The novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can cause a different spectrum of illnesses, ranging from asymptomatic infection to atypical acute respiratory distress syndrome in 4–16% of cases.1,2 When hypoxemic acute respiratory failure (ARF) occurs, supplemental oxygen is the first-line medication. For ARF patients developing hypoxemia despite receiving conventional oxygen therapy, with no indications for invasive mechanical ventilation (IMV), international guidelines recommend the use of high flow nasal cannula (HFNC) instead of conventional oxygen therapy or noninvasive positive pressure ventilation (NPPV).3 If HFNC is not available and there is no urgent indication for endotracheal intubation, current guidelines recommend a trial of NPPV with close monitoring and short-interval assessment for deterioration, even though this procedure is supported by very low-quality evidence.3 Furthermore, in acute respiratory distress syndrome (ARDS) patients, the effectiveness of noninvasive continuous positive airway pressure (CPAP) remains largely undocumented.

The Italian Respiratory Society (SIP/IRS) and the Italian Thoracic Society (AIPO-ITS) have recently issued a clinical management algorithm for the treatment of COVID-19 patients requiring oxygen escalation through helmet CPAP. More recently, a multicentre observational study has shown that CPAP along with HFNC and NPPV can be readily applied outside of the ICU environment during COVID-19.4 However, intubation after CPAP failure has been shown to occur in 22–25%5 of patients, and other studies on noninvasive ventilation have reported that delayed intubation might worsen patient survival.6 For this reason, the use of CPAP in COVID-19 patients is still under debate. Hence, the identification of early predictors of CPAP failure represents an urgent unmet clinical need.

In the present study, we identify for the first time early factors associated with intubation in COVID-19 patients treated with CPAP out-of-ICU and candidate for intubation in case of CPAP failure. This information may help clinical decision making in these challenging pandemic times.

From March 1st to April 15th, 2020, a total of 397 patients were enrolled in the study. Of these, 30 (7.6%) patients were excluded from the study as they had been intubated on the same day as that of CPAP initiation. Two-hundred-seventeen patients were successfully treated with CPAP, while 150 patients failed CPAP and were thus subjected to endotracheal intubation. Table 1 shows the distribution of demographic and clinical characteristics of the patients stratified by “successful CPAP” or “failed CPAP” as well as the P-value derived from the univariable models. Table 1 s shows the same data stratified by centre. Among demographic and clinical patients’ characteristics, age (65 vs. 68 years, P = 0.005), gender (male 68% vs. 80%, P = 0.010), LDH (518 vs. 654 U/L, P < 0.001) and white blood cells (6.75 vs. 7.04 × 103/µL, P = 0.01) resulted significantly different between successful and failed CPAP. Helmet was applied in 77% and 85% of CPAP successes and failures, respectively. Fifteen patients received CPAP trough both helmet and mask. CPAP duration was significantly different among CPAP successes and failures (8 vs. 4 days, P < 0.0001). Most patients received hydroxychloroquine and prophylactic enoxaparin. Noteworthy, steroids were administered more frequently in the CPAP success group (49% vs. 25%, P < 0.0001).

Among respiratory variables, ΔPaO2 (59% vs. 19%, P = 0.007), ΔFiO2 (0% vs. 0%, P = 0.035), Δlactate (-12% vs. 0%, P < 0.001), ΔPaO2/FiO2 (87 vs. 44%, P = 0.001), ΔROX index (19 vs.7%, P = 0.039), resulted significantly different between CPAP success and failure. Table 2 reports ABG values obtained during Venturi mask oxygen therapy prior to CPAP initiation and those obtained at 2–24 h after CPAP, while Table 2s reports ABG values stratified by centres.

The values of PaO2, SpO2, lactate and RR, recorded during Venturi mask oxygen therapy, as well as those of PaO2, SpO2, FiO2, lactate, RR, PaO2/FiO2 and SpO2/FiO2, measured during CPAP, resulted significantly associated with CPAP failure.

Multivariable analysis of both clinical data on admission and the Δ% ABG values between Venturi mask and CPAP, allowed us to develop 4 models (Table 3). The best models were those obtained by adjusting for age, gender, CCI, white blood cell and lymphocyte count, LDH and C-reactive protein levels, Δlactate and either ΔPaO2/FiO2 or ΔPaO2 and ΔFiO2, separately, both leading to an area under the curve of 0.818.

From the model analysis, we found that age (RR, 1.026; 95% CI, 1.009−1.044, model 3), gender (RR, 1.718; 95% CI, 1.094−2.699, model 3), LDH (RR, 1.001; 95%CI, 1.000−1.001, model 3), and PaO2/FiO2 (RR, 0.998; 95%CI, 0.996−0.999, model 3), were independent predictors of intubation.

Our study, comprising 367 COVID-19 positive patients treated with out-of-ICU CPAP due to hypoxemic ARF and candidate to intubation in case of CPAP failure, shows that gender, LDH on admission and percentage of increase in PaO2/FiO2 between Venturi mask and CPAP are independent predictors of CPAP failure once corrected for the major clinical variablesi.e., age, CCI, white blood cell and lymphocyte count and C-reactive protein levels on admission.

A number of investigations seeking to find relevant outcome predictors among hospitalized patients with COVID-19 have shown that, on admission, increased D-dimer concentration11 and neutrophil-to-lymphocyte ratio12 as well as enhanced levels of C-reactive protein,13 creatinine13 and cardiac troponin I14 are all associated with a higher risk of intubation. Similarly, a body mass index (BMI) ≥35 kg/m,15 increasing age,15 male sex,15 comorbid status,13 respiratory rate,13 and SpO213 have been shown to be independently associated with worse in-hospital outcomes.

To the best of our knowledge, there has only been one out-of-ICU study, performed in high-dependency units, evaluating some potential predictors associated with CPAP failure, defined as death or intubation, in CPAP-treated patients (n = 157).5 The results of this study show that severity of pneumonia on admission and enhanced baseline IL-6 levels are both associated with death and intubation.5 Of note, 41.4% of the patients included in that study had a do-not-intubate order.

Here, we report the first large multicenter study on predictors of intubation in out-of-ICU COVID-19 patients (n = 367) candidate for intubation in the case of CPAP failure. Among the clinical and laboratory characteristics considered, we show that male gender is associated with a higher risk of intubation, in good agreement with what reported by previous studies on COVID-19 patients admitted to hospital16–18 or ICU.19

Our results identify LDH as a bona fide predictor of out-of-ICU CPAP failure. LDH is a ubiquitous intracellular enzyme, which catalyzes the interconversion of pyruvate and lactate, with concomitant interconversion of NADH and NAD+. High LDH values, resulting from multiple organ injury and decreased oxygenation paralleled by upregulation of the glycolytic pathway,20 have been associated with worse outcomes in patients with viral infections, such as severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS).21–24 In line with our findings, elevated LDH values have been recently shown to be associated with increased risk of severe COVID-19 pneumonia and mortality.25

A known predictor of intubation in NPPV-treated patients is PaO2/FiO2 <200 mmHg, 1 h after NPPV initiation.26 In our study, during Venturi mask oxygen therapy, the PaO2/FiO2 values were similar in both the “failed CPAP” and “successful CPAP” patient groups. In contrast, during CPAP therapy, those patients who failed had a median PaO2/FiO2 value of 121 mmHg, while those who succeeded had one above 200 mmHg.

In our study cohort, the intubation rate was 45%, also taking into account those patients (n = 30) excluded from the study because intubated on the same day CPAP was started. This intubation rate is similar to that observed in non-COVID-19 ARF patients treated with NPPV,26 but almost twice as high as that reported by two large studies on CPAP-treated COVID-19 patients (25%4 and 22%5), comprising both “do-not-intubate” and “candidate for intubation” patients. In the cohort of patients included in the study by Franco et al.,4 the severity of hypoxemic ARF, as assessed by PaO2/FiO2 is lower than in our investigation, which might explain the lower rate of CPAP failure. In the study by Aliberti et al.,5 when "do-not-intubate" patients are excluded, as in our study, the intubation rate increases up to 37%.

Finally, though beyond the aims of the present investigation, it is worth mentioning that droplets and large aerosols generated by all the respiratory therapies included CPAP in patients with COVID-19 may represent a risk factor for healthcare worker contamination despite using the recommended personal protection equipment.27,28 In a multicentre observational study, in healthcare workers treating 670 consecutive patients with confirmed COVID-19 with CPAP, NPPV or HFNC, 11.4% tested positive for infection.4 In our study, although we were not able to collect data on healthcare workers positive swabs, all strategies aimed at containing healthcare workers contamination i.e., appropriate personal protection equipment and to minimize droplet spread from CPAP i.e., minimize leaks, exhalation filter, have been undertaken.

One limitation of our study is that we were not able to assess the predictive value of the respiratory rate and ROX index and of immune system activation markers (e.g., ferritin, D-dimer) in our models due to the large number of missing data. We were also unable to record some other potential early predictors of intubation, such as ventilation-related markers (e.g., expiratory tidal volume). Missing data cause weakness in detecting statistically significant association. Nonetheless, we found LDH and PaO2/FiO2 to be associated with higher intubation probability, which, assuming that what is missing is random, corroborates their value as outcome predictors. Evidence in medical treatment at the time the study was conducted, was not against hydroxychloroquine and not yet in favor of corticosteroids as subsequently. Therefore, our data need to be confirmed in patients treated differently. Lastly, ABG after CPAP was performed over a time period ranging from 2 to 24 h, which under normal circumstances may be considered a very long time span, but it was not regarded as such in our case because we were operating during the early phase of the COVID-19 pandemic.

Our results reveal that gender, LDH on hospital admission and percentage of increase in PaO2/FiO2 changing from Venturi mask to CPAP therapy are independent predictors of intubation in out-of-ICU CPAP-treated COVID-19 patients.

RV, NDV, FDC and GC, had the idea for and designed the study, had full access to all of the data in the study, took responsibility for the integrity of the data, contributed to the design of the study and to the interpretation of data for the work and drafted the paper. FBA and LS had the idea for and designed the study, had full access to all of the data in the study and, took the responsibility for the accuracy and the data analysis and drafted the paper. PN contributed to the design of the study and to the interpretation of data for the work and drafted the paper. FR, CP, CM, DC, CO, ES, LC, TC, MT, LG, MAM, GA, SB, MB, SBai, PB, SBa, VB, SC, FC, VD, LDC, MMae, MM, FMo, RP, MP, VR, DR, LV, and FV contributed to the acquisition, integrity and analysis of the data. FM contributed to the interpretation of data for the work. All authors 1) agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved; 2) revised the work critically for important intellectual content; and 3) gave final approval of the version to be published.

FR reports personal fees for lectures from Philips Respironics, outside the submitted work.

CO has a patent, No. 102016000114357, with royalties paid from Intersurgical SpA.

DC reports personal fees from Nestlé Healthcare nutrition, outside the submitted work.

FM received fees for lectures from GE Healthcare, Hamilton Medical, Seda Spa; consulting agreement between University of Pavia and Hamilton Medical.

PN reports personal fees from Intersurgical SpA, Resmed, Philips, Novartis, MSD, Getinge, Orion Pharma and non-financial support from Draeger, outside the submitted work. In addition, PN has a pending patent, No. 102020000008305, filed to the Università di Padova, and a patent, No. 102016000114357, with royalties paid from Intersurgical S.p.A.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.