Abstract
Background
Identifying effective therapies in heterogeneous conditions like acute hypoxemic respiratory failure (AHRF) depends on defining sub-phenotypes with distinct prognosis or therapeutic response. Prior efforts have focused on acute respiratory distress syndrome (ARDS), although ARDS is a minority of all AHRF patients, has limited reliability in research, and is similarly heterogeneous. We propose a novel AHRF sub-phenotype called persistent hypoxemic respiratory failure (PHRF), defined by PaO2:FiO2 ratio ≤ 300 in individuals still requiring mechanical ventilation on day 3 following intubation. We hypothesized individuals with PHRF (+ PHRF) have greater mortality than individuals without PHRF (-PHRF), irrespective of ARDS (+/-ARDS).
Methods
We included mechanically ventilated AHRF patients (n = 768) from a single-center prospective cohort of medical and surgical ICU patients. We estimated the relative risk of 28-day inpatient mortality associated with + PHRF compared to -PHRF using generalized linear models. We also compared mortality and baseline log-transformed plasma biomarkers of inflammation and endothelial activation/dysfunction in + PHRF/-ARDS, -PHRF/+ARDS, and + PHRF/+ARDS compared to -PHRF/-ARDS.
Results
Cumulative incidence of + PHRF was 53% (n = 408), of whom 51% were + ARDS by ICU day 3 (n = 209). +PHRF was associated with a 1.55-fold higher risk of death (95% CI: 1.02, 2.34) compared to -PHRF, adjusting for demographics, chronic respiratory disease, and APACHE-III. Absolute mortality was higher in + PHRF/+ARDS (23%) and + PHRF/+ARDS (15%) patients than -PHRF/+ARDS (12%) and -PHRF/-ARDS (7%) patients. Interleukin-6 was 2.36-fold (95% CI: 1.47, 3.80) and 2.62-fold (1.63, 4.20) higher in + PHRF/-ARDS and + PHRF/+ARDS compared to -PHRF/-ARDS; granulocyte-colony stimulating factor was 1.96-fold (95% CI: 1.28, 3.01) and 1.82-fold (95% CI: 1.16, 2.85) higher; angiopoeitin-2 was 1.32-fold (95% CI: 1.01, 1.73) and 1.59-fold (95% CI: 1.21, 2.09) higher. In contrast, -PHRF/+ARDS patients did not have significantly different mortality or plasma biomarkers from -PHRF/-ARDS patients in adjusted models.
Conclusions
PHRF represents a common sub-phenotype of patients with AHRF, characterized by higher mortality and higher biomarkers of inflammation and endothelial dysfunction than -PHRF. PHRF captures many high-risk patients not included in current ARDS definition who may share biologic features with ARDS. Identifying patients with PHRF can support clinical prognostication and targeted trial enrollment for investigational therapies in the broad AHRF population.