Prone positioning in mechanically ventilated patients with severe acute respiratory distress syndrome and coronavirus disease 2019

Abstract

Background: The management of COVID-19 ARDS is debated. Although current evidence does not suggest an atypical acute respiratory distress syndrome (ARDS), the physiological response to prone positioning is not fully understood and it is unclear which patients benefit. We aimed to determine whether proning increases oxygenation and to evaluate responders.

Methods: This case series from a single, tertiary university hospital includes all mechanically ventilated patients with COVID-19 and proning between 17 March 2020 and 19 May 2020. The primary measure was change in PaO₂ :FiO₂ .

Results: Forty-four patients, 32 males/12 females, were treated with proning for a total of 138 sessions, with median (range) two (1-8) sessions. Median (IQR) time for the five sessions was 14 (12-17) hours. In the first session, median (IQR) PaO₂ :FiO₂ increased from 104 (86-122) to 161 (127-207) mm Hg (P < .001). 36/44 patients (82%) improved in PaO₂ :FiO₂ , with a significant increase in PaO₂ :FiO₂ in the first three sessions. Median (IQR) FiO₂ decreased from 0.7 (0.6-0.8) to 0.5 (0.35-0.6) (<0.001). A significant decrease occurred in the first three sessions. PaO₂ , tidal volumes, PEEP, mean arterial pressure (MAP), and norepinephrine infusion did not differ. Primarily, patients with PaO₂ :FiO₂ approximately < 120 mm Hg before treatment responded to proning. Age, sex, BMI, or SAPS 3 did not predict success in increasing PaO₂ :FiO₂ .

Conclusion: Proning increased PaO₂ :FiO₂ , primarily in patients with PaO₂ :FiO₂ approximately < 120 mm Hg, with a consistency over three sessions. No characteristic was associated with non-responding, why proning may be considered in most patients. Further study is required to evaluate mortality.

Keywords: COVID-19; acute respiratory distress syndrome; intensive care; oxygenation; prone position; responders.

Figure 1

PaO₂:FiO₂ during five consecutive prone positioning sessions. Displayed as medians with IQR. ***P < .001, **P < .005

Figure 2

(a) PaO₂:FiO₂ (PFI) at the end of the first proning session as a function of the initial PaO₂:FiO₂. The line of no response is shown in black. (b) The predicted effect of initial PaO₂:FiO₂ (PFI) on the change in PaO₂:FiO₂ (ΔPFI) from the ordinal regression model, taken at the median values of the other covariables