| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Department of Physical Therapy (Drs. Arena and Peberdy), Virginia Commonwealth University, Medical College of Virginia, Health Sciences Campus; and the Cardiology Division (Drs. Myers, Aslam, and Varughese), Veterans Affairs Palo Alto Health Care System, Stanford University, Palo Alto, CA.
Correspondence to: Ross Arena, PhD, PT, Assistant Professor, Department of Physical Therapy, Box 980224, Virginia Commonwealth University, Medical College of Virginia, Health Sciences Campus, Richmond, VA 23298-0224; e-mail: raarena{at}mail2 vcu.edu
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
E)-carbon dioxide output (CO2) relationship has recently been demonstrated to have prognostic significance in the heart failure (HF) population. However, the method by which the E/CO2 slope is expressed has been inconsistent.
Methods: One hundred eighty-eight subjects, who had received diagnoses of HF, underwent exercise testing. Two E/CO2 slope calculations were made, one using exercise data prior to the ventilatory threshold (VT), and one using all data points from rest to peak exercise. Four separate peak exercise E/CO2 slope calculations also were derived with unaveraged, 10-s, 30-s, and 60-s ventilatory expired gas sampling intervals.
Results: Although univariate Cox regression analysis demonstrated pre-VT and peak E/CO2 slope calculations to both be significant predictors of cardiac-related mortality and hospitalization (p < 0.001), the peak classification scheme was significantly better (p < 0.01). The ventilatory expired gas-sampling interval that was used did not impact the predictive ability of the peak E/CO2 slope.
Conclusion: Although both the pre-VT and peak E/CO2 slope calculations were prognostically significant, the peak expression was superior. The sampling interval did not appear to have a significant impact on prognostic utility. We hope that the results of the present study will contribute to the standardization of the E/CO2 slope and will enhance its clinical application.
Key Words: cardiac-related hospitalization • cardiac-related mortality • exercise testing • ventilatory expired gas
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
E)-carbon dioxide output (CO2) relationship, which is commonly expressed as the E/CO2 slope, continues to mount in the heart failure (HF) population.1
2
3
4
5
6
7
8
9
10
11
12
This noninvasive measure is significantly correlated with aerobic capacity,1
2
13
cardiac function,1
and pulmonary perfusion.8
14
The E/CO2 slope has been shown consistently to be a significant predictor of mortality4
5
6
7
9
12
15
and hospitalization3
in patients with HF. Interestingly, the evidence to date has suggested that the E-CO2 relationship may be superior to the more established ventilatory expired gas measure peak oxygen output (O2) in predicting outcome.3
4
7
12
16
While evidence demonstrating the clinical value of the E/CO2 slope is now robust, investigation into technical aspects, which would help to standardize the expression of this measure, has been lacking. This may be one reason why peak O2, a measure in which the expression is well-standardized,17
18
remains the exercise test "gold standard" for assessing prognosis in the HF population. A primary area of debate involves whether to calculate the E/CO2 slope using only exercise data prior to the onset of the ventilatory threshold (VT)15
/acidotic drive to ventilation6
12
or to include all data points from the onset of exercise to peak exercise.3
4
5
7
The hypothesized reason for not using data points beyond the VT is that there is an acidosis-induced dislinearity in the E-CO2 relationship and that the inclusion of this portion has no clinical value. Both techniques, however, consistently produce a prognostically significant marker, although a direct comparison of pre-VT/acidotic ventilatory drive and all-inclusive E/CO2 slope calculations has yet to be performed. Moreover, a compelling theoretical rationale for choosing one calculation method over the other has not been proposed. A second area lacking exploration is the effect that different ventilatory expired gas-sampling intervals have on the E/CO2 slope calculation and its prognostic value. Previous investigations have been reported using unaveraged (ie, breath-by-breath),1
10-s,3
7
6
and 30-s5
averaged sampling intervals to calculate this relationship. Again, while the aforementioned studies have consistently demonstrated the significant prognostic value of the E/CO2 slope, the potential for an optimal sampling interval has not been explored.
Addressing these technical issues would lead to standardizing the expression of this variable and potentially to improving its clinical utility. The purpose of the present investigation was to compare the prognostic ability of the E/CO2 slope calculated as follows: (1) using data from the initiation of exercise to the point of VT (ie, pre-VT) and from the initiation of exercise to peak exercise; and (2) using different ventilatory expired gas-sampling intervals.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Equipment Calibration
Ventilatory expired gas analysis was obtained by one of several metabolic systems depending on the clinic and time frame for exercise testing (at the Medical College of Virginia Hospital: Vmax29, SensorMedics, Yorba Linda, CA; Medical College of Virginia Hospital and Veterans Affairs Hospital: CPX-D, Medgraphics, Minneapolis, MN; Veterans Affairs Hospital: CS-100, Schiller, Baar, Switzerland; and Orca Diagnostics, Santa Barbara, CA). The oxygen and carbon dioxide sensors were calibrated using gases with known oxygen, nitrogen, and carbon dioxide concentrations prior to each test. The flow sensor also was calibrated before each test. Testing was not conducted unless the ventilatory expired gas system passed calibration. Both centers followed similar calibration procedures. The basic principles governing the function of modern-day ventilatory expired gas systems are similar. Given that system calibration was conducted prior to each test, equipment bias was not considered to be a limitation.
Testing Procedure and Data Collection
Symptom-limited exercise tests with ventilatory expired gas analysis were performed using either a treadmill or a cycle ergometer. The treadmill was the only mode of exercise used at the Medical College of Virginia Hospital. The Veterans Affairs Hospital likewise used a treadmill in the majority of exercise tests (ie, approximately 90%). The mode of exercise was dependent on equipment availability as opposed to subject characteristics. Furthermore, there is no hypothetical rationale to support the possibility that the relationship between E and CO2 is dependent on exercise mode in a given subject. Both centers solely employ ramping protocols for exercise testing in the HF population, which were similar in stage time and incremental workload adjustments. Monitoring consisted of continuous ECG measurements, manual BP measurements, heart rate recordings at every stage via an ECG, and rating of perceived exertion (Borg scale, 6 to 20) at each stage. The testing procedures were explained prior to testing, and subjects were encouraged to give a maximal effort. The intention of explaining the procedures was to alleviate any apprehension a given subject may have had in putting forth a maximal effort. During testing, subjects were given further verbal encouragement to put forth a maximal effort by the clinician conducting the test. Test termination criteria were in accordance with American College of Sports Medicine guidelines.18
Monitoring and test termination guidelines were similar at both centers.
Data for O2, CO2, and E were collected continuously throughout the exercise test. All 188 subjects had these variables, averaged at 
10-s intervals, either in hard copy or computerized format. The 10-s averaged E/CO2 and E/O2 data were input into a spreadsheet (Excel; Microsoft; Redmond, WA), and the VT was determined by the ventilatory equivalent method.19
Two separate E/CO2 slope calculations were derived, one using data from the initiation of exercise to the point of the VT (ie, pre-VT E/CO2 slope), and the other using data from the initiation of exercise to peak exercise (ie, peak E/CO2 slope). Fourteen subjects did not demonstrate a detectable VT and were excluded from this analysis. The difference between pre-VT and peak E/CO2 slope was also calculated. One hundred and fifty-three subjects had data maintained on the computer in a format permitting four separate E/CO2 slope calculations; unaveraged (breath-by-breath), 10-s, 30-s, and 60-s averaged sampling intervals. The E/CO2 slope was again determined by inputting E and CO2 data for each of the ventilatory expired gas sampling intervals into a spreadsheet (Excel; Microsoft). Only the peak E/CO2 slope calculation was used for the comparison of sampling intervals (the rationale in the "Discussion" section). All E/CO2 slope calculations were derived via least squares linear regression (ie, y = mx + b, m = slope). To improve the uniformity of the data, the primary investigator (RA) was the sole processor of exercise test data from both centers.
End Points
Subjects were followed up for cardiac-related mortality and hospitalization 2 years following exercise testing via medical chart review. Subjects not regularly observed by the Medical College of Virginia Hospital or the Veterans Affairs Hospital were excluded from the study to help ensure that end points of interest were not overlooked. Cardiac-related mortality was also separately tracked without a 1-year time constraint. Any death or hospital admission that was precipitated by cardiac dysfunction, as per the hospital discharge diagnosis, was considered to be an event. The most common causes of mortality, as per the hospital discharge diagnosis, were cardiac arrest, myocardial infarction, and end-stage HF. The most common causes of hospitalization were decompensated HF and coronary artery disease. Subjects in whom mortality was of a noncardiac etiology were treated as censored cases. To ensure uniformity from both centers, the primary investigator (RA) collected all end point data.
Statistical Analysis
Paired t tests were used to compare differences between pre-VT and peak E/CO2 slopes. Repeated-measures analysis of variance (ANOVA) with Tukey honest significant difference post hoc testing was used to compare differences among unaveraged, 10-s, 30-s, and 60-s averaged E/CO2 slope calculations.
Univariate Cox regression analysis was used to determine the ability of peak E/CO2 slope, pre-VT E/CO2 slope, pre-VT-peak E/CO2 slope difference, and E/CO2 slope calculations, using different sampling intervals (ie, unaveraged, 10 s, 30 s, and 60 s), to predict the 1-year cardiac-related hospitalization rate, the 1-year cardiac-related mortality rate, and the overall cardiac-related mortality rate.
Receiver operating characteristic (ROC) curves were constructed for the E/CO2 slope calculations described above. A z test compared the pre-VT to the peak E/CO2 slope classification schemes and the four sampling interval classification schemes.20
The E/CO2 slope was assessed as a continuous variable throughout. All statistical tests with a p value of < 0.05 were considered to be significant.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
O2 and respiratory exchange ratio were 16.6 ± 8.0 mL/kg/min and 1.1 ± 0.2, respectively. The mean pre-VT and peak E/CO2 slope values were 29.9 ± 6.8 and 32.5 ± 8.0, respectively. The mean difference between these calculations was 2.6 ± 4.1. The difference between the pre-VT and peak E/CO2 slope calculations was statistically significant (p < 0.0001). The ranges for the pre-VT and peak E/CO2 slopes were 17.0 to 55.1 and 16.7 to 68.6, respectively. The E/CO2 slope, irrespective of the calculation method, was expressed exclusively as a continuous variable in all of the following prognostic assessments.
Univariate Cox regression analysis for the pre-VT E/CO2 slope, the peak E/CO2 slope, and the pre-VT-peak E/CO2 slope difference revealed that all were significant predictors of the overall cardiac-related mortality rate, the 1-year cardiac-related mortality rate, and the 1-year cardiac-related hospitalization rate (Table 2
).
|
E/CO2 slope classification schemes are listed in Table 3 . A z test revealed that the peak E/CO2 slope classification scheme was superior to the pre-VT E/CO2 slope classification scheme for all three end points (p < 0.0005).
|
E/CO2 slope calculation, all data-sampling comparisons were performed using the peak E/CO2 slope. The mean values for unaveraged, 10-s, 30-s, and 60-s averaged peak E/CO2 slope calculations were 31.8 ± 8.1, 32.7 ± 8.4, 32.7 ± 9.1, and 32.5 ± 8.9, respectively. Repeated-measures ANOVA revealed that there was a significant difference among the four calculations (p = 0.005). The unaveraged peak E/CO2 slope calculation was significantly less than the other three averaged calculations (p < 0.05 [Tukey honest significant difference test]), whereas the other methods did not differ from one another.
Univariate Cox regression analysis revealed the unaveraged, 10-s, 30-s, and 60-s averaged peak E/CO2 slope calculations to be significant predictors of the overall cardiac-related mortality rate, the 1-year cardiac-related mortality rate, and the 1-year cardiac-related hospitalization rate (Table 4
).
|
E/CO2 slope classifications are listed in Table 5
. A z test revealed that the peak E/CO2 slope classifications were not significantly different for any of the end points.
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
E/CO2 slope calculations.3
4
5
7
Evidence indicating an elevated E/CO2 slope, when present in a patient with HF, manifests itself at rest and continues into exercise.1
21
22
The fact that this ventilatory expired gas abnormality occurs early and does not normalize at any point during exercise may be a primary reason why both pre-VT and peak E/CO2 slope calculations are significant predictors of outcome. An argument has been raised regarding the calculation of the E/CO2 slope using only the linear (pre-VT/acidosis) portion of the data, since after the VT there is a nonlinear increase in ventilation. The alternative is to express the E/CO2 relationship using a best-fit line between all the data points during exercise, since this would include the hyperventilatory response and potentially would produce a relatively higher slope in patients who exhibit, for example, early metabolic acidosis and ventilation/perfusion mismatching.
While both calculations were significant predictors of mortality and hospitalization in the present study, the peak E/CO2 slope calculation was clearly a superior predictor compared to its pre-VT counterpart, as indicated by the ROC curve comparisons in Table 3
. The significant prognostic difference between the pre-VT and peak E/CO2 slopes prompted the investigators to assess the value of the difference between these measures independently. It was discovered that the greater the increase in E/CO2 slope from pre-VT to peak, the worse the outcome for a given subject. The fact that the difference between the pre-VT and the peak E/CO2 slope calculations is in itself a significant predictor of mortality and hospitalization (Table 2)
suggests that capturing the physiologic response from VT to peak exercise is clinically important with respect to this measure. Figure 1
illustrate the unique responses of two subjects from this group.
|
E/CO2 slope calculations shown in Figure 1
, top, were identical and fell within the normal range (slope, 26). This subject did not experience a cardiac-related event during the follow-up period (ie, mortality or hospitalization). The pre-VT and peak E/CO2 slope calculations shown in Figure 1
, bottom, were, however, significantly different, with the pre-VT calculation falling within normal limits (slope, 27), while the peak calculation was elevated (slope, 37). The subject illustrated in Figure 1
, bottom, died from cardiac complications 4 months following the exercise test. In this instance, the peak E/CO2 slope calculation appropriately identified this subject as being at high risk, while the pre-VT calculation did not.
Previous evidence has indicated that an elevated E/CO2 slope is at least partly attributable to heightened ventilation associated with reduced cardiac output and thus to poor pulmonary perfusion in HF.8
14
Reindl et al1
reported a significant inverse correlation (r = -0.66) between the E/CO2 slope (measured to peak exercise) and cardiac output at rest in a group of patients with HF. One reason that the peak E/CO2 slope was a superior predictor of outcome in the present study may be that individuals with poor outcome exhibit further declines in cardiac function and therefore pulmonary perfusion at high levels of exercise (eg, from VT to peak exercise). In this situation, a E/CO2 slope calculated to peak exercise would better reflect prognosis because it captures this decline in central function, which is missed by the pre-VT/acidosis expression.
The E/CO2 slope is commonly calculated via linear regression, as was the case in this investigation. Figure 1
, bottom, does, however, illustrate the potential for this relationship to become dislinear when calculated to peak exercise in those individuals with poorer prognoses. This finding may support the use of nonlinear regression to better express the E/CO2 slope, although applying such a calculation uniformly would be difficult given that a number of subjects with HF demonstrate a linear response from the onset of exercise to peak exercise. The prognostic superiority of the peak E/CO2 slope calculated via linear regression does, however, support its general use over pre-VT/acidosis calculations, which was a primary objective of this study. Comparing the prognostic efficacy of linear and nonlinear regression techniques is beyond the scope of this investigation and should be an avenue for future research.
It has been suggested that breath-by-breath data should be used to calculate the E/CO2 slope, given that the greater density of data and greater precision potentially provide a truer reflection of the E/CO2 relationship. In clinical practice, few laboratories acquire breath-by-breath data during exercise, and the majority of the prognostic studies have employed various averaged samples to express the E/CO2 slope. We observed that the sampling interval chosen for the E/CO2 slope had a minimal impact on either the value obtained or its prognostic utility. Repeated-measures ANOVA indicated that the mean breath-by-breath E/CO2 slope calculation was slightly, but significantly, less that its averaged counterparts. This difference may be attributed to the inherent fluctuation associated with breath- by-breath data compared to averaged calculations. All sampling interval classification schemes were, however, similar in terms of predicting end points (Table 5)
, suggesting that the difference detected by repeated-measures ANOVA has an insignificant impact on the prognostic accuracy and clinical value of the E/CO2 slope. The data from previous investigations3
4
7
11
using different sampling intervals to calculate the E/CO2 slope may therefore be comparable.
Overall cardiac-related mortality without a defined time period was used as the end point in the present study in order to make it consistent with previous investigations assessing the prognostic utility of the E/CO2 slope5
6
12
15
This was considered important, given that the conception of the present study was partially based on the findings of these investigations. Limiting the end points tracked to a 1-year time period may, however, be clinically optimal given the fluid nature of cardiac function in HF. Because the stability of HF can change quickly, limiting the follow-up period to 1 year makes the results more relevant for clinical practice. A 1-year tracking period may strike a sufficient balance between avoiding outdated information and the economic constraints of multiple exercise tests. Additionally, most research examining the prognostic value of ventilatory expired gas variables has not used hospitalization as an end point. Given that HF is the primary hospital diagnostic-related group among Medicare patients,23
analysis of the measures predicting hospitalization in this population seems to be warranted. The ability of the E/CO2 slope to effectively predict the risk of hospitalization may provide clinicians the unique opportunity to identify high-risk patients, to provide appropriate interventions, to prevent hospitalization, and to reduce the cost of care. Last, it is the author’s opinion that the E/CO2 slope, being closely related to cardiac function, is a better predictor of cardiac-related events as opposed to all-cause events. Variables from an exercise test should not be viewed as "all-inclusive" predictors. One would not expect an individual with cancer of the liver and normal cardiac function to have an elevated E/CO2 slope. While prognosis may be poor, the E/CO2 slope would not reflect this.
Enthusiasm for the use of the E/CO2 slope has evolved faster than the methodological issues related to it have been clarified. Future investigations should be conducted to confirm or refute the present findings. An avenue for future research may entail the retrospective analysis of the data collected by other institutions that have examined the prognostic ability of the E/CO2 relationship in the HF population.4
5
6
7
11
12
15
Duplication of the methods of the present study should not be difficult given the computerized nature of current metabolic measuring systems. Such investigations will either lend support or challenge the findings of the present study and are needed to reach a consensus on these technical issues. A second area requiring prospective analysis relates to the proposed hypothesis suggesting that a further decline in cardiac function from VT to peak exercise is captured only by a peak E/CO2 slope calculation and is the reason for its prognostic superiority compared to the pre-VT method. Establishing a relationship among cardiac function, pulmonary perfusion, and the E/CO2 slope at high levels of exercise is required. A prognostic analysis of post-VT changes in cardiac output and pulmonary perfusion also would be useful to determine the mechanism for the superiority of the peak E/CO2 slope observed in the present study.
In conclusion, the results of the present study lay the groundwork for standardizing the E/CO2 slope. The E/CO2 slope, using data from rest to peak exercise, has greater prognostic power than pre-VT data in patients with HF. The ventilatory expired gas sampling interval used for the E/CO2 slope calculation appears to have little impact on its prognostic utility. It is hoped that this investigation will compel other groups to perform similar investigations leading to the standardization of the E/CO2 slope and to wider clinical application of this measure.
|
Footnotes |
|---|
CO2 = carbon dioxide output; E = minute ventilation; O2 = oxygen output; VT = ventilatory threshold Received for publication February 11, 2003. Accepted for publication February 12, 2003.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
E/VCO(2)slope and peak VO(2). Eur Heart J 2000;21,154-161This article has been cited by other articles:
|
|
 |
|
  M. Guazzi, J. Myers, and R. Arena Cardiopulmonary Exercise Testing in the Clinical and Prognostic Assessment of Diastolic Heart Failure J. Am. Coll. Cardiol., November 15, 2005; 46(10): 1883 - 1890. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Arena, J. Myers, J. Abella, and M. A. Peberdy Influence of Heart Failure Etiology on the Prognostic Value of Peak Oxygen Consumption and Minute Ventilation/Carbon Dioxide Production Slope Chest, October 1, 2005; 128(4): 2812 - 2817. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Wonisch, P. Lercher, D. Scherr, R. Maier, R. Pokan, P. Hofmann, and S. P. von Duvillard Influence of Permanent Right Ventricular Pacing on Cardiorespiratory Exercise Parameters in Chronic Heart Failure Patients With Implanted Cardioverter Defibrillators Chest, March 1, 2005; 127(3): 787 - 793. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. L. Bard Cardiopulmonary exercise testing in patients with heart failure J. Am. Coll. Cardiol., January 4, 2005; 45(1): 163 - 163. [Full Text] [PDF]
|
|
|
|
|
|
P. d. Groote, J. Dagorn, B. Soudan, N. Lamblin, E. Mc Fadden, and C. Bauters Reply J. Am. Coll. Cardiol., January 4, 2005; 45(1): 163 - 164. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
