This is a retrospective study analyzing data obtained in patients with precapillary PH due to pulmonary vascular disease (PAH/CTEPH) during clinically indicated RHC at the University Hospital Zurich, 490 m altitude. This study was conducted in concordance with the amended Helsinki declaration, approved by the cantonal ethics committee Zurich (No. 2020-01541) and all patients provided written informed consent.

We included men and women, aged ≥18 years with mean PAP (mPAP) > 20 mmHg, pulmonary artery wedge pressure (PAWP) ≤ 15 mmHg) and revealing a PVR > 2.0 WU (20–22) referred to our clinic between March 2016 and June 2020 for RHC and who had undergone measurements of hemodynamic parameters while breathing ambient air (normoxia FiO₂ 0.21), hypoxia (FiO₂ 0.15), and hyperoxia (FiO₂ 1.0) and who were classified as PAH or CTEPH according to a thorough investigation, including pulmonary function tests, CT-pulmonary angiography, ventilation-perfusion scan, serology, and other measures as required. We excluded severely hypoxemic patients (PaO₂ <7.3 kPa on ambient air) who could not undergo hypoxia testing (21), patients after lung transplantation, and patients with other PH-classifications (2).

Supine resting RHC was performed via jugular venous access with a balloon-tipped, triple-lumen, fluid-filled 7.5 French Swan-Ganz catheter (Baxter/Edwards, Deerfield, IL, USA) as described (6, 23). Transducers were placed at the mid-thoracic level and zeroed to atmospheric pressure (24, 25). Systolic, mean, and diastolic PAP, right atrial pressure (RAP), and PAWP were averaged over several respiratory cycles (Dräger, Liebefeld, Switzerland) (23, 25, 26). CO was measured by thermodilution (Baxter/Edwards, Deerfield, IL, USA) and indexed for body surface area (BSA) to obtain cardiac index (CI). Blood samples were drawn from radial artery puncture and pulmonary artery catheter tips were analyzed directly (RapidPoint 500, Siemens, Zurich, Switzerland) to obtain arterial partial pressures of oxygen and carbon dioxide (PaO₂, PaCO₂), arterial saturation of oxygen (SaO₂), mixed venous partial pressure (PmvO₂) and saturation of oxygen (SmvO₂), hemoglobin (Hb), and lactate.

Patients were asked to rest supine for at least 10 min and baseline hemodynamic parameters and blood gas samples were obtained under normoxia (FiO₂ 0.21). Patients then received hypoxia (FiO₂ 0.15) over an airtight mouthpiece (AltiTrainer, SMTEC, Nyon, Switzerland) for at least 10 min, after which hemodynamics and blood gas samples were obtained a second time while still breathing hypoxia. Following a washout phase of at least 10 min, patients received pure oxygen (hyperoxia, FiO₂ 1) for 10 min administered via a non-rebreathing valve from a reservoir bag (AmbuSPUR-II, Synmedic, Zurich, Switzerland), after which hemodynamics and blood samples were obtained for the third time while still breathing hyperoxia and shunt fraction was calculated with the following formula (11):

The standard barometric atmospheric pressure at 490 m (P_B) was assumed to be 717.7 mmHg, the partial pressure of water at body temperature (P_H2O) assumed to be 47.1 mmHg at 37°C, the respiratory quotient (R) was assumed to be 0.8, and pulmonary capillary oxygen saturation (ScapO₂) was assumed to be 100% under FiO₂ 1.0. The factor 0.0031 is the solubility coefficient of oxygen in the blood and the factor 1.34 is the ratio of oxygen in milliliters per gram of Hb.

The main outcomes consisted of the acute change of mPAP and PVR under hypoxia (FiO₂ 0.15) and hyperoxia (FiO₂ 1.0). Further outcomes included the size of Qs/Qt and the influence of Qs/Qt, PaO₂, and PmvO₂ on hemodynamic parameters (mPAP, PVR, CI, and PAWP) measured under normoxia, hypoxia, and hyperoxia, and subgroup-analysis separately for PAH and CTEPH.

The statistical analysis included all available measurements under normoxia, hypoxia, and hyperoxia. Single missing values were not replaced. All statistical analyses were performed using R Studio (version 1.2.1578, R Studio Inc., San Francisco, CA, USA). Analyses of hemodynamic measures (mPAP, PVR, PVR Index, CI, and PAWP) were performed in univariable (FiO₂ = normoxia vs. hypoxia vs. hyperoxia) and multivariable mixed linear regression models, with Qs/Qt, PH group (PAH or CTEPH), age, sex, carbon monoxide transfer coefficient (K_CO), PaO₂, and PmvO₂ as independent variables, and grouped by subject. Coefficients are expressed as mean change and 95% CI. We tested regression model fitting with QQ-plots and fitted residual and random intercept plots. Multi-collinearity was excluded with variance inflation factor testing. Continuous data are reported as mean ± SD and categorical data as number (percentage). A two-sided p-value of < 0.05 or a 95% CI not crossing 0 was considered statistically significant.

We included 149 patients with PH, 79 PAH, and 70 CTEPH (Figure 1). Patients with PAH were predominantly women (i.e., 70% women), whereas sex distribution was balanced in patients with CTEPH (47% women) (Table 1; Supplementary Table 1). The vast majority of patients were incident without any therapy (90% PAH and 87% CTEPH), a minority of patients with CTEPH had RHC 6–12 months after pulmonary endarterectomy and some patients with PAH were medically pretreated. Pulmonary function tests from the patient records revealed mildly decreased forced expiration in the first second of expiration (FEV₁) and forced vital capacity (FVC), as well as mildly decreased diffusing capacity of lung carbon monoxide (DLCO) and K_CO values.

Under normoxia, patients with PH showed an increased mPAP and PVR and mild hypoxemia according to inclusion criteria (Table 2 and Figure 2). Under hypoxia, a univariable mixed linear regression analysis adjusted for FiO₂ showed that the mPAP did not change significantly compared with normoxia, while the PVR increased by 0.35 WU and the PAWP decreased by 0.5 mmHg. Arterial and mixed venous oxygen and carbon dioxide decreased.

During acute exposure to hyperoxia, univariable mixed models showed a significant reduction in mPAP of 4.4 mmHg and PVR decreased by 0.43 WU. PAWP (−0.5 mmHg) and CI (−0.15 l/min/m²) significantly decreased compared with normoxia. PaO₂ and PmvO₂ rose (Tables 2, 3). These changes in hemodynamic parameters persisted after adjustment for age, sex, PH-classification, K_CO, and Qs/Qt in multivariable mixed models (Table 3). The magnitude of these hemodynamic changes remained largely similar if the two large subgroups, PAH and CTEPH, were analyzed separately (Supplementary Tables 1–4).

Univariable regression models adjusted for arterial and mixed venous oxygen partial pressures (independent of administered FiO₂), revealed that mPAP and PVR were negatively associated with PaO₂ and PmvO₂ (Table 4), also when PAH and CTEPH were analyzed separately. Whereas, CI was negatively associated with PaO₂ but not PmvO₂, PAWP was independent of both (Table 4). Combined in a multivariable mixed linear regression analysis adjusted for age, sex, PH-class, and Qs/Qt, neither mPAP and PVR were significantly inversely associated with PaO₂, while both were independently associated with PmvO₂ (Table 4).

Fraction of right-to-left shunt (Qs/Qt) calculated under hyperoxia was 14.1 ± 5.7%, with 138/139 (99%) subjects showing Qs/Qt levels above 5% and 105/139 (75%) above 10%. There was no significant difference in Qs/Qt between patients with PAH (14.5 ± 5.5%) and CTEPH (14.2 ± 6.7%). The Qs/Qt values could not be calculated in 10 subjects due to missing blood gas values (Table 2). Both mPAP and PVR showed a significant negative association with K_CO, but were unaffected by age, sex, PH class, and Qs/Qt, while mPAP and PAWP were independent of all adjusting variables, including K_CO. CI showed a positive association with Qs/Qt, a negative association with age and was decreased in patients with CTEPH compared with PAH, but independent of sex (Table 4).

This hitherto largest study investigating pulmonary hemodynamics by RHC in patients with pulmonary vascular diseases showed that changes in FiO₂ had distinct effects on resting hemodynamics. Under hypoxia, we confirmed our previous and potentially unforeseen finding that in this large collective of patients with PH, hypoxia had no effect on mPAP (6). We did observe an increase in PVR, but as CI did not change under hypoxia, the increase in PVR under hypoxia was driven by a minor decrease in PAWP, so that the total pulmonary resistance remained unchanged. However, as PAWP was independent of both PaO₂ and PmvO₂, it seems that the change of PAWP was unrelated to the changes in oxygen partial pressures. Moreover, we observed a slight, but significant decrease of PAWP during hyperoxia, further indicating that the changes in PAWP are likely unrelated to FiO₂ or oxygen partial pressures. We hypothesize that breathing through a tube connected to the airtight mouthpiece might have slightly modified intrathoracic pressures which could have contributed to the slightly lowered venous return and thus the PAWP during both hypoxia and hyperoxia exposure. Thus, the present results indicate that patients with PH with a chronically diseased pulmonary vasculature, acute HPV as found in healthy subjects may be blunted (29–32). Whether this is due to the altered mitochondrial redox signaling or altered response of potassium or calcium channels of diseased pulmonary artery smooth muscle cells in PH remains to be determined (33). It equally remains to be seen, whether long-term exposure to hypoxia, as during mountain travel, would worsen the pulmonary hemodynamics in patients with PH.

On the other hand, when patients with PH breathed pure oxygen (hyperoxia), we observed a large decrease in mPAP and PVR. The decreased PVR is especially remarkable as also the CI and the PAWP decreased during oxygen breathing, which would both, in turn, increase the PVR. Both the decrease in CI and PAWP is consistent with previous reports (7). A possible explanation for the largely improved pulmonary hemodynamics with oxygen might be the high FiO₂ used, both in the present and in previous studies (7). However, a previous study by Leuchte et al. found that oxygen given at a flow rate of 5 l/min to achieve a SpO₂ > 90% improved mPAP and PVR in patients with PH, suggesting that lower flow rates could already achieve clinically important improvements (8). It is further remarkable, that pure oxygen breathing in this acute setting gave for 10' comparably improved hemodynamics in PAH and CTEPH. Whether oxygen therapy would reveal a beneficial effect on pulmonary hemodynamics and especially functional parameters in long-term studies and which level of FiO₂ would be sufficient remains to be determined in future well-designed randomized-controlled trials.

The mechanism of action of oxygen therapy in patients with pulmonary vascular disease is not exactly known. A recently published review suggested that acute oxygen induced pulmonary vasodilation is predominantly mediated by alveolar mechanisms and may not even require correction of hypoxemia, although data on mixed venous oxygen content levels were not reported (7). To further explore this mechanism, we calculated linear regression models focusing on the effects of PaO₂ and PmvO₂ on pulmonary hemodynamics. We found that mPAP and PVR were inversely and linearly dependent on PaO₂ and PmvO₂, with the effect of PmvO₂ on both mPAP and PVR being greater in magnitude than PaO₂. While there is no evidence that HPV is both dependent on PaO₂ and PmvO₂, with PaO₂ as directly related to the alveolar PO₂ being the more important factor (34), these models suggest that the mixed venous hypoxemia plays a role in pulmonary vasoconstriction. Patients with PH reveal mostly normal lung functions so that alveolar hypoxia inducing HPV should not majorly contribute to hypoxemia, which is usually attributed to a low mixed venous oxygenation induced by a low cardiac output (35). In precapillary PH, the culprit vascular lesions that contribute to the increased pressure and resistance are mainly located upstream of the pulmonary capillaries and alveoli before the most gas exchange takes place, although capillary and pulmonary venular remodeling is also found (36, 37). Whether changes in pulmonary precapillary vasculature modify mixed venous oxygen sensing and thus pulmonary vasoconstriction and dilation remain to be investigated.

To summarize, although the duration of exposure to oxygen was short and the FiO₂ high, our findings may support the therapeutic value of oxygen in pulmonary vascular diseases in a short-term setting and may provide some physiological explanation of the effect of nocturnal or domiciliary oxygen therapy on improvement of exercise capacity and symptoms as previously shown, but future studies are needed to investigate the long-term effects on pulmonary hemodynamics (9, 10).

In the multivariable regression analyses, we saw that both mPAP and PVR are negatively associated with the carbon monoxide transfer coefficient (Kco). These findings indicate that patients with more severely compromised pulmonary hemodynamics reveal higher impairments of pulmonary gas exchange, or vice versa, which might be associated with more severe vascular remodeling. Strikingly, even after adjustment for K_CO, the effect size of both hypoxia and hyperoxia remained the same, indicating that the effect of different FiO₂ in our regression model remains robust.

Almost all patients (99%) showed Qs/Qt above the physiological threshold of 5 and 75% above the clinically more frequently used threshold of 10% (11). This is comparable with a previous study in which all 87 patients with PH revealed increased Qs/Qt (17). Interestingly, in the present study, changes in resting hemodynamic parameters under hypoxia and hyperoxia were all independent of the size of Qs/Qt. Thus, our data did not support the hypothesis that patients with an increased right-to-left shunt fraction, e.g., through a PFO or IPAVA, would reveal exaggerated responses in mPAP and PVR to hypoxia or hyperoxia, for instance by acting as an escape valve mechanism. This is in line with retrospective studies in subjects with PH and PFO that did not show any association with mPAP or PVR (16–18). However, due to the retrospective nature of the present study, we cannot provide data on the PFO as we do not systematically screen patients with saline contrast echocardiography. As we only calculated Qs/Qt under hyperoxia, we cannot exclude the possibility that Qs/Qt might change depending on the current FiO₂. Future approaches could include contrast echography under normoxia and hypoxia to detect potential changes in Qs/Qt at different FiO₂ (38) or investigate whether therapies for pulmonary vascular disease would influence shunt fractions.

Adjustment for age, sex, and PH disease classification (PAH vs. CTEPH) (2) in the multivariable regression model revealed that mPAP, PVR, and PAWP were independent of age and PH-classification. CI showed a negative association with age.

Exposure to hypoxia and hyperoxia was only of short duration (at least 10 min) and further studies are required to investigate whether longer exposure would consistently change hemodynamics and whether different oxygen fractions used would result in hemodynamic or clinical improvements on long-term. We assumed that a washout period of at least 10 min between different FiO₂ levels would suffice to avoid a significant carry-over effect of the interventions. Additionally, as all patients received the same order of FiO₂ intervention (first hypoxia and then hyperoxia), we cannot exclude any order or time effects, which could potentially affect results. For physiological reasons, the Qs/Qt from blood gases can only be calculated under FiO₂ 1.0 and we do not know whether a herewith-calculated shunt fraction remains stable under different FiO₂, an assumption that has been questioned (11, 19).