Critical care subphenotypes: prospective identification and extension beyond syndromic definitions

Reproducible inflammatory subphenotypes (the hypoinflammatory and hyperinflammatory subphenotypes) carry prognostic and predictive significance, but they have typically been identified retrospectively using latent class analysis (LCA) and studied only within ARDS, limiting their translation into clinical practice and prospective trials.<br><br>This thesis examines whether inflammatory subphenotypes can be identified in clinically relevant, real-time settings, extended beyond ARDS, characterised across disease severity and over time, and operationalised to inform precision-medicine trial design.

Following a review of the field, its terminology, and the barriers to clinical translation, inflammatory subphenotypes are examined in patients with severe respiratory failure requiring veno-venous extracorporeal membrane oxygenation. The prospective, multicentre PHIND study then evaluates bedside identification of subphenotypes in 30 intensive care units across the United Kingdom and Ireland, using a near-patient immunoanalyser and a parsimonious model (interleukin-6, soluble tumour necrosis factor receptor-1, and bicarbonate) to assign subphenotypes in real time and to test their association with 60-day mortality.