Skip to main content

Trial Summaries & Critiques

RESCUE-3

Image: Shutterstock / Kiryl Liss

ICU patient with multi-organ failure

Publication

  • Title: Temporary Transvenous Diaphragm Neurostimulation for Weaning from Mechanical Ventilation (RESCUE-3): A Randomized Clinical Trial
  • Acronym: RESCUE-3
  • Year: 2026
  • Journal published in: American Journal of Respiratory and Critical Care Medicine
  • Citation: Dres M, Ewert R, Conrad SA, Ataya A, Shrager J, Mortaza S, et al; RESCUE-3 Trial Investigators. Temporary transvenous diaphragm neurostimulation for weaning from mechanical ventilation (RESCUE-3): a randomized clinical trial. Am J Respir Crit Care Med. 2026;212(1):86-94.

Context & Rationale

  • Background
    • Prolonged invasive mechanical ventilation and difficult liberation are common, heterogeneous problems internationally, with major impacts on mortality, morbidity, ICU length of stay, and resource use1.
    • Diaphragm weakness (including ventilator-induced diaphragmatic dysfunction and critical-illness–associated respiratory muscle weakness) is prevalent in patients with prolonged ventilation and is biologically plausible as a contributor to weaning failure.
    • Temporary transvenous diaphragm neurostimulation (TTDN) aims to activate the diaphragm via phrenic nerve stimulation delivered through a central venous catheter, potentially improving inspiratory muscle function during a period of established weaning difficulty.
    • Early feasibility work (RESCUE 1) demonstrated technical feasibility of TTDN in prolonged mechanically ventilated patients, with physiological signals suggesting potential to improve inspiratory strength2.
    • A subsequent multicentre randomised study (RESCUE-2) did not demonstrate superiority for its primary clinical endpoint in a difficult-to-wean population, but maintained interest in whether patient selection, timing, and trial design could reveal a clinically meaningful effect3.
    • The transvenous diaphragm pacing programme (including design considerations around endpoint choice, weaning definitions, and procedural delivery) was pre-specified in earlier protocol work, highlighting persistent uncertainty about efficacy and safety in this population4.
  • Research Question/Hypothesis
    • In adults receiving invasive mechanical ventilation for ≥96 hours who have failed ≥2 weaning attempts and have reduced inspiratory strength (maximal inspiratory pressure ≤50 cm H2O), does adding twice-daily TTDN to protocolised standard care increase the probability of successful liberation from mechanical ventilation by Day 30 compared with standard care alone?
  • Why This Matters
    • Even modest improvements in liberation outcomes in a prolonged-ventilation population could translate into clinically meaningful reductions in ventilator exposure and ICU resource utilisation.
    • TTDN introduces a device-based rehabilitation concept that, if effective and safe, could represent a distinct therapeutic class beyond traditional spontaneous breathing trials, sedation minimisation, and conventional rehabilitation strategies.
    • Because prolonged weaning is internationally prevalent and strongly associated with adverse outcomes, robust evidence is needed before adoption of invasive device-based approaches1.

Design & Methods

  • Research Question: In difficult-to-wean adults on invasive mechanical ventilation with inspiratory weakness, does temporary transvenous diaphragm neurostimulation plus standard care increase successful weaning by Day 30 versus standard care alone?
  • Study Type: International, multicentre, open-label, randomised controlled trial conducted in ICUs at 48 centres (United States and Europe); Bayesian group sequential design incorporating prespecified borrowing from a prior related trial.
  • Population:
    • Setting: ICU patients receiving invasive mechanical ventilation.
    • Key inclusion: adults ≥18 years; invasive mechanical ventilation >96 hours; met predefined readiness-to-wean criteria; failed ≥2 weaning attempts occurring ≥48 hours after initiation of mechanical ventilation and on different days; maximal inspiratory pressure (absolute value) ≤50 cm H2O.
    • Key exclusion: invasive mechanical ventilation >90 days; extracorporeal membrane oxygenation; weaning failure judged secondary to volume overload; contraindication to central venous catheter placement; implanted electrical devices that could interfere with stimulation.
  • Intervention:
    • Temporary transvenous diaphragm neurostimulation via a dedicated stimulation catheter placed in the left subclavian or internal jugular vein.
    • Stimulation schedule: twice daily sessions, at least 3 hours apart; each session comprised 6 sets of 10 stimulations (12 sets/day; ~120 stimulations/day).
    • Stimulation parameters: pulse frequency 15 Hz; pulse duration 200–300 microseconds; intensity titrated up to the highest tolerated level (maximum 27 mA) to achieve visible diaphragm contraction.
    • Continued daily until successful weaning, Day 30, a safety event, withdrawal, or another stopping condition.
  • Comparison:
    • Protocolised standard care for weaning, including once-daily weaning trials when readiness criteria were met; sedation minimisation and mobilisation encouraged.
    • Inspiratory muscle strength training was prohibited.
    • After successful weaning trial and liberation from invasive ventilation, clinicians could use non-invasive ventilation and/or high-flow nasal oxygen to prevent post-extubation respiratory failure.
  • Blinding: Open-label (no sham catheter); primary efficacy outcome used protocolised criteria, but key clinical decisions (e.g., extubation, tracheostomy management, reconnection) could not be blinded.
  • Statistics: Bayesian group sequential design with a maximum planned sample size of 400; interim analyses beginning at n=150 and after each additional 50 participants; superiority assessed using posterior probability thresholds (98.25% at interim 1 increasing to 99.25% at later analyses) rather than a frequentist alpha; operating-characteristic simulations targeted >85% power to detect an absolute increase in Day-30 successful weaning from 68% to 80%; primary analysis prespecified in a modified intention-to-treat population (randomised to intervention with successful catheter placement and ability to stimulate ≥1 phrenic nerve), with sensitivity analyses in per-protocol and intention-to-treat populations.
  • Follow-Up Period: 30 days for clinical outcomes; serious adverse events tracked until 48 hours after catheter removal.

Key Results

This trial was stopped early. Enrolment was halted after the first interim analysis because of slower-than-expected enrolment and the need for substantial additional funding to complete the planned sample size.

Outcome TTDN + standard care Standard care Effect p value / 95% CI Notes
Successful weaning by Day 30 (≥48 h off invasive ventilation) 71/102 (69.6%) 69/114 (60.5%) HR 1.34 95% CrI 1.01 to 1.78; posterior probability 97.9% Primary endpoint; modified intention-to-treat; Bayesian model with prespecified borrowing from RESCUE-2.
Duration of invasive ventilation to successful weaning or Day 30 (days) 13.0 (6.0–29.0) 18.0 (7.0–29.0) Absolute mean difference −2.4 d 95% CrI −5.0 to 0.1; posterior probability 97.1% Modified intention-to-treat; censored at Day 30 for those not weaned.
Ventilator-free days to Day 30 (days) 16.0 (0.0–22.8) 6.0 (0.0–22.0) OR 1.61 95% CrI 1.03 to 2.41; posterior probability 98.2% Modified intention-to-treat; model-based estimate reported by authors.
Day 30 mortality 10/102 (9.8%) 12/114 (10.5%) HR 0.74 95% CrI 0.37 to 1.46; posterior probability 80.6% Modified intention-to-treat; not powered for mortality.
Change in maximal inspiratory pressure from baseline to last measurement (cm H2O) 12.8 (0.0–23.6) 7.7 (0.0–19.1) Difference 5.9 95% CrI 1.8 to 10.0; posterior probability 99.8% Modified intention-to-treat; physiological endpoint supporting separation of mechanism.
Serious adverse events (patients with ≥1 SAE) 39/109 (35.8%) 27/114 (23.7%) Not reported Not reported Intention-to-treat safety population; includes 2 deaths assessed as possibly related to catheter placement.
Cardiac serious adverse events 12/109 (11.0%) 2/114 (1.8%) Not reported Not reported Intention-to-treat safety population.
  • TTDN increased the Bayesian probability of Day-30 successful weaning (69.6% vs 60.5%; HR 1.34; 95% CrI 1.01 to 1.78; posterior probability 97.9%), but the trial did not reach its planned sample size because of early termination.
  • Ventilator-free days to Day 30 were higher with TTDN (median 16.0 vs 6.0 days; OR 1.61; 95% CrI 1.03 to 2.41), while Day-30 mortality was similar (9.8% vs 10.5%; HR 0.74; 95% CrI 0.37 to 1.46).
  • Safety signals favoured standard care: serious adverse events occurred in 35.8% with TTDN vs 23.7% with standard care, and two deaths were assessed as possibly related to catheter placement.

Internal Validity

  • Randomisation and allocation:
    • Central electronic randomisation with site stratification and random permuted blocks; allocation concealment likely adequate up to assignment.
    • Open-label delivery after allocation created potential for performance and detection bias, particularly for clinician-driven weaning decisions.
  • Dropout/exclusions and analysis sets:
    • 223 patients randomised (TTDN n=109; control n=114), but the prespecified primary analysis used modified intention-to-treat (TTDN n=102; control n=114) excluding 7/109 (6.4%) allocated to TTDN with unsuccessful catheter placement or withdrawal before the procedure.
    • Per-protocol population required ≥50% of stimulations; 90/102 (88.2%) met this threshold.
  • Protocol adherence and separation of the variable of interest:
    • TTDN delivery: 81/102 (79.4%) continued stimulation until the end of the study; mean proportion of required stimulations delivered was 79.3% (±25.0).
    • Dose/intensity: mean stimulation intensity 21.4 (±9.1) mA; mean stimulations per day 120.7 (±18.2); mean total stimulations delivered 1,415 (±945) versus required 1,799 (±1,073).
    • Control group: no study stimulation catheter; this creates clear intervention separation but also introduces differential exposure to an invasive catheter procedure.
  • Baseline comparability and illness severity:
    • Groups were broadly similar: age 64.6 (±13.3) vs 63.8 (±14.2) years; male sex 67.6% vs 66.7%; baseline maximal inspiratory pressure 30.0 (±11.9) vs 29.0 (±12.9) cm H2O.
    • Prolonged ventilation at randomisation: days of invasive mechanical ventilation 27.8 (±18.3) vs 30.3 (±18.9); tracheostomy present in 56.9% vs 63.2%.
  • Outcome assessment:
    • Primary endpoint (≥48 hours off invasive ventilation) is relatively objective once liberation occurs, but the timing and decision to attempt liberation remain clinician-mediated and unblinded.
    • Serious adverse events were adjudicated; differential catheter exposure may influence event ascertainment and true event rates.
  • Statistical rigour:
    • Prespecified Bayesian design with borrowing and interim monitoring was implemented, but early termination reduced precision and left credibility heavily dependent on modelling assumptions.
    • Effect estimates were reported with 95% credible intervals and posterior probabilities rather than p values, consistent with the design.

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation and protocolisation support causal inference, but open-label delivery, the prespecified modified intention-to-treat exclusion of unsuccessful catheter placements, early stopping, and reliance on Bayesian borrowing introduce non-trivial risks of bias and imprecision.

External Validity

  • Population representativeness:
    • Participants were a selected, severely prolonged ventilation cohort (mean ~28–30 days ventilated at randomisation), with a high prevalence of tracheostomy and repeated failed weaning attempts, representing a specific and resource-intensive subgroup of ICU patients.
    • Exclusions (e.g., contraindication to central venous catheter placement, ECMO, very prolonged ventilation >90 days) limit generalisability to some high-risk or specialised populations.
  • Applicability:
    • Requires availability of trained operators for catheter placement, stimulation titration, and monitoring; this may limit uptake outside centres with established prolonged-weaning expertise and device access.
    • Findings do not directly extrapolate to earlier phases of mechanical ventilation, to patients with preserved inspiratory strength, or to units where protocolised weaning practices differ substantially.

Conclusion on External Validity: Overall generalisability is moderate for ICUs managing prolonged, difficult-to-wean patients with access to invasive neuromodulation capability, and limited for routine short-duration ventilation pathways or resource-limited settings.

Strengths & Limitations

  • Strengths:
    • International, multicentre randomised design with a prespecified protocolised approach to readiness-to-wean and weaning trials.
    • Clinically meaningful primary endpoint focused on liberation from invasive mechanical ventilation by Day 30, complemented by ventilator-free days and mechanistic measures (maximal inspiratory pressure; rapid shallow breathing index).
    • Transparent reporting of Bayesian credible intervals and posterior probabilities aligned to a prespecified group sequential framework.
  • Limitations:
    • Stopped early for operational/financial reasons, leading to reduced precision and incomplete fulfilment of prespecified sample size and monitoring plan.
    • Open-label design without sham procedure; potential for performance bias in clinician-driven liberation decisions.
    • Primary analysis in a modified intention-to-treat population excluding unsuccessful catheter placement/early withdrawal in the intervention arm (7/109), with potential to bias efficacy estimates.
    • Safety concerns: higher serious adverse event rates in the intervention arm and two deaths assessed as possibly related to catheter placement.

Interpretation & Why It Matters

  • Clinical signal
    • TTDN produced a favourable Bayesian signal for successful weaning by Day 30 (69.6% vs 60.5%) and for ventilator-free days (median 16.0 vs 6.0), without an apparent mortality signal at Day 30.
  • Mechanistic plausibility
    • Between-group separation in inspiratory strength was modest (maximal inspiratory pressure change difference 5.9 cm H2O), raising questions about whether measured inspiratory strength improvements fully explain the observed liberation signal6.
  • Implementation caution
    • Any clinical adoption must weigh potential liberation benefits against procedural complications and higher serious adverse event rates, particularly given the invasive catheter-based delivery and two possibly procedure-related deaths.

Controversies & Other Evidence

  • Early stopping and interpretability
    • The trial was terminated for operational and funding reasons rather than crossing a prespecified superiority or futility boundary, leaving the final inference dependent on an under-recruited dataset with wide uncertainty.
    • Editorial commentary highlighted that early termination complicates interpretation and increases the importance of replication in a fully powered trial, particularly for patient-centred outcomes and longer-term endpoints6.
  • Bayesian borrowing and “evidence amplification”
    • The prespecified primary analysis incorporated downweighted, propensity-matched external information from RESCUE-2, which can improve efficiency but also makes conclusions sensitive to assumptions about transportability across trials.
    • This approach is methodologically defensible when pre-specified, but the degree to which posterior probability reflects RESCUE-3’s own data versus borrowed information remains a central interpretive issue for trialists and methodologists37.
  • Analysis population and potential bias
    • Use of modified intention-to-treat excludes patients allocated to TTDN who could not be stimulated because catheter placement was unsuccessful or the procedure did not occur; if these exclusions correlate with prognosis, efficacy estimates may be biased.
    • Open-label weaning decisions may interact with the protocolised weaning framework, potentially favouring earlier liberation attempts in the intervention arm.
  • Safety signal and procedural risk
    • Serious adverse events occurred in 35.8% with TTDN vs 23.7% with standard care, including more cardiac SAEs (11.0% vs 1.8%).
    • Two deaths were assessed as possibly related to catheter placement (haemo/pneumohaemothorax following catheter placement; and acute coronary syndrome associated with a tension pneumothorax after catheter insertion).
  • Subsequent evidence and positioning within the wider field
    • Earlier programme evidence (RESCUE 1 feasibility and RESCUE-2 randomised study) provides important context for interpreting RESCUE-3’s signals and uncertainties23.
    • Related rehabilitation strategies show mixed efficacy: high-intensity inspiratory muscle training did not improve weaning success versus sham in IMweanT, underscoring the difficulty of translating physiological improvements into robust liberation benefits5.
    • Beyond the weaning phase, ongoing work explores diaphragm neurostimulation-assisted ventilation to maintain diaphragm activity during passive ventilation (e.g., STIMULUS phase 1 feasibility), which informs mechanistic plausibility but does not establish weaning efficacy1011.
    • Contemporary guideline documents emphasise protocolised liberation practices, mobilisation, and rehabilitation; none currently provides specific recommendations for catheter-based diaphragm neurostimulation in weaning (guidelines largely pre-date RESCUE-3)89.

Summary

  • RESCUE-3 tested twice-daily temporary transvenous diaphragm neurostimulation (TTDN) versus standard care in a prolonged, difficult-to-wean mechanically ventilated population with inspiratory weakness.
  • The trial was stopped early for operational and funding reasons, before reaching its planned maximum sample size.
  • Primary outcome favoured TTDN: Day-30 successful weaning (≥48 h off ventilation) occurred in 69.6% vs 60.5% (HR 1.34; 95% CrI 1.01 to 1.78; posterior probability 97.9%).
  • TTDN was associated with higher ventilator-free days (median 16.0 vs 6.0) and a modest increase in inspiratory strength (maximal inspiratory pressure change difference 5.9 cm H2O).
  • Safety concerns were prominent: serious adverse events were more frequent with TTDN (35.8% vs 23.7%), and two deaths were assessed as possibly related to catheter placement.

Overall Takeaway

RESCUE-3 provides the strongest randomised evidence to date that temporary transvenous diaphragm neurostimulation may improve liberation outcomes in a highly selected prolonged-weaning population, but its early termination and open-label, model-dependent inference limit certainty. The liberation signal must be weighed against an important safety signal and invasive procedural risks, making confirmatory, adequately powered trials with longer-term outcomes and robust safety evaluation essential before routine adoption.

Overall Summary

  • TTDN showed a favourable Bayesian signal for Day-30 weaning success and ventilator-free days in prolonged weaning, but with increased serious adverse events and early trial termination.

Bibliography