ARTICLE IN PRESS. Determining the Best Ventilatory Efficiency Measure to Predict Mortality in Patients with Heart Failure

Transcription

1 ARTICLE IN PRESS Determining the Best Ventilatory Efficiency Measure to Predict Mortality in Patients with Heart Failure Robert L. Bard, MA, a Brenda W. Gillespie, PhD, b Nicholas S. Clarke, a Timothy G. Egan, MD a and John M. Nicklas, MD, Background: Ventilatory efficiency, the relationship between ventilation (VE) and carbon dioxide production (VCO 2 ), predicts mortality in heart failure patients, but its determination has not been standardized. Additionally, it is unclear if the prognostic power of ventilatory efficiency is independent of exercise intensity. Therefore, we investigated the relative prognostic power of different measures of ventilatory efficiency calculated from maximal and sub-maximal exercise in patients with heart failure. Methods and Heart failure patients (n 355, 72% males, age years) had follow-up for at least 5 years from Results: an exercise test. There were 145 events (133 deaths and 12 emergent cardiac transplants). Ventilatory efficiency calculations were not equivalent. Of the different measures of ventilatory efficiency, the VE/VCO 2 slope to peak exercise was the most significant predictor of mortality in a multivariable Cox model, including ejection fraction, systolic blood pressure, peak oxygen consumption (VO 2 ), gender, etiology, and heart rate.a5uincrement of the VE/VCO 2 slope to peak exercise corresponded to a 9% increase in mortality risk. When tests were grouped by peak exercise intensity, by quartiles of peak respiratory exchange ratio, the VE/VCO 2 slope to peak exercise was always a better predictor than peak VO 2. Conclusion: Peak and sub-maximal measures of ventilatory efficiency were not equivalent, and the VE/VCO 2 slope to peak exercise was the best predictor of mortality in patients with heart failure. Thus, the prognostic power of ventilatory efficiency is enhanced when exercise extends beyond the ventilatory threshold and includes all of the available exercise data. J Heart Lung Transplant. Patients with heart failure often use relatively high ventilatory volumes to exchange carbon dioxide during exercise. 1 This abnormality of ventilatory efficiency is commonly calculated from the relationship between ventilation (VE) and carbon dioxide production (VCO 2 ) or VE/VCO 2. Elevated VE/VCO 2 during graded exercise is frequent among symptomatic patients and has been identified as an independent prognostic marker of mortality. 2 VE/VCO 2 is independent of peak oxygen consumption (VO 2 ) as a prognostic marker, and patients with a high VE/VCO 2 and a low peak VO 2 have From the a Division of Cardiovascular Medicine and the b Center for Statistical Consultation and Research, University of Michigan, Ann Arbor, Michigan. Submitted April 7, 2005; revised November 10, 2005; accepted November 10, Presented as a poster at the annual scientific sessions of the American Heart Association, Orlando, Florida, November 9 12, Correspondence: Robert L. Bard, MA, University of Michigan, 24 Frank Lloyd Wright Drive, Division of Cardiovascular Medicine, Lobby M, 3 rd Floor, Ann Arbor, MI Telephone: Fax: bbard@umich.edu Copyright 2006 by the International Society for Heart and Lung Transplantation /06/$ see front matter. doi: / j.healun been reported to have especially high mortality rates. 3 A low VO 2 at ventilatory threshold and a high VE/VCO 2 slope has also been associated with increased mortality rates. 4 There is no standard definition of ventilatory efficiency, however. 5,6 Ventilatory efficiency has been variably calculated in different studies as slopes, ratios, and averages of VE/VCO 2 ratios. Many investigators represent ventilatory efficiency by a plot of exercise ventilation vs carbon dioxide production and report the slope of this relationship (VE/VCO 2 slope). 2,7 9 Some investigators omit data past their identification of a respiratory compensation point. This introduces subjectivity, because only tests with an identifiable respiratory compensation point have data omitted and every subject s slope is not calculated in the same manner. Some investigators report only a single sub-maximal 3 or peak 10,11 ratio of VE/VCO 2, which can be mistaken as the slope. A newer calculation defined as the mean of the lowest 3 consecutive of VE/VCO 2 ratio points has been proposed as the best calculation of ventilatory efficiency, despite being tested in only healthy volunteers. 12 Clinicians have been interested in obtaining the same prognostic information from sub-maximal exercise that 1

2 2 Bard et al. ARTICLE IN PRESSThe Journal of Heart and Lung Transplantation is currently obtained from maximal exercise to make the testing more applicable to activities of daily living. Many clinicians believe that maximal and sub-maximal VE/VCO 2 slopes are equivalent in both value and prognostic ability despite a lack of published evidence. Metra et al 1 reported a significant correlation (r 0.83) between peak and sub-maximal slopes, but this relationship is expected by definition because the calculation of peak slope includes every point that was used in the calculation of the sub-maximal slope. Ventilatory efficiency, by any measure, is an excellent predictor of mortality in heart failure patients, but only 2 studies 13,14 compare the prognostic difference between the peak and sub-maximal slopes. Gitt et al 4 attempted to use sub-maximal exercise (ventilatory threshold 11ml kg 1 min 1 ) in combination with VE/VCO 2 slope to assess risk, but the submaximal measure of ventilatory efficiency has been shown in 2 different studies 13,14 to be inferior to a maximal measure of ventilatory efficiency. Both of these studies were relatively small, observed few mortality end points, and did not involve a multivariable model including other mortality predictors. The purpose of our study was to determine if there were statistical and prognostic differences among the different determinations of ventilatory efficiency and compare their relative abilities to predict mortality in a heart failure population by using a multivariable model. METHODS The University of Michigan Medical School Institutional Review Board approved this project, and each patient gave written informed consent before exercise testing. All patients referred for clinical cardiopulmonary exercise testing during a cardiac transplant evaluation from January 1997 to September 1998 were considered. Patients were prospectively included in the database if they were taking a loop diuretic and had a dilated cardiomyopathy for at least 3 months with a documented left ventricular ejection fraction of 40%. A primary diagnosis of heart failure was required, and patients with other significant or end-stage diseases were excluded. Patients were classified as those with an ischemic or non-ischemic cardiomyopathy, and patients with hypertrophic cardiomyopathy were excluded. In cases where patients were serially tested, only the first test was included. Medical histories were reviewed with each patient, and body weight and height were measured. Patients rested for at least 10 minutes as their chests were prepared for electrodes, testing procedures were explained, the metabolic cart (Medical Graphics Corpora- Figure 1. Representative patient example of the calculation of the ventilatory equivalent for carbon dioxide (VE/VCO 2 ) slope and VE/VCO 2 ratio at ventilatory threshold and at peak exercise (PkEx).

3 The Journal of Heart and Lung Transplantation ARTICLE IN PRESS Bard et al. 3 tion CPX/D, St. Paul, MN) was calibrated, and a 12-lead electrocardiogram and blood pressure were obtained. Exercise gas exchange was collected breath-bybreath from a ramping treadmill exercise test. Peak VO 2 assessment was standardized by using the graphical method 15 and ventilatory threshold (VT) was determined by using the V-slope technique. 16 Predicted oxygen uptake was calculated based on the gender and body weight of each patient. 17 Every test was conducted, analyzed, and standardized by the same investigator, who encouraged each patient to exercise to exhaustion. Five different methods of determining ventilatory efficiency were compared. Breath-by-breath gas exchange data were averaged by the 15-second sampling interval, and VE/VCO 2 slope was determined with the metabolic cart s Breeze 3.06 software (Medical Graphics Corporation) by building a graph of ventilation and carbon dioxide production and fitting a linear regression equation to the relationship. Data collected from Breeze 3.06 software were imported into SAS 12.0 statistical software (SAS Institute, Inc, Cary, NC) for confirmation of the results provided by Breeze. The VE/VCO 2 slope to VT was based on data from the onset of exercise to VT, and VE/VCO 2 slope to peak exercise (PkEx) was determined from the onset of exercise to PkEx (Figure 1). Also calculated was the VE/VCO 2 ratio at VT, which was defined as the point-specific ratio of ventilation to carbon dioxide production at VT (Figure 1). The point-specific VE/VCO 2 ratio was also calculated at PkEx. The lowest VE/VCO 2 ratios were determined in the same manner as reported by Sun et al, 12 defined as the smallest average of 3 consecutive point-wise VE/ VCO 2 ratios during exercise. Statistical Analysis and Prognostic Assessment Descriptive statistics are reported as the mean and standard deviation (SD). The medication profiles were not tracked, the medications are those prescribed at the time of the exercise test. Paired samples t-tests compared the sub-maximal and PkEx intensity variables. Predictors of patient survival from the time of the exercise test were investigated by using Cox regression. Follow-up data included mortality and cardiac transplantation from the date of the exercise test; status 1A transplantation patients 18 were treated as deaths, and other transplants were censored. A secondary analysis considered all transplants as censored. All patients had at least 5 years of potential follow-up from the date of their exercise test. Follow-up was performed using the Social Security Death Index and the patients medical records. The hazard ratios and corresponding 95% confidence intervals were obtained from Cox regression. Linearity of continuous covariates was investigated by including each variable and its square in the base model. Possible threshold values of mortality predictors were investigated by treating each variable as categoric, grouping patients by tertiles and quartiles, to determine the linearity of the hazard ratios obtained from Cox regression. A risk function was calculated by graphing the variable s hazard ratio by tertile or quartile to investigate linear or threshold relationships. Akaike information criterion was used to compare non-nested models. 19 A base survival model was built that included variables that were significantly predictive of mortality but did not include the ventilatory efficiency variables under study. Variables considered for the base model included age, gender, etiology, cigarette smoking, diabetes mellitus, left ventricular ejection fraction, body mass index, resting systolic blood pressure, resting diastolic blood pressure, resting heart rate, atrial fibrillation, attainment of VT, -blockers, angiotensin-converting enzyme (ACE) inhibitors peak VO 2, percent predicted peak VO 2, peak respiratory exchange ratio (RER), peak systolic blood pressure, peak diastolic blood pressure, and peak heart rate. Once a base model was established, each ventilatory efficiency variable under investigation was added individually. The predictive abilities of each added variable were compared using Wald and model chi-square values. To further investigate the effect of exercise intensity on the model, the predictive ability of peak VO 2 and VE/VCO 2 slope to PkEx were compared within each quartile of RER. Table 1. Patient Characteristics All subjects N (men/women) 355 (255/100) Follow-up (y) Ischemic/non-ischemic 159/197 Body mass index (kg/m 2 ) Weight (kg) Age (y) Active cigarette smokers (n) 43 (12) Current diabetes mellitus (n) 75 (21) Current atrial fibrillation (n) 33 (9) Left ventricular ejection fraction (%) Peak VO 2 (ml kg 1 min 1 ) Percent predicted peak VO 2 (%) Ventilatory threshold (ml kg 1 min 1 ) (n 305, 86%) Peak respiratory exchange ratio Resting systolic blood pressure (mm Hg) Resting diastolic blood pressure (mm Hg) Peak systolic blood pressure (mm Hg) Peak diastolic blood pressure (mm Hg) Resting heart rate (beats/min) Peak heart rate (beats/min) VO 2, oxygen consumption.

4 4 Bard et al. ARTICLE IN PRESSThe Journal of Heart and Lung Transplantation Table 2. Results of the Different Evaluations of Ventilatory Efficiency Mean SD Range VE/VCO 2 slope to PkEx Lowest average of VE/VCO 2 ratios Peak VE/VCO 2 ratio VE/VCO 2 ratio at VT VE/VCO 2 slope to VT VE/VCO 2, ventilatory equivalent for Carbon Dioxide, VT, ventilatory threshold. RESULTS Table 1 displays the characteristics of the patients in this study that included 28% women and 72% men with mean age of years, left ventricular ejection fraction of , and peak VO 2 of ml kg 1 min 1 at an RER of Exercise tests were performed on 355 patients during the study. Two patients VE/VCO 2 slope to PkEx data were omitted from the analyses because of outlying values (values of 136 and 101 compared with a range of 22 to 77, with a median of 36 for the remaining subjects). Follow-up events included 12 patients with urgent cardiac transplantation, 17 patients with elective cardiac transplantation, and 133 patients who died. The medication profile at the time of the exercise test was loop diuretics (n 355, 100%), ACE inhibitors (n 305, 86%), angiotensin receptor blockers (n 29, 8%), digoxin (n 280, 79%), and -blockers (n 91, 26%). There was no difference in VE/VCO 2 slope to PkEx for those taking each medication. For example, the presence or absence of an ACE inhibitor (slope with, ; without, ) or a -blocker (with, ; without, ) did not reveal different peak VE/VCO 2 slopes. Table 2 summarizes the values for the ventilatory efficiency variables under investigation. Of the 355 patients, 305 patients achieved VT (86%). Overall, the mean VE/VCO 2 slopes were significantly greater at PkEx than at VT in the 305 patients who achieved VT (paired samples t-test, p 0.001). The VE/VCO 2 slope to PkEx was greater than the VE/VCO 2 slope at VT in 80% of the cases; the average absolute difference between the 2 slopes was The correlation between VE/VCO 2 slopes at PkEx and at VT was The VE/VCO 2 ratio was also significantly greater at PkEx than at VT in the 305 patients who achieved VT (paired samples t-test, p 0.001). The VE/VCO 2 ratio at PkEx was greater than the VE/VCO 2 ratio at VT in 75% of the cases; the average absolute difference between the 2 ratios was The ventilatory efficiency method using the average of the 3 lowest VE/VCO 2 ratios was similar to the ratio at VT (Table 2). Of every variable, only resting systolic blood pressure and resting heart rate had functional relationships that led to fitting these variables as dichotomous (threshold effects) rather than continuous variables. The base model that best predicted mortality included resting systolic blood pressure 100 mm Hg, resting heart rate 80 beats/min, gender, etiology, peak VO 2, and left ventricular ejection fraction. The predictive ability of all variables in the model did not change when status 1A transplants were considered as a death instead of a censored event. For example, the hazard ratio for the VE/VCO 2 slope to PkEx was for both scenarios, corresponding to a hazard ratio of 1.2 for a 10 U increase in the VE/VCO 2 slope to PkEx (approximately 1 SD). Every measure of ventilatory efficiency was a significant predictor of mortality by univariate Cox regression analysis, including the VE/VCO 2 slope to PkEx, VE/ VCO 2 slope at VT, VE/VCO 2 ratio at VT, peak VE/VCO 2 ratio at PkEx, and the lowest average of VE/VCO 2 ratios (all p 0.001). When each ventilatory efficiency parameter was considered individually with the base model, the VE/VCO 2 slope to PkEx gave the best mortality prediction based on the overall chi-square and its individual Wald chi-square value (Table 3). Patients who did not achieve VT had significantly higher VE/VCO 2 slopes than patients who achieved VT Table 3. Comparison of the Predictive Ability Between Different Ventilatory Efficiency Measures* VT achieved All Tests Model 2 Wald chi-square (for each variable) p Model 2 Wald chi-square (for each variable) p VE/VCO 2 slope to PkEx Lowest average of VE/VCO 2 ratios Peak VE/VCO 2 ratio VE/VCO 2 ratio at VT na na na VE/VCO 2 slope to VT na na na VE/VCO 2, ventilatory equivalent for Carbon Dioxide. *Results of Cox regression analysis when considered with the base model. The statistical computations were made with the ventilatory efficiency variable of interest, and adjusted for dichotomous resting systolic blood pressure 100 mm Hg, dichotomous resting heart rate 80 beats/min, sex, etiology, peak VO 2, and left ventricular ejection fraction. The data are separated out by the attainment of ventilatory threshold (VT) to evaluate the effect of exercise intensity for different ventilatory efficiency techniques. Subjects in the All Tests comparison did not necessarily achieve ventilatory threshold. Variables are listed in order according to their predictive ability based on the Wald chi-square.

5 The Journal of Heart and Lung Transplantation ARTICLE IN PRESS Bard et al. 5 ( vs , p by independent samples t-test). Also, mortality was marginally different for patients (n 51) who did not achieve VT compared with those (n 305) who achieved VT (p Cox regression). Mortality was similar at 1 year after the exercise test (11% mortality for those achieving VT vs 12% for not attaining VT) but 2-year mortality was increased for patients who did not achieve VT compared with patients who achieved VT (30% vs 19% mortality). Mortality remained higher in those who did not achieve VT through Year 5. In this sub-group of 51 patients who did not achieve VT, the VE/VCO 2 slope to PkEx was higher among the patients who died ( ) vs those who survived ( ) and a nearly significant predictor of mortality by Cox regression (hazard ratio, 1.014; p 0.074). In a bivariate Cox regression model, both peak VO 2 (p 0.004) and VE/VCO 2 slope to PkEx (p 0.001) were statistically significant. The predictive ability of the peak slope was the same when peak VO 2 was dichotomized at a cut point of 14 ml kg 1 min 1. There was no threshold in the VE/VCO 2 slope to PkEx because categorizing the VE/VCO 2 slope into tertiles or quartiles resulted in linearly increasing coefficients. The effect of the VE/VCO 2 slope to PkEx on mortality was independent of exercise intensity. When data were cut at quartiles of peak RER ( 1.050, , , 1.210), the VE/VCO 2 slope to PkEx was a better mortality predictor than peak VO 2 in each quartile. Peak VO 2 was not significant in these quartiles, partly because peak VO 2 and peak RER were significantly correlated (r 0.22, p 0.001). DISCUSSION This study demonstrated that the VE/VCO 2 slope to PkEx was the most predictive means of expressing ventilatory efficiency and the most predictive variable of mortality among the 20 variables that were considered. The VE/ VCO 2 slope to PkEx and the sub-maximal VE/VCO 2 slope to VT differed both in value and prognostic power. In a large cohort of heart failure patients, the average VE/VCO 2 slope to PkEx was significantly greater than the average sub-maximal VE/VCO 2 slope to VT. Although other calculations of ventilatory efficiency were significant univariate predictors of mortality, the VE/VCO 2 slope to PkEx was the superior mortality predictor compared with other mortality predictors in a multivariable model (Table 3). This study also demonstrated that ventilatory efficiency is an exceptionally powerful and independent predictor of mortality compared with other predictors, including peak VO 2, which is the only variable obtained from cardiopulmonary exercise testing that is commonly used to assess mortality risk and to triage patients for cardiac transplantation. Maximal exercise testing has superior prognostic ability compared with sub-maximal exercise testing, but patients who did not achieve VT also had a high incidence of mortality. Perhaps these patients were unable to exercise to VT because of advanced illness, as suggested by an elevated VE/VCO 2 slope to PkEx. Mortality risk assessment was improved with maximal testing, but an inability to achieve VT while having a high VE/VCO 2 slope despite encouragement during the exercise test was associated with a poor prognosis. Comparisons with the Literature Contrary to other reports, 1,7,18,19 the VE/VCO 2 slope was not independent of exercise intensity. It is unclear why the slope has been considered to be the same at maximal or sub-maximal exercise, although this belief may derive from conclusions from a 1992 study by Metra et al 1 where sub-maximal and maximal VE/VCO 2 slopes were measured in 68 patients with heart failure. Sub-maximal exercise was defined as the first 30% and 60% of the exercise test. The authors reported a significant correlation between the peak slope and both sub-maximal slopes (r 0.62 for 30%, r 0.83 for 60%, both p 0.001) and concluded that the slope did not change throughout exercise, despite slopes reported as for PkEx, for the first 30% of exercise, and 32 9 for the first 60% of exercise. This study did not prove that submaximal and maximal VE/VCO 2 slopes are equivalent but only showed that a relationship existed by definition, since a high correlation is expected because the calculation of the VE/VCO 2 slope to PkEx contains all of the data from the calculations of sub-maximal slopes. Gitt et al 4 did not compare different determinations of ventilatory efficiency but found that the VE/VCO 2 slope to VT was a powerful predictor of mortality independent of peak VO 2 in a multivariable model. Our results indicate that the predictive ability of ventilatory efficiency would be improved by using the VE/VCO 2 slope to PkEx. The details of the calculations of ventilatory efficiency are important because there were statistical and prognostic differences between slopes and ratios as well as between peak and sub-maximal exercise. Ratios and slopes should not be used interchangeably, because we showed a considerable prognostic difference. In addition, Sun et al 12 reported that slopes and ratios are equivalent only 5% of the time. The measurement of the slope is preferable to other measures of ventilatory efficiency because it represents a trend over the exercise duration and eliminates outliers or excessively large VE/VCO 2 ratios that may be associated with increased hyperventilation during respiratory compensation. Our study confirms and extends the findings of Arena et al, 13 Tabet et al, 14 and Gitt et al. 4 Although 2 of these studies 13,14 reported significant prognostic differences between peak and sub-maximal slopes, the studies were

6 6 Bard et al. ARTICLE IN PRESSThe Journal of Heart and Lung Transplantation relatively small and unable to assess the contribution of ventilatory efficiency to other variables in a model predicting mortality. Compared with Arena et al, 13 our study was better controlled, and the heart failure patients were more ill. Our testing was performed by 1 investigator at 1 site over a relatively short period (1.5 years), whereas Arena et al 13 deployed multiple technicians and equipment at 2 sites over 8 years. Although the average peak VO 2 s were similar, Arena et al 13 reported lower mean peak VE/VCO 2 slopes (33 vs 37) and had 33% of the mortality rate (22 deaths/174 patients 13% vs 135 deaths/355 patients 38%) despite an older population (51 vs 56 years). Tabet et al 14 evaluated the prognostic ability of peak and sub-maximal slopes in 97 heart failure patients where sub-maximal exercise was defined as the first 3 minutes of exercise. Mortality rates were closer to those of the current study (32% vs our 38%). Interestingly, both peak VO 2 ( , ml kg 1 min 1 ) and the peak slope (47 vs 42) of non-survivors was greater than those of survivors. Despite the differences between studies, we agree with the conclusions of Tabet et al, 14 who state that VE/VCO 2 slopes should be calculated with all of the exercise data because valuable prognostic information is obtained during the last minutes of exercise. The patients studied by Gitt et al 4 differed from our patients. There was a large average age difference of more than 12 years (our 50.6 vs their 62.9 years) and the mode of exercise (our treadmill vs their bicycle) varied, which can explain differences in both the peak VO 2 and RER attained. Each patient in this study was being triaged for cardiac transplantation, and the sample is representative of that population. To ensure patients were clinically similar, they were only included if they were taking a loop diuretic and had a documented left ventricular ejection fraction of less than 40%. Patients did not have significant comorbidities that would affect their prognosis. Compared with other prognostic VE/ VCO 2 studies in patients with heart failure, the current study considered multivariable modeling, had the longest average follow-up (3.8 years, 5 years since the exercise test), and included a large sample size (355). The ratio of patients taking vs not taking -blockers at the time of exercise testing was 1:3, and we did not note any significant effect of -blockers on mortality or ventilatory efficiency, which agrees with 1 prospective study 21 involving heart failure patients who underwent exercise testing before and after -blocker treatment. Our study, like many studies, was not designed to address medications. One major limitation in studies that claim to investigate medications is that the medications are not prospectively tracked, rather they are the medications present at the time of the exercise test. 22 It is not appropriate to make conclusions unless the medications are closely tracked at short intervals. This was especially true with heart failure patients during this study because of the initiation of the -blocker era and the fact that many different medications are prescribed by a variety of different physicians. Exercise Intensity There is no threshold value of VE/VCO 2 slope to PkEx that clinicians can use to quantify patients as having a high or a low slope that correlates with mortality risk, because greater slopes were associated with greater risk. Low and high values of VE/VCO 2 slope have been proposed 2 and implemented 4 but these values were based on values outside of a normal range for healthy subjects and these proposed threshold values have not been shown to predict morbidity or mortality in any patient cohort. Regardless of the exercise intensity achieved, the VE/VCO 2 slope to PkEx was always a better mortality predictor than peak VO 2. This statement does not mean that peak and submaximal slopes are the same, it means that the VE/VCO 2 slope of all of the available data in any exercise test was always a powerful mortality predictor, independent of peak VO 2. This information may help the clinician in determining prognosis in patients with sub-maximal exercise data. Despite valiant efforts by the clinician, patients who are unable to achieve VT have a very poor prognosis if their VE/VCO 2 slope is elevated, but it is best to try to obtain a quality maximal test since maximal slopes are superior predictors of mortality than are sub-maximal slopes. Nevertheless, peak VO 2 remains an independent predictor of mortality and Figure 2 illustrates that the combination of VE/VCO 2 slope to PkEx and peak VO 2 is a powerful method of stratifying mortality risk in patients with heart failure. Figure 2. Survival risk determined by grouping patients into 4 different groups based on the median values for peak oxygen consumption (VO 2 ) (16.5 ml kg 1 min 1 ) and the ventilatory equivalent for carbon dioxide (VE/VCO 2 ) slope to peak exercise (PkEx) (35.6), and displayed with Kaplan-Meier survival curves.

7 The Journal of Heart and Lung Transplantation ARTICLE IN PRESS Bard et al. 7 These data do not support the use of sub-maximal exercise testing for prognostic purposes. The studies that evaluated sub-maximal and maximal VE/VCO 2 slopes in heart failure patients all agree that the best prognostic data are obtained from PkEx, despite the differences between the studies. Mezzani et al 23 further support the use of maximal exercise testing, since exercise intensities with an RER 1.15 were superior mortality predictors to tests with an RER Based on these data, exercise test technicians should strive to achieve maximal exercise in all patients and repeated efforts should be attempted in patients who do not achieve maximal exercise to best assess prognosis and to stratify for cardiac transplantation because predictive power is greatest when PkEx is obtained. CONCLUSION The VE/VCO 2 slope to PkEx is the most powerful predictor of mortality within a multivariable model, and it is the best method of determining ventilatory efficiency. These results support the utility and superiority of maximal compared with sub-maximal exercise testing for prognostic purposes in patients with heart failure. Regardless of the exercise intensity achieved, the VE/VCO 2 slope to PkEx was the best predictor of mortality; therefore, this variable should be considered in consensus statements involving the evaluation of heart failure patients for cardiac transplantation evaluation. The authors thank Mona Vekaria, Nhien Duong, Marguerite Wickman, and the University of Michigan Undergraduate Research Opportunities Program for assistance with data collection. REFERENCES 1. Metra M, Dei Cas L, Panina G, et al. Exercise hyperventilation chronic congestive heart failure, and its relation to functional capacity and hemodynamics. Am J Cardiol 1992; 70: Chua TP, Ponikowski P, Harrington D, et al. Clinical correlates and prognostic significance of the ventilatory response to exercise in chronic heart failure. J Am Coll Cardiol 1997;29: MacGowan GA, Janosko K, Cecchetti A, et al. Exerciserelated ventilatory abnormalities and survival in congestive heart failure. Am J Cardiol 1997;79: Gitt AK, Wasserman K, Kilkowski C, et al. Exercise anaerobic threshold and ventilatory efficiency identify heart failure patients for high risk of early death. Circulation 2002;106: Bard RL. Cardiopulmonary exercise testing in patients with heart failure [Correspondence]. J Am Coll Cardiol 2005;45: de Groote P, Dagorn J, Soudan B, Lamblin N, Mc Fadden E, Bauters C [Author reply]. J Am Coll Cardiol 2005;45: Dimopoulou I, Tsintzas OK, Alivizatos PA, et al. Pattern of breathing during progressive exercise in chronic heart failure. Int J Cardiol 2001;81: Cicoira M, Zanolla L, Franceschini L, et al. Skeletal muscle mass independently predicts peak oxygen consumption and ventilatory response during exercise in noncachectic patients with chronic heart failure. J Am Coll Cardiol 2001;37: Ponikowski P, Francis DP, Piepoli MF, et al. Enhanced ventilatory response to exercise in patients with chronic heart failure and preserved exercise tolerance: marker of abnormal cardiorespiratory reflex control and predictor of poor prognosis. Circulation 2001;103: Robbins M, Francis G, Pashkow FJ, et al. Ventilatory and heart rate responses to exercise: better predictors of heart failure mortality than peak oxygen consumption. Circulation 1999;100: de Jonge N, Kirkels H, Lahpor JR, et al. Exercise performance in patients with end-stage heart failure after implantation of a left ventricular assist device and after heart transplantation: an outlook for permanent assisting? J Am Coll Cardiol 2001;37: Sun XG, Hansen JE, Garatachea N, et al. Ventilatory efficiency during exercise in healthy subjects. Am J Respir Crit Care Med 2002;166: Arena R, Myers J, Aslam SS, et al. Technical considerations related to the minute ventilation/carbon dioxide output slope in patients with heart failure. Chest 2003;124: Tabet JY, Beauvais F, Thabut G, et al. A critical appraisal of the prognostic value of the VE/VCO2 slope in chronic heart failure. Eur J Cardiovasc Prev Rehab 2003;10: Bard RL, Nicklas JM. New graphical method for evaluating gas exchange in congestive heart failure. Med Sci Sports Exerc 2000;32: Beaver WL, Wasserman K, Whipp BJ. A new method for detecting anaerobic threshold by gas exchange. J Appl Physiol 1986;60: Wasserman K, Hansen JE, Sue DY, Whipp BJ, Casaburi R. Principles of exercise testing and interpretation, 2nd ed. Philadelphia: Lea and Febiger, p Mudge GH, Goldstein S, Addonizio LJ, et al. Twentyfourth Bethesda conference: cardiac transplantation, Task Force 3: recipient guidelines/prioritization. J Am Coll Cardiol 1993;22: Marubini E, Valsecchi MG. Analysing survival data from clinical trials and observational studies. New York: John Wiley and Sons, p Kleber FX, Vietzke G, Wernecke KD, et al. Impairment of ventilatory efficiency in heart failure: prognostic impact. Circulation 2000;101: Witte KK, Thackray S, Nikitin NP, Cleland JG, Clark AL. The effects of long-term beta-blockade on the ventilatory responses to exercise in chronic heart failure. Eur J Heart Failure 2005;7: Bard RL. Letter regarding article by O Neill et al, Peak oxygen consumption as a predictor of death in patients with heart failure receiving beta-blockers Circulation 2005;112:e Mezzani A, Corra U, Bosimini E, et al. Contribution of peak respiratory exchange ratio to peak VO 2 prognostic reliability in patients with chronic heart failure and severely reduced exercise capacity. Am Heart J 2003;145: