studies,15,16 neither E : Ea nor E : Vp was diagnostically useful in the prediction of increased LVFP. The rationale behind the use of combined indices such as E : IVRT, but also E : Ea and E : Vp, is to “correct” for the effect of relaxation on a variable that is largely dependent on filling pressure, but also influenced by relaxation. By combining Peak E, a variable that is determined mainly by filling pressure and relaxationn,18,40,41 with a variable that is most dependent on relaxation (eg, IVRT, Peak Ea, and Peak Vp),18,40 the effect of relaxation on Peak E can be minimized. Because increased filling pressure, a main hemodynamic characteristic of CHF,2,13 is associated with both increased Peak E and decreased IVRT,13,14,19,21,41 the ratio of E : IVRT should be high in dogs with CHF and low in dogs with compensated heart disease.14 The latter was confirmed in this study for both dogs with MVD or DCM.

The IVRT is an index of relaxation and is linearly related to tau, the time constant of LV isovolumic relaxation. However, it is also influenced by a multitude of other factors including preload, afterload, heart rate, and age.13,14 Therefore, IVRT represents the net effect of many determinants, only one of which is relaxation.42 Whereas a mild increase in LVFP as seen in the early stages of heart failure shortens tau (ie, improves relaxation) but is not associated with shortening of IVRT, moderate and severe increase in filling pressure as seen in CHF prolongs tau (ie, makes relaxation worse) but shortens IVRT in a linear manner.13,14,41,42 Shortened IVRT is by definition an integral part of restrictive LV filling, a transmitral flow pattern considered specific for advanced diastolic dysfunction, high filling pressure, and CHF.13,41,43 That is, high filling pressure can minimize the effect of relaxation on IVRT, turning it into a more specific indicator of LVFP and thus CHF.42 This was confirmed in our study, in particular in dogs with DCM. Both IVRT and Peak E (and thus E : IVRT) are relatively easy to measure DE variables and might provide, in concert with historical and clinical findings, immediate information on heart failure status in dogs with DCM or MVD.

The E : Ea ratio and the E : Vp ratio have been reported to be useful DE indices of LVFP and CHF in clinical trials in people19,40 with E : Ea preferred by most.18,19 Although studies on the use of such variables in experimental dogs are numerous,15,16,20,21,29 data on dogs with naturally acquired heart disease are sparse and its diagnostic value largely unproven.m,44 Both Ea and Vp have been shown to be relatively preload-independent indices of LV relaxation, making them suitable for correcting Peak E for the effects of relaxation.18,40 In healthy dogs, Vp and even more Ea are dependent on “lengthening load” (load that the myocardium experiences during relaxation).15,16,20,21,45 Absence of diastolic abnormalities primarily concerning myocardial relaxation and chamber compliance, typically seen in dogs with MVD,1 are potential reasons why Ea is sensitive to changes in preload under such circumstances,16,21 thereby limiting the global use of E : Ea as an index of LVFP and CHF. In contrast, a close linear relationship between mean left atrial pressure and E : Ea (r= 0.83, P < .05) was reported from studies17 using a dog model of acute LV volume overload secondary to severe iatrogenic mitral regurgitation. In another study44 involving 39 dogs with naturally acquired MVD, a correlation between E : Ea and class of heart failure was documented. Using a cut-off of 13.0, E : Ea identified CHF with a sensitivity of 0.80 and a specificity of 0.83, which is similarly low compared with our findings. Conflicting results on the use of E : Ea as a reliable index of congestion and increased filling pressures have been reported in people with primary mitral valve regurgitation, with most studies rejecting the use of E : Ea under such circumstances.43,46 In contrast to DCM, the preload dependency of Ea in hearts with preserved systolic function and normal to only mildly affected diastolic function in concert with disproportionate volume overload may limit the use of E : Ea in the prediction of CHF in dogs with MVD.20,43,45,46

Diagnosis of CHF by Doppler Transmitral Flow Patterns and Class of Diastolic Function

The mitral inflow velocity profile is determined in a complex manner by multiple factors, which include left atrial pressure, relaxation, LV systolic pressure, ventricular suction, preload, heart rate, and atrial function.47 The pattern of LV filling determined by Doppler transmitral flow is used to noninvasively evaluate LV diastolic performance and has been described in detail in people22,41 and validated in experimental dogs.13,14,47 Early stages of LV dysfunction most often seen in asymptomatic dogs with heart disease (or healthy dogs older than 10 years)33 commonly lead to a delayed relaxation transmitral flow pattern.14,22,48 At this stage, LV relaxation is prolonged but filling pressure still normal or only minimally increased.48 With advancement of disease, LVFP rises and overrides the relaxation abnormality dominant influence on LV filling, leading to a renormalized (yet “pseudonormal”) flow pattern.14,49 The final stage in the natural history of LV diastolic dysfunction is restrictive filling, a flow pattern indicating markedly increased LVFP, most commonly associated with CHF.49 From a clinical perspective it is of utmost importance to distinguish normal from pseudonormal flow for which variables such as Aduration : ARduration or E : Ea may be instrumental.22,41 The results of the present study proved the clinical usefulness of assessing LV diastolic function in the DE diagnosis of CHF in dogs with MVD or DCM. Diastolic Functional Class, in particular restrictive filling, was highly predictive of CHF status. However, it also became obvious that because of differences in disease characteristics, LV filling patterns in dogs with DCM and MVD need different clinical decision-making cut-offs and require differential interpretation. Whereas all dogs with preclinical DCM had DE evidence of normal LV diastolic function or delayed LV relaxation and all dogs with symptomatic DCM had restrictive LV filling, the situation was less definitive in dogs with MVD. Owing to the fundamental structural and functional differences between MVD and DCM it is unlikely that a universal approach to both conditions can be advanced with regard to interpretation of DE variables. In DCM, there is primarily a systolic dysfunction-dominant influence on LV filling and LV stiffness is increased.1 Early work done in experimental dogs and people revealed that the diagnostic ability to assess LV diastolic function using transmitral flow patterns is best when ejection fraction is reduced.13,22 This was confirmed in our study in which 9 DE variables had an AUC of ≥0.90 indicating excellent diagnostic performance in the diagnosis of CHF in dogs with DCM. In contrast, the volume-overload dominant influence on LV filling combined with often normal global LV systolic function, chamber compliance, and relaxation limits the value of transmitral flow patterns in the diagnosis of CHF in the setting of MVD.21,22,42 Pseudonormal and even restrictive LV filling can both be the sole consequence of volume overload.21 True restrictive filling is characterized by markedly increased E : A, a very short DTE, and a reduced Peak Ea.21,49 However, if E : A is increased but DTE is only mildly reduced and Peak Ea is normal or even increased, the effect of volume (ventricular “overfilling”), and not pressure, is considered the main mechanism in the generation of the filling pattern termed “pseudorestrictive.”21 Therefore, the diagnosis of CHF for an individual dog with MVD requires a stepwise approach incorporating all available data, and caution is advised in the isolated use of transmitral flow patterns for such purpose.21,42,43,46,49

Diagnosis of CHF Using Natriuretic Peptides

Circulating natriuretic peptides are increased in dogs with CHF because of MVD4,5,7,10,12 and DCM4,9–12; and dogs with CHF have higher blood concentrations than dogs with preclinical disease.4,5,8–11 Results of the present study indicate that (a) NT-proANP concentrations in dogs with decompensated MVD were significantly (P < .05) higher than in dogs with decompensated DCM despite similar concentrations in dogs with preclinical MVD or DCM; (b) NT-proBNP concentrations in dogs with preclinical and decompensated DCM were significantly (P < .05) higher than in dogs with preclinical and decompensated MVD; (c) both NT-proANP and NT-proBNP can be used in the prediction of CHF in dogs with MVD although a clinically important overlap between dogs with preclinical and decompensated disease was found; (d) only NT-proBNP but not NT-proANP can be used in the prediction of CHF in dogs with DCM; and (e) both natriuretic peptides have less accuracy in the prediction of CHF as compared with many DE variables and respiration rate. Our findings with regard to differences in natriuretic peptide concentrations between compensated and decompensated dogs are similar to findings from a recent studies in 156 dogs with MVD5,8 and 15 dogs with DCM11 but dissimilar, at least in part, from findings in 20 Doberman Pinschers with DCM9 and 137 dogs with MVD or DCM.4 The identification of reasons for such discrepancies was beyond the scope of the present study but may indicate that more research is needed to fully appreciate the clinical usefulness and potential incremental value of natriuretic peptide analysis in the diagnosis and management of patients with MVD and DCM.

Diagnosis of CHF Using Respiration Rate

Respiration rate and the pattern of ventilation have a long-standing clinical basis for identification of CHF. However, effects of CHF on respiration rate have not yet been prospectively studied in dogs to our knowledge. Respiration rate is controlled by involuntary and voluntary mechanisms and, among others, correlates closely to the amount of lung water.r,2 In the present study, respiration rate taken at the hospital outperformed most DE and laboratory variables in the diagnosis of CHF. A diagnostic cut-off of 41 and 34 per minute in dogs with MVD and DCM, respectively, was useful in the prediction of presence or absence of CHF with high accuracy (sensitivity and specificity between 92 and 100%). These results are very encouraging as respiration rate is very simple to obtain, does not add additional costs, and can be done repeatedly by instructed owners in the home environment. The clinical relevance of this finding is several-fold: determination of respiration rate might be beneficial in the early diagnosis of CHF in dogs with MVD and DCM; it might allow for earlier therapeutic intervention; it could be a very cost-effective tool in the assessment of success of treatment of CHF; and it might allow both clinicians and owners to more accurately tailor home treatment to a target RR and to avoid excessive or insufficient diuresis. Although unproven, respiration rate might also be useful in the discrimination on respiratory distress because of either CHF or primary respiratory disease. This study suggests that veterinarians should use respiration rate in the diagnosis of CHF and instruct owners to monitor RR in dogs with MVD and DCM at risk for developing CHF.


Certain weaknesses of this study need emphasis. One limitation is the use of thoracic radiography in the diagnosis of CHF because of its limited diagnostic accuracy, owing to a multitude of factors.6,7,25 Pulmonary capillary wedge pressure, a surrogate measure of left atrial pressure and thus filling pressure,50 was not measured in our dogs. Therefore, the presence or absence of increased LVFP could only be assumed on clinical and radiographic ground and a quantitative relationship between LVFP and echocardiographic and biochemical variables could not be determined. Furthermore, because of the mode of selection of dogs, pretest probability of CHF in symptomatic dogs was high, limiting the more global applicability of the results of this study to all dogs with clinical signs of CHF. Dogs with CHF were sedated, which may have influenced central hemodynamics and thus DE variables. The number of dogs studied was small, in particular dogs with DCM, rendering the study underpowered to detect differences among groups. Moreover, echocardiography and blood sampling were done only once in each dog, neglecting day-to-day and circadian rhythms of filling pressure and circulating natriuretic peptide concentrations.51,52 In addition, variables and cut-offs used for classification of LV diastolic function were chosen based on recommended flow patterns for use in people41 and data obtained from a prior pilot studym in dogs with MVD and DCM. Age, body weight, and heart rate were not specifically considered for diastolic classification and determination of discrimination limits, although such variables may influence LV diastolic function.33 Sample handlingq including short-term (<4 weeks) storage at −80°C might have affected analysis for natriuretic peptides. Finally, prior treatment, in particular administration of furosemide, could have confounded interpretation of DE and radiographic findings.


Despite these limitations, the clinical implications of our findings are that in dogs with CHF because of degenerative MVD and DCM, E : IVRT, Diastolic Functional Class, and IVRT are the Doppler indices best predicting presence of left-sided CHF but requiring disease-specific discrimination limits for clinical use. Respiration rate has a comparably high predictive power, is simple to obtain, but needs further study to fully appreciate its diagnostic value. Whether goal-directed transthoracic DE focusing on evaluation of filling variables and CHF can be used to guide in the management of patients with MVD and DCM remains to be determined. If useful under such circumstances, E : IVRT, Diastolic Functional Class, IVRT, and respiration rate may become simple to perform invaluable diagnostic studies for assessing CHF status and optimizing preload in dogs with left-sided CHF.


a Ultrasonic Doppler flow detector, Model 811-B, Parks Medical Electronics Inc, Aloha, OR

b Sedecal, Sedecal USA, Arlington Heights, IL

c Prestige II, GE Medical Systems, Milwaukee, WI

d Digital radiography system EDR-6, Sound-Eklin, Carlsbad, CA

e Viewing software E-film, Merge Healthcare, Milwaukee, WI

f Vivid 7 Vantage with EchoPac software package version BT05, GE Medical Systems

g Acepromazine maleate injection, Boehringer Ingelheim Vetmedica Inc, St Joseph, MO

h Butorphenol injection, IVX Animal Health Inc, St Joseph, MO

i Diana A, Sanacore A, Guglielmini C, et al. Radiographic features of pulmonary edema associated with mitral regurgitation in dogs. Vet Radiol Ultrasound 2008;49:213 (abstract)

j IDEXX Laboratories, Westbrook, MA

k VetSign Canine CardioSCREEN proANP, IDEXX Laboratories

l VetSign Canine CardioSCREEN NT-proBNP, IDEXX Laboratories

m Schober KE, Bonagura JD. Doppler echocardiographic assessment of the E : Ea ratio as an indicator of left ventricular filling pressure in normal dogs and dogs with heart disease. J Vet Intern Med 2005;931 (abstract)

n Sigma Stat, Version 3.5, SPSS Inc, Chicago, IL

o Prism 4, Graph Pad Software Inc, San Diego, CA

p SPSS Statistics version 9.2, SPSS Inc

q Farace G, Beardow A, Carpenter C, et al. Effect of shipping temperature on canine N-terminal pro-hormone atrial natriuretic peptide and N-terminal pro-hormone brain natriuretic peptide. J Vet Intern Med 2008;22:756 (abstract)

r DaCunha DNQ, Pedraza A, Kuenzler R, et al. Trends in respiration rate as an indicator of worsening heart failure. J Card Fail 2007;13 (Suppl 2): S173 (abstract)


The authors gratefully acknowledge Kathryn Meurs, John Mattoon, Nicole Ponzio, Laura Spayd, Agnieszka Kent, Patty Mueller, Becky Conners, and Richard Cober for their contributions.

This study was supported by a grant from the Morris Animal Foundation.


Table Appendix 1..   Radiographic composite score on congestive heart failure (CHF).
Left atrial enlargementNone0
Moderate to severe3
Pulmonary venous congestionNone0
Pulmonary infiltrates compatible with cardiogenic edemaNone0
Mild interstitial1
Diffuse interstitial2
Pleural effusionNone0
Final assessmentScore 0–2CHF not likely
Score 3–4CHF possible
Score >4CHF likely