Individualized treatment effects of high flow nasal oxygen in immunocompromised patients
In brief
A model predicts HFNO cuts intubation risk by 20 points in selected patients
In 704 immunocompromised patients with acute respiratory failure, high-flow nasal oxygen had no significant overall effect on intubation by day 28. A machine-learning model identified a subgroup with a predicted 20.2-percentage-point lower risk; the ranking of likely benefit held in a separate observational cohort, but estimates were less precise. Prospective studies must confirm who benefits.
- Journal
- Intensive care medicine (Q1)
- Published
- 2 November 2026
- Study design
- Randomized controlled trial
- Evidence level
- Level 2, Moderate (CEBM 2b)
- Authors
- Alexis Maillard, Mélanie Métais, Virginie Lemiale, Djamel Mokart, Anne-Sophie Moreau, Emmanuel Canet, et al.
- PMID
- 42970133
- DOI
- 10.1007/s00134-026-08635-8
Why clinicians should know about it
- Picked for Critical Care and Intensive Care Medicine (paper of the day, 4 November 2026): HFNO treatment-effect analysis for acute hypoxemic respiratory failure
Abstract
RATIONALE: High flow nasal oxygen (HFNO) has become a common first-line practice for acute hypoxemic respiratory failure (ARF). In immunocompromised patients, ARF is the leading cause of ICU admission, yet a randomized trial found no overall benefit of HFNO in this population. OBJECTIVES: To assess whether baseline characteristics influence HFNO's effectiveness in preventing invasive mechanical ventilation (IMV) in immunocompromised patients with ARF. METHODS: A causal forest machine-learning algorithm was derived in the HIGH trial (776 patients randomized to HFNO or conventional oxygen) to predict the HFNO effects on day-28 IMV incidence. Patients with a "do-not-intubate" order (n = 72) were excluded. Model performance was validated through 500 bootstrap iterations and patients were categorized into terciles based on predicted HFNO benefit. Transportability of the model was assessed in an independent observational cohort (818 patients). MAIN RESULTS: Among the 704 participants, HFNO's average effect on day-28 IMV was not significant. However, substantial heterogeneity in treatment responses was identified (predicted benefit: from 20.2% [IQR 15.5%-23.9%] absolute IMV risk reduction with HFNO to 7.4% risk reduction [2.9%-11.2%] with conventional oxygen). Patients with the highest predicted benefit were older, more frequently solid organ transplant recipients, less neutropenic, more severely hypoxemic/tachycardic, but with lower non-respiratory sequential organ failure assessment (SOFA) scores. In this subgroup, HFNO was associated with a significant lower incidence of IMV. In the observational cohort, the benefit-ranking performance was preserved, although absolute effect estimates were less calibrated. CONCLUSIONS: These results suggest HFNO might benefit specific immunocompromised patients with ARF. Further prospective studies are needed to validate these findings.
Abstract as published, via PubMed.
For healthcare professionals. The summary is generated by AI from the published abstract, and the evidence level is assigned automatically from the study design on the Oxford CEBM hierarchy. Neither is medical advice. Read the full paper before changing practice.