In this observational study out of 213 patients with Covid-19 related severe hypoxemic respiratory failure and instructed to lie in the prone position while receiving HFNC, 35 patients did not tolerate prone position while the remaining 178 (83.5%) had HFNC in awake prone position. Intubation (HFNC failure) was required in similar proportions in patients who tolerated prone position (52/178; 29%) and in those who had HFNC in the supine position during their ICU stay (9/25; 26%). Delta ROX corresponding to the difference between ROX index calculated at ICU admission and after the first HFNC session was the best predictor of the outcome of HFNC. A ROX cut-off ≤1.8 had the best operative characteristics in predicting the failure of HFNC. HFNC outcome was independently associated with indicators of ARF severity (such as SOFA score, baseline PaO2/FiO2, or baseline ROX) and not with body position.
Since the start of the pandemics, Covid-related severe hypoxemia revealed a great challenge to intensivists, with evolving paradigm regarding the place of non-invasive ventilation in this setting. From the ban unanimously applied on NIV during the first wave, evidence has progressively emerged on the reduced risk incurred by health workers and environment contamination when airways are managed without intubation, authorizing therefore the use of different NIV tools [10, 15, 36–40]. In early ARF, bilevel NIV delivered through helmet interface, and HFNC are deemed more effective than bilevel NIV delivered via facemask . CPAP has recently emerged as the preferred tool of oxygenation in hypoxemic Covid-19 patients.
Combining prone position in awake patients under spontaneous ventilation whichever the tool, is an idea that progressively gained scientific ground [12, 17, 19, 20, 42, 43]. Prone positioning is indeed a first-line intervention in patients invasively ventilated for severe ARDS leading to substantial reduction in mortality through both, oxygenation improvement and VILI reduction[25, 44, 45]. Its use in awake and spontaneously breathing patients with ARDS secondary to COVID-19 has recently been reported to improve oxygenation in few retrospective or small sample prospective cohorts[19, 20, 43, 46, 47]. One of the mechanisms that could explain the beneficial effect of ventilation in prone position is the homogenization of the VA/Q mismatch as shown in a case report by Zarontonello et al .
The observational nature of our study does not allow reliable assessment of prone impact on hard outcomes such as HFNC success or ICU mortality. The lack of randomization precludes random assignment to prone position and does not exclude statistical imbalance in variables with outcome relevance between patients who did tolerate prone position and those who did not. However, studies of varying sizes can provide information on the magnitude of the benefits associated with the application of awake prone HFNC in hypoxemic Covid-19. Small recent studies on the use of prone position in spontaneously breathing patients with Covid-19 related hypoxemia reported lower success rates of HFNC: 34% in Zucman et al study, and 46.7% in Blez et al study, compared to 64% in our cohort [32, 48]. Conversely, in a retrospective cohort analysis of Covid-19 patients oxygenated with HFNC whether in supine or in prone position, Ferrando et al reported similar rates of intubation in both groups (41% and 40%, respectively) and no significant risk in 28-day mortality.
In a landmark multinational meta-trial, Ehrmann et al have recently provided the evidence for the superiority of awake prone positioning over supine, in Covid-19 related hypoxemic respiratory failure. The authors pooled and analyzed in advance the individual observations of patients already included in prospective studies in progress spread across the world. These studies had in common the objective of evaluating the impact of delivery of awake prone HFNC in patients with Covid-19 acute hypoxemic respiratory failure. The meta-trial included 1,121 patients and reported a reduced HFNC failure rate (intubation or death) in patients assigned to awake prone positioning (40%), compared to patients assigned to standard care (46%; Relative risk 0·86 [95% CI 0·75−0·98]). They recommended accordingly routine awake prone position in patients with COVID-19 who require HFNC. They even claimed that equipoise surrounding the prone positioning of patients treated with HFNC does no longer exist recommending therefore to stop ongoing RCTs on the issue as performing HFNC in supine position becomes no longer ethical. Although observational in its design, our study was conducted before the publication of the meta-trial by Ehrmann et al and raises no ethical issue. Moreover, our observation of no significant difference between HFNC in supine or in prone position should not be considered in contradiction with that of Ehrmann et al since 95% CI boundaries surrounding point estimates overlap. Several explanations might account for the lack of statistical significance in our study. First, the lack of randomization cannot exclude an imbalance in baseline risk in patients who tolerated and those who did not tolerate prone position. Second, we cannot exclude a type II error and an underpowered study. Third, the multinational character of Ehrmann's study which on one hand reinforces the result’s extrapolation might in the other hand, introduce a center bias. Mexico (and France) provided the highest number of participants whereas Spain, Ireland, and Canada contributed only modestly to the study participants. Discrepancy in study protocols or in clinical practice for procedures that were not standardized by protocols (such as actual prone duration which varied from 1.6 hour in Spain to 8.6 h in Mexico) might have impacted to a certain extent the failure/success rates. Besides, given its sample size, the Mexican subgroup had the highest weight among included subgroups and substantially impacted the final meta-trial result as it was the only positive trial per se. One might speculate that the effect of awake prone position is risk-dependent requiring therefore a risk-stratified analysis. In this connection, the lower failure rate reported in our study compared to that reported in the meta-trial by Ehrmann et al, might suggest a lower overall failure risk in our population compared to that of the meta-trial. However, the rate of HFNC failure reported in our study compares fairly with that usually reported outside the setting of Covid-19 (34.3% in the meta-analysis by Rochwerg et al), or in Covid-19 .
In addition to information on feasibility and outcome of awake prone in real life, our study provides information on predictors of HFNC outcome in the specific setting of Covid-19 related hypoxemia. Our study confirms the performance of the ROX index evaluated elsewhere outside the Covid-19 setting. We show that the variation of the ROX index between baseline and the end of the first HFNC session, rather than a unique baseline measurement, is the best indicator of HFNC outcome. An increase by less than a threshold of 1.8 appears to have the best operative characteristics to predict HFNC failure. Early following setting up oxygenation with HFNC, clinicians have a reliable tool on outcome prediction and decision-making.
The usual questions in matters of HFNC arise therefore: in what proportion failure occurs, and how to predict them early enough to avoid the loss of chance related to delayed intubation in these patients. Although non-randomized, our study could answer most these previously enumerated relevant questions. It allowed precision of the rate of ROX index proved effective in the prediction of HFNC failure. In patients ventilated either in prone or in supine position, the cut-off of ROX variation between baseline and the end of the first HFNC session had the best operative characteristics among other physiologic variables such as severity of hypoxemia or general severity scores (SOFA). Our study further clarified the impact of HFNC combined to prone position on patient’s survival.