Difference in PaO2/FiO2 between high-flow nasal cannula and Venturi mask in hypoxemic COVID-19 patients

During the SARS-CoV-2 pandemic, the ratio between arterial blood partial pressure of oxygen and fraction of inspired oxygen in the inspired mixture (PaO₂/FiO₂, mmHg) was largely used for defining the severity of the respiratory failure and its progression, for deciding the appropriate respiratory support, and, consequently, for using specific pharmacological therapy [1]. Therefore, a careful assessment of PaO₂/FiO₂ is fundamental. As it is well known, PaO₂/FiO₂ is not the ideal variable for measuring the PO₂ alveolar-arterial gradient because it does not consider PCO₂ and its not linear relationship with FiO₂ regardless of the alveolar-arterial gradient [2]. Moreover, PaO₂/FiO₂ was initially proposed in mechanically ventilated patients, where FiO₂ is carefully measured. In non-mechanically ventilated patients, as patients in conventional O₂ mask or high-flow nasal cannula (HFNC), the assessment of the true FiO₂ in the inspired mixture may be problematic, specifically in dyspneic/tachypneic patients with high inspiratory peak flow or with the mouth breathing during HFNC [3, 4]. Therefore, PaO₂/FiO₂ values during the transition through the different methods of O₂ delivery could lead to misinterpretation of the degree of respiratory dysfunction. For the above reasons, we decided to investigate whether the transitions from HFCN and Venturi mask (VM), or vice versa, alter PaO₂/FiO₂ values because of potential undetected differences in true FiO₂.

Thirty consecutive patients admitted to our COVID-19 intensive care unit (ICU) because of severe respiratory failure due to SARS-CoV-2 infection and undergoing weaning from respiratory supports were included. As for internal protocol, after weaning from mechanical ventilation and SpO₂ > 90% with FiO₂ < 0.7 in HFNC (flow rate 60 L/min), the patients alternated HFNC and VM (FIAB S.p.A., model OS/60K, Florence, Italy) at the same FiO₂ levels with a progressive de-escalating time scheme. Twenty minutes after the transition from HFNC to VM or vice versa, we collected respiratory rate (RR), mean arterial pressure, heart rate, FiO₂, PaO₂, the partial pressure of carbon dioxide (PaCO₂), and pH in the arterial blood and calculated alveolar-arterial PO₂ gradient (P(A-a)O₂) [5]. To evaluate the differences between HFNC and VM, analysis of variance, linear regression analysis, and Bland-Altman method were used [6]. The study was approved by the ethical committee (658/2020/OSS*/AOUMO SIRER ID 417), and informed consent was obtained from participants.

Patients’ characteristics were detailed in Table 1. Twenty-nine patients received HFNC first and only one VM first. The PaO₂/FiO₂ measured in HFNC were like those measured in VM (Table 2) with a significant linear relationship (R² = 0.51, p < 0.01). In the Bland-Altman analysis, the mean value and standard deviation of the bias were 0 and 23 mmHg (Fig. 1). The RR and PaCO₂ values resulted to be different (p = 0.013 and p = 0.001) with RR higher and PaCO₂ lower in VM compared to HFNC (Table 2). In patients previously treated with HFNC, PaCO₂ was lower during VM in 25 patients (86.2%), and the difference was up to 5 mmHg in only 4 patients (13.8%).

Bild vergrößern

Our data indicate that after 20 min from HFNC to VM transition and vice versa, PaO₂/FiO₂ remain similar with only a small and not significant difference. Previous studies suggested that HFNC may improve oxygenation and decrease work of breathing compared to conventional O2 therapy [7‐9]. In contrast with previous reports, our results may suggest that there may be an underestimation of FiO₂ with Venturi device yielding overestimation of PaO₂/FiO₂ in contrast with the accurate delivery of FiO₂ with HFNC. Anyway, in this specific population under these circumstances, both these two phenomena have clinically acceptable limits of agreement. Moreover, we observed that RR differed between the two methods despite the very short period of exposure (20 min). The main part of the sample transitioned to VM after having received HFNC, so it cannot exclude a possible carry-over effect for PaCO₂ in this transition. Moreover, a possible carry-over effect cannot be excluded also for PaO₂/FiO₂ ratio. Rather than reducing work of breathing, the association of lower RR with a consensual increase of PaCO₂ supports the hypothesis that HFNC, compared to VM, could improve oxygenation with less requirement of alveolar ventilation for maintaining PaO₂ level. However, the low number of patients and the lack of assessment of the true FiO₂ limit any further speculation on this point.

In conclusion, our study demonstrated that although in HFNC and VM the FiO₂ is only estimated and may vary with the patient's respiratory pattern, PaO₂/FiO₂ measured with a VM may be considered a reliable parameter for respiratory dysfunction evaluation in severely hypoxemic patients during the transitions from different O₂ delivery methods.