Abstract
BACKGROUND: COVID-19 pneumonia has been responsible for many ICU patients’ admissions with hypoxemic respiratory failure, and oxygen therapy is one of the pillars of its treatment. The current pandemic scenario has limited the availability of ICU beds and access to invasive ventilation equipment. High-flow nasal cannula (HFNC) can reduce the need for orotracheal intubation compared with conventional oxygen therapy, providing better results than noninvasive respiratory support. However, HFNC use has been controversial due to concerns about the benefits and risks of aerosol dispersion. In this context, we evaluated the performance of the HFNC therapy in patients with COVID-19 and investigated factors that can predict favorable responses. METHODS: A prospective observational study was conducted, which included hospitalized adult subjects with COVID-19 in the respiratory wards who needed oxygen therapy. Clinical and laboratory parameters were collected to compare HFNC therapy use and the outcomes. RESULTS: In 6 months, 128 subjects were included and the success rate of HFNC therapy was 53%. Logistic regression analysis showed that the Charlson comorbidity score, need for oxygen flow, 
, and breathing frequency predicted therapy failure. The mortality rate increased among the non-responders versus the responders (47% vs 3%), 48% of failure occurred in the first 24 h of the HFNC therapy. A ROX (respiratory frequency – oxygenation) index > 4.98 in 6 h and > 4.53 in 24 h predicted success of the HFNC therapy with an area under the curve of 0.7, and a ROX index < 3.47 predicted failure with 88% of specificity. CONCLUSIONS: HFNC in the subjects with COVID-19 was associated with reduced mortality and improved oxygenation in the subjects with respiratory distress. Close monitoring of specific parameters defines eligible patients and rapidly identifies those in need of invasive ventilatory support.
Introduction
COVID-19 is a potentially fatal infection caused by SARS-CoV-2. The highly contagious nature and exponential spread of SARS-CoV-2, along with its potential for rapid progression to ARDS, has burdened health-care systems worldwide, which contributes to high mortality rates.1-4 The initial approach to respiratory support for severe COVID-19 pneumonia was based on invasive mechanical ventilation. This measure has some detrimental effects for many patients due to the extended hospital stay, high rates of health-care–related infections, and ventilator-induced lung injury, among others. Thus, it is essential to evaluate other strategies to improve oxygenation in patients with hypoxemic respiratory failure who require ventilatory support.1-4 The high-flow nasal cannula (HFNC) is composed of an oxygen source and a flow generator that mixes the gases, a single heated circuit, and a nasal interface. It delivers adequately heated and humidified medical gas at flows up to 60 L/min and 
from 0.21 to 1.0.4,5 HFNC is considered to have several physiologic benefits, including reducing anatomic dead space and ventilatory requirements, providing a constant 
with adequate humidification, and a degree of PEEP up to 7 cm H2O.2,5-8
After the trend of increasing COVID-19 cases in Brazil since January 2021, there was a rise in the number of patients with severe hypoxemia admitted to the respiratory wards associated with a reduction in the availability of ICU beds during the second wave of the pandemic in our region. With reports that show that the use of HFNC reduces the need for ICU care in patients with hypoxemic respiratory failure when compared with conventional oxygen therapy and, therefore, increases the ability to manage patients with severe respiratory distress secondary to COVID-19,8-10 ventilation systems that use the HFNC were purchased for use in the respiratory wards. Thus, the objective of this study was to analyze HFNC therapy in the treatment of patients with COVID-19 pneumonia. The primary outcome was the need for invasive mechanical ventilation. The secondary outcome was to examine factors associated with an improved clinical response to this therapy.
Quick Look
Current Knowledge
High-flow nasal cannula (HFNC) therapy has several physiologic benefits and has been an alternative in treating hypoxemic respiratory failure. The use of HFNC may reduce the need for invasive mechanical ventilation and may provide improvements in oxygenation levels compared with standard oxygen therapy.
What This Paper Contributes to Our Knowledge
HFNC allowed noninvasive respiratory support of subjects with COVID-19 on the wards and reduced the need for invasive ventilation by 50%. Subjects requiring invasive ventilation had a greater incidence of comorbidities and required increasing oxygen. Those subjects with progressive disease and the need for invasive ventilation had a significant increase in mortality.
Methods
Study Design and Setting Description
A single-center prospective observational study was conducted in 3 respiratory wards of the Complexo Hospital de Clínicas, Universidade Federal do Paraná (CHC/UFPR), a tertiary-care hospital and an academic health center in Curitiba, southern Brazil. CHC/UFPR has been one of the referral hospitals for patients with COVID-19 from the public health network in Curitiba and the metropolitan region. Approximately 100 ward beds and 88 ICU beds were available to provide care to these patients, with increasing occupancy rates since the beginning of the pandemic. With regard to human resources, the permanent staff members were relocated to COVID-19 areas and temporary staff members were hired to expand assistance: physicians, nurses, nursing technician, physiotherapists, and social workers. These professionals worked exclusively on the front line to care for patients with COVID-19. In the wards, we had a team that involved physicians from different specialties: general practitioners, infectious disease specialists, pulmonologists, and surgeons. In addition, our work schedule included the following: 1 nurse to every 7 patients; 1 nursing technician to every 3 or 4 patients, and 1 physiotherapist to every 7 patients. This health-care team used bedside patient monitoring systems. The institutional research and ethics committee of CHC/UFPR approved the study, and the informed consent signature was waived (45476221.2.0000.0096).
Subject Selection and Follow-up
This study enrolled subjects hospitalized in the respiratory wards of the CHC/UFPR with a diagnosis of COVID-19 confirmed by real time-polymerase chain reaction. Patients who exhibited signs of hypoxemic respiratory failure when using non-rebreathing masks with an O2 reservoir and a progressive need for oxygen supply > 60% were eligible for HFNC therapy and were eligible to participate in the study. The following parameters were considered to initiate the HFNC therapy: (1) the patient is breathing spontaneously; (2) age >18 years; (3) alert and aware; (4) with airway protection (effective cough); (5) dyspnea (Borg scale score > 4 [moderate, moderately severe to severe dyspnea] and Medical Research Council grade 4 [dyspnea during usual activities, such as bathing and changing clothes])11; (6) 
/
< 300 mm Hg and > 200 mm Hg; and (7) breathing frequency > 25 breaths/min and heart rate > 120 beats/min. All the subjects who met the above criteria underwent HFNC therapy and were included in the analysis.
Patients with a clinical and laboratory need for immediate invasive mechanical ventilation after admission to the ward and those with severe anxiety who were unable to cooperate with HFNC therapy delivery were excluded from the study. HFNC oxygen therapy was delivered by using an AIRVO 2 respiratory humidifier (Fisher & Paykel Healthcare, Auckland, New Zealand). The handling of the equipment was conducted by a previously trained professional (VCN). The subjects were monitored as follows: the ROX (respiratory frequency – oxygenation) index was evaluated sequentially within the first 24 h of the therapy (in 2, 6, 12, and 24 h), and every 24 h afterward.
The oxygen need, breathing frequency, and additional data were recorded in a specific form, as briefly described below: (1) the initial settings were temperature, 31°C; flow, 40 L/min; 
, 0.8; (2) reassessment of the comfort level and the breathing frequency after 30 min (ideally keeping frequency ≤ 30 breaths/min); (3) recalculation of the ROX index in 2, 6, 12, and 24 h, and daily according to protocol; (4) on clinical improvement (progressive decrease in the volume of oxygen needed to maintain an oxygen saturation between 92% and 95%), progressive reduction of 
until < 0.4 first, followed by a reduction of flow; and (5) return to a low-flow nasal catheter when 
≤ 0.4.
The ROX index is calculated by the ratio of peripheral oxygen saturation and 
to the breathing frequency. It was first described in a cohort study that involved subjects with pneumonia and/or ARDS admitted to the ICU and treated with HFNC. This index has been applied as a predictor of successful HFNC weaning and as an indicator to properly titrate 
and set the optimal flows in patients with acute respiratory failure treated with HFNC.9,12,13 In addition to the above-mentioned HFNC therapy control parameters, demographic and clinical data were retrieved from the subjects’ medical records: (1) patient identification, age, and sex; (2) date of the symptoms onset; (3) the presence of signs and symptoms of acute respiratory infection, including cough, runny nose, respiratory effort, pulmonary auscultation; (4) date of hospitalization; (5) Charlson comorbidity score; (6) ventilatory support needed; (7) oxygen saturation < 95%; (8) ICU care needed; and (9) pharmacologic treatment: antibiotics, corticosteroids, and prophylactic anticoagulant.
The Charlson comorbidity score is a standardized method proposed in 1987 to assess comorbidities, originally developed to predict 1-year all-cause mortality based on 17 underlying conditions. It has been validated in many clinical settings, and it is the most widely used comorbidity index. Due to advances in treatment and disease management, updated versions have been adapted, validated, and reported in different databases.14 Also, follow-ups were performed on the subjects until hospital discharge or death. No subjects were lost to follow-up. The outcomes assessed were the following: (1) primary, need for invasive ventilatory support; and (2) secondary, transfer to the ICU, death, or hospital discharge. Patients who required invasive mechanical ventilation or who required vasoactive drugs due to hemodynamic instability or who had a reduced level of consciousness were referred to the ICU.
While conducting this study, no antiviral drug or specific monoclonal antibody with anti-SARS-CoV-2 activity had been approved for use by ANVISA (Sanitary Surveillance Agency) in Brazil; therefore, all hospitalized patients with COVID-19 in the units were treated by using the standard of care approach, including symptomatic medication, prophylactic anticoagulants, and oxygen therapy. The only adjunctive treatment used was dexamethasone 6 mg (orally or intravenously, as available) for 10 d (for patients who required oxygen support or mechanical ventilation).
Statistical Analysis
A descriptive analysis of the subjects’ profiles included in this study and the follow-up results were presented. For the analysis of categorical variables, absolute and relative frequencies were obtained, and, for quantitative variables, the median (interquartile range) and average with the standard deviation were calculated. The chi-square test or the Fisher exact test was used to analyze differences between the groups, whereas the Mann-Whitney U test was used for continuous variables when appropriate. A logistic regression model was fitted to investigate the relationship between the outcome (success or failure) and potential explanatory variables. The stepwise forward procedure was performed with the explanatory variables that presented P < .2 in the bivariate analysis. The estimated measure of association for the outcome was the odds ratio with a corresponding 95% CI. The discriminative power of the ROX index value was assessed to identify the subjects with a higher probability of HFNC therapy failing by calculating the receiver operating characteristic area under the curve, sensitivity, and specificity. The ROX index mean values obtained between the different times were considered.
An area under the curve > 0.7 was considered useful, whereas an area under the curve between 0.8 and 0.9 indicated good diagnostic accuracy. To investigate the change in the ROX index over time, the Gaussian copula marginal regression model15 was used for the longitudinal data analysis. The gamma distribution was assumed for the response variable (ROX index), logarithmic link function for the linear predictor, and the autoregressive structure of order 1 was used for the correlation matrix. In this analysis, there were 2 explanatory variables: (1) HFNC response, which is a factor with 2 levels (success and failure), and (2) the evaluation time, which was treated as a factor with 7 levels (pre-HFNC, 2 h, 6 h, 24 h, 48 h, 72 h, and 96 h). The linear predictor comprised both the main and the interaction effects between time and the HFNC response (success or failure). Also, a multiple comparison test was conducted, in which the P values were obtained through the Bonferroni correction.16 The statistical data analysis was performed by using the R software version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) with the gcmr package.17
Results
From March to August 2021, a total of 1,487 patients were admitted to the respiratory wards; of these, 130 patients (8.7%) presented with criteria for ventilatory support with HFNC, and 128 subjects were included in the study. Two subjects were excluded because data were not collected in all periods determined for the calculation of the ROX index. Sixty-eight (68/128 [53%]) had success when using the HFNC therapy. Observing the distribution of success and failure cases by month, it was noticeable that the proportion of success increased in the subsequent months (Fig. 1). When comparing the subjects in whom HFNC therapy failed with those in whom HFNC therapy succeeded, a homogenous distribution among the groups was observed in relation to some demographic and clinical data, mainly the presence of comorbidities and the Charlson comorbidity score (Table 1).
Distribution of high-flow nasal cannula success and failure in study population. Points show the total number of subjects for each month.
Clinical and Laboratory Findings Among Subjects With HFNC Success and Failure (N = 128)
The weighted median of the subjects in whom the therapy succeeded was significantly higher than those in whom the therapy failed. In addition, the duration of illness (onset of symptoms until hospitalization) was similar between both groups. With regard to the respiratory and oxygen need before HFNC, only oxygen saturation was not a response surrogate. The need for ICU care was high in both groups, but the mortality rate was significantly higher among the subjects in whom the therapy failed versus those in whom the therapy succeeded (47% vs 3%). Variables that presented P < .2 were stepwise forward selected for the multivariate logistic regression analysis. The final regression model contained the following explanatory variables: length of hospitalization, length of HFNC therapy, oxygen flow, breathing frequency, and Charlson comorbidity score, as shown in Table 1.
The ROX index seemed to be an important predictive response factor, and its progression in the first hours can be helpful to determine those who may benefit from HFNC therapy. In this study, in 48% and 86% of the subjects who did not respond to HFNC therapy, the therapy failed within 24 and 48 h after the therapy onset, respectively. However, in multivariate analysis, the ROX index was not statistically significant, probably due to multicollinearity, because the value is calculated based on other explanatory variables: oxygen saturation, 
, and breathing frequency. An evaluation was also conducted by applying a Gaussian copula marginal regression model to the longitudinal data of the ROX index, which compared both groups of subjects (Fig. 2). The ROX index of the between-group evaluation was consistently and significantly lower among the subjects in whom the HFNC therapy failed compared with those in whom the HFNC therapy succeeded. In the within-group analysis, there was a significant difference in the mean ROX index values but only up to the first 24 h (Supplementary Table 1 and Supplementary Fig. 1 [see the supplementary materials at http://www.rcjournal.com]).
ROX index mean (±SE) progression during HFNC therapy, N = 128. SE = standard error.
In addition, when considering the importance of the early definition of the subjects who were more likely to respond adequately to HFNC therapy and, mainly aiming not to delay invasive mechanical ventilation, we assessed the receiver operating characteristic curve generated to evaluate a ROX index cutoff value to identify failing probability among the included subjects. The receiver operating characteristic curves that presented better results corresponded to the ROX index that ranged from 3.47 to 4.53 obtained in pre-HFNC and the first 24 h, respectively, are shown in Figure 3. Both had an area under the curve of 0.73, which is helpful in discriminating subjects in whom therapy failed in comparison with those in whom the HFNC therapy succeeded.
Receiver operating characteristic curves of ROX (respiratory frequency – oxygenation) index values obtained A: pre high-flow nasal cannula (HFNC); B: 2 h after HFNC initiation; C: 6 h after HFNC initiation; and D: 24 h after HFNC initiation.
Discussion
This prospective observational study showed that HFNC could successfully provide respiratory support for the subjects with severe COVID-19 respiratory failure. All the subjects received HFNC in a non-critical care ward–based environment, which demonstrated the feasibility of HFNC outside of the ICU. The assessment of the applicability of this intervention in emergency care units and wards is essential at this pandemic time due to the limited number of ICU beds to face this public health emergency. Overall, 53% of the subjects responded favorably to this therapy. Although there is a concern with regard to the use of HFNC therapy in patients with COVID-19, the evidence of the efficacy of HFNC in reducing the requirement for intubation and invasive mechanical ventilation is consistent with several observational studies.18,19
However, in this study, HFNC therapy failed in 47% of the subjects who subsequently needed invasive mechanical ventilation, and the mortality rate in this group of subjects was high (the mortality was 47% and 3% in the group that failed or succeeded with HFNC therapy, respectively). Zhou et al,20 in a retrospective study, showed that the rate of non-survivors was significantly higher among those who used HFNC (the rate of non-survivors and survivors was 61% and 6% among those who used or did not use CNAF, respectively). Similarly, Wang et al21 reported that 41% of the subjects experienced HFNC treatment failure, which was directly associated with 
/
≤ 200mm Hg. Thus, training health professionals to use this modality and defining cutoff points that indicate the risk of failure for this therapy is essential to avoid delays in ICU referral or immediate invasive mechanical ventilation.8,13,21
When comparing both groups in the study, success and failure, a homogeneous distribution with regard to age, body mass index, comorbidities, and disease length, factors that could influence the outcome, were observed. Interestingly, it was found that the subjects who successfully underwent therapy had a higher body mass index, a risk factor for a worse COVID-19 outcome. However, the benefits of HFNC therapy in patients with obesity had already been reported.7 With regard to the Charlson comorbidity score, it was observed only in the logistic regression model that subjects with a higher score had a slightly higher risk of failure (odds ratio 1.5, 95% CI 1.06-2.44). In our experience, the criteria to start HFNC therapy was primarily the non-response to using a non-rebreathing mask with an O2 reservoir, which requires an oxygen supply > 60%. The physiotherapist and the assistant physician shared the decision, and both were also responsible for data collection to calculate the ROX index and recommend mechanical ventilation. This joint action provided greater security to the health-care group and contributed to a progressive increase in success rates.
Most reports of HFNC use come from retrospective studies or case series, which possibly contributes to the discrepancy in results with concern to the benefits of this therapy. In our analysis, we observed that the evaluation of the ROX index in a longitudinal design allowed the identification of subjects who were more likely to respond to this therapy.10,13,18,19,22 The evolutionary dynamics of the ROX index is of particular importance and should be calculated regularly to evaluate the patient’s improvement (score increasing) or deterioration (score stable or decreasing) to not miss the exact moment when the patient may need to be intubated.12,13,23 In both the success and the failure groups, longitudinal evaluation of the ROX index showed the lowest ROX index values in individuals in whom the therapy failed. This difference was significant in all the measurements, but, by observing the curve shown in Figure 2, after 24 h, there was a substantial divergence of these values. In the within-group longitudinal evaluation, the ROX values of the group in whom the therapy failed were significantly different in all measurements in the first 24 h.
By using the mean values of the ROX index (pre-HFNC, 2 h, 6 h, and 24 h), we constructed receiver operating characteristic curves for each of these times. The cutoff values obtained for the ROX index were between 3.97 and 4.98, with an area under the curve of ∼0.7, which demonstrated that such values can be used to define patients at risk of failure and, therefore, have an indication for invasive mechanical ventilation. It is noteworthy that the sensitivity and specificity of these values were approximately 50% and 85%, respectively. Ricard et al,24 showed that HFNC therapy in some patients with servere actute hypoxemic respiratory failure improves several parameters of respiratory function. Such beneficial effects can prevent intubation and have a good outcome. The investigators found that an ROX index < 5.37 in the first 4 h after starting HFNC was an indication for subsequent intubation in subjects at high risk.24 In addition, previous studies that included subjects with COVID-19 and subjects without COVID-19 showed that an ROX index > 4.88 indicated success of HFNC treatment and a low risk of intubation.23,24 Patients with values between 3.85 and 4.87 should be closely monitored for the need of intubation, between 2.85 and 3.84 should be monitored in the ICU due to highly increased risk of intubation, and for an ROX index < 2.85, intubation should be considered.23,25,26
To date, it has not been proven that early initiation of therapy with HFNC has better results.27 It is considered that early-onset are cases in which patients present 
/
< 300 mm Hg or 
< 93% with O2 > 5 L/min or 
< 94% with 
of 0.4, whereas the late-onset would be those with 
< 92% with O2 at 15 L/min and/or 
/
< 150 mm Hg. The use of this therapy in our service occurred at a critical moment of the pandemic when we had a high number of patients with serious conditions and with restrictions on HFNC equipment and ICU bed availability. Thus, we gradually evaluated oxygen therapy, which culminated in the late start of this therapy. The preliminary studies showed that early use might have better results, which is suggested to be evaluated in prospective studies.10,27,28 The generation of contaminated droplets and aerosol during HFNC therapy is significant and has been considered an occupational risk. HFNC use is recommended in isolation rooms and with negative pressure, which was unfeasible at this time of the pandemic.29 In our service, the health-care team was advised to wear personal protective equipment, such as N95 masks, eye protection, and specific clothing for aerosol protection. In addition, patients were also advised to wear surgical masks to reduce the emission of droplets and aerosols, which contributed to improve the patients’ oxygenation.
This study had some limitations: (1) in the initial phase of the study, the team was trained to use the HFNC equipment at a time of severe lack of beds for patients who were critically ill and with limited availability of mechanical ventilators; due to a lack of a clear cutoff value for the ROX index, some subjects remained on HFNC therapy for a longer period as an alternative, which may have contributed to delaying intubation; (2) sample-size estimation parameters were not calculated, and there was a low percentage of subjects included (<10%) because many eligible subjects were unable to use this therapy due to a shortage of equipment; (3) due to the high demand of the care team in the wards as a result of the pandemic, some clinical characteristics of included participants were missing; and (4) it was a single-center experience, therefore, the external validity of the results may be affected.
However, the study’s strengths were the training of the entire health-care team in the management of these cases and the shared decision making as well as the prospective data collection, which allowed the application of this therapy to patients in the wards. We were rapidly able to safely identify patients at high risk for early referral to the ICU and potentially contributed to favorable outcomes. Further, it should be noted that the success of HFNC therapy increased in the subsequent months, which was undoubtedly associated with greater staff training and more precise indications of the HFNC usage. It has been shown that noninvasive respiratory support modalities may cause intubation delay and may be associated with a longer hospital length of stay and high morbidity and mortality rates. However, in this study, we observed that, when correctly indicated and with a follow-up performed by a trained team for this therapy, the response can be excellent, which contributed to reduction in ICU mortality rate and patients’ exposure to complications that arise from invasive mechanical ventilation.
Conclusions
HFNC is an alternative in treating patients with signs of hypoxemic respiratory failure, and its benefits include reduced need for invasive ventilation, patient comfort and ease of application. Nevertheless, it is important to define non-response parameters for HFNC therapy, particularly to not delay the recommendation for orotracheal intubation, thus reducing the risk of severe hypoxemia and its consequences. HFNC correctly prescribed and followed up by a trained health-care team can contribute to a reduction in ICU mortality and patients’ exposure to complications from invasive mechanical ventilation.
Acknowledgments
The authors thank the team of health-care professionals who helped the patients with COVID-19: nurses, laboratory specialists, nutritionists, occupational therapists, speech therapists, and psychologists. Their commitment and professionalism have profoundly contributed to the care of the patients affected by this new disease.
Footnotes
- Correspondence: Sonia M Raboni MD, Infectious Diseases Division, Complexo Hospital de Clínicas, Universidade Federal do Paraná, Rua General Carneiro, 180, 3o. andar, Curitiba, Paraná, Brazil 80060–900. E-mail: sraboni{at}ufpr.br
Drs Raboni and Neves contributed equally to the manuscript.
Dr Raboni receives a CNPq Research Productivity Scholarship.
The authors have disclosed no conflicts of interest.
Supplementary material related to this paper is available at http://www.rcjournal.com.
- Copyright © 2022 by Daedalus Enterprises
