The Partial Pressure of Resting End-Tidal Carbon Dioxide Predicts Major Cardiac Events In Patients with Systolic Heart Failure

Abstract

Background

The resting partial pressure of end tidal carbon dioxide (P_ETCO₂) has been shown to reflect cardiac performance in acute care settings in patients with heart failure (HF). The purpose of the present study was to compare the prognostic ability of the partial pressure of P_ETCO₂ at rest to other commonly collected resting variables in patients with systolic HF.

Methods

A total of 353 patients (mean age: 58.6 ±13.7, 72% male) with systolic HF were included in this study. All patients underwent cardiopulmonary exercise testing where NYHA class, resting P_ETCO₂, peak oxygen consumption (VO₂) and the minute ventilation/carbon dioxide production (VE/VCO₂) slope were determined. Subjects were then followed for major cardiac events (mortality, LVAD implantation, urgent heart transplantation).

Results

There were one hundred and four major cardiac events during the 23.6 ± 17.0 month tracking period. Multivariate Cox regression analysis revealed NYHA class (Chi-square: 28.7, p<0.001), LVEF (Residual chi-square: 21.7, p<0.001) and resting P_ETCO₂ (Residual chi-square: 14.1, p<0.001) were all prognostically significant and retained in the regression. In a separate Cox regression analysis, LVEF (residual chi-square: 8.8, p=0.003), NYHA class (residual chi-square: 7.7, p=0.005) and resting P_ETCO₂ (residual chi-square: 5.7, p=0.02) added prognostic value to the VE/VCO₂ slope (Chi-square: 26.0, p<0.001).

Conclusion

Resting P_ETCO₂ can be non-invasively collected from subjects in a short period of time, at a low cost and with no risk or discomfort to the patient. Given the prognostic value demonstrated in the present study, the clinical assessment of resting P_ETCO₂ in the HF population may be warranted.

Introduction

Ventilatory expired gas analysis during aerobic exercise testing clearly provides prognostic value in the heart failure (HF) population.(1) The predictive value of ventilatory efficiency and aerobic capacity is a function of their relationships with other measures of pathophysiology and disease severity in this population. While a great deal of attention has been devoted to measures obtained during exercise, particularly peak oxygen consumption (VO₂) and the minute ventilation/carbon dioxide production (VE/VCO₂) slope, little focus has been directed towards the clinical value of resting ventilatory expired gas analysis. The partial pressure of resting end-tidal carbon dioxide (P_ETCO₂) is one such variable that may provide important clinical information, particularly given its previously demonstrated relationship to cardiac function in other patient populations.(2–10) Furthermore, Matsumoto et al.(11) found significantly lower resting P_ETCO₂ values in HF patients with a higher NYHA class, indicating a link between this expired gas variable and an established clinical measure of disease severity.

To our knowledge, we are the only research group to previously report on the reliability(12) and prognostic significance of resting P_ETCO₂ in patients with HF.(13) In the prior analysis, we found resting P_ETCO₂ was a significant univariate predictor of adverse events in a small HF cohort (n=121), with the majority of events being cardiac-related hospitalization, over only a one-year tracking period. Furthermore, resting P_ETCO₂ added prognostic value to other resting and exercise variables in a multivariate analysis. While these previous results are compelling, more work in this area is required to corroborate and expand upon our initial findings. The purpose of the present investigation is to therefore examine the ability of resting P_ETCO₂ to predict cardiac mortality, left ventricular assist device implantation (LVAD) and urgent heart transplantation, all hard endpoints, in a large HF cohort over an extended tracking period.

Methods

This study is a multi-center analysis including HF patients from the exercise testing laboratories at San Paolo Hospital, Milan, Italy Wake Forrest University Baptist Medical Center, Winston-Salem, North Carolina and Virginia Commonwealth University, Richmond, Virginia, USA. A total of 353 patients with chronic HF, secondary to systolic dysfunction, were included. Inclusion criteria consisted of a diagnosis of HF(14) and evidence of left ventricular systolic dysfunction by two-dimensional echocardiography [left ventricular ejection fraction (LVEF) ≤40%] obtained within one month of data collection. Furthermore, none of the subjects included in this analysis suffered from arterial hypoxemia or had a previous diagnosis of pulmonary disease or pulmonary embolism as per medical chart review. Subjects from Virginia Commonwealth University assessed in the original resting P_ETCO₂ investigation by our group, described in the introduction(13), were also included in this analysis. All subjects completed a written informed consent and institutional review board approval was obtained at each institution.

Ventilatory expired gas analysis was performed using a metabolic cart at all three centers (Medgraphics CPX-D, Minneapolis, MN or Sensormedics Vmax29, Yorba Linda, CA). Before each test, the equipment was calibrated in standard fashion using reference gases and a three-liter syringe. Prior to symptom-limited exercise testing, ventilatory expired gas data was collected at rest in the seated position for two minutes. The subject was allowed to adjust to the mouthpiece for a minimum of 30 seconds before resting data collection was initiated. The partial pressure of end-tidal carbon dioxide (in mmHg) and the minute ventilation (VE)/carbon dioxide production (VCO₂) ratio at rest were both expressed as two-minute averaged values. Resting respiratory exchange ratio (RER) was monitored to ensure subjects were not hyperventilating at rest and data was only collected when resting RER was below 1.0 for the entire two-minutes (average = 0.87). All subjects then underwent a symptom-limited exercise test using a conservative ramping protocol. Peak oxygen consumption (VO₂), the minute ventilation (VE)/carbon dioxide production (VCO₂) slope, P_ETCO₂ at ventilatory threshold (VT)(15), the change in P_ETCO₂ from rest to VT (P_ETCO_2Delta) and P_ETCO₂ at peak exercise were also determined for this analysis. Peak VO₂ P_ETCO₂ at VT, P_ETCO₂ at peak exercise and P_ETCO_2Delta were all expressed as 10-second averaged sample obtained during the exercise test. Percent-predicted peak VO₂ was calculated using the equations derived by Wasserman et al.(16) 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).

Subjects were followed for major cardiac events (mortality, LVAD implantation, urgent heart transplantation) via hospital and outpatient medical chart review for a four year period. Subjects were followed by the HF programs at their respective institution providing a high likelihood that all events were captured. 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 urgent heart transplant/LVAD implantation.

All continuous data are reported as mean values ± standard deviation (SD). Pearson Product moment correlation assessed the relationship between resting P_ETCO₂ and other continuous variables included in this analysis (age, LVEF, resting VE/VCO₂, peak VO₂, P_ETCO₂ at ventilatory threshold, P_ETCO_2Delta, P_ETCO₂ at peak exercise and the VE/VCO₂ slope). A correlation analysis between the VE/VCO₂ slope and both resting VE/VCO₂ and P_ETCO₂ measurements during exercise was also performed. One-way analysis of variance assessed differences in resting P_ETCO₂ amongst NYHA classes I, II and III/IV (III/IV combined secondary to the low number of subjects in the latter NYHA class). Tukey’s honestly significant difference was used to determine differences in subgroups when the one-way ANOVA p-value was <0.05. Univariate and multivariate (forward stepwise method, entry and removal values 0.05 and 0.10, respectively) Cox regression analysis assessed the prognostic characteristics of LVEF, resting P_ETCO₂, NYHA class, age and HF etiology. The combined prognostic value of LVEF, resting P_ETCO₂ and exercise variables (peak VO₂, P_ETCO_2Delta and the VE/VCO₂ slope) was also assessed. Measures of P_ETCO₂ at ventilatory threshold and peak exercise were not included in the survival analysis secondary to collinearity with resting P_ETCO₂ (see correlation analysis in the results). Kaplan-Meier analysis was used to assess survival characteristics of dichotomous expressions of resting variables retained in the multivariate regression analyses both separately and in combination. Kaplan-Meier analysis was also used to assess the combined prognostic value of resting and cardiopulmonary exercise test variables retained in a multivariate regression analysis. The log-rank test determined statistical significance of the Kaplan-Meier analysis. Statistical differences with a p-value <0.05 were considered significant.

Results

Subject characteristics are listed in Table 1. The majority of this cohort was male, NYHA class II or III and prescribed both an ACE inhibitor and diuretic. The mechanism of HF was balanced between ischemic and non-ischemic etiology. Mean resting P_ETCO₂ and peak VO₂ were lower and the VE/VCO₂ slope was elevated compared to expected responses in apparently healthy individuals. The correlation between resting P_ETCO₂ and LVEF (r = 0.17, p=0.001), resting VE/VCO₂ (r = −0.43, p<0.001), P_ETCO₂ at VT (r = 0.84, p<0.001), P_ETCO₂ at peak exercise (r = 0.78, p<0.001), peak VO₂ (r = 0.21, p<0.001) and the VE/VCO₂ slope (r = −0.51, p=0.001) were statistically significant while the relationship with age (r = −0.07, p=0.19) and P_ETCO_2Delta (r = −0.03, p=0.60) were not. Moreover, the correlation between the VE/VCO₂ slope and resting VE/VCO₂ (r = 0.59, p<0.001), P_ETCO₂ at VT (r = −0.60, p<0.001), P_ETCO₂ at peak exercise (r = 0.75, p<0.001) and P_ETCO_2Delta (r = −0.31, p<0.001) were all statistically significant. Mean resting P_ETCO₂ values for subjects in NYHA class I, II and III/IV were 35.6 ± 3.6, 34.6 ± 3.6 and 33.7 ± 4.3 mmHg, respectively. Post-hoc analysis revealed that while the difference in resting P_ETCO₂ values between NYHA class II and III/IV approached a p-value <0.05 (p=0.09), only differences between NYHA class I and III/IV reached statistical significance (p=0.01).

Table 1.

Subject Characteristics

Variable	Value
Age (mean in years)	58.6 ±13.7
Sex (percent male/female)	72.0/28.0
NYHA class (percent I/II/III/IV)	13.3/42.2/42.2/2.3
HF etiology (percent ischemic/non-ischemic)	50.1/49.4
LVEF (mean in percent)	28.0 ±8.5
Resting PETCO2 (mean in mmHg)*	34.3 ±4.0
Resting VE/VCO2	43.2 ±11.3
Beta-blocker (percent yes)	58.4
ACE inhibitor (percent yes)	81.3
Diuretic (percent yes)	87.4
Peak VO2 (mean in mlO2•kg−1•min−1)	14.5 ±5.6
Peak VO2 Percent Predicted (mean %)	56.3 ±19.1
PETCO2 at ventilatory threshold (mean in mmHg)	37.0 ±4.7
PETCO2 at peak exercise(mean in mmHg)	33.5 ±5.6
PETCO2Delta (mean in mmHg)	2.7 ±2.6
VE/VCO2 slope (mean)£	36.4 ±8.7

^*Normal Resting P_ETCO₂: 36–42 mmHg

^£Normal VE/VCO₂ slope: <30

There were 104 major cardiac events (92 deaths, 3 LVAD implantations and 9 urgent heart transplantations) during the mean 23.6 ± 17.0 month tracking period. The annual event rate was 13.3%. Cox regression results for resting variables are listed in Table 2. NYHA class, LVEF and resting P_ETCO₂ were all prognostically significant univariate markers and retained in the multivariate regression. Neither age nor HF etiology were significant predictors of risk. NYHA, LVEF and resting P_ETCO₂ were again retained in the multivariate Cox regression when cardiac-related mortality was the only event considered.

Table 2.

Cox Multiple Regression Analysis for Resting Variables

Variable	Chi-square	p-value
NYHA Class	28.7	<0.001
Variable	Residual Chi-square	p-value
LVEF	21.7 (Univariate chi square: 27.3, p<0.001)	<0.001
Resting PETCO2	14.1 (Univariate chi square: 27.2, p<0.001)	<0.001
Age	0.51 (Univariate chi square: 0.93, p=0.34)	0.47
HF etiology	0.07 (Univariate chi square: 0.26, p=0.68)	0.80

For Kaplan-Meier analysis, dichotomous thresholds of ≤ />25%, ≤ />33 mmHg and I/II vs. III/IV were set for LVEF(14), resting P_ETCO₂(13) and NYHA class, respectively. Hazard ratios for the NYHA class, LVEF and resting P_ETCO₂ were 2.20 (95% CI: 1.49–3.26, p<0.001), 2.26 (95% CI: 1.54–3.33, p<0.001) and 2.17 (95% CI: 1.48–3.19, p<0.001), respectively. The independent prognostic characteristics of these dichotomous thresholds are illustrated in Figure 1. Using these thresholds, there was a significant difference in event rates between groups. Figure 2 illustrates event-free survival characteristics when these thresholds for resting P_ETCO₂, LVEF and NYHA class were considered collectively. This latter approach improved the accuracy of identifying low (no abnormal values) and high-risk (3 abnormal values) groups. Furthermore, use of all three resting variables allowed for the identification of subjects at intermediate risk (1–2 negative values).

Figure 1.

Individual Kaplan-Meier analyses for LVEF, NYHA class and resting P_ETCO₂ thresholds

Resting PETCO2: ≤/>33 mmHg
Group	Characteristics	Subjects meeting criteria	Major Cardiac Events	Percent Event Free
A	>33 mmHg	237	54	77.2%
B	≤ 33 mmHg	116	50	56.9%
Log rank: 16.3, p<0.001
NYHA class: I/II vs. III/IV
Group	Characteristics	Subjects meeting criteria	Major Cardiac Events	Percent Event Free
A	I/II	196	41	79.1%
B	III/IV	157	63	59.9%
Log rank: 16.3, p<0.001
LVEF: ≤/>25%
Group	Characteristics	Subjects meeting criteria	Major Cardiac Events	Percent Event Free
A	>25%	216	48	77.8%
B	≤25%	137	56	59.1%
Log rank: 18.0, p<0.001

Figure 2.

Kaplan-Meier analysis for combined LVEF, NYHA class and resting P_ETCO₂ thresholds

Group	Characteristics*	Subjects meeting criteria	Major Cardiac Events	Percent Event Free
A	No negative values	94	14	85.4%
B	1–2 Negative Values	221	71	67.9%
C	3 Negative Values	36	19	47.2%

Log rank: 29.8, p<0.001

^*

Negative Values:

• Resting P_ETCO₂: ≤33 mmHgM
• NYHA Class: III/IV
• LVEF:≤25%

In a separate Cox regression analysis including exercise test variables, the VE/VCO₂ slope (Threshold </≥36.0; chi-square: 26.0, p<0.001), LVEF (Threshold ≤ />25%; chi-square: 18.0, p<0.001), resting P_ETCO₂ (Threshold ≤ />33 mmHg; chi-square: 16.3, p<0.001), NYHA class (Threshold I/II vs. III/IV; chi-square 16.3, p<0.001), peak VO₂ (Threshold </ ≥ 14 mlO₂•kg⁻¹•min¹ ; chi-square: 7.4, p=0.006) and P_ETCO_2Delta (Threshold </ ≥ 2 mmHg; chi-square: 3.9, p=0.047) were all significant univariate predictors of major cardiac events. Left ventricular ejection fraction (residual chi-square: 8.8, p=0.003), NYHA class (residual chi-square: 7.7, p=0.005), resting P_ETCO₂ (residual chi-square: 5.7, p=0.02), but not P_ETCO_2Delta (residual chi-square: 0.53, p=0.47) or peak VO₂ (residual chi-square: 0.02, p=0.90), added prognostic value to the VE/VCO₂ slope in a multivariate model. A second Kaplan-Meier analysis, illustrated in Figure 3, considered the VE/VCO₂ slope in addition to NYHA class, LVEF and resting P_ETCO₂. Assessment of this cardiopulmonary exercise test variable, in conjunction with resting measures, further improved risk stratification.

Figure 3.

Kaplan-Meier analysis for combined LVEF, NYHA class, resting P_ETCO₂ and VE/VCO₂ slope thresholds

Group	Characteristics*	Subjects meeting criteria	Major Cardiac Events	Percent Event Free
A	No negative resting values and positive VE/VCO2 slope	65	5	92.3%
B	1 Negative Resting Value or Negative VE/VCO2 slope	136	38	74.0%
C	2 Negative Resting Values and Positive VE/VCO2 slope or 1 Negative Resting Value and Negative VE/VCO2 slope	108	44	59.3%
D	3 Negative Resting Values and Negative VE/VCO2 slope	30	17	43.3%

Log rank: 40.5, p<0.001

^*

Negative Values:

• Resting P_ETCO₂: ≤33 mmHg
• NYHA Class: III/IV
• LVEF: ≤25%
• VE/VCO₂ Slope:≥36.0

Discussion

Our group previously demonstrated resting P_ETCO₂ was a significant predictor of cardiac events in 121 subjects with HF.(13) In this relatively small analysis, the majority of recorded events were hospitalization secondary to decompensated HF (41 of 50), a soft endpoint. While the initial findings were compelling, confirmation was required in a larger cohort with a hard composite endpoint. The results of the present study, using hard endpoints (death, urgent transplantation or LVAD implant), further affirm the prognostic value of resting P_ETCO₂ in combination with other baseline measures with well-established prognostic value [NYHA class(17–20), LVEF(20–22)]. Moreover, resting variables, including resting P_ETCO₂, were prognostically superior to peak VO₂ and added predictive value to the VE/VCO₂ slope. While peak VO₂ is still most frequently considered for prognostic purposes, evidence demonstrating other ventilatory expired gas measures, such as the VE/VCO₂ slope(23) and now resting P_ETCO₂, may provide superior predictive information. The measurement and clinical consideration of resting P_ETCO₂ may therefore be warranted during resting evaluations and in combination with cardiopulmonary exercise testing.

Carbon dioxide elimination, and therefore P_ETCO₂, decreases when blood flow to the lungs is reduced. Numerous investigations have demonstrated a significant relationship between resting P_ETCO₂ and cardiac output.(2–10) Shibutani et al.(10) found P_ETCO₂ was significantly correlated with changes in cardiac output during surgery (r² = 0.82, p<0.01) in patients undergoing abdominal aortic aneurysm repair. Wahba et al.(7) reported P_ETCO₂ was effectively able to detect changes in cardiac index (r = 0.75, p<0.001) in subjects undergoing open heart surgery. Lastly, in a group of 100 critically ill surgery patients, Domsky et al.(24) reported a 55% mortality rate in patients with a P_ETCO₂ less than or equal to 28 mmHg. This is compared to a 17% mortality rate for those patients with a P_ETCO₂ greater than 28 mmHg. While none of these investigations were performed in patients with HF, they do provide a basis for the notion that resting P_ETCO₂ provides some reflection of varying degrees of cardiac output/function and therefore possesses prognostic value in this patient population. The results of the present study and our previous investigation¹³ appear to support this hypothesis as it relates to the prognostic value of resting P_ETCO₂.

Mechanisms other than cardiac function for a diminished resting P_ETCO₂ in advanced HF are, however, highly plausible and must be considered. Previous investigations have demonstrated physiologic dead space ventilation continues to increase while the partial pressure of carbon dioxide in arterial blood continues to decline as HF severity progresses.(25;26) Both of these physiologic abnormalities, either independently or synergistically, can also lead to a reduction in resting P_ETCO₂. It is therefore likely that the mechanism for the reduction in resting P_ETCO₂ observed in some patients with HF is multi-factorial (i.e. reduced cardiac output, reduced partial pressure of carbon dioxide in arterial blood and increased physiologic dead space ventilation) with a continued decline in this resting expired gas variable reflecting higher levels of overall HF severity and worse prognosis. Future research is required to better elucidate the reason(s) why resting P_ETCO₂ is reduced in HF.

Previous investigations have reported that the VE/VCO₂ slope is significantly correlated with other measures reflecting disease severity in HF such as invasively assessed cardiac output and pulmonary hemodynamics(27), brain natriuretic peptide(28), heart rate variability(29) and alveolar-capillary membrane conductance(30). While we were able to demonstrate a significant correlation between resting P_ETCO₂ and the VE/VCO₂, the collection of other disease severity indicators was limited. It would appear that the relationship between LVEF and resting P_ETCO₂, while statistically significant, is rather weak. Likewise, the difference in resting P_ETCO₂ according to NYHA class was only significant between class I and III/IV. Future investigations should be directed toward examining the relationship between resting P_ETCO₂ and other objective markers of HF severity.

Resting P_ETCO₂ is quickly and easily collected at a low cost, making this measure a particularly attractive assessment option in clinical practice. In addition to its prognostic potential, other possible clinical applications for resting P_ETCO₂ include assessing the response to therapeutic interventions and gauging HF stability. Interventions, such as cardiac resynchronization therapy and aerobic exercise training, that result in an improvement in cardiac function(31;32) may also increase resting P_ETCO₂. In this instance, resting P_ETCO₂ may be valuable in distinguishing between responders and non-responders. Furthermore, resting P_ETCO₂ likely declines as a patient with HF worsens from a compensated to decompensated state. Serial assessment of resting P_ETCO₂ would allow for the rapid detection of declining HF stability, assisting in the formulation of a treatment plan. Future research should be directed toward determining the value of broader clinical applications for resting P_ETCO₂ in patients with HF.

The fact that we collected resting P_ETCO₂ data in a select HF cohort referred for cardiopulmonary exercise testing is a study limitation. The characteristics of subjects included in the present investigation most likely differ from the HF population seen in the out-patient setting. For example, while patients in NYHA class IV are seen in the clinic, few are referred for exercise testing. To support the broader application of resting P_ETCO₂, our findings must be confirmed in HF patients seen on an out-patient basis, but not necessarily selected to undergo exercise testing. In addition, other resting variables with prognostic value, such as b-type natriuretic peptide(33;34), should be included in the multivariate analyses of subsequent investigations. Lastly, application of our findings to patients with diastolic HF is unknown. Future research should also determine if the prognostic characteristics of resting P_ETCO₂ is similar in patients with diastolic HF.

In conclusion, a diagnosis of HF continues to carry a poor prognosis(35), making the identification of variables with predictive value an important area of study. Resting P_ETCO₂ appears to add prognostic value to variables that are well-established and commonly collected in clinical practice. The fact that resting P_ETCO₂ is easily, cheaply and non-invasively obtained portends high clinical promise for this measurement.