Multicenter target trial emulation to evaluate corticosteroids for sepsis stratified by predicted organ dysfunction...

Area of Science:

• Critical care medicine
• Infectious diseases
• Pharmacology

Background:

• Corticosteroids are used in sepsis, septic shock, pneumonia, and ARDS.
• The precise risk-benefit balance of corticosteroids, particularly regarding organ dysfunction duration, remains unclear.
• Existing research lacks a nuanced understanding of how corticosteroid use impacts mortality across different patient trajectories.

Purpose of the Study:

• To evaluate the association between corticosteroid use and 28-day mortality in patients with sepsis, pneumonia, and ARDS.
• To stratify patients based on predicted organ dysfunction trajectories.
• To emulate a target trial framework for analyzing treatment effects.

Main Methods:

• Retrospective, multicenter, proof-of-concept study.
• Two-stage machine learning methodology for patient subphenotyping and trajectory prediction.
• Target trial emulation framework applied after patient classification.

Main Results:

• The association between corticosteroid use and 28-day mortality varied significantly based on the predicted organ dysfunction trajectory.
• Different patient cohorts exhibited distinct responses to corticosteroid treatment.
• Machine learning successfully subphenotyped patients according to their organ dysfunction progression.

Conclusions:

• Corticosteroid utility in critical illnesses is trajectory-dependent.
• Matching corticosteroid treatment strategies to empirically observed pathobiology offers a more nuanced approach.
• Personalized medicine strategies may optimize corticosteroid use in sepsis, pneumonia, and ARDS.