Relationship Between Phenotyping and Individualized Absolute Risk Differences in Sepsis: A Secondary Analysis of Two...

Area of Science:

• Critical Care Medicine
• Clinical Trial Design
• Biostatistics

Background:

• Sepsis patient populations exhibit variable responses to therapeutic interventions, complicating treatment optimization.
• Current methods for identifying differential treatment responses in sepsis clinical trials are insufficient for individual patient-level prediction.

Purpose of the Study:

• To explore the relationship between established clinical and biologic subgrouping approaches and a novel individualized absolute risk difference (iARD) model.
• To assess the variability of iARDs within identified patient subgroups in sepsis trials.
• To evaluate the potential of subgrouping for personalizing sepsis treatment, specifically early goal-directed therapy (EGDT).

Main Methods:

• Secondary analysis of the Protocolized Care for Early Septic Shock (ProCESS) and Australasian Resuscitation in Sepsis Evaluation (ARISE) randomized controlled trials.
• Application of clinical (α, β, γ, δ) and biologic (hyperinflammatory, nonhyperinflammatory) subphenotypes to patient cohorts.
• Supervised learning models were used to predict iARDs for 90-day mortality, utilizing clinical variables and biomarker data.

Main Results:

• The average treatment effect of EGDT varied within both clinical and biologic subgroups, showing potential benefit in some (β, nonhyperinflammatory) and harm in others (γ, hyperinflammatory).
• Crucially, predicted iARDs within each subgroup demonstrated a wide range, from significant harm to considerable benefit.
• For instance, in the β-subtype, while the average EGDT effect suggested a 8.5% mortality reduction, individual predictions ranged from a 29% increase to a 16% decrease in mortality, with 39% of patients predicted to experience harm.

Conclusions:

• While clinical and biologic phenotyping can identify subgroups with differing average treatment effects, significant individual variability in risk and benefit persists within these groups.
• These findings raise concerns about the reliability and safety of using current phenotyping strategies for personalizing sepsis treatment.
• Further research is needed to develop more accurate methods for individualized treatment prediction in sepsis.