Inflammatory Phenotypes, Early Clinical Trajectories, and Outcomes in Acute Respiratory Distress Syndrome: A Multicenter Derivation and Validation Study

Abstract Background: Acute respiratory distress syndrome (ARDS) is biologically and clinically heterogeneous. Whether baseline inflammatory phenotypes correspond to distinct early clinical trajectories, how those trajectories evolve during the first 72 hours, and whether trajectory information adds prognostic value beyond baseline inflammatory phenotype alone remain unclear. Methods: We conducted a multicenter study of clinically recognized ARDS across three Chinese centers. Ruijin Hospital served as the derivation cohort, Fujian Provincial Hospital as the retrospective validation cohort, and The Second People’s Hospital of Fuyang City as the prospective validation cohort. Baseline inflammatory phenotypes were defined in biomarker-rich cohorts using a prespecified parsimonious biomarker framework based on interleukin-6 (IL-6), tumor necrosis factor receptor 1 (TNFR1), procalcitonin, and lactate. Early clinical trajectories over the first 72 hours were derived in the Ruijin cohort and assigned in the validation cohorts using a fixed-reference framework. We examined associations between baseline phenotype and trajectory class, phenotype transition and speed of early change, clinical outcomes across trajectories, and the incremental prognostic value of trajectory beyond phenotype alone. Results: A total of 995 patients with clinically recognized ARDS were included, including 150 in the Ruijin derivation cohort, 505 in the Fujian retrospective validation cohort, and 340 in the Fuyang prospective validation cohort. In the derivation cohort, three early trajectory classes were identified: rapid improvers (54/150, 36.0%), intermediate resolvers (53/150, 35.3%), and refractory patients (43/150, 28.7%). In Ruijin, PaO₂/FiO₂ changed from 238.4 ± 54.1 to 320.4 ± 70.3 mmHg in rapid improvers, from 186.3 ± 41.8 to 192.1 ± 49.7 mmHg in intermediate resolvers, and from 103.1 ± 28.5 to 87.0 ± 24.1 mmHg in refractory patients over the first 72 hours (all P < 0.001 across groups). Compared with the lower-inflammatory phenotype, the higher-inflammatory phenotype was associated with lower odds of a rapid improver trajectory (adjusted OR 0.54, 95% CI 0.29–0.98) and higher odds of a refractory trajectory (adjusted OR 2.81, 95% CI 1.59–4.95). In transition analyses, persistent lower-inflammatory status was associated with improving oxygenation (ΔPaO₂/FiO₂ +78.4 ± 64.2 mmHg) and lower ICU mortality (11.9%), whereas persistent higher-inflammatory status was associated with worsening oxygenation (ΔPaO₂/FiO₂ -8.7 ± 58.6 mmHg) and higher ICU mortality (32.2%). Rapid improvers consistently had the most favorable outcomes across cohorts. Addition of baseline inflammatory phenotype improved discrimination for 28-day mortality from 0.71 to 0.78, and further addition of early trajectory improved discrimination to 0.84. Conclusions: In ARDS, baseline inflammatory phenotype captures only part of clinically relevant heterogeneity. Early 72-hour clinical trajectory, including the direction and speed of change, provides complementary dynamic information and improves prognostic stratification beyond baseline phenotype alone.