Exercise capacity determined by peak ventilatory oxygen uptake (VO₂) has attracted considerable interest in recent years as a way to help estimate prognosis in chronic heart failure [4-6, 9-12]. Peak VO₂ has been shown to complement other clinical and invasive markers of disease severity, including ejection fraction, pulmonary capillary wedge pressure, type of heart failure, and cardiac index. Functional impairment has therefore evolved as an important criterion for selecting patients for transplantation. It has been demonstrated that a peak VO₂ less than 10 mL/kg per minute identifies a group of patients with a particularly poor 1-year prognosis (1-year mortality rate, 24% to 77%) [6, 9, 12, 13]. Conversely, patients in whom the peak VO₂ is greater than 14 mL/kg per minute have a 1-year survival rate similar to that of patients receiving transplantation (>90%); this suggests that transplantation can be safely deferred in these patients [9]. However, numerous other criteria have also been proposed [11, 13-16]. Previous studies have tended to include small numbers of observed events; have used end points other than death (such as transplantation or change in status for transplant listing); and have been limited by short follow-up (mean, 2 years), resulting in relatively wide CIs in the odds ratios for death rates. Moreover, there have been few integrated approaches to assessing survival (including clinical, exercise, echocardiographic, and hemodynamic data) in the modern era, in which most patients are aggressively treated with angiotensin-converting enzyme inhibitors and survival seems to be improving [6].

Exercise test, ventilatory gas exchange, clinical, and catheterization data have been gathered from patients with severe heart failure at Stanford University over the past 10 years. The high volume of referrals to Stanford over this period has yielded a relatively robust database, including 187 deaths during a mean follow-up of 4 years, with which to address these issues. We sought to determine clinical, hemodynamic, and cardiopulmonary exercise test predictors of survival in patients referred for evaluation of severe heart failure.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

The study group included 644 ambulatory patients with advanced heart failure who were referred for evaluation at Stanford University between 1986 and 1995 and underwent a valid cardiopulmonary exercise test. The duration of heart failure was at least 1 month, and most of the patients had had one or more hospital admissions for decompensated heart failure. Evaluation included history, cardiovascular examination, and maximal exercise testing with ventilatory gas exchange analysis. Patients who were unable to undergo exercise testing were not included in the study. Standard medical treatment, including digoxin, diuretic agents, and angiotensin-converting enzyme inhibitors, was administered. Many patients were taking warfarin (46%); antiarrhythmic agents (23%); or a ß-blocker, usually metoprolol (12%). All patients were stabilized clinically before undergoing exercise testing, right-heart catheterization, and echocardiography.

Variables from the initial examination that were included in the analysis included age, sex, body surface area, and type of underlying heart disease. Underlying heart disease was coded as ischemic or idiopathic cardiomyopathy. Ischemic cardiomyopathy was defined on the basis of previous myocardial infarction, percutaneous transluminal coronary angioplasty, bypass surgery, or history of significant coronary artery disease by catheterization. All other patients were considered to have idiopathic cardiomyopathy.

Exercise Testing

Before undergoing exercise testing, all patients provided written informed consent according to a protocol established by the institutional review board at Stanford University. Symptom-limited maximal exercise tests were performed while patients were upright by using an electrically braked bicycle ergometer at a constant cadence of 60 rpm. A continuous ramp protocol was used in which the work rate was increased by 10 W/min. Ventilatory oxygen uptake was measured by using a Medical Graphics Corporation 2001 system (St. Paul, Minnesota). Gas exchange data were obtained breath by breath and were recorded every 30 seconds. Ventilatory oxygen uptake per minute, carbon dioxide production, expiratory volume per minute, and respiratory exchange ratio were calculated online. Peak VO₂ was defined as the highest VO (₂) achieved during exercise. The ventilatory threshold was determined by two experienced reviewers using the V-slope method [17] and was confirmed by ventilatory criteria. Heart rate was recorded continuously by electrocardiography, and blood pressure was recorded at regular intervals throughout the test with a semiautomated recorder (Quinton STBP-680, Quinton Instruments, Seattle, Washington).

Other Procedures

Right-heart catheterization with measurement of right atrial, pulmonary artery, and pulmonary wedge pressures and calculation of cardiac output done using the Fick method [18] was performed in 63% of patients. Left ventricular ejection fraction was measured with radionuclide or angiographic ventriculography [19] in 74% of the cohort. Left ventricular dimensions and fractional shortening were obtained by using two-dimensional echocardiography in 51% of patients.

Follow-up

Patients were followed at the Stanford Heart Failure Clinic or by the referring physician. Patient status was determined from medical records, telephone interview of the patient or the patient's family, or an information resource research service (IRSC, Fullerton, California) as of December 1995. Twelve patients (1.7%) were lost to follow-up. The analysis was done with death as the end point; transplantation was considered a censored event (removed from the study at the time of transplantation).

Statistical Analysis

Data are expressed as the mean ±SD unless otherwise stated. Group differences (survivors compared with nonsurvivors) were compared by using unpaired t-tests. A Cox proportional-hazards model was used to determine the effect of a given independent variable on time to death. The time to event was expressed in days. Predicted peak VO₂ was determined by using a sex-, age-, weight-, height-, and protocol-specific formula outlined by Wasserman and colleagues [17] and was expressed as a percentage of normal by using the following equation: achieved peak VO_2/predicted peak VO₂ x 100.

Results

Top Methods Results Discussion Author & Article Info References

Baseline Characteristics

Descriptive characteristics of the study sample are presented in Table 1. Eighty percent of patients were men. Idiopathic dilated cardiomyopathy was the predominant underlying cause of chronic heart failure (383 patients [59%]). Ischemic heart disease was the underlying cause in 221 patients (34%), and the remaining 40 patients (6%) had other causes. Despite optimal medical treatment, all patients had symptomatic heart failure with a marked decrease in ventricular function as evidenced by a mean ejection fraction of 0.196 ± 0.096, a peak VO₂ of 16.1 ± 5.7 mL/kg per minute, and echocardiographic data demonstrating ventricular enlargement with decreased fractional shortening. Eighty percent of patients received angiotensin-converting enzyme inhibitors, 76% received digoxin, and 82% received diuretic agents. After initial examination, 225 patients (34.9%) were accepted for transplantation and put on the waiting list, and 419 (51.4%) were rejected because they were too well, were too sick, were too old, or had pulmonary hypertension or other contraindications to transplantation.

Survival and Transplantation

During the follow-up period of 48.2 ± 28.3 months, 187 patients (29%) died. Actuarial 1-year and 5-year survival rates were 90.5% and 73.4%, respectively. Of the 225 patients listed for heart transplantation, 104 (16.1% of the total cohort) underwent transplantation (34 of the 104 died during follow-up) and 48 (7.5% of the total cohort) died while awaiting transplantation. Of the patients rejected for transplantation, 81 (12.6% of the total cohort) died during follow-up. The 1-year and 5-year survival rates for patients who received transplants were 85% and 68%, respectively. Patients who were listed for and underwent transplantation were similar to those not listed for transplantation in age, body surface area, and cardiac output but had lower ejection fractions (0.164 ± 0.062 compared with 0.203 ± 0.10), lower peak VO₂ (14.1 ± 5 mL/kg per minute compared with 16.5 ± 6 mL/kg per minute), and greater pulmonary capillary wedge pressure (25.6 ± 8 mm Hg compared with 20.0 ± 7 mm Hg).

Determinants of Survival

The clinical, hemodynamic, and exercise data of survivors and nonsurvivors are shown in Table 2. All measurements of exercise capacity (peak VO₂, exercise time, watts achieved, peak heart rate, and VO₂ expressed as a percentage of the predicted maximum) were higher among survivors than among nonsurvivors (P < 0.001 for each variable). The VO₂ and heart rate at the ventilatory threshold were also higher among survivors (P = 0.02 for each). Resting heart rate, pulmonary wedge pressure, cardiac index, and ejection fraction did not differ between the groups.

Univariate analysis showed that age and type of heart failure were significant predictors of death among clinical variables. Patients with underlying coronary artery disease had an increased risk for death compared with patients who had idiopathic dilated cardiomyopathy (relative risk, 1.73 [95% CI, 1.35 to 2.20]; P < 0.01) (Figure 1). Sex, body weight, and body mass index were not related to survival. Among exercise variables, peak VO₂, VO₂ at the ventilatory threshold, and peak watts achieved were significant predictors of death. Peak VO₂ expressed as a percentage of normal and as less than the median predicted normal value of 55% were of borderline significance (P = 0.06 for both variables). Hemodynamic variables (cardiac index, ejection fraction, and pulmonary capillary wedge pressure) were not significant univariate determinants of death, although survival analysis for resting pulmonary capillary wedge pressure and left ventricular end-diastolic dimensions demonstrated small but significant differences for patients in whom values were above the median compared with those in whom values were below the median.

View larger version (10K): [in this window] [in a new window]   Figure 1. Survival curves in patients with chronic heart failure and idiopathic cardiomyopathy (squares) or underlying coronary artery disease (circles). The dotted line represents the annual mortality rate expected on the basis of sex and age. The difference between the two curves was significant (P < 0.001). Numbers given along the curves are numbers of patients evaluated at each time point; numbers in parentheses are cumulative numbers of deaths.

Separate univariate analyses were performed for patients with underlying coronary artery disease and cardiomyopathy. In both groups, the strongest and most consistent predictors of death were measures of exercise capacity, including peak VO₂, VO₂ at the ventilatory threshold, percentage of age-related peak VO (₂) achieved, and peak watts achieved. With the exception of age and pulmonary capillary wedge pressure in the cardiomyopathy group, no clinical or hemodynamic variables were significant univariate predictors of death.

By multivariate analysis in the coronary artery disease and cardiomyopathy groups, peak VO₂ was the only significant predictor of death. The maximal systolic blood pressure achieved (<130 mm Hg) was almost significant in the coronary artery disease group (P = 0.07), and age was almost significant in the idiopathic cardiomyopathy group (P = 0.11). Because these variables in combination provided a clinically significant widening of the survival curves, a regression model was developed for each group and was expressed as a score. Figure 2 shows the survival curve for patients with coronary artery disease in whom scores were greater than and less than the median of 13, calculated as [score = 5 x (1 if maximal systolic blood pressure 130 mm Hg or 0 if maximal systolic blood pressure <130 mm Hg) + peak VO₂] and for patients with cardiomyopathy in whom scores were greater than and in whom scores were less than 18, calculated as [score = 3.6 x (peak VO₂) –age].

View larger version (15K): [in this window] [in a new window]   Figure 2. Top. Survival curves for patients who had a regression score greater than 13 (squares) and those who had a regression score less than 13 or equal to (circles) among patients with chronic heart failure and underlying coronary artery disease (P < 0.001 between curves). Bottom. Survival curves for patients who had a regression score greater than 18 (squares) and those who had a regression score 18 (circles) among patients with chronic heart failure and underlying cardiomyopathy (P = 0.006 between curves).

Survival curves between patients who achieved a peak VO₂ greater than compared with less than 12, 14, and 16 mL/kg per minute are shown in Figure 3. Each peak VO₂ between 11 and 16 mL/kg per minute significantly estimated risk for death, and each measurement was similar in its ability to estimate risk for death at 1, 2, and 3 years.

View larger version (17K): [in this window] [in a new window]   Figure 3. Survival curves for patients achieving above compared with below 12, 14, and 16 mL/kg per minute for peak oxygen uptake (VO2). Numbers refer to patients evaluated at each time point for each survival curve; numbers in parentheses are cumulative numbers of deaths.

Figure 4 shows survival curves for patients in whom the peak VO₂ was greater than 14 mL/kg per minute and in whom the peak VO₂ was 14 mL/kg per minute or less, as predicted from the work rate (in watts) [18] and as measured directly by using gas exchange techniques. Only peak VO₂ measured directly discriminated between survivors and non-survivors.

View larger version (15K): [in this window] [in a new window]   Figure 4. Top. Survival curves for patients in whom the peak oxygen uptake (VO2) predicted from the work rate was greater than 14 mL/kg per minute (circles) and those in whom the predicted peak VO2 was 14 mL/kg per minute or less (squares). Bottom. Survival curves for patients in whom the peak VO2 measured directly was greater than 14 mL/kg per minute (circles) and those in whom the peak VO2 was 14 mL/kg per minute or less (squares). Patients in whom the predicted peak VO2 was greater than 14 mL/kg per minute did not differ significantly from those in whom the predicted peak VO2 was 14 mL/kg per minute or less, but for measured peak VO2, the difference in survival between patients achieving greater than and those achieving less than 14 mL/kg per minute was significant (P < 0.001).

Discussion

Top Methods Results Discussion Author & Article Info References

To our knowledge, this study is the largest evaluation of initial clinical, exercise, and hemodynamic profiles and outcomes among ambulatory patients referred to a heart failure and transplantation center. Data were collected over 10 years, permitting considerably more "hard" end points [death] and longer follow-up, which in turn allowed more robust statistical analysis. Our major findings are that 1) peak VO₂ was a stronger predictor of risk for death than were other exercise, clinical, or hemodynamic variables, 2) no single cut-point for peak VO₂ was superior to another in predicting risk for death, and 3) right-heart catheterization data were not important in the assessment of risk in patients with chronic heart failure.

Exercise Capacity

It has been known for several decades that exercise capacity, expressed as exercise time or workload achieved, plays an important role in risk stratification of patients with heart disease [20]. In recent years, interest has grown in peak VO₂ directly measured by using gas exchange techniques in patients with heart failure. In theory, directly measured VO₂ should estimate prognosis more accurately than other functional indices do because it yields a more precise and reproducible measure of exercise capacity [18, 21] and is an expression of the functional health of the combined cardiovascular, pulmonary, and skeletal muscle systems. Among the recent American Heart Association/American College of Cardiology task force recommendations was the recommendation that peak VO₂ should be directly measured as part of the routine work-up when patients are being considered for transplantation [14]. In the past several years, a peak VO₂ of 14 mL/kg per minute has gained acceptance as a cut-point for medical and surgical decision making in patients with chronic heart failure [9, 10, 14]. However, it was not previously demonstrated that directly measured peak VO₂ is superior to VO₂ predicted from the peak work rate in terms of prognosis. In our study, peak VO₂ was a more significant predictor of death than peak watts achieved, although both variables were significant univariate predictors of death. Of interest, when expressed using the common cut-point of 14 mL/kg per minute, measured peak VO₂ was a significant risk factor for death, whereas exercise capacity predicted from the work rate was not (Figure 4). These findings seem to confirm the importance of gas exchange techniques in the assessment of risk in patients with heart failure.

As have other recent studies [9, 10, 12], we found that a peak VO₂ of 14 mL/kg per minute discriminated between survivors and nonsurvivors. However, this criterion ( 14 mL/kg per minute compared with >14 mL/kg per minute) was a relatively modest predictor of death in the Cox proportional-hazards model and was not a significant predictor of death by multivariate analysis. In addition, the differences in risk for death were similar in patients in whom peak VO₂ was greater than 11, 12, 13, 14, 15, and 16 mL/kg per minute compared with those in whom peak VO₂ was 11, 12, 13, 14, 15, and 16 mL/kg per minute or less (Figure 3); mortality was markedly different regardless of the cut-point used. This suggests that a peak VO₂ of 14 mL/kg per minute, although commonly used as a cut-point, may not necessarily be the optimal criterion for determining survival in patients with heart failure. Moreover, whereas we and Mancini and colleagues [9] found similar 1-year survival rates for patients in whom the peak VO₂ was 14 mL/kg per minute or greater (94% compared with 88%), the survival rate that we found for patients in whom the peak VO₂ was less than 14 mL/kg per minute was substantially higher than that found by Mancini and colleagues (65% compared with 47%). Thus, although our data confirm previous observations that an exercise capacity of 14 mL/kg per minute or more portends excellent survival [9, 10], they suggest that most patients in whom the exercise capacity is less than 14 mL/kg per minute also survive. The higher survival rate that we found is consistent with the improving survival rate for these patients in recent years and is generally attributed to the broader use of angiotensin-converting enzyme inhibitors, more aggressive reduction of filling pressures, and avoidance of type I antiarrhythmic agents [1, 2, 6].

We also found that peak VO₂ was a stronger predictor of death than the percentage of normal peak VO₂ achieved. This contrasts with the recent findings of Stelken and coworkers [11], who found that a value of 50% or less of the predicted maximum VO₂ was the strongest multivariate predictor of cardiac death or listing as status 1 priority for transplantation. Our study and that by Stelken and coworkers [11] both used the Wasserman equation, which considers age, sex, exercise mode, height, and body weight in developing a normal standard [17]. In our study, the median achieved value was 55% of the maximum predicted VO₂, and we found that this outperformed other percentages of predicted normal values. Aaronson and Mancini [12] recently reported that two different equations used to estimate the percentage of normal peak VO₂ in patients with severe heart failure did not improve the prediction of survival over that achieved with peak VO₂ alone. In theory, expressing peak VO₂ relative to a normal standard should improve its predictive accuracy because VO₂ is influenced by sex; body weight; and, most important, age. However, the relation between peak VO₂ and age in the various published regression equations is relatively weak, and the regression between these two factors varies considerably [18]. Moreover, peak VO₂ is related to other variables that are not so easily measured, such as psychological factors, muscle mass, and strength [18, 22]. These factors increase the importance of which regression Equation is used as a relative standard for normal, the specificity of the population, and the exercise mode and protocol used, all of which strongly influence normal reference values [17, 18]. Because the available evidence is conflicting, this seems to be an area in need of further study.

We know of no previous studies that apply VO₂ at the ventilatory threshold in the context of prognosis. We found that VO₂ at the ventilatory threshold was a significant univariate predictor of death, although it did not outperform peak VO₂. Although this measurement carries inherent problems related to definition, determination, and application [23], the observation that it may have a complementary role in prognosis is intriguing. Because of safety concerns, muscle weakness, deconditioning, or proximity to myocardial infarction, some patients with chronic heart failure are not appropriate candidates for, or may not be able to perform, a symptom-limited maximal exercise test. In such patients, submaximal testing is more appropriate and can be useful for making activity recommendations, modifying the medical regimen, and evaluating the need for further interventions [6]. Moreover, by definition, patients hyperventilate at exercise levels beyond this point, making the ventilatory threshold a useful upper limit for daily activities. These findings suggest that even when submaximal testing is appropriate, the test may nevertheless yield prognostic information in this population.

Hemodynamic Data

The observation that peak VO₂ outperformed clinical and hemodynamic variables in determining risk for death is provocative. Ejection fraction, left ventricular dilation, and pulmonary capillary wedge pressure are generally considered to be major markers for risk in heart failure [4, 6, 14, 16], and patients with severe heart failure frequently undergo right-heart catheterization to optimize assessment of risk. In our study, cardiac index, ejection fraction, and pulmonary capillary wedge pressure were relatively poor predictors of outcome, and minimal differences were seen between survivors and nonsurvivors for these variables (Table 2). This is in general agreement with the findings of Mancini and colleagues [9] and Stevenson and associates [6], who also compared resting hemodynamic observations with exercise responses as determinants of risk in heart failure, and it underscores the fact that exercise capacity and hemodynamic variables are poorly related [24]. Of note, Connors and coworkers [25] found that right-heart catheterization did not favorably influence outcomes in critically ill patients, including a subgroup of 456 patients with severe chronic heart failure. Chomsky and colleagues [26] recently reported that hemodynamic data obtained during exercise provides more valuable prognostic information than resting hemodynamic data or other clinical and exercise test variables. Finally, Aaronson and coworkers [27] developed multivariate models for predicting death or transplantation in patients with heart failure and identified a wider spectrum of variables, including eight invasive (such as blood pressure, peak VO₂, ejection fraction, and pulmonary capillary wedge pressure) and seven noninvasive (including underlying coronary artery disease, resting heart rate and blood pressure, and peak VO₂) variables, as significant predictors of outcome. Using these models, marked differences between low-risk (approximately 90% 1-year survival) and high-risk (approximately 40% 1-year survival) subgroups were seen.

Limitations

Our study sample differed from those in similar recent studies in that all patients had severe heart failure but not all patients were candidates for transplantation. How aggressively patients are referred for transplantation, as well as the intensity of medical therapy, depends on the treatment philosophy at different institutions. Our study assessed mortality in a referral population, and the findings may not be entirely generalizable to all patients with chronic heart failure. In the present climate of aggressive therapy, the natural history of heart failure is difficult to determine because it is altered by such interventions as transplantation, which we appropriately censored. Our sample included patients accepted for transplantation after evaluation, which raises the possibility of selection bias. The cause of heart failure can be difficult to determine and may not have been categorized accurately in all patients. Serum sodium and plasma catecholamine levels have been shown to predict outcome in heart failure, and these data were not available to us. Left ventricular ejection fraction measured by catheterization or echocardiography was used, and results obtained with these two techniques may differ. Our entry criteria required a symptom-limited exercise test with gas exchange analysis; patients who are unable to exercise are known to have an elevated risk for death [6], and we do not know the mortality rate among these patients in our sample. Finally, patients were evaluated after clinical stabilization, and both hemodynamic and exercise observations would be strongly influenced by the degree of stability achieved in each individual patient.

Clinical Implications

Our study further defines factors associated with shortened life expectancy 1 to 5 years after clinical evaluation in patients with severe heart failure. In this population, exercise test data, including gas exchange analysis, are clearly important components of the risk profile; peak VO₂ outperformed all other clinical, exercise, and hemodynamic data in determining risk for death in patients with severe heart failure. This suggests that all patients being evaluated for heart failure should undergo a cardiopulmonary exercise test; for evaluation of risk for death, this test seems to be more important than right-heart catheterization. Given the current short-age of donor hearts and the costs and morbidity associated with transplantation, the cardiopulmonary exercise test may have a crucial, cost-effective place in the evaluation of patients with heart failure.

From Palo Alto Veterans Affairs Health Care System, Palo Alto, California; and Stanford University School of Medicine, Stanford, California.

Dr. Gullestad: Medical Department B, Rikshospitalet, 0027 Oslo, Norway.

Drs. Vagelos and Fowler: Stanford University Medical Center, Cardiovascular Research Building, Stanford, CA 94304-5246.

Mr. Do: 4900 Medical Drive, Apartment 1204, San Antonio, TX 78229.

Dr. Ross: The Toronto Hospital, NVW 10-129, 200 Elisabeth Street, Toronto, Ontario M5G 2C4, Canada.