From physiological innovation to clinical precision: defining the right target for diaphragm neurostimulation

Difficult weaning from mechanical ventilation remains one of the most challenging situations in intensive care: prolonged dependency, accumulated complications, and a persistent sense that the therapeutic toolbox is running thin (1). Once patients fail two or more spontaneous breathing trials after more than four days on the ventilator, the options available are largely reactive. Diaphragm neurostimulation has been positioned, for some time now, as a potential way forward, an approach grounded in physiology.

RESCUE-3 gives that rationale new standing (2). This randomized trial enrolled 223 patients (216 randomized) across 48 centers in the United States and Europe and found that twice-daily transvenous phrenic nerve stimulation increased successful weaning at Day 30: 70% of the treatment group versus 61% of controls were successfully weaned, with a median adjusted hazard ratio (HR) of 1.34 (95% credible interval 1.01–1.78) and a posterior probability of superiority of 97.9%. Duration of mechanical ventilation was reduced by approximately 2.5 days. These are important numbers in a population where patients had already been ventilated for a mean of 29 days at enrollment, and where 64% had a tracheostomy.

Yet, the trial deserves comprehensive unpacking. Although the results are exciting, they lie in a context that generates as many questions as it resolves. The population is broad and heterogeneous, the timing of intervention is late, the statistical approach involves borrowing from a prior trial, and serious adverse events were more frequent in the treatment group. This does not dismiss the findings; it is a reason to think about where the technique should be positioned: early versus late, continuous versus intermittent, in a more precise, homogenous population, versus in the diverse population studied in the trial.

Clinical perspective

Difficult weaning is not a diagnosis (3). It is a clinical circumstance, a syndromic entity, a collection of patients who keep failing spontaneous breathing trials or extubation attempts on their own for reasons that vary considerably between individuals. The RESCUE-3 inclusion criteria capture this well enough: mechanical ventilation for more than 96 hours, two failed weaning attempts, readiness-to-wean criteria met. These criteria are clinically recognizable, but mechanistically nonspecific.

Among the 216 randomized patients, half were ventilated for acute respiratory distress syndrome (ARDS) or acute respiratory failure, roughly 12% for chronic obstructive pulmonary disease (COPD), 12% for pneumonia, and 15% for surgical reasons (2). Mean baseline maximal inspiratory pressure (MIP) was 29 cmH₂O, below the 50 cmH₂O exclusion threshold but variable enough across individuals to suggest that diaphragm weakness was neither uniformly severe nor uniformly the dominant problem (4). A potential issue may lie in the fact that an intervention targeting one specific mechanism was tested in a population where that mechanism coexists with other drivers of weaning failure: residual lung injury, cardiac dysfunction, intensive care unit (ICU)-acquired weakness, altered respiratory drive (5). Do we observe a signal showing the success of neurostimulation to treat or to prevent critical illness associated diaphragmatic dysfunction? Observing a significant effect under these conditions is encouraging. But the effect is probably larger in a more precisely identified subgroup, and the current design cannot reveal that. The trial makes no attempt to stratify based on diaphragm-specific phenotyping, while standardized noninvasive measures are available (6). No systematic assessment of contractility by ultrasound, no characterization of inspiratory effort at baseline, no effort to identify patients in whom diaphragm dysfunction was the primary weaning barrier rather than one contributor among several (7). Such characterization would allow patients to be differentiated according to the dominant mechanism of weaning failure and would reduce the heterogeneity that currently limits interpretation of the treatment effect.

Moreover, the trial tested a late intervention in a late stage: a mean of 29 days on mechanical ventilation at randomization, ranging up to 86 days. By that point, diaphragm atrophy is not a risk to be prevented; it is an established reality for many patients. Replacement fibrosis has been documented histologically in mechanically ventilated patients (8). Myofibrillar hibernation from super-relaxed myosin accumulation has been described in this population. The physiological literature frames diaphragm neurostimulation largely as a preventive or early-phase strategy: maintaining contractile activity during controlled ventilation, before significant atrophy consolidates (9-11). RESCUE-3 tests something different, a rescue intervention in patients who are already deep into the consequences of prolonged ventilation. Both questions matter, but answering the second one does not tell us what would happen if the technique was applied earlier. The fact that a consistent signal emerges in patients ventilated for nearly a month on average suggests that rehabilitative capacity persists even late. But it may mean that the technique is applied under less favorable biological conditions.

For an invasive approach, the risk-benefit balance needs a thorough assessment. Serious adverse events occurred in 36% of treatment patients and 24% of controls. The dominant difference was cardiac: 11% versus 2% for cardiac serious adverse events. Two patients in the treatment group died from events adjudicated as possibly related to the device or procedure, one from hemothorax and pneumohemothorax, one from acute coronary syndrome secondary to tension pneumothorax at catheter placement (2). The authors raise ascertainment bias as a partial explanation. Treatment patients all had a central venous catheter in place, which was not true for all controls, and catheter-related events would naturally concentrate in the treatment arm. In RESCUE-2, where the Clinical Events Committee adjudicated with blinding to group assignment, the rates of severe adverse events were similar between arms (12). These are legitimate mitigating arguments, that may suggest that the open-label design contributed to the apparent difference. They do not fully resolve the concern. Although a reduction of 2.5 days of ventilation is clinically meaningful, it is not clear whether it justifies a 36% rate of serious adverse events. The risk-benefit balance probably varies meaningfully across patients in such a heterogeneous population. If a substantial proportion of included patients did not need the intervention in the first place, the observed benefit is diluted while the procedural risk is not. Better patient selection and safer device are probably the most important next steps.

Physiological perspective

The physiological rationale of diaphragm neurostimulation is better established than the clinical evidence at this point, and understanding it helps explain some findings in RESCUE-3 that might otherwise seem paradoxical. Positive pressure ventilation inverts normal respiratory mechanics (13). Inflation is driven by airway pressure rather than diaphragm contraction. Under sedation, the diaphragm becomes quiescent and undergoes a cephalic displacement, particularly in dorsal regions (14). Dorsal regions, the most dependent in the supine patient, receive less ventilation and are prone to atelectasis (Figure 1). Mechanical stress and strain concentrate in non-dependent zones (15). This spatial heterogeneity is a driver of ventilator-induced lung injury. Higher positive end-expiratory pressure (PEEP) recruits some dependent regions but cannot recreate the mechanical pattern of diaphragmatic inflation, and it does so at the cost of hemodynamic compromise (15,16).

Figure 1 Physiological consequences of positive-pressure mechanical ventilation with and without diaphragm neurostimulation. Without neurostimulation, diaphragm inactivity causes atrophy and dorsal atelectasis, concentrating ventilation in non-dependent zones and promoting regional strain and lung injury. Neurostimulation restores diaphragm contraction, redistributes ventilation to dependent regions, and improves hemodynamic tolerance of higher PEEP. PEEP, positive end-expiratory pressure.

The STIMULUS trial, a physiological study in 19 patients, provides the clearest data on this (17,18). Diaphragm neurostimulation redistributed tidal ventilation to dorsal regions in a dose-dependent fashion: at maximum stimulation, 63% of tidal volume reached dependent lung zones, compared with 43% during passive ventilation (P<0.001). Increasing PEEP alone, without stimulation, barely altered this distribution. Ventilation patterns during each breath depend primarily on the diaphragm contraction; pressure-based interventions cannot adequately replicate this. Earlier findings from Parfait et al. in 12 ARDS patients had pointed in the same direction, with continuous neurostimulation during volume-controlled ventilation: driving pressure decreased by 21%, and the dorsal-to-ventral ventilation ratio shifted from 0.70 to 1.20 (19).

These are substantial physiological effects for only 120 stimulations per day, a dose the RESCUE-3 investigators describe as a tiny fraction of the approximately 28,000 contractions a healthy diaphragm performs daily. However, the stimulation protocol itself deserves further consideration. The parameters used in RESCUE-3 were directly inherited from the RESCUE-2 trial, which was primarily designed to establish feasibility and safety, not to define an optimal dosing strategy. As such, the current approach appears more empirical than physiology-driven. Neurostimulation was delivered through a temporary transvenous catheter capturing the phrenic nerve, applied in two daily sessions for a total of approximately 120 stimulations per day, with stimulation intensity titrated in each patient to achieve phrenic nerve capture (2). The delivered stimulation represents less than 1% of physiological diaphragm activity and is administered intermittently, raising questions about its ability to sustain physiological effects. Moreover, stimulation intensity is titrated to achieve phrenic nerve capture, and no dose-response relationship has been established. In this context, neurostimulation remains a proof of concept rather than a fully optimized therapy.

Neurostimulation restores negative-pressure physiology within a positive-pressure system. The diaphragm contracts, dependent lung preferentially inflates, ventilation becomes more homogeneous, and regional stress and strain are redistributed. Recruitment is achieved through mechanics rather than through pressure. However, in STIMULUS, increasing stimulation was associated with progressively increasing Pendelluft, gas movement between lung regions during a single breath, reaching 20% of tidal volume at maximum stimulation (17). This reflects dynamic differences in regional pressures that could injure the lung in distinct ways from those the technique is preventing (20). Low-to-moderate stimulation levels may provide beneficial redistribution of ventilation, while higher levels introduce potential harm. This is an important design consideration for future trials, and it is relevant to interpreting RESCUE-3: the 120 daily stimulations used represent a specific dose choice, not necessarily the optimal one.

The findings of the STIMULUS dataset on PEEP deserve attention (17). The lung recruitment effects of neurostimulation were present at higher PEEP (median 14 cmH₂O) and largely absent at lower PEEP (median 8 cmH₂O). End-expiratory lung volume increased by a median of 1014 mL at high PEEP versus 86 mL at low PEEP (P-for-interaction =0.03). Respiratory system elastance decreased from 1.4 to 1.2 cmH₂O per mL per kg at high PEEP and was unchanged at low PEEP (P-for-interaction =0.02). At the same time, neurostimulation mitigated the hemodynamic cost of higher PEEP (18). Cardiac index increased with stimulation to a median of 2.7 L/min/m² at high PEEP, versus 2.1 L/min/m² without stimulation at the same PEEP level (P=0.03). The two interventions appear physiologically synergistic: PEEP stabilizes recruited alveoli during expiration; neurostimulation facilitates their opening during inspiration through diaphragmatic contraction and preserves the hemodynamics that higher PEEP would otherwise compromise. In patients with right ventricular dysfunction, a common complication in ventilated patients, the cardiovascular effects of neurostimulation could carry clinical relevance.

In RESCUE-3, the MIP increase was 5.9 cmH₂O greater in the treatment group, statistically significant with a posterior probability of 99.8% but modest in terms of clinical effect (2). That raises an important question: how can such a small improvement in inspiratory muscle strength explain such a significant difference in weaning success (21)? A likely explanation is that the clinical benefit operates through a broader physiological pathway: improved ventilation distribution, reduced regional mechanical stress, better cardiopulmonary interaction. RESCUE-3 was not designed to disentangle these mechanisms. The mismatch between a 5.9 cmH₂O MIP gain and a clinically meaningful weaning outcome is perhaps the most interesting empirical observation the trial produces. It points toward a mechanism more complex, and more interesting, than simple muscle rehabilitation, and it suggests that measuring strength alone is insufficient to understand the intervention’s consequences (22). Future trials should therefore incorporate longitudinal functional imaging of the diaphragm to track recovery over time and to clarify whether the benefit reflects muscle rehabilitation, improved regional ventilation, or both.

Statistical perspective

RESCUE-3 used a Bayesian adaptive design with a power prior borrowing from RESCUE-2 (23,24). The downweighting procedure, based on propensity matching, was designed to account for population differences between the two trials. The critical sensitivity analysis is conducted without any borrowing; in that scenario, the posterior probability of superiority decreases to 91.5%, and the HR shifts to 1.26, with a 95% credible interval of 0.91 to 1.75 (2). It clarifies the evidence available from RESCUE-3 data alone, without the prior: a compatible but weaker signal. The trial enrolled 223 patients out of 400 planned, halted at the first interim analysis for financial and logistical reasons related in part to the coronavirus disease 2019 (COVID-19) pandemic. The 95% credible interval in the primary analysis spans 1.01 to 1.78. The true effect could be modest, or could be substantially larger if the trial had reached completion. The per-protocol sensitivity analysis, restricted to patients who received more than 50% of planned stimulations, finds a posterior probability of 99.3%, which is reassuring for the biological plausibility of the intervention. The statistical picture is best read as a directional signal of moderate strength, rather than as a definitive estimate of treatment effect.

Once again, a solution to strengthen the statistical signal would be to target a more physiologically enriched population (25). Difficult weaning is clearly the right problem; it is practical, prevalent, and hard to treat. However, using a broad clinical definition as enrollment criterion for a mechanistically specific intervention introduces a dilution. Different physiologically relevant targets can be considered. First, patients early in the course of controlled mechanical ventilation, within the first 48 to 96 hours of intubation for ARDS or severe hypoxemic respiratory failure, represent the prevention scenario. The goal here is maintaining diaphragm contractile activity before atrophy establishes itself as tested in the STIMULUS scenario (17). To avoid patients in which the intervention is unlikely to benefit, one could focus on patients phenotyped as hypoinflammatory, whose outcomes are most correlated to the trajectory of the respiratory failure (26,27). In hyperinflammatory patients, death is most often mediated by systemic inflammation and extra-respiratory organ failure, leading to dilution of potential neurostimulation effects (28). Second, patients with significant dorsal atelectasis, identifiable by electrical impedance tomography, computed tomography (CT) morphology, or lung ultrasound (29-31), would be those most likely to respond to the ventilation redistribution effect. Patients with right ventricular dysfunction or hemodynamic compromise attributable in part to positive-pressure ventilation represent a third category, in which the cardiovascular effects of neurostimulation could provide additive value. These categories are not mutually exclusive: a patient in the early phase of hypoinflammatory ARDS, with a focal morphological phenotype, and right ventricular dysfunction driven by high PEEP, sits at the intersection of all three. That overlap is exactly the profile in which the effect might be largest.

Conclusions

RESCUE-3 is a meaningful step. In a patient population that is especially difficult to treat, a signal favoring diaphragm neurostimulation emerges across multiple outcomes. What the trial cannot tell us is whether this is the optimal population for the intervention, whether 120 daily stimulations are the right dose, or whether the technique is best used to prevent diaphragmatic atrophy or to treat it. RESCUE-3 establishes diaphragm neurostimulation as a clinically active intervention in difficult weaning, but the gap between a 5.9 cmH₂O MIP gain and a meaningful weaning benefit signals a mechanism that extends well beyond muscle rehabilitation. Realizing its full potential will require earlier use, physiology-guided selection, and dose optimization.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Journal of Thoracic Disease. The article has undergone external peer review.

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1292/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1292/coif). J.P. received funding from the Société Française d’Anesthésie-Réanimation, the Philippe Foundation, the Beth Israel Deaconess Medical Center (Boston, MA), and the University Hospital of Montpellier. M.C. received funding from the Société Française d’Anesthésie-Réanimation, and the Beth Israel Deaconess Medical Center (Boston, MA). The other author has no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Pham T, Heunks L, Bellani G, et al. Weaning from mechanical ventilation in intensive care units across 50 countries (WEAN SAFE): a multicentre, prospective, observational cohort study. Lancet Respir Med 2023;11:465-76. [Crossref] [PubMed]
2. Dres M, Ewert R, Conrad SA, et al. Temporary Transvenous Diaphragm Neurostimulation for Weaning from Mechanical Ventilation (RESCUE-3): A Randomized Clinical Trial. Am J Respir Crit Care Med 2026;212:86-94. [PubMed]
3. Béduneau G, Pham T, Schortgen F, et al. Epidemiology of Weaning Outcome according to a New Definition. The WIND Study. Am J Respir Crit Care Med 2017;195:772-83. [Crossref] [PubMed]
4. Dres M, Goligher EC, Heunks LMA, et al. Critical illness-associated diaphragm weakness. Intensive Care Med 2017;43:1441-52. [Crossref] [PubMed]
5. Capdevila M, Pensier J, De Jong A, et al. Impact of Underassisted Ventilation on Diaphragm Function and Structure in a Porcine Model. Anesthesiology 2025;142:896-906. [Crossref] [PubMed]
6. Capdevila M, De Jong A, Belafia F, et al. Ultrasound-guided Transcutaneous Phrenic Nerve Stimulation in Critically Ill Patients: A New Method to Evaluate Diaphragmatic Function. Anesthesiology 2025;142:522-31. [Crossref] [PubMed]
7. Aarab Y, Flatres A, Garnier F, et al. Shear Wave Elastography, a New Tool for Diaphragmatic Qualitative Assessment: A Translational Study. Am J Respir Crit Care Med 2021;204:797-806. [Crossref] [PubMed]
8. Shi Z, van den Berg M, Bogaards S, et al. Replacement Fibrosis in the Diaphragm of Mechanically Ventilated Critically Ill Patients. Am J Respir Crit Care Med 2023;207:351-4. [Crossref] [PubMed]
9. Heunks L, Donker DW, Oppersma E, et al. Diaphragm Pacing in Early Critical Illness? A Plea for a Super-Relaxed Approach. Am J Respir Crit Care Med 2025;211:316-8. [Crossref] [PubMed]
10. Etienne H, Morris IS, Hermans G, et al. Diaphragm Neurostimulation Assisted Ventilation in Critically Ill Patients. Am J Respir Crit Care Med 2023;207:1275-82. [Crossref] [PubMed]
11. Levine S, Nguyen T, Taylor N, et al. Rapid disuse atrophy of diaphragm fibers in mechanically ventilated humans. N Engl J Med 2008;358:1327-35. [Crossref] [PubMed]
12. Dres M, de Abreu MG, Merdji H, et al. Randomized Clinical Study of Temporary Transvenous Phrenic Nerve Stimulation in Difficult-to-Wean Patients. Am J Respir Crit Care Med 2022;205:1169-78. [Crossref] [PubMed]
13. De Jong A, Capdevila M, Aarab Y, et al. Incidence, Risk Factors, and Long-Term Outcomes for Extubation Failure in ICU in Patients With Obesity: A Retrospective Analysis of a Multicenter Prospective Observational Study. Chest 2025;167:139-51. [Crossref] [PubMed]
14. Pensier J, Guerrero MA, Berger-Estilita J, et al. Perioperative ventilation support, what clinicians and searchers must know. Anaesth Crit Care Pain Med 2025;44:101554. [Crossref] [PubMed]
15. Pensier J, de Jong A, Hajjej Z, et al. Effect of lung recruitment maneuver on oxygenation, physiological parameters and mortality in acute respiratory distress syndrome patients: a systematic review and meta-analysis. Intensive Care Med 2019;45:1691-702. [Crossref] [PubMed]
16. Writing Group for the Alveolar Recruitment for Acute Respiratory Distress Syndrome Trial (ART) Investigators. Effect of Lung Recruitment and Titrated Positive End-Expiratory Pressure (PEEP) vs Low PEEP on Mortality in Patients With Acute Respiratory Distress Syndrome: A Randomized Clinical Trial. JAMA 2017;318:1335-45. [Crossref] [PubMed]
17. Morris IS, Bassi T, Bellissimo CA, et al. Continuous On-Demand Diaphragm Neurostimulation to Prevent Diaphragm Inactivity During Mechanical Ventilation: A Phase 1 Clinical Trial (STIMULUS). Am J Respir Crit Care Med 2025;211:1442-51. [Crossref] [PubMed]
18. Morris IS, Bassi T, Castellvi-Font A, et al. Diaphragm neurostimulation mitigates the adverse cardiopulmonary effects of positive pressure ventilation. Crit Care 2026;30:182. [Crossref] [PubMed]
19. Parfait M, Rohrs E, Joussellin V, et al. An Initial Investigation of Diaphragm Neurostimulation in Patients with Acute Respiratory Distress Syndrome. Anesthesiology 2024;140:483-94. [Crossref] [PubMed]
20. Rohrs EC, Bassi TG, Nicholas M, et al. Negative-pressure-assisted ventilation lowers driving pressure and mechanical power in an ARDS model. J Appl Physiol (1985) 2022;133:1237-49. [Crossref] [PubMed]
21. Mauri T, Heunks L. Diaphragm Neurostimulation: An Innovative Approach to Treat Diaphragm Weakness in Difficult to Wean Patients? Am J Respir Crit Care Med 2026;212:11-2. [Crossref] [PubMed]
22. Aleva F, Heunks L, Doorduin J. Respiratory muscle training and neurostimulation in critical illness: evidence and future directions. Curr Opin Crit Care 2026;32:87-92. [Crossref] [PubMed]
23. Goligher EC, Heath A, Harhay MO. Bayesian statistics for clinical research. Lancet 2024;404:1067-76. [Crossref] [PubMed]
24. Naudet-Lasserre A, Pensier J, de Jong A, et al. Effect of noninvasive ventilation on tracheal reintubation among patients with hypoxemic respiratory failure following abdominal surgery: a bayesian post-hoc analysis of the NIVAS trial. Crit Care 2025;29:533. [Crossref] [PubMed]
25. Munroe ES, Spicer A, Castellvi-Font A, et al. Evidence-based personalised medicine in critical care: a framework for quantifying and applying individualised treatment effects in patients who are critically ill. Lancet Respir Med 2025;13:556-68. [Crossref] [PubMed]
26. Pensier J, Fosset M, Paschold BS, et al. Temporal stability of phenotypes of acute respiratory distress syndrome: clinical implications for early corticosteroid therapy and mortality. Intensive Care Med 2025;51:1784-96. [Crossref] [PubMed]
27. Renard Triché L, Fosset M, Jabaudon M, et al. Temporal stability of inflammatory subphenotypes of acute respiratory distress syndrome: 28-day insights from the ICAR trial. Anaesth Crit Care Pain Med 2025;44:101559. [Crossref] [PubMed]
28. Pensier J, Fosset M, Paschold BS, et al. Effects of inflammatory phenotypes in acute respiratory distress syndrome on mortality and partitioning of lung and chest wall mechanics in patients in the USA and Canada: a retrospective cohort study. Lancet Respir Med 2026;14:609-19. [Crossref] [PubMed]
29. He H, Zhao Z, Becher T, et al. Recommendations for lung ventilation and perfusion assessment with chest electrical impedance tomography in critically ill adult patients: an international evidence-based and expert Delphi consensus study. EClinicalMedicine 2025;89:103575. [Crossref] [PubMed]
30. Sinnige JS, Smit MR, Alam MJ, et al. Personalized mechanical ventilation guided by lung ultrasound in patients with ARDS: a pilot phase of a randomized clinical trial. Intensive Care Med Exp 2025;13:135. [Crossref] [PubMed]
31. Pensier J, Touaibia M, Meerun MA, et al. Morphological subphenotypes of acute pancreatitis-related acute respiratory distress syndrome. Crit Care 2026;30:65. [Crossref] [PubMed]