Author + information
- Received July 25, 1996
- Revision received June 5, 1997
- Accepted June 20, 1997
- Published online October 1, 1997.
- Bradley A Bart, MDA,*,
- Linda K Shaw, ABA,
- Charles B McCants Jr., BSA,
- Donald F Fortin, MDA,
- Kerry L Lee, PhDB,
- Robert M Califf, MD, FACCA and
- Christopher M O’Connor, MD, FACC, FACPA
- ↵*Dr. Bradley A. Bart, Cardiology Division, Hennepin County Medical Center 865A, 701 Park Avenue South, Minneapolis, Minnesota 55415.
Objectives. We sought to characterize the clinical determinants of mortality in patients with angiographically diagnosed ischemic or nonischemic cardiomyopathy.
Background. Patients with ischemic cardiomyopathy may have a worse prognosis than patients with nonischemic cardiomyopathy. Few studies have assessed the effect of ischemic versus nonischemic etiology on outcomes.
Methods. We analyzed prospectively collected data on 3,787 patients with a left ventricular ejection fraction ≤40% who underwent coronary angiography. Patients were considered to have ischemic cardiomyopathy (n = 3,112) if they had a history of myocardial infarction, percutaneous transluminal coronary angioplasty, coronary artery bypass graft surgery or at least one major epicardial coronary artery with ≥75% stenosis; all others were considered to have nonischemic cardiomyopathy (n = 675).
Results. The median age, ejection fraction and proportion of patients with New York Heart Association functional class III or IV symptoms for the nonischemic and ischemic groups were 55 years versus 63 years, 27% versus 32% and 57% versus 25%, respectively. After adjustment for baseline clinical risk factors and presenting characteristics, ischemic etiology remained an important independent predictor of 5-year mortality (p < 0.0001). The extent of coronary artery disease was a better predictor of survival than ischemic or nonischemic etiology (log likelihood chi-square 700 vs. 675, respectively).
Conclusions. Ischemic etiology is a significant independent predictor of mortality in patients with cardiomyopathy. However, the extent of coronary artery disease contributes more prognostic information than the clinical diagnosis of ischemic or nonischemic cardiomyopathy. Further research is needed to refine the clinical definition of ischemic cardiomyopathy so that physicians can appropriately prescribe treatment and accurately predict outcome.
Patients with left ventricular systolic dysfunction are commonly divided into two major groups—those with ischemic and nonischemic cardiomyopathy. Ischemic cardiomyopathy, a result of the complications of coronary artery disease (CAD) , is one of the most common causes of heart failure in the Western world .
It is important to distinguish between ischemic and nonischemic cardiomyopathy because the diagnosis influences management. Smoking cessation, lipid-lowering therapy and hormone replacement therapy are important interventions in patients with ischemic cardiomyopathy. Revascularization in patients with low ejection fractions and significant CAD is strongly associated with improved survival [3–5]and should be considered in all patients with ischemic cardiomyopathy . In addition to risk factor modification and revascularization, recent clinical trials suggest that the etiology of heart failure may influence the response to medical therapy. Trials of amlodipine , amiodarone [7, 8], bisoprolol and digoxin [10, 11]suggest that patients with ischemic cardiomyopathy may not respond to medical therapy as favorably as patients with nonischemic cardiomyopathy. Although differential treatment effects are difficult to interpret owing to potential misclassification of heart failure etiology, these findings highlight the importance of evaluating outcomes in patients with well established ischemic and nonischemic cardiomyopathy.
To better describe the clinical and angiographic characteristics and outcomes associated with ischemic versus nonischemic cardiomyopathy, over an 11-year period we evaluated all patients referred for cardiac catheterization at Duke University Medical Center who had a left ventricular ejection fraction ≤40%.
1.1 Patient group.
We identified 4,519 patients with a left ventricular ejection fraction ≤40% who underwent cardiac catheterization at Duke University Medical Center between January 1984 and January 1995 (Fig. 1). We then excluded patients with myocardial infarction within 30 days before index catheterization (n = 472); primary valvular heart disease, as defined by severe aortic or pulmonary insufficiency, or severe stenosis in any heart valve (n = 246); congenital heart disease (n = 14); or previous heart transplantation (n = 0). After exclusions, the final study group included 3,787 patients. Ischemic etiology (n = 3,112) was defined as left ventricular dysfunction with previous myocardial infarction, percutaneous transluminal coronary angioplasty, coronary artery bypass graft surgery or significant CAD (see discussion of cardiac catheterization) at index catheterization; all other patients were considered to have nonischemic cardiomyopathy (n = 675).
1.2 Data collection.
Baseline variables from each patient’s clinical history and physical examination were collected prospectively and stored in the Duke Databank for Cardiovascular Disease [12, 13]. Follow-up procedures have been described [13–15]. Briefly, follow-up information on ischemic cardiomyopathy in patients with significant CAD, coronary angioplasty or bypass surgery was collected at 6 months, 1 year and annually after the index cardiac catheterization. Follow-up information was collected by mailed, self-administered questionnaires, followed by telephone contact to nonresponders. Subsequent clinic visits were also monitored to determine survival status. Identifying information on all patients not contacted by these methods was submitted to the National Death Index . Death certificates were ordered on all potential matches, and each death certificate was examined to confirm the match. The end date for follow-up assessment was January 1, 1995. The median follow-up time was 5.7 years; follow-up for the entire study group was 98.4% complete.
1.3 Cardiac catheterization.
The results of cardiac catheterization were prospectively collected [13, 15, 17, 18]. Significant CAD was defined as ≥75% occlusion of a major epicardial coronary artery . Stenotic lesions were graded subjectively by visual consensus of at least two experienced observers on the following ordinal scale: 0%, 25%, 50%, 75%, 95% or 100% occlusive. The extent of CAD was characterized by the traditional one-, two- or three-vessel disease classification and the Coronary Artery Disease Prognostic Index (Table 1) [14, 19]. The CAD index accounts for the number of diseased vessels but also considers left anterior descending coronary artery disease, the number of 95% diseased vessels, 75% and 95% proximal left anterior descending coronary artery stenosis and 75% and 95% left main coronary artery stenosis.
Left ventricular ejection fraction was determined by biplane ventriculography using the modified area–length method .
1.4 Data analysis.
Medians with 25th and 75th percentiles were calculated for continuous baseline characteristics. Categoric variables are expressed as percentages. We developed Cox regression models for estimating survival by examining univariable tests for interactions between baseline characteristics and ischemic etiology. Next, we constructed a multivariable Cox regression model that included all baseline characteristics and all significant interaction variables. A final model was generated that eliminated nonsignificant model components and retained all significant interactions with ischemic etiology. All candidate variables were examined graphically to ensure that their relation with the outcome was modeled appropriately. Nonlinear relations were recoded or transformed to adhere to model assumptions. The mortality model was developed without variables relating to the extent of CAD, because this information is not typically used in the clinical diagnosis of ischemic cardiomyopathy. Rather, the presence (in contrast to the extent) of significant CAD commonly results in such a diagnosis.
To compare the prognostic contribution of ischemic etiology to that of the extent of CAD, several models were developed. The intent of these modeling procedures was to compare the prognostic contribution of etiology to that of extent of CAD after adjustment for important clinical characteristics. This is in contrast to the survival model described earlier—the intent of which was to develop the best predictive model. The prognostic contribution of etiology and extent of CAD, as measured by the CAD index, was evaluated by comparing overall model log likelihood chi-square tests.
We used Kaplan-Meier survival estimates to describe the survival patterns for all study patients, stratified by disease etiology and extent of CAD. Patients in the ischemic group were subdivided into mild (CAD index ≤42), moderate (CAD index 48 to 63) and severe CAD (CAD index 67 to 100). Follow-up time was computed from the index catheterization date and was based on all-cause mortality. The Cox proportional hazards regression model was used for statistical comparison of survival predictors.
2.1 Clinical characteristics.
The clinical characteristics of patients with ischemic and nonischemic cardiomyopathy are shown in Table 2. There were more men in the ischemic cohort than in the cohort of patients with nonischemic cardiomyopathy. Patients with ischemic cardiomyopathy were older and had a higher incidence of risk factors commonly associated with CAD. Patients with nonischemic cardiomyopathy had more severe heart failure symptoms, lower left ventricular ejection fractions and symptoms of congestive heart failure for a longer period before the index cardiac catheterization. A history of coronary artery revascularization or myocardial infarction was common in the ischemic group. Typical angina pectoris was more common in patients with ischemic cardiomyopathy than in those with nonischemic cardiomyopathy; however, the incidence remained high in the nonischemic cohort (34%).
2.2 Angiographic characteristics.
The coronary anatomy of patients with ischemic and nonischemic cardiomyopathy is shown in Table 3. Seventy-five percent of patients in the ischemic cohort had significant stenosis in the left anterior descending coronary artery; 46% had three-vessel CAD. Seven percent of patients included in the ischemic cohort had no significant stenotic lesions at the time of index catheterization. Of these patients, 93% had a history of myocardial infarction and 58% had a history of coronary angioplasty or bypass surgery. As specified in the subgroup definitions, no patient in the nonischemic group had ≥75% stenosis in any epicardial artery. Sixty-six percent of nonischemic patients had completely normal coronary arteries, whereas 19% had 25% stenotic lesions and another 16% had 50% stenotic lesions.
Unadjusted Kaplan-Meier survival curves for nonischemic patients, according to the degree of coronary artery stenosis, are shown in Fig. 2. Survival in patients with nonischemic cardiomyopathy was similar, regardless of the degree of coronary stenosis present (0%, 25% or 50%, p = 0.39).
The results of a multivariable Cox proportional hazards model are shown in Table 4. After adjustment for differences in baseline characteristics, ischemic etiology remained a significant, independent predictor of mortality in these patients. Other independent predictors of mortality included older age, lower ejection fraction, previous symptomatic heart failure, hypertension, diabetes, peripheral vascular disease, New York Heart Association functional class IV heart failure symptoms and typical angina.
Three interaction variables were also independent predictors of mortality—ejection fraction by etiology (chi-square statistic 7.9, p < 0.01), hypertension by etiology (chi-square 7.6, p < 0.01) and typical angina by etiology (chi-square 8.8, p < 0.01). The prognostic effect of hypertension or typical angina was reduced in the nonischemic group; that is, the relation between these clinical features and mortality was more pronounced in the ischemic group than in the group with nonischemic cardiomyopathy. The prognostic effect of decreasing ejection fraction was reduced in the nonischemic group; the relation between lower ejection fraction and mortality was more pronounced in the ischemic group than in the nonischemic group.
Adjusted Kaplan-Meier survival curves depicting 5-year survival in patients with ischemic and nonischemic cardiomyopathy are shown in Fig. 3. Five-year survival was 0.59 and 0.69 for patients with ischemic and nonischemic cardiomyopathy, respectively (p < 0.0001).
2.4 Prognostic contribution of etiology and extent of CAD.
The prognostic contribution of etiology and extent of CAD, as measured by the CAD index, is shown in Table 5. The CAD index contributed more prognostic information to a model including age, ejection fraction, history of symptomatic heart failure, peripheral vascular disease, functional class IV symptoms and typical angina than did etiology of heart failure (total model log likelihood chi-square 700 vs. chi-square 680, respectively).
Adjusted Kaplan-Meier survival curves depicting 5-year survival in patients with nonischemic cardiomyopathy and ischemic cardiomyopathy divided into groups of mild, moderate and severe CAD are shown in Fig. 4. The 5-year survival rate of ischemic patients with mild CAD was similar to that of the nonischemic group (p = 0.18). Ischemic patients with moderate and severe CAD had increasing mortality rates.
We followed up 3,787 patients with angiographically diagnosed ischemic (n = 3,112) and nonischemic (n = 675) cardiomyopathy over 11 years. Five-year survival in patients with ischemic cardiomyopathy was significantly lower than that in nonischemic patients. In a multivariable Cox proportional hazards model that adjusted for differences in baseline clinical characteristics, ischemic etiology remained a significant predictor of mortality. However, the extent of CAD, as measured by the CAD index, contributed more prognostic information than did the clinical diagnosis of ischemic or nonischemic cardiomyopathy.
3.1 Ischemic cardiomyopathy and mortality.
Several investigators have studied the influence of heart failure etiology on mortality [21–25]. The Studies Of Left Ventricular Dysfunction (SOLVD) investigators followed over 6,000 patients with congestive heart failure for 1 year, 69% of whom had ischemic cardiomyopathy. Their group had a clinical risk profile similar to that of the current study group. Overall mortality for patients in the SOLVD registry at 1 year was 18%, similar to the mortality rate in our investigation. In contrast to the findings of the current investigation, however, the etiology of heart failure (ischemic or nonischemic) in the SOLVD registry did not influence mortality.
One possible explanation for why ischemic etiology was not associated with increased mortality rates in the SOLVD registry relates to their definition of ischemic cardiomyopathy. The diagnosis of ischemic cardiomyopathy in the SOLVD registry was based on clinical judgment. The protocol did not require angiographic or noninvasive evidence of ischemia. Our data highlight the difficulties of making a diagnosis without angiographic results. Angina and the presence of risk factors associated with ischemic heart disease are common in patients with nonischemic cardiomyopathy and should not be exclusively relied on to determine the etiology of left ventricular dysfunction.
Andersson and Waagstein studied a large group of Swedish patients with congestive heart failure and the group was identified retrospectively with an administrative data base. Patients with the International Classification of Disease, ninth revision (ICD-9) diagnostic code for congestive heart failure (n = 2,711) were followed up for 5 years. The mean age of this cohort was 58 years; 68% were men. Because of the retrospective design of this study, no information on left ventricular function or functional class symptoms was reported. Cardiac catheterization was not required to establish the etiology of congestive heart failure. Subgroup analysis of 5-year survival among patients with ischemic and nonischemic cardiomyopathy showed that mortality was significantly higher in the ischemic group (p < 0.0001). In agreement with our results, the presence of ischemic heart disease was an independent predictor of mortality after multiple regression analysis (p < 0.0001). The 5-year mortality rate for all patients in this study was 50%, which was higher than that in our investigation. This probably reflects the method used to identify patients; all were identified at the time of an emergency hospital admission. In contrast, patients in our study were referred for routine cardiac catheterization, many of which were performed on an outpatient basis.
Both the SOLVD registry and the Swedish study contribute to a better understanding of the natural course of congestive heart failure. Both studies are limited, however, in their ability to draw conclusions about the clinical course of ischemic and nonischemic cardiomyopathy. It is very difficult to diagnose ischemic cardiomyopathy reliably without performing coronary angiography, because many clinical features commonly associated with this condition frequently appear in patients with nonischemic disease. Only one other investigation has compared the clinical course of angiographically diagnosed ischemic with that of nonischemic cardiomyopathy. Likoff et al. followed the clinical course of 201 patients with heart failure to determine which clinical variables were the best predictors of mortality. All patients underwent coronary angiography, the results of which were used to determine ischemic or nonischemic etiology. During the 10.8-month follow-up period, the overall mortality rate was 42%; the mortality rate among ischemic patients (n = 121) was significantly higher than that among nonischemic patients (n = 80) (p = 0.005). This difference in survival between ischemic and nonischemic patients is consistent with the findings of our investigation.
3.2 Extent of CAD and ischemic cardiomyopathy.
Although the angiographic diagnosis of ischemic and nonischemic cardiomyopathy does provide valuable prognostic information, assessing the extent of CAD improves the ability to assess risk in patients with left ventricular dysfunction. Our analysis indicates that after adjusting for important baseline characteristics, the prognostic contribution of the extent of CAD disease is greater than that of the clinical diagnosis of ischemic or nonischemic cardiomyopathy. Moreover, the mortality rate of some patients with the clinical diagnosis of ischemic cardiomyopathy and only mild CAD is similar to that of patients in the nonischemic group. In these cases—patients with left ventricular dysfunction “out of proportion” to the extent of CAD—the clinical diagnosis of ischemic cardiomyopathy provides misleading information about expected outcomes.
These findings suggest that coronary angiography is an important part of the assessment of patients with cardiomyopathy and should be considered in all patients with left ventricular dysfunction. Such a diagnostic approach provides important diagnostic and prognostic information; assists in identifying therapies likely to result in improved clinical outcomes such as lipid-lowering therapy, smoking cessation, hormone replacement therapy and revascularization for patients with ischemic cardiomyopathy; and may be useful in evaluating medications that may have important etiology-specific interactions, such as amlodipine, amiodarone, bisoprolol and digoxin.
3.3 Clinical determinants of mortality.
Prognostic models provide a means of assessing a patient’s risk of adverse clinical outcomes and may influence clinical management. The model we developed in this study is particularly important because it represents a large group of patients with left ventricular dysfunction who were followed for a relatively long period. Moreover, the variables that were independent predictors of mortality are commonly collected in patients with heart failure (age, ejection fraction, history of symptomatic heart failure, ischemic or nonischemic etiology, hypertension, diabetes mellitus, peripheral vascular disease, functional class IV symptoms and angina).
Increasing age was the most powerful predictor of mortality in our study cohort. This finding is supported by data from the Framingham Heart Study and other heart failure trials [2, 9, 24]. The Framingham Heart Study assessed the influence of advancing age on mortality in an unselected group of 652 patients who developed heart failure between 1948 and 1988. The risk of death among men and women increased by 27% and 61%, respectively, per decade of age, establishing advancing age as an important predictor of mortality in such a population.
Left ventricular ejection fraction and the severity of heart failure symptoms were also important independent predictors of mortality in our study. These findings are supported by several heart failure trials [2, 9, 27–30]. Cohn et al. assessed the influence of reduced left ventricular function on mortality in 1,446 patients participating in the Veterans Administration Heart Failure Trials (V-HeFT I and V-HeFT II). Reduced ejection fraction was shown to be a powerful predictor of mortality in patients with heart failure, especially for those patients with very low ejection fractions (<25%).
Hypertension, diabetes mellitus and peripheral vascular disease were all independent predictors of mortality in our study cohort. Hypertension and diabetes are important precursors to the development of heart failure and are associated with increased mortality [2, 24, 31]. Such findings highlight the importance of treating these conditions and preventing the development of heart failure. Peripheral vascular disease and angina reflect the presence of atherosclerosis and ischemic heart disease. Their prognostic contribution to our mortality model is therefore not surprising.
3.4 Study limitations.
The major limitation of this study is referral bias. The fact that all the patients with left ventricular dysfunction were referred for cardiac catheterization limits our ability to generalize with all such patients. This limitation, however, is also one of the strengths of this study—the etiology of left ventricular dysfunction in all patients was established by cardiac catheterization (generally considered the “reference standard”). Other studies directly comparing survival in these two patient groups have not required cardiac catheterization to establish the diagnosis [22, 24, 25]. The study by Likoff et al. , one of the few studies that did require cardiac catheterization in all patients, included only 201 patients followed <1 year. The present study, with >3,700 patients and much longer follow-up, represents the largest data set studied to date comparing outcomes in angiographically confirmed ischemic and nonischemic cardiomyopathy.
Ischemic etiology is an independent predictor of mortality in patients with left ventricular dysfunction. However, the extent of CAD contributes more prognostic information than the clinical diagnosis of ischemic or nonischemic cardiomyopathy. Coronary angiography should be considered in all patients with left ventricular dysfunction, because the results substantially contribute to diagnosis, prognosis and management decisions. Further research is needed to refine the clinical definition of ischemic cardiomyopathy so physicians can appropriately prescribe treatment and accurately predict outcome.
☆ To discuss this article on-line, visit the ACC Home Page at www.acc.org/membersand click on the JACC Forum
- Received July 25, 1996.
- Revision received June 5, 1997.
- Accepted June 20, 1997.
- The American College of Cardiology
- Alderman EL,
- Fisher LD,
- Litwin P,
- et al.
- O’Connor CM,
- Puma JA,
- Gardner LH,
- Califf RM,
- Jones RH
- Singh SN,
- Fisher SG,
- Singh BN,
- Massie BM,
- Fletcher RD
- CIBIS Investigators and Committees
- Gheorghiade M,
- Young JB,
- Uretsky BF,
- Packer M
- Yusuf S,
- Garg R,
- Smith T,
- et al.
- Harris PJ,
- Harrell FE Jr.,
- Lee KL,
- Behar VS,
- Rosati RA
- Mark DB,
- Nelson CL,
- Califf RM,
- et al.
- ↵Patterson JE, Bilgrad R. The National Death Index experience. In: Record Linkage Techniques—1985: Proceedings of the Workshop on Exact Matching Methodologies (Arlington, Virginia, May 9–10, 1985). Washington, DC: Department of the Treasury, Internal Revenue Service, Statistics of Income Division, 1995, publication no. 1299 (2–86).
- Pryor DB,
- Harrell FE Jr.,
- Rankin JS,
- et al.
- Smith LR,
- Harrell FE Jr.,
- Rankin JS,
- et al.
- van den Broek SA,
- van Veldhuisen DJ,
- de Graeff PA,
- Landsman ML,
- Hillege H,
- Lie KI
- Ho KK,
- Anderson KM,
- Kannel WB,
- Grossman W,
- Levy D
- for the V-HeFT VA Cooperative Studies Group,
- Cohn JN,
- Johnson GR,
- Shabetai R,
- et al.
- Massie BM,
- Conway M
- for the V-HeFT VA Cooperative Studies Group,
- Johnson G,
- Carson P,
- Francis GS,
- Cohn JN
- Scrutinio D,
- Lagioia R,
- Ricci A,
- Clemente M,
- Boni L,
- Rizzon P