Longitudinally Stable, Clinically Defined Clusters of Patients with Asthma Independently Identified in the ADEPT and U-BIOPRED Asthma Studies.

Background: ADEPT (Airways Disease Endotyping for Personalized Therapeutics) and U-BIOPRED (Unbiased Biomarkers for the Prediction of Respiratory Disease Outcome Consortium) are independent asthma biomarker studies that aim to enable personalization of therapies.

Methods: Patients in both studies were identified by similar criteria, and similar clinical parameters and biomarkers were assessed in blood, sputum, and airway samples. Fuzzy partition-around-medoid clustering was performed on the ADEPT dataset (n = 154) and independently on the U-BIOPRED asthma dataset (n = 82), filtered to match ADEPT inclusion criteria. For both studies, the same eight easily measurable clinical variables were used, and ADEPT also included methacholine airway hyperresponsiveness. Models for cluster classification probabilities were derived and applied to the 12-month longitudinal ADEPT data and the full U-BIOPRED adult asthma dataset (n = 397) as independent external validation.

Measurements And Main Results: Four clusters were identified in the ADEPT-asthma study population with distinct clinical and biomarker profiles. In general, Cluster 1 consists of patients with mild asthma not treated with steroids and well controlled with preserved lung function and a low-inflammatory phenotype; Cluster 2 is partially controlled, with mild airflow obstruction but severe airway hyperresponsiveness and a Th2 phenotype (brittle phenotype); Cluster 3 is partially controlled with mild airflow obstruction but reduced vital capacity, less bronchodilator reversibility, and a non-Th2 phenotype with neutrophilic inflammation (chronic obstructive pulmonary disease-like); and Cluster 4 is poorly controlled, with marked airflow obstruction, marked bronchodilator reversibility, and a mixed inflammatory phenotype. Overall, the ADEPT clusters were stable over 12 months and reproduced by identifying four analogous clusters in the U-BIOPRED asthma dataset, with distributions for most clustering and nonclustering variables similar to ADEPT.

Conclusions: We report four clinical clusters in ADEPT and confirmed these by external validation in U-BIOPRED. The ADEPT clusters have distinct clinical and molecular characteristics, are stable over 12 months, and present opportunities for the development of tailored therapeutics for asthma.