Temporal stability of inflammatory subphenotypes of acute respiratory distress syndrome : outcomes and implications for early corticosteroids

Position du problème et objectif(s) de l’étude

Inflammatory subphenotypes of acute respiratory distress syndrome (ARDS) identify patients with distinct outcomes and may lead to possible individualized treatment (1). However, temporal evolution of the initial subphenotype is unclear (2,3), and may impact targeted treatment strategies (4). This study assessed the temporal stability of inflammatory subphenotypes of ARDS and potential clinical implications using the example of early corticosteroids.

Matériel et méthodes

Data and biomarkers from six multicenter randomized controlled trials (RCTs, n=3,958: ARMA, ALVEOLI, FACTT, SAILS, ROSE, EPVent-2) were analyzed to develop and validate an open-source AI Clinical Classifier for classification of inflammatory subphenotypes of ARDS using routinely available data (IRB#2024D000283). Then, it was used to classify a retrospective cohort including 5,578 adult ARDS patients from the seven intensive care units in a hospital network between 2008 and 2024 (IRB#2025P000018). A discrete-time Bayesian Markov model was used to evaluate temporal stability at 3-day intervals from baseline until day 30 (5). Multivariable logistic regression models assessed the effect of early corticosteroids on 30-day mortality across subphenotypes, after adjustment for age, sex, ethnicity, body mass index, hematologic malignancy, COVID-19, chronic obstructive pulmonary disease, PaO2/FiO2 ratio, Sequential Organ Failure Assessment score, neuromuscular blockade administration, and sedation depth.

Résultats & Discussion

An open-source AI Clinical Classifier (https://bostonmontpelliercare.shinyapps.io/AIClarity) was developed and externally validated across six RCTs before longitudinal validation using biomarkers until day 7. Among 18 predictors, the most important class-defining variables were bicarbonate, creatinine, and vasopressors. In the retrospective cohort, 2,169 (39%) hyperinflammatory and 3,409 (61%) hypoinflammatory patients were identified. The 30-day mortality was 49% (1,053/2,169) and 24% (826/3,409, p<0.001), respectively. Over 30 days, 1,072 (49%) hyperinflammatory patients at baseline transitioned to hypoinflammatory, compared to 229 (7%) hypoinflammatory patients at baseline transitioning to hyperinflammatory (p<0.001, Figure 1.). Every 3 days, patients initially classified as hyperinflammatory had a 51% probability of staying hyperinflammatory, a 24% probability of transitioning to hypoinflammatory, a 13% probability of death, and a 12% probability of extubation. Every 3 days, patients initially classified as hypoinflammatory had a 62% probability of staying hypoinflammatory, a 7% probability of transitioning to hyperinflammatory, a 6% probability of death, and a 25% probability of extubation. Early corticosteroids were associated with a lower 30-day mortality (adjusted odds ratio, aOR=0.76 [95%ConfidenceInterval(95%CI):0.59-0.99], p=0.040) in hyperinflammatory patients at baseline, and a higher 30-day mortality (aOR=1.46 [95%CI:1.17-1.82], p<0.001) in hypoinflammatory patients (p-for-interaction<0.001, Figure 2.). Only patients who stayed hyperinflammatory at day 3 exhibited a persistent response to corticosteroids (aOR=0.51, 95%CI:0.32-0.80, p=0.004).

Conclusion

While the hypoinflammatory subphenotype of ARDS remains stable over time, the hyperinflammatory subphenotype displays a more unpredictable trajectory. Early corticosteroids may benefit hyperinflammatory ARDS patients until transition to hypoinflammatory, supporting re-classification to ensure treatment success.

Auteurs

Joris PENSIER (1) , Maxime FOSSET (2), Béla-Simon PASCHOLD (3), Dario VON WEDEL (4), Simone REDAELLI (3), Ben BRAEUER (3), Victor NOVACK (5), Boris JUNG (2), Samir JABER (1), Daniel TALMOR (6), Elias BAEDORF KASSIS (6), Maximilian S SCHAEFER (3) - (1)Anesthesiology And Intensive Care, Anesthesia And Critical Care Department B, Saint Eloi Teaching Hospital, Phymedexp, University Of Montpellier, Inserm U1046, Montpellier, France, Montpellier, France, (2)Medical Intensive Care Unit, Premedical Inria-Inserm Team Idesp, Montpellier University And Montpellier University Health Care Center, Montpellier, Montpellier, France, (3)Center For Anesthesia Research Excellence, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts, Boston, États-Unis, (4)Institute Of Medical Informatics, Charité-Universitätsmedizin Berlin, Berlin, Germany, Berlin, Allemagne, (5)Clinical Research Center, Soroka University Medical Center And Faculty Of Health Sciences, Ben-Gurion University Of The Negev, Beer-Sheva, Israel, Tel Aviv, Israël, (6)Department Of Anesthesia, Critical Care And Pain Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, Boston, Ma, 02215, Usa, Boston, États-Unis

Orateur(s)

Joris PENSIER (Montpellier)