The study will be performed as a randomized controlled non-inferiority trial. HFA has been increasingly used in the last years to treat hypoxic respiratory failure (i.e. type I failure), and numerous studies have shown its efficiency in this indication.

Despite this good evidence for HFA in hypoxic respiratory failure, it has only reluctantly been used for hypercapnic respiratory failure. HFA has been shown to generate PEEP, despite not being a closed system, and to improve CO2 clearance by flushing anatomical dead space. It might also help to reduce inspiratory resistance and facilitate secretion clearance from humidified gas. A study on COPD patients showed an increase in breathing pressure amplitude and mean pressure, as well as tidal volume, with a trend towards reduction of carbon dioxide partial pressure.

Intervention consists of HFA using standard equipment at the department. A gas flow of 60 litres per minute and a FiO2 as clinically feasible will be used. Therapy will be continued until a pCO2-level of 50 mmHg or less is reached, or therapy has to be aborted because of lack of tolerance by the patient or indication for intubation.

Control consists of non-invasive continuous positive airway pressure ventilation support using a tight mask and standard respirator equipment of the Department of Emergency Medicine. A positive airway pressure of 3,67 mmHg and a FiO2 as clinically feasible will be used. Therapy will be continued until a pCO2-level of 50 mmHg or less is reached, or therapy has to be aborted because of lack of tolerance by the patient or indication for intubation.

Respiratory failure is a leading cause of morbidity and mortality, and one of the most frequently encountered problems at the emergency department. Hypercapnic or hypercarbic respiratory failure, also dubbed respiratory failure type II, is characterized by failure of the respiratory system in one of its two gas exchange functions: carbon dioxide (CO2) elimination. It is thus usually defined via partial pressure of CO2 in the arterial blood gas (PaCO2), a value greater than 50 mmHg being a commonly used cut-off. Hypercapnic respiratory failure is often associated with severe airway disorders, such as asthma and chronic obstructive pulmonary disease (COPD), but might also be found in other conditions, where respiratory drive is restricted, such as intoxications, neuromuscular diseases or chest wall abnormities. Hypercapnic respiratory failure, i.e. respiratory failure type II, is often associated with hypoxemic respiratory failure, i.e. respiratory failure type I, failure of oxygenation. However, type I failure is also often observed without hypercapnia. Respiratory failure, both type I and II, may be further classified into either acute or chronic. Distinction between both forms is often challenging, and, in addition to arterial blood gas analysis, might require additional tests to identify clinical markers such as polycythemia or pulmonary heart disease. For practical reasons, acute respiratory failure is often defined as a condition, in which respiratory failure develops too fast to allow for renal compensation and an increase in bicarbonate (HCO3-) levels, and thus leading to acidosis (pH less than 7.3). Therapeutic strategies for hypercapnic respiratory failure include non-invasive CPAP ventilation support, intubation and mechanical ventilation (both assisted and controlled forms), and, in very severe cases, extracorporeal methods, such as extracorporeal life support systems.

Non-invasive CPAP ventilation support via either helmets, or different kinds of tight masks, is the current method of choice for the treatment of patients with acute respiratory failure in the intensive care setting. Eligible patients include those with an intact airway, airway-protective reflexes, who are alert enough to follow commands, whereas patients who lack those criteria require immediate endotracheal intubation. Non-invasive CPAP ventilation support improves both oxygenation (by providing an inspired fraction of oxygen (FiO2) of 100%, which is not possible via a simple venturi-mask, and the possibility of positive end-expiratory pressure (PEEP), preventing collapsing of the alveoli), and decarboxylation (by increasing tidal volume). It has been shown to decrease both need for intubation and in-hospital mortality. Non-invasive CPAP ventilation support, however, requires a high grade of skill from providers, and intensive communication with the patient to explain the usefulness of a tight-sitting device in the face in a situation of perceived massive dyspnea. Although severe adverse effects of non-invasive CPAP ventilation support are very rare, pain and pressure marks may occur. Despite all efforts of care providers, there is a relevant proportion of patients who do not tolerate ventilation support via a tight mask at all. About 15% of patients not tolerating the therapy, with an additional 25% of patients presenting with contraindications. These might include general contraindications against non-invasive techniques, such as aforementioned lack of airway-protective reflexes, but also such specific for tight masks, such as anatomical abnormities of the face.

High-flow Air via Nasal Cannula (HFA) therapy is usually applied via a wide-bore nasal cannula. It provides up to 60 litres per minute of a heated and humidified gas mixture (with an adjustable FiO2). This therapy is much less invasive for the patient, and thus often better tolerated. HFA has been increasingly used in the last years to treat hypoxic respiratory failure (i.e. type I failure), and numerous studies have shown its efficiency in this indication both at the intensive care unit and at the emergency. A recent systematic review and meta-analysis has concluded in improved patient comfort and reduced dyspnea scores. Despite this good evidence for HFA in hypoxic respiratory failure, it has only reluctantly been used for hypercapnic respiratory failure. This might be explained in a large part by the fact that patients with chronic hypercapnia are known to diminish their respiratory drive when exposed to hyperoxia. However, evidence has begun to change on this indication in recent time. HFA has been shown to generate PEEP, despite not being a closed system, and to improve CO2 clearance by flushing anatomical dead space. It might also help to reduce inspiratory resistance and facilitate secretion clearance from humidified gas. A study on COPD patients showed an increase in breathing pressure amplitude and mean pressure, as well as tidal volume, with a trend towards reduction of pCO2. Based on these findings, the use of HFA has increased in clinical practice, and a number of case reports and -series indicate successful use. Fraser et al. successfully investigated the use of HFA in patients with chronic COPD changes in arterial blood gases during use of HFA in the ED for both hypercapnic and non-hypercapnic patients were analyzed in previous studies, and found a significant reduction of pCO2.

There is, however, to date no randomized controlled trial investigating the effect of HFA in acute hypercapnic respiratory failure.

The study will be performed as a randomized controlled non-inferiority trial. The study site is the Department of Emergency Medicine (ED) at the Vienna General Hospital, a leading academic research center for emergency medicine at a large, tertiary care hospital. Around 90,000 patients are being treated at the department each year, approximately 150-200 of them suffering from hypercapnic respiratory failure, and requiring non-invasive ventilation support. The department features its own ICU and intermediate-care unit, with 7 positions each, for a total of 14 positions capable of providing CPAP therapy. The HFA-device is also at regular use at the department.

Patients who are temporary not able to give informed consent due to hypercapnia will be randomized and treated. They will be informed post-hoc as soon as they are able to give informed consent, and will have the possibility to give consent to the use of their data.

A random sequence will be generated by a person not involved in the enrolment of patients using standard software. Randomisation will be performed in variable blocks of 4 to 6, to yield an unpredictable allocation yet warranting balanced group sizes. Sequentially numbered sealed opaque envelopes (SNOSE), containing allocation either to the intervention or the control group, will be pre-produced. The envelopes will be opened after consent immediately before the start of the intervention to allow for allocation concealment and reduce the risk of immediate post-random exclusion.

Intervention consists of HFA using standard equipment at the department. A gas flow of 60 litres per minute and a FiO2 as clinically feasible will be used. Therapy will be continued until a pCO2-level of 50 mmHg or less is reached, or therapy has to be aborted because of lack of tolerance by the patient or indication for intubation.

Control consists of non-invasive CPAP ventilation support using a tight mask and standard respirator equipment of the Department of Emergency Medicine. A positive airway pressure of 3,67 mmHg and a FiO2 as clinically feasible will be used. Therapy will be continued until a pCO2-level of 50 mmHg or less is reached, or therapy has to be aborted because of lack of tolerance by the patient or indication for intubation.

Based on treating physician's discretion, both intervention and control treatments might be aborted at any time, and any other therapy (simple Venturi-Mask, HFA, non-invasive CPAP-ventilation support, intubation, extracorporeal methods) might be initiated.

Baseline characteristics and demographic variables will include age, sex, smoking status, prior diseases, especially any history of COPD or asthma, and duration of treatment of those, medication, body size, pre-hospital treatment. pCO2 levels will be measured using blood gas analysis at 0 -30 -60 (and every 60 minutes thereafter) minutes after the beginning of the therapy, and at the end of the therapy. Patient's perception of the therapy will be assessed after the end of the therapy using a 10-point Likert-Scale from very uncomfortable to very comfortable.

Sample size considerations are based on the primary outcome pCO2 reduction per hour. The investigators assume baseline pCO2-levels of 50 to 100 mmHg. Based on the published literature, the investigators assume that the outcome in the control group is 4±3mmHg/hour. Based on our clinical judgment the investigators assumed a limit of non-inferiority of 2 mmHg, which lies well within one standard deviation of the outcome. Hence, the investigators would need 28 experimental subjects and 28 control subjects to be able to reject the null hypothesis that the lower limit of a two-sided 90% confidence interval of the true difference between two groups is above the non-inferiority limit, at a power of 80%. Formally the investigators will have to enroll 56 patients. To allow for potential loss to follow up, missing data, measurement issues, or other design factors the investigators will increase actual sample size to 62. In terms of feasibility, the investigators expect approximately 150 to 200 eligible patients within one year, resulting in an expected study-duration of approximately 6 months.

Due to the relatively small sample size, there are no preplanned interim analyses.

Baseline data and demographics will be tabulated for the intervention- and control-group to assess success of randomisation. Reduction of pCO2 per hour will be compared between individuals in the intervention and individuals in the control group. The investigators will calculate effects as differences with 95% confidence intervals. This will be done using a linear mixed model with pCO2 as the outcome, treatment group as a factor variable, and baseline pCO2 and treatment time as covariates. The investigators will assume non-inferiority if the two-sided 95% confidence interval lies within predefined limits of non-inferiority. As a sensitivity analysis the investigators will also test for time/treatment-interaction in the model. As another sensitivity analysis, the investigators will analyze data during first 6 hours of treatment at maximum. In case of therapy failure, the investigators will use the last-observation-carried-forward method to 6 hours. Primary analysis will follow the intention-to-treat principle. The unit of analysis will be single persons.

Categorical secondary outcomes will be analysed by calculating relative risks with exact standard error based 95% confidence intervals. The investigators will assume non-inferiority if the two-sided 95% confidence interval lies within predefined limits of non-inferiority. Continuous secondary outcomes will be analyzed like the primary outcome. Length of stay data are expectedly lognormally distributed, therefore the investigators plan to use log-transformed values for further calculations. For data analysis the investigators will use Stata 11. A two-sided p-value less 0.05 is generally considered statistically significant. Reporting will follow the standards of the CONSORT extension for non-inferiority and equivalence trials.

Privacy and Data safety Directly and indirectly patient-related data will be stored physically and logically separated. In addition, only members of the study-group will have access to study data. Data will be stored on a secured computer of the department of emergency medicine and will be accessible only via restricted access for members of the study-group.

Inclusion Criteria:

• Adult patients (e.g. at least 18 years old) treated at the Emergency Department
• Acute hypercapnic respiratory failure defined as a pCO2 >50mmHg and a pH<7.30 on admission

Exclusion Criteria:

• Patients being comatose on admission, with no intact airway, lack of airway-protective reflexes, or those who are not alert enough to follow commands
• Patients intubated by Emergency Medical Service
• Patients requiring intubation on admission
• Pregnant women