Lung- and diaphragm-protective strategies in acute respiratory failure: an in silico trial

Damian Ratano, Binghao Zhang, José Dianti, D. Georgopoulos, Laurent Brochard, Timothy C. Y. Chan, Ewan C. GoligherIntensive 2024 · DOI

This study uses a sophisticated "digital twin" approach to tackle one of the most difficult balancing acts in intensive care: how to support a patient's breathing without damaging their lungs or allowing their diaphragm to wither away. By simulating 5,000 patients, the researchers demonstrate that while standard ventilation modes often fail to meet these dual protective targets, adding extracorporeal $CO_2$ removal ($ECCO_2R$) can significantly bridge the gap.

TL;DR

Researchers conducted a large-scale in silico (computer-simulated) trial to show that achieving both lung- and diaphragm-protective (LDP) targets is difficult with standard mechanical ventilation alone (success rates ~40%). However, integrating extracorporeal $CO_2$ removal ($ECCO_2R$) as an adjunct increases the success rate to over 80%, particularly for patients with moderate physiological derangement.

What Problem Does This Paper Address?

Mechanical ventilation is a double-edged sword. While it saves lives in acute respiratory failure, it can cause Ventilator-Induced Lung Injury (VILI) if pressures are too high, and Ventilator-Induced Diaphragm Dysfunction (VIDD) if the ventilator does too much (causing atrophy) or too little (causing overwork) of the work.

The "holy grail" is Lung- and Diaphragm-Protective (LDP) ventilation, which aims for a "Goldilocks" zone: enough respiratory effort to keep the diaphragm active, but not so much that the resulting lung expansion (transpulmonary pressure) becomes injurious. Because clinical trials to find the best settings are logistically difficult and potentially risky, the authors used a mathematical model to explore the feasibility of different strategies.

Methods

The authors employed a previously validated physiology-based mathematical model to conduct an in silico clinical trial.

  • The Model: It simulates the complex interplay between respiratory mechanics, $CO_2$ kinetics, acid-base homeostasis (using the Stewart approach), and the effects of sedation (propofol) on respiratory drive.
  • The Population: They generated 5,000 "virtual patients" with physiological characteristics (compliance, resistance, metabolic rate) sampled from distributions found in real-world clinical literature.
  • The Interventions: They tested two common ventilation modes—Pressure Support Ventilation (PSV) and Proportional Assist Ventilation (PAV+)—using a specific titration algorithm to adjust support and sedation. They also tested these modes with adjunctive $ECCO_2R$ and evaluated "awake $ECCO_2R$" (no mechanical ventilation).
  • Validation: To ensure the simulation reflected reality, they compared the in silico results against data from a real-world pilot clinical trial (LANDMARK 1).

Datasets

  • Simulated Dataset: A newly generated population of 5,000 virtual patients based on parameters from several landmark ARDS and respiratory failure studies.
  • Validation Dataset: Data from the LANDMARK 1 clinical trial (n=30), used to verify that the model's predicted success rates matched observed clinical outcomes.

Key Insights and Ideas

  • Mode Neutrality: Surprisingly, PAV+ (which is theoretically more "natural") did not outperform the more common PSV in achieving LDP targets after titration. This suggests that how you titrate may be more important than the specific mode used.
  • The $ECCO_2R$ "Unloader" Effect: $ECCO_2R$ doesn't just fix blood gases; by removing $CO_2$ extracorporeally, it lowers the patient's biological drive to breathe. This makes it much easier for clinicians to dial in the "Goldilocks" zone of effort.
  • The Failure of "Awake $ECCO_2R$": There is a growing interest in managing patients with $ECCO_2R$ alone to avoid intubation entirely. However, the simulation found this was successful in only 9% of cases. In severe failure, the "wakefulness drive" and poor lung mechanics still force the patient to work too hard, leading to potential lung injury even if $CO_2$ is being cleared by a machine.
  • Predictive Power: Success in achieving LDP targets is largely determined by three factors: lung compliance, ventilatory ratio (a surrogate for dead space), and strong ion difference (metabolic acid-base status).

Key Results

  • Success Rates: PSV achieved LDP targets in 43% of patients, while PAV+ achieved them in 37%.
  • The $ECCO_2R$ Boost: Adding $ECCO_2R$ to either mode increased the success rate to ~85%.
  • Clinical Score: The researchers developed an "ECCO2R score" based on ventilatory ratio and elastance. Patients with intermediate scores saw the greatest benefit from adding $ECCO_2R$.
  • Model Accuracy: The model's predictions discriminated between success and failure in real patients with high accuracy (AUC of 0.83).

Limitations and Future Work

  • Simplified Drive: The model primarily focuses on chemoreceptive control (blood gases). It does not fully account for non-chemical triggers like pain, anxiety, or lung stretch reflexes (Hering-Breuer), which can significantly influence a real patient's breathing pattern.
  • Fixed Respiratory Rate: The simulation assumed a fixed respiratory rate for simplicity, which might not capture the full variability of patient-ventilator interaction.
  • PEEP Effects: The model does not explicitly simulate the complex effects of Positive End-Expiratory Pressure (PEEP) on lung recruitment and gas exchange.
  • Next Steps: The authors suggest that the "ECCO2R score" derived here should be prospectively validated in clinical trials to identify which patients are the best candidates for this expensive and invasive therapy.

Summary updated

Abstract

In this in silico trial, many patients required ECCO<sub>2</sub>R in combination with mechanical ventilation and sedation to achieve LDP targets. ECCO<sub>2</sub>R increased the probability of achieving LDP targets in patients with intermediate degrees of derangement in elastance and ventilatory ratio.
BibTeX
@inproceedings{ratano2024lung,
  author = {Damian Ratano and Binghao Zhang and José Dianti and D. Georgopoulos and Laurent Brochard and Timothy C. Y. Chan and Ewan C. Goligher},
  title = {{Lung- and diaphragm-protective strategies in acute respiratory failure: an in silico trial}},
  booktitle = {Intensive},
  year = {2024},
  doi = {10.1186/s40635-024-00606-x},
}
Figure