Publication

• Title: Effect of Protocolized Weaning With Early Extubation to Noninvasive Ventilation vs Invasive Weaning on Time to Liberation From Mechanical Ventilation Among Patients With Respiratory Failure: The Breathe Randomized Clinical Trial
• Acronym: Breathe
• Year: 2018
• Journal published in: JAMA
• Citation: Perkins GD, Mistry D, Gates S, Gao F, Snelson C, Hart N, et al; for the Breathe Collaborators. Effect of Protocolized Weaning With Early Extubation to Noninvasive Ventilation vs Invasive Weaning on Time to Liberation From Mechanical Ventilation Among Patients With Respiratory Failure: The Breathe Randomized Clinical Trial. JAMA. 2018;320(18):1881-1888.

Context & Rationale

• Background

  • Prolonged invasive mechanical ventilation is associated with increased morbidity and mortality, and evidence-based weaning strategies emphasise daily readiness assessment, spontaneous breathing trials (SBTs), sedation minimisation, weaning protocols, and early mobilisation.
  • Most patients have “simple” weaning, but about one-third require more than one SBT (“difficult” weaning), accruing disproportionate ICU resource use and higher risk of complications (eg, ventilator-associated pneumonia, immobility, discomfort, inability to speak).
  • Noninvasive ventilation (NIV) can provide ventilatory support while avoiding the endotracheal tube, plausibly reducing sedation exposure and endotracheal-tube–related harms.
  • Prior NIV-facilitated weaning trials were often small, focused on selected (frequently COPD/hypercapnic) populations, used non-standard readiness/weaning criteria, and had variable risk of bias, limiting generalisability to contemporary UK ICU practice.
• Research Question/Hypothesis

  • Among adult ICU patients ventilated >48 hours who failed an SBT, does protocolised early extubation to NIV (vs protocolised continued invasive weaning) reduce the time to successful liberation from all forms of mechanical ventilation among survivors?
• Why This Matters

  • If effective, an NIV-facilitated approach could reduce invasive ventilator days, sedation exposure, ventilator-associated complications, and ICU length of stay.
  • If ineffective (or harmful), it would argue for focusing on optimising conventional protocolised invasive weaning rather than adding complexity and resource demands.
  • Methodologically, it tests whether shifting “where” ventilatory support is delivered (ETT vs mask) translates into clinically meaningful acceleration of liberation, and clarifies which endpoints best capture that benefit.

Design & Methods

• Research Question: In adults invasively ventilated for >48 hours who fail an SBT, does protocolised early extubation to NIV (vs protocolised invasive weaning) reduce time to successful liberation from all mechanical ventilation among survivors?
• Study Type: Randomised, allocation-concealed, open-label, pragmatic, multicentre parallel-group clinical trial; 41 UK National Health Service ICUs; enrolment March 2013 to October 2016; follow-up to April 2017.
• Population:
  • Adults (>16 years) invasively ventilated via endotracheal tube for >48 hours, screened daily and judged ready to wean, who then failed an SBT (T-piece, CPAP, or low-level pressure support 5–7 cm H₂O; planned duration ≥30 minutes, up to 120 minutes in higher-risk patients).
  • Key exclusions included: tracheostomy present; unplanned extubation; clinical need for immediate extubation (eg, only intubated for airway protection); post–upper airway surgery; receiving long-term home oxygen or home NIV; and other protocol-defined exclusions.
• Intervention:
  • Protocolised noninvasive weaning: early extubation soon after failed SBT to NIV delivered via mask; initial settings aligned to pre-extubation support, with protocolised stepwise reduction (reducing inspiratory support and/or breaks from NIV) as tolerated over a minimum 12-hour period; standardised criteria for reintubation and tracheostomy were prespecified.
• Comparison:
  • Protocolised invasive weaning: continued invasive ventilation with sequential pressure support reduction and once-daily SBTs; extubation after successful SBT; standardised criteria for reintubation and tracheostomy were prespecified.
• Blinding: Open-label (clinicians and participants not blinded); outcome assessors not blinded; primary outcome was largely objective (time to liberation from ventilation), but some secondary outcomes (eg, antibiotic treatment decisions) could be susceptible to performance/detection bias.
• Statistics: A total of 280 patients were required to detect a 1-day reduction in time to liberation from ventilation with 90% power at the 5% significance level; inflated to 364 to allow for losses and modelling assumptions (including a Weibull shape parameter of 0.737). Primary analysis was intention-to-treat among survivors using time-to-event methods; sensitivity analyses included competing risk methods for death, per-protocol analysis, and multiple imputation for missing primary outcome data.
• Follow-Up Period: In-hospital outcomes plus follow-up to 6 months (survival status assessed at 30, 90, and 180 days; health-related quality of life at 3 and 6 months).

Key Results

This trial was not stopped early. Recruitment and sample size were revised (from the original plan) because projected recruitment would not meet the initial target; analyses were completed at the revised target (n=364).

• Protocolised early extubation to NIV did not shorten the primary endpoint (time to liberation from any mechanical ventilation among survivors), despite marked acceleration of extubation.
• The NIV strategy achieved substantial separation in exposure to invasive ventilation (median 1 vs 4 days) and reduced sedation days (mean 4.1 vs 5.5).
• Reintubation was numerically higher after early extubation to NIV (37.0% vs 28.7%), but no statistically significant difference in reintubation, tracheostomy, or survival was reported.

Internal Validity

• Randomisation and allocation: Central randomisation with allocation concealment; baseline characteristics were broadly balanced (eg, APACHE II mean 18.9 vs 18.8; age mean 64.3 vs 61.8 years; prior ventilation median 5.3 vs 4.7 days).
• Drop out / exclusions: 364 randomised; 319 evaluable for the primary effectiveness outcome because 41 died before liberation, 2 withdrew, and 2 were discharged with ongoing ventilation.
• Performance/detection bias: Open-label design could influence co-interventions and clinician-driven outcomes; primary outcome is objective, but antibiotic prescribing and some adverse event reporting could be affected.
• Protocol adherence: High adherence to allocated strategy (received allocated intervention: 175/182 [96.1%] noninvasive vs 176/182 [96.7%] invasive).
• Separation of the variable of interest: Randomisation-to-extubation median 0.5 vs 3.0 days; invasive ventilation median 1 vs 4 days; sedation mean 4.1 vs 5.5 days; all demonstrating meaningful separation between strategies.
• Baseline illness severity and case-mix: Heterogeneous diagnoses (eg, pneumonia/respiratory infection 31.3% vs 40.1%; cardiac 14.8% vs 9.9%); heterogeneity may dilute benefit if effect is concentrated in specific physiological phenotypes.
• Heterogeneity and site effects: Conducted across 41 centres; 44% recruited from 3 centres (potential clustering and learning-curve effects), with reported sensitivity analyses not showing material differences between high-recruiting and other sites.
• Timing: Randomised after failed SBT (a clinically relevant decision point); intervention delivered promptly in NIV arm as shown by extubation timing.
• Dose (intervention intensity): Protocolised NIV step-down included a minimum 12-hour period of sequential reduction/breaks, which may influence total “any ventilation” time (by design, NIV remains counted as mechanical ventilation).
• Outcome assessment: Primary outcome conditioned on survival; competing-risk sensitivity analysis produced a similar hazard ratio; HRQoL follow-up had substantial missingness (eg, EQ-5D data at 3 months in 175/247 survivors and at 6 months in 159/224 survivors).
• Statistical rigour: Effect estimates were supported by multiple sensitivity analyses (per-protocol and imputation-based approaches) with similar conclusions; the revised sample size limited power to detect smaller-than-planned differences.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong given allocation concealment, high protocol adherence, and objective primary outcome measurement, but is tempered by the open-label design, the survivor-conditioned primary estimand, and reduced power after sample size revision.

External Validity

• Population representativeness: Pragmatic enrolment from 41 UK NHS ICUs in adults ventilated >48 hours with a failed SBT reflects a common “difficult weaning” subgroup in high-resource ICUs.
• Key exclusions: Patients requiring long-term home oxygen or home NIV and those with tracheostomy at baseline were excluded, limiting applicability to chronic hypercapnic COPD cohorts and long-term ventilator-dependent populations where NIV-facilitated weaning may have different effects.
• Intervention deliverability: Requires staff skilled in NIV initiation, titration, interface management, and monitoring; translation may be limited in settings with lower NIV experience or staffing constraints.
• Comparator realism: Control arm was a protocolised best-practice invasive strategy (including once-daily SBTs); incremental benefit of NIV may differ where usual care is less protocolised or SBT frequency differs.

Conclusion on External Validity: Generalisability is moderate to similar adult, high-resource ICUs using protocolised weaning; applicability is more limited in chronic hypercapnic COPD/home-NIV populations and in units without established NIV infrastructure.

Strengths & Limitations

• Strengths: Large pragmatic multicentre randomised trial; allocation concealment; high adherence; robust separation in invasive ventilation exposure; clinically relevant decision point (failed SBT); comprehensive secondary outcomes including sedation exposure, adverse events, survival, and longer-term HRQoL.
• Limitations: Open-label (risk of performance/detection bias for clinician-driven outcomes); primary outcome conditioned on survival and counts NIV as “ventilation” (potentially obscuring benefit of earlier extubation); revised sample size and recruitment concentration in a few sites; heterogeneity of underlying causes of weaning failure; substantial missing HRQoL data at follow-up time points.

Interpretation & Why It Matters

• Clinical meaning of a neutral primary endpoint

  In a broadly selected “difficult weaning” ICU population, shifting ventilatory support from endotracheal tube to NIV shortened invasive ventilation exposure without clearly accelerating liberation from any mechanical ventilation or improving survival.
• Pragmatic practice impact

  For many ICUs, the most reproducible signal is reduced invasive ventilator days (median 1 vs 4) and reduced sedation exposure (mean 4.1 vs 5.5), suggesting value where minimising endotracheal-tube burden is a priority and NIV delivery is safe and well-supported.
• Trial-methods message

  Endpoint selection matters: “time to liberation from any ventilation” may underweight clinically important reductions in invasive ventilation, while survivor-conditioned analyses require explicit competing-risk thinking and careful interpretation.

Controversies & Subsequent Evidence

• The accompanying editorial highlighted that the trial compared two protocolised strategies, meaning the control arm likely approximated best-practice invasive weaning; this raises the possibility that any incremental effect of NIV is smaller in well-protocolised ICUs than in settings with more variable usual care.¹
• Primary endpoint choice remains debated in NIV-facilitated weaning trials: counting NIV time as “mechanical ventilation” can attenuate apparent benefit of earlier extubation if the key patient-relevant advantage is reduced endotracheal-tube exposure rather than faster cessation of all ventilatory support.¹
• Systematic reviews/meta-analyses consistently note heterogeneity by underlying physiology and trial design; Bayesian synthesis of NIV-facilitated weaning trials supports variation in treatment effect across populations and weaning protocols, and reinforces the need for careful patient selection and protocol details when applying findings to practice.²
• Cochrane syntheses have framed NIV-facilitated weaning as potentially beneficial in selected groups (notably hypercapnic COPD) while emphasising limits of generalisability and the importance of protocolised co-interventions; BREATHE adds a large pragmatic dataset in a mixed ICU population where the primary endpoint remained neutral.³⁴
• Guideline recommendations on liberation from mechanical ventilation prioritise protocolised approaches (SBTs, sedation minimisation) and recognise NIV immediately after extubation in selected contexts; BREATHE supports avoiding a blanket assumption that NIV-facilitated weaning will shorten time to complete liberation in unselected difficult-to-wean patients.⁵

Summary

• In 364 adults ventilated >48 hours who failed an SBT across 41 UK ICUs, early extubation to NIV did not shorten time to liberation from any mechanical ventilation among survivors (median 4.3 vs 4.5 days; adjusted HR 1.10; 95% CI 0.89–1.40).
• The NIV strategy achieved earlier extubation (median 0.5 vs 3.0 days) and reduced invasive ventilation exposure (median 1 vs 4 days; IRR 0.60; 95% CI 0.47–0.87).
• Total ventilator days were slightly reduced (median 3 vs 4 days; IRR 0.80; 95% CI 0.62–1.00), but this did not translate into a clear reduction in the primary time-to-liberation endpoint.
• No statistically significant differences in reintubation, tracheostomy, or survival were reported; reintubation was numerically higher after early extubation to NIV (37.0% vs 28.7%).
• Meaningful secondary signals included fewer sedation days (mean 4.1 vs 5.5) and shorter time in critical care (mean 10.8 vs 12.2 days), supporting reduced invasive-support burden even with a neutral primary endpoint.

Overall Takeaway

BREATHE is a landmark pragmatic UK multicentre trial showing that, in unselected adults who fail an initial SBT after prolonged ventilation, protocolised early extubation to NIV substantially reduces invasive ventilation exposure but does not shorten time to complete liberation from all mechanical ventilation or improve survival. Its key contribution is clarifying that the clinical value of NIV in weaning may lie more in reducing endotracheal-tube and sedation burden than in accelerating total liberation, and that effects are highly sensitive to population physiology and endpoint choice.

Bibliography

• 1Munshi L, Ferguson ND. Weaning from mechanical ventilation: what should be the best strategy? JAMA. Published online October 22, 2018.
• 2Yeung CH, Couper K, Ryan EG, et al. Non-invasive ventilation as a strategy for weaning from invasive mechanical ventilation: a systematic review and Bayesian meta-analysis. Intensive Care Med. 2018;44(11):1694-1704.
• 3Burns KE, Meade MO, Premji A, Adhikari NK. Noninvasive positive pressure ventilation as a weaning strategy for intubated adults with respiratory failure. Cochrane Database Syst Rev. 2013;(12):CD004127.
• 4Osadnik CR, Tee VS, Carson-Chahhoud KV, et al. Non-invasive ventilation for the management of acute hypercapnic respiratory failure due to exacerbation of chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2017;(7):CD004104.
• 5Ouellette DR, Patel S, Girard TD, et al. Liberation from mechanical ventilation in critically ill adults: an official American College of Chest Physicians/American Thoracic Society clinical practice guideline: inspiratory pressure augmentation during spontaneous breathing trials, protocols minimizing sedation, and noninvasive ventilation immediately after extubation. Chest. 2017;151(1):166-180.