Official ERS/ATS clinical practice guidelines: noninvasive ventilation for acute respiratory failure

Question 1: Should NIV be used in COPD exacerbation?

Chronic obstructive pulmonary disease (COPD) exacerbations are a very common reason for admission to hospital. Approximately 20% of patients hospitalised for COPD present with [6] or develop hypercapnic respiratory failure [7], which is an indicator of an increased risk of death [6–8]. ARF leading to acute or acute-on-chronic respiratory acidosis, later referred to in this guideline as acute respiratory acidosis, develops when the respiratory muscles fail to achieve adequate alveolar ventilation despite high levels of diaphragmatic activity [9]. It is likely that hyperinflation during exacerbation also contributes to respiratory muscle compromise. When faced with a load that exceeds the capacity of the respiratory muscle pump, a rapid shallow breathing pattern develops which is characterised by an increased respiratory rate with small tidal volumes. Although this is not always beneficial in terms of respiratory muscle energy expenditure, this pattern seems to have a complex pathophysiological mechanism and occurs at the expense of adequate alveolar ventilation. As a consequence, the level of arterial carbon dioxide (CO₂) rises and respiratory acidosis ensues. Measuring respiratory rate, observing chest and abdominal wall movement, and obtaining a sample of arterial blood are therefore key in the initial assessment of the patient at risk of acute respiratory acidosis (pH ≤7.35). Many patients with COPD are hypercapnic at baseline and the development of acidosis indicates acute-on-chronic hypercapnic respiratory failure.

Bilevel NIV may be considered in COPD patients with an acute exacerbation in three clinical settings [10]:

• 1) To prevent acute respiratory acidosis, i.e. when the arterial CO₂ tension (P_aCO2) is normal or elevated but pH is normal (see Question 1a).

• 2) To prevent endotracheal intubation and invasive mechanical ventilation in patients with mild to moderate acidosis and respiratory distress, with the aim of preventing deterioration to a point when invasive ventilation would be considered (see Question 1b).

• 3) As an alternative to invasive ventilation in patients with severe acidosis and more severe respiratory distress (see Question 1b).

Bilevel NIV may also be used as the only method for providing ventilatory support in patients who are not candidates for or decline invasive mechanical ventilation.

Question 1a: Should NIV be used in ARF due to a COPD exacerbation to prevent the development of respiratory acidosis?

One randomised study [11] of 52 patients with COPD comparing bilevel NIV to standard oxygen therapy used inclusion criteria of recent onset of shortness of breath and a pH >7.30 (average pH of randomised patients was within the normal range), suggesting that most enrolled patients had not developed acute respiratory acidosis. NIV was poorly tolerated and there was no effect of NIV on intubation rate (8% in NIV arm and 7% in control arm) or mortality (4% in NIV arm and 7% in control arm). However, there was a decrease in Borg dyspnoea score with NIV compared with standard oxygen therapy at 1 h and on day 2. When one outlier was removed from the analysis, there was a statistically significant reduction in hospital length of stay with bilevel NIV compared with the control group (5 versus 7 days).

In three other studies (two randomised controlled trials (RCTs) and one prospective observational study [12–14]) the mean pH was only very mildly acidotic, suggesting again that patients without acute respiratory acidosis were included; none of these studies showed an advantage to bilevel NIV. In a large study (n=342) in which bilevel NIV was started within 24–48 h of admission compared with standard oxygen therapy, in patients with a mean pH at randomisation of 7.35, there was a decrease in patients meeting criteria for intubation with NIV (1.4% versus 4.6%; p=0.002), but no difference in mortality [15]. Subgroup analysis of patients with a pH >7.35 (n=151) showed a beneficial effect of NIV for patients meeting clinical criteria for intubation versus those that did not (9 out of 80 versus 2 out of 71; p=0.045) but not for mortality. However, NIV was started much later than in other studies (24–48 h after admission) and the number of patients deteriorating to the point at which endotracheal intubation was considered necessary was surprisingly high, considering the mild acidosis. The main focus in hypercapnic patients with COPD who are not acidotic should be medical therapy and, most importantly, oxygen targeted to a saturation of 88–92% [16]. Austin et al. [17] showed a better survival in COPD patients being transferred to hospital when oxygen was delivered to achieve this range of saturation.

Question 1b: Should NIV be used in established acute hypercapnic respiratory failure due to a COPD exacerbation?

Question 2a: Should NIV be used in ARF due to cardiogenic pulmonary oedema?

Over 30 trials have been published on the use of CPAP and/or NIV compared with standard therapy or each other for patients with acute cardiogenic pulmonary oedema. The majority are small single-centre trials that span a period of 30 years, during which cardiogenic pulmonary oedema management, particularly in the setting of acute coronary syndrome, has evolved along with trial inclusion and exclusion criteria. Patients in cardiogenic shock have been almost universally excluded from trials and thus are not included in our NIV recommendations. Many trials have also excluded patients requiring acute revascularisation and some have excluded patients with acute coronary syndrome.

The pathophysiology of respiratory failure during cardiogenic pulmonary oedema includes decreased respiratory system compliance and alveolar flooding due to high capillary pressure associated or not with left ventricular systolic dysfunction [41]. NIV (including both bilevel and CPAP) in this context has the ability to improve respiratory mechanics and facilitate left ventricular work by decreasing left ventricular afterload [42]. This is facilitated by the decrease in negative pressure swings generated by the respiratory muscles.

In 2008, Gray et al. [43] published the largest multicentre trial from 26 emergency departments, in which 1069 patients were randomised to CPAP, bilevel NIV or standard oxygen therapy. This trial found physiological improvement in the CPAP and bilevel NIV groups compared with the standard group, but no difference in intubation rate or mortality at 7 and 30 days. However, interpretation of the results is limited by the high crossover rate (56 out of 367 patients (15%) in the oxygen group crossed over to bilevel NIV). Subsequently, five systematic reviews [44–48] that have incorporated the data from Gray et al. [43], as well as other new trials, have been published. They consistently conclude that: 1) NIV decreases the need for intubation, 2) NIV is associated with a reduction in hospital mortality, 3) NIV is not associated with increased myocardial infarction (a concern raised by the first study comparing NIV and CPAP [49]), and 4) CPAP and NIV have similar effects on these outcomes.

Question 2b: Should a trial of CPAP prior to hospitalisation be used to prevent deterioration in patients with ARF due to cardiogenic pulmonary oedema?

Six single-centre RCTs have evaluated the use of CPAP or bilevel NIV to treat cardiogenic pulmonary oedema in the pre-hospital setting. Four trials focused on acute cardiogenic pulmonary oedema and two included a subset of cardiogenic pulmonary oedema patients within a general population of patients with ARF (Thompson et al. [50]: 44 out of 69 cardiogenic pulmonary oedema; Roessler et al. [51]: 25 out of 49 cardiogenic pulmonary oedema). The studies were heterogeneous, with four using CPAP and two using bilevel NIV, and they were conducted in different settings. Five studies were conducted in Europe with a physician supporting the ambulance staff; one study was conducted in Canada, with pre-hospital care and study procedures managed by ambulance staff alone. The intubation rates varied among studies, suggesting potential differences in severity of illness or experience of the ambulance attendants. One study compared early versus delayed NIV [52], while the rest used a control arm of usual medical therapy.

Question 3: Should NIV be used in ARF due to acute asthma?

The main feature of acute asthma is a sudden and reversible episode of bronchoconstriction, leading to an increase in airways resistance that varies in severity. The acute change in mechanical load (mainly resistive) generates hyperinflation, increased respiratory muscle effort and dyspnoea. Hyperinflation also reduces respiratory muscle efficiency and thus the respiratory muscle pump may ultimately become exhausted, leading to hypercapnia [53].

NIV is used, together with conventional pharmacological treatment, with the aim to decrease the respiratory muscle work that is much increased during the episodes of acute bronchoconstriction, to improve ventilation, decrease the sensation of dyspnoea, and ultimately avoid intubation and invasive mechanical ventilation.

Possibly because the magnitude of the problem is small (i.e. episodes of acute asthma requiring ICU admission are uncommon), there is not much published research in this field. One physiological study gives some support to the use of CPAP. This study [54] showed that in asthmatic individuals subjected to histamine challenge causing acute bronchoconstriction, the use of CPAP (average 12 cmH₂O), compared with spontaneous breathing, substantially decreased the pressure–time product of the respiratory muscles (an index of energy utilisation), which was explained by a change in mechanics and breathing pattern.

A few uncontrolled studies and RCTs have compared NIV versus routine care in patients with acute asthma [55, 56]. Uncontrolled studies have shown physiological improvements in selected patients, whereas the RCTs and a meta-analysis have not shown differences in clinically relevant outcomes when comparing NIV with usual care [57]. A retrospective observational study by Fernandez et al. [58] described three types of patients admitted in the ICU for acute asthma: those with acute severe near-fatal asthma requiring emergency intubation, those responding well to medical therapy (the majority) and a small subgroup not responding well to medical therapy, who in this study were treated with NIV, mostly with success. RCTs in asthma mostly considered physiological end-points such as improvement in peak flows and suggested some beneficial effects.

In a recent retrospective cohort study [59] conducted in 97 US hospitals during a 4-year period, the use of NIV in patients with acute asthma was 4% (556 out of 13 930) and the use of invasive ventilation was 5% (668 out of 13 930). Failure rate (defined as intubation) of NIV was 4.7% (26 patients). Hospital mortality rates in those who received invasive mechanical ventilation without a previous trial of NIV, those who failed a NIV attempt and those who succeeded on NIV were 14.5%, 15.4% and 2.3%, respectively.

Question 4: Should NIV be used for ARF in immunocompromised patients?

ARF is the main indication for ICU admission in immunocompromised patients. Currently available literature supports the use of NIV as a first-line approach for managing mild to moderate ARF in selected patients with immunosuppression of various aetiologies. Several studies have reported clinical benefits of NIV, although strict monitoring in the ICU and prompt availability of invasive mechanical ventilation are mandatory [60–62]. A large observational multicentre Italian survey investigated the clinical impact of NIV use in 1302 haematological patients admitted to the ICU with ARF [63]. The authors, after a propensity score analysis, confirmed the role of NIV as an independent predictor of survival. Similarly, early CPAP has been reported as a practical, simple and inexpensive method to prevent deterioration of respiratory function and complications in such patients [64]. However, in a recent multicentre randomised trial of 374 immunocompromised subjects, early NIV, compared with standard oxygen, was not associated with clinical benefits in mortality, ICU-acquired infections, duration of mechanical ventilation or length of ICU stay [65].

In addition, a more recent post hoc analysis of immunosuppressed patients enrolled in a larger trial of patients with ARF suggested no benefit of NIV, either for intubation or survival [66].

Question 5: Should NIV be used in de novo ARF?

De novo respiratory failure refers to respiratory failure occurring without prior chronic respiratory disease. Most patients in this category have hypoxaemic respiratory failure, usually defined as significant hypoxaemia (arterial oxygen tension/inspiratory oxygen fraction ratio (P_aO2/F_IO2) ≤200), tachypnoea (respiratory rate >30–35 breaths·min^–1) and a non-COPD diagnosis (e.g. pneumonia and/or acute respiratory distress syndrome (ARDS)). Although included in some of the studies, patients with cardiogenic pulmonary oedema or with post-operative respiratory distress are not considered here because their pathophysiology is different and the indications and outcomes for NIV differ.

Defining whether a patient has all the features of ARDS [68] may be difficult or impossible in nonintubated patients because precise measurement of the actual inspired fraction of oxygen is unavailable and positive end-expiratory pressure is not used or its level is uncertain (e.g. with high-flow oxygen). Therefore, we considered studies on patients with hypoxaemic respiratory failure, community-acquired pneumonia and ARDS in the same category. NIV is used in these patients with the aims of improving oxygenation, facilitating ventilation, decreasing the work of breathing and dyspnoea, avoiding intubation, and reducing the complications associated with invasive mechanical ventilation.

Limitations of NIV in achieving some of these aims relative to invasive ventilation in patients with de novo ARF include its lack of efficacy in reducing work of breathing, in contrast to hypercapnic respiratory failure where its ability to reduce work of breathing has been clearly demonstrated. In ARDS patients, it has been shown that the use of noninvasive inspiratory pressure support can decrease the inspiratory effort compared with no inspiratory assistance only if sufficient pressure support is added [69]. Of concern, the tidal volume can also be significantly higher during NIV, especially when substantial inspiratory pressure is delivered, and further exacerbated by the high inspiratory demand seen in patients with acute hypoxic respiratory failure [70]. Therefore, the total pressure dissipated to inflate the lungs can be excessive during NIV. Such large transpulmonary pressures and the resulting large tidal volumes may exacerbate lung injury if prolonged over time. It is possible, although not proven, that NIV is especially useful in patients who do not substantially increase their tidal volume, but further work is needed in this area [71].

Invasive ventilation for hypoxaemic respiratory failure clearly reduces work of breathing and permits paralysis if total control of breathing is desired, effects that can redistribute blood flow from the respiratory muscles to other organs in patients with shock and hence help to treat shock itself [72]. The ability of NIV to achieve optimal pressures to reduce the work of breathing reliably in acute hypoxaemic respiratory failure is challenging because the high pressures often required increase air leaks, gastric insufflation and patient intolerance [69]. Thus, the ability to use lung protective ventilator strategies (such as maintaining a low tidal volume of 6 mL·kg⁻¹ of predicted body weight) may be more difficult with NIV than with invasive ventilation [73, 74]. Some evidence even supports the idea that spontaneous ventilation can induce harm similar to ventilator-induced lung injury in situations of severe lung injury [75–77], which raises a note of caution when using NIV that combines spontaneous effort with ventilator support.

Hypoxaemia and high work of breathing return immediately when NIV is removed, explaining the risk associated with NIV interruptions. This may be ameliorated, as suggested by some recent trials, by the use of high-flow nasal cannula therapy (using a specialised nasal cannula to deliver heated and humidified oxygenated gas at flows between 30 and 60 L·min^–1) during breaks or using longer sessions of NIV [78]. Also, the duration of NIV and its tolerance may depend on the type of interface used; recent evidence suggests that the use of a helmet may offer better tolerance over prolonged periods [79–82].

Potential uses of NIV for de novo ARF include as a preventive strategy for avoiding intubation. One pilot study on patients with “early” ARDS (P_aO2/F_IO2 >200 and ≤300) showed avoidance of intubation and reduced cytokine levels as favourable outcomes [83]. However, this study has not been replicated. NIV has also been studied as an alternative to intubation, with occasional reports showing benefit [84]. Positive studies on hypoxaemic, nonhypercapnic respiratory failure, mainly caused by community- or hospital-acquired pneumonia, have enrolled carefully selected patients who are cooperative with no associated major organ dysfunction, cardiac ischaemia or arrhythmias, and with no limitations in clearing secretions [29, 60, 85–87], which may explain the benefits seen.

Until recently, almost all studies on NIV for de novo ARF compared it with oxygen delivered with standard air entrainment (Venturi masks) or reservoir masks. Recently, high-flow nasal cannula therapy has been shown to offer several advantages compared with NIV, including better tolerance and dead space reduction [88]. One recent RCT reported a survival benefit of high-flow nasal cannula over standard oxygen therapy and bilevel NIV, although the primary end-point of intubation was not significantly different [67]. Although high-flow nasal cannula therapy is not specifically addressed in these recommendations, it may assume an important role in the therapy of de novo respiratory failure in the future.

The main risk of NIV for the indication of de novo ARF is to delay a needed intubation [86]. Early predictors of NIV failure include higher severity score, older age, ARDS or pneumonia as the aetiology for respiratory failure, or a failure to improve after 1 h of treatment [89]. Although the reasons for a poorer outcome are not completely understood, patients with NIV failure have higher tidal volumes before intubation [71] and develop more complications after intubation [90]. Studies have shown that NIV failure is an independent risk factor for mortality specifically in this population, although careful patient selection seems to reduce this risk [91, 92].

Question 6: Should NIV be used in ARF in the post-operative setting?

Surgery, particularly that approaching the diaphragm, anaesthesia and post-operative pain may all have deleterious effects on the respiratory system, causing hypoxaemia, a decrease in lung volume and atelectasis due to diaphragm dysfunction [93]. These modifications of respiratory function occur early after surgery and diaphragm dysfunction may last up to 7 days, leading to an important deterioration in arterial oxygenation [94]. Maintenance of adequate oxygenation and avoidance of symptoms of respiratory distress in the post-operative period are of major importance, especially when pulmonary complications such as ARF occur [95, 96].

Both bilevel NIV and CPAP are frequently used in these clinical situations. Imaging studies have shown that the use of NIV may increase lung aeration and decrease the amount of atelectasis during the post-operative period of patients undergoing major abdominal surgery [93]. Physiological studies have shown that CPAP and bilevel NIV are effective at improving lung aeration and arterial oxygenation and decreasing the amount of atelectasis without adverse haemodynamic effects during the post-operative period after extubation [93, 97].

Question 7: Should NIV be used in patients with ARF receiving palliative care?

In palliative care, the intensity of breathlessness frequently worsens as death approaches. Patients and their families expect symptomatic relief of this devastating symptom. Clinicians often respond by providing opioids, a highly effective treatment for this symptom but with a number of potentially undesirable side-effects, including excessive sedation.

The Society of Critical Care Medicine charged a Task Force to provide guidance for the use of NIV in palliative care settings [101]. Apart from type 1 patients in whom NIV is a life support with no limitation of therapy, the authors identified distinct goals for patients in palliative settings; identifying a “type 2” scenario in which a patient has decided to forego intubation, but still wants to receive salvage NIV therapy with the goal of surviving the hospitalisation. “Type 3” patients seek symptom alleviation, mainly dyspnoea, and survival is not a goal for these patients. Most of these type 3 patients, with the support of their families, are interested in ensuring comfort while dying, but some may also be interested in prolonging their lives for a few hours while maintaining cognition and the ability to communicate as they await relatives or to finalise their affairs. In this context, NIV would be considered effective if it improves breathlessness and respiratory distress without causing other troubling consequences, such as mask discomfort or unduly prolonging life. One large observational study on patients using NIV in ICUs found that those in whom NIV was a ceiling for therapy had the same quality of life as patients with no treatment limitations if they survived to day 90 [102].

Two RCTs in patients with advanced cancer have evaluated the efficacy of NIV in reducing dyspnoea. Hui et al. [103] showed a similar improvement in dyspnoea scores between NIV and high-flow oxygen, while a larger multicentre study [104] demonstrated a significantly greater reduction in breathlessness using NIV, especially in the hypercapnic subgroup of patients. Interestingly, the latter investigation showed that NIV might reduce the dose of morphine necessary to palliate dyspnoea, maintaining better cognitive function. Overall, NIV had a similar rate of acceptance by patients compared with oxygen therapy (∼60%).

Question 8: Should NIV be used in ARF due to chest trauma?

The utility of bilevel NIV or CPAP for treatment of respiratory failure in chest trauma patients has been investigated in three RCTs [105–107] and in a small subgroup of patients enrolled in a fourth trial [85]. Two of these studies compared NIV versus supplemental oxygen [85, 107], while the other two used invasive mechanical ventilation as control [105, 106].

A single-centre RCT randomised 69 trauma patients with at least four unilateral rib fractures to receive either CPAP or invasive mechanical ventilation [105]. Two out of 33 patients in the CPAP group died in hospital compared with none of the invasively ventilated patients (p>0.05). Patients randomised to CPAP had a significantly shorter ICU length of stay and less nosocomial pneumonia. A second RCT randomised 52 patients with multiple rib fractures and hypoxia to either CPAP or invasive ventilation [106], but no statistical difference was observed in any of the outcomes considered.

A third study randomised chest trauma patients with any injury and respiratory failure to NIV or high-flow oxygen therapy (>10 L·min^–1) [107]. The study was stopped early after a second interim analysis after 50 patients had been included. Patients enrolled in this study were significantly more hypoxaemic (mean P_aO2/F_IO2 100±34.5) than those in other studies. Significantly fewer intubations were performed in patients randomised to receive NIV (RR 0.20, 95% CI 0.05–0.87), resulting in a significant decrease in ICU length of stay with NIV.

The last study was a multicentre RCT that randomised patients with respiratory failure from multiple aetiologies to receive NIV versus standard oxygen therapy [85]. 17 out of 105 hypoxaemic patients (16%) that were enrolled had a presenting diagnosis of chest trauma. Outcome data for mortality and intubation was presented for the chest trauma subgroup, but no significant difference between groups was seen for either of these outcomes.

Question 9: Should NIV be used in ARF due to pandemic viral illness?

The use of NIV for severe acute respiratory syndrome and other airborne diseases has been assessed in several observational studies and remains controversial. These studies reported NIV failure rates of 30% and 33% [108, 109], with no evidence of viral spread to caregivers who took appropriate precautions. More recently, NIV was also used in patients with ARF due to influenza A H1N1 infection, with failure rates ranging between 13% and 77% [110–112]. However, no randomised clinical trial has assessed the efficacy of NIV in such pandemics.

Question 10: Should NIV be used in ARF following extubation from invasive mechanical ventilation?

Re-intubation following extubation failure is a major clinical problem. Rates of extubation failure as high as 23.5% have been reported [113]. In contrast to patients who are successfully extubated, those who are re-intubated because of post-extubation respiratory failure are characterised as having a worse prognosis, even after controlling for level of severity, which suggests a direct adverse effect of re-intubation itself on patient outcomes [114].

As re-intubation is associated with increased mortality, any strategy aimed at reducing the rate of post-extubation respiratory failure and avoiding re-intubation deserves consideration. NIV might be a means to avoid re-intubation either by treating post-extubation respiratory failure when it develops or preventing it from developing at all by instituting NIV immediately after extubation. As these are different questions, they are best considered and discussed separately. For the sake of clarity, post-operative patients are not considered in this section.

Question 10a: Should NIV be used to prevent respiratory failure post-extubation?

The benefits of early application of NIV soon after extubation have been assessed in unselected patients (i.e. any patients after planned extubation) and in at-risk patients. For most included studies, at-risk included patients >65 years or those with underlying cardiac or respiratory disease.

Question 10b: Should NIV be used in the treatment of respiratory failure that develops post-extubation?

Following the positive findings of case series and case–control studies, two RCTs compared NIV with conventional treatment (oxygen therapy) in patients who developed respiratory failure after planned extubation [123, 124].

Keenan et al. [123] investigated the role of NIV, compared with standard treatment, to avert re-intubation in 78 patients who developed respiratory distress, defined as tachypnoea and use of respiratory accessory muscles, within 48 h of extubation. This single-centre RCT did not find any beneficial effect of NIV on re-intubation, ICU and hospital mortality or length of ICU and hospital stay. In a multicentre RCT, Esteban et al. [124] enrolled 221 patients with respiratory failure defined by at least two of the following: hypoxaemia, respiratory acidosis, tachypnoea or respiratory distress within 48 h of extubation; patients were randomised to NIV or standard treatment. The trial showed no benefit from NIV on re-intubation and ICU length of stay. ICU mortality was higher in the NIV group, suggesting that it was harmful, probably by delaying intubation. A meta-analysis performed by Lin et al. [125] in 2014 on these two studies indicates no benefit of NIV compared with standard treatment with respect to re-intubation rate and mortality.

Question 11: Should NIV be used to facilitate weaning patients from invasive mechanical ventilation?

NIV has been shown to be as effective as invasive mechanical ventilation in improving breathing pattern, reducing inspiratory effort and maintaining adequate gas exchange during the weaning phase in selected patients intubated and ventilated for hypercapnic ARF [126]. Based on this physiological rationale, NIV has been utilised in these patients as a means to speed up the weaning process, while avoiding the side-effects and complications of invasive ventilation.

Burns et al. [127] identified 16 RCTs enrolling a total of 994 participants, mostly with COPD. Of the 16 studies, nine included only patients with COPD [128–136], six included mixed populations (predominantly hypercapnic) [137–142] and one small pilot study included only hypoxaemic patients [143]. Compared with conventional weaning using either a progressive reduction of inspiratory support or SBTs, NIV was associated with a significant decrease in mortality (RR 0.53, 95% CI 0.36–0.80; 994 patients). The pooled data showed a significant reduction in the proportion of weaning failures with NIV (RR 0.63, 95% CI 0.42–0.96). There were also significant reductions in ventilator-associated pneumonia (RR 0.25, 95% CI 0.15–0.43; 953 patients), ICU (mean difference −5.59 days, 95% CI −7.90 to −3.28) and hospital (mean difference −6.04 days, 95% CI −9.22 to −2.87) length of stay, total duration of mechanical ventilation (mean difference −5.64 days, 95% CI −9.50 to −1.77), and duration of invasive ventilation (mean difference −7.44 days, 95% CI −10.34 to −4.55). Weaning time and weaning failure were no different between NIV and conventional weaning (mean difference −0.25 days, 95% CI −2.06 to 1.56).

Of the 16 studies, the largest multicentre RCT (including 208 patients randomised to three groups receiving either conventional weaning through the endotracheal tube, early extubation followed by NIV or early extubation followed by oxygen therapy) resulted in similar rates of re-intubation in all three groups (30%, 32%, and 37%, respectively; p-value nonsignificant) [138]. However, the cumulative rates of weaning success were significantly better in patients randomised to the NIV group. The reason for this discrepancy in re-intubation versus weaning success was related to the use of NIV as a rescue treatment for patients who failed conventional weaning and oxygen therapy.