Determining the Preferred Percent-Predicted Equation for Peak Oxygen Consumption in Patients with Heart Failure

Abstract

Background

Peak oxygen consumption (VO₂) is routinely assessed in patients with heart failure (HF) undergoing cardiopulmonary exercise testing (CPX). The purpose of the present investigation is to determine the prognostic ability of several established peak VO₂ prediction equations in a large HF cohort.

Methods and Results

One thousand one hundred and sixty-five subjects (70% male, age: 57.0 ±13.8 years, ischemic etiology: 43%) diagnosed with HF underwent CPX. Percent-predicted peak VO₂ was calculated according to normative values proposed by Wasserman and Hansen (equation), Jones (equation), the Cooper Clinic (below low fitness threshold), a Veteran’s Administration male referral data set (four equations) and the St. James Take Heart Project for women (equation). The prognostic significance of percent-predicted VO₂ values derived from the two latter, sex-specific equations were assessed collectively (VA-St. James). There were 179 major cardiac events (117 deaths, 44 heart transplantations and 18 left ventricular assist device implantations) during the two year tracking period (annual event rate: 10%). Measured peak VO₂ and all percent-predicted peak VO₂ calculations were significant univariate predictors of adverse events (Chi-square: ≥31.9, p<0.001) and added prognostic value to ventilatory efficiency (VE/VCO₂ slope), the strongest CPX predictor of adverse events (Chi-square: 150.7, p<0.001), in a multivariate regression. The Wasserman/Hansen prediction equation provided optimal prognostic information.

Conclusions

Actual peak VO₂ and the percent-predicted models included in this analysis all were significant predictors of adverse events. It appears that the percent-predicted peak VO₂ value derived from the Wasserman/Hansen equations may outperform other expressions of this CPX variable.

Introduction

Peak oxygen consumption (VO₂) is a clinically accepted and important variable in the prognostic evaluation of patients with heart failure (HF) undergoing cardiopulmonary exercise testing (CPX).¹ The actual value of peak VO₂, typically expressed relative to body weight, is the most common approach to reporting aerobic capacity in apparently healthy individuals as well as different patient populations, including HF.² Reporting peak VO₂ as a percent-predicted value has also been advocated. Moreover, a number of different approaches to estimating normal aerobic capacity are readily available.^3-9 These prediction equations have used various independent variables such as height, weight and mode of exercise but the inclusion of age and consideration of sex is a shared commonality.

The body of evidence demonstrating the prognostic utility of the actual peak VO₂ value is robust; collectively these investigations have included thousands of patients and hundreds of adverse events, consistently demonstrating the ability of this CPX variable to identify those patients with HF at increased risk for poor outcomes.¹⁰ A limited number of investigations have also examined the prognostic value of percent-predicted peak VO₂ in HF patients with mixed results. For example, using two different prediction equations, Aaronson et al.¹¹ found neither percent-predicted peak VO₂ value was a superior prognostic marker compared to the actual value in 272 patients with HF. However, using one equation, Stelken et al.¹² reported percent-predicted peak VO₂ was superior to the actual value in predicting mortality in a separate group of 181 patients with HF.

We are unaware of any investigation that has simultaneously compared the prognostic utility of the most commonly utilized peak VO₂ prediction equations to each-other and the actual aerobic capacity value in a large HF cohort undergoing CPX within the past 10 years. Moreover, a prognostic comparison of percent-predicted VO₂ values to other important CPX measures, such as ventilatory efficiency, has not been performed. The purpose of the present investigation was to address these issues in an effort to determine the clinical relevance of expressing peak VO₂ relative to a normative value in patients with HF.

Methods

This study was a multi-center analysis including HF patients from the CPX laboratories at San Paolo Hospital, Milan, Italy, Wake Forest University Baptist Medical Center, Winston-Salem, North Carolina, USA, LeBauer Cardiovascular Research Foundation, Greensboro, North Carolina, USA, VA Palo Alto Health Care System, Palo Alto, California, USA and Virginia Commonwealth University, Richmond, Virginia, USA. A total of 1,165 patients with chronic HF were included. Inclusion criteria consisted of a diagnosis of HF¹³ and evidence of left ventricular dysfunction by two-dimensional echocardiography obtained within one month of data collection. All subjects completed a written informed consent and institutional review board approval was obtained at each institution. The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written

CPX Procedure and Data Collection

Symptom-limited CPX was performed on all subjects and pharmacologic therapy was maintained during exercise testing. Conservative ramping protocols were employed at all centers. Ventilatory expired gas analysis was performed using a metabolic cart at all five centers (Medgraphics CPX-D and Ultima, Minneapolis, MN, Sensormedics Vmax29, Yorba Linda, CA or Parvomedics TrueOne 2400, Sandy, UT). Before each test, the equipment was calibrated in standard fashion using reference gases and a three-liter syringe. Standard 12-lead electrocardiograms were monitored throughout exercise; blood pressure was measured using a standard cuff sphygmomanometer. Heart rate (HR) at rest and maximal exercise were obtained from the electrocardiogram. Heart rate recovery at one minute (HRR₁) was the difference between HR at peak exercise and one-minute post. All subjects performed an active cool-down, at the initial workload of the exercise protocol, for at least one minute following the cessation of exercise. Minute ventilation (VE), oxygen uptake (VO₂), and carbon dioxide output (VCO₂) and end-tidal carbon dioxide production (P_ETCO₂) were acquired breath-by-breath, and averaged over 10-second intervals. Resting P_ETCO₂ was expressed as a two-minute averaged value in mmHg. Peak VO₂ and peak respiratory exchange ratio (RER) were expressed as the highest 10-second averaged samples obtained during the exercise test. VE and VCO₂ values, acquired from the initiation of exercise to peak, were input into spreadsheet software (Microsoft Excel, Microsoft Corp., Bellevue, WA) to calculate the VE/VCO₂ slope via least squares linear regression (y = mx + b, m=slope).^{14, 15} The oxygen uptake efficiency slope (OUES) was also determined via least squares linear regression (VO₂ = a log₁₀VE + b)¹⁶ by spreadsheet software (Microsoft Excel, Seattle, Washington) using all of the exercise data. Percent-predicted peak VO₂ was calculated according to normative values proposed by Wasserman and Hansen et al. (one of six equations according to sex and bodyweight) ^{4, 5}, Jones et al. (equation)³, the Cooper Clinic (below low fitness threshold)^{8, 9}, a Veteran’s Administration male referral data set (four equations; A-D)¹⁷ and the St. James Take Heart Project for women (equation)⁷. The prognostic significance of percent-predicted VO₂ values derived from the two latter, sex-specific, prediction equations were assessed collectively (VA A-D/St. James).

Endpoints

Subjects were followed for major cardiac events (mortality, LVAD implantation, heart transplantation) via hospital and outpatient medical chart review for a maximum of two years at all centers. Any death with a cardiac-related discharge diagnosis was considered an event. Clinicians conducting the exercise test were not involved in decisions regarding cause of death or heart transplant/LVAD implantation.

Statistical Analysis

All continuous data are reported as mean ± standard deviation. Unpaired t-testing assessed differences in continuous baseline and CPX variables between those subjects suffering an event and those who were event free. Chi-square analysis assessed differences in the distributions of HF etiology, gender, race and pharmacologic management between these groups. One-way analysis of variance (ANOVA) compared differences in the percent-predicted peak VO₂ values according to the various equations in the overall group. A mixed-model two-way ANOVA assessed differences between percent-predicted peak VO₂ values (within subject factors) according to adverse event status (between subject factors). When a significant difference between percent-predicted peak VO₂ values was detected, post-hoc analysis was performed by multiple paired-testing with a Bonferroni correction [p<0.007 (0.05/7)]. Receiver operating characteristic (ROC) curves were constructed for the prognostic classification schemes of all percent-predicted peak VO₂ values. A z-test was used to assess for significance of differences amongst areas under the ROC curve for the prognostic models.¹⁸ Univariate Cox regression analysis assessed the prognostic value of percent-predicted peak VO₂ calculations for cardiac mortality alone and in subgroups according to sex, mode of exercise [treadmill vs. lower extremity (LE) ergometer], peak RER (< vs. ≥1.05) and age (< vs. ≥50 years old). Separate multivariate Cox regression analyses (forward stepwise method, entry and removal values 0.10 and 0.05, respectively) then assessed the combined prognostic value of the VE/VCO₂ slope and each expression of peak VO₂ (actual and percent-predicted values). Multivariate Cox regression analysis was also used to assess the prognostic value of the VE/VCO₂ slope, peak VO₂ expressions and an expanded list of baseline variables including age, HF etiology, left ventricular ejection fraction, and NYHA class. For the peak VO₂ expression demonstrating the highest prognostic value in the expanded multivariate Cox regression, Kaplan-Meier analysis was performed including all variables retained in that particular regression. The log-rank test determined statistical significance of the Kaplan-Meier analysis. With the exception of post-hoc testing for the mixed-model, two-way ANOVA, all statistical tests with a p-value <0.05 were considered significant.

Results

There were 179 major cardiac events (117 deaths, 44 heart transplantations and 18 left ventricular assist device implantations) during the two-year tracking period (annual event rate: 10%). Baseline characteristics for the overall group as well as subgroups according to adverse event status are listed in Table 1. The percentages of subjects with an ischemic HF etiology and prescribed a diuretic were significantly higher in subjects suffering a major cardiac event. In addition, mean NYHA class was significantly higher while left ventricular ejection fraction was significantly lower in the major cardiac event subgroup.

Table 1.

Baselineand Pharmacotherapy Characteristics

	All Patients (n=1,165)	Event Free (n=986)	Major Cardiac Event (n=179)
Baseline Characteristics
Age, years	57.0 ±13.8	57.0 ±13.7	57.0 ±14.1
Sex, M/F%	70/30	69/31	75/25
Race, Caucasian/African- American/Other%	76/23/1	76/23/1	78/21/1
Etiology, Isch./Non-Isch.	43/57	42/58	50/50*
NYHA Class	2.5 ±0.61	2.4 ±0.60	2.9 ±0.55**
Resting Heart Rate, beats/min	75.1 ±13.4	74.8 ±13.2	77.0 ±14.4
LVEF, %	31.4 ±14.0	32.7 ±14.1	24.4 ±11.6**
Therapy Distribution, %
ACE Inhibitor	72	72	73
Diuretic	75	73	87**
Beta-Blocker	67	67	65

^*p<0.05

^**p<0.01

CPX results for the overall group as well as subgroups according to adverse event status are listed in Table 2. In the overall group, all percent-predicted peak VO₂ values were significantly different, with the exception of values derived from the VA-B/St. James and Cooper clinic equations. Aside from mode of exercise characteristics and peak RER, all variables of interest were significantly different according to adverse event status. Peak VO₂, maximal heart rate, and all percent-predicted peak VO₂ calculations were significantly lower while the VE/VCO₂ slope was significantly higher in subjects suffering a major cardiac event. Moreover, all percent-predicted peak VO₂ calculations were significantly different between those with and without a major event. Resting P_ETCO₂, the OUES and HRR₁ data was available in 737 (major events: 120), 452 (major events: 62) and 612 (major events: 82) subjects, respectively. For these subgroups, resting P_ETCO₂ (34.3 ±4.5 vs. 32.5 ±4.5 mmHg), the OUES (1.8 ±0.9 vs. 1.3 ±0.6) and HRR₁ (19.2 ±11.8 vs. 12.0 ±9.6 beats per minute) were all significantly higher (p<0.001) in subjects who did not suffer a major cardiac event.

Table 2.

Cardiopulmonary Exercise Test Data and Percent-Predicted Peak VO₂ Calculations^£

	All Patients (n=1,165)	Event Free (n=986)	Major Cardiac Event (n=179)
Mode of Exercise, Treadmill/Ergometer %	57/43	57/43	55/45
Peak VO2 , ml• kg-1•min-1	15.3 ±5.6	15.9 ±5.6	12.0 ±4.0*
VE/VCO2 slope	35.9 ±9.5	34.7 ±8.6	42.8 ±11.6*
Peak RER	1.09 ±0.14	1.09 ±0.14	1.09 ±0.14
Maximal Heart Rate, beats/min	125.9 ±22.8	128.0 ±22.1	114.3 ±22.7*
Wasserman/Hansen, %	59.4 ±22.5	62.1 ±22.2	45.0 ±18.3*
Jones, %	65.6 ±30.8	68.5 ±31.2	49.1 ±22.0*
VA-A/St. James, %	57.2 ±20.4	59.5 ±20.2	44.8 ±16.7*
VA-B/St. James, %	55.2 ±19.9	57.4 ±19.6	43.4 ±16.8*
VA-C/St. James, %	62.4 ±25.9	64.8 ±26.0	49.5 ±21.2*
VA-D/St. James, %	51.5 ±18.9	53.5 ±18.7	40.4 ±16.5*
Cooper Clinic, %	55.4 ±18.9	57.6 ±18.6	43.6 ±15.7*

^*No-event vs. adverse event subgroups, p<0.01

^£With the exception of VA-B/St. James vs. Cooper Clinic (p=0.35), all other percent predicted values significantly different from one-another for the overall group (p<0.001).

Receiver operating characteristic curve analysis results for the different peak VO₂ prediction equations is listed in Table 3. All prognostic classification schemes were statistically significant. With the exception of the VA-D/St. James equation, all optimal threshold values were within five percentage points. The area under the ROC curve was greatest for the Wasserman/Hansen equation, although statistical significance in ROC areas was only reached in comparison between the VA-C and VA-D/St. James equations. The Jones, Cooper Clinic, VAA/St. James and VA-B/St. James equations also demonstrated a significantly greater area under the ROC curve compared to VA-C/St. James equation. All other area under the ROC curve comparisons did not reach statistical significance.

Table 3.

Receiver Operating Characteristic Curve Analysis

	ROC curve area (95% CI)	Optimal Threshold	Sensitivity/Specificity	Hazard Ratio (95% CI)β	p-value
Wasserman/Hansen	0.74 (0.71-0.78)*	<47%	74/59	4.0 (3.0-5.4)	<0.001
Jones	0.73 (0.69-0.77)£	<51%	73/60	4.0 (2.9-5.3)	<0.001
VA-A/St. James	0.72 (0.69-0.76)£	<47%	70/60	3.6 (2.6-4.8)	<0.001
VA-B/St. James	0.72 (0.68-0.76)Ω	<46%	70/61	3.5 (2.6-4.7)	<0.001
VA-C/St. James	0.69 (0.64-0.73)	<50%	67/62	3.3 (2.5-4.4)	<0.001
VA-D/St. James	0.71 (0.67-0.75)	<43%	68/62	3.4 (2.5-4.6)	<0.001
Cooper Clinic	0.73 (0.70-0.78)£	<47%	70/62	3.5 (2.6-4.8)	<0.001

^*Significantly greater than VA-C/St. James (p<0.01) and VA-D/St. James (p<0.05)

^£Significantly greater than VA-C/St. James (p<0.01)

^ΩSignificantly greater than VA-C/St. James (p<0.05)

^βHazard ratios generated from univariate Cox regression using optimal threshold value

Table 4 lists the prognostic value of percent-predicted VO₂ values according to cardiac mortality as the only endpoint and sex, mode of exercise, exercise effort and age-based subgroups. The ability of all percent-predicted VO₂ values to predict major adverse events remained statistically significant when cardiac mortality was considered the only endpoint and in all subgroup analyses.

Table 4.

Prognostic Value of Percent-Predicted Oxygen Consumption According to Relevant Subgroups^*

	Cardiac Mortality Alone	Sex		Mode of Exercise		Subject Effort		Age
	All subjects (n=1165; events=117)	Male (n=813; events=134)	Female (n=352; events=45)	Treadmill (n=665; events=98)	LE Ergometer (n=419; events=81)	Peak RER ≤1.05 (n=486; events=76)	Peak RER >1.05 (n=576; events=103)	<50 Years Old (n=314; events=42)	≥50 Years Old (n=851; events=137)
Wasserman /Hansen	33.8, p<0.001	84.7, p<0.001	11.5, p=0.001	75.9, p<0.001	23.7, p<0.001	31.3, p<0.001	66.3, p<0.001	23.5, p<0.001	78.5, p<0.001
Jones	28.3, p<0.001	53.4, p<0.001	8.6, p=0.003	42.2, p<0.001	21.8, p<0.001	15.7, p<0.001	48.9, p<0.001	7.9, p=0.005	67.8, p<0.001
VA-A/St. James	30.3, p<0.001	88.8, p<0.001		76.8, p<0.001	14.7, p<0.001	23.8, p<0.001	65.2, p<0.001	20.1, p<0.001	71.8, p<0.001
VA-B/St. James	27.4, p<0.001	82.8, p<0.001	7.6, p=0.006£	76.8, p<0.001	13.5, p<0.001	22.8, p<0.001	64.0, p<0.001	21.2, p<0.001	71.4, p<0.001
VA-C/St. James	16.6, p<0.001	59.8, p<0.001		60.6, p<0.001	5.8, p=0.016	16.7, p<0.001	43.6, p<0.001	21.6, p<0.001	51.1, p<0.001
VA-D/St. James	25.1, p<0.001	76.4, p<0.001		74.5, p<0.001	13.7, p<0.001	21.2, p<0.001	62.9, p<0.001	21.1, p<0.001	71.2, p<0.001
Cooper Clinics	31.6, p<0.001	82.1, p<0.001	8.1, p=0.004	73.7, p<0.001	19.2, p<0.001	23.8, p<0.001	69.1, p<0.001	18.7, p<0.001	76.1, p<0.001

^*Univariate Cox regression analysis results reported as chi-square and respective p-value

^£Same St. James Female equation used for VA-A — VA-D

Multivariate Cox regression analysis including the VE/VCO₂ slope and each expression of aerobic capacity is listed in Table 5. The VE/VCO₂ slope was the superior predictor of major cardiac events in each analysis while measured peak VO₂ and each percent-predicted peak VO₂ calculation added predictive value and were thus retained in the regression. The residual chi-square value was greatest for the percent-predicted peak VO₂ value derived from the Wasserman/Hansen equation.

Table 5.

Multivariate Cox Regression Analyses for the Combination of Ventilatory Efficiency and Different Aerobic Capacity Expressions^Ω

Variable	Chi-square	p-value
VE/VCO2 slope	150.7	<0.001

Variable	Residual Chi-square	p-value
Peak VO2	32.9* (Univariate chi square: 75.1, p<0.001)	<0.001
Wasserman/Hansen£	52.3* (Univariate chi square: 96.9, p<0.001)	<0.001
Jones	42.9* (Univariate chi square: 62.8, p<0.001)	<0.001
VA-A/St. James	43.0* (Univariate chi square: 86.9, p<0.001)	<0.001
VA-B/St. James	43.0* (Univariate chi square: 84.3, p<0.001)	<0.001
VA-C/St. James	31.9* (Univariate chi square: 58.9, p<0.001)	<0.001
VA-D/St. James	42.8* (Univariate chi square: 81.3, p<0.001)	<0.001
Cooper Clinic	42.9* (Univariate chi square: 90.4, p<0.001)	<0.001

^*Retained in multivariate regression

^ΩMultivariate regression run total of eight separate times: VE/VCO₂ slope plus each aerobic capacity expression

^£Highest residual chi-square

Two expanded multivariate Cox regression analyses that included the VE/VCO₂ slope, either measured peak VO₂ or percent-predicted peak VO₂ according the Wasserman/Hansen equation and key baseline characteristics are presented in Table 6. The VE/VCO₂ slope was again the superior prognostic marker in both assessments. Measured peak VO₂ and percent-predicted peak VO₂ according the Wasserman/Hansen equation were retained in the separate regression analyses in addition to NYHA class and left ventricular ejection fraction. Based on the residual chi-square values, the Wasserman/Hansen equation provided superior predictive value compared to measured peak VO₂ and other consistent demographic variables.

Table 6.

Multivariate Cox Regression Analysis Using Actual Peak Oxygen Consumption Or Percent-Predicted Peak Oxygen Consumption Using the Wasserman/Hansen Equation

Actual Peak Oxygen Consumption
Variable	Chi-square	p-value
VE/VCO2 slope	150.7	<0.001

Variable	Residual Chi-square	p-value
NYHA Class*	28.5	<0.001
Left Ventricular Ejection Fraction*	26.7	<0.001
Peak VO2*	8.1	0.005
HF Etiology	1.7	0.20
Age	0.44	0.51

Percent-Predicted Peak Oxygen Consumption
Variable	Chi-square	p-value
VE/VCO2 slope	150.7	<0.001

Variable	Residual Chi-square	p-value
Wasserman/Hansen Equation*	28.1	<0.001
Left Ventricular Ejection Fraction*	14.8	<0.001
NYHA Class*	13.0	<0.001
HF Etiology	2.4	0.12
Age	0.81	0.37

^*Retained in multivariate regression

The Jones (Residual chi-square: 10.8, p=0.001), VA-A/St. James (Residual chi-square: 8.0, p=0.005), VA-B/St. James (Residual chi-square: 7.9, p=0.005), VA-C/St. James (Residual chi-square: 5.2, p=0.02), VA-D/St. James (Residual chi-square: 8.0, p=0.005) and Cooper Clinic (Residual chi-square: 7.9, p=0.005) equations were all retained in the same expanded multivariate Cox regression depicted in Online Supplemental data. The residual chi-square values were comparable to that found with measured peak VO₂ and below that provided by the Wasserman/Hansen equation. Moreover, as found with the analysis including measured peak VO₂, the residual chi-square values for left ventricular ejection fraction and NYHA class were higher (Residual chi-square: 12.2, p<0.001) in each of these latter scenarios.

Kaplan-Meier analysis curves are illustrated in Figure 1. Dichotomous thresholds of </≥36.0, </≥47%, ≤/>25% and I/II vs. III/IV were set for the VE/VCO₂ slope¹⁹, percent-predicted peak VO₂ according to the Wasserman/Hansen equation (determined by ROC curve analysis in the present investigation), left ventricular ejection fraction¹³ and the NYHA class, respectively. Using these thresholds, there was a significant difference in adverse event rates between subgroups according to the number of abnormal characteristics.

Figure 1.

Kaplan-Meier analysis for combined VE/VCO₂ slope, Percent-Predicted Peak VO₂, Left Ventricular Ejection Fraction and NYHA class threxsholds

Discussion

A number of equations designed to estimate the percentage of normal aerobic capacity achieved during exercise testing are presently available for clinical application. The present study demonstrates that several well-known methods for the determination of percent-predicted aerobic capacity: 1) provide values that are for the most part significantly different from one-another, limiting the portability of a given equation to different populations of patients with HF; 2) are all prognostically significant in a large HF cohort from a univariate perspective; 3) are all prognostically significant when only cardiac mortality was considered as an end-point; 4) all provide prognostic value in subgroups according to sex, mode of exercise, exercise effort and age; and 5) all provide additional prognostic value in a multivariate model including other clinically established exercise and resting variables.

In the prognostic comparison between equations, it appears the Wasserman/Hansen calculations provided better resolution, although differences in areas under the ROC curve were not significantly different compared to most other equations assessed. The potential value of the Wasserman/Hansen equations is more so apparent in the multivariate Cox regression analysis that included the VE/VCO₂ slope and key resting variables (Tables 5 and 6). In these analyses, percent-predicted values derived from the Wasserman/Hansen approach possessed the highest univariate chi-square value (compared to measured peak VO₂ and other percent-predicted calculations), was retained in the multivariate regression and outperformed both left ventricular ejection fraction and NYHA class. While measured peak VO₂ and all other percent-predicted calculations were also retained in their respective multivariate Cox regression analyses, their added prognostic value was not as powerful compared to that derived from inclusion of the Wasserman/Hansen equations. The findings of the present study are consistent with the previous investigation by Stelken et al.¹² in that percent-predicted peak VO₂ according to the Wasserman/Hansen equations prognostically outperformed an abbreviated list of other equations and measured peak VO₂ in 181 patients with HF. Similalry, Osada et al.²⁰ found percent-predicted peak VO₂, again derived from the Wasserman/Hansen equations, prognostically outperformed measured peak VO₂ in 500 patients with HF. Aaronson et al.¹¹, however, found that the Wasserman/Hansen percent-predicted VO₂ calculations and measured peak VO₂ performed similarly in a HF cohort comprised of 272 patients. An advantage of the present investigation compared to these previous studies is the larger sample size, possibly lending more credence to our findings and supporting the previous investigations by Stelken¹² and Osada et al²⁰.

Previous investigations^{12, 20} have solely used a percent-predicted peak VO₂ threshold of 50% in their dichotomous prognostic assessments. Determination of this threshold appears to have been more arbitrary than the optimal sensitivity/specificity determination via ROC curve analysis. Despite differences in determination of the dichotomous threshold employed in the current and previous investigations, the cut-point was for the most part similar. This concordance of research indicates that patients with a percent-predicted VO₂ value below approximately 50% have poorer outcomes compared to those surpassing this threshold, although the optimal cut-point may slightly deviate from 50% for a given predicted peak VO₂ equation.

There is considerable variation in how presently available prediction equations are defined. While all equations considered age and sex, the Wasserman/Hansen male/female equations have taken the greatest number of additional factors into consideration, including body weight (underweight/normal weight/overweight), mode of exercise (treadmill/lower extremity ergometer) and sedentary lifestyle. Moreover, the Wasserman/Hansen, Jones, VA-A and VA-B equations, used peak VO₂ determined by ventilatory expired gas analysis to develop their normative values. The VA-C, VA-D, St. James and Cooper Clinic equations all estimated aerobic capacity from treadmill speed/grade or exercise time. Given the differences in equation development, the improved prognostic utility of the Wasserman/Hansen equations may be the result of their ability to account for more explanatory variables and provide a truer depiction of aerobic capacity for an apparently healthy, but sedentary individual. It should be noted, however, that despite the potential limitations of the other prediction equations assessed in the present investigation, they all provided significant prognostic value both independently and in combination with clinically important variables.

It could be argued that the additional steps required for the Wasserman/Hansen equations are not worth the relatively modest increment gained in prognostic information. However, these equations are easily incorporated into the software packages that operate CPX systems. The percent-predicted peak VO₂ value according to the Wasserman/Hansen equation can therefore be automatically generated by manually inputting age, sex, height, weight and mode of exercise, a procedure which is already commonplace in preparation for a CPX. The ease by which all percent-predicted peak VO₂ calculations are derived by presently available software packages eliminates the need for consideration of a given equations complexity.

There is a wealth of research supporting the prognostic strength of the VE/VCO₂ slope in HF, which commonly outperforms measured peak VO₂.^{19, 21-23} While the prognostic value of the VE/VCO₂ slope is gaining clinical recognition, peak VO₂ continues to be the sole or primary CPX variable considered in the prognostic assessment of patients with HF undergoing this procedure. The results of the present study indicate measured or percent-predicted peak VO₂ should continue to be considered as a secondary CPX variable, complimenting the insight gained from the VE/VCO₂ slope. Irrespective of which variable provides the highest level of prognostic information, it appears clear from the recent literature that the ability to predict adverse events improves with multivariate modeling that includes CPX variables as well as key resting measures such as NYHA class and left ventricular ejection fraction.²²

Subjects included in the present investigation were referred for CPX at their respective institution, creating the potential for selection bias. Caution must therefore be taken in extrapolating our findings to the HF population as a whole or to CPX laboratories assessing HF patients with differing characteristics, such as, for example, younger individuals with a congenital heart defect being considered for transplantation. Moreover, while the overall number of subjects in the present investigation exceeded 1,000, the majority were male. While all percent-predicted VO₂ equations were prognostic in the female subgroup, particular caution should be taken in extrapolating our findings to the female population with HF. Given the heterogeneous nature of this disease process, future investigations should be performed in other HF cohorts to determine if the prognostic value of percent-predicted peak VO₂ is universally applicable to this chronic disease population. While peak VO₂ and the VE/VCO₂ slope are well-established, other variables, such as resting P_ETCO₂²⁴, the OUES²⁵ and HRR²⁶ have demonstrated prognostic value. While these variables were unfortunately not available for the entire cohort in the present investigation, subgroup analysis revealed all three demonstrated significantly better characteristics in subjects who did not suffer a major cardiac event. In addition, peak VO₂ adjusted for body fat has demonstrated prognostic value in patients with HF.^{27, 28} Analysis of this expression of aerobic capacity could not be performed in the present study as body fat assessment was not performed in any of the subjects. Future research should be directed toward a more comprehensive multivariate survival analysis to determine all clinically relevant CPX variables as well as optimal expression.

In conclusion, variables obtained from CPX provide important prognostic insight in patients with HF. The findings of the present investigation further confirm the prognostic superiority of ventilatory efficiency (VE/VCO₂ slope) and suggest equations used to determine percent-predicted peak VO₂ provide similar and in some instances better predictive information compared to the measured value obtained from CPX. Although many laboratories conducting CPX in patients with HF report percent-predicted peak VO₂ values in their written report, they do not commonly consider its prognostic significance. Our results suggest that: 1) peak VO₂ should be expressed as a percentage of the predicted normal value and this should be a routine part of the summary report; and 2) the Wasserman/Hansen equation is superior to other equations in terms of prognostic power in patients with HF.