Skip to main content

Foundational Trials

ACTiVE

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Effect of Automated Closed-Loop Ventilation vs Protocolized Conventional Ventilation on Ventilator-Free Days in Critically Ill Adults: A Randomized Clinical Trial
  • Acronym: ACTiVE
  • Year: 2025
  • Journal published in: JAMA
  • Citation: Sinnige JS, Buiteman-Kruizinga LA, Horn J, Paulus F, Schultz MJ, Serpa Neto A, et al; ACTiVE Investigators and the Protective Ventilation Network. Effect of Automated Closed-Loop Ventilation vs Protocolized Conventional Ventilation on Ventilator-Free Days in Critically Ill Adults: A Randomized Clinical Trial. JAMA. Published online December 8, 2025.

Context & Rationale

  • Background
    • Invasive mechanical ventilation is ubiquitous in ICU care, yet preventing ventilator-associated lung injury requires continuous titration of tidal volume, pressure, and oxygenation targets.
    • Closed-loop ventilation systems aim to automate adjustments (e.g., VT/RR/FiO2/PEEP) using physiological feedback (e.g., SpO2, end-tidal CO2) to improve target adherence and reduce workload.
    • Prior studies of closed-loop systems largely focused on process measures (time in target ranges, number of manual setting changes) and were not definitive for patient-centred outcomes in heterogeneous ICU populations.
  • Research Question/Hypothesis
    • Does early, automated, closed-loop ventilation (INTELLiVENT-ASV) increase ventilator-free days (VFDs) at day 28 compared with protocolised conventional ventilation in critically ill adults?
    • Secondary mechanistic hypothesis: closed-loop ventilation improves “ventilation quality” (time within predefined physiologic zones) relative to protocolised conventional care.
  • Why This Matters
    • Automation could standardise lung-protective practice and reduce avoidable hypoxaemia/hyperoxaemia/hypercapnia across variable staffing, workload and expertise.
    • Demonstration of benefit (or harm) in a broad ICU population is essential before closed-loop ventilation is adopted as default rather than a niche tool.

Design & Methods

  • Research Question: In invasively ventilated critically ill adults, does automated closed-loop ventilation (INTELLiVENT-ASV) improve ventilator-free days at day 28 versus protocolised conventional ventilation?
  • Study Type: Multicentre, international, investigator-initiated, randomised, open-label, superiority trial conducted in 7 ICUs (Netherlands and Switzerland).
  • Population:
    • Setting: adult ICUs; randomisation within 1 hour of initiation of invasive ventilation in the ICU.
    • Key inclusion features: critically ill adults receiving invasive ventilation, expected to remain invasively ventilated for at least 24 hours.
    • Key exclusion features: not reported in full in the main manuscript (see protocol for full list); major exclusions included situations where the study mode could not be applied safely or reliably (e.g., practical/device constraints, inability to obtain/maintain consent).
  • Intervention:
    • Automated closed-loop ventilation using INTELLiVENT-ASV on a single ventilator platform, with automated titration of ventilation and oxygenation targets using end-tidal CO2 and SpO2.
    • Implemented immediately after randomisation and used throughout invasive ventilation during the ICU course per study allocation (with clinician oversight and safety overrides available).
  • Comparison:
    • Protocolised conventional ventilation using standard ventilator modes with predefined targets for lung-protective ventilation and oxygenation/ventilation goals.
    • Co-interventions (sedation/analgesia and weaning approach) were protocolised and intended to be similar between groups.
  • Blinding: Open-label (no participant/clinician blinding); outcomes were largely objective, but extubation timing and rescue therapies are potentially clinician-influenced.
  • Statistics: A total of 1200 patients were required to detect an increase in mean ventilator-free days at day 28 of 1.5 days (from 20.0 to 21.5) assuming SD 9.0, with 80% power at the 5% significance level; primary analysis was modified intention-to-treat using an ordinal (cumulative logistic) mixed-effects model with centre as a random effect.
  • Follow-Up Period: 90 days (primary endpoint at day 28).

Key Results

This trial was not stopped early. Recruitment proceeded to the planned sample size for the modified intention-to-treat analysis.

Outcome INTELLiVENT-ASV (closed-loop) Protocolised conventional ventilation Effect p value / 95% CI Notes
Ventilator-free days at day 28 (primary; median [IQR]) 16.7 (0.0–25.0) 16.3 (0.0–25.0) OR 0.91 95% CI 0.77 to 1.06; P=0.23 Ordinal model; OR >1 would favour more VFDs (benefit).
28-day mortality (No./total, %) 224/602 (37.2) 218/597 (36.5) HR 1.04 95% CI 0.86 to 1.25; P=0.69 Holm–Bonferroni corrected P=1.000.
90-day mortality (No./total, %) 246/600 (41.0) 237/592 (40.0) HR 1.04 95% CI 0.87 to 1.24; P=0.66 Holm–Bonferroni corrected P=1.000.
Duration of invasive ventilation among survivors (median [IQR], days) 3.3 (1.0–9.4) 2.6 (1.0–8.7) MdD 0.7 d 95% CI −0.0 to 1.4; P=0.05 Holm–Bonferroni corrected P=0.810.
Ventilation quality (ordered zones; first 6 h; subset) Not reported Not reported OR 1.50 95% CI 1.43 to 1.57; P<0.001 Higher odds of being in a better ventilation zone; Holm–Bonferroni corrected P<0.001.
Time in critical ventilation zone (first 6 h; subset; mean [SD], %) 11.6 (26.9) 35.8 (45.0) MD −24.6% 95% CI −36.6 to −12.6; P<0.001 Subset n=78 vs 64 from 3 centres; multiplicity not applied to zone-level CIs.
Severe hypoxaemia: PaO2 <55 mm Hg (No./total, %) 96/599 (16.0) 126/596 (21.1) AD −5.1% 95% CI −9.4 to −0.7; P=0.02 Holm–Bonferroni corrected P=0.352.
Need for rescue strategies (composite) (No./total, %) 86/599 (14.4) 121/596 (20.3) AD −6.0% 95% CI −10.2 to −1.7; P=0.006 Holm–Bonferroni corrected P=0.108; components included recruitment manoeuvres, prone positioning, bronchoscopy for atelectasis.
Ventilator-associated pneumonia (No./total, %) 13/599 (2.2) 21/596 (3.5) AD −1.4% 95% CI −3.3 to 0.4; P=0.13 Holm–Bonferroni corrected P=1.000.
Pneumothorax (No./total, %) 6/599 (1.0) 2/596 (0.3) AD 0.7% 95% CI −0.3 to 1.6; P=0.16 Holm–Bonferroni corrected P=1.000.
  • Primary outcome: no evidence that closed-loop ventilation improved VFDs at day 28 (OR 0.91; 95% CI 0.77 to 1.06; P=0.23).
  • Mechanistic/process signal: closed-loop ventilation improved ventilation “quality” (better zone distribution; OR 1.50; 95% CI 1.43 to 1.57; P<0.001), driven by less time in the critical zone (11.6% vs 35.8%; MD −24.6%; 95% CI −36.6 to −12.6; P<0.001) in a centre-limited subset.
  • Clinical secondary signals (severe hypoxaemia; rescue strategies) did not remain statistically significant after Holm–Bonferroni adjustment (corrected P=0.352 and 0.108, respectively).

Internal Validity

  • Randomisation and Allocation:
    • Central randomisation with centre as a stratification factor; centre included as a random effect in analyses.
  • Drop out or exclusions:
    • Randomised: 1514 participants.
    • Included in modified intention-to-treat analysis: 1201 (602 closed-loop; 599 conventional); 313 excluded post-randomisation due to withdrawal/non-obtained consent (150 vs 163).
    • Primary outcome denominators were slightly lower due to missing data (e.g., VFDs: n=601 vs 595; as reported in the trial tables).
  • Performance/Detection Bias:
    • Open-label design creates potential for clinician-influenced co-interventions (e.g., rescue strategies, extubation timing), although major outcomes were objective and co-interventions were protocolised.
  • Protocol Adherence:
    • Clear separation in exposure: patient-level percentage of ventilated days on INTELLiVENT-ASV was 100.0% (IQR 80.0–100.0) in the closed-loop group vs 0.0% (0.0–0.0) in conventional ventilation.
    • Cross-use at 1 hour: INTELLiVENT-ASV was used in 23/594 (3.9%) of patients in the conventional group.
  • Baseline Characteristics:
    • Baseline physiology and illness severity were broadly similar between groups (e.g., height median 175 cm in both; weight median 80 kg in both; APACHE IV median 85 vs 82).
  • Heterogeneity:
    • Broad ICU population; low incidence of ARDS developing after 48 hours (2.5% vs 2.0%), which may limit ability to detect benefit in subphenotypes with greater ventilator-management sensitivity.
  • Timing:
    • Early initiation: randomisation and delivery occurred within 1 hour of starting invasive ventilation in ICU.
  • Dose:
    • At 1 hour post-randomisation, ventilation modes differed substantially (INTELLiVENT-ASV 72.3% vs 3.9%; pressure control 20.8% vs 76.1%), supporting early separation in the delivered exposure.
  • Separation of the Variable of Interest:
    • Mode exposure: INTELLiVENT-ASV 100.0% (80.0–100.0) vs 0.0% (0.0–0.0) of ventilated days.
    • Ventilation quality subset: critical-zone time 11.6% (SD 26.9) vs 35.8% (SD 45.0) over the first 6 hours.
  • Outcome Assessment:
    • VFDs combine survival and ventilation duration; the trial also reported mortality and ventilation duration components to support interpretation.
  • Statistical Rigor:
    • Primary analysis used a prespecified ordinal mixed model appropriate for the bounded/zero-inflated VFD distribution; multiplicity control applied to secondary outcomes (Holm–Bonferroni procedure).

Conclusion on Internal Validity: Moderate: robust randomisation and strong intervention–control separation support causal inference, but the open-label design and substantial post-randomisation exclusions (modified ITT) introduce plausible risks of bias and reduce interpretability for patient-centred outcomes.

External Validity

  • Population Representativeness:
    • High-income, protocol-driven ICUs in 2 European countries; findings most applicable to similar settings with established lung-protective ventilation culture.
    • Heterogeneous ICU case-mix with relatively low ARDS frequency; applicability to ARDS-enriched cohorts remains uncertain.
  • Applicability:
    • Intervention is platform-specific (INTELLiVENT-ASV on a single ventilator family), limiting direct generalisability to other devices/algorithms.
    • Potentially greater impact in settings with less protocolised ventilation or where staffing/workload constrains frequent manual titration.

Conclusion on External Validity: Moderate and context-dependent: generalisable to high-resource ICUs using similar ventilator platforms and protocolised care, but uncertain transfer to low-resource settings, different devices, or ARDS-dominant populations.

Strengths & Limitations

  • Strengths:
    • Large, multicentre, international RCT with early initiation of the assigned ventilation strategy.
    • Clear separation of the intervention exposure with minimal cross-use in the control group.
    • Prespecified statistical approach for a bounded, skewed primary endpoint and multiplicity control for secondary outcomes.
    • High follow-up completion through 90 days as reported in the manuscript.
  • Limitations:
    • Open-label design with clinician-mediated components (rescue therapies; extubation decisions) that could influence composites such as VFDs.
    • Substantial post-randomisation exclusions due to consent withdrawal/non-obtainment, resulting in a modified ITT analysis set.
    • Ventilation quality analysis was conducted in a subset from 3 centres, potentially limiting generalisability of mechanistic findings.
    • Control arm used protocolised conventional ventilation (not “usual care”), potentially reducing detectable effect size in a high-performing system.

Interpretation & Why It Matters

  • Clinical effect
    • Automated closed-loop ventilation did not improve VFDs at day 28 and did not reduce mortality (28-day: HR 1.04; 95% CI 0.86 to 1.25; P=0.69).
  • Mechanistic/process effect
    • Closed-loop ventilation improved early ventilation “quality” in a subset (critical-zone time 11.6% vs 35.8%; MD −24.6%; 95% CI −36.6 to −12.6; P<0.001), supporting that the algorithm meaningfully altered delivered ventilation.
  • Practice implication
    • In high-performing ICUs with protocolised conventional ventilation, automation may improve process metrics without translating into more ventilator-free days.
    • Implementation decisions should consider device availability, staffing/workload, and whether local usual care leaves meaningful room for improvement.

Controversies & Subsequent Evidence

  • Post-randomisation exclusions due to deferred/withdrawn consent (313/1514 randomised) shift interpretation away from a strict intention-to-treat framework and increase vulnerability to bias if exclusions are not exchangeable between groups. 123
  • Ventilator-free days are a composite with structural zeros (death) and a bounded, often skewed distribution; ACTiVE’s ordinal modelling approach aligns with modern methodological recommendations, but component outcomes (mortality and duration among survivors) remain essential for clinical interpretation. 45
  • A key “success” signal (improved ventilation quality) was derived from a centre-limited subset, raising questions about representativeness and the extent to which early process improvements translate into patient-centred outcomes at scale. 1
  • Systematic reviews suggest closed-loop strategies can reduce manual setting changes and improve time in target ranges, with uncertain effects on mortality; ACTiVE materially strengthens the evidence base by testing a clinically meaningful endpoint in a large heterogeneous ICU cohort. 67
  • International observational data demonstrate wide variation in liberation practices and outcomes; automation might plausibly have greater marginal benefit where adherence to lung-protective targets or weaning protocols is inconsistent. 8
  • Contemporary guidelines for ARDS management and ventilator liberation continue to emphasise lung-protective targets and protocolised liberation rather than recommending closed-loop ventilation as routine default, reflecting the current evidentiary gap for patient-centred benefit. 910

Summary

  • ACTiVE randomised 1514 patients and analysed 1201 (modified ITT) across 7 ICUs in 2 countries, comparing INTELLiVENT-ASV closed-loop ventilation with protocolised conventional ventilation.
  • There was no improvement in ventilator-free days at day 28 (median 16.7 vs 16.3; OR 0.91; 95% CI 0.77 to 1.06; P=0.23).
  • Mortality was similar at 28 days (37.2% vs 36.5%; HR 1.04; 95% CI 0.86 to 1.25; P=0.69) and 90 days (41.0% vs 40.0%; HR 1.04; 95% CI 0.87 to 1.24; P=0.66).
  • Closed-loop ventilation improved early ventilation “quality” (better zone distribution; OR 1.50; 95% CI 1.43 to 1.57; P<0.001) with markedly less time in the critical zone (11.6% vs 35.8%) in a centre-limited subset.
  • Signals for fewer rescue strategies and severe hypoxaemia did not remain statistically significant after multiplicity correction (corrected P=0.108 and 0.352, respectively).

Overall Takeaway

ACTiVE is a landmark pragmatic evaluation of a commercially available, fully automated closed-loop ventilation system in a broad ICU population: it demonstrates that, in protocolised high-performing ICUs, automation improves early physiologic “quality” metrics yet does not increase ventilator-free days or reduce mortality. The trial therefore reframes closed-loop ventilation as a potential process optimiser whose patient-centred benefit is likely to be context-, population-, and implementation-dependent rather than universal.

Overall Summary

  • No improvement in ventilator-free days at day 28 with INTELLiVENT-ASV vs protocolised conventional ventilation.
  • Ventilation quality improved early (better zone distribution) in a centre-limited subset, but did not translate into better patient-centred outcomes.
  • Signals for reduced rescue strategies and severe hypoxaemia were attenuated after multiplicity correction.

Bibliography