We have read the consensus document of the SEMICyUC, published in your journal, regarding noninvasive ventilatory support in adults with acute respiratory failure (ARF) due to COVID-19.1 The document states that «extrapolating the evidence in de novo ARF (sic)», high-flow oxygen therapy (HFO) would be the first-choice modality. Noninvasive ventilation is established as the second option in the event of insufficient patient response in the absence of immediate intubation criteria. This recommendation is based on two literature references. The first2 is the interim guidance of the World Health Organization (WHO), which positions HFO and NIV at the same level (yellow traffic light, conditional recommendation), since both therapies «should be used only in selected patients with hypoxemic respiratory failure». Curiously, in Remark 3, the WHO states that «compared with standard oxygen therapy», HFO reduces the need for intubation. This observation is based on the European/American clinical practice guide.3 However, this guide, in Question 5 on de novo ARF, explains that «the primary endpoint of intubation was not significantly different» in the FLORALI-REVA trial,4 and is not able to establish any recommendation because the evidence is of low quality.

The second reference is the FLORALI-REVA study.4 This was a clinical trial involving three cohorts (HFO, NIV and conventional oxygen), and with the proportion of patients requiring intubation as the primary endpoint. A statistical power of 80% in identifying a relevant difference (defined as 20%) in the frequency of intubation was calculated for this purpose. No statistically significant differences in the primary endpoint were recorded. In the rest of the study, analyses were made of post hoc comparisons between groups of patients, with Cox regression models to explain the primary endpoint and mortality. Both analyses could be biased: no adjustment for multiple comparisons was made, no model with time-dependent variables (HFO and NIV were interchanged) was used, and there may have been over-adjustment. They consequently could only serve to generate hypotheses that would have to be confirmed by future trials.

However, the SEMICyUC document does not cite a clinical trial5 specifically designed (power 80%) to detect a relevant decrease (now defined as 30%) in the intubation rate. In the mentioned study, NIV versus standard oxygen therapy was seen to significantly reduce the intubation rate in patients with de novo hypoxemic ARF. This experiment has not been replicated, though the preliminary data on the experience with COVID-19 in China appear to confirm its results. With a beta-binomial model, using an a priori non-informative construct, the probability that the intubation rate is lower with NIV versus HFO was 0.9993 (difference in rates = 0.444; 95%CI = 0.097−0.706).6

The current health emergency situation requires full dedication on the part of intensivists, but also a rational distribution of the available resources. If we seek to avoid intubations, perhaps NIV and HFO should be positioned at the same level as first choice option. The WHO has done so.