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

Key Points

Question  In adults in whom weaning from invasive mechanical ventilation is difficult, does early extubation using a protocolized noninvasive weaning regimen reduce the time to liberation from ventilation compared with protocolized invasive weaning?

Findings  In this randomized clinical trial that included 364 adults, the median time to liberation from ventilation for patients randomized to noninvasive weaning vs invasive weaning was 4.3 days vs 4.5 days, a difference that was not statistically significant.

Meaning  Protocolized weaning with early extubation to noninvasive ventilation compared with invasive weaning did not significantly shorten time to liberation from all forms of mechanical ventilation.

Importance  In adults in whom weaning from invasive mechanical ventilation is difficult, noninvasive ventilation may facilitate early liberation, but there is uncertainty about its effectiveness in a general intensive care patient population.

Objective  To investigate among patients with difficulty weaning the effects of protocolized weaning with early extubation to noninvasive ventilation on time to liberation from ventilation compared with protocolized invasive weaning.

Design, Setting, and Participants  Randomized, allocation-concealed, open-label, multicenter clinical trial enrolling patients between March 2013 and October 2016 from 41 intensive care units in the UK National Health Service. Follow-up continued until April 2017. Adults who received invasive mechanical ventilation for more than 48 hours and in whom a spontaneous breathing trial failed were enrolled.

Interventions  Patients were randomized to receive either protocolized weaning via early extubation to noninvasive ventilation (n = 182) or protocolized standard weaning (continued invasive ventilation until successful spontaneous breathing trial, followed by extubation) (n = 182).

Main Outcomes and Measures  Primary outcome was time from randomization to successful liberation from all forms of mechanical ventilation among survivors, measured in days, with the minimal clinically important difference defined as 1 day. Secondary outcomes were duration of invasive and total ventilation (days), reintubation or tracheostomy rates, and survival.

Results  Among 364 randomized patients (mean age, 63.1 [SD, 14.8] years; 50.5% male), 319 were evaluable for the primary effectiveness outcome (41 died before liberation, 2 withdrew, and 2 were discharged with ongoing ventilation). The median time to liberation was 4.3 days in the noninvasive group vs 4.5 days in the invasive group (adjusted hazard ratio, 1.1; 95% CI, 0.89-1.40). Competing risk analysis accounting for deaths had a similar result (adjusted hazard ratio, 1.1; 95% CI, 0.86-1.34). The noninvasive group received less invasive ventilation (median, 1 day vs 4 days; incidence rate ratio, 0.6; 95% CI, 0.47-0.87) and fewer total ventilator days (median, 3 days vs 4 days; incidence rate ratio, 0.8; 95% CI, 0.62-1.0). There was no significant difference in reintubation, tracheostomy rates, or survival. Adverse events occurred in 45 patients (24.7%) in the noninvasive group compared with 47 (25.8%) in the invasive group.

Conclusions and Relevance  Among patients requiring mechanical ventilation in whom a spontaneous breathing trial had failed, early extubation to noninvasive ventilation did not shorten time to liberation from any ventilation.

Trial Registration  ISRCTN Identifier: ISRCTN15635197

Invasive mechanical ventilation is a lifesaving intervention. However, prolonged ventilation is associated with increased morbidity and mortality.^1,2 Optimal processes for weaning from ventilation have been studied for many years and have led to evidence-based clinical practice guidelines to facilitate early liberation from invasive mechanical ventilation.³ These guidelines recommend using spontaneous breathing trials, minimizing sedation, using weaning protocols, and early mobilization to promote liberation from ventilation.

Although most invasively ventilated patients have an uncomplicated (simple) weaning pathway, about one-third require more than 1 spontaneous breathing trial and are considered difficult to wean.^1,4,5 Patients with difficulty weaning face the physical discomfort of ongoing tracheal intubation, are often unable to speak,⁶ and are at increased risk of ventilator-associated pneumonia.^7,8 Mobilization is often delayed because of concurrent sedation and concerns about unintentional extubation.^9,10 This group of patients consume a disproportionate amount of intensive care unit (ICU) resources.¹¹

Noninvasive mechanical ventilation, which is being used increasingly as an alternative to invasive ventilation,^12,13 may have a role in supporting early liberation from invasive mechanical ventilation in patients who have difficulty weaning. Although the use of noninvasive ventilation as an adjunct to weaning has been tested in previous studies, the patient populations and interventions tested are not generalizable to contemporary clinical ventilation practice.¹⁴

In this multicenter randomized clinical trial conducted in the United Kingdom, it was hypothesized that weaning protocols that directed clinicians to extubate patients who were difficult to wean to noninvasive ventilation, compared with conventional weaning protocols for invasive mechanical ventilation, would reduce the time to liberation from ventilation.

We conducted this randomized, allocation-concealed, controlled, open-label, multicenter trial in 41 general adult ICUs in the United Kingdom. The trial protocol was designed by the trial investigators (Supplement 1) and was approved by South Central C Research Ethics Committee (reference 12/SC/0515). It was endorsed by the UK Intensive Care Foundation. Written consent was obtained from patients, their next of kin, or a physician who was independent from the trial prior to randomization in accordance with national laws. The study included an internal pilot spanning the first 6 months of the trial, at which point progress was reviewed by the funder. The same trial protocol was used for the internal pilot as for the main study. Patients enrolled in the internal pilot were included as part of the main trial.

Adult patients who had received invasive mechanical ventilation through an endotracheal tube continuously for more than 48 hours and were ready to commence weaning were considered for enrollment. Exclusion criteria were pregnancy, presence of a tracheostomy, contraindications to noninvasive ventilation, profound neurological deficit, home ventilation prior to admission, treatment limitations, need for further surgery or sedation, or no noninvasive ventilator available. Readiness to wean was assessed by the treating clinician before randomization according to prespecified criteria.¹⁵ Patients judged ready to start weaning underwent a spontaneous breathing trial (eAppendix in Supplement 2). Patients in whom the spontaneous breathing trial failed were defined as difficult to wean and were eligible for randomization. After obtaining consent, eligible patients were randomized using web-based secure electronic randomization designed by the study statistician. The minimization method was used to randomize patients in a 1:1 (noninvasive or invasive) allocation. The stratifying factors used in the minimization algorithm were center, presence or absence of chronic obstructive pulmonary disease (COPD), and postoperative/nonoperative reason for ICU admission, and these ensured equal balance between treatment groups. Chronic obstructive pulmonary disease was defined by a preadmission diagnosis of COPD requiring pharmacological treatment, evidence of a ratio of forced expiratory volume in the first second to forced vital capacity of less than 0.7 (FEV₁/FVC <0.7) and an FEV₁ less than 80% of predicted, or presence of respiratory symptoms. Patients admitted to the ICU after surgery were defined as the postoperative group. Following the spontaneous breathing trial, pressure support ventilation was reestablished using the previous settings. If necessary, the level of pressure support was further titrated to achieve patient comfort and a respiratory rate less than 30/min.

When a treating clinician judged that a patient was ready to wean, the patient underwent extubation and immediately was provided with noninvasive ventilation via face mask. The noninvasive ventilator was configured to deliver an equivalent level of inspiratory positive airway pressure to the level of pressure support that was being provided by the invasive ventilator and expiratory positive airway pressure equivalent to the level of positive end-expiratory pressure. The level of inspiratory positive airway pressure was then titrated to achieve patient comfort and a respiratory rate less than 30/min. Every 2 hours, the patient was assessed for signs of distress or fatigue. In the absence of distress or fatigue, the treating clinician either removed the noninvasive ventilation mask to allow a self-ventilation trial or reduced the level of positive airway pressure by 2 cm H₂O. The noninvasive weaning protocol was discontinued when the patient tolerated 12 hours of unsupported spontaneous ventilation.

Every 2 hours, a clinician assessed a patient for signs of distress or fatigue. In the absence of distress or fatigue, pressure support was reduced by 2 cm H₂O. This cycle was repeated every 2 hours as tolerated. If at any point the patient developed signs of distress or fatigue, then reversible causes were sought and corrective treatments initiated as appropriate. If this failed to resolve the situation, the level of pressure support was increased by 2 cm H₂O. Spontaneous breathing trials were repeated daily to assess extubation readiness. This cycle continued until either the patient underwent extubation after a successful spontaneous breathing trial or a tracheostomy was performed.

In both groups, the fraction of inspired oxygen was titrated to maintain arterial oxygen saturations greater than 90%. Both active weaning protocols were implemented between 8 am and 10 pm. Unless a patient developed signs of fatigue or distress, ventilator settings remained unchanged overnight.

The protocol encouraged use of a ventilator bundle (head-up position; oral decontamination; sedation hold; peptic ulcer prophylaxis) and recommended deferral of tracheostomy until after 7 days of ventilation. Guidance was provided for the criteria for reintubation, but the decision to reintubate was made by patients’ physicians. The decision to initiate antibiotic therapy and other treatments was at the discretion of patients’ physicians.

The primary outcome was time from randomization to successful liberation from ventilation, defined as the time point at which a patient was alive and free of ventilator (invasive or noninvasive) support for more than 48 hours. Secondary outcomes were duration of invasive ventilation and total ventilator days (invasive and noninvasive); proportion of patients receiving antibiotics for presumed respiratory infection; total days receiving antibiotics; rate of reintubation; mortality at 30, 90, and 180 days; time to meeting ICU discharge criteria; and rate at which patients fulfilled predefined criteria indicating the need for reintubation irrespective of whether they underwent reintubation. The predefined criteria were cardiac or respiratory arrest, respiratory pauses with loss of consciousness or gasping for air, severe psychomotor agitation inadequately controlled by sedation, persistent inability to remove respiratory secretions, heart rate of 50/min or lower or respiratory rate of 140/min or higher with loss of alertness, hemodynamic instability unresponsive to fluids and vasoactive drugs, requirement for surgery or other interventional procedure requiring deep sedation or anesthesia, proportion of patients receiving a tracheostomy, and mortality at 30, 90, and 180 days after randomization. Post hoc key process variables (weaning pathway, sedation use, length of ICU stay) are also reported. Outcomes were extracted from the ICU hospital clinical records and from questionnaires returned by patients. Because of the nature of the intervention and clinical record designs (which typically record mode of ventilation alongside respiratory variables), it was not possible for those assessing core ventilation outcomes to be blinded to treatment allocation. Adverse events were defined as development of skin or mucosal damage, vomiting, gastric distension, non–respiratory tract infection, and cardiac dysrhythmias. Health-related quality of life was assessed by the EQ-5D-5L (Euro Quality of Life 5 Dimensions Questionnaire with 5 levels of severity for each of the 5 dimensions) and the Short Form 12 at baseline (estimated retrospectively) and 90 and 180 days after randomization. All reported outcomes are postrandomization results.

The original sample size was 920 patients, but after a formal review requested by the funder, the sample size was revised to reflect a shorter than anticipated period of weaning. A median duration of weaning of 2.9 days and a difference of 1 day provided an associated hazard ratio of 1.53 and a minimum sample size of 280 with 90% power at an α=.05 significance level. One day was defined by the investigators and patient and public representatives as the minimal clinically important difference. The sample size was inflated by 23% to account for the rate of loss to follow-up seen up to the interim review of the data. It also accounted for the shape parameter, p, which was estimated by the data to be 0.918 and which allowed for nonconstant hazards (as modeled by the Weibull distribution), resulting in a final sample size of 364 (182 patients in each group). Revision of the sample size meant that the primary outcome would be analyzed using a Cox proportional hazards model as opposed to the competing risks regression model that was prespecified in the protocol.

The primary analysis method was intention to treat. Analysis of the primary outcome, time from randomization to liberation from ventilation, and other time to event outcomes used a Cox proportional hazards regression model to estimate hazard ratios and 95% confidence intervals. In addition, we used a competing risks regression model to account for the competing risk of death. Prior to the competing risk regression analysis, the cumulative incidence of liberation and death was plotted as basic descriptive data to understand the overall pattern over time. Mixed-effects logistic regression models were used to estimate the difference in mortality at 30, 90, and 180 days between the 2 groups, for which odds ratios and 95% confidence intervals are reported. Mixed-effects linear regression models were used to estimate mean treatment differences and 95% confidence intervals for continuous outcomes including the health-related quality-of-life measures (change from baseline). Mixed-effects negative binomial models were used to estimate incidence rate ratios and 95% confidence intervals for overdispersed count data; eg, number of days on invasive ventilation with zero inflation where several participants had no days on invasive ventilation. The study was not powered to detect treatment differences in secondary outcomes; hence, secondary analyses are considered exploratory.

We performed a per-protocol analysis and 2 predefined subgroup analyses (presence or absence of COPD; postoperative/nonoperative reason for ICU admission). It was not possible to perform the third planned subgroup analysis (physician-led vs nurse-led weaning) because all sites used a multiprofessional approach involving both physicians and nurses. Multiple imputation by chained equations was used to impute missing primary outcome data, and the imputed data set was analyzed as a sensitivity analysis.

All of the analyses used mixed-effects models adjusted for age, sex, center, post–spontaneous breathing trial Paco₂, presence or absence of COPD, and postoperative/nonoperative reason for ICU admission, where center was included as a random effect in the models. Modeling assumptions were assessed for all models fitted. The proportional hazards assumption was assessed for the Cox proportional hazards regression model and the competing risks model using plots of the log(−log) survival function and the Schoenfeld residuals and by assessing the influence of time-varying covariates. Linear, logistic, and negative binomial regression models were checked to ensure that the assumptions of linearity and constant variance were satisfied using residual plots. In addition to this, all the covariates included in the model were assumed to be independent of the outcome. All statistical tests were 2-sided using a P<.05 significance threshold. Statistical analyses were performed using Stata version 15.1 (StataCorp).

Figure 1 shows the flow of patients through the trial. Recruitment took place between March 2013 and October 2016, during which 364 patients were recruited from across 41 hospitals. There were 182 patients randomized to each group. Most patients received their randomized intervention (invasive group, 96.7% [176/182]; noninvasive group, 96.1% [175/182]).

Participant follow-up ended in April 2017. Overall baseline and physiological characteristics of patients were well matched (Table 1). For most patients, pneumonia (35.7%) or postsurgery respiratory failure (21.4%) was the main reason for mechanical ventilation.

The primary outcome, time from randomization to liberation from ventilation, was a median of 4.3 days (95% CI, 2.63-5.58 days) in the noninvasive group compared with a median of 4.5 days (95% CI, 3.46-7.25 days) in the invasive group (adjusted hazard ratio, 1.1; 95% CI, 0.89-1.40) (Figure 2). The competing risks regression analysis produced a similar result (adjusted hazard ratio, 1.1; 95% CI, 0.86-1.34) (Figure 3).

The noninvasive group required less invasive ventilation (median, 1 day vs 4 days; incidence rate ratio, 0.6; 95% CI, 0.47-0.87) and required fewer total ventilator days (median, 3 days vs 4 days; incidence rate ratio, 0.8; 95% CI, 0.62-1.0). Fewer patients in the noninvasive group received antibiotics for respiratory infection (60.4% vs 70.3%; unadjusted absolute difference, 9.9%; 95% CI, 0.17%-19.61%). The total number of days on which antibiotics were administered (respiratory and nonrespiratory) was not significantly different, with a mean of 9.1 days (SD, 12.0 days) in the noninvasive group and 10.4 days (SD, 13.2 days) in the invasive group (mean difference, 1.3 days; 95% CI, −1.31 to 3.88 days).

A higher proportion of patients underwent extubation in the noninvasive group (181/182) compared with the invasive group (143/182). Sixty-seven (37.0%) of 181 undergoing extubation in the noninvasive group underwent reintubation compared with 41 (28.7%) of 143 in the invasive group (odds ratio, 1.54; 95% CI, 0.89-2.41). For the end point of meeting the criteria for reintubation, there were 63 of 181 patients (34.8%) in the noninvasive group compared with 42 of 143 (29.4%) in the invasive group (odds ratio, 1.3; 95% CI, 0.78-2.12).

The rate of tracheostomy was 23.6% in the noninvasive group and 30.2% in the invasive group (odds ratio, 0.7; 95% CI, 0.44-1.15). Survival rates were not significantly different at 30 days (86.8% in the noninvasive group vs 86.3% in the invasive group; odds ratio, 1.1; 95% CI, 0.58-1.96) or at 180 days (78% in the noninvasive group vs 73.1% in the invasive group; odds ratio, 1.4; 95% CI, 0.85-2.27) (eTable 2 in Supplement 2). There were no significant differences in the proportions of patients who experienced adverse events and serious adverse events. Adverse events occurred in 45 patients (24.7%) in the noninvasive group compared with 47 (25.8%) in the invasive group. The distributions of adverse events and serious adverse events were similar (Table 2).

Post hoc key process measures showed that patients in the noninvasive group underwent extubation earlier than those in the invasive group (median, 0.5 day [interquartile range {IQR}, 0.5-1 day]) vs 3 days [IQR, 2-10 days]; adjusted hazard ratio, 2.5; 95% CI, 2.01-3.15; P < .001). Among those requiring reintubation, the noninvasive group underwent reintubation at a median of 2 days (IQR, 0.9-3.0 days) after randomization compared with 3.2 days (IQR, 2.3-4.7 days) in the invasive ventilation group (P < .001). The noninvasive group received sedation for fewer days (mean, 4.1 [SD, 5.0] days vs 5.5 [SD, 5.1] days; incidence rate ratio, 0.7; 95% CI, 0.61-0.91) and spent less time in critical care (mean, 10.8 [SD, 8.8] days vs 12.2 [SD, 8.4] days; P = .02). The median time from randomization to tracheostomy was 5.8 days (IQR, 3.71-8.46 days) in the invasive group and 5.6 days (IQR, 3.43-8.46 days) in the noninvasive group. There was no significant difference between the 2 groups (nonparametric P = .65).

Although health-related quality of life was impaired (eTable 3 in Supplement 2), there was no significant difference between the 2 groups at 3 months or at 6 months.

The per-protocol analysis produced results similar to the primary analysis (hazard ratio, 1.1; 95% CI, 0.90-1.44). The explored subgroups showed no significant difference in treatment effect (eTable 4 in Supplement 2). The sensitivity analysis using multiple imputation for the 45 participants with missing (censored) primary outcome data found no difference between the 2 groups (hazard ratio, 1.1; 95% CI, 0.90-1.36). A further sensitivity analysis found no significant difference in outcome between the 3 highest recruiting centers (who recruited 161 patients [44%]) and the other participating centers. There were no major departures from the modeling assumptions for all of the regression models fitted.

In this multicenter randomized trial, early extubation to noninvasive ventilation compared with protocolized invasive weaning with sequential pressure support reduction prior to extubation did not reduce the time to liberation from all forms of ventilation. Consistent with the protocol design, patients in the noninvasive ventilation group underwent extubation earlier and spent less time receiving invasive ventilation. Mortality rates, the requirement for reintubation or tracheostomy, and adverse event rates were not significantly different.

Spontaneous breathing trials are used to identify patients who are ready for extubation.¹⁶ The 59% to 86% of invasive ventilation patients in whom a spontaneous breathing trial fails are classified as difficult to wean.^1,4,17-19 These patients contribute to a disproportionate amount of ICU resource utilization to achieve successful liberation.¹¹ Noninvasive ventilation has been suggested to be a useful tool to facilitate weaning, but most previous studies recruited predominantly patients with COPD.^20-24 In that patient group, noninvasive weaning reduced mortality, duration of invasive ventilation, reintubation, and ICU length of stay.¹⁴ The patients enrolled in the present study better reflect contemporary ICU practice, as fewer patients with COPD now undergo invasive ventilation.^25,26

The rate of reintubation in this study was expected to be higher than among patients with simple weaning needs, in whom reintubation rates of 10% to 20% are reported.²⁷ The 30% overall rate of reintubation is consistent with findings in previous studies that recruited patients with difficulty weaning.^21,22,24 Because more patients underwent extubation in the noninvasive group, more were at risk of reintubation. One of the major concerns about reintubation is the association with increased mortality seen in some observational studies.^28,29 The survival rates in the present study were not significantly different in noninvasive and invasive weaning groups, although these findings should be interpreted with caution because the study was not powered to show a difference in this outcome and was not designed to assess equivalence.

The design of this study afforded several advantages to previous studies. First, a protocolized weaning regimen in both groups allowed clear separation of the intervention from the effect of protocolization.³⁰ Best practice guidelines (ventilation bundle, daily spontaneous breathing trials, tracheostomy insertion) reduced heterogeneity between treatment groups. Second, antibiotic use was selected as a surrogate for ventilator-associated pneumonia to limit the risk of detection bias arising from different approaches to obtaining respiratory samples for culture; this outcome is arguably more relevant than ventilator-associated pneumonia diagnosis as it better reflects antibiotic stewardship and exposure.

The study has several limitations. First, the nature of the intervention prevented blinding of clinicians, patients, or outcome assessors. This may have led to performance and/or detection bias. Second, the noninvasive weaning protocol mandated sequential reductions in respiratory support (either a decrease in inspiratory pressure support or a break from noninvasive ventilation) as tolerated over a minimum of a 12-hour period. It is possible that this may have extended the period of ventilatory support for some patients. Third, in the invasive ventilation group, the protocol required once-daily spontaneous breathing trials. It is possible that more frequent spontaneous breathing trials may have led to earlier recognition of readiness for extubation in some patients. Fourth, the patients enrolled were a heterogeneous group of patients with differing relative contributions of respiratory, cardiac, neuromuscular, metabolic, pharmacological, and neuropsychological impairment. Whether a more physiologically based assessment process could identify a group more likely to benefit from noninvasive ventilation remains to be determined in future studies. Fifth, 44% of the patients were recruited from 3 centers, which could limit generalizability. It is possible that performance and outcomes may have improved as centers became more experienced in the use of the noninvasive weaning intervention.

Among patients requiring mechanical ventilation in whom a spontaneous breathing trial had failed, early extubation to noninvasive ventilation did not shorten time to liberation from any ventilation.

Corresponding Author: Gavin D. Perkins, MD, Warwick Clinical Trials Unit, University of Warwick, Coventry CV4 7AL, United Kingdom (g.d.perkins@warwick.ac.uk).

Accepted for Publication: September 14, 2018.

Published Online: October 22, 2018. doi:10.1001/jama.2018.13763

Author Contributions: Drs Mistry and Lall had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Perkins, Gates, Gao, Snelson, Hart, Hoddell, McAuley, Walsh, Blackwood, Rose, Petrou, Young.

Acquisition, analysis, or interpretation of data: Perkins, Mistry, Gates, Snelson, Hart, Camporota, Varley, Carle, Paramasivam, McAuley, Walsh, Blackwood, Rose, Lamb, Petrou, Young, Lall.

Drafting of the manuscript: Perkins, Mistry, Hart, Varley, Paramasivam, Blackwood, Lamb, Petrou, Young, Lall.

Critical revision of the manuscript for important intellectual content: Perkins, Mistry, Gates, Gao, Snelson, Hart, Camporota, Varley, Carle, Paramasivam, Hoddell, McAuley, Walsh, Blackwood, Rose, Lamb, Petrou, Young.

Statistical analysis: Mistry, Gates, Lamb, Petrou, Lall.

Obtained funding: Perkins, Gates, Gao, Hart, McAuley, Walsh, Blackwood, Petrou, Young.

Administrative, technical, or material support: Perkins, Gao, Snelson, Hart, Camporota, Hoddell, Walsh, Young.

Supervision: Perkins, Gates, Paramasivam, Walsh, Young.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr McAuley reports receipt of personal fees from GlaxoSmithKline, SOBI, Peptinnovate, Boehringer Ingelheim, and Bayer; his institution has received funds from grants from the UK National Institute for Health Research (NIHR), Wellcome Trust, and others and from GlaxoSmithKline for Dr McAuley undertaking bronchoscopy as part of a clinical trial. In addition, Dr McAuley is 1 of 4 named inventors on patent US8962032 covering the use of sialic acid–bearing nanoparticles as anti-inflammatory agents issued to his institution, the Queen’s University of Belfast. Dr Varley reports receipt of nonfinancial support from La Jolla Pharmaceuticals and personal fees from EMAS Pharma. Dr Hart reports receipt of unrestricted research grants from Guy’s & St Thomas’ Charity and lecture fees from Fisher-Paykel, Philips, and Resmed. Dr Hart has a Myotrace patent pending and is on the Pulmonary Research Advisory Board for Philips; Philips-Respironics and Philips Research are contributing to the development of the Myotrace technology. Dr Hart’s Lane Fox Clinical Respiratory Physiology Research Group has received unrestricted grants (managed by Guy’s & St Thomas’ Foundation Trust) from Philips-Respironics, Philips, Resmed, Fisher-Paykel, and B&D Electromedical. No other disclosures were reported.

Group Information: The Breathe Collaborators: Addenbrookes Hospital–John Farman ICU: Adrian James Varley (principal investigator [PI]), Charlotte Bone, Petra Polgarova, Amy Scullion, Charlotte Summers, Katarzyna Zamoscik; Basildon and Thurrock University Hospital: Agilan Kaliappan (PI), Mark Vertue; Bedford Hospital: Sarah Snape (PI), Bindumal Jophy, Aoife McKnight, Catherine O’Brien; Birmingham Heartlands Hospital: Fang Gao-Smith (PI), Joanne Gresty, Teresa Melody, Peter J. Sutton; Blackpool Teaching Hospitals: Jason Cupitt (PI), Robert Thompson (PI), Melanie Caswell, Emma Stoddard; Bradford Royal Infirmary: Stephen Fletcher (PI), Tom Lawton (PI), Martin Northey; Edinburgh Royal Infirmary: Michael A. Gillies (PI), Heidi Dawson, Jane Whitehorn, Gosha Wojcik; Freeman Hospital, Newcastle upon Tyne: Claire Randell (PI), Verity Calder; George Eliot Hospital: Sam George (PI), Vivek Poongavanam (PI); Glenfield Hospital: Rakesh Vaja (PI), Dawn Hales, Gary Lau, Natalie Rich; Hillingdon Hospitals: Elisa Kam (PI), Sohan Bissoonauth, Lourdes Opimo, Hull Royal Infirmary: Ian Smith (PI), Caroline Abernethy, Victoria Martinson, Neil Smith; Intensive Care Foundation (ICF): Tim Gould (ICF chair); James Paget University Hospital: Andreas Brodbeck (PI), Lynn Everett; Leeds General Infirmary: E. Paramasivam (PI), Elizabeth Wilby; Leicester Royal Infirmary: Neil Flint (PI), Prem Andreou, Dawn Hales, Natalie Rich; Leighton Hospital: Daniel Saul (PI), Philip Chilton, Clare Hammell, Alistair Martin; Manchester Royal Infirmary: Mike Sharman (PI), Katie Ball, Andrew Brown, Richard Clark, Elaine Coughlan, Rachel Pearson, Sheeba Pradeep; Milton Keynes Hospital: Richard Stewart (PI), Jane Adderley, Rebecca Hinch, Cheryl Padilla Harris, Sara Beth Sutherland; Musgrove Park Hospital: Richard Innes (PI), Patricia Doble, Moira Tait; Papworth Hospital: Alain Vuylsteke (PI), Fiona Bottrill, Antonio Rubino; Peterborough City Hospital: Coralie Carle (PI), Theresa Croft, David Hannen, Alan Pope; Poole Hospital: Henrik Reschreiter (PI), Helena Barcraft-Barnes, Julie Camsooksai, Sarah Jenkins, Sarah Patch; Queen Alexandra Hospital, Portsmouth: Dave Pogson (PI), Steve Rose; Queen Elisabeth Hospital Birmingham: Catherine Snelson (PI), Toni Brunning, Ronald Carrera, Philip Pemberton, Martin Pope, Arlo Whitehouse; Queen Elisabeth Hospital King’s Lynn: Darcy Pearson (PI), Parvez Moondi (PI); Royal Blackburn Hospital: Srikanth Chukkambotla (PI), Caroline Aherne, Martin Bland, Lynne Bullock, Donna Harrison-Briggs; Royal Gwent Hospital: Nicholas Mason (PI), Una Gunter; Royal Sussex Hospital: Owen Boyd (PI), Laura Ortiz-Ruiz de Gordoa; Royal Surrey County Hospital: Ben Creagh-Brown (PI); Royal Victoria Hospital Belfast: Bronagh Blackwood (PI), Danny McAuley (PI), Leona Bannon, Laura Creighton, Pauline McElhill, Lia McNamee, Vanessa Quinn, Grainne White; Royal Victoria Infirmary Newcastle: Simon Baudouin (PI), Carmen Scott (PI), Ian Clemnt; Russells Hall Hospital: Michael Reay (PI), Clare Allcock, Sarah Fullwood, Ranjit Gidda; St Thomas’s Hospital: Luigi Camporota (PI), Kathryn Chan, Kate Flynn, Katie Lei, Nicola Purchase, John Smith, Samantha Smith; University Hospital Coventry and Warwickshire: Christopher Bassford (PI), Jeff Ting, Geraldine Ward; University Hospitals of North Midlands: Stephan Krueper (PI); University Hospital Southampton: Rebecca Cusack (PI), Clare Bolger, Karen Salmon; Western General Hospital: Anthony Bateman (PI), Heidi Dawson, Gosha Wojcik; West Middlesex University Hospital: Amandeep Gupta (PI), Barbara Walczynska; Yeovil District Hospital: Agnieszka Kubisz-Pudelko (PI), Nicholas Craw, Sarah Craw, Jeremy Reid; York Hospital: Joseph Carter (PI); Trial Steering Committee: Steve Goodacre, Simon Finney, Shelia Harvey, Gary Mills, Catherine Plowrigh, Duncan Wells, Barry Williams; Data Monitoring Committee: Charles Hinds (chair), David Harrison, Mark Griffiths; Warwick CTU: Nicola Cashin, Adam de Paeztron, Sarah Rumble, Laura Blair, Julia Sampson, Adam de Paeztron, Claire Jacques, Karoline Munro, Jess Smith, Kimberley White, Adam de Paeztron.

Funding/Support: This project was funded by the NIHR Health Technology Assessment Programme (project HTA 10/134) and supported by the Intensive Care Foundation. The trial was co-sponsored by the Heart of England NHS Foundation Trust and the University of Warwick.

Role of the Funder/Sponsor: This trial was commissioned and funded by the NIHR. The Heart of England NHS Foundation Trust and the University of Warwick took legal responsibility for all aspects of the research but did not provide specific funding. The sponsor and funder approved the design of the study and monitored the conduct of the study. They played no direct role in the design, data collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Disclaimer: The views expressed are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health.

Meeting Presentation: Presented at the European Society of Intensive Care Medicine 31st Annual Congress; October 22, 2018; Paris, France.

Data Sharing Statement: See Supplement 3.