All consecutive patients with acute respiratory failure and laboratory-confirmed SARS-CoV-2 infection admitted to one of the four participating dedicated COVID-19 ICUs in Paris, France, between February 21 and April 24, 2020, were enrolled. Acute respiratory failure was defined as respiratory rate ≥25, bilateral pulmonary infiltrates on chest X-ray or computed tomography scan, and need for standard oxygen ≥3 L/min⁻¹ to maintain peripheral arterial oxygen saturation ≥92%. Laboratory confirmation of SARS-CoV-2 was defined as a positive result of real-time RT-PCR assay of nasal and pharyngeal swabs (5). The study was approved by the ethics committee of the French Intensive Care Society (n. 20–23), which waived the need for informed consent from individual patients because of the retrospective nature of this chart review. Data were abstracted from the medical charts and electronic reports by attending intensivists at each hospital. Fi_O2 was calculated as 0.21+ (oxygen flow [L/min⁻¹] × 0.03) in patients receiving standard oxygen and was actual Fi_O2 in those receiving HFNC. In the four participating units, HFNC targeted a flow ≥50 L/min, which could be reduced in case of poor tolerance. Need for invasive mechanical ventilation and mortality 28 days after ICU admission were recorded. Continuous variables were described as median (interquartile range) and were compared between groups using the nonparametric Wilcoxon rank-sum test. Categorical variables were described as frequency (percentages) and were compared between groups using Fisher’s exact test. Mortality was assessed using survival analysis; Kaplan-Meier graphs were used to express the probability of death from inclusion to Day 28, and comparisons were performed using the log-rank test. We used a competing risks model to account for the risk of invasive mechanical ventilation while taking into account discharge alive and death as time-dependent competing risks. Comparisons were performed using the Gray test. Risk of death was assessed using Cox model including variables at ICU admission, such as oxygenation modality.

In a sensitivity analysis, a propensity score (PS)-matched analysis was performed according to factors associated with receiving HFNC. On the basis of a conditional backward model, the following variables were selected for inclusion into the PS model: immunosuppression, ICU admission within 7 days from symptom onset, vasopressors, and acute kidney injury. A case-matching procedure was performed on 1:1 ratio without replacement and according to the nearest neighbor method. The adequacy of the matching procedure was assessed by plotting PS across groups and assessing differences across groups using standardized mean difference. Univariate analysis and then double adjustment by Cox model were performed on relevant variables associated with outcome and those poorly matched.

Statistical analyses were performed with R statistical software, version 3.4.4 (available online at http://www.r-project.org/), and “Survival,” “Cmprisk,” and “MatchIt” package were used. P < 0.05 was considered significant.

Over the study period, 379 patients with COVID-19 (age, 66 [53–68] yr; 77% men) were admitted to the four ICUs for acute hypoxemic respiratory failure. Comorbidities included hypertension (50%), diabetes (30%), immunosuppression (18%), chronic kidney disease (17%), cardiovascular disease (8%), asthma (6%), or chronic obstructive pulmonary disease (5%). Median body mass index was 28 (25–32) kg/m⁻².

Overall, 146 (39%) patients received HFNC (all within the first 24 h after ICU admission) and were compared with 233 patients who did not. Table 1 shows the patients characteristics. None of the variables depicting patients’ characteristics at baseline significantly differed between the two groups. Patients who received HFNC were admitted after a longer period since symptoms onset, but time since hospital admission was not different. Pa_O2/Fi_O2 ratio was 126 (86–189) mm Hg and 130 (97–195) mm Hg in the patients who received HFNC and those who did not, respectively (P = 0.43). Sequential Organ Failure Assessment score at Day 1 was significantly lower in patients with HFNC (4 [3–5] vs. 6 [3–9], P = 0.001), which was consistent with a lower proportion of patients with acute kidney injury (40% vs. 60%, P < 0.0001) and vasopressors (29% vs. 53%, P < 0.0001).

The proportion of patients requiring invasive mechanical ventilation at Day 28 was 56% (95% confidence interval [CI], 47–64) vs. 75% (95% CI, 70–81; P < 0.0001 [Gray test]). Mortality at Day 28 was 21% in the HFNC group versus 30% in those who did not receive HFNC (hazard ratio [HR], 0.69; 95% CI, 0.45–1.07).

After adjusting on a PS to receive HFNC, 137 patients who received HFNC were matched to 137 patients who did not. Change in standardized mean difference before and after matching was excellent or good for most variables (Figure 1). HFNC was associated with a reduced proportion of patients requiring invasive mechanical ventilation at Day 28 (55% [95% CI, 46–63] vs. 72% [95% CI, 64–79]; P < 0.0001 [Gray test]; Figure 1B). Day 28 mortality was similar between the two groups (21% in the HFNC group vs. 22% in the other group; HR, 1.35; 95% CI, 0.56–3.26). These findings were similar in various sensitivity analyses adjusting for frailty effect on center (HR for mortality, 1.04; 95% CI, 0.62–1.73; and subdistribution HR for mechanical ventilation, 0.54; 95% CI, 0.39–0.75) and adjusting for frailty effect on center and remaing poorly matched variables, namely, Sequential Organ Failure Assessment score and body mass index (HR for mortality, 1.41; 95% CI, 0.82–2.44; and subdistribution HR for mechanical ventilation, 0.61; 95% CI, 0.44–0.85).

Figure 1. (A) Change in standardized mean difference before and after matching. (B) Cumulative incidence of invasive mechanical ventilation (blue line) while accounting for ICU mortality (green line) or discharge alive from ICU (orange line). AKI = acute kidney lung injury; BMI = body mass index; COPD = chronic obstructive pulmonary disease; ECLS = extracorporeal lung support; HFNC = high-flow nasal cannula; NSAIDs = nonsteroidal antiinflammatory drugs; RRT = renal replacement therapy; SOFA = Sequential Organ Failure Assessment score.

Download Figure | Download Powerpoint

[More] [Minimize]

Symptomatic management to restore oxygenation of severe acute respiratory failure is a major issue in this COVID-19 outbreak. This study suggests that HFNC significantly reduces intubation and subsequent invasive mechanical ventilation but does not affect case fatality (6). These findings are in line with a previous trial that demonstrated reduced intubation rates in the most hypoxemic patients (3) and that mortality is not affected by HFNC put forward the complexity of SARS-CoV-2 infection, in which the underlying lungs do not hold typical features of acute respiratory distress syndrome (7, 8). Instead, acute fibrinous and organizing pneumonia with organizing intraalveolar fibrin associated with notorious endothelial injury can be found in postmortem biopsies (7, 8). Moreover, the proportion of pulmonary embolism and acute kidney and myocardial injury are reported in much higher proportions than in typical acute respiratory distress syndrome (9). Finally, this study highlights that HFNC was as safe as standard oxygen in a large cohort of patients with COVID-19.