Author + information
- Received May 25, 2001
- Revision received September 10, 2001
- Accepted October 18, 2001
- Published online January 16, 2002.
- ↵*Reprint requests and correspondence:
Dr. G. Michael Felker, Duke University Medical Center, Box 31185, Durham, North Carolina, USA 27710.
Objectives We sought to evaluate the association between the extent of coronary artery disease (CAD) and survival in patients with symptomatic heart failure (HF) and to create the most prognostically powerful clinical definition of ischemic cardiomyopathy.
Background An ischemic etiology of HF is known to be a predictor of adverse outcome; however, there is no uniform definition for ischemic cardiomyopathy.
Methods We assessed the clinical history and coronary anatomy of patients with symptomatic HF and ejection fraction ≤40% undergoing diagnostic coronary angiography between 1986 and 1999 (n = 1,921). Five classification schemes were tested to identify the most prognostically powerful method for defining the extent of CAD and to develop the best definition of ischemic cardiomyopathy for prognostic purposes.
Results A more extensive CAD was independently associated with shorter survival. When the various classification schemes were compared, a modified number-of-diseased-vessels classification, in which patients with single-vessel disease and no prior history of revascularization or myocardial infarction (MI) were classified as nonischemic, provided the most prognostic power. A definition of ischemic cardiomyopathy that incorporated this definition had more prognostic power than the traditional definition.
Conclusions Angiographically diagnosed ischemic HF is associated with shorter survival than nonischemic HF. A more extensive CAD is independently associated with shorter survival, and patients with single-vessel disease and no history of MI or revascularization should be classified as nonischemic for prognostic purposes. Standardization of the definition of ischemic cardiomyopathy will be useful in the conduct and interpretation of clinical research in HF.
Accurate prognostication is an important goal in the management of patients with chronic heart failure (HF), particularly given the limited availability of some therapies such as cardiac transplantation. Among the most important clinical determinants of prognosis in HF is the etiology of left ventricular (LV) dysfunction. The last several decades have seen a shift in the most common underlying etiology of HF from hypertension and valvular disease to ischemic heart disease (1). Ischemic etiology has been shown to be independently associated with worse long-term outcome in patients with LV dysfunction in a variety of studies (2–5). The etiology of HF also impacts the decision to pursue revascularization and may effect the response to some pharmacologic therapies (6). Finally, recent data suggest that the mechanism of sudden death may differ between ischemic and nonischemic HF patients, further emphasizing the potential importance of accurate differentiation between etiologies (7).
Clinically, patients are classified as having HF of ischemic or nonischemic etiology based on a history of myocardial infarction (MI) or based on objective evidence of coronary artery disease (CAD) such as angiography or functional testing. As a practical matter, however, the determination of etiology in an individual patient may be difficult because patients with HF and no CAD may have typical angina or regional wall motion abnormalities on echocardiography (8), whereas patients with severe CAD may have no symptoms of angina or history of MI. This has lead to the recommendation for either noninvasive functional testing or coronary angiography in assessing patients with systolic dysfunction, depending on the clinical presentation and risk factor profile (9). Even with objective information such as angiography, the appropriate classification for a given patient is not always clear. Although HF etiology represents an important variable in the design and interpretation of clinical trials (10), the means by which etiology was determined are often not clear from published reports, and criteria may differ between studies (6). This may be of particular importance in the design and interpretation of clinical trials where etiology of HF is an important covariate and frequently defines subgroups of interest. The lack of a standardized definition of ischemic etiology may lead to unnecessary variability in defining subgroups and inconsistency between different studies. We sought to use data from the Duke Databank for Cardiovascular Diseases to develop the most prognostically powerful clinical definition of ischemic etiology.
Patient data was obtained from the Duke Databank for Cardiovascular Diseases, an ongoing databank of all patients undergoing diagnostic cardiac catheterization at Duke University Medical Center. Patients were included in the study population if they had an LV ejection fraction (EF) of ≤40% and a history of symptomatic HF (New York Heart Association [NYHA] functional class II or greater). Patients were excluded from analysis if they had experienced MI within 30 days of the index catheterization, had primary valvular heart disease (defined as severe aortic or mitral insufficiency or severe stenosis of any heart valve) or congenital heart disease. Baseline clinical variables for each patient were stored in the Duke Databank using methods previously described (11). Follow-up on patients with significant coronary disease or revascularization procedures was obtained through self-administered questionnaires, with telephone follow-up to nonresponders. Patients not contacted through this mechanism had vital status determined through a search of the National Death Index (12).
Data from the index catheterization was prospectively collected. Coronary stenoses were graded by visual consensus of at least two experienced observers. In order to determine the best method for quantifying the angiographic data, several classification schemes were evaluated for their prognostic power. For the traditional binary classification (ischemic vs. nonischemic), an ischemic etiology of HF was defined as the presence of any epicardial coronary vessels with ≥75% stenosis or any history of MI or coronary revascularization (either percutaneous transluminal coronary angioplasty or coronary artery bypass grafting). The number-of-diseased-vessels classification was defined as the number of vessels with ≥75% stenosis (0, 1, 2 or 3). The CAD prognostic index (see Appendix), which has been previously described and validated in an overlapping HF population (2), considers the number of diseased vessels, the number of vessels with >95% stenosis and the presence of left anterior descending (LAD) or left main coronary disease. In addition, a modified binary definition of ischemic etiology was also evaluated, which reclassified those patients with single-vessel disease who did not have a history of MI or coronary revascularization into the nonischemic group. Patients with ≥75% stenosis of the left main coronary artery or proximal LAD were included in the ischemic group. This also led to a modified number-of-diseased-vessels classification, wherein patients with single-vessel disease were combined with those without any significant coronary lesions (0/1, 2, 3). These modified definitions attempted to define a group of patients with LV dysfunction “out of proportion” to their degree of CAD.
This study sought to develop the definition of ischemic etiology of HF that would provide the most power to distinguish the differing prognoses between ischemic and nonischemic patients. In so doing, we compared five classification schemes for assessing the extent of CAD: the traditional binary classification (ischemic vs. nonischemic), the number of diseased vessels (0, 1, 2, 3), the CAD prognostic index, the modified binary classification and the modified number-of-diseased-vessels classification (0/1, 2, 3).
Baseline characteristics are described with medians and interquartile ranges for continuous variables and percentages for discrete variables. Pearson chi-square tests were used for group comparisons of unordered categorical variables, and Kruskal-Wallis tests were used for continuous measures. Survival curves for various groups were constructed using the method of Kaplan and Meier, and comparisons were made using the log-rank test. A p value ≤0.05 was used to indicate statistical significance for all comparisons. The association between baseline characteristics and mortality was calculated using a Cox proportional hazards model. Increased risk associated with a given variable was described using hazard ratios, which gives the relative increase in risk associated with the presence of a given characteristic. Multivariable Cox proportional hazards analysis was used to adjust for baseline differences between groups. The overall prognostic power of a given model was assessed by comparing the overall log likelihood ratio chi-square value for each model. The best definition of ischemic cardiomyopathy was considered to be the definition that maximized the chi-square test statistic for the model. All statistical analysis was performed with SAS version 6.1 (SAS Institute, Cary, North Carolina).
A total of 1,921 patients met criteria for the study between January 1, 1986, and December 31, 1999. Median follow-up for the cohort was 3.1 years (interquartile range 1.0 to 6.2).
Baseline characteristics for the ischemic and nonischemic groups are provided in Table 1. Using the traditional definition of ischemic etiology, 1,304 patients (68%) were classified as ischemic and 617 (32%) as nonischemic. As expected, a number of baseline characteristics differed significantly between the ischemic and nonischemic cohorts, with ischemic patients more likely to be older, male, Caucasian, hypertensive, diabetic and smokers. Ejection fraction was slightly higher in the ischemic cohort than it was in the nonischemic cohort. Notably, over 40% of patients classified as nonischemic based on angiography reported a history of anginal chest pain.
Overall five-year survival for the study population was 49%. As expected, patients with an ischemic etiology of HF had shorter survival than those with nonischemic HF (five-year survival: 45% for the ischemic group vs. 62% for the nonischemic group, p < 0.001 by log-rank test). Unadjusted survival curves for the two groups are shown in Figure 1. Using unadjusted Cox proportional hazards analysis, a greater extent of CAD was significantly associated with shorter survival regardless of the classification scheme used. Unadjusted relationships between all tested variables and survival are shown in Table 2.
Variables that were significant in the unadjusted analysis or thought to have clinical relevance were included as candidates in a multivariable survival model. After adjustment for baseline characteristics, ischemic etiology remained a significant independent predictor of long-term survival. Other independent predictors of survival were increasing age, male gender, lower EF, increasing NYHA functional class, diabetes mellitus and valvular heart disease.
Comparing etiology classification schemes
In order to assess the most prognostically powerful method for describing ischemic etiology, several models were compared. Each model included the independent baseline predictors of survival (age, gender, EF, NYHA functional class, diabetes and valvular heart disease) while varying the etiology classification scheme. Etiology remained independently associated with outcome regardless of which classification scheme was used. The log likelihood ratio chi-squares for each overall model are provided in Table 3, and adjusted survival curves for each of the five methods of classifying ischemia are shown in Figure 2. The modified number-of-diseased-vessels model (zero to one-, two- or three-vessel disease) provided the best prognostic power overall. Among the binary classification models, the modified definition of ischemic etiology provided an increase in the prognostic power of the model (as measured by the overall model chi-square) over the traditional definition. Hazard ratios for each of the independent predictors of outcome are shown in Figure 3.
The association of extent of CAD in a large population with both symptomatic and asymptomatic LV dysfunction has been previous reported by Bart et al. (2). That study demonstrated that the extent of CAD provides improved prognostic power over a binary definition in patients with LV dysfunction, with or without the clinical syndrome of HF. The current study examines those relationships in a more clinically applicable population with symptomatic HF (NYHA functional class II or greater), evaluating a large cohort of such patients with up to 14 years of follow-up. In so doing, we sought to develop and assess a more prognostically accurate definition of ischemic etiology that could be used for classification of HF patients in population studies and clinical trials, where etiology of HF is an important variable with substantial prognostic and therapeutic implications.
Our study identified a population of patients who would have traditionally been classified as having an ischemic etiology of HF but who had a prognosis similar to those with nonischemic etiology. These patients, who had single-vessel disease but no history of revascularization or MI, represent patients with HF “out or proportion” to their degree of CAD. Of the 1,304 patients initially classified as ischemic, 140 (11%) were reclassified into the nonischemic category based on this definition. By reclassifying such patients into the nonischemic group, we added prognostic power to the binary classification model and the more detailed number-of-diseased-vessels model. Since a binary classification system remains attractive due to its simplicity, we propose a modified definition of ischemic etiology of HF that includes patients with single-vessel coronary disease only if present in the left main or proximal LAD artery in patients without a history of MI or prior revascularization (Table 4).
Assigning etiology of HF
Although the accurate assessment of the etiology of HF may be obvious in many cases, it may be a difficult clinical problem in others. Clinicians may use varying definitions or criteria to assign etiology, resulting in heterogeneity in clinical research as well as clinical practice. Our data point out the difficulty of using clinical assessment for this purpose, as 40% of the patients in the nonischemic group reported a history of typical angina. Quantification of coronary disease by angiography remains the most definitive means for assessing both the presence and the extent of CAD. Although a specific histologic pattern (replacement fibrosis) has been reported in endomyocardial biopsy specimens of patients with ischemic cardiomyopathy, this method lacks sensitivity and is significantly invasive (13). Traditional noninvasive means, such as nuclear imaging and echocardiography, have not been shown to consistently differentiate ischemic from nonischemic HF patients (14,15). Novel noninvasive means for differentiating ischemic and nonischemic cardiomyopathy such as positron emission tomography and electron-beam computed tomography are under investigation but not yet validated (16,17). Despite the potential attractiveness of noninvasive testing, our data suggest that detailed assessment of the extent of CAD, as can be done only with angiography, can contribute additional prognostic data in patients with symptomatic HF.
The current study has several limitations. The study cohort was taken from the Duke Databank for Cardiovascular Disease, which contains data only on patients undergoing diagnostic cardiac catheterization. Thus, patients in both the ischemic and nonischemic groups would be weighted toward those with a higher “index of suspicion” for CAD. Although patients thought to be at very low risk for CAD would have been excluded from this analysis, our study population does reflect the patient population in which decision-making about etiology of HF is most difficult. Additionally, the pathophysiology of HF is complex, and many pathologic processes (such as ischemia, hypertension and diabetes mellitus) may contribute to the development of LV dysfunction in a given patient. Although binary classification systems are attractive due to simplicity and clarity, they may represent an oversimplification of a complex biologic phenomenon. Our study is strengthened by the requirement of coronary angiography in all patients, the large cohort of patients and the long-term follow-up.
We propose a new definition of ischemic cardiomyopathy that reclassifies patients with single-vessel disease as nonischemic unless they have left main or proximal LAD disease or a history of revascularization or MI. The use of such a standardized definition will help limit variability in defining etiologic subgroups for clinical trials and population-based studies. Accurate ascertainment of etiology and its impact on prognosis is important for risk stratification of individual patients and for planning appropriate subgroups for clinical research. Based on these data, coronary angiography should remain a cornerstone of the evaluation of patients with newly diagnosed systolic dysfunction and symptomatic HF.
The Coronary Artery Disease (CAD) Index
|The CAD prognostic index is a hierarchical scoring system to describe the extent of CAD on a continuous scale from 1 to 100|
|Extent of CAD||Prognostic Weight(0 to 100)|
|No CAD ≥50%||0|
|One-VD 50% to 74%||19|
|One-VD 50% to 74%||23|
|One-VD ≥ 95%||32|
|Two-VD (both ≥ 95%)||42|
|One-VD ≥ 95%, proximal (LAD)||48|
|Two-VD ≥ 95% LAD||48|
|Two-VD ≥ 95% proximal LAD||56|
|Three-VD ≥ 95% in at least one vessel||63|
|Three-VD 75% proximal LAD||67|
|Three-VD ≥ 95% proximal LAD||74|
|Left main (75%)||82|
|Left main (≥95%)||100|
CAD = coronary artery disease; LAD = left anterior descending coronary artery; VD = vessel disease.
- coronary artery disease
- ejection fraction
- heart failure
- left anterior descending coronary artery
- left ventricle, left ventricular
- myocardial infarction
- New York Heart Association
- Received May 25, 2001.
- Revision received September 10, 2001.
- Accepted October 18, 2001.
- American College of Cardiology
- Gheorghiade M.,
- Bonow R.O.
- Bart B.A.,
- Shaw L.K.,
- McCants C.B.,
- et al.
- Adams K.F.,
- Dunlap S.H.,
- Sueta C.A.,
- et al.
- Follath F.,
- Cleland J.G.,
- Klein W.,
- Murphy R.
- Uretsky B.F.,
- Thygesen K.,
- Armstrong P.W.,
- et al.
- Wallis D.E.,
- O’Connell J.B.,
- Henkin R.E.,
- Costanzo-Nordin M.R.,
- Scanlon P.J.
- Williams J.F.,
- Bristow M.R.,
- Fowler M.B.,
- et al.
- Harris P.J.,
- Lee K.L.,
- Harrell F.E.,
- Behar V.S.,
- Rosati R.A.
- Boyle C.A.,
- Decoufle P.
- Hare J.M.,
- Walford G.D.,
- Hruban R.H.,
- Hutchins G.M.,
- Deckers J.W.,
- Baughman K.L.
- Greenberg J.M.,
- Murphy J.H.,
- Okada R.D.,
- Pohost G.M.,
- Strauss H.W.,
- Boucher C.A.
- Mody F.V.,
- Brunken R.C.,
- Stevenson L.W.,
- Nienaber C.A.,
- Phelps M.E.,
- Schelbert H.R.
- Budhoff M.J.,
- Shavelle D.M.,
- Lamont D.H.,
- et al.