Definition of robust subphenotypes of Hospital-Acquired Pneumonia associated with all-cause mortality in data from four national observational cohorts and validation in one international randomized trial

AUTHORS

Florian Pierre Martin(1)*, Cécile Poulain(1,2)*, Jelle Haitsma Mulier(3,4), Ana Motos(1,5,6,7), Ismaël Ogan(2), Emmanuel Montassier(1,8), Yoann Launey(9), Sigismond Lasocki(10), Raphaël Cinotti(2,11), Lennie Derde(3,4), Antoni Torres(5,6,7), Olaf Cremer(3), Antoine Roquilly(1,2) and the Atlanrea Study group

* Equal first authors

AFFILIATIONS

1 Nantes Université, Inserm, CHU Nantes, Center for Research in Transplantation and Translational Immunology, UMR 1064, F-44000 Nantes, France
2 Nantes Université, CHU Nantes, Service d’Anesthesie Réanimation, F-44000, Nantes, France
3 Department of Intensive Care Medicine, University Medical Center Utrecht, Utrecht, the Netherlands
4 Julius Centre for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands
5 Servei de Pneumologia. Hospital Clinic. Universitat de Barcelona,Fundació Clinic IDIBAPS. ICREA. CIBERES. Spain
6 Department of Pulmonology, Thorax Institute, Hospital Clinic de Barcelona, Barcelona Spain
7 Department of Medicine, School of Medicine University of Barcelona, Barcelona Spain
8 Service des Urgences, Nantes Université, CHU Nantes, Nantes, France
9 Reanimation, CHU Rennes, Rennes, France
10 Reanimation, CHU Angers, Angers, France
11 Nantes Université, Univ Tours, CHU Nantes, CHU Tours, INSERM, MethodS in Patients-centered outcomes and HEalth Research, SPHERE F-44000 Nantes, France

Corresponding author. Antoine Roquilly. Institut de Recherche en Santé 2 Nantes Biotech, Nantes University, 44000 Nantes, France. Ph: +33 253482230 e-mail: [email protected]
Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Antoine Roquilly ([email protected])

ABSTRACT

Background. Despite optimal antimicrobial therapy, the rate of treatment failure of hospital-acquired pneumonia (HAP) routinely reaches 40% in critically ill patients. Subphenotypes have been identified within sepsis and acute respiratory distress syndrome with important therapeutic implications. We assessed whether HAP sub-phenotypes exist and if they are associated with response to treatment.
Methods. We used data from four independent cohorts of critically ill patients in France (PNEUMOCARE, n=511, ATLANREA, n=401), Netherlands (MARS, n=1351) and Europe (ENIRRI, n=900) to investigate HAP heterogeneity using unsupervised clustering based on clinical and routine biological variables available at HAP diagnosis. To evaluate HAP subphenotype robustness, we developed a machine learning-based workflow to build a simplified classification model in the discovery datasets, then applied this classification model to a replication dataset from an international randomised clinical trial which has compared linezolid to tedizolid for HAP (VITAL, n=726 patients). The primary outcome was the subphenotype association with 28-day all-cause mortality. We also studied the subphenotype associations with treatment failure at test-of-cure, with the respiratory microbiome and cytokinome in an ATLANREA subgroup and the response to tedizolid in the VITAL study.
Findings. The derivation cohort included 3163 patients with HAP (mean age 58 (±17) years; 2316 (73%) males; 1742 (55%) early-onset HAP), and the validation cohort included 726 patients (mean age 58 (±18) years; 503 (69%) males; 333 (46%) early-onset HAP). We tested twelve metrics and determined that a two-cluster model best fits all cohorts. HAP phenotype 2 was characterised by severe respiratory microbiome dysbiosis, high inflammatory chemokine blood levels, and lower body temperature and PaO2/FiO2 than phenotype 1 patients. Across all cohorts, the rates of 28-day mortality and treatment failure at test-of-cure were higher among phenotype 2 vs phenotype 1 (P<0.01 for all comparisons). In the VITAL trial, the relative risk of treatment failure with tedizolid versus linezolid was increased in phenotype 1 (RR= 1.52; 95%CI 1.12-2.06) but not in phenotype 2 (RR= 0.98; 95%CI 0.7-1.38, pinteraction = 0.14).
Interpretation. In this retrospective analysis of HAP data sets, two robust clinical phenotypes that correlated with respiratory microbiome composition, systemic inflammation and clinical outcomes were identified. Identifying HAP phenotypes might help implement personalized antimicrobial therapy and patient selection for future clinical trials.\

Funding. European Union’s Horizon 2020 research and innovation program under grant agreement number 847782 (HAP2 project).

KEYWORDS Hospital-acquired pneumonia, sub-phenotype, lung microbiome, cytokines, machine learning, tailored treatment.

Code DOI: 10.5281/zenodo.14186470