Respiratory infection due to SARS-CoV-2, COVID-19 disease, can cause severe hypoxemic acute respiratory failure (ARF), requiring admission to intensive care unit (ICU) and respiratory support.1 At the beginning of the pandemic, the use of non-invasive ventilatory devices (NIVD) was discouraged due to the lack of clear evidence of their efficacy in the treatment of severe hypoxemic ARF, the possible increase in the spread of the virus to the environment, the hypothetical risk of developing patient self-induced lung injury, and the worse prognosis derived from a delay in intubation.2–4 Thus, high-flow oxygen therapy through nasal cannula (HFNC), CPAP, and non-invasive ventilation in bilevel positive airway pressure (BiPAP) were rarely used, with initial recommendations favoring early intubation and invasive mechanical ventilation (IMV).5 After the first wave of the pandemic, and when there was evidence of a very high mortality in patients on IMV, together with a low contagion in duly protected health personnel, the use of CPAP and BiPAP became generalized.6,7

Multiple observational series as well as randomized controlled trials analyzing the use of NIVD in COVID-19 have shown very different results. In observational series, the CPAP or BiPAP failure rate is very variable, depending on multiple factors,7 but it is considerably high, around 30–50%,8,9 reaching 88% in a multicenter study.10 In clinical trials, the intubation rate of patients randomized to CPAP or BiPAP ranged from 30% to 47%.11–15 Due to these high failure rates, a recent consensus of different Spanish scientific societies concluded that during viral pandemics, the use of CPAP or BiPAP can be considered in carefully selected patients treated in centers with extensive experience and with optimized measures to prevent contagion.16 In addition, it is common to use different types of NIVD sequentially in the same patient, in relation to the clinical response or the presence of complications associated with the device, fundamentally those related to the interface and pressure levels.6,17

In the treatment of hypoxemic ARF, CPAP or BiPAP can be used. CPAP improves arterial oxygenation by recruiting non-functioning areas of the lung without providing inspiratory support. BiPAP, in addition to recruiting non-functioning alveoli by applying positive pressure at the end of expiration, can reduce inspiratory effort by applying inspiratory positive pressure and thus be the mode of choice in patients with chronic respiratory disease with hypercapnia.18

Different consensus recommended the use of CPAP as the first line treatment in patients with severe ARF.19,20 Despite this, multiple clinical series using BiPAP have been published.6 This variability may be related to the preferences and trust of the physicians and the availability of resources, but also to the degree of impairment of the patient's respiratory function.3,8

We hypothesize that BiPAP as first line of treatment is as effective as CPAP in hypoxemic ARF due to severe COVID-19. The primary objective of this study was to compare the rate of NIVD failure in patients with ARF due to COVID-19. As secondary objectives, we analyzed differences between the following four groups: patients who only received CPAP, those who only received BiPAP, those who initially received CPAP and were crossover to BiPAP, and those initially with BiPAP and who were crossover to CPAP. In addition, the differences between ICU and in-hospital mortality, and complications related to NIVD were also analyzed.

We conducted a retrospective observational study with a prospective database in an ICU of a University Hospital. The study was approved by the Institution's Ethical Committee.

During the study period 478 patients were admitted with positive PCR, of which 49 were excluded (Fig. 1). Of the 452 patients admitted with COVID-19 related ARF, 429 (94.9%) were analyzed. Three hundred and twenty-eight (76.5%) patients initially received CPAP, and 101 (23.5%) BiPAP. During the ICU stay, some patients crossover from one type of respiratory device to the other, in such a way that 120 (28%) patients only received CPAP, 64 (14.9%) only BiPAP, and 245 (57.1%) received both types of NIVD: 208 (48.5%) crossover from CPAP to BiPAP and 37 (8.6%) from BiPAP to CPAP.

Figure 1.

Flow chart ICU patients.

(Definitions of abbreviations: ARF: acute respiratory failure, CPAP: continous positive airway pressure, CVC: central venous catheter; HFNC: high flow nasal cannula, IMV: invasive mechanical ventilation, PCR: real-time polymerase chain reaction)

(0.21MB).

In this study, it was shown that the management of COVID-19 associated ARF requiring NIVD can be performed using CPAP or BiPAP with similar effectiveness and safety.

NIV has shown efficacy in preventing endotracheal intubation in patients with acute on-chronic respiratory failure.21 However, evidence of benefit has only been demonstrated in mild-to-moderate hypoxemic ARF.21 Several recent systematic reviews have evaluated the role of NIV in the treatment of hypoxemic ARF, in both COVID-19 and non-COVID-19 settings.22–24 The results of a network meta-analysis suggest that NIV, HFNC, and conventional oxygen therapy are comparable in terms of treatment failure and mortality in COVID-19.22 Pitre et al. demonstrated that Helmet NIV strategies are probably effective at reducing mortality, while HFNC, facial-mask NIV, helmet bilevel ventilation, and CPAP reduce the need for IMV compared to standard oxygen-therapy.23 In the study by Sakuraya et al., a decrease in mortality was observed with the use of CPAP, but not with NIV.24

Some studies associate the use of NIV with a worse prognosis.25–27 In a multicenter study performed on patients in and outside ICU during the first wave COVID-19, CPAP failure rate was 36.8%, while NIV failure rate was 60.8%.25 Franco et al., in a multicenter Italian study, conducted during the first wave, found a higher rate of mortality/intubation in patients receiving NIV than CPAP (47.3% versus 53%), but the differences disappeared after adjusting for confounding variables.26 In another multicenter observational study, carried out in high dependency units, with 54% of patients having a DNI order, Crisafulli et al. showed a higher failure rate, defined as intubation or death, when NIV was used initially or during the hospital stay, compared to CPAP (34% versus 60%).27 Like these studies, our work also showed a higher failure rate in the BiPAP group, however, it was lower than in the previous studies: 34,7%, while CPAP failure rate was 25%. However, the adjustment of variables showed that mortality did not differ between the two types of NIVD.

Although the characteristics of the patients in the mentioned studies were quite similar, the management protocols of NIVD clearly differed between them, as well as the type of hospital unit to which the patients were admitted. In a recent observational study in COVID-19 patients with mild-moderate ARF, the failure rate of CPAP was 3.9%.28 The lower failure rate of CPAP in this series compared to the one from our study may be related to the lower severity of the patients (initial PaO2/FiO2 of 193&#¿;mmHg versus 120&#¿;mmHg).

Currently, there is no clear standardization for the use of NIVD in the treatment of ARF, neither in relation to the time of initiation nor with the type of respiratory device to be used, nor with the programmed parameters. A wide variability is shown in this area,6,11–15 both in randomized controlled trials and in observational studies. Most studies that have compared CPAP with BiPAP in adult patients have been in cardiogenic pulmonary edema,29 and those performed in patients with ARF of other etiologies have been scarce and have not shown differences between both modalities.30

We advocate for a sequential treatment of ARF, in which HFNC, CPAP and BiPAP are used depending on the patient's condition. During ARF, variations in the patient's clinical situation are frequent, due to the evolution of the disease itself, and due to the presence of complications associated with the NIVD and interfaces used, fundamentally the presence of claustrophobia, pain or intolerance.17 Because of this, even though most patients initially receive CPAP, crossover to NIV is frequent when the clinical situation does not improve. Likewise, it is frequent crossover from NIV to CPAP in the presence of complications related to NIV. In a recently published pilot study on the timing of starting CPAP or NIV in patients with ARDS due to COVID-19, the initial treatment with CPAP was crossover to NIV due to an inadequate respiratory response.31 Crossover from CPAP to NIV, and from NIV to CPAP was frequent in the study by Crisafulli et al.27 The failure rate of BiPAP in our study was slightly higher than CPAP. However, this higher failure rate with BiPAP may be explained, at least partially, by a greater severity of the patients who initially received BiPAP. The pressure support used during BiPAP can also cause an increase in tidal volume (promoting volutrauma), which has been related to BiPAP failure.32 Finally, a greater effect of patient self-induced lung injury cannot be ruled out due to an increase in inspiratory effort that could occur during the use of NIVD. In the case of BiPAP it may be related to patient-ventilator desynchrony, whose effect may be potentiated by higher tidal volumes associated with pressure support.33 However, the difficulty of measuring inspiratory effort during NIVD, as well as the lack of a definitive treatment to reduce patient self-induced lung injury make it difficult to know the exact role that this process plays in NIVD failure.34 Despite these considerations, when adjusting for different factors that may condition a worse prognosis, through paired propensity analysis, the differences in the failure rate between BiPAP and CPAP diminished (OR 1.37, CI 95% 0.72–2.62).

However, there were two groups of patients with a worse prognosis: those with continuous need for BiPAP, which had a failure rate of 54.7% and in-hospital mortality of 32.8%, and those with a need for crossover from CPAP to BiPAP, with a failure of 38% and in-hospital mortality of 22%. These hallmarks were also shown in a study by Nevola et al., where the need for crossover from CPAP to NIV was a risk factor for NIV failure in the univariate analysis, although not in the multivariate analysis, while the need for continuous NIV was an independent predictor of poor prognosis.31 Crossover to more complex therapies usually indicate greater severity and worse prognosis.35 In our series, few patients were treated with ECMO. Only four patients received this therapy, all of them under 60 years old and without relevant comorbidities. All four were intubated and died in the hospital.

Few studies have analyzed the relationship between the type of NIVD and complications. In a small study, Pontes et al. found a greater number of complications in patients treated with BiPAP than with CPAP.36 The use of NIV leads in many cases to the appearance of asynchrony between the patient and the ventilator, which is one of the factors that can cause intolerance. Despite this consideration, in a series of patients with NIVD, intolerance was slightly more frequent in patients treated with CPAP than with NIV.37 In our study, total intolerance was very low (1.6%), despite the duration of NIVD, probably related to the systematic use of opiates.

One of the reasons for not using NIVD at the beginning of the COVID-19 pandemic was the possible association between a delay in endotracheal intubation and a worse prognosis. This relationship, shown in de novo ARF patients of different etiologies treated with and without NIV3,38 has been showing contradictory results in ARF related to COVID-19.30,38–40 In our work, the average time to intubation was high, being greater in patients initially treated with CPAP (median of 5 days in overall population and 5.5 days in population adjusted through propensity) even though mortality did not differ between both groups.

This study has several limitations. First, it is an uncontrolled observational study. The effectiveness of a treatment must be evaluated through a well-designed controlled and randomized trial, where the confusion variables, known and not known, are controlled through an appropriate randomization. In observational studies, control of some of the confusion variables can be attempted to imply causality through different techniques, including propensity score-matched analysis, but even in these circumstances, since there is the possibility that some of the analyzed variables were not considered adequately (and that there might be other confusion variables which were not measured), the conclusions must be evaluated carefully, and the results of these analyzes must not be considered definitive. On the other hand, it must be noted that the comparison between the adjusted groups may lead to problems arising from over-adjustment. Also, the number of patients in the sample may not be sufficient to ensure adequate statistical power. Secondly, our patient management protocol includes several novel points, such as the use of full-face mask, the systematic use of perfusion and bolus of fentanyl, the use of HFNC in weaning or in the presence of complications, which may be controversial since there is no clear evidence for its use. This makes it difficult to extrapolate the results to other protocols on the use of non-invasive respiratory devices. Finally, the study was carried out in a single center, and in an ICU with extensive experience in the use of NIV in hypoxemic ARF patients, with highly trained nursing staff, an adequate nurse-patient ratio, and high-performance ventilators and devices, which can make it difficult to extrapolate the results to other units with fewer resources or experience.

Our results point to the possibility of using CPAP or BiPAP in patients with moderate-severe COVID-19 associated ARF. Bearing in mind that the use of CPAP is more common than that of BiPAP, the selection between them will tend to involve the experience of healthcare teams, the severity and antecedents of the patient, as well as the availability of resources, ventilators, monitoring and health staff, as well as the place/unit to which the patients was admitted to. Also, respiratory management must be individualized, depending on the pathological and clinical characteristics of the patient, and on the type of NIVD (and the parameters to be programmed), to ensure that the patient is not intubated and can be discharged alive from the hospital.

COVID-19 associated hypoxemic ARF can be treated initially with non-invasive positive pressure respiratory devices with a high success rate and avoidance of endotracheal intubation and death. These findings suggest that while both CPAP and BiPAP can be effective in managing acute respiratory failure, the choice of treatment should be tailored to the patient's clinical profile, with close monitoring for those who may require escalation of care.

Andrés Carrillo Alcaraz: Data mining Statistical analysis. Elaboration and drafting of the article.

Miguel Guia: Revision and drafting of the article.

Laura Lopez-Gomez: Data mining. Drafting.

Pablo Bayoumy: Data mining. Drafting.

Aurea Higon-Cañigral: Data mining. Drafting.

Elena Carrasco González: Data mining. Drafting.

Pilar Tornero Yepez: Data mining. Drafting.

Juan Miguel Sánchez-Nieto: Data mining. Drafting.

None.