Introduction

Acute respiratory failure (ARF) is a life-threatening condition frequently found in patients admitted to the emergency department (ED). ARF is defined as the inability to maintain normal haematosis. Many underlying diseases can cause ARF (such as pneumonia or heart failure).

ARF can be either hypercapnic, hypoxaemic or both.

ARF requires rapid identification and prompt implementation of resuscitation measures. ARF’s first treatment is oxygen therapy. Its purpose is to alleviate respiratory failure and to restore a satisfactory haematosis. Most experts stress the importance of having a Oxygen saturation (SpO2) target of more than 90% in the majority of patients.1 While non-invasive ventilation (NIV) is the standard of care in the initial management of hypercapnic acidosis,2 recommendations for oxygen therapy in acute hypoxic respiratory failure in the ED are currently lacking.

The choice of oxygen delivery device depends on the severity of the hypoxaemia, underlying physiology, dyspnoea type and patient device tolerance. The most commonly used devices are nasal probes, face masks and non-rebreather face masks. They offer some oxygen inspired fraction (fraction of O2 in the gas mixture breathed, FiO2) depending on the oxygen flow rate instituted (with a maximum of 15 L/min).

However, conventional oxygen devices may be insufficient because of the uncontrolled FiO2 and the discrepancy between the volume of oxygen administered and the patient’s needs, needs that can fluctuate from 30 to 100 L/min.3

High-flow nasal oxygen (HFNO) is increasingly used in EDs for hypoxaemic ARF according to national and international recommendations.4 These recommendations are still weak and are based mainly on results obtained outside the ED.5

With HFNO, a much higher flow of oxygen, up to 60–70 L/min, can be administered. This allows to generate low levels of positive pressure in the upper airways and to adapt the FiO2 delivered up to 100%.6 Its use was initially developed in paediatric and neonatal intensive care units (ICUs).2 6 It was then gradually extended to adult ICU in the treatment of hypoxaemic, non-hypercapnic ARF with no indication for orotracheal intubation.3

Most of these studies were conducted in intensive care and resuscitation units. They underline HFNO’s superiority compared with conventional oxygen therapy (COT) in improving SpO2 and in decreasing dyspnoea and dry mouth, respiratory frequency and improving biological parameters.7–10 However, despite the increasing number of studies, the potential advantages of HFNO regarding the need for therapeutic escalation and mortality have not been precisely evaluated.6 11 12 Although the majority of these studies advocate the use of HFNO, almost all are monocentric, retrospective, with very small study populations, heterogeneous methodologies and with disparate use of HFNO.

Messika et al retrospectively analysed data from patients placed on HFNO in their resuscitation unit.13 87 patients benefited from this support (including 51 first-line patients). HFNO was used in 45 subjects with acute respiratory distress syndrome (ARDS), of whom only 40% required secondary intubation. The authors recommended the use of HFNO as a first-line treatment for patients presenting with ARF, particularly for those with ARDS. In their FLORALI study published in 2015, Frat et al compared HFNO to NIV and COT in hypoxaemic patients with an arterial oxygen tension/fractional inspired oxygen (PaO2/FiO2) ratio of less than 300.6 The rate of intubation was similar in the three groups, mortality at 90 days was significantly lower in the HFNO group. While this study reinforced the indication for HFNO in hypoxaemic patients managed in ICU, many questions remain unanswered: the small number of patients included and heterogeneity in patients’ comorbidities as much as the treatments received, notably antibiotics in patients with hypoxaemic pneumonia, do not allow for a firm recommendation for the use of HFNO.

Two meta-analyses compared HFNO and COT in immunocompromised patients.14 15 Both agreed that HFNO was associated with less intubation. However, they differed in terms of mortality, with Huang et al reporting a benefit with HFNO while Cortegiani et al found none. Both teams concluded that there was a need for randomised controlled trials.

EDs are in the first line for patients with hypoxaemic ARF and where oxygenation is quickly initiated. The management of these particular patients requires significant staff resources and time. Therefore, that pathology has a major impact on the department’s operation and the care provided to all its users. Few studies were conducted to evaluate the HFNO specifically in the ED.11 16–21 They suffer from the same limitations as those conducted in ICUs and many essential questions remain unanswered regarding the potential benefits of HFNO over COT and NIV in terms of intubation requirement, treatment failure, hospitalisation or mortality.22 23

For example, Lenglet et al showed in an ED, on a series of 17 patients oxygenated by HFNO, a clinical improvement 15 min after replacement of the non-rebreather oxygen mask initially used and a significant increase in SpO2.16 However, the prospective randomised trial HOT-ER, conducted in 303 patients treated in the ED with HFNO or COT, failed to demonstrate a statistical difference between the two treatments in terms of mortality or escalation to mechanical ventilation. However, neither arterial blood gas (ABG) values nor the PaO2/FIO2 ratio have been presented, and one weakness of this study would probably lie in the recruitment of patients with a low severity pattern.18

In summary, all these studies show a benefit of HFNO in hypoxic ARF. They are of different methodologies and often of limited statistical power. Up until now, we do not know the optimal conditions for the use of HFNO. In particular, the optimal time for oxygenation is not known. Whereas it has been shown that the longer we delay intubation for those patients, the higher the mortality rate is.24 The average time would usually be more than 90 min.16 18

Our purpose is to evaluate the contribution of early administered HFNO for patients with acute non-hypercapnic respiratory failure presenting to the ED, with the aim of obtaining rapid correction of haematosis. The objective is to compare COT delivered by nasal probes or masks at flow rates up to 15 L/min to HFNO, with the hypothesis that HFNO would reduce the need for ventilatory therapy escalation. The other hypotheses concern the interest of HFNO in reducing the use of intensive care admission and thus the healthcare costs.

Materials and analysisDesign

We designed HIFLOWED (High Flow Nasal Oxygen Therapy Started in the Emergency Room Versus Conventional Oxygen Therapy in Patients With Acute Hypoxaemic Respiratory Distress study) as a multicentric, prospective, open and randomised superiority study. HIFLOWED is recruiting with the first enrolment on 23 October 2020. However, the COVID-19 pandemic has had a great impact on starting and deploying this study in the participating centres.

Study objectives and outcomes

The primary objective is to prove the efficacy of HFNO therapy initiated within the first 60 min after the ED’s admission (HFNO arm) compared with COT (CO arm) in the initial management of patients with acute hypoxaemic respiratory failure.

The primary outcome is a failure in the oxygen therapy. It is defined as the need for a therapeutic escalation within 4 hours after therapy initiation. In the COT group, therapeutic escalations are the switch to HFNO or to non-invasive or mechanical ventilation. In the HFNO group, failure is the switch to non-invasive or mechanical ventilation.

Considering the emergency settings, the identification of deterioration or lack of improvement in respiratory status and, therefore, the potential switch to another therapy (the primary endpoint of the study) is left to the discretion and decision of the clinician.

The secondary objectives are

To identify possible early prognostic factors for HFNO failure, defined as the need for a therapeutic escalation within 1 hour after its initiation, such as the persistence of clinical signs of acute respiratory distress (respiratory rate, SpO2, pulling, dyspnoea score, laboured breathing) and/or persistence of PaO2/FIO2 ratio below 300 mm Hg.

To assess the effectiveness of HFNO on patient outcomes: total time with HFNO, ED length of stay (LOS), hospital LOS, the ward where the patient is admitted and 30-day mortality.

To describe the aetiology of dyspnoea.

Secondary outcomes are

Possible early prognostic factors for HFNO failure will be investigated based on the following criteria, which will be assessed 60 min after initiation of oxygen therapy:

Clinical criteria (respiratory rate, SpO2, signs of respiratory failure (pulling, laboured breathing), dyspnoea score measured by the modified Borg scale.25

Paraclinical criteria: PaO2/FIO2 ratio <300 mm Hg.

To assess the effectiveness of HFNO on patient outcomes, we will use:

The total time on oxygen therapy is defined as the time from the start of its initial application to its discontinuation or change to another oxygen therapy device.

The ED LOS is defined as the time between the administrative admission at the ED and the notified time in the patient’s chart for the transfer from the ED.

Hospitalisation wards and LOS are defined by the place and number of days of hospitalisation notified on the patient’s hospital records.

Outcome at 30 days defined by the patient’s status: deceased, still hospitalised or discharged from the hospital.

ARF can have multiple origins that we will identify for each patient: impairment of pulmonary neuromuscular function (eg, Guillain-Barré syndrome, acquired myasthenia), secondary conditions related to chronic or acute pulmonary obstructive disease (eg, asthma, chronic obstructive pulmonary disease), alveolar processes such as cardiogenic or non-cardiogenic pulmonary oedema, pneumopathy and vascular diseases like acute or chronic pulmonary embolism.

Randomisation

Eligible patients are randomised through an eCRF (cleanweb) to either the HFNO arm or the conventional oxygenation arm (1:1 ratio with a block size of 4 and stratification according to site). HFNO is administered at a flow of 50 L/min with a baseline FiO2 of 100%.

In France, physicians are used to giving the maximum oxygen to hypoxic patients and quickly deescalate according to SpO2. In order to standardise our protocol, we kept the same approach. However, the attending physician has the discretion to adjust the flow rate as necessary.

In the CO arm, the method of CO is at the discretion of the investigator in charge of the patient. CO is defined as the use of any device (nasal probes, masks or non-rebreather masks) other than HFNO.

Randomisation is performed during the first 60 min following the patient’s admission to the ED. The randomising investigator will get access to the internet site through a personal password.

After randomisation, the clinician in charge of the patient will be totally free to adapt the management.

Patient’s selection

Consecutive patients aged 18 years or older admitted to the ED with ARF are eligible to participate in the study. Inclusion and non-inclusion criteria are presented in box 1. Patients will be recruited from 10 hospitals, 8 teaching hospitals (hospitals of Besançon, Dijon, Poitiers, Lariboisière, Montpelier, Clermont-Ferrand, Toulouse and Rennes) and 2 general hospitals (Vesoul and Chartres). Before starting trial participation, written informed consent will be obtained by the clinician from all participants. Depending on the patient’s consciousness status, the clinician can adopt an emergency consent procedure. In that case, the patient will be included, and consent will be asked as soon as possible.

Box 1 Eligibility criteria for participants in the HIFLOWED (High Flow Nasal Oxygen Therapy Started in the Emergency Room Versus Conventional Oxygen Therapy in Patients With Acute Hypoxaemic Respiratory Distress) studyNon-inclusion criteria:

Hypercapnic patients (PaCO2>45 mm Hg) with respiratory acidosis (pH<7.30).

Indication for non-invasive ventilation or early invasive mechanical ventilation according to current scientific recommendations (acute pulmonary oedema or others).

Dyspnoea of traumatic origin.

Traumatic pneumothorax.

Haemodynamic instability (mean arterial pressure <65 mm Hg).

Patients managed by Emergency medical services (Service Mobile d'Urgence et Réanimation–Mobile Emergency and Resuscitation Unit) who have already received cardiac or pulmonary treatment.

Patients with cognitive deterioration (Glasgow score less than 13, dementia or mental failure that would prevent good cooperation).

Patients with ear nose and throat lesions contraindicated for the implementation of high-flow nasal oxygen therapy.

Patients unlikely to cooperate in the study and/or low cooperation is anticipated by the investigator.

Patients without health insurance.

Pregnant women.

Patients being in the exclusion period of another study or included in the ‘national volunteer file’.

PaO2/FIO2, arterial oxygen tension/fractional inspired oxygen; PaCO2, arterial carbon dioxide tension; SpO2, oxygen saturation.

InterventionsDescription of the products used

HNFO is delivered with the usual device of each participating team. It is used in accordance with standard practice and current knowledge. A different set of interfaces (probes cannula, tubing and humidifier) packaged by the manufacturer in individual packs will be used for each person. HFNO is set up at a fixed flow rate of 50 L/min with a variable FiO2 depending on the defined SpO2 target ie, ≥95%).

The most commonly used devices for conventional oxygenation are the nasal probes, the face mask and the non-rebreather face mask. They offer different FiO2 depending on the oxygen flow rate set.

Recruitment and data collection

We informed about the trial and involved key medical personnel in the different ED participating in this study.

At baseline, patients are managed as usual (clinical examination…). The physician in charge gives an information note, presents the study and invites eligible patients to participate in the study. The randomisation process, done by the physician in charge, occurs once the patient has signed the informed consent.

The oxygen therapy randomised (HFNO or CO) is then initiated within 30–60 min following the admission (T0).

Patients are managed as usual. We collect the data describing the patient’s evolution at T0+60 min, T0+4 hours and T0+24 hours (oxygen-therapy status (device and flow) as well as ABGs tests, clinical laboratory tests and Borg score (if required by the patients).

30 days after admission, we collect the patients’ status (hospitalised, discharged or deceased) (table 1).

Safety issues

No major risks are expected to be induced with HNFO.

Adverse events

All adverse events occurring since the signing of the informed consent are reported by the investigator according to the regulatory reporting procedures established in each centre.

Data management and monitoring

In each centre, research assistants collect the data. All data are entered in pseudonymised form into an eCRF that is periodically checked for missing values and errors. Monitoring, including data management and data quality assurance, will be conducted by the study sponsor. The scheduling of visits will be determined by the study sponsor. The sponsor will assume responsibility for monitoring the safety of all study participants. Paper documents are collected in a study file archiving and supervised by the principal study investigator. The final dataset will be password protected and accessible only to the study personnel directly involved in the data analysis. Depending on the missing data, a procedure for replacing missing data will be considered. The archiving will be the responsibility of the sponsor in accordance with recommendations, current regulations or the agreement signed between the partners, for the longest archiving period covered by these texts.

Statistical analysis

The analysis of this clinical superiority trial will be conducted on an intention-to-treat basis. Quantitative variables will be described using the following parameters: mean and SD, median, minimum and maximum values. Qualitative variables will be described by the frequency and proportion of each class. Categorical variables will be compared by a χ2 test or Fisher’s exact test. Quantitative variables will be compared by a Student’s t-test or a Mann and Whitney test. The comparability of the two arms at inclusion will be tested. In case of non-comparability of certain factors, an adjustment on these factors could be implemented in the multivariable analysis. An exploratory analysis will be conducted on different subgroups of patients. If the number of events and the number of cofactors in the model allows it, an iterative/sensitivity analysis will be conducted. Additionally, we will assess for potential centre effects using a random effect logistic model and will account for it if necessary.26

Granting the attending physician discretion in the decision to escalate treatment (main endpoint) could lead to rely on the main endpoint of the physician decision. Considering the open-label status of the trial, potential information bias could occur and would be discussed. We will assess the conditions of the switch in both arms and compare the conditions of failure or non-failure including ABG and vital signs. The proportion of subjects requiring ventilatory therapy escalation in the first 4 hours of oxygen therapy will be studied using a multivariate logistic regression model. A similar approach will be conducted on prognostic factors for failure and patient outcome (secondary endpoints).

Based on an unpublished retrospective pilot study led at the University Hospital of Besançon in 2018, we estimated the therapeutic escalation rate at 20%. In the HFNO arm, a 50% reduction in this proportion is expected.12 Using a one-sided test with an alpha of 4.9% (an intermediate analysis is planned, threshold according to O'Brian and Fleming) and a power of 90%, the number of subjects required is 218 per arm, that is, 436 subjects in total. The number of subjects recruited will be 500 to accommodate lost to follow-up and possible withdrawal of consent.

The significance level of the statistical tests is set at 5% and the formulation of the hypotheses is one sided.

The rate of inclusion was calculated to be well below the number of eligible patients. This estimation was based on feasibility questionnaires completed by the centres and an observational study conducted in the Besançon ED. Additionally, we designed the study to align as closely as possible with current practice, with the only specific tasks added being the identification of eligible patients and randomisation, aimed at limiting additional workload for caregivers. These measures seem adequate to achieve participant enrolment and reach the target sample size.

Patient and public involvement

No patients were involved (patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research).

Ethics and dissemination

This trial will be conducted according to good clinical research practice and the latest Declaration of Helsinki.27

The study has been submitted and approved by the Comité de Protection des Personnes Nord Ouest IV (20 October 2020). As required, a notification was sent to the Agence nationale de sécurité du médicament et des produits de santé (22 October 2020).

Any important protocol modification will be communicated to all relevant parties by the sponsor.

Signed informed consent is obtained by the physician from all participating patients. Patient anonymity is protected by the use of subject identification codes.

Participants can withdraw their consent at any time point for any reason and will be invited to still participate in the follow-up assessments.

Trial results will be submitted in peer-reviewed journals and international congresses. Authorship will be granted according to the recommendations from the International Committee of Medical Journal Editors.28

Stopping guidelines

Any participant who wishes to leave the study is free to do so at any time. The investigator may also remove a subject from the study for safety, protocol deviations or administrative reasons.

The total discontinuation of the study will be in agreement with the ethics committee.

Discussion

Robust studies demonstrating the effectiveness of HFNO in non-hypercapnic hypoxaemic ARF exist within ICUs. However, such evidence is lacking in the ED, where oxygen treatments are frequently initiated. To address this gap, we initiated the HIFLOWED study. Conducting research within the ED’s constraints is, however, a challenge. In order to be feasible, the study-specific procedures have been reduced to a minimum. They concern patients’ identification, informed consent and the randomisation within the first hour of admission. ABG is performed according to the usual practices, that is, at least once at patient’s arrival and as needed based on their condition. This approach minimises the frequency of potentially painful procedures while maintaining the integrity of our analyses. Physicians will be free to decide the best management for their patients.

A blinded study was hardly conceivable. In HIFLOWED, patients and physicians are not blinded to the intervention. However, given that our criteria are quite objective, we believe this will have a limited impact on our results.

There will be many aetiologies for ARF in our study population. They have in common that they require oxygen administration that can be delivered by means of CO or HFNO. In order to prevent the common bias of a too heterogeneous study population, we have chosen not to include patients with a suspicion of acute heart failure and/or acute cardiogenic pulmonary oedema if they require NIV.

HIFLOWED have several strengths. Participants will be recruited from several teaching and general hospitals and none of them is new to HFNO. The multicentric design will increase the external validity.

The primary outcome, therapeutic escalation 4 hours post-ED admission, is simple and adapted to ARF that often requires rapid therapeutic adaptations. Patients are also followed at medium (24 hours) and long term (30 days).

We believe that our study will provide innovative and original insights to address the question of the usefulness and the necessity of early HFNO initiation in the ED for non-hypercapnic hypoxaemic failure patients. We believe that, if the superiority of HFNO is demonstrated in the emergency settings with a high number of patients, a change in practice can be expected. HFNO could be the first-line emergency treatment instead of COT for hypoxic ARF, whatever the aetiology. For emergency medicine, this study is expected to have a significant impact on patient treatment and we expect the results will be considered while establishing future recommendations.