Sepsis-associated acute respiratory distress syndrome (ARDS) is one of the deadliest and most biologically heterogeneous forms of respiratory failure. Despite uniform diagnostic criteria, patients with septic ARDS show wide variability in inflammatory intensity, alveolar epithelial and endothelial injury, alveolar fluid composition, ventilatory mechanical properties, and clinical evolution. Early identification of these differences may enable better prognostication and more precise treatment.

This prospective observational study aims to deeply characterize the earliest phases of septic ARDS by integrating serial bronchoalveolar lavage (BAL) at 0, 24 and 72 hours with parallel plasma biomarker profiling and detailed mechanical ventilation data. This design captures the evolving biological and physiological landscape of septic ARDS during its most dynamic window. The central goal is to identify systemic, alveolar, and hybrid bio-mechano-inflammatory subphenotypes that can inform personalized approaches to support, risk stratification, and future interventional trials.

Acute respiratory distress syndrome caused by sepsis is not a uniform disease but a syndrome comprising multiple biological and mechanical states. The Berlin Definition provides a useful clinical entry point but does not capture the profound heterogeneity found in molecular pathways, alveolar immune activation, epithelial disruption, endothelial dysfunction, and mechanical ventilation responses.

The first 72 hours of ARDS represent a critical, rapidly evolving biological landscape where systemic cytokine release, alveolar epithelial injury, endothelial activation, and capillary-alveolar permeability all fluctuate dramatically. These early shifts are believed to determine downstream trajectories such as ventilator dependence, multiorgan dysfunction, and mortality. Capturing this dynamic process requires repeated sampling at predefined intervals rather than the traditional single one-time measurement.

ARDS emerges primarily within the alveolar space; yet most studies rely exclusively on plasma biomarkers, which provide only a partial window into the alveolus. The pulmonary compartment often behaves independently of the systemic circulation due to the compartmentalization of inflammatory mediators. BAL sampling therefore offers a unique opportunity to interrogate the lung directly. On the other hand, ventilatory mechanics-in particular driving pressure, plateau pressure, lung compliance, and ventilatory ratio-reflect the biomechanical stress conditions imposed on the lung, which may interact with or even exacerbate biological injury.

Thus, inflammation, epithelial damage, endothelial leak, alveolar flooding, and mechanical stress constitute interdependent dimensions of the early ARDS process.

This study integrates:

• Serial BAL to measure alveolar cytokines, epithelial and endothelial markers, proteins and permeability markers.
• Serial plasma biomarkers to identify systemic inflammatory phenotypes and compare them with alveolar phenotypes.
• Mechanical ventilation parameters to quantify biomechanical stress and its role in injury amplification.
• Clinical endpoints and physiology to connect biological patterns with real outcomes.

This multidimensional dataset is designed to reveal biological, mechanical, and hybrid subphenotypes which could explain why patients who are clinically similar diverge into starkly different trajectories.

Inclusion Criteria:

• Age ≥18
• ARDS diagnosis per Berlin definition
• Sepsis per Sepsis-3 criteria
• Invasive mechanical ventilation
• Expected to remain intubated ≥72 hours
• Consent from surrogate

Exclusion Criteria:

• Contraindications to bronchoscopy/BAL
• Refractory hemodynamic instability
• Pregnancy
• Pulmonary transplant patients
• Surrogate declines participation

Participants are adult ICU patients diagnosed with ARDS secondary to sepsis, requiring intubation and mechanical ventilation. Because the great majority will be sedated or unconscious, informed consent will be sought from legally authorized representatives. This population is particularly prone to adverse outcomes, making biological and mechanical profiling both valuable and clinically meaningful.

No plan to share IPD because of data sensitivity and institutional data protection requirements