Highlights
- Two distinct inflammatory subphenotypes (hyper- and hypoinflammatory) were identified in patients with COVID-19-related moderate-to-severe ARDS.
- The hyperinflammatory subphenotype was associated with significantly higher 28-day mortality (52% vs. 25%) and fewer ventilator-free days.
- Bayesian discrete-time Markov modeling confirmed high temporal stability, with a 70% probability of remaining in the hypoinflammatory state and a 52% probability of remaining in the hyperinflammatory state over 7-day intervals.
- Transitions between subphenotypes were rare, suggesting that these biological profiles represent sustained pathological states rather than transient fluctuations.
Background: The Challenge of Heterogeneity in ARDS
Acute Respiratory Distress Syndrome (ARDS) remains one of the most significant challenges in intensive care medicine, characterized by high morbidity and mortality despite decades of research. One of the primary hurdles in developing effective pharmacotherapies has been the inherent heterogeneity of the syndrome. ARDS is not a single disease but a clinical construct arising from various insults, including pneumonia, sepsis, and trauma. In recent years, the identification of biological subphenotypes—specifically the hyperinflammatory and hypoinflammatory profiles—has offered a pathway toward precision medicine in the ICU.
While previous retrospective analyses of randomized controlled trials (RCTs) have consistently identified these two subphenotypes, a critical question remained: how stable are these profiles over time? If a patient’s inflammatory state fluctuates rapidly, a single baseline assessment might be insufficient for guiding long-term therapy or enrichment in clinical trials. The ICAR trial reanalysis seeks to address this gap by evaluating the temporal stability of these subphenotypes over a 28-day period in patients with COVID-19-related ARDS.
Study Design and Methodology
The study was a secondary analysis of the ICAR (Immunomodulatory Therapy in COVID-19-Related ARDS) trial, a multicenter randomized controlled trial conducted across 43 centers. The study population included 146 patients with COVID-19-related moderate-to-severe ARDS.
Subphenotype Identification
To identify subphenotypes, the researchers employed K-means clustering, a machine learning technique, at multiple time points: inclusion (day 0), and subsequently at 7-day intervals (day 7, 14, 21, and 28). The clustering was based on a variety of clinical and biological parameters, including plasma levels of cytokines, chemokines, adhesion molecules, and proangiogenic factors.
Stability Assessment
The core of the analysis involved a Bayesian discrete-time Markov model. This statistical approach allowed the researchers to calculate the probability of a patient transitioning between states (hypoinflammatory, hyperinflammatory, extubated, or deceased) over 7-day intervals. This model is particularly robust for longitudinal data in critical care, where competing risks (such as death or discharge) must be accounted for when assessing biological stability.
Key Findings: Biological Profiles and Clinical Outcomes
At the time of inclusion, 121 patients (83%) were classified as hypoinflammatory, while 25 patients (17%) were classified as hyperinflammatory. This distribution is consistent with previous subphenotyping studies in non-COVID ARDS cohorts.
The Hyperinflammatory Signature
Patients in the hyperinflammatory group exhibited a markedly different biological and clinical profile compared to the hypoinflammatory group. Key characteristics included:
- Higher levels of pro-inflammatory cytokines (e.g., IL-6, IL-8) and chemokines.
- Increased levels of adhesion molecules and proangiogenic factors, suggesting significant endothelial dysfunction.
- Higher rates of multi-organ failure and lower endothelial stability.
Prognostic Significance
The clinical impact of these subphenotypes was profound. The hyperinflammatory subphenotype was associated with a 28-day mortality rate of 52% (13/25), compared to only 25% (30/121) in the hypoinflammatory group (p = 0.001). Furthermore, hyperinflammatory patients had significantly fewer ventilator-free days (VFDs) through day 28 (p < 0.01), underscoring the severity of the lung injury and the systemic inflammatory response in this cohort.
Longitudinal Dynamics: Are Subphenotypes Stable?
The most significant contribution of this study is the confirmation of temporal stability. The Bayesian Markov model revealed that the probability of a patient remaining in their assigned subphenotype over a 7-day period was high.
Stability Probabilities
- Hypoinflammatory Stability: Patients in this group had a 70% probability of remaining hypoinflammatory. They also had a 17% probability of being extubated, while only 7% progressed to the hyperinflammatory state.
- Hyperinflammatory Stability: Patients in this group had a 52% probability of remaining hyperinflammatory. The risk of death was high (23%), while the probability of transitioning to a hypoinflammatory state (20%) or being extubated (5%) was relatively low.
These findings suggest that once a patient is categorized into a subphenotype, that biological trajectory is likely to persist throughout the acute phase of the illness. This challenges the notion that ARDS subphenotypes are merely transient snapshots of a rapidly evolving condition.
Expert Commentary: Mechanistic Insights and Clinical Implications
The stability of these subphenotypes has major implications for the future of ARDS management. From a mechanistic perspective, the hyperinflammatory state appears to be driven by a sustained cycle of endothelial injury and cytokine release that does not easily resolve. The high levels of adhesion molecules and proangiogenic factors in the hyperinflammatory group point toward the endothelium as a primary driver of the persistent pathology.
Precision Medicine and Trial Design
For decades, ARDS trials have failed, likely because they treated all patients as a homogeneous group. The ICAR trial data suggests that future trials should consider “predictive enrichment.” By identifying hyperinflammatory patients early, researchers can target them with specific anti-inflammatory or immunomodulatory therapies. Because these phenotypes are stable, a treatment initiated based on a day 0 assessment is likely to remain biologically relevant for the duration of the patient’s ICU stay.
Limitations and Generalizability
While the study is robust, it is important to note that the cohort consisted entirely of COVID-19 patients. While COVID-19 ARDS shares many features with traditional ARDS, the inflammatory kinetics may differ. Additionally, the sample size (n=146) is relatively small for complex clustering, though the multicenter nature and the use of Bayesian modeling strengthen the validity of the results.
Conclusion
The ICAR trial reanalysis provides compelling evidence that inflammatory subphenotypes in COVID-19-related ARDS are stable over 28 days. The hyperinflammatory subphenotype identifies a high-risk population with distinct biological drivers and a poor prognosis. Monitoring these subphenotypes is not only valuable for assessing patient trajectories but is also essential for the development of targeted, evidence-based therapies. As we move toward a more personalized approach in the ICU, understanding the longitudinal stability of these biological signatures will be paramount.
References
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 Sep;44(5):101559. doi: 10.1016/j.accpm.2025.101559. Epub 2025 May 26. PMID: 40436270.
