Oxygenation improvement and duration of prone positioning are associated with ICU mortality in mechanically ventilated COVID-19 patients

Abstract

Background: Prone position has been diffusely applied in mechanically ventilated COVID-19 patients. Our aim is ascertaining the association between the physiologic response and the length of the first cycle of prone position and intensive care unit (ICU) mortality.

Methods: International registry including COVID-19 adult patients who underwent prone positioning. We measured the difference for arterial partial pressure of oxygen to inspired fraction of oxygen ratio (PaO2/FiO2), ventilatory ratio, and respiratory system compliance (Crs) between baseline supine position and at either the end of the first cycle of prone position (Delta-PP) or re-supination (Delta-PostPP).

Results: We enrolled 1816 patients from 53 centers. Delta-PP and Delta-PostPP for PaO2/FiO2 were both associated with ICU mortality [OR (95% CI) 0.48 (0.38, 0.59), and OR (95% CI) 0.60 (0.52, 0.68), respectively]. Ventilatory ratio had a non-linear relationship with ICU mortality for Delta-PP (p = 0.022) and Delta-PostPP (p = 0.004). Delta-PP, while not Delta-PostPP, for Crs was associated with ICU mortality [OR (95% CI) 0.80 (0.65, 0.98)]. The length of the first cycle of prone position showed an inverse relationship with ICU mortality [OR (95% CI) 0.82 (0.73, 0.91)]. At the multivariable analysis, the duration of the first cycle of prone position, Delta-PP and Delta-PostPP for PaO2/FiO2, and Delta-PostPP for ventilatory ratio were independently associated with ICU mortality.

Conclusion: In COVID-19 patients with acute respiratory failure receiving invasive mechanical ventilation and prone positioning, the physiological response to prone position is associated with ICU mortality. Prolonging the duration of the first cycle of prone position is associated with improved survival.

Keywords: Acute respiratory failure; Arterial partial pressure of oxygen to inspired fraction of oxygen ratio (PaO2/FiO2); Mechanical ventilation; Prone position; Respiratory system compliance; Ventilatory ratio.

Fig. 1

Study design and timeline. These time windows were desirable but not obligatory. The real timing of variable collection was recorded and variables registered more than 30 min later than required by the study design were excluded. PaO₂/FiO₂ arterial partial pressure of oxygen to inspire fraction of oxygen ratio. Vent Ratio, ventilatory ratio. Crs, static compliance of the respiratory system

Fig. 2

Relationship between ICU mortality and the response to the first cycle of prone position on the basis of arterial oxygenation at the end of the prone position cycle (Delta-PP) (A) and after re-supination (Delta-PostPP) (B), on the basis of ventilatory ratio at the end of the prone position cycle (Delta-PP) (C) and after re-supination (Delta-PostPP) (D), and on the basis of respiratory system static compliance at the end of the prone position cycle (Delta-PP) (E) and after re-supination (Delta-PostPP) (F). The x-axis shows the predictor and the y-axis shows the effect of the predictor on the outcome in log-odds. The solid line represents the regression line estimated by the logistic regression model, and the gray bands show the confidence intervals. When the relationship was found to be non-linear, the change-point was identified (red dotted line). Log-odds quantify the influence of factors on the probability of an outcome: a positive value indicates that the probability of the event increases with the factor, while a negative value suggests a decrease. If we suppose the logistic regression model gives a log-odds for Delta-PP for PaO₂/FiO₂ of −0.05, this log-odds value indicates that for each additional point of Delta-PP for PaO₂/FiO₂, the log-odds of death in the ICU decreases by 0.05. In panel A, the odds ratio of 0.48 for ICU mortality associated with an increase of Delta-PP for PaO₂/FiO₂ indicates that an improvement in the PaO₂/FiO₂ ratio within the interquartile range (i.e., from 26 to 124 mmHg) significantly reduces the odds of ICU mortality. Specifically, this increase from the 25th to the 75th percentile is associated with a 52% reduction in the odds of ICU mortality. The same rule could be applied for the interpretation of the linear relationships depicted in panels B, E, and F. When the relationship is non-linear, as in panels C and D, the odds ratio for ICU mortality refers to an increase of the variable within the range indicated in the table below the chart. For example, in panel C, the odds ratio of 1.27 for ICU mortality associated with the increase of Delta-PP for ventilatory ratio from 0.0 to 0.5 indicates that the odds of ICU mortality are significantly increased if Delta-PP for ventilatory ratio passes from 0.0 to 0.5. PaO₂/FiO₂ arterial partial pressure of oxygen to inspire fraction of oxygen ratio. Crs, static compliance of the respiratory system. OR, odds ratio. 95%CI, 95% confidence interval

Fig. 3

Relationship between ICU mortality and the length of the first cycle of prone position (A) and the total numbers of prone position cycles (B). The x-axis shows the predictor and the y-axis shows the effect of the predictor on the outcome in log-odds. The solid line represents the regression line estimated by the logistic regression model, and the gray bands show the confidence intervals. Log-odds quantify the influence of factors on the probability of an outcome: a positive value indicates that the probability of the event increases with the factor, while a negative value suggests a decrease. If we suppose the logistic regression model gives a log-odds for the length of prone position of −0.05, this log-odds value indicates that for each additional point of prone position duration, the log-odds of death in the ICU decreases by 0.05. OR, odds ratio. 95%CI, 95% confidence interval