The work was carried out at the Royal Veterinary College, the Beaumont Animals’ Hospital, Camden, and the Blue Cross Animal Hospital, Victoria.
The Combined Prognostic Potential of Serum High-Sensitivity Cardiac Troponin I and N-Terminal pro-B-Type Natriuretic Peptide Concentrations in Dogs with Degenerative Mitral Valve Disease
Article first published online: 28 FEB 2012
Copyright © 2012 by the American College of Veterinary Internal Medicine
Journal of Veterinary Internal Medicine
Volume 26, Issue 2, pages 302–311, March-April 2012
How to Cite
Hezzell, M.J., Boswood, A., Chang, Y.-M., Moonarmart, W., Souttar, K. and Elliott, J. (2012), The Combined Prognostic Potential of Serum High-Sensitivity Cardiac Troponin I and N-Terminal pro-B-Type Natriuretic Peptide Concentrations in Dogs with Degenerative Mitral Valve Disease. Journal of Veterinary Internal Medicine, 26: 302–311. doi: 10.1111/j.1939-1676.2012.00894.x
- Issue published online: 20 MAR 2012
- Article first published online: 28 FEB 2012
- Manuscript Accepted: 12 JAN 2012
- Manuscript Revised: 13 DEC 2011
- Manuscript Received: 18 JUL 2011
- Petplan Charitable Trust
- IDEXX Ltd
- Risk stratification;
- Valvular disease
Identification of factors associated with decreased survival in dogs with degenerative mitral valve disease (DMVD) will allow more accurate prognosis. N-terminal pro-B-type natriuretic peptide (NT-proBNP) is negatively associated with survival in dogs with DMVD. In human patients, multimarker strategies provide superior risk stratification compared with single markers.
High-sensitivity cardiac troponin I (hscTnI) and other clinical variables will be associated with survival time in dogs with DMVD. Measuring hscTnI and NT-proBNP in combination will be prognostically superior to measurement of either marker alone. The rate of change of these markers will vary according to cause of death.
Client-owned dogs (n = 202) with DMVD of varying severity and age-matched healthy control dogs (n = 30) recruited from first opinion private practice.
Prospective cohort study relating clinical variables at enrollment in dogs with DMVD to survival time (all-cause, cardiac, and noncardiac mortality). Multivariable Cox regression analysis was used to identify factors associated with survival. Measurements were obtained approximately every 6 months. Repeated measures models were constructed to assess changes over time.
hscTnI, LVEDDN, heart rate, and age were independently associated with decreased survival time (all-cause mortality). Survival times were shortest in dogs in which both serum hscTnI and NT-proBNP were increased. hscTnI and NT-proBNP increased more rapidly in dogs that died of cardiac disease.
Conclusions and Clinical Importance
Serum hscTnI has prognostic value in dogs with DMVD. Measurement of NT-proBNP and hscTnI is prognostically superior to measuring either alone. Serial measurement strategies provide additional prognostic information.
degenerative mitral valve disease
N-terminal pro-B-type natriuretic peptide
high-sensitivity cardiac troponin I
left atrial to aortic root ratio
left ventricular end-systolic diameter, normalized for body weight
left ventricular end-diastolic diameter, normalized for body weight
left ventricular end-diastolic diameter to left ventricular free wall thickness in diastole ratio
Cavalier King Charles Spaniel
receiver operating characteristic
angiotensin converting enzyme
Degenerative mitral valve disease (DMVD) is the most common cardiac disease in dogs. The rate of progression of the disease is variable, with many dogs never developing heart failure.[2, 3] The ability to identify dogs at higher risk of more rapid progression would be of clinical benefit, both for prognosis and to target therapy toward dogs with progressive disease. In dogs with DMVD it has been shown previously that left ventricular end-diastolic diameter, normalized for body weight (LVEDDN), and serum N-terminal pro-B type natriuretic peptide (NT-proBNP) are independently predictive of survival. In human patients, multimarker strategies for risk stratification in heart disease have been shown to be more effective than the use of single markers. It is likely that measurements of multiple biomarkers will be similarly superior in estimating prognosis in veterinary medicine.
High-sensitivity cardiac troponin I (hscTnI) previously has been measured in dogs with DMVD and is a marker of myocardial damage. Dogs with moderate and severe DMVD have significantly higher concentrations of plasma hscTnI than do healthy controls, and severely affected dogs have significantly higher concentrations than do mildly and moderately affected dogs. Chronic low-grade loss of troponins and myocardial cells may be reflected by increased hscTnI and may result in diminished myocardial contractile capacity, contributing to systolic failure. The utility of hscTnI as a prognostic marker has not been assessed previously in dogs. In human patients, increased hscTnI is associated with decreased survival times, even for concentrations <99th percentile for the normal population. We postulated that increasing myocardial damage, and hence circulating hscTnI, would be associated with decreased survival times in dogs with DMVD.
The aim of this prospective study was to evaluate the survival of dogs with DMVD managed in first opinion practice in order to identify prognostic indicators. Clinical characteristics, treatment with cardiac medications, echocardiographic measurements, and circulating markers were investigated. We aimed to investigate whether hscTnI was associated with survival in dogs with DMVD and, in addition, whether assessment of both hscTnI and NT-proBNP has superior predictive potential to assessing either marker alone. We also aimed to determine whether measurement of the rates of change of serum hscTnI and NT-proBNP over time could discriminate between cardiac and noncardiac causes of death in dogs with DMVD.
Materials and Methods
The study was approved by the Royal Veterinary College ethical committee, and informed owner consent was obtained. Two hundred and two dogs with evidence of DMVD and 30 age-matched healthy controls were prospectively and consecutively recruited from 2 London-based first opinion practices between December 2004 and October 2010. Dogs were referred to the study by the first opinion veterinarians after detection of a murmur consistent with mitral regurgitation at any stage in the natural history of the disease. Dogs with any other cardiac disease or clinically relevant organ-related or systemic disease were not enrolled. Dogs developing concurrent disease after enrollment and dogs receiving medication for heart failure were included in the study. For each dog, the history was taken followed by physical examination, blood sampling, and electrocardiographic and echocardiographic examination. All examinations were performed in a quiet room without sedation. Radiography was not performed routinely as part of the study protocol, but was recommended to the primary veterinarians if clinically indicated.
Blood was collected by jugular venipuncture into 5 mL serum gel tubes. Samples were chilled at 4°C for up to 6 hours before separation by centrifugation. Serum was stored at −80°C for batched analysis. Serum aliquots were transported to a commercial laboratory on dry ice for analysis.
Before analysis, the frozen serum was allowed to thaw slowly at room temperature. Concentrations of hscTnI were measured by an enzyme-linked immunosorbent assay (ELISA)1 according to the manufacturer's instructions. The use of this assay has been reported previously for canine samples. Concentrations of NT-proBNP were measured by a previously validated canine NT-proBNP ELISA2 according to the manufacturer's instructions. All measurements were performed by the same version of the assay. Serum urea and creatinine were measured by a commercial laboratory.2
Electrocardiography was performed in right lateral recumbency. Heart rate was measured from a 60-second recording of lead II.
Echocardiography was performed to confirm the diagnosis of DMVD and to exclude the presence of other cardiac diseases. Dogs in which no murmur was detected on auscultation did not undergo echocardiography and were assigned to the healthy control group. Echocardiographic examinations were performed by 1 of 2 board-certified cardiologists. Dogs were placed in right and then left lateral recumbency on an ultrasound examination table. The echocardiographic examination was performed by an ultrasound unit4 equipped with 2–4 MHz and 3–7 MHz phased array transducers and ECG monitoring. Standard imaging planes were digitally stored. Assessment of mitral valve structures was performed from the right parasternal long-axis view and the left apical 4-chamber view. Dogs continued to be examined in the same manner at 6-month intervals until they died or were euthanized, they were lost to follow-up, or the end of the study period was reached.
Diagnosis of DMVD was on the basis of characteristic abnormalities of the valve leaflets (thickening, prolapse, or both) and evidence of regurgitant flow across the valve detected by Doppler. The left atrial to aortic root ratio (LA/Ao) was measured from the right parasternal short axis view. M-mode measurements of the left ventricle were obtained by standard techniques. M-mode measurements were used to derive left ventricular end-systolic diameter (LVESD), left ventricular end-diastolic diameter (LVEDD), and left ventricular free wall thickness in systole diastole (LVFWd). The LVEDD/LVFWd ratio was calculated from these measurements. LVESD was normalized for body weight (LVESDN) by the formula: LVESD/(body weight [kg])0.315. LVEDD was normalized for body weight (LVEDDN) by the formula: LVEDD/(body weight [kg])0.294.
Survival times were calculated from the date of enrollment to the date of death or euthanasia. The latter was obtained either from the records of the primary veterinarians or by contacting the owners by telephone to request follow-up information. Dogs were censored in the survival analysis if they were alive at the end of the study period (April 2011) or if they had been lost to follow-up. Dogs were categorized as lost to follow-up if they had not been presented to the clinic for 9 months or if their owners could not be contacted by telephone. All clinical and echocardiographic measurements used in the survival analysis were made on the date of enrollment. Dogs that died were assigned to groups (cardiac and noncardiac death) according to the reason for death or euthanasia stated by the primary veterinarian. Dogs that died suddenly were assigned to the cardiac death group. If the reason for euthanasia was unspecified, but the dogs were known to have previously had mild DMVD and no clinical signs, they were assigned to the noncardiac death group.
Statistical analyses were performed by commercially available software.5 A value of P < .05 was considered significant. Values of hscTnI below the limit of detection of the assay (0.01 ng/mL) were assigned values of 0.01 ng/mL. Values of NT-proBNP below the lower (50 pmol/L) or above the upper (3000 pmol/L) limits of detection were assigned values of 50 pmol/L and 3000 pmol/L, respectively.
Data were assessed graphically for normality. Medians (interquartile ranges) were used to provide descriptive statistics for continuous variables. All statistical analyses were performed by means of measurements obtained at the baseline visit, corresponding to enrollment in the study. The exceptions were the repeated measures linear model analyses, which also used measurements obtained from subsequent visits at 6-month intervals. Comparisons of continuous variables between groups were performed by Mann-Whitney U-tests. Comparisons of proportions between groups were performed by Chi-squared tests. Linear regression analysis was performed to assess the effects of storage time on serum NT-proBNP, hscTnI, urea, and creatinine. The assumption of linearity was assessed graphically by plotting storage time against the dependent variable (serum NT-proBNP, hscTnI, creatinine, or urea). The distributions of residuals were graphically tested for normality by plotting the predicted values against the residuals.
For the survival analysis, data for the following potential risk factors were entered into univariate Cox proportional hazards models to determine whether they were associated with survival: serum hscTnI, NT-proBNP, urea and creatinine; LA/Ao ratio, LVEDDN, LVEDD/LVFWd ratio, LVESDN, heart rate; treatment, with angiotensin-converting enzyme (ACE) inhibitors, furosemide or pimobendan; and breed (Cavalier King Charles Spaniel [CKCS]: yes or no), sex, and age. The hazard ratio (HR) and 95% confidence interval (CI) for the HR are reported for the model. The relationships between the continuous predictors and outcome of interest in the Cox proportional hazards models were examined by categorizing continuous data into quartiles and computing the log odds of outcome against the category means. If a linear trend was identified, risk factors were treated as continuous variables in subsequent analyses. If no linear trend was identified receiver operating characteristic (ROC) curve analysis was used to determine cut-off values to assign variables to categories for subsequent analysis. Sensitivity was plotted against 1-specificity for identification of dogs that died (all-cause mortality) and the value selected, which provided the highest value for the mean of sensitivity and specificity. Variables with P < .2 in the univariate analysis were included in a manual backward-selection, stepwise multivariable Cox proportional hazards analysis for survival (all-cause mortality). Multivariable analysis for cardiac and noncardiac mortality was performed by selection of the x variables (where x = the smaller of the 2 numbers of end-points/10) of greatest interest. The variables of greatest interest were chosen on the basis of previously published risk factors and the hypotheses of the study. These variables were entered into a single step, multivariable Cox proportional hazards analysis for survival. Continuous variables were transformed into dichotomous variables (above and below median) and first-order interactions were assessed by entering the product terms of the dichotomous variables into Cox proportional hazards models.
The log rank test was used to compare survival between dogs with increases in both hscTnI and NT-proBNP and those with one or neither variable increased. ROC curve analysis was used to determine cut-off values for hscTnI and NT-proBNP. Posthoc comparisons between groups were made by the log rank test with Bonferroni corrections for multiple comparisons. Cox proportional hazards analysis was used to calculate HRs.
Repeated measures linear models were used to study the effect of status (cardiac death, noncardiac death or alive), time, and the interaction of status and time on NT-proBNP and hscTnI over a maximum of 17 months. A first-order autoregressive (co)variance structure was assumed between residuals of the same subject (dog) in the model. Dogs that had been monitored for <17 months, because of death, loss to follow up or recruitment <17 months before the end of the study period were included in the model. The coefficient (b) and the 95% CI for the coefficient are reported for the model. The assumption of normality of the model residuals was assessed graphically.
Thirty healthy control dogs (14 males and 16 females) with a median age of 9.0 years (6.0, 11.0) and a median body weight of 11.0 kg (7.2, 13.8) were included in the study. Cross-breeds were most frequently represented (n = 9), followed by CKCS (n = 7), 2 each of Border Collie, Miniature Poodle, and Yorkshire Terrier and 1 each of 8 other breeds. Two hundred and two dogs with DMVD (116 males and 86 females) with a median age of 10.0 years (8.0, 11.6) and a median body weight of 10.4 kg (7.8, 13.9) were included in the study. The most frequently represented breed was the CKCS (n = 77), followed by cross breeds (n = 52), Jack Russell Terrier (n = 14), Yorkshire Terrier (n = 11), 4 each of Bichon Frise, Chihuahua, Miniature Poodle, and Shih Tzu, 3 each of Lurcher, Pomeranian, and Whippet, 2 each of Border Collie, Chinese Crested, Maltese Terrier, and Norfolk Terrier, and 1 each of 15 other breeds. Summary statistics are shown in Table 1. Detectable concentrations of hscTnI were measured in 226/232 dogs. Detectable concentrations of NT-proBNP were measured in 230/232 dogs.
|Characteristic||Dogs with DMVD||Healthy Control Dogs||Significance|
|Sex (male/female) [% male]||116/86 [57.4%]||14/16 [46.7%]||.362|
|CKCS (yes/no) (% yes)||77/125 [38.1%]||7/23 [23.3%]||.171|
|Age (years)||10.0 (8.0,11.6)||9.0 (6.0,11.0)||.059|
|Body weight (kg)||10.4 (7.8,13.9)||11.0 (7.2,13.8)||.760|
|LVEDD/LVFWd ratio||3.92 (3.41,4.59)||N/A|
|LA/Ao ratio||1.27 (1.15,1.46)||N/A|
|Heart rate||126.0 (107.5,139.0)||N/A|
|Treatment with ACE inhibitors (yes/no) [% yes]||42/160 [20.8%]||0/30 [0%]|
|Treatment with furosemide (yes/no) [% yes]||28/174 [13.9%]||0/30 [0%]|
|Treatment with pimobendan (yes/no) [% yes]||20/182 [9.9%]||0/30 [0%]|
|Creatinine (mg/dL)||0.82 (0.70,1.00)||0.90 (0.70,1.04)||.213|
|Urea (mg/dL)||16.2 (12.9,21.9)||17.5 (13.1,21.1)||.973|
|hscTnI (NG/ML)||0.03 (0.02,0.06)||0.03 (0.02,0.03)||.020|
|NT-proBNP (pmol/L)||575.0 (348.0,1067.5)||324.5 (166.8,530.0)||<.001|
There was no evidence for an effect of storage time on serum NT-proBNP (P = .587), hscTnI (P = .281), or creatinine (P = .157). A modest effect of storage on serum urea (P = .026, B = 0.001 [0.000–0.003]) was detected.
Ninety-nine dogs died or were euthanized during the study period, 43 of which were considered to have died or been euthanized as a direct result of their cardiac disease. Seventeen dogs were lost to follow-up (median time to loss to follow up, 189 days; 95% CI, 0, 480.8) and 86 dogs remained alive at the end of the study period. Overall, median follow-up time was 407 days.
The relationship between hscTnI and survival was found to be nonlinear. A ROC curve was constructed to determine a cut-off value so that hscTnI could be entered into Cox proportional hazards models as a categorical covariate. The area under the curve was 0.629. The cut-off value for hscTnI selected for use in survival analysis was 0.025 ng/mL, which had a sensitivity of 82% and a specificity of 43%.
Univariate analysis for death or euthanasia caused by any cause indicated that age, serum hscTnI, NT-proBNP, urea and creatinine, LA/Ao ratio, LVEDDN, LVEDD/LVFWd ratio, LVESDN, heart rate, treatment with ACE inhibitors, and treatment with furosemide were negatively associated with survival at the 20% level (P < .2) (Table 2). LVEDDN, LVESDN, and LA/Ao ratio were multiplied by 10 and NT-proBNP divided by 100 in order to calculate clinically meaningful HRs. Serum urea and creatinine were highly correlated, and so only urea was considered in the multivariable model. Both LVEDD/LVFWd ratio and LVESDN were highly correlated with LVEDDN, and so only LVEDDN was considered in the multivariable model. The following factors at enrollment were excluded from further analysis because they were not significantly associated with survival (P < .2): sex (male/female), breed (CKCS: yes/no), body weight (kg), and treatment with pimobendan.
|Variable||HR [95% CI for HR]||Probability|
|Sex (male/female)||0.90 [0.61–1.34]||.602|
|CKCS (yes/no)||1.06 [0.71–1.59]||.787|
|Age (years)||1.23 [1.15–1.32]||<.001|
|Body weight (kg)||1.02 [0.98–1.05]||.357|
|LVEDDN (×10)||1.26 [1.17–1.37]||<.001|
|LVEDD/LVFWd ratio||1.25 [1.11–1.41]||<.001|
|LVESDN (×10)||1.12 [1.00–1.26]||.050|
|LA/Ao ratio (×10)||1.20 [1.12–1.28]||<.001|
|Heart rate||1.02 [1.01–1.03]||.001|
|Treatment with ACE inhibitors (yes/no)||1.57 [0.99–2.49]||.058|
|Treatment with furosemide (yes/no)||1.88 [1.12–3.14]||.017|
|Treatment with pimobendan (yes/no)||1.08 [0.59–1.98]||.799|
|Creatinine (mg/dL)||1.98 [0.86–4.58]||.111|
|Urea (mg/dL)||1.02 [1.01–1.04]||<.001|
|hscTnI > 0.025 (ng/mL)||3.23 [1.98–5.27]||<.001|
|NT-proBNP (pmol/L) (/100)||1.06 [1.04–1.08]||<.001|
Multivariable analysis indicated that LVEDDN, age, heart rate, and serum hscTnI concentration were independently associated with survival. An interaction between LVEDDN and age was identified (P < .05). The product term for this interaction was not significant when entered into the model. In the final multivariable model, survival was negatively associated with age (P < .001), serum hscTnI concentration (P = .024), heart rate (P = .001), and LVEDDN (P < .001) (Table 3). There was no significant association between serum concentrations of NT-proBNP, urea, LA/Ao ratio, treatment with ACE inhibitors or treatment with furosemide, and survival. Graphical assessment of categorical variables by plotting survival time against the log cumulative hazard confirmed that the assumption of proportional hazards was satisfied.
|HR||95% CI for HR||Probability|
|hscTnI > 0.025 (ng/mL)||1.82||1.08–3.06||.024|
Multivariable analysis was repeated, excluding echocardiographic variables. Serum hscTnI and NT-proBNP, heart rate, and age were found to be independently associated with survival. No significant interactions were identified among these variables. In the final multivariable model, survival was negatively associated with age (P < .001), heart rate (P = .001), serum NT-proBNP (P = .010), and hscTnI (P = .016) (Table 4).
|HR [95% CI for HR)||Probability|
|Age (years)||1.16 [1.07–1.25]||<.001|
|NT-proBNP (pmol/L) (/100)||1.03 [1.01–1.05]||.010|
|hscTnI > 0.025 (ng/mL)||1.91 [1.13–3.23]||.016|
|Heart rate||1.02 [1.01–1.03]||.001|
Univariate analysis for death or euthanasia as a result of cardiac disease indicated that serum hscTnI, NT-proBNP and urea, LVEDDN, LVEDD/LVFWd ratio, LVESDN, LA/Ao ratio, heart rate, age, breed (CKCS: yes/no), treatment with ACE inhibitors, and treatment with furosemide were associated with survival at the 20% level (P < .2) (Table 5). Serum creatinine, sex, body weight, and treatment with pimobendan were not associated with survival. The 4 variables of greatest interest (serum hscTnI, NT-proBNP, LVEDDN, and age) were carried forward for consideration in the multivariable model (number of events = 43).
|Variable||Cardiac Mortality||Noncardiac Mortality|
|HR [95% CI for HR]||Probability||HR [95% CI for HR]||Probability|
|Sex (male/female)||0.91 [0.49–1.66]||.749||1.30 [0.77–2.20]||.332|
|CKCS (yes/no)||2.07 [1.13–3.81]||.019||0.59 [0.32–1.07]||.080|
|Age (years)||1.11 [1.00–1.24]||.043||1.34 [1.22–1.48]||<.001|
|Body weight (kg)||0.98 [0.93–1.04]||.565||1.03 [1.00–1.07]||.089|
|LVEDDN (×10)||1.52 [1.35–1.71]||<.001||1.07 [0.95–1.19]||.274|
|LVEDD/LVFWd ratio||1.35 [1.19–1.53]||<.001||1.03 [0.77–1.38]||.851|
|LVESDN (×10)||1.31 [1.10–1.55]||.002||0.99 [0.85–1.16]||.993|
|LA/Ao ratio (×10)||1.36 [1.23–1.45]||<.001||1.05 [0.94–1.17]||.375|
|Heart rate||1.02 [1.01–1.04]||<.001||1.01 [1.00–1.02]||.240|
|Treatment with ACE inhibitors (yes/no)||2.04 [1.05–3.94]||.034||1.24 [0.64–2.41]||.527|
|Treatment with furosemide (yes/no)||2.47 [1.21–5.05]||.013||1.44 [0.68–3.07]||.344|
|Treatment with pimobendan (yes/no)||1.61 [0.71–3.64]||.215||0.74 [0.29–1.86]||.519|
|Creatinine (mg/dL)||1.74 [0.49–6.20]||.395||2.20 [0.72–6.71]||.167|
|Urea (mg/dL)||1.02 [1.01–1.04]||.008||1.02 [1.01–1.04]||.015|
|hscTnI > 0.025 (ng/mL)||3.18 [1.50–6.74]||.002||3.27 [1.72–6.23]||<.001|
|NT-proBNP (pmol/L) (/100)||1.09 [1.08–1.13]||<.001||1.03 [1.00–1.06]||.084|
Multivariable analysis indicated that serum NT-proBNP and LVEDDN were independently associated with survival. No first order interactions were identified. In the final multivariable model, survival was negatively associated with serum NT-proBNP (P < .001) and LVEDDN (P < .001) (Table 6).
|HR||95% CI for HR||Probability|
Multivariable analysis was repeated, excluding echocardiographic variables. Serum NT-proBNP was found to be independently associated with survival (P < .001) (Table 7).
|HR||95% CI for HR||Probability|
Univariate analysis for death or euthanasia as a result of noncardiac disease indicated that serum hscTnI, NT-proBNP, creatinine and urea, CKCS (yes/no), body weight, and age were associated with survival at the 20% level (P < .2) (Table 5). The 4 variables of greatest interest (serum hscTnI, NT-proBNP, LVEDDN, and age) were carried forward for consideration in the multivariable model (number of events = 56). Sex, LVEDDN, LVEDD/LVFWd ratio, LVESDN, LA/Ao, heart rate, treatment with ACE inhibitors (yes/no), treatment with furosemide (yes/no), and treatment with pimobendan (yes/no) were not associated with survival.
Multivariable analysis suggested that serum hscTnI and age were independently associated with survival. No first-order interactions were identified. In the final multivariable model, survival was negatively associated with age (P < .001). A trend for a negative association between survival time and serum hscTnI was detected (P = .054) (Table 8). Graphical assessment of categorical variables by plotting survival time against the log cumulative hazard confirmed that the assumption of proportional hazards was satisfied. Multivariable analysis was not repeated excluding echocardiographic variables because no echocardiographic variable was significant in the final model. Noncardiac causes of death are summarized in Table 9.
|HR||95% CI for HR||Probability|
|hscTnI > 0.025 (ng/mL)||1.99||0.99–3.85||.054|
|Cause of Death or Reason for Euthanasia||Frequency||Percentage of Total Noncardiac Deaths|
|Chronic kidney disease||3||5.4|
|Multiple health problems||17||30.4|
|Idiopathic vestibular disease||1||1.8|
An ROC curve was constructed to determine a cut-off value for NT-proBNP and survival. The cut-off value for NT-proBNP selected as a predictor of all-cause mortality was 524 pmol/L, which had a sensitivity of 69% and a specificity of 55%. Dogs were assigned to 1 of 4 categories according to their combined serum hscTnI and NT-proBNP measurements (Table 10). Dogs with high hscTnI and low NT-proBNP or vice versa were combined to form a “medium” category (Table 11). Kaplan-Meier survival curves were constructed to compare survival times among these 3 groups (Fig 1). There were significant differences in survival time between low and medium categories (P = .003), between low and high categories (P < .003), and between medium and high categories (P = .006) (Table 11).
|Low||45||21||n = 66|
|High||47||89||n = 136|
|n = 92||n = 110||Total = 202|
|Category||Serum hscTnI (ng/mL) and NT-proBNP (pmol/L)||Median Survival (days) [95% CI for median]||Number of Dogs||Number of Deaths||Hazard Ratio (HR) [95% CI for HR] (relative to low group)|
|Low||hscTnI < 0.025 and NT-proBNP < 524||1503 [1257.5–1748.5]||45||12 (26.7%)|
hscTnI < 0.025 and NT-proBNP > 524 OR
hscTnI > 0.025 and NT-proBNP < 524
|981 [435.8–1526.2]||68||31 (45.6%)|
(P = .003)
|High||hscTnI > 0.025 and NT-proBNP > 524||475 [409.7–540.3]||89||56 (62.9%)|
(P < .003)
Serum NT-proBNP and hscTnI were not normally distributed and were logarithmically transformed before construction of repeated measures models. A repeated measures linear model demonstrated a significant interaction between time (months) and status (alive, cardiac death or noncardiac death) on serum NT-proBNP concentrations (P = .036), and the rate of increase of NT-proBNP was significantly higher for dogs in the cardiac death group when compared with the noncardiac death and alive groups (b = 0.023; 95% CI, 0.014, 0.032 for cardiac death, b = 0.011; 95% CI, 0.004, 0.019 for noncardiac death, and b = 0.009; 95% CI, 0.003, 0.015 for alive) (Fig 2). This is equivalent to a doubling of serum NT-proBNP concentration every 13.1 months in dogs that died of cardiac disease. There was a significant interaction between time and cause of death (alive, cardiac death, or noncardiac death) on serum hscTnI concentration (P < .001), and the rate of increase of hscTnI was significantly higher for dogs in the cardiac death group when compared with the noncardiac death and alive groups (b = 0.038; 95% CI, 0.028, 0.048 for cardiac death, b = 0.010; 95% CI, 0.002, 0.019 for noncardiac death, and b = 0.003; 95% CI, −0.003, 0.001 for alive) (Fig 3). This is equivalent to a doubling of serum hscTnI concentration every 7.9 months in dogs that died because of cardiac disease.
This study demonstrates the prognostic utility of hscTnI in dogs. The results of this study indicate that LVEDDN, age, heart rate, and hscTnI are significantly and independently associated with decreased survival time for all-cause mortality in dogs with naturally occurring DMVD.
cTnI is a marker of myocardial damage. In DMVD, myocardial damage is likely to be a late process in the pathophysiology of the disease as ventricular remodeling becomes advanced. This is consistent with the findings of the repeated measures model, in which the doubling interval is short and concentrations appear to diverge from the noncardiac death group only in the last 6 months of life (Fig 3). hscTnI previously has been shown to be significantly higher in dogs with severe DMVD than in those with mild or moderate disease and a negative correlation between hscTnI and survival was predicted. This relationship was not found to be linear, such that dogs with serum hscTnI concentrations > 0.025 ng/mL had shorter survival times regardless of the exact concentration. An interaction between LVEDDN and serum hscTnI concentration was identified, in agreement with previous demonstration of a correlation between hscTnI and percentage increase in LVEDD. In the previous study, the 25th percentile of hscTnI measurements in the severely affected group was 0.031 ng/mL such that the majority of dogs in this group would have been classified as being at increased risk of death or euthanasia according to the results of the present study. In contrast, in the previous study, the 75th percentiles of hscTnI concentrations in the mildly and moderately affected groups were 0.024 and 0.029 ng/mL, respectively, such that the majority of animals in these groups would not have been classified as being at increased risk of death or euthanasia. In the present study, the group of healthy control dogs had apparently higher circulating hscTnI concentrations (0.03 ng/mL; 0.02, 0.03) than those in the previous study (0.001 ng/mL; 0.001, 0.004). However, the normal group in the previous study had a median age of 4.5 years (interquartile range, 3.7–7.5), which is considerably younger than those in the current study (9.0 years; 6.0, 11.0). Circulating hscTnI concentrations are positively associated with age, which might explain this difference. It is notable that 14 dogs in the healthy control group had serum hscTnI concentrations of 0.03 ng/mL, which is higher than the cut-off derived to indicate a higher risk of all-cause mortality (0.025 ng/mL). These dogs tended to be older than other dogs in the healthy control group that had serum hscTnI measurements < 0.03 ng/mL (P = .088), further supporting the theory that an age-related increase in hscTnI contributed to this observation. In both dogs and human patients, circulating troponins are known to be increased in a variety of noncardiac diseases,[13-16] and so it is possible that some of these apparently healthy dogs were suffering from subclinical disease. Indeed, there was a trend toward an independent association between serum hscTnI and survival in dogs that died or were euthanized because of noncardiac causes, suggesting that this marker does not specifically predict cardiac outcomes. Serum hscTnI was not found to be independently associated with survival time in dogs that died or were euthanized because of cardiac disease, which might reflect this lack of specificity. However, alternative explanations include a lower power of the analysis to show a difference because of the lower number of observations contributing to this subanalysis and that the median follow-up time (407 days) was too long for a baseline measurement to detect the divergence in serum hscTnI measurements in the cardiac mortality group from the noncardiac mortality group, because this does not occur until the last 6 months of life.
Before development of the high-sensitivity assay for cTnI, an assay with a lower limit of detection of 0.1 ng/mL was available. This assay therefore would enable measurement of cTnI in only the most severely affected dogs with DMVD. A detectable concentration of cTnI by this assay is likely to be associated with decreased survival times as suggested by the results of a small pilot study.
Serum NT-proBNP is a marker of increased cardiac wall stress. Although its concentration also increases with severity in DMVD, information provided by this marker may be complementary to that of hscTnI. NT-proBNP was not found to be independently associated with survival time in the multivariable model for all-cause mortality when echocardiographic variables were included. However, the findings of a multivariable model are only valid if all variables are measured. Given the frequency with which DMVD is diagnosed and the technical challenges associated with obtaining reliable echocardiographic measurements, circulating biomarkers may be of greater utility for veterinarians in general practice. When echocardiographic variables were excluded from the multivariable model, both NT-proBNP and hscTnI were found to be independently predictive of survival, supporting this assertion. This implies that the 2 markers provide complementary and different predictive information about the patient rather than the value of 1 marker subsuming the value of the other. Indeed, in the multivariable analysis for cardiac death, NT-proBNP (but not hscTnI) was independently associated with survival time.
LVEDDN, heart rate and age previously have been reported to be associated with decreased survival time in this disease, although the present study is the first in which heart rate and age have been shown to be independently associated with survival. hscTnI concentrations are known to be associated with age. However, in the present study hscTnI was found to be independently associated with survival when age was included in a multivariable model. This demonstrates that the relationship between hscTnI and survival is not simply related to the age-dependent increase in hscTnI concentration. The previously reported HR for LVEDDN was 1.20 (95% CI, 1.04–1.37), which is similar to that of the present study and within the 95% CI of HR prediction. This is not surprising, because information from 100 animals was used in both studies. However, concentrations of NT-proBNP in the previous study were determined by the original version of the canine-specific ELISA,6 whereas in the present study all concentrations were determined by the current version of the assay.3 The similarity in HRs suggests that the current version of the assay provides performance that is comparable to outcome prediction to that of the original version.
In human patients with heart disease, measurement of multiple biomarkers has been shown to provide more accurate risk stratification compared with individual markers. In human patients, the combination of high-sensitivity cardiac troponin T with BNP has been shown to be of greater prognostic value than BNP alone. In the present study, combining serum hscTnI and NT-proBNP measurements more accurately identified those dogs with a shorter survival time than did either measurement alone.
The cut-off of NT-proBNP used in this dual marker approach is lower than the previously suggested concentration of 740 pmol/L. A higher cut-off value has the advantage of increased specificity, decreasing the risk of a false positive result. If only NT-proBNP is measured, therefore, the higher cut-off should be used. Nevertheless, the decrease in specificity is compensated for by the combination of measurements of NT-proBNP and hscTnI providing improved overall risk stratification.
ROC curves were used to derive cut-off values of serum hscTnI and NT-proBNP for use in survival analysis rather than to produce a cut-off predictive of death. For this reason, the relatively low sensitivities and specificities are tolerable.
A limitation of the study was that serum samples were stored at −80°C for up to 62 months before analysis. However, linear regression did not demonstrate any effect of storage on serum NT-proBNP, cTnI, or creatinine and only a modest effect of storage on serum urea. It is unlikely, therefore, that storage at −80°C significantly influenced the findings of the present study. NT-proBNP concentrations in canine blood samples are known to decrease over time after collection because of protease degradation. Current recommendations therefore include mixing plasma with a protease inhibitor in a commercially available tube.7 In this study, NT-proBNP was measured in serum samples that had not been mixed with a protease inhibitor, because sample collection began before these tubes were available. Another limitation of the study was that dogs in the healthy control group did not undergo echocardiography. However, none of these dogs were found to have an audible cardiac murmur on careful auscultation by a board-certified cardiologist, and so it is unlikely that any clinically relevant cardiac pathology was present in any of these dogs. This study identified risk factors associated with decreased survival times in dogs with naturally occurring mitral valve disease. These risk factors should be validated prospectively in a separate population to determine their predictive value. Postmortem examinations were not performed to confirm the cause of death or reason for euthanasia. The primary veterinarian had not recorded a specific cause of death or reason for euthanasia in the case of 5 (8.9%) of the animals assigned to the noncardiac mortality group. A proportion of these deaths may have been a direct result of cardiac disease, leading to inaccurate classification. Echocardiographic measurements, but not serum NT-proBNP or hscTnI concentrations, were available to the primary veterinarians at the time of euthanasia and may have played a part in the decision-making process. Older dogs commonly die because of neoplasia or neurological disease, which was reflected in the causes of death seen in the noncardiac mortality group (Table 9).
Repeated measures models indicated that both NT-proBNP and hscTnI increased more rapidly in dogs that died or were euthanized because of cardiac disease than in those that died because of noncardiac disease. However, visual inspection of Figure 3 suggests that the increase in hscTnI in dogs that died because of cardiac disease occurs late in the disease, whereas Figure 2 suggests that the curve of NT-proBNP measurements in dogs that die because of cardiac disease diverges earlier from that of those that die of noncardiac disease. Specific investigation of whether NT-proBNP identifies dogs that die because of cardiac disease earlier in the course of the disease was not performed in this study. Nevertheless, serial measurements of NT-proBNP and hscTnI made every 6 months to monitor the rates of change of these markers may provide further valuable prognostic information.
In conclusion, LVEDDN, circulating hscTnI concentration, heart rate, and age are independently associated with survival (all-cause mortality) in dogs with DMVD. Measurement of NT-proBNP and hscTnI is prognostically superior to measuring either in isolation. However, when other variables, including echocardiographic variables, are known, the independent prognostic value of including both of these variables is lost. LVEDDN and NT-proBNP are independently associated with survival (cardiac mortality), whereas age and hscTnI are independently associated with survival (noncardiac mortality). Hence, both single and serial measurements of these markers have the potential to provide valuable prognostic information in clinical cases of DMVD. Additional studies are needed to validate these risk factors prospectively.
The work was supported by a grant from the Petplan Charitable Trust. Laboratory work was funded by IDEXX Ltd.
This information has not been previously presented.
Access Systems AccuTnI Assay, Beckman Coulter Inc, Fullerton, CA
VETSIGN Canine Cardiopet Nt-proBNP, IDEXX Laboratories, Westbrook, ME
IDEXX Laboratories, Wetherby, UK
Acuson Cypress, Siemens Medical Solutions, Siemens House, Oldbury, Bracknell, UK
PASW version 18.0 for Windows, SPSS Inc, Chicago, IL
Vetsign™ Canine CardioScreen Nt-proBNP, Guildhay Limited, Biomedica, Guildford, Surrey, UK
Cardiopet proBNP transport tube, IDEXX Laboratories, Westbrook, ME
- 2Association of plasma N-terminal pro-B-type natriuretic peptide concentration with mitral regurgitation severity and outcome in dogs with asymptomatic degenerative mitral valve disease. J Vet Intern Med 2009;23:984–994., , , et al.
- 11Veterinary Epidemiologic Research. Charlottetown, PE: AVC Inc; 2003:706., , .