Acute respiratory distress syndrome (ARDS) varies wildly between patients, leading to unpredictable outcomes and inconsistent responses to ventilator settings like Positive End-Expiratory Pressure (PEEP).

Researchers used machine learning to validate two distinct biological profiles (subphenotypes) in 1,532 ARDS patients across multiple independent clinical trials, using routine bedside data on lung mechanics and gas exchange.

The Two Subphenotypes: Patients naturally fell into one of two categories:
– “Efficient”: Patients with better lung function and lower baseline risk.
– “Restrictive”: Patients with severe lung restriction who faced a significantly higher risk of death (a 75% increase in 28-day mortality).

The two groups responded oppositely to ventilator settings. Higher PEEP increased mortality in “Efficient” patients but reduced mortality in “Restrictive” patients.

By using basic, readily available bedside data to identify a patient’s specific ARDS subphenotype, doctors can move away from a one-size-fits-all approach and use personalized medicine to save lives on ventilators.

Read More