Ischemic heart disease is the most prevalent risk factor for left ventricular dysfunction (LVD) and chronic heart failure (HF), both for women and men (1). Impaired pulmonary gas exchange, quantified as an increased ventilation (VE) to carbon dioxide exhalation (VCO₂) slope during exercise, and low end-tidal pressure of carbon dioxide (P_ETCO₂) are landmarks of heart failure patients (2, 3). These parameters received attention when several studies found higher VE/VCO₂ slopes and lower P_ETCO₂ to be associated with poorer prognosis (4–6). The components of the VE/VCO₂ slope are the arterial CO₂ partial pressure (P_aCO₂), which in CHF is affected mainly by ventilation, and the pulmonary dead space to tidal volume ration (VD/VT) mainly by pulmonary perfusion abnormalities (3, 7). Exercise hyperventilation is a hallmark of a failing heart (2). Several mechanisms have been proposed as mediators of this excessive ventilatory response, including: (1) alveolar ventilation-perfusion mismatching, (2) increased humoral stimuli (e.g., lactate and H⁺) due to skeletal muscle hypoperfusion and deconditioning, (3) juxta-capillary receptor stimulation consequent to pulmonary vascular congestion and/or hypertension, (4) augmented central and peripheral chemosensitivity, and (5) an inordinately high degree of afferent neural traffic originating from within the locomotor muscles (i.e., the ergoreflex or “skeletal muscle” hypothesis). Hyperventilation is well known to stimulate sympathetic nervous activity (8). Chronic sympathetic nervous hyperactivity, in turn, may decrease the aerobic capacity of skeletal muscles by reducing capillarisation (9) and red blood cell flux (10), which lead to a shift in muscle fibre type towards a lower content on type I fibres (11). The ensuing anaerobic muscle metabolism leads to increased muscle fatiguability (12) and acidosis already at low levels of exercise, which trigger exaggerated responses in ventilation (13). While exaggerated ventilation has been found to be a strong predictor of mortality and adverse cardiac events in patients with established heart failure, this has not been assessed in patients with LVD or with ischemic heart disease (IHD). There is a continuum from patients with IHD to LVD and HF (14), with asymptomatic LVD being twice as common as HF (15). However, inefficient ventilation has not been assessed as a prognostic marker in patients with IHD or LVD.

Comprehensive exercise-based cardiac rehabilitation (exCR) has the potential to improve cardiorespiratory fitness (CRF) and quality of life (16), cardiovascular risk factors (17), and reduce hospitalisations and cardiovascular mortality (17, 18). In HF patients with reduced ejection fraction, studies assessing the effects of exercise have focussed largely on circulatory parameters (19), peak oxygen uptake (VO₂) and muscle strength (20). Some studies have also reported changes in ventilatory efficiency in response to exercise during exCR (21, 22). However, data in elderly patients with LVD are sparse and no data exists on how exCR influences breathing patterns at rest.

The purpose of the present study in elderly patients with and without ischemic LVD was to (1) describe breathing patterns and pulmonary gas exchange parameters at rest and during exercise; (2) assess the association of exCR with a change of these parameters; and (3) assess the prognostic value of these parameters on major adverse cardiac events (MACE). We hypothesised that ventilation would be less efficient in patients with compared to those without LVD, and that, similarly to patients with HF, exCR would improve ventilatory efficiency, and that ventilatory inefficiency would be associated with MACE.

The EU-CaRE observational longitudinal study was a European project focusing on the effectiveness and sustainability of exCR programs in the elderly (65 years or above). EU-CaRE was conducted from 2016 to 2019 and involved eight participating CR sites in seven countries, with five offering ambulatory CR of 6–12 weeks duration (Denmark, Italy, the Netherlands, Spain, and Switzerland) and two offering stationary CR of 3–4 weeks duration (France, Germany) (23).

Of 1,633 enrolled EU-CaRE patients, 986 patients had an ACS or CCS with percutaneous coronary intervention or no revascularisation. Of these, 79.8% had an ACS and 90.2% a PCI as an indication for exCR. LVEF was known in 912 patients. Of these, valid CPET at T0 was available in 867 patients of whom 707 had no LVD and 166 had LVD. Amongst these, data on breathing patterns were available from at least one time point in 151 with and 667 patients without LVD (Figure 1). The baseline data of the two patient groups are shown in Table 1. Patients with LVD were older, were less often women, and had lower BMI and FVC relative to body surface area. An equal prevalence of 86% of patients with New York Heart Association (NYHA) functional class I existed in patients with and without LVD, and NYHA III class was present in 2% and 3% of patients with and without LVD, respectively. Patients with LVD had significantly lower physical component scores of the SF36 indicating poorer health status. Patients with LVD had a higher percentage of pre-existing chronic heart failure and anaemia, and a higher prescription rate of beta-blockers and ACE inhibitors or ARBs.

With regard to ventilatory parameters, two centres did not record data on BF and VT, therefore, these parameters were missing in 5 and 51 patients with and without LVD, respectively. Due to an insufficient ventilation monitoring duration (≤1 min) during the 3-min resting phase, VE and P_ETCO₂ at rest were missing in 9 and 68 patients with and without LV dysfunction at T0, respectively, in 6 and 82 at T1, respectively, and in 4 and 46 at T2, respectively. The first ventilatory threshold could not be determined in 38 and 130 patients with and without LV dysfunction, respectively, at T0, in 20 and 71 at T1, respectively, and in 16 and 61 at T2, respectively. The largest relative adjusted (for sex, age, and BMI) differences of the LVD group compared to the noLVD group at baseline were found for VO₂ peak (−12.6%), VE/VCO₂ slope (11.0%), nadir VE/VCO₂ (8.8%), and resting VE (8.1%) (Table 2).

Regarding the changes in ventilatory parameters from before to after CR, patients with LVD had significantly greater reductions in resting VE (Table 2). They also had greater reductions in nadir VE/VCO₂, and a greater increase in P_ETCO₂ and RER at peak exercise.

Results from the mixed linear models including data also for the 1-year follow-up measurement as well as confounding factors age, sex, and BMI showed a significantly greater reduction of VE at rest in the group of patients with LVD not only from before to after exCR but also to 1-year follow-up (Figure 2, Supplementary Table S1). BF at rest was higher in patients with LVD at all time points (Figure 2, Supplementary Table S1), and resting P_ETCO₂ increased significantly at 1-year follow-up in both groups (Figure 3, Supplementary Table S1). At peak exercise, VE was reduced in the patients with LVD due to a significantly lower VT in the patients with LVD but increased with exCR and to 1-year follow-up in both groups (Figure 2, Supplementary Table S2). Resting P_ETCO₂ was lower by trend (p = 0.0537) in patients with LVD and significantly lower at peak exercise at all time points (Figure 3, Supplementary Tables S1 and S2). Accordingly, VE/VCO₂ slope was increased in patients with LVD at all time points (Figure 3, Supplementary Table S3). VO₂ relative to body weight was greatly reduced in patients with LVD at peak exercise (Figure 3, Supplementary Table S3) at all time points. Peak VO₂ increased in both groups from before to after exCR and VE/VCO₂ slope decreased over time. The physical component score was consistently lower by 3 points in patients with LVD but significant consistent improvements in SF36 PCS between time points were comparable in both groups with nearly 3 points between T0 and T2 (Supplementary Table S3).

Approximately 20% of patients with and without LVD had a MACE within 1-year follow-up with no difference between groups (p-value for Chi-square test 0.069). Nadir of VE/VCO₂ significantly discriminated patients with and without MACE in patients with and without LVD, while VE/VCO₂ slope and peak VO₂ discriminated for MACE only in patients without LVD, and peak P_ETCO₂ and resting BF discriminated for MACE only in patients with LVD (Table 3).

This study provides four clinically relevant and novel aspects in a representative European cohort of elderly patients with coronary artery disease undergoing exCR: (1) Nearly one-fifth of patients had asymptomatic LVD at the start of exCR; (2) LVD was associated with elevated ventilation at rest and impaired pulmonary gas exchange during exercise; (3) exCR was associated with an improvement of ventilation at rest and gas exchange parameters during exercise, but ventilatory efficiency and P_ETCO2 remained significantly reduced in patients with LVD compared to patients without LVD in the 1-year follow-up; and (4) BF at rest, ventilatory efficiency and P_ETCO₂ during exercise were associated with worse outcome in patients with LVD.

This is the first large study demonstrating increased resting ventilation and impaired pulmonary gas exchange during exercise in patients with LVD without major signs and symptoms of congestive heart failure. NYHA functional class did not differ between patients with and without LVD and 86% of patients in both groups were in NYHA functional class I. Furthermore, self-reported physical activity did not differ between the groups. Although the median physical component score of the SF36 questionnaire was 3 points lower in patients with LVD compared to patients without LVD, the values of 44.3 at T0 and 47.5 at T2 were clearly above the median of 35.3 found in patients with chronic heart failure (33).

We suggest that signs of an exaggerated ventilatory drive are already apparent in patients with LVD at rest (higher resting VE and BF), during exercise (increased VE/VCO₂ slope and nadir VE/VCO₂), and at peak exercise (lower VO₂, VT, and P_ETCO₂). Reductions of the exaggerated ventilatory drive seen with exCR were similarly beneficial in patients with and without LVD. The decrease in nadir VE/VCO₂ was significantly greater in patients with LVD than in those without LVD, with a similar magnitude of the difference between changes in VE/VCO₂ slope, albeit not reaching statistical significance. This is in accordance with some previous studies that found ventilatory efficiency to improve most in patients with the most severe ventilatory inefficiency at baseline. For example, in a study with 131 patients after acute myocardial infarction without HF participating in exCR, VE/VCO₂ slope improved most in patients with VE/VCO₂ slope > 32 (34), which was also confirmed by a study comparing the effect of aerobic training to a combination of aerobic and strength training (35). Similarly, a study in patients with HFrEF by Servantes et al. found large decreases in VE/VCO₂ slopes with mean pre-exercise values of 36 (36). In part, a larger decrease in patients with higher starting values may be due to a phenomenon referred to as “regression to the mean” (37). In a study including 123 CAD patients, patients in the group with peak VO₂< 17.5 ml/kg/min also had the highest nadir VE/VCO₂ (38). This group also had the highest improvement in peak VO₂ and nadir VE/VCO₂ after exCR. Likewise, in 37 CAD patients completing exCR, P_ETCO₂ at the first and second ventilatory threshold and peak exercise were improved after the programme. On the contrary, Fu et al. did not demonstrate an improvement of the VE/VCO₂ slope with exercise training in patients with HFrEF and a VE/VCO₂ slope of 35 at baseline (21).

At peak exercise, peak ventilation was 3 L/min (6%) lower in patients with compared to patients without LVD, while peak VO₂ was reduced by 2.2 ml/kg/min (12%). The disproportionate reduction in gas exchange was likely to be due to the significantly smaller peak VT (by 150 ml or 8%) in patients with LVD at an insignificantly increased peak BF, resulting in a larger anatomical dead space. This is in line with observation in patients with HFrEF (39), and the emerging evidence that HF patients adopt a “rapid shallow breathing pattern” to avoid the adverse effects of large intrathoracic pressure swings on cardiac pre- and afterload (2). Importantly, our study extends the evidence on abnormal breathing patterns and hyperventilation in patients with HFrEF to patients with LVD. The significantly higher resting HR in patients with LVD compared to patients without LVD is compatible with sympathetic hyperactivity (2, 40). This finding suggests that several proposed mechanisms for exaggerated ventilation in HFrEF (e.g., augmented peripheral and central chemosensitivity and an inordinately high degree of afferent neural traffic originating from within the locomotor muscles) may already be present in patients with asymptomatic LVD starting the vicious cycle that may progress LVD to HFrEF (2). Interestingly, the models corrected for age, sex, and BMI showed a significant positive relationship of BMI with P_ETCO₂ at rest and at AT, and a significant inverse relationship with nadir VE/VCO₂. This raises the question of whether the obesity paradox not only applies to patients with HF but already to elderly patients with (and without) LVD (41).

The observed normalisation of the excessive ventilation of patients with LVD after exCR may partly be ascribed to the beneficial effects of exercise training on ergoreflex sensitivity, and/or chemoreceptor activation (40). Importantly, guideline-directed medical and device therapies for HFrEF improve chemo- and baroreceptor function insufficiently, highlighting the importance of exercise in this population (42).

While the VE/VCO₂ slope is an established prognostic parameter in patients with HFrEF, only a few studies have reported the physiologically inversely related parameter P_ETCO₂ at rest and at peak exercise (43, 44). Matsumoto and colleagues have found the severity of HF in 112 patients with cardiac disease to be negatively associated with P_ETCO₂ at rest and at peak exercise (43). They found P_ETCO₂ at peak exercise to be negatively associated with VE/VCO₂ slope and VD/VT, but not with P_aCO₂. Arena and colleagues found a resting P_ETCO₂ of <33 mmHg to discriminate significantly for cardiac events at 1-year follow-up and to significantly add discriminative power to VE/VCO₂ slope (44). Similarly, Schäper and colleagues found resting P_ETCO₂≤ 31 mmHg to be associated with mortality (45). Our study demonstrated consistent findings on the prognostic value of pulmonary gas exchange parameters in patients with LVD. The nadir VE/VCO₂ and peak P_ETCO₂ were associated with MACE. Moreover, as a novel finding, BF at rest was associated with MACE. The discriminative value of resting ventilation parameters for MACE deserves further investigation as both resting P_ETCO₂ and BF have been found to be predictive for cardiovascular complications also in lung resection patients (46). Interestingly, NYHA class was not associated with MACE.

The important role of low P_ETCO₂ in patients with HFrEF and periodic breathing has been pointed out by a study by Apostolo and colleagues who restored a normal breathing pattern with inhalation of 2% CO₂ in 95% of their study patients (47). Likewise, inhalation of CO₂ has been shown to stabilise breathing in patients with periodic breathing and Cheyne-Stokes respiration during sleep (48–50). Breathing instabilities are recognised markers for poor prognosis in patients with HFrEF and highlight the central role of breathing patterns in the progression of HF (51). Nevertheless, few therapies have been developed to address breathing patterns. A few small studies with encouraging results have employed slow breathing training in patients with heart failure (52, 53).

Strengths of this study are the large sample size of a representative European cohort of elderly patients with coronary artery disease undergoing exCR, and the longitudinal as well as comprehensive assessment of ventilatory parameters at rest, during ramp exercise, and at peak exercise. Furthermore, this is the first documentation of improved breathing patterns and pulmonary gas exchange both at rest and with exercise after completion of exCR in both patients with and without LVD.

One limitation of this study are its observational design which does not allow the identification of cause-effect relationships. This explorative study is also limited by multiple testing. Therefore, confidence intervals and p-values have to be interpreted accordingly. Unfortunately, we did not have data on beta-blocker classes. Previous studies showed that carvedilol, an unselective ß1- and ß2-blocker, was associated with a lower VE/VCO₂ slope and nadir VE/VCO₂ (54, 55). However, the standard treatment of ACS and CCS includes cardioselective beta-blockers like metoprolol. We can therefore only speculate that only a small proportion of the 93% of our patients have received an unselective beta-blocker. Further, ACE inhibitors have been shown to improve gas transfer and ventilation perfusion matching (56). Nevertheless, the higher percentage of ACE in patients with LVD could not completely offset their less efficient ventilation.

Logistic regressions for the identification of CPET variables significantly improving MACE prediction were performed for the whole EU-CaRE cohort (including patients with open heart surgery and valve replacement) recently (27). Due to the small sample size of our subgroup of patients with LVD, we did not perform logistic regression models in the present study.

Patients with LVD had an exaggerated BF at rest and impaired pulmonary gas exchange during exercise. Abnormal breathing patterns may be an early and clinically relevant sign of LVD and linked to increased ergoreflex sensitivity and/or abnormal chemosensitivity. ExCR may contribute to improvements in breathing patterns and pulmonary gas exchange in this population. However, interventions aimed at specifically improving altered breathing patterns at rest and during exercise may have additive value and should be investigated.