Current literature clearly shows the benefit of pulmonary rehabilitation in symptomatic COPD (Trooster,2005). However, these patients are frequently unable to sustain a work-load sufficiently high to obtain a full benefit on exercise tolerance (Trooster,2005). Especially in patients with Chronic Respiratory Failure (CRF), the development of hypoxia (O'Donnel,2001) and the increase of dead space (Elbehairy,2015) during effort explain the out-of-proportion increase in ventilation leading to an early achievement of the ventilatory reserve. Recent studies on heated and humidified high flow oxygen (HFO) delivered through nasal cannula, show several positive effects on breathing pattern and ventilatory efficiency, mostly in critical care setting and at rest (Spoletini,2015). Some recent physiological studies have evidenced that high flows of humidified oxygen improve exercise performance in patients with COPD and severe oxygen dependency, in part by enhancing oxygenation (Chatila,2004). Recently, a pilot study by our group showed that HFO may improve the exercise performance in severe COPD patients with ventilatory limitation. This effect is associated to an improvement of oxygen saturation (SatO2) and perceived symptoms at iso-time (Cirio,2016). No clinical studies are available about the use of HFO during exercise training. The investigators hypothesize that, in severe COPD patients with CRF and exercise limitation , the use of HFO could improve the efficiency of ventilation, leading to an increase in the patient's exercise performance and outcome.

Primary aim will be to evaluate in patients COPD with CRF the difference in the endurance tolerance improvement (expressed in minutes) after an high intensity training program, at iso-FiO2, using HFO with respect to usual oxygen administration by " Venturi Mask" .

Secondary objectives will be to study effectiveness of HFO with respect to "Venturi Mask" in terms of improvement of meters of 6 Minute Walking Test, dyspnea at rest, peripheral and respiratory muscle strength,blood gases, motor and respiratory disability,quality of life,impact of the disease and patients satisfaction.

160 patients will be enrolled in 4 different hospitals. One of those, ICS Maugeri, will contribute with 5 different Institutes [Brescia, Pavia, Tradate (VA),Veruno (NO) and Cassano delle Murge (BA)]. Thus the total recruiting centers will be eight.

During the whole rehabilitation period, the patients will perform one training session/day.

The patients will use medications and oxygen therapy at rest as prescribed. The patients will perform no less than 20 training sessions within a period of one month. The rehabilitative intervention will be constantly supervised by a respiratory therapist.

The cycle load started from the 50% of the theoretical maximum watt (evaluated by 6 minutes walking test [6MWT] through the Luxton's equation) (Luxton,2008) and increased by Maltais's protocol (Maltais,1997). Run-in training session: all patients will be submitted to a preliminary 30 minute training under a Venturi mask to set the FiO2 able to maintain SpO2 constantly up to 93%. Usual nocturnal ventilation and nocturnal oxygen therapy under nasal cannula will be allowed.

Patients will be randomized to 2 groups:

• OXYGEN (Control Group, n = 80): the training will be performed using the Venturi mask with the FiO2 set during the run -in session.
• HFO (TREATMENT Group, n = 80): the training will be performed using the HFO device. Air-flow will be set at the highest value tolerated by the patient, until a maximum value of 60 l/min and with the same FiO2 set of the Control Group (iso-FiO2) during the run-in session.

The HFO will be administered using the AIRVO2® (Fisher&Paykel- NewZealand, CE0123 - 93/42/CEE). Afterwards, the oxygen flow provided into the system will be progressively increased until the pre-fixed FiO2 will be reached, as displayed by the AIRVO2 monitor.

the investigators will expect an enrollmnet rate of at least 23 patients/center. An enrollment rate less than 12 patients/center will not guarantee 1 author name on the possible Scientific publication. All the centers will be named in a possible congress/poster presentation. The possible publication will consider that a) the study is under the endorsement of AIPO and ARIR societies and b) Authors acknowledge the Industry contribution.

Measures

At baseline (T0), for both groups the following clinical evaluations will be performed:

• Anthropometric (age, BMI, diagnosis)
• Scale of comorbidity (CIRS)
• Spirometry (FEV1 %prd, FVC %prd, FEV1/FVC)

At baseline (T0) and at the end of rehabilitative program (T1), an operator, unblinded to the study, will perform the following outcome measures:

• Constant load cycle-ergometer endurance [at work load of 80% of maximal load predicted by Luxon's equation (Luxton,2008)] under usual oxygen supply with nasal cannula
• Arterial blood gas analysis (ABG) under room air
• 6-min walking test (6MWT) under usual oxygen supply with nasal cannula
• Scale of the MRC dyspnea
• Quality of Life (MRF26 scale)
• Respiratory muscle strength (MIP and MEP)
• Biceps and Quadriceps muscle strength tested by manual dynamometer and MRC muscle scale
• Disability (Barthel index, Barthel dispnea)
• Impact of the disease (CAT scale)

During each training session the investigators will evaluate:

• Side effects (discomfort, severe dyspnea, dryness of mucosa , etc)
• delta increase in training prescription (% of variation in watts compared with the previous session)

Only at the end of the program (T1) the investigators will evaluate :

1. patient satisfaction for training sessions with a Likert scale (0= no discomfort; 1 minimum discomfort; 2= moderate discomfort; 3 = high discomfort; 4= maximum discomfort)
2. drops out (number and reasons: exacerbation, intolerance of treatment, refuse, early discharge, etc)

Statistical analysis will be performed using STATA 12. Descriptive data will be shown as mean ± SD.

The analyses will be conducted on a intention-to-treat (all randomized patients) or per-protocol (all completers) basis. A two-sample t test was used to explore differences in: 1. baseline characteristics intervention and control groups, 2. between improvers and non improvers, 3.between completers and 4. dropouts and to assess differences in changes of parameters following the rehabilitation program between intervention and control groups. Wilcoxon matched-paired tests and Mann-Whitney U test will be employed for nonparametric data. Frequency distributions will be analyzed with χ2 test. Data will be considered significant for p<0.05.

The sample estimated size for two sample comparison of means on primary outcome (endurance time evaluated by cicloergometer at costant load) (alpha error 0.05, beta error=0.90 considering mean control group=150 s, mean treatment group=280s, standard deviation for both groups 250s ) is 156 patients

Inclusion Criteria:

• COPD and CRF diagnosis under LTOT,
• clinical stability (pH > 7.35 and < 7.46, no change in respiratory drugs therapy during the last seven days)

Exclusion Criteria:

• orthopedic or neurological disease,
• cognitive impairment (Mini Mental State Examination < 22)
• anamnestic history of ischemic heart disease or heart failure, COPD+ fibrosis and COPD+ OSAS.