Robust airway microbiome signatures in acute respiratory failure and hospital-acquired pneumonia

Abstract

Respiratory microbial dysbiosis is associated with acute respiratory distress syndrome (ARDS) and hospital-acquired pneumonia (HAP) in critically ill patients. However, we lack reproducible respiratory microbiome signatures that can increase our understanding of these conditions and potential treatments. Here, we analyze 16S rRNA sequencing data from 2,177 respiratory samples collected from 1,029 critically ill patients (21.7% with ARDS and 26.3% with HAP) and 327 healthy controls, sourced from 17 published studies. After data harmonization and pooling of individual patient data, we identified microbiota signatures associated with ARDS, HAP and prolonged mechanical ventilation. Microbiota signatures for HAP and prolonged mechanical ventilation were characterized by depletion of a core group of microbes typical of healthy respiratory samples, and the ARDS microbiota signature was distinguished by enrichment of potentially pathogenic respiratory microbes, including Pseudomonas and Staphylococcus. Using machine learning models, we identified clinically informative, three- and four-factor signatures that predicted ARDS, HAP and prolonged mechanical ventilation with relatively high accuracy (area under the curve of 0.751, 0.72 and 0.727, respectively). We validated the signatures in an independent prospective cohort of 136 patients on mechanical ventillation and found that patients with microbiome signatures associated with ARDS, HAP or prolonged mechanical ventilation had longer times to successful extubation than patients lacking these signatures (hazard ratios of 1.56 (95% confidence interval (CI) 1.07-2.27), 1.51 (95% CI 1.02-2.23) and 1.50 (95% CI 1.03-2.18), respectively). Thus, we defined and validated robust respiratory microbiome signatures associated with ARDS and HAP that may help to identify promising targets for microbiome therapeutic modulation in critically ill patients.