Our most important finding was that after the first 24 hours in the prone position, the percentage of PaO2/FiO2 increase over baseline beyond its absolute value was independently associated with higher mortality. That might suggest that patients who raise PaO2/FiO2 less than 50% above baseline are refractory to, or failed, prone treatment. In addition, with each consecutive, additional cycle of prone the chances of survival decreased significantly. Older age was also a factor that was associated with lower survival after the first 24 hours in prone positioning.
A higher mortality in severe ARDS with PaO2/FiO2 values lower than 100 mmHg has been described in the literature (15, 16, 17). In our cohort, the initial value of PaO2 /FiO2 was not related to prognosis; instead, the determinant was the change in the PaO2/FiO2 after having been in prone position.
To our knowledge, ours is the study including the largest population of patients undergoing prone position, as well as the one presenting the more exhaustive analysis of oxygenation and respiratory mechanics changes, and of the number of prone cycles.
Many studies have analyzed the physiological response to prone position in patients with ARDS. Additionally, they included a small number of patients and a mix of mild, moderate and severe ARDS, so the final effect of prone position on oxygenation might have been attenuated (18, 19)
Previously, in a subgroup of COVID patients who received prone position, Langer and colleagues considered those who presented a 20 mmHg increase in PaO2/FiO2 as responders to the treatment. However, this improvement can be considered as having little significance in clinical practice(18). In our cohort of patients, the subgroup that increased PaO2/FiO2 by 20 mmHg (which corresponded to a 25% increase in PaO2/FiO2 from baseline approximately) had 87% mortality, and a 30% higher risk of dying than those who increased PaO2/FiO2 by 50%. Patients who were unable to raise PaO2/FiO2 values by more than 50% exhibited a very high mortality rate, between 72 and 87%.
Our findings regarding the physiological behavior of oxygenation in response to prone position are consistent with the results of other researchers, both in COVID-19 and non-COVID-19 populations. A recently published observational study comparing both populations demonstrates, after a sensitivity analysis, that the percentage change in PaO2/FiO2 after the first prone cycle predicted weaning of mechanical ventilation at 90 days, with an AUROC of 0.89 (20). Similarly, a retrospective study in moderate and severe ARDS in non-COVID-19 patients suggests a similar response to oxygenation after administering the first prone position cycle (21).
The dose of prone position required to improve oxygenation in a sustained manner is unknown. Short cycles have been described as ineffective (22); but it has not been established how many cycles, and how prolonged they should be, to consistently reverse hypoxemia. Guerin et al. utilized in the PROSEVA study 4.4 cycles per patient of 17±3 hours each (9). It has been suggested that SARS-CoV-2 as a cause of ARDS behaves more aggressively and might require a higher number of cycles.
Similarly, Langer et al, in a multicenter study of COVID-19 patients, reported an utilization of 3 [1-4] prone position cycles, although the percentage of patients who received this therapeutic maneuver – despite there was a number of patients who presented mild ARDS which may have interfered with the results(18).
On the other hand, the need to extend prone treatment identifies patients with poor response in terms of oxygenation or with an adequate response that cannot be sustained over time. Patients with a higher need for prone cycles or with a requirement for longer cycles are likely to have more severe lung disease. In addition, the chance of surviving with each subsequent cycle of prone treatment decreases. The delay in early identification of these patients may exclude them from other rescue therapies for severe hypoxemia which might have been applied earlier, such as extracorporeal membrane oxygenation techniques (13, 14). Therefore, patients who fail to increase PaO2/FiO2 by more than 25%, or who increase but subsequently deteriorate oxygenation and require additional prone cycles, could be candidates for other treatments for refractory hypoxemia, since the mortality rate in these patients might exceed 80% (17).
Patient age might be another factor contributing to decision-making. In our study, older patients had higher mortality, in agreement with what other researchers found: from the beginning in the pandemic, increasing age was a risk factor for severe illness due to COVID-19 (4, 23,24,25). Even in the subgroup of patients requiring critical care, age was associated with higher mortality. Recently, a study showed that elderly patients with COVID-19 who required life-sustaining treatments in the ICU had higher mortality compared to other elderly patients with similar treatments who were admitted for other diseases than COVID-19. (24)
Our study has strengths and weaknesses. This is an observational study, and as such is subject to biases and possible residual confounders. Part of the recruitment was done retrospectively. However, most of the data were recorded prospectively, using the same collection instrument used for routine daily monitoring in the ICU. Our findings are in line with what was found by other researchers, include a higher number of patients, and provide novel information regarding prone treatment in patients with ARDS. This might be useful for early decision-making in severely compromised patients, underscoring the need of escalating therapies in this group.