Correlation of Hemodynamic and Respiratory Parameters in Invasive Cardiopulmonary Exercise Testing (iCPET)

Background: Invasive cardiopulmonary exercise testing (iCPET) is an integral part in the advanced diagnostic workup of pulmonary hypertension (PH). Our study evaluated the relation between hemodynamic and respiratory parameters at two different resting conditions and two defined low exercise levels with a close synchronization of measurements in a broad variety of dyspnea patients. Subjects and methods: We included 146 patients (median age 69 years, range 22 to 85 years, n = 72 female) with dyspnea of uncertain origin. Invasive hemodynamic and gas exchange parameters were measured at rest, 45° upright position, unloaded cycling, 25 and 50 W exercise. All measurements were performed in a single RHC procedure. Results: Oxygen uptake (VO2/body mass) correlated significantly with cardiac index (all p ≤ 0.002) at every resting and exercise level and with every method of cardiac output measurement (thermodilution, method of Fick). Mean pulmonary arterial pressure (PAPmean) correlated with all respiratory parameters (respiratory rate, partial end-tidal pressures of oxygen and carbon dioxide [petCO2 and petO2], ventilation/carbon dioxide resp. oxygen ratio [VE/VCO2, VE/VO2], and minute ventilation [VE], all p < 0.05). These correlations improved with increasing exercise levels from rest via unloaded cycling to 25 W. There was no correlation with right atrial or pulmonary arterial wedge pressure. Summary: In dyspnea patients of different etiologies, the cardiac index is closely linked to VO2 at every level of rest and submaximal exercise. PAPmean is the only pressure that correlates with different respiratory parameters, but this correlation is highly significant and stable at rest, unloaded cycling and at 25 W.


Introduction
The invasive measurement of hemodynamic parameters by right heart catheterisation (RHC) is the decisive diagnostic method in pulmonary hypertension (PH). It is mandatory for therapeutic decisions and allows a stratification of the patients' prognosis. The indication, procedure and interpretation of RHC have been standardized to a large extent, for instance by the European Society of Cardiology in 2017 [1]. Cardiopulmonary exercise testing (CPET) evaluates different aspects of pathophysiology, namely regarding exercise, and is, in contrast to RHC, a non-invasive method. Exercise limitation and response to therapy in pulmonary arterial hypertension (PAH) can be graded by CPET [2,3]. As with RHC, the application and performance of CPET in PH were subject of a standardization process, and expert consensus has been published [4,5]. The combination of both methods is termed "invasive cardiopulmonary exercise testing" (iCPET) and provides a very differentiated analysis of the etiology of dyspnea and the response to exercise. It has been shown to predict prognosis in patients with chronic systolic heart failure [6], heart failure with preserved ejection fraction (HFpEF) [7], and with exertional dyspnea in general [8]. iCPET is an integral part in the advanced diagnostic workup of PH [9].
One must consider that iCPET with repeated measurement of thermodilution is difficult under exercise conditions and requires a trained and well-cooperating staff. This might be one of the reasons why the mode of exercise for iCPET is not well standardized [5], despite the several proposals that have been published in recent years [9][10][11]. Aside from the still necessary standardization, the allocation of RHC parameters to respiratory parameters at exercise was mostly evaluated in selected groups of patients and at various levels of exercise, i.e., at unloaded cycling, anaerobic threshold, 25 W or peak exercise. About 30 years ago, early studies found a significant relation between the respiratory parameters VE/VCO 2 -slope and peak oxygen uptake (peakVO 2 ) and the invasive measured hemodynamic parameters cardiac index (CI), mean pulmonary artery pressure (PAPmean), and pulmonary arterial wedge pressure (PAWP), but all correlations still referred to patients with chronic heart failure [12]. By contrast, a pilot study using a micro-manometer-tipped catheter in patients with PAH at exercise showed that ventilatory equivalents for oxygen (VE/VO 2 ) and carbon dioxide (VE/VCO 2 ) correlated with PAPmean at exercise, but oxygen pulse and oxygen uptake (VO 2 ) did not [13].
Most parameters have been evaluated at rest and at peak exercise only, and the abovementioned technical and logistical demands might have prevented a stepwise analysis. Against this background, our study evaluated the relationship between hemodynamic and respiratory parameters at two different resting conditions and two defined low exercise levels. We defined a protocol that was aimed at the greatest possible synchronization of measurements and applied the method to a broad variety of dyspnea patients.

Patients
This single-centre retrospective study recruited from a total of 306 patients that were referred to the Greifswald University hospital between 2016 and 2020 due to dyspnea of uncertain etiology. All of these patients underwent right heart catheterization during exercise with measurement of respiratory parameters (iCPET). We performed these procedures with predefined and close synchronization of thermodilution, blood gas sampling and measurement of respiratory parameters. In case of multiple testing of patients in long-time follow-up, only the first iCPET was included, so that the final study population consisted of 146 patients. The study was approved by the Ethics Committee of Greifswald University (No. BB 215/20 of 17 November 2020 , which was amended on 28 January 2021. All patients gave their written informed consent prior to the procedure.

Right Heart Catheterization
The indication for exercise RHC with simultaneous measurement of oxygen uptake was the suspicion of a multi-factorial etiology of dyspnea as a result of the preceding noninvasive examinations. The latter included ECG, X-ray of the thorax, echocardiography, CPET, and bodplethysmography. All catheter procedures were performed in hospitalized patients and followed the guidelines of the ESC/ERS [14] and German recommendations [15]. We have published the detailed course of the exercise RHC previously [16]. In brief, this procedure starts with a hemodynamic measurement at rest in a supine position (labelled in the following "baseline") and a second measurement at rest in a semi-supine position ("45°at rest"), followed by unloaded cycling in a semi-supine position for 3 to 5 min ("unloaded cycling"), and 3 to 5 min cycling at 25 W. If both anaerobic threshold (in the literature, synonymously "first ventilatory threshold") was not reached and dyspnea reported was only mild to moderate, a 50 W step was appended. On average, this was 59% of the peak VO 2 that the patient achieved on a cycle ergometer 3 to 5 days before. This pre-test was performed in a manner that was symptom limited, in a sitting position and without RHC. At rest we measured cardiac output (CO), respectively CI, by thermodilution (TD) and the direct and indirect method of Fick (dFM resp. iFM, with the iFM method using the table of LaFarge [17]). At exercise, CO was calculated at every level by two methods in each patient: by TD using the mean of 3 (up to 5) cold water injections and verification of the visualized temperature curve, and simultaneously by the dFM method. The latter was performed after reaching a steady state (mostly after 3-5 min) of VO 2 at the respective exercise level. PH was defined according to the definition at time of study enrolment as PAPmean ≥ 25 mmHg, and pulmonary arterial hypertension (PAH) was defined as PAPmean ≥ 25 mmHg plus PAWP < 15 mmHg. The recent definition with a cutoff at PAPmean >20 mmHg and inclusion of pulmonary vascular resistance (PVR) ≥ 3 Wood units (≥240 dyn·s·cm −5 ) [18] was published in 2019 and therefore not applicable for our study that started in 2016.

Cardiopulmonary Exercise Test and Blood Gas Analysis
VO 2 and VCO 2 were measured with a 10 s average by a cardiopulmonary exercise test system (SentrySuite; Viasys Healthcare GmbH, Höchberg, Germany) via a face mask at room air. Blood samples for the measurement of mixed-venous and arterial oxygen saturation and arterial partial pressures of oxygen (p a O 2 ) and carbondioxide (p a CO 2 ) were taken at the same time as the exhaled gas measurement and within a 30 s frame of the cold water injection for TD measurement. Blood gases were analyzed immediately (ABL 90; Radiometer, Copenhagen, Denmark). A marker was set at the exact time of blood sampling in the electronic file, and CPET data were extracted later from the measurement data file at this very point and 10 s before and 10 s after. In order to avoid outliers by very deep or shallow breathing on the one hand and to be very close to hemodynamic measurement on the other hand, we transferred the −10 s/0/+10 s datasets to a database program and calculated the mean for each respiratory parameter. These data were analyzed statistically as described below. The situation of the assessment is shown in Figure 1. All data were acquired, calculated and saved electronically.

Statistics
Nominal data are given in absolute numbers and percent, and continuous data are given as median (25th; 75th percentile). Pearson and Spearman correlation coefficients (r) were computed to assess the relationships between gas exchange and hemodynamic measures at rest and at different levels of exercise. A value of p < 0.05 was considered statistically significant. The analyses were performed with the SAS 9.4 program (SAS Institute Inc., Cary, NC, USA).

Oxygen Uptake and Hemodynamic Parameters
Oxygen uptake VO 2 /body weight (in mL/min/kg) correlated significantly with cardiac index (all p ≤ 0.002). This correlation was very robust because it was reproducible at every level, i.e., at rest (supine and 45°), at unloaded cycling and at 25 W. The correlation between CI and VO 2 /body weight was proven for every method of hemodynamic measurement, with somewhat lower correlation coefficients for CI by TD (r ≤ 0.37; p < 0.001) and the best correlation for the dFM (r ≤ 0.54; p < 0.001). The absolute, i.e., not body weight-related, values of VO 2 (in mL/min) correlated with cardiac output and with lower correlation coefficients also with CI at rest, 45°and unloaded cycling (Figure 2). A special pattern was proven at 25 W with correlations only between the body-weight-related (VO 2 /body weight and CI) and body-weight-unrelated (VO 2 and CI) parameters. All correlations lost significance at 50 W, in parallel with the lower number of patients. Figure 2. Cardiac output, oxygen uptake and correlation coefficients. Oxygen uptake in milliliters per minute (VO 2 ) and per kg body weight (VO 2 /kg). Thick frames indicate significance level p < 0.05.

Respiratory Parameters and PAPmean, PAWP and PVR)
The PAPmean correlated with all respiratory parameters (respiratory rate (Rr), p et CO 2 , p et O 2 , VE/VCO 2 , VE/VO 2 , and VE). Considering the exercise levels, these correlations improved with increasing exercise levels from rest via unloaded cycling to 25 W. Aside from Rr, the correlations with PAPmean were consistent at both resting positions (supine and 45°). The pressures RAPmean and PAWP did not correlate with any of these parameters, except at 25 W (Figure 3). PVR correlated with none of these respiratory parameters except respiratory rate (at unloaded cycling with r = 0.18; at 25 W with r = 0.27; both p ≤ 0.05) and inversely with the breathing reserve in percent (%BR, from r = −0.40 at rest to −0.45 at 25 W, all p < 0.001). As above, there were no correlations at 50 W. Figure 3. Right heart pressures, respiratory values and correlations. PAWP, pulmonary arterial wedge pressure; RAPmean mean, right arterial pressure; PAPmean, mean pulmonary arterial pressure; Rr, respiratory rate; p et CO 2 , end-tidal partial pressure of carbon dioxide; p et O 2 , end-tidal partial pressure of oxygen; VE/VCO 2 , ventilation-to-carbon-dioxide output ratio; VE/VO 2 , ventilation-to-oxygenuptake ratio; VE, ventilation per minute; V t , tidal volume. Thick frames indicate significance level p < 0.05.

Discussion
Our study proved significant correlations (a) between VO 2 /body weight and CI and (b) between all respiratory parameters and PAPmean. These correlations remained significant at supine rest, 45°rest, unloaded cycling and 25 W exercise. To our knowledge, such a constant relationship through different levels of rest and exercise has not been shown before. Aside from the pathopysiological causality, it was necessary to apply our method with a maximum synchronization of CO, ventilatory parameters and blood gas measurements to unravel these correlations. The literature on iCPET covers a large spectre of assessed parameters but left relevant gaps. In a recent review [3], we identified 53 studies, of which only 7 studies included more than 100 patients. None of these studies applied both methods of CO measurement (TD and dFM), and the included exercise variables were limited. Comparably to [3], the method of iCPET has been summarised by the Boston working group in 2013 [21] and recommendations on measuring CO during exercise have been updated in recent years [10]. Following these recommendations, iCPET has a central role in the advanced work-up of dyspnea and the according diagnostic algorithms [11]. However, these recommendations do not refer to correlations between gas exchange and exercise hemodynamics.

iCPET and Correlations
Correlations have been shown for defined patient groups and mostly at peak exercise level. In a very early study on 23 COPD patients, Schrijen and co-workers found a slight correlation between peakVO 2 and hemodynamics (namely PVR) at rest and during light exercise [22]. In patients with chronic heart failure, the working group around Marco Metra published correlations in 1992: they showed that in 33 patients, peakVO 2 and VE/VCO 2 -slope correlated with CI and PVR but not with PAPmean [12]. Some years later and based on a larger number of patients (n = 219) this group could also give an equation to calculate CO from peakVO 2 as follows: CO = 0.00625 · VO 2 + 1.51 [6]. The application of iCPET was extended to patients with connective tissue diseases [13], systemic sclerosis [23], and hypertrophic cardiomyopathy [24]. The latter study showed interesting correlations between PAPmean and the VE/VCO 2 -ratio at peak exercise and p et CO 2 at anaerobic threshold, but there was no correlation to CO, and with a time delay of at mean 3 days between RHC and CPET, this is not strictly comparable to an iCPET protocol.
Another not literally iCPET study combined consecutively performed CPET and RHC in patients with systemic sclerosis and found that peakVO 2 , nadir VE/VCO 2 , and peak p et CO 2 correlate with PAPmean, TPG and PVR at rest [23]. The potential value of iCPET for the discrimination of combined pre-and postcapillary PH and isolated post-capillary PH was shown in patients with chronic heart failure (CHF). In this group, PAPmean correlated again with VE/VCO 2 , TPG, and also with the dead space/tidal volume ratio (V d /V t ) [25]. Recently, two great studies from the Boston working group proved the prognostic value of iCPET. Based on the physiological coupling of PAPmean to the corresponding increase in CO, they assessed 714 patients with exertional dyspnea and proved that the PAPmean/COslope was predictive for survival. A cut-off of PAPmean/CO-slope > 3 mmHg/L/min separated all respiratory parameters. Correlations with hemodynamic parameters were not the aim of this study and were not published [8]. The second study with 493 patients and exertional dyspnea was restricted to HFpEF and revealed, comparably to the systemic sclerosis study mentioned above, that VE/VCO 2 at its nadir is a decisive parameter for prognosis and associated with CO, PVR and PAPmean/CO-slope [7].

Physiological Considerations
Studies that assessed iCPET at a low level exercise are rare. In an older small study (n = 10) on pulmonary hypertension, PAPmean correlated with VE/VCO 2 at 30 W but not with VO 2 [13]. The Boston/Yale working group showed that in PH patients, the RV function deteriorates already at 50% of peak oxygen uptake. This is in line with our finding that PAPmean correlates with many respiratory parameters at unloaded cycling and even better at 25 W. However, a direct comparison to our study is prevented by their use of a steeper exercise protocol and elaborated computing algorithms [26]. Under physiological conditions, PVR must decrease at exercise to cushion the up to fivefold increase in right ventricular output [11]. This may explain the prognostic cut-off for the PAPmean/CO-slope > 3 mmHg/L/min [8]. A late rise in PAPmean beyond the anaerobic threshold ("takeoff-type") is associated with higher CO and peakVO 2 despite no differences in PAPmean at peak exercise [27].
Inequalities between ventilation and perfusion make gas exchange less efficient than normal and increase the physiological dead space and the V d /V t ratio, respectively. This leads to a lower p et CO 2 at rest and exercise, as shown in patients with PAH [28] and CHF [29]. Lactate acidosis at already low work rates is a second factor that might lead to an increased ventilatory drive and consecutively increased dead space ventilation and lowered p et CO 2 . Hence the decrease of p et CO 2 was more marked in patients with more severe heart failure [29]. This increased V d /V t ratio might explain the correlation of lower respiratory reserve and higher PVR in our study, as well as the correlation of respiratory rate and PVR. By contrast, a study on COPD and PH (Nice class 3) found that a higher PAPmean was associated with a higher breathing reserve [30]. This illustrates, in our opinion, the shift from a strict pulmonary to a pulmonary-circulatory limitation of exercise.
In an attempt to find an alternative test for subjects who are unable to exercise, a very recent study compared the pulmonary vascular distensibility measured by iCPET with the distensibility after passive leg raise [31]. This study found a correlation of both methods for distensibility, but we still consider iCPET as indispensable because this correlation becomes blurred if the healthy control subjects in this study are omitted. This is in line with our own findings [16] that exercise and fluid challenge activate different pathophysiological pathways and unmask diastolic dysfunction in different patient groups.

Strengths of the Study
This is the first greater study that assessed iCPET at different and lower levels of exercise. It studies a typical cohort of patients with unexplained dyspnea and typical comorbidities. It provides robust correlations between hemodynamic and respiratory parameters at different exercise levels. Moreover, the study minimized the time delay between hemodynamic and respiratory measurement. Further shortening of the respiratory interval would make the measurement susceptible for outliers in the form of shallow or sigh breathing.

Limitations
We did not calculate the new and prognostic parameter PAPmean/CO-slope. The study took more than 4 years, and a new definition of PH was presented in the middle of this time. A further evaluation of the prognostic value of iCPET at low-level exercise will require the application of this defintion.

Conclusions
Our study shows that in dyspnea patients of different etiologies, the cardiac index is closely linked to VO 2 at every level of rest and submaximal exercise. PAPmean is the only pressure that correlates with different respiratory parameters, but this correlation is highly significant and stable at rest, unloaded cycling and at 25 W.