European Heart Journal (2005) 26, 472–480 doi:10.1093/eurheartj/ehi060

Clinical research

Exercise ventilation inefficiency and cardiovascular mortality in heart failure: the critical independent prognostic value of the arterial CO2 partial pressure Marco Guazzi1*, Giuseppe Reina2, Gabriele Tumminello1, and Maurizio D. Guazzi3 1

Cardiopulmonary Laboratory, Cardiology Division, University of Milano, San Paolo Hospital, Via A. di Rudinı`, 8, 20142, Milano, Italy 2 Institute of Statistics and Biometry, University of Milano, Italy 3 Institute of Cardiology, University of Milano, Italy Received 3 March 2004; revised 16 October 2004; accepted 28 October 2004; online publish-ahead-of-print 14 December 2004

See page 426 for the editorial comment on this article (doi:10.1093/eurheartj/ehi141)

KEYWORDS Heart failure; PaCO2; Prognosis; Ventilation

Aims In chronic heart failure (CHF) patients, the ventilation (VE ) needed to eliminate metabolically produced CO2 during exercise (i.e. the VE /VCO 2 slope) is a strong prognosticator. VE /VCO 2 slope determinants are the dead space–tidal volume (VD /VT ) ratio and the arterial CO2 partial pressure (PaCO 2). We aimed at defining the respective prognostic role of these two variables. Methods and results One hundred and twenty-eight stable CHF patients (average left ventricular ejection fraction 34 + 10%) underwent cardiopulmonary exercise testing and blood gas analysis. The prognostic relevance of the VE /VCO 2 slope, VD /VT, and PaCO 2 at peak exercise was evaluated by the Kaplan–Meier approach with log-rank testing and by multivariate Cox regression analysis. During a mean period of 31.3 + 20 months, 24 patients died from cardiac causes. In univariate analysis, predictors of death included the use of anti-aldosterone drugs, low peak VO 2, peak VE /VO 2, peak PaCO 2 and high VE /VCO 2 slope, and peak VD /VT. Multivariate analysis identified a low peak PaCO 2 (,35 mmHg) as the strongest independent prognostic indicator [hazard ratio 4.65, 95% confidence interval (CI) (1.695 2 12.751), P ¼ 0.003] that primarily accounts for the VE /VCO 2 slope prognostic power. Conclusion These findings imply that regulatory mechanisms involved in the tight control of ventilatory command and blood gas tension, rather than lung function abnormalities, play a critical pathophysiological role in the exercise ventilation inefficiency of CHF patients.

Introduction Identification of chronic heart failure (CHF) patients at high risk for early death remains a basic challenge. This challenge has two major complementary approaches:

* Corresponding author. Tel/fax: þ39 02 50323144. E-mail address: [email protected]

(i) the search for new and more sensitive prognostic indicators, and (ii) the refinement of already known prognostic markers through a more in-depth study and characterization of their determinants. In recent years, there has been growing and convincing evidence that inefficient ventilation during physical performance is a very sensitive indicator of poor outcome in CHF patients.1–7 Interestingly, a number of reports

& The European Society of Cardiology 2004. All rights reserved. For Permissions, please e-mail: [email protected]

Prognostic value of PaCO 2 in heart failure

suggest that the excessive amount of ventilation (VE ) needed to eliminate metabolically produced CO2 (i.e. the VE /VCO 2 slope) is a prognostic indicator even more powerful than oxygen consumption measured at peak exercise (peak VO 2).2,3,7 The VE /VCO 2 slope determinants are the physiological dead space–tidal volume (VD /VT ) ratio and the arterial CO2 partial pressure (PaCO 2). Studies investigating the pathophysiological basis for an increased VE /VCO 2 slope in CHF have reported conflicting results. Some investigators have stressed the importance of an increased VD /VT secondary to ventilation/perfusion mismatching, especially in patients with moderate to severe heart failure with a preserved or minimally impaired neural control of ventilation.8,9 Others have provided evidence that overactive chemoreflex and ergoreflex responses drive the ventilatory pattern during exercise.1,4 Information regarding PaCO 2 changes during exercise is limited, and despite balanced evidence in favour for5,10–13 and against8,9,14 a significant decrease in PaCO 2 at peak exercise, the assumption is generally accepted that ventilation inefficiency occurs in the absence of significant changes in CO2 tension.15 A salient point that has not been addressed before and that represents the primary objective of this study is the exploration of the relative contribution of VD /VT and PaCO 2 to the prognostic information provided by the VE /VCO 2 slope. This may represent a step forward to a more precise identification of the mechanisms underlying an excessive exercise ventilation, and possibly toward more appropriate therapeutic interventions.

Methods Study population The study comprised patients with CHF due to either ischaemic or idiopathic dilated cardiomyopathy, who were referred to the Cardiopulmonary Laboratory at San Paolo Hospital or at the Institute of Cardiology, University of Milan, for CHF evaluation. Assessment included echocardiography, lung function, and symptom-limited cardiopulmonary exercise testing. We restricted the analysis to patients who had been in a stable clinical condition for at least 4 months before evaluation, and under a stable therapeutic regimen prescribed by the referring physician, which was optimized during the hospital stay. We excluded patients who presented with anginal symptoms, had undergone a coronary artery bypass procedure in the previous 6 months, had primary pulmonary disorders and/or a forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC) ratio , 70% of the predicted normal value, had primary valvular heart disease, or had a history of smoking more than 10 cigarettes per day during one of the past 5 years. The primary endpoint was death for cardiac reasons. To assess vital status we resorted to the combined use of administrative and clinical databases; records during re-admission and outpatient records during follow-up (most patients attended our outpatient clinic) were reviewed. When vital status could not be determined by these methods, patients or their families were interviewed by telephone.

473

Echocardiography Two-dimensional and Doppler echocardiography was performed by standard methods. Left ventricular end-systolic and enddiastolic chamber dimensions and volumes were quantified by standard techniques, using the area–length method to measure ejection fraction.

Pulmonary function tests Spirometry was performed with equipment (Vmax Spectra, Sensomedics, Yorba Linda, CA, USA.) that met the American Thoracic Society performance criteria.16 To adjust for height, age, and sex, published prediction equations for FEV1 and FVC17 were used. Lung diffusion capacity for carbon monoxide (DLCO ) was determined twice with washout intervals of at least 4 min (the average was taken as the final result) with a standard single breath technique.

Cardiopulmonary exercise testing Patients underwent symptom-limited cardiopulmonary exercise testing with a respiratory gas exchange measurement. A personalized ramp protocol was used.18 A 12-lead electrocardiogram, heart rate, and blood pressure were obtained at rest and at each minute during exertion. For breath-by-breath gas exchange measurements, a Sensor Medics metabolic cart (Vmax Spectra, Sensomedics, Yorba Linda, CA, USA) was utilized. Minute ventilation [VE , BTPS (body temperature, atmospheric pressure saturated with water vapour)], O2 uptake [VO 2, STPD (standard temperature and pressure dry)], CO2 output (VCO 2, STPD), and other exercise variables were computer-calculated breath-by-breath, interpolated second-by-second, and averaged at 10 s intervals. The VD /VT ratio was derived from PaCO 2, according to the following formula:19 VD ðPaCO2 2 PECO2 Þ=ðPaCO2 2 VDappÞ ¼ VT VT where PECO 2 is the mean expiratory pressure of CO2 and VD app is the dead space of the breathing apparatus. The VE /VCO 2 slope was measured by linear regression, excluding the non-linear part of the data after the onset of ventilatory compensation for metabolic acidosis. Peak VO 2 was defined as the highest VO 2 observed during the exercise test. Age-, gender-, and weight-adjusted predicted VO 2 values were also determined by using the regression equations of Wasserman et al. 18 Anaerobic threshold (AT) was determined using the V-slope method.20 Blood gases (PaO 2, PaCO 2) and pH were measured at rest and just before peak exercise, on arterialized capillary blood samples from the hyperaemic earlobe.

Statistical methods The prognostic value of VE /VCO 2 slope, peak PaCO 2 and peak VD / VT, and other clinical variables (age, ejection fraction, drug therapy, peak VO 2, and peak VT ) were analysed by means of the Kaplan–Meier approach with log-rank testing and by univariate Cox regression analysis. Considering that, for the assessment of ventilatory efficiency, some authors have simply used the VE / VCO 2 ratio at peak exercise2 or the VE /VCO 2 ratio at AT,21,22 these two variables were also included in the univarate approach. The cut-off values for high VE /VCO 2 slope, peak VE /VCO 2 ratio, VE /VCO 2 ratio at AT, age, VD /VT, and those for low EF, peak VO 2 and peak VT were based on median values derived from the

474

M. Guazzi et al.

heart failure cohort. Selection of median values as cut-off was motivated by the lack of an established clinical predictive cut-off for exercise peak PaCO 2 and peak VD /VT (this is the first study investigating the prognostic power of these variables). Multivariate Cox regression models together with Shoenfeld residual analysis was used to assess the prognostic relevance of the VE /VCO 2 slope, of its determinants PaCO 2 and VD /VT, and of other possible predictors. Model construction was based on a backward approach with initial selection of covariates on the basis of results of univariate analyses. In order to evaluate the independent prognostic value of VE / VCO 2 slope, of peak PaCO 2 and peak VD /VT, two multivariate Cox regression analyses were carried out. Specifically, the first model was performed in order to confirm the VE /VCO 2 slope prognostic power. As VE /VCO 2 is a function of VD /VT and PaCO 2, the second analysis was performed by excluding VE /VCO 2 and including its two determinants PaCO 2 and VD /VT. Both models were adjusted for the clinical variables statistically significant at the univariate analysis. Four Kaplan–Meier curves for 6.5 year survival were plotted: one, for the patients with normal and for those with high VE /VCO 2 slope (Figure 1 ); two, using as the discriminatory parameter peak PaCO 2 (Figure 2 ) and peak VD /VT (Figure 3 ), respectively; one relating survival to the combination of both peak PaCO 2 and peak VD /VT (Figure 4 ). The Student’s t-test for unpaired values was used to compare the means of groups for quantitative variables. Data are presented as means + SD. The level of statistical significance was set at two-tailed P value , 0.05.

Figure 3 Kaplan–Meier plot relating survival to peak exercise VD /VT.

Figure 4 Kaplan–Meier plot relating survival to combination of peak PaCO 2 and peak VD /VT.

Pearson correlation analyses were used to assess the association between VE /VCO 2 slope and peak VD /VT and peak PaCO 2. All the analyses were performed with the statistical package STATA 7.0 (New Station, TX, USA).

Results Figure 1 Kaplan–Meier plot relating survival to VE /VCO 2 slope.

Follow-up on survival One hundred and twenty-eight consecutive patients who met the entry criteria were selected for follow-up. No patients were lost to follow-up. The mean duration of follow-up was 31.3 + 20 months (median 25 months). During this period there were 24 deaths for cardiac reasons. Three patients with an implantable cardioverter defibrillator (ICD) had ventricular fibrillation successfully terminated by the ICD. None of the patients underwent heart transplantation.

Baseline characteristics

Figure 2 Kaplan–Meier plot relating survival to peak exercise PaCO 2.

Table 1 reports the clinical characteristics of the patient population. Mean age of participants was 60+9 years and left ventricular ejection fraction averaged 34 + 10%. Ischaemic cardiomyopathy was the predominant (56%)

Prognostic value of PaCO 2 in heart failure

475

Table 1 Clinical characteristics of the study population Age, years Gender, M/F, % Height, cm Weight, kg Ischaemic, non-ischaemic, % NYHA class LV ejection fraction, % Therapy distribution, % Digoxin Diuretics ACE-inhibitors Nitrates b-blockers AT 1blockers Aldosterone antagonists Amiodarone Pulmonary function FEV1, L % predicted FVC, L % predicted DLco, mL . min21 . mmHg21 % predicted

60 + 9 79/21 169 + 7 76 + 12 56/44 2.0 + 0.8 34 + 10 34 72 76 16 30 8 33 32 2.4 + 0.6 85 + 17 3.0 + 0.7 75 + 15 20 + 6 76 + 20

Blood gases PaCO 2 at rest, mmHg PaO 2 at rest, mmHg pH at rest Peak PaCO 2, mmHg Peak PaO 2, mmHg pH at peak exercise

39.5 + 5 104 + 8 7.43 + 0.02 38 + 5 110 + 15 7.37 + 0.03

Cardiopulmonary exercise testing Heart rate at rest, b.p.m. Heart rate at peak exercise, b.p.m. Peak VO 2, mL . min21 . kg21 % predicted VO 2 AT, mL . min21 . kg21 Peak O2 pulse, mL . beat21 VE , L Peak VD /VT Peak VE /VO 2 VE /VCO 2 AT Peak VE /VCO 2 VE /VCO 2 slope

80 + 17 139 + 30 16.5 + 4.4 60 + 16 10 + 3.0 9.5 + 2.9 49 + 19 0.22 + 0.05 44 + 10 39 + 9 40 + 8 34 + 8

Data are presented as means + SD.

cause of CHF. All patients had symptomatic heart failure with a mean New York Heart Association (NYHA) class of 2.0 + 0.8. Current medical therapy included ACE-inhibitors (76%), digoxin (34%), diuretics (72%) b-blockers (30%), amiodarone (32%), and aldosterone antagonists (33%).

Lung function test and blood gases The mean FEV1, FVC, and DLco were 85, 75, and 76% of predicted normal value, respectively. When patients were grouped according to the median value of peak VD /VT (0.22) and of peak PaCO 2 (35 mmHg), significant differences in FEV1, FVC, and DLco were detected

between the two groups (P , 0.05; Tables 2 and 3 ). In the whole population, average values of arterial blood gases both at rest and at peak exercise were within normal limits. However, when patients were grouped according to the median peak VD /VT and peak PaCO 2 cut-off, there were significant differences in PaCO 2 and pH at peak exercise (Tables 2 and 3 ).

Cardiopulmonary exercise testing As shown in Table 1, patients presented with a moderate exercise limitation (average peak VO 2: 16.5 + 4.4 mL . min21 . kg21 corresponding to 60% of maximum predicted). Again, when patients were grouped according to peak VD /VT and peak PaCO 2 median value, those with the lower VD /VT and higher PaCO 2 presented with significantly higher peak VO 2, VO 2 at AT O2 pulse and peak tidal volume, and lower peak

Table 2 Pulmonary function, cardiopulmonary exercise testing data, and arterial blood gases in the study population according to exercise peak VD /VT median value Peak VD /VT , 0.22 (n ¼ 64)

Peak VD /VT
0.22 (n ¼ 64)

Pulmonary function FEV1, L % predicted FVC, L % predicted DLco, mL . min21 . mmHg21 % predicted

2.55 + 0.6 89 + 14 3.2 + 0.8 79 + 15 22 + 5 83 + 20

2.1 + 0.5 79 + 18 2.7 + 0.7 70 + 15 18 + 5 70 + 18

Blood gases PaCO 2 at rest, mmHg PaO 2 at rest, mmHg pH at rest Peak PaCO 2, mmHg Peak PaO 2, mmHg pH at peak exercise

40.5 + 5.0 104 + 8 7.45 + 0.02 37 + 5 111 + 8 7.37 + 0.03

39.4 + 4.0 103 + 7 7.43 + 0.03  34 + 5 109 + 8 7.29 + 0.02

Cardiopulmonary exercise testing Heart rate at rest, b.p.m. Heart rate at peak exercise, b.p.m. Peak VO 2, mL . min21 . kg21 % predicted VO 2 AT, mL . min21 . kg21 Peak O2 pulse, mL . beat21 VE , L Tidal volume, L Peak VD /VT Peak VE /VO 2 VE /VCO 2 AT Peak VE /VCO 2 VE /VCO 2 slope

77 + 15 141 + 26 18.2 + 4.5 64 + 15 11.0 + 2.4 10.2 + 2.4 52 + 20 1.84 + 0.5 0.19 + 0.04 42 + 10 36.0 + 6.4 37.0 + 7.0 32.0 + 6.0

Data are presented as means + SD.  P , 0.05 vs. VD /VT , 0.22.  P , 0.01 vs. VD /VT , 0.22.

82 + 18 133 + 33 

13.7 + 3.6 54.4 + 15.7 9.22 + 2.3 8.39 + 3.15 42 + 16 1.56 + 0.3  0.27 + 0.03  47.5 + 10.9  42.5 + 10.4  42.6 + 9.9  36.8 + 9.96

476

M. Guazzi et al.

VD /VT, peak VE /VO 2, VE /VCO 2 AT, peak VE /VCO 2, and VE /VCO 2 slope (Tables 2 and 3 ).

Table 3 Pulmonary function, cardiopulmonary exercise testing data, and arterial blood gases in the study population according to exercise peak PaCO 2 median value Peak PaCO 2
35 (n ¼ 64) Pulmonary function FEV1, L % predicted FVC, L % predicted DLco, mL . min21 . mmHg21 % predicted Blood gases PaCO 2 at rest, mmHg PaO 2 at rest, mmHg pH at rest Peak PaCO 2, mmHg Peak PaO 2, mmHg pH at peak exercise Cardiopulmonary exercise testing data Heart rate at rest, b.p.m. Heart rate at peak exercise, b.p.m. Peak VO 2, mL . min21 . kg21 % predicted VO 2 AT, mL . min21 . kg21 Peak O2 pulse, mL . beat21 VE , L Tidal volume, L Peak VD /VT Peak VE /VO 2 VE /VCO 2 AT Peak VE /VCO 2 VE /VCO 2 slope

Univariate analysis

Peak PaCO 2 ,35 (n ¼ 64)

2.5 + 0.7 88 + 15 3.1 + 0.7 78 + 15 21 + 6 80 + 20

2.2 + 0.4  80 + 17  2.8 + 0.7 71 + 15 19 + 5 72 + 18

40 + 4 105 + 7 7.44 + 0.03 38 + 5 112 + 7 7.38 + 0.03

39 + 5 102 + 8 7.42 + 0.02 33 + 5 109 + 8 7.28 + 0.02

80 + 17 143 + 32

80 + 18 136 + 29

17.6 + 4.0 63 + 15 10.8 + 2.5 9.7 + 2.8 49 + 15 1.8 + 0.4 0.21 + 0.05 38.5 + 6.5 34 + 5 34.7 + 5.5 30.0 + 4.6

15.4 + 4 57 + 16 9.7 + 2.6 9.1 + 3.0 49 + 23 1.6 + 0.3 0.23 + 0.05 50.0 + 9.5 43 + 10 44 + 8 38 + 9

Results of the univariate analysis of factors known to influence prognosis are reported in Table 4. VE /VCO 2 slope and peak exercise PaCO 2 emerged as stronger predictors. Among the therapeutic drug regimen, aldosterone antagonists were found to be the only significant predictors of survival.

Multivariate analyses Tables 5 and 6 report the results of the Cox regression analysis and Shoenfeld test of proportional hazard assumption. Two models have been used: the first model (Table 5 ) included ejection fraction, use of aldosterone antagonists, peak VE/VO2, and peak VO 2; an abnormally steep VE /VCO 2 slope was found to be the only predictor of death [HR 5.84, CI (1.692–20.197), P ¼ 0.005]. In the second model (Table 6 ), after adjusting for ejection fraction, aldosterone antagonist therapy, peak VE /VO 2, peak VO 2, peak PaCO 2, and peak VD /VT, the strongest independent predictors of death were peak PaCO 2 [HR 4.65, CI (1.695–12.751), P ¼ 0.003], peak VD /VT [HR 2.51, CI (1.067–5.931), P ¼ 0.035] and ejection fraction [HR 2.96, CI (0.151–7.646), P ¼ 0.024].

Survival analysis The Kaplan–Meier 6.5-year survival approach evidenced a survival of 38% for patients with a VE /VCO 2 slope median value
32.6 compared with 94% survival for those with a median value ,32.6 (Figure 1 ). When analyses were performed according to the VE /VCO 2 slope determinants, PaCO 2 and VD /VT at peak exercise, the survival was 40% (Figure 2 ) and 46% (Figure 3 ), respectively. Remarkably,

Data are presented as means + SD.  P , 0.05 vs. PaCO 2
35.  P , 0.01 vs. PaCO 2
35.

Table 4 Univariate predictors of death (n ¼ 24)

a

Age, years LV ejection fraction, %b ACE-inhibitorsc b-Blockersc Aldosterone antagonistsc Peak VO 2, mL . min21 . kg21b VE /VCO 2 ATa Peak VE /VCO a2 VE /VCO 2 slopea Peak VE /VO a2 Tidal volume, Lb Peak VD /VTa Peak PaCO 2, mmHgb a

Discrete dummy variable ‘high vs. low’. Discrete dummy variable ‘low vs. high’. Discrete dummy variable ‘no vs. yes’.

b c

Hazard ratio

SE

z

P.z

95%

1.4429 3.1776 0.5685 0.6185 3.2992 2.5426 2.2093 2.2090 7.7488 2.6249 1.6629 2.2356 1.6629

0.6001 1.5021 0.2478 0.3153 1.3842 1.1599 0.9939 1.0458 4.7896 1.1996 0.7038 0.9694 2.6360

0.882 0.014 21.295 0.942 2.845 2.046 1.762 1.674 3.313 2.112 1.202 1.855 3.207

0.378 0.014 0.195 0.346 0.004 0.041 0.078 0.094 0.001 0.035 0.230 0.064 0.001

0.638–3.260 1.258–8.025 0.2419–1.336 0.2277–1.680 1.449–7.508 1.039–6.217 0.914–5.335 0.873–5.587 2.307–26.023 1.071–6.429 0.725–3.811 0.955–5.229 1.892–14.043

Prognostic value of PaCO 2 in heart failure

477

Table 5 Multivariate predictors of death: results of Cox regression analysis (model 1) and Shoenfeld test of proportional hazard assumption after adjusting for clinical variables significant in the univariate approach

LV ejection fraction, %a VE /VCO 2 slopeb Peak VO 2, mL . min21 . kg21a

Hazard ratio

SE

z

P.z

95%

2.2066 5.8464 2.0988

1.0640 3.6979 0.9694

1.641 2.792 1.605

0.101 0.005 0.108

0.857–5.677 1.692–20.197 0.848–5.1896

rho

x2

df

Prob . x2

1 1 1 3

0.5198 0.8993 0.6292 0.8846

LV ejection fraction, % VE /VCO 2 slope Peak VO 2, mL . min21 . kg21 Global test a

0.13285 20.2637 20.09650

0.41 0.02 0.23 0.65

Discrete dummy variable ‘low vs. high’. Discrete dummy variable ‘high vs. low’.

b

Table 6 Multivariate predictors of death: results of Cox regression analysis (model 2) and Shoenfeld test of proportional hazard assumption after adjusting for clinical variables significant in the univariate approach

LV ejection fraction, % Peak PaCO 2, mmHga Peak VD /VT b

a

Hazard ratio

SE

z

P.z

95% CI

2.9677 4.6502 2.5168

1.4330 2.2934 1.1009

2.253 2.986 2.110

0.024 0.003 0.035

0.151–7.646 1.695–12.751 1.067–5.931

rho

x2

df

Prob . x2

0.16102 0.00923 20.08212

0.63 0.00 0.16 0.86

1 1 1 3

0.4288 0.9625 0.6884 0.8358

LV ejection fraction, % Peak PaCO 2, mmHg Peak VD /VT Global test a

Discrete dummy variable ‘low vs. high’. Discrete dummy variable ‘high vs. low’.

b

survival of patients with both a low peak exercise PaCO 2 and a high peak VD /VT was as low as 25% (Figure 4 ).

Correlation analyses As shown in Figure 5A, a strong inverse correlation was found between VE /VCO 2 slope and peak PaCO 2 (r ¼ 20.69; P ¼ 0.0001); 85% of non-survivors were distributed in the area with a lower peak PaCO 2 (,35 mmHg) and a higher VE /VCO 2 slope (
32.4). A weaker positive correlation was found between VE /VCO 2 slope and peak VD /VT (r ¼ 0.37; P ¼ 0.001) (Figure 5B ); 58% of non-survivors were distributed in the area with a higher peak VD /VT (
0.22) and a higher VE /VCO 2 slope (
32.4) (Figure 5B ). No correlation was found between peak PaCO 2 and peak VD /VT (r ¼ 20.152, P ¼ 0.086).

Discussion There are two novel findings in the present study. First, a low peak exercise PaCO 2 and an elevated peak VD /VT are strong independent predictors of mortality in stable CHF patients. Second, a low peak exercise PaCO 2 is the

more significant determinant of the prognostic value of a steep VE /VCO 2 slope. A consistent association between increased levels of ventilation and cardiovascular mortality has been documented in several studies across different population groups.1–7 Mostly, it has been shown that an abnormally high VE /VCO 2 slope predicts an increased risk of death among CHF patients to an even greater degree than peak VO 2.2,3,6,7 The relationship between a high VE /VCO 2 slope and survival has been further strengthened by the recent landmark finding that even in patients with normal exercise performance and peak VO 2 (
18 mL . min21 . kg21), an abnormal exercise ventilation significantly discriminates survival.4 In addition, the VE /VCO 2 slope is independent of motivation, showing minimal variability at 30, 60, and 100% of exercise capacity,23 and retains a high prognostic power when measured at submaximal constant workloads.24 The current study confirms that VE /VCO 2 slope is a stronger predictor than peak VO 2, and expands, in some important respects, information provided by previous reports. In fact, it is the only study aimed at analysing the prognostic significance of the physiological determinants of ventilation in CHF patients. Furthermore, when peak PaCO 2, peak VD /VT, and peak VO 2 were considered together, the ventilatory variables emerged

478

M. Guazzi et al.

Figure 5 Correlation between VE /VCO 2 slope and peak PaCO 2 (A ), and peak VD /VT (B ). Survivors (open circles); non-survivors (filled circles).

as stronger and independent predictors of death, indicating that they are not simply a function of lower workload achieved.

Increased VE /VCO 2 slope: pathophysiological bases The precise pathophysiological substrates that predispose to, and sustain an excessive ventilation in, CHF patients remain controversial.15 Mathematically, the VE /VCO 2 slope is determined by three factors: the rate of CO2 production, the physiological VD /VT, and the PaCO 2. Therefore, for a given VCO 2, an increased VE /VCO 2 slope has multiple possible substrates: (i) an augmented central and/or peripheral command to ventilation, which drives the PaCO 2 below the physiological range; (ii) a large dead space which requires an increase in ventilation to maintain a normal PaCO 2; and (iii) an early occurrence of metabolic acidosis which demands ventilatory compensation. Initial studies by Sullivan et al.8 reported that an increased VD /VT, in the face of a preserved neural control of ventilation and normal blood gases, is responsible for the augmented VE /VCO 2 slope. In 130 patients with different CHF severity, Wasserman et al. 9 reproduced the same findings, identifying structural changes intrinsic to the lung (restrictive lung changes) and reduced lung perfusion, as responsible for the occurrence of a high VD /VT and consequent ventilation/perfusion mismatch. In these reports, changes in pH and PaCO 2 between rest and peak exercise were minimal, suggesting no differences in the PaCO 2 set point compared with healthy subjects. In contrast, other studies in CHF patients have reported a significant reduction in PaCO 2 at peak exercise.5,10,11,13 Notably, in a population similar to ours, Hachamovitch et al. 12 reported a significant reduction in PaCO 2 even during a submaximal constant workload at 50 W. The link, however, between an increased VD /VT and a reduced PaCO 2 remains elusive. In the present report, both an increased VD /VT and a reduced

peak exercise PaCO 2 were documented and for both of them a significant correlation with VE /VCO 2 slope was found. Chua et al. 25 observed that an augmented ventilatory response to exercise in CHF is significantly correlated with an impaired central and peripheral control of ventilation. The same group of investigators has expanded these observations providing impressive evidence of an abnormal cardiorespiratory reflex control, related to both activation of chemoreceptors and fibres originating from the working muscle (i.e. ergoreflex stimuli),26,27 and an abnormal autonomic baroreflex control of circulation.28

PaCO 2 vs. VD /VT at peak exercise: prognostic insights It is remarkable that in CHF the only link between augmented exercise ventilation and prognosis is provided by studies assessing the ergoreflex and chemoreflex activation.1,4,27 In these reports, however, neither were blood gas measurements obtained, nor peak exercise VD /VT calculated, and their possible relationship with survival has remained undefined. Our study sheds light on the prognostic power of these variables, and multivariate Cox regression analysis has identified both arterial PaCO 2 and VD /VT at peak exercise as survival predictors. PaCO 2 at peak exercise, however, retained a greater prognostic significance than peak VD /VT and was associated with a higher hazard ratio (4.65 vs. 2.51). The decrease in peak exercise PaCO 2 may be both the trigger (development of metabolic acidosis due to inadequate cardiac output), or the consequence (excessive ventilatory response to increasing CO2 output) of the increased ventilatory response to exercise. It is very likely that these two mechanisms were additive in influencing PaCO 2 changes. Although we cannot rule out an excessive ventilatory drive secondary to chemoreflex and/or ergoreflex activation, it is noteworthy that patients who developed hypocapnia and metabolic

Prognostic value of PaCO 2 in heart failure

acidosis were exposed to a higher risk of death. Whatever the underlying mechanisms, our data are in favour of a peripheral substrate, rather than of a pulmonary ventilation/perfusion mismatching as a determinant of the prognostic significance of the ventilatory response to exercise in CHF.

Study limitations The number of patients in this study was relatively small. It should be considered, however, that the average follow-up was quite long. The analysis included only CHF patients who were in stable clinical conditions and were able to perform a symptom-limited exercise test. However, this was an unselected ambulatory CHF population. The average peak VO 2 of 16.5 + 4.4 mL . min21 . kg21 reflected a mild to moderate exercise limitation, and results might not apply to advanced CHF. Nevertheless, because patients with advanced CHF and severe exercise limitation may exhibit hypocapnia and metabolic acidosis from the very early exercise stages, it is conceivable that an abnormally low arterial PaCO 2 at peak exercise may bear an even more significant prognostic power in this subset of patients.

Conclusions An excessively low peak exercise PaCO 2 and an abnormally high peak VD /VT (the two determinants of ventilation for a given CO2) are both strong predictors of death among stable CHF patients and their prognostic value is independent of peak VO 2. PaCO 2 emerges as the more significant determinant of the VE /VCO 2 slope, suggesting that regulatory mechanisms involved in the tight control of ventilatory command and blood gas tension, rather than lung function abnormalities, play a critical pathophysiological role in the exercise ventilation inefficiency of CHF patients.

Acknowledgements This study was supported, in part, by a grant from the Luigi Berlusconi Foundation.

References 1. Chua TP, Ponikowski P, Harrington D, Anker SD, Webb-People K, Clark AL, Poole-Wilson PA, Coats AJ. Clinical correlates and prognostic significance of the ventilatory response to exercise in chronic heart failure. J Am Coll Cardiol 1997;29:1585–1590. 2. Robbins M, Francis G, Pashkow FJ, Snader CE, Hoerker K, Young JB, Lauer MS. Ventilatory and heart rate response to exercise. Better predictors of heart failure mortality than peak oxygen consumption. Circulation 1999;100:2411–2417. 3. Kleber FX, Vietzke G, Wernecke KD, Bauer U, Opitz C, Wensel R, Sperfeld A, Glaser S. Impairment of ventilatory efficiency in heart failure. Circulation 2000;101:2803–2809.

479 4. Ponikowski P, Francis DP, Piepoli MF, Davies LC, Chua TP, Davos CH, Florea V, Banasiak W, Poole-Wilson PA, Coats AJ, Anker SD. 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;103:967–972. 5. Gitt A, Wasserman K, Kikowski C, Kleeman T, Kikowski A, Bangert M, Schneider S, Schwarz A, Senges J. Exercise anaerobic threshold and ventilatory efficiency identify heart failure patients for high risk of early death. Circulation 2002;106: 3079–3084. 6. Francis DP, Shamin W, Davies LC, Piepoli MF, Ponikowski P, Anker SD, Coats AJ. Cardiopulmonary exercise testing for prognosis in chronic heart failure: continuous and independent prognostic value from VE/VCO(2) slope and peak VO(2). Eur Heart J 2000; 21:154–161. 7. Guazzi M, De Vita S, Cardano P, Barlera P, Guazzi MD. Normalization for peak oxygen uptake increases the prognostic power of the ventilatory response to exercise in patients with chronic heart failure. Am Heart J 2003;146:542–548. 8. Sullivan M, Higginbotham M, Cobb F. Increased exercise ventilation in patients with chronic heart failure: intact ventilatory control despite hemodynamic and pulmonary abnormalities. Circulation 1988;77:552–559. 9. Wasserman K, Zangh YY, Gitt A, Belardinelli R, Koike A, Lubarsky L, Agostoni PG. Lung function and exercise gas exchange in chronic heart failure. Circulation 1997;96:2221–2227. 10. Wilson JR, Ferraro N. Exercise intolerance in patients with chronic heart failure: relation to oxygen transport and ventilatory abnormalities. Am J Cardiol 1983;51:1358. 11. Wilson JR, Ferraro N, Weber KT. Respiratory gas analysis during exercise as a noninvasive measure of lactate concentration in chronic congestive heart failure. Am J Cardiol 1983; 51:1639. 12. Hachamovitch R, Brown HV, Rubin SA. Respiratory and circulatory analysis of CO2 output during exercise in chronic heart failure. Circulation 1991;84:605–612. 13. Perego GB, Marenzi G, Guazzi M, Sganzerla P, Assanelli E, Palermo P, Conconi B, Lauri G, Agostoni PG. Contribution of PO2, P50 and Hb to changes in arteriovenous O2 content during exercise in heart failure. J Appl Physiol 1996;80:623–631. 14. Franciosa JA, Leddy CL, Wilen M, Schwartz DE. Relation between hemodynamic and ventilatory responses in determining exercise capacity in severe congestive heart failure. Am J Cardiol 1984; 53:127–134. 15. Johnson RL Jr. Gas exchange efficiency in chronic heart failure. Circulation 2001;103:916–918. 16. American Thoracic Society. Standardization of spirometry: 1987 update. Am Rev Resp Dis 1987;136:1285–1298. 17. Hankinson JL, Odenerantz JR, Fedan KB. Spirometric reference values from a sample of the general US population. Am J Resp Crit Care Med 1999;159:179–187. 18. Wasserman K, Hansen JE, Sue DY et al. Principles of Exercise Testing and Interpretation. 2nd ed. Philadelphia: Lea & Febiger; 1994. 19. Guazzi M, Marenzi GC, Assanelli E, Perego GB, Cattadori G, Doria E, Agostoni PG. Evaluation of dead space to tidal volume ratio in patients with chronic heart failure. J Cardiac Failure 1995;5: 405–408. 20. Beaver WL, Wasserman K, Whipp BJ. A new method for detecting the anaerobic threshold by gas exchange. J Appl Physiol 1986;60: 2020–2027. 21. Wada O, Asanoi H, Miyagi K, Ishikaza S, Kameyama T, Seto H, Sasayama S. Importance of abnormal lung perfusion in excessive exercise ventilation in chronic heart failure. Am Heart J 1993;125: 790–798. 22. Sun XG, Hansen JE, Garatochea N, Storer TW, Wasserman K. Ventilatory efficiency during exercise in healthy subjects. Am J Respir Crit Care Med 2002;166:1443–1448. 23. Metra M, Dei Cas L, Panina G, Visioli O. Exercise hyperventilation in chronic congestive heart failure, and its relation to functional capacity and hemodynamics. Am J Cardiol 1992,70: 622–628.

480 24. Koike A, Itoh H, Kato M, Sawada H, Aizawa T, Tai Fu L, Watanabe H. Prognostic power of ventilatory responses during submaximal exercise in patients with chronic heart failure. Chest 2002;121:1581–1588. 25. Chua TP, Clark AM, Amadi AA, Coats AJ. The relationship between chemosensitivity and the ventilatory response to exercise in chronic heart failure. J Am Coll Cardiol 1996;27:650–657. 26. Piepoli M, Clark AM, Volterrani M, Adamopoulos S, Sleight P, Coats AJ. Contribution of muscle afferents to the hemodynamic, autonomic, ventilatory responses to exercise in patients with chronic heart failure. Circulation 1996;93:940–952.

M. Guazzi et al. 27. Ponikowski PP, Chua TP, Francis DP, Capucci A, Coats AJ, Piepoli MF. Muscle ergoreceptor overactivity reflects deterioration in clinical status and cardiorespiratory reflex control in chronic heart failure. Circulation 2001;6:2324–2330. 28. Ponikowski P, Chua TP, Piepoli M, Ondusova D, Webb-People K, Harrihgton D, Anker SD, Volterrani M, Colombo R, Mazzuero G, Giordano A, Coats AJ. Augmented peripheral chemosensitivity as a potential input to baroreflex impairment and autonomic imbalance in chronic heart failure. Circulation 1997;96: 2586–2594.