Risk heterogeneity within hypoinflammatory acute respiratory failure: continuous probabilities identify high-risk patients masked by binary classification.

Pubmed ID: 42043553

Journal: Intensive care medicine

Publication Date: April 1, 2026

MeSH Terms: Humans, Male, Female, Aged, Middle Aged, Risk Assessment, Prognosis, Critical Illness, Probability, Inflammation, Interleukin-6, Respiratory Insufficiency, Registries, Biomarkers, COVID-19

Grants: P01 HL114453, 1R21 HL168070, 5R01 GM141081, R21 HL168070, R01 176668, R01 GM141081

Authors: Morris A, Nouraie SM, Baron RM, Yang Z, Bain W, Kitsios GD, McVerry BJ, Venkatesan N, Shah FA, Dela Cruz CS, Zuchelkowski B, Rizzo AN, Joshi S, Arciniegas A, Zambrano K, Aneis H, Mayr FB, Talisa VB

Cite As: Venkatesan N, Shah FA, Bain W, Yang Z, Dela Cruz CS, Baron RM, Zuchelkowski B, Rizzo AN, Joshi S, Arciniegas A, Zambrano K, Aneis H, Mayr FB, Morris A, Talisa VB, McVerry BJ, Nouraie SM, Kitsios GD. Risk heterogeneity within hypoinflammatory acute respiratory failure: continuous probabilities identify high-risk patients masked by binary classification. Intensive Care Med 2026 Apr;52(4):673-686. Epub 2026 Apr 27.

Studies:

Abstract

PURPOSE: Binary inflammatory subphenotype classification (hyperinflammatory vs. hypoinflammatory) may guide trial enrollment in acute hypoxemic respiratory failure (AHRF), but assumes within-category homogeneity. We determined whether continuous probabilities reveal clinically meaningful heterogeneity. METHODS: We analyzed 575 critically ill adults with AHRF (Pittsburgh Acute Lung Injury Registry) and validated findings in 1134 patients from the EDEN trial, the COVID-19 cohorts, and the RoCI registry. Continuous subphenotype probabilities were calculated using a parsimonious biomarker model (IL-6, sTNFR-1, bicarbonate; probability threshold 0.5). The primary outcome was 90-day mortality. RESULTS: Among 575 patients, 77 patients (13%) were hyperinflammatory and 498 (87%) hypoinflammatory. Hyperinflammatory patients demonstrated prognostic homogeneity (mortality overall 55%, p = 0.72, across tertiles). Hypoinflammatory patients exhibited marked heterogeneity: 90-day mortality increased from 19 to 31% to 40% across probability tertiles (P < 0.001). Restricted cubic spline modeling demonstrated a strong non-linear relationship between continuous probabilities and mortality, with the steepest risk increases occurring below the 0.5 threshold. Clinical severity scores and biomarkers of immune activation increased progressively across hypoinflammatory tertiles (all P < 0.001). Among 330 patients with longitudinal sampling, rising probability trajectories within hypoinflammatory groups predicted 50-100% mortality vs. 16-40% for stable or declining trajectories (all P < 0.001); hyperinflammatory patients had poor outcomes regardless of trajectory. External validation confirmed heterogeneity and preserved non-linear probability-mortality patterns across cohorts. Similar patterns were observed with the procalcitonin-based model. CONCLUSIONS: Binary classification obscures substantial prognostic heterogeneity within hypoinflammatory AHRF patients. Continuous probability-based stratification may identify additional trial-eligible high-risk patients and improve enrollment strategies for subphenotype-guided trials.