Effect of Automated Closed-Loop Ventilation vs Protocolized Conventional Ventilation on Ventilator-Free Days in Critically Ill Adults: A Randomized Clinical Trial

Sinnige J. JAMA 2025. doi:10.1001/jama.2025.24384

Clinical Question

• Among critically ill mechanically ventilated adults, does early use of an automated closed-loop ventilation system (INTELLiVENT-ASV) increase ventilator-free days at day 28 (VFD-28) compared with protocolled conventional ventilation (CV)?

Background

• Mechanical ventilation carries the risk of causing lung injury and respiratory muscle weakness.
• Employing mandatory modes of ventilation in spontaneously breathing patients can cause discomfort, dyssynchrony, and increased requirements for sedation
• Automated closed-loop ventilators are able to continually analyse and react to patient respiratory parameters and respiratory effort and alter ventilatory parameters
• The rapidly changing conditions of critically ill patients make accurate, appropriate and timely adjustments of ventilator settings challenging and time consuming
• As such, the use of automated closed-loop systems may improve weaning, alongside promoting adherence to best practices in lung and diaphragm protection, with the potential added benefit of decreasing medical and nursing workload via automation
• A systematic review in 2024 by Goossen et al. found closed-loop models are at least as effective at choosing correct ventilator settings as ICU professionals. However, the studies were limited by underpowered study designs to show efficacy
• To date there has not been a large scale randomised controlled trial comparing automated closed-loop mechanical ventilation to conventional ventilation

Design

• International, multi-centre, investigator-initiated, open-label randomized superiority trial with concealed allocation
• Trial statisticians blinded to allocation
• Web-based central randomisation (1:1) stratified by centre using permuted blocks of 4, 6, or 8
• Deferred consent model with written consent within 72h of randomisation
• Modified intention-to-treat analysis
• Power calculation:
  • Based on previous studies in which mean VFD-28 was 20 (SD 9) days
  • Sample size estimation 1200 pts for beta of 80% and 2-tailed alpha 0.05
  • To detect mean difference of 1.5 VFD-28, allowing for 5% dropout

Setting

• 7 hospitals in 2 countries (Netherlands and Switzerland)
• Oct 2020 – June 2025

Population

• Inclusion:
  • Adults with < 1 hr since the start of invasive mechanical ventilation in the ICU and anticipated duration of invasive ventilation > 24hrs
• Exclusion:
  • Patient previously randomized in the study
  • Patient received invasive ventilation for > 1hr after ICU admission or start of ventilation
  • Patient received invasive ventilation for > 6hr directly preceding the current ICU admission
  • Patient aged < 18yrs
  • Confirmed or suspected pregnancy
  • Morbidly obese (BMI > 40)
  • Receiving or planned to receive ECMO
  • INTELLiVENT-ASV ventilator not available
  • Recent pneumonectomy or lobectomy
  • Premorbid restrictive pulmonary disease
  • Unreliable pulse oximetry (ie. secondary to carbon monoxide or sickle cell)
  • Neuromuscular diagnosis that can prolong duration of mechanical ventilation (eg. GBS, ALS, MS, MG, high spinal lesion)
  • No written deferred consent by patient or substitute decision-markers
• Participant Flow:
  • 3641 adults with < 1hr mechanical ventilation assessed for eligibility
  • 1514 randomised, 1201 analysed after post-randomisation exclusions
    • Post randomisation exclusion due to failure to obtain written deferred consent
  • 752 randomised to closed-loop ventilation
    • 602 analysed, 150 excluded due to no consent (19.9%)
  • 762 randomised to protocolised conventional ventilation
    • 599 analysed, 163 excluded due to no consent (21.4%)
• Comparing baseline characteristics of closed-loop vs. conventional group
  • Age: 63 vs 63
  • Female sex: 36.4% vs 35.7%
  • BMI: 26.2 vs 25.9
  • SAPS II on admission: 54 vs 52
  • SOFA: 8 vs 8
  • Sepsis presentation: 14.5% vs 18.6%
  • Hypoxaemic respiratory failure: 15.1% vs 16.7%
  • Medical ICU admission: 86 vs 86.3%
  • Emergency surgery admission: 12.5 vs 12 %
  • Planned surgery admission:  1.5 vs 1.7%
  • Reason for intubation:
    • Respiratory failure 30.4% vs 33.4%
    • Cardiac arrest 27.7% vs 25%
    • Neurological 25.9% vs 24.9%
    • Post-op 8.8% vs 9.2%
    • Airway protection 5% vs 4.7%
  • Ventilation before randomisation: 0.6 vs 0.6 hrs
  • Ventilation modes at randomisation:
    • Pressure control 63.4% vs 69.2%
    • Volume control 21.3% vs 21.5%
    • INTELLiVENT-ASV 9.1% vs 4.1%
    • PSV 4.8% vs 3.5%
  • Ventilation variables at randomisation:
    • Tidal volume: 6.6 vs 6.4 ml/kg
    • Plateau pressure: 19 vs 19 cmH2O
    • Respiratory rate: 22 vs 22
    • PEEP 7.7 vs 7.0 cmH2O
    • Driving pressure: 13 vs 13 cmH2O
    • FiO2: 0.6 vs 0.6
    • PF ratio: 190 vs 184 mm Hg

Intervention

• Settings:
  • Switched to INTELLiVENT-ASV within 1 hour
  • Optimal presets for underlying conditions that could be overridden at any time:
    • Normal: EtCO2 35-45, SpO2 92-96%, PEEP typically 5 cmH2O
    • ARDS: higher %minute volume ~120%, tighter SpO2 targets 88-94%, automated PEEP favouring higher 8-16 cmH2O
    • Chronic hypercapnia: higher EtCO2 45-55, manual PEEP control
    • Brain injury: stricter EtCO2 ~35, higher SpO2 95-99%, manual PEEP
  • Otis and Mead formulas used to determine optimal combination of tidal volume and respiratory rate that minimises work of breathing and mechanical power for a given MV target based on predicted body weight and continually adjusted according to EtCO2 targets
  • FiO2 and PEEP continually adjusted according to SpO2 targets and able to be fine-tuned by clinicians
• Weaning:
  • QuickWean™ monitored continuously for spontaneous breathing and assessed readiness for weaning by progressively reducing %MV targets:
    • Dynamic adjustment EtCO2 targets – shift to +5 higher than during full support
    • Monitor spontaneous breathing frequency: once ≥5 consecutive spontaneous breaths the ventilator adapts supports accordingly.
    • Continuous safety supervision restores previous support level if hypoventilation, apnoea, desaturation, or haemodynamic instability
• SBT:
  • Automated or conventional SBT at clinician’s discretion
  • Integrated Automated SBT
    • Transition to minimal support (PSV 5, PEEP 5), trial duration ~ 30 mins and classified as successful if stability criteria maintained for duration of SBT
    • Failure triggers automatic termination and return to previous support with all changes logged in ventilator history

Control

• Settings:
  • Conventional volume-, pressure-control, or pressure support ventilation modes
  • Adherence to lung-protective ventilation at all times (tidal volumes 6-8 ml/kg, limitation of plateau pressure ≤30 cmH2O,  maintenance of adequate PEEP and maintenance of oxygenation targets (92-96%) using minimum FiO2)
• Weaning:
  • Readiness to transition to PSV assessed at least 3x/day
  • Weaning guided by gradual, protocolled reductions
• SBT:
  • Conventional SBT using T-piece, or PS < 10
  • Successful when RR < 35, SpO2>90%, increase <20% in HR and BP, and without anxiety or diaphoresis sustained at least 30 mins

Management common to both groups

• Readiness for extubation assessed at least 3x/day
• Analgosedation preferred over hypnosedation, daily sedation interruptions actively encouraged
• Clear extubation criteria but this decision was at the discretion of the treating clinician
• Tracheostomy generally avoided within 10 days unless expected prolonged ventilation, impaired consciousness, severe ICU-acquired weakness, or repeated extubation failures occurred

Outcome

• Primary outcome:
• Median number of ventilator-free days at day 28 [death was assigned 0 days]:
  • 16.7 days (IQR 0-26.1 days) ASV vs 16.3 days (IQR 0-26.5 days) conventional group
    • OR: 0.91 [95%CI: 0.77-1.06]
  • No treatment x subgroup interactions identified in pre-specified subgroups (eg. type of admission, cardiac arrest, BMI, severity of illness)
• Secondary outcomes:
• No difference in:
  • Duration of ventilation in survivors: 3.3 vs 2.6 days (MD 0.7, [95%CI -0.0 to 1.4])
  • 28-day mortality 37.2 vs 36.5% (HR 1.04, [95%CI 0.86 – 1.25])
  • ICU, hospital or 90 day mortality
  • ICU or hospital length of stay
  • Extubation failure
  • ARDS, pneumothorax or VAP
• Increased in closed loop group:
  • In patients where enough data was available (13% of patients), closed-loop ventilation was associated with higher odds of being in a more favourable predefined ventilation zone than conventional ventilation ie. ventilation quality (OR 1.5, [95%CI 1.43-1.57], eFigure 7)
    • Largely driven by increased numbers of patients in a critical ventilatory zone in the conventional group (26.9 vs 45.0%, MD -24.6 [95%CI -36.6 – 12.6])
• Reduced in closed loop group on unadjusted analysis:
  • Rates of severe hypoxaemia (PaO2 < 55 mmHg):  16.0 vs 21.1% (AD -4.1%  [95%CI -9.4 – -0.7%]
  • Rates of rescue strategies: 14.4 vs 20.3% (AD -6.0% [95%CI -10.2 – -1.7%])
    • Largely driven by reductions in need for prone positioning
  • Note after adjusting for multiplicity these were no longer significant

Authors’ Conclusions

• In this multicentre trial, there was no difference in ventilator-free days at day 28 between closed-loop ventilation and conventional ventilation when ventilator parameters were systematically managed, and liberation protocols were uniformly applied.

Strengths

• Well conducted trial on an important area in critical care (especially as mechanical ventilation trials can be challenging to run)
• Bi-national trial improves external validity
• Minimal loss to follow-up (1 in closed-loop, 4 in conventional)
• Pre-published trial protocol and statistical analysis plan
• Patient-centred primary outcome
• Balanced co-interventions and processes of care across groups such as sedation, fluid balance (eTable 4, eTable 6)
• Early randomisation minimises any contamination bias
• High adherence to intervention (median of 96% of patient hours in closed loop arm ventilated in correct mode, and 100% in conventional arm, eTable 4)

Weaknesses

• High rate of post-randomisation exclusions, although balanced between groups (~21%), necessitated the use of a modified intention-to-treat analysis however these were for governance reasons and not at the discretion of the clinician
  • No difference in per-protocol analysis (eTable 7)
• Heterogenous population of patients included (25% with neurological dysfunction) with PF ratio of ~190 mmHg and driving pressures of 13 cm H20 – whilst this improves external generalisability this may dilute treatment effect if you believe the greatest group that may benefit from closed loop ventilation are those with pulmonary pathologies (as opposed to those who remain intubated for airway protection) who may require more frequent ventilatory assessment and changes
• Secondary outcomes exploratory only and the outcome of ventilation quality obtained from a subset of patients may not be representative of whole study population
  • Quality of ventilation based on subjective criteria
• Higher rate of patients already started on automated mode prior to randomisation may be due to the participating centres already adopting ASV may lead to performance bias
• Un-blinded and whilst clear parameters for extubation provided this still remained at the discretion of the clinician – no data provided as to the adherence of these criteria
• Control arm group is arguably the provision of high quality ventilatory care and as such this study compares automation to expert protocolised care by clinicians interested in participating in mechanical ventilation trials rather and may not be reflective of usual practice elsewhere
• External validity: both participating countries have high quality healthcare services and high health literacy
  • The trial findings could potentially favour automated closed-loop ventilation in more resource limited settings where staffing and expertise may differ from the trial centres
• Not generalisable to extremes of physiology – trial excluded high BMI, restrictive lung disease, pulmonary resections, and neuromuscular disease

The Bottom Line

• This study found that automated closed-loop ventilation using INTELLiVENT-ASV does not improve ventilator-free days compared to high quality, protocolled conventional ventilation
• Automated closed-loop ventilation could potentially alleviate the need for intense clinician/nursing input while still delivering safe mechanical ventilation
• This trial establishes safety of automated closed-loop modes and opens the door to further research comparing safety and effects on workload

External Links

• article Effect of Automated Closed-Loop Ventilation vs Protocolized Conventional Ventilation on Ventilator-Free Days in Critically Ill Adults
• editorial Automated Modes to Improve Mechanical Ventilation Outcomes—The Ghost in the Machine

Metadata

Summary author: Tim Law
Summary date: 14 March 2026
Peer-review editor: George Walker

Picture by: Pexels/S Starostin