LAUSR

inspired oxygen (FIO2) ratio, higher CRP (C-reactive protein) concentrations, and lower platelet counts were independently associated with an increased risk of NIV failure. However, we believe that using the PaO2/FIO2 ratio is a rather questionablemeansofpredictingNIV failure. First, PaO2/FIO2 values are inherently inaccurate because theFIO2 in anonintubatedpatient is quite difficult to determine (2). And this is especially true in the context of non-ICU departments when using free-flow systems under the conditionofagas leakaroundthemaskorhelmet.Second, thePaO2/FIO2 thresholdof150mmHgisalsorathercontroversial. InastudybyFranco andcolleagues (3), therewerenodifferences in the30-daymortality rate between patients with COVID-19 with baseline PaO2/FIO2 values of 101–150mmHg and patients with COVID-19 with baseline PaO2/FIO2 values of 151–200 mmHg (24% and 26%, respectively). Third, in two other studies (4, 5) conducted outside the ICU, the mortality rates of patients with COVID-19 with even lower baseline PaO2/FIO2 values duringNIVwere better than that in the studybyBellani and colleagues: the rate was 17% in a study by Brusasco and colleagues (initial median PaO2/FIO2 of 119 mmHg) (4), and the rate was 21% in a study by Nightingaleandcolleagues (initialmedianPaO2/FIO2of122mmHg)(5). In our recent study, in patients with COVID-19 receiving NIV outside the ICU, thebaselinePaO2/FIO2 indexalsodidnotdiffer between the success and failure groups, but we found that patients who experiencedNIVfailurehadhigherminuteventilationonthefirstdayof NIV (due to a slightly higher tidal volumeandahigher respiratory rate) (6), which, of course, may increase the risk of self-inflicted lung injury. These findings are in line with Bellani and colleagues’ data, which also showed that the patients experiencing NIV failure had lower arterial carbon dioxide pressure levels. However, we must admit that today in non-ICU settings, it is rather difficult to identify robust markers of possible self-inflicted lung injury. Bellani and colleagues also identified elevated levels of CRP as an independent predictor of NIV failure. Interestingly, in our study, an elevated D-dimer level was an indicator of the increased possibility of NIV failure (6). Both high CRP levels and high D-dimer levels are associated with the progression of COVID-19 and a higher mortality rate. So, progressive underlying processes in COVID-19 might need prolonged respiratory support and can be associated with NIV failure. In conclusion, because the appropriate patient selection is the key to the successful application of NIV, further research is needed to identify reliable predictors of NIV failure in COVID-19–associated AHRF.