High-flow nasal cannula versus noninvasive ventilation in patients with hypoxemic respiratory failure: a prospective cohort study

High-flow nasal cannula (HFNC) and non-invasive ventilation(NIV) are commonly used for hypoxemic respiratory failure, but their comparative efficacy remains unclear. This prospective cohort study enrolled 259 non-hypercapnic hypoxemic respiratory failure patients (PaO₂ <60 mmHg, PaCO₂ <50 mmHg, pH ≥ 7.30) in a Chinese ICU. Patients were allocated to HFNC (n = 128) or NIV (n = 131) based on physician judgment. Primary outcome was 28-day intubation rate; secondary outcomes included 28-day mortality, treatment withdrawal, and ICU/hospital stay duration. Baseline characteristics were similar except for higher respiratory/heart rates and disease severity in the NIV group. Before adjustment, the HFNC group had a lower intubation rate (P < 0.05) and fewer events of death or treatment withdrawal(P<0.01). After propensity score matching, all group differences became non-significant: intubation(aOR 2.61,95%CI 0.58-11.68,P = 0.21), hospital mortality (aOR 0.91,95%CI 0.31-2.62,P = 0.86),death or withdrawing treatment༈aOR 0.54,95%CI 0.26-1.13,P = 0.10),and the composite endpoint (aOR = 1.14,95%CI 0.42-3.08, P = 0.80).The durations of ICU stay (8 vs. 8 days, P = 0.87) and hospital stay (13 vs. 12 days, P = 0.10) showed no significant differences both before and after matching.Patients transitioning from HFNC to NIV had longer ICU stays than NIV-only patients (12 vs. 8 days, P < 0.05). Outcomes did not differ between HFNC-failure patients transitioning to NIV or intubation (P > 0.05). HFNC and NIV show similar efficacy in preventing intubation and reducing mortality. NIV may shorten ICU stay in severe cases, but escalation to NIV or intubation after HFNC failure does not improve outcomes. Treatment should be individualized based on disease severity and patient response. However, given the non-randomized design and potential for residual confounding despite multivariate adjustment, these findings should be interpreted with caution and require validation in randomized controlled trials.