OUP user menu

Exercise oscillatory breathing in diastolic heart failure: prevalence and prognostic insights

Marco Guazzi , Jonathan Myers , Mary Ann Peberdy , Daniel Bensimhon , Paul Chase , Ross Arena
DOI: http://dx.doi.org/10.1093/eurheartj/ehn437 2751-2759 First published online: 4 October 2008


Aims Exercise intolerance occurs in both systolic and diastolic heart failure (HF). Exercise oscillatory breathing (EOB) is a powerful predictor of survival in patients with systolic HF. In diastolic HF, EOB prevalence and prognostic impact are unknown.

Methods and results A total of 556 HF patients (405 with systolic HF and 151 with diastolic HF) underwent cardiopulmonary exercise testing (CPET). Diastolic HF was defined as signs and symptoms of HF, a left ventricular ejection fraction ≥50%, and a Doppler early (E) mitral to early mitral annulus ratio (E′) ≥8. CPET responses, EOB prevalence and its ability to predict cardiac-related events were examined. EOB prevalence in systolic and diastolic HF was similar (35 vs. 31%). Compared with the patients without EOB, patients with EOB and either systolic or diastolic HF had a higher New York Heart Association class, lower peak VO2 and higher E/E′ ratio (all P < 0.01). Univariate Cox regression analysis demonstrated that peak VO2, VE/VCO2 slope and EOB all were significant predictors of cardiac events in both systolic and diastolic HF. Multivariable analysis revealed that EOB was retained as a prognostic marker in systolic HF and was the strongest predictor of cardiac events in diastolic HF.

Conclusion EOB occurrence is similar in diastolic and systolic HF and provides relevant clues for the identification of diastolic HF patients at increased risk of adverse events.

  • Diastolic heart failure
  • Exercise oscillatory breathing


The occurrence of symptoms and signs of heart failure (HF) with normal ejection fraction (EF) and predominant anomalies of active relaxation and passive stiffness (i.e. diastolic HF) is increasingly recognized.1 This diagnosis carries a mortality risk that is currently estimated to be similar to that seen in systolic HF, ∼15% per year in older patients.2,3 Remarkably, in recent year's survival rate of systolic HF has improved, whereas the prognosis for diastolic HF remains unchanged.2 This is at least in part attributable to the fact that the definition, diagnosis, and natural history of diastolic HF are less well characterized.46

Limitation to physical activity occurs in conjunction with both systolic and diastolic HF and cardinal manifestations of exercise intolerance are a reduction in peak oxygen consumption (VO2),7 an excessive ventilation to CO2 output rate (VE/VCO2),811 and the occurrence of exercise oscillatory breathing (EOB).1214 These variables are established markers in the clinical and prognostic assessment of systolic HF, but their significance in diastolic HF has not been fully explored.

There has been a recent appreciation for the clinical relevance of EOB, a complex and intriguing pathophysiological phenomenon characterized by waxing and waning of tidal volume that leads to an oscillatory kinetics in measured VO2 and VCO2, reflecting an instability in central and peripheral controllers of ventilation. Clinically, interest in EOB has arisen because of its strong prognostic power; in fact, recent reports have demonstrated that EOB is actually superior to both peak VO2 and the VE/VCO2 slope1517 and may be valuable in the prediction of arrhythmic cardiac death.18

EOB has been substantially characterized in patients with systolic HF, but no information is available on its prevalence or prognostic utility in diastolic HF. Such information would be helpful in order to better stratify diastolic HF patients at increased risk. Therefore, the present study was aimed at elucidating the prevalence and the prognostic relevance of EOB in diastolic HF compared with systolic HF.


This was a multicentre study consisting of HF patients referred to the cardiopulmonary exercise laboratories at San Paolo Hospital, Milan, Italy, Virginia Commonwealth University, Richmond, Virginia, USA, VA Palo Alto Health Care System, Palo Alto, California, USA and LeBauer Cardiovascular Research Foundation, Greensboro, NC, USA. A total of 556 consecutive subjects diagnosed with HF who underwent a symptom-limited cardiopulmonary exercise testing (CPET) between June 1998 and June 2007 were included. Subjects with significant obstructive lung disease, evidenced by a forced expiratory volume in 1 s ≤70%, or who were unable to perform a maximal exercise test, were excluded from the study.

Smokers were 15–20% of the population and were similarly distributed across different subgroups. All patients were in New York Heart Association (NYHA) functional classes II–III and 40 had been already hospitalized for acute HF.

Inclusion criteria for systolic HF were signs and symptoms of HF with a left ventricular (LV) EF ≤ 40%.

Inclusion criteria for diastolic HF consisted of a diagnosis based on (i) signs and symptoms of HF (ii) presence of preserved LV systolic function (EF ≥ 50%) as assessed by two-dimensional echocardiography, and (iii) documentation of a mitral inflow early (E) velocity to mitral annulus early velocity (E′) ≥8.19,20 Approval of the Institutional Review Board at each institution was obtained.

Cardiopulmonary exercise testing procedure and data collection

Symptom-limited CPET was performed in all patients after written informed consent had been obtained. Each centre used individualized ramp protocols and maximal tests were designed to obtain exercise duration between 8 and 10 min. USA centres utilized a treadmill and the Italian centre used a cycle ergometer. Potential bias related to different exercise modes can reasonably be excluded given that a previous study has shown that comparing these two modes of exercise yields identical predictive cut-off values for both peak VO2 and the VE/VCO2 slope.21 Ventilatory expired gas analysis was performed using a metabolic cart at all four centres (Medgraphics CPX-D, Minneapolis, MN or Sensormedics Vmax, Yorba Linda, CA, USA).

Standard 12-lead electrocardiograms were obtained at rest, each minute during exercise and for at least 5 min during the recovery phase; blood pressure was measured using a standard cuff sphygmomanometer. Minute ventilation (VE, BTPS), oxygen uptake (VO2, STPD), carbon-dioxide output (VCO2, STPD), and other cardiopulmonary variables were acquired breath-by-breath, averaged over 30 s, and printed using rolling averages every 10 s. The V-slope method was used to measure the anaerobic threshold (AT).22 Different methods for calculating the VE/VCO2 slope have been proposed. We measured this variable by including all data points from the beginning to the end of exercise, in agreement with previous studies by our group23 and others.24 Measurement of VE/VCO2 slope by using the entire data points may offer some advantages over the calculation of the first slope of VE/VCO2 relationship (from rest to the isocapnic buffering at ventilatory threshold). Specifically, in patients with EOB, calculation of first VE/VCO2 slope linear relationship may represent an issue, because of the scattered distribution of breath-by-breath values, making it impossible to detect the ventilatory threshold. It follows that the use of the whole VE/VCO2 relationship for both EOB and no-EOB patients allows a comparable measurement between groups. In addition, we deemed this calculation useful to prevent variability among laboratories.

Ten-second averaged VE and VCO2 data, from the initiation to peak exercise, were input into spreadsheet software (Microsoft Excel, Microsoft Corp., Bellevue, WA, USA) to calculate the VE/VCO2 slope through least squares linear regression (y= mx + b, m= slope). Ten-second averaged VE data were also analysed to determine if a subject demonstrated EOB during testing.

EOB was defined accordingly to the following criteria: (i) oscillations of ≥60% of exercise data at an amplitude of >15%; (ii) minimal average amplitude of ventilatory oscillation of 5 L (peak value minus the average of two in-between consecutive nadirs); (iii) a regular oscillation as defined by a SD of three consecutive cycle lengths (time between two consecutive nadirs) within 20% of the average.17,18 Different patters of oscillations may be encountered across the wide spectrum of HF populations. As depicted in Figure 1, some noisy oscillations (A) may occur at early exercise that are considered as no-EOB (absence of criteria above-reported); an EOB pattern that may last for the initial and intermediate phases of exercise and disappears after >60% of whole exercise (B) or EOB with cycles of very wide amplitude that last for the entire exercise duration (C). In patients with EOB that persisted for the whole exercise duration VO2 at peak exercise was calculated as average of the last 60 s of exercise.

Figure 1

Examples of different patterns of oscillatory ventilation. (A) Some noisy oscillations may occur at early exercise that are considered as no-exercise oscillatory breathing. (B) An exercise oscillatory breathing pattern that may last for the initial and intermediate phases of exercise and disappearing after 60% of whole exercise. (C) Exercise oscillatory breathing with very wide amplitude of cycles that persist for the entire exercise.

Test termination criteria consisted of symptoms (i.e. dyspnoea and/or fatigue), ventricular tachycardia, ≥2 mm of horizontal or down sloping ST segment depression, drop of systolic blood pressure ≥20 mmHg during progressive exercise. A qualified exercise physiologist with physician supervision conducted each exercise test.


LV chamber dimensions were evaluated using standard procedures. LVEF was calculated from two-dimensional apical images according to the Simpson method. LV mass was calculated with the formula proposed by Devereux et al.25

Conventional Doppler and tissue Doppler imaging measurements

Mitral inflow measurements included peak early (E) and peak late (A) flow velocities, the deceleration time (DT) of early mitral flow velocity, the E/A ratio and the isovolumic relaxation time (IVRT). The tissue Doppler imaging of the mitral annulus was obtained from the apical four-chamber view. A 1.5 sample was placed sequentially at the lateral and septal annular sites. Analysis was performed for the early (E′) and late (A′) diastolic peak velocities. The ratio of early transmitral flow velocity to annular mitral velocity of the lateral LV wall (E/E′) was taken as an estimate of LV filling pressure.26


Subjects were followed for cardiac-related death after CPET through hospital and outpatient medical chart review at the respective centres. Cardiac-related mortality was defined as death directly resulting from failure of the cardiac system. An example fitting this definition is myocardial infarction followed by cardiac arrest. Subjects were followed by the institution where CPET was conducted providing for the high likelihood that all major events were captured. Clinical follow-up to ascertain predefined outcome was obtained every 6 months. The last follow-up assessment was obtained at 4 years. Any death with a cardiac-related discharge diagnosis was considered an event. Clinicians conducting the CPET were not involved in decisions regarding the cause of death or heart transplant/LVAD implantation.

Statistical analysis

In an earlier study,11 the incidence of cardiac mortality at 1 year in patients with EOB vs. no-EOB was 30 vs. 12%, respectively; thus, if we hypothesized a similar association in this study, a population of 126 patients would be needed to detect this difference with α = 0.05 and a power of 0.90.

All continuous data are reported as means ± SD. Categorical data are reported as percentages or frequencies. Unpaired t-testing was used to compare continuous data between subjects with no-EOB and that with EOB. Differences between no-EOB and EOB patients for systolic vs. diastolic HF were tested by two-way analysis of variance (ANOVA). Both main (no-EOB vs. EOB and systolic vs. diastolic HF) and interaction (EOB*HF type) effects were assessed. All tests were two-sided.

Variables included in the Cox regression model were only exercise variables (peak VO2, VE/VCO2 slope and EOB). Peak VO2 and VE/VCO2 slope are the most established prognostic exercise indicators as suggested by several consistent studies. EOB is the new variable we are assessing, which has recently shown a strong prognostic value primarily in systolic HF.1618

Univariate Cox regression analysis was used to assess the prognostic ability of EOB in the overall group. Multivariable Cox regression analysis was performed by using the forward stepwise method with entry and removal of P-value for multivariate analyses set at 0.05 and 0.10, respectively.

For variables included in the Cox regression analysis, proportional hazards assumption was assessed by plotting log–log survival curves. These plots revealed all variables included in the analysis were parallel, satisfying the proportional hazards assumption. Kaplan–Meier analysis assessed the differences in survival between subjects with and without EOB in the overall group. The log-rank test determined statistical significance between groups in the Kaplan–Meier analysis. Statistical differences with a P < 0.05 were considered significant.


Follow-up on survival

Among the 630 patients initially assessed, 48 did not fulfil the eligibility criteria; 15 patients did not provide their consent to be included in the study follow-up and 11 patients were excluded because lost at follow-up. The median follow-up duration was 13 months with an interquartile range of 8 (25%) and 34 (75%) months for diastolic HF and of 6 (25%) and 31.3 (75%) months for systolic HF. During this period, there were 99 major events (78 deaths, 5 LVADs, 16 transplants), with an overall annual mortality rate of 12.4% for the systolic HF group, and 17 major events (16 deaths, 1 transplant) with an annual mortality rate of 5.7% for diastolic HF group. For diastolic HF group, the mean follow-up for survivors was 22.4 ± 19.5 months and that for non-survivors was 19.0 ± 18.4 months. For systolic HF group, the mean follow-up for survivors was 17.6 ± 15.4 months and that for non-survivors was 17.6 ± 17.5 months.

Baseline characteristics

The EOB prevalence in patients with systolic vs. diastolic HF was 35 and 31%, respectively. A total of 251 patients were tested on a bike and 305 on a treadmill. Interestingly, EOB prevalence was 25% among those tested on a bike vs. 41.3% among those tested on a treadmill. Nonetheless, the prognostic characteristics of EOB was the same for bike (hazard ratio: 3.6, 95% CI: 2.1–6.3, P < 0.001) and treadmill (hazard ratio: 3.2, 95% CI: 1.9–5.2, P < 0.001) tests.

Comparisons of key variables between no-EOB and EOB among subjects with systolic and diastolic HF are listed in Tables 1 and 2, respectively. For patients with systolic HF, no significant differences were detected between those with and without EOB in terms of age, gender distribution, aetiology of cardiomyopathy, and EF. Conversely, EOB patients exhibited a higher NYHA functional class and fewer were taking beta-blockers. ACE-inhibitors and diuretics were prescribed similarly between the subgroups. Among patients with diastolic HF, there were fewer females with EOB and NYHA functional class was higher. In addition, patients with EOB were more frequently prescribed RAS inhibitors and diuretics.

View this table:
Table 1

Demographic and clinical characteristics according to presence or absence of exercise oscillatory breathing in systolic heart failure

Overall group (n = 405)No-EOB (n = 263)EOB (n = 142)P-value (EOB vs. no-EOB)
Age, years56.0 ± 13.455.3 ± 13.257.1 ± 13.80.34
Gender (male/female), %71/2975/2569/310.26
LV ejection fraction, %25.3 ± 7.925.5 ± 8.025.2 ± 7.70.88
NYHA (average)2.4 ± .762.3 ± 0.762.6 ± .72<0.001
Heart rate at rest, beats/min74 ± 2472 ± 2874 ± 300.47
Blood pressure at rest, mmHg
 Systolic118 ± 25120 ± 26117 ± 260.43
 Diastolic72 ± 2474 ± 2570 ± 270.81
 Ischaemic (%)203 (50)123 (47)80 (56)0.12
 Hypertensive (%)122 (30)82 (31)40 (28)0.35
 Post-myocarditis (%)40 (10)28 (11)12 (8)0.44
 Idiopathic (%)40 (10)30 (11)10 (7)0.31
Drug therapy distribution
 Prescribed ACE-inhibitor (%)312 (77)200 (76)111 (78)0.60
 Prescribed diuretic (%)311 (76)184 (70)99 (70)0.98
 Prescribed beta-blocker (%)296 (73)200 (76)95 (67)0.007
 Prescribed antialdosterone agents (%)202 (50)131 (50)71 (50)0.95
 Prescribed inotropic agents (%)162 (40)121 (46)77 (54)0.55
  • NYHA, New York Heart Association.

View this table:
Table 2

Demographic and clinical characteristics according to presence or absence of exercise oscillatory breathing in diastolic heart failure

Overall group (n = 151)No-EOB (n = 103)EOB (n = 48)P-value (EOB vs. no-EOB)
Age, years58.5 ± 13.457.0 ± 12.661.8 ± 14.40.04
Gender (male/female), %67/3363/3775/250.005
Left ventricular ejection fraction, %47.8 ± 7.847.8 ± 7.947.6 ± 7.80.83
NYHA (average)2.0 ± 0.851.8 ± 0.782.6 ± 0.0<0.001
Heart rate at rest, beats/min78 ± 2877 ± 3078 ± 290.06
Blood pressure at rest, mmHg
 Systolic125 ± 26126 ± 26124 ± 260.67
 Diastolic84 ± 2583 ± 2484 ± 260.62
 Ischaemic (%)75 (50)50 (48)25 (52)0.73
 Hypertensive (%)45 (30)34 (33)11 (23)0.24
 Post-myocarditis (%)8 (5)6 (6)2 (4)0.1
 Idiopathic (%)23 (15)13 (13)10 (21)0.08
Therapy distribution
 Prescribed ACE-inhibitor (%)108 (72)46 (45)26 (55)0.015
 Prescribed diuretics (%)92 (61)52 (51)23 (49)0.12
 Prescribed beta-blockers (%)84 (56)46 (45)26 (55)0.018
 Prescribed antialdosterone agents (%)60 (40)49 (48)25 (52)0.52
 Prescribed inotropic agents (%)27 (18)51 (50)24 (50)0.64
  • NYHA, New York Heart Association.

Echocardiographic characteristics

Table 3 reports the echocardiographic data. The two-way ANOVA analysis documented a significant EOB main effect for both left ventricular end-diastolic volume (LVEDV) and left ventricular end-systolic volume (LVESV) (P = 0.04) with a significant main effect for systolic vs. diastolic HF (P < 0.001). Diastolic HF patients had a significantly lower LVEDV and LVESV (P < 0.01) vs. systolic HF. An EOB and HF type interaction was detected for diastolic HF patients with EOB who presented significantly lower LVEDV and LVESV compared with no-EOB systolic HF patients (P = 0.001). As to left ventricular mass (LVM), there was a main effect for EOB and patients with no-EOB had a significantly lower LVM compared with EOB patients (P = 0.02). The main effect for systolic vs. diastolic HF was also significant (P < 0.001). Systolic HF patients had a significantly higher LVM. A significant interaction effect between EOB and HF type was observed and diastolic HF patients with EOB had a significantly lower LVM vs. systolic HF patients and EOB (P = 0.003). A significant main effect for EOB was observed for LVM/LVEDV just in diastolic HF patients (P = 0.002). A significant interaction effect emerged between EOB and HF type with diastolic HF patients and EOB showing a higher LVM/LVEDV compared with systolic HF patients and EOB. No main or interaction effects emerged for IVRT, DT, and E/A ratio (P > 0.05). As to E/E' lateral, there was a main effect for EOB. No-EOB vs. EOB patients had a significantly lower E/E' lateral (P < 0.001). The main effect for systolic vs. diastolic HF was also significant (P = 0.001) in that patients with diastolic HF and no-EOB had a significantly lower E/E′ lateral vs. systolic HF with EOB. There was, however, no interaction effect between EOB and HF type (P = 0.61).

View this table:
Table 3

Echocardiographic data characteristics of systolic and diastolic heart failure

Systolic HFDiastolic HF
LVESV, mL107.0 ± 29.0122.0 ± 30.0*100.3 ± 32.0#88.6 ± 28.6*,§,#
LVEDV, mL185.0 ± 29.0200.0 ± 30.0*160.0 ± 31.0#140 ± 29.0*,§,#
LVM, g221.0 ± 21.0237.0 ± 20.0*219.0 ± 23.0§,#215.0 ± 22.7*,§,#
LVM/LVEDV, g/mL1.19 ± 2.01.20 ± 1.01.37 ± 1.51.53 ± 1.2*,§,#
Mitral inflow pattern and tissue Doppler
IVRT, ms99.0 ± 7.092.0 ± 6.092.0 ± 4.090.0 ± 3.0
E/A ratio1.13 ± 0.91.30 ± 0.81.22 ± 0.41.13 ± 0.3
DT, ms220 ± 30170 ± 28200 ± 27190 ± 27
E/E′ lateral9.0 ± 1.012.0 ± 3.0*9.0 ± 1.0§14.0 ± 2.0@,#
  • *P < 0.05 vs. no-EOB; @P < 0.01 vs. no-EOB; §P < 0.01 vs. EOB systolic HF; #P < 0.01 vs. no-EOB systolic HF. Abbreviations: LVESV, left ventricular end-systolic volume; LVEDV, left ventricular end-diastolic volume; LVM, left ventricular mass; IVRT, isovolumic relaxation time; E, early mitral flow velocity; A, late mitral flow velocity; DT, deceleration time; E′, early annular mitral velocity.

Cardiopulmonary data

Two representative cases of systolic and diastolic HF either showing a similar EOB pattern are reported in Figure 2.

Figure 2

Representative cases of exercise oscillatory breathing patterns in systolic (A) and diastolic (B) heart failure.

Exercise gas exchange data are summarized in Table 4. All patients exercised above their AT and the mean peak respiratory exchange ratio (RER) was ≥1.05 in all groups, suggesting that they developed significant metabolic acidosis and exercised close to maximal intensity. At the two-way ANOVA analysis, there was no main or interaction effect for peak RER. As to peak VO2 and VO2 at AT, there was a main effect for EOB. Patients with no-EOB had significantly higher values compared with the subjects with EOB (P < 0.001). The main effect for systolic vs. diastolic HF was not significant (P = 0.07). There was, however, an interaction effect between EOB and HF type. Patients with diastolic HF and no-EOB had a significantly higher peak VO2 compared with the subjects with systolic HF and no-EOB (P = 0.005).

View this table:
Table 4

Cardiopulmonary exercise testing data of systolic and diastolic heart failure

Systolic HFDiastolic HF
Peak RER1.06 ± 0.131.06 ± 0.121.06 ± 0.111.05 ± 0.12
Peak VO2, mL O2 . kg−1 . min−115.5 ± 5.012.1 ± 3.8@17.8 ± 5.5§,#11.9 ± 3.9@,#
VO2 at anaerobic threshold, mL O2 . kg−1 . min−110.6 ± 5.09.2 ± 4.0*14.5 ± 5.5§,#9.6 ± 3.9@
VE/VCO2 slope35.3 ± 10.139.7 ± 9.1@32.8 ± 7.9§,#36.9 ± 6.7@,#
  • *P < 0.05 vs. no-EOB; @P < 0.01 vs. no-EOB; §P < 0.01 vs. EOB systolic HF; #P < 0.01 vs. no-EOB systolic HF. RER, respiratory exchange ratio; VE, ventilation; VCO2, CO2 output.

As to VE/VCO2 slope, there was a main effect for EOB; patients with no-EOB had a significantly lower VE/VCO2 slope compared with EOB patients (P < 0.001). The main effect for systolic vs. diastolic HF was also significant (P = 0.009). Patients with diastolic HF and EOB had a significantly lower VE/VCO2 slope compared with the subjects with systolic HF. There was, however, no interaction effect between EOB and HF type (P = 0.90).

Cox regression analyses

Univariate Cox regression analysis demonstrated that EOB was a significant predictor of cardiac mortality in both populations (Table 5A and B). At multivariable analyses, EOB was retained as prognostic indicator in systolic HF and emerged as the most powerful indicator in diastolic HF (Table 6A and B). The cardiac-related event rate was higher in systolic than diastolic HF. Specifically, 99 (24%) major cardiac-related events were observed among patients with systolic HF (78 deaths, five LVADs, 16 transplants) and 17 (11%) major events were observed for those with diastolic HF (16 deaths, one transplant).

View this table:
Table 5

Univariate Cox regression analysis for systolic and diastolic heart failure

χ2Hazard ratio (95% CI)P-value
Systolic heart failure
 VE/VCO2 slope38.41.05 (1.03–1.07)<0.001
 Peak VO216.30.91 (0.86–0.95)<0.001
 EOB22.82.5 (1.6–3.7)<0.001
Diastolic heart failure
 VE/VCO2 slope11.91.10 (1.04–1.16)0.001
 Peak VO26.10.87 (0.77–0.97)0.01
 EOB14.15.9 (2.1–16.9)<0.001
  • RER, respiratory exchange ratio; VE, ventilation; VCO2, CO2 output.

View this table:
Table 6

Multivariable Cox regression analysis for systolic and diastolic heart failure

χ2 bHazard ratio (95% CI)P-value
Systolic heart failure
 VE/VCO2 slope38.41.05 (1.03–1.07)<0.001
 EOBa13.02.0 (1.3–3.1)<0.001
 Peak VO23.00.08
Diastolic heart failure
 EOB14.15.9 (2.1–16.9)<0.001
 VE/VCO2 slopea6.51.09 (1.03–1.16)0.01
 Peak VO20.230.63
  • RER, respiratory exchange ratio; VE, ventilation; VCO2, CO2 output. aRetained in the regression. bResidual χ2.

When Cox regression analysis was performed looking at cardiac mortality without LVADs and transplants, EOB was prognostic in both diastolic (χ2 = 12.0, hazard ratio: 5.4, 95% CI: 1.9–15.7, P = 0.001) and systolic (χ2 = 21.3, hazard ratio: 3.0, 95% CI: 1.9–4.8, P < 0.001) HF groups.

Looking also at predictive differences according to a peak RER cut-off of 1.05, it emerged that in the diastolic HF group, 44% of patients had RER ≥1.05. EOB was prognostic in this group (χ2 = 12.3, hazard ratio: 9.7, 95% CI: 2.1–45.7, P < 0.001) but not in those with a RER < 1.05 (χ2 = 1.2, hazard ratio: 2.4, 95% CI: 0.48–12.5, P = 0.27). The latter finding may be due to the less number of events in the latter group (only six).

In the systolic HF group, 53% had RER ≥1.05 and EOB was prognostic in both patients with EOB and peak RER≥1.05 (χ2 = 7.6, hazard ratio: 2.1, 95% CI: 1.2–3.7, P = 0.007), and below (χ2 = 11.7, hazard ratio: 2.8, 95% CI: 1.5–5.2, P = 0.001).

Kaplan–Meier analyses

Kaplan–Meier analysis results for the four subgroups are illustrated in Figure 3. Survival rate was significantly lower in those with EOB and either systolic or diastolic dysfunction. Patients with diastolic HF and EOB had a lower survival rate than systolic HF patients without EOB.

Figure 3

Kaplan–Meier analysis for patients with no-exercise oscillatory breathing and exercise oscillatory breathing for systolic vs. diastolic heart failure subgroups.


Results of the present study define, for the first time, the prevalence and the prognostic significance of exercise oscillatory gas exchange kinetics in patients with diastolic HF. EOB, a phenomenon that has been traditionally linked to advanced systolic HF,12,27 is also part of the syndrome of diastolic HF and its recognition may help to identify a subset of diastolic HF patients at increased risk of cardiac mortality.

Notably, as in systolic HF, EOB in diastolic HF showed a prognostic power superior to peak VO2 and the VE/VCO2 slope, parameters that have become standards regarding heart transplantation urgency in advanced HF patients (peak VO2), and risk stratification across HF populations with varying disease severity (VE/VCO2 slope).811

Diastolic heart failure, exercise gas exchange and ventilation

The natural history of diastolic HF is less well characterized than that of systolic HF; thus, precise management guidelines and optimal therapeutic approaches can present a challenge for these patients. In particular, information on diastolic HF and CPET performance are limited. Six studies have investigated exercise pathophysiology by gas exchange analysis,2833 although just one has addressed prognosis.29 In these studies, disparate criteria were used for diastolic HF definition, and there was a wide heterogeneity among populations concerning age and gender distribution. However, although exercise performance and peak VO2 varied widely across these studies, the cardiopulmonary exercise pattern common to all studies but one33 was an increased VE/VCO2 slope and an excessive dyspnoea sensation. The present observations further support this evidence given that multivariable analysis demonstrated that the VE/VCO2 slope outperformed peak VO2 in predicting cardiac events. In addition, these findings underscore the importance of addressing abnormalities in the exercise ventilatory pattern to better characterize symptomatic diastolic HF. Interestingly, patients with diastolic HF and a high VE/VCO2 slope were more commonly observed in the EOB subgroup. Indeed, this population presented with an average VE/VCO2 slope of 37 that, although a little lower than the average value of 40 observed in systolic EOB patients, is strongly predictive of worse prognosis.11

Exercise oscillatory breathing: clinical and prognostic characteristics

Based on our numbers, the prevalence of abnormal ventilatory oscillation in patients with diastolic HF approximates 30%, and is similar to that observed in systolic HF in both the present and previous reports.17,18 Likewise, as in systolic HF, a close link between EOB and clinical severity exists. Aside from the above-reported differences in VE/VCO2 slope, the clinical picture of patients with diastolic HF and EOB overlaps that observed among systolic HF patients with EOB; overall exercise performance and peak VO2 were similar and LV end-diastolic pressure (E/E′) was similarly increased. EOB diastolic HF patients exhibited normal LV chamber size, increased LV mass and a higher LV mass to volume ratio, providing evidence of a remodelling pattern at high risk. The 1-year mortality rate for diastolic HF is reported to be ∼15% in patients ≥65 years.3 The average age of present diastolic HF population was 62 years and this may explain the observed 1-year mortality rate of 5.7%. Almost all cardiac events occurred in the EOB subgroup, with a 1-year event rate of 10%. It is also interesting to note that in diastolic HF, presence of gas exchange oscillatory kinetics during exercise yielded an event-free survival rate worse than that in systolic HF patients without EOB (83 vs. 75%; P < 0.001).

In a recent study by our group, primarily involving systolic HF patients, we found EOB to be a potent predictor of sudden cardiac death.18 In this study, we were unable to define whether this may also be the case in diastolic HF patients. Considering the unquestionable impact of sudden death on the natural history of HF, further investigation in this area would be desirable.

Study limitations

Definition of the mechanisms underlying EOB in diastolic HF was beyond the purposes of the study. Although our findings suggest similar EOB prognostic characteristics in diastolic as systolic HF, the pathogenetic mechanisms of the phenomenon in diastolic HF are unclear. However, it is tempting to speculate that an increased filing pressure at rest would lead to a further increase during exercise, leading to pulmonary interstitial fluid accumulation and stimulation of the J-receptors beyond normal reflexogenic control.34

A potential selection bias is represented by the fact that only patients who were referred to CPET were included. Thus, in a study of prevalence, one may argue that, in most clinical settings, physician may favour referral of patients with systolic dysfunction to CPET and even more, that populations of diastolic HF patients referred to a CPET evaluation may not be entirely representative of the whole spectrum of diastolic HF patients.

Gender distribution was predominantly male for both systolic and diastolic HF, although EOB appeared to be independent of gender. In addition, the average age of our diastolic HF patients was somewhat lower compared with previous reports. This may be unusual because of the well recognized higher diastolic HF rate in female and older populations. However, the lower propensity of primary care providers to exercise older or female patients to exhaustion must be considered. These considerations all together may offer a conceivable explanation on the differences between our sample of diastolic HF study population and the broader unselected samples of diastolic HF followed-up in larger epidemiological studies.

Conclusions and perspectives

The current findings suggest that (i) in diastolic HF, EOB occurrence and prevalence is similar to systolic HF; (ii) EOB recognition implements the ability to identify diastolic HF patients at increased risk; and (iii) confirmatory data and definitive strategies to treat oscillatory breathing patterns are warranted.


This work was supported by a grant provided by the Monzino Foundation, Milano, Italy.

Conflict of interest: none declared.


  • Presented in part at the 80th American Heart Association Scientific Session, Orlando, Florida, November 3–7, 2007.


View Abstract