Multiview Cluster Analysis Identifies Variable Corticosteroid Response Phenotypes in Severe Asthma

Abstract

Rationale: Corticosteroids (CSs) are the most effective asthma therapy, but responses are heterogeneous and systemic CSs lead to long-term side effects. Therefore, an improved understanding of the contributing factors in CS responses could enhance precision management. Although several factors have been associated with CS responsiveness, no integrated/cluster approach has yet been undertaken to identify differential CS responses. Objectives: To identify asthma subphenotypes with differential responses to CS treatment using an unsupervised multiview learning approach. Methods: Multiple-kernel k-means clustering was applied to 100 clinical, physiological, inflammatory, and demographic variables from 346 adult participants with asthma in the Severe Asthma Research Program with paired (before and 2-3 weeks after triamcinolone administration) sputum data. Machine-learning techniques were used to select the top baseline variables that predicted cluster assignment for a new patient. Measurements and Main Results: Multiple-kernel clustering revealed four clusters of individuals with asthma and different CS responses. Clusters 1 and 2 consisted of young, modestly CS-responsive individuals with allergic asthma and relatively normal lung function, separated by contrasting sputum neutrophil and macrophage percentages after CS treatment. The subjects in cluster 3 had late-onset asthma and low lung function, high baseline eosinophilia, and the greatest CS responsiveness. Cluster 4 consisted primarily of young, obese females with severe airflow limitation, little eosinophilic inflammation, and the least CS responsiveness. The top 12 baseline variables were identified, and the clusters were validated using an independent Severe Asthma Research Program test set. Conclusions: Our machine learning-based approaches provide new insights into the mechanisms of CS responsiveness in asthma, with the potential to improve disease treatment.

Keywords: asthma phenotype; corticosteroids; eosinophils; severe asthma.

Figure 1.

Heatmap of the four clusters identified among 346 patients by the multiple-kernel k-means clustering method. Rows represent the patients, and columns represent the variables in each view (for an ordered list of variables, see Table E1).

Figure 2.

Five demographic variables differed by cluster. (A and E) Cluster 3 had the oldest subjects (A) with the latest age at onset (E). (B) Cluster 1 had the largest percentage of males, and cluster 4 had the lowest. (C) Cluster 4 had the highest body mass index (BMI). (D) Cluster 4 had the highest percentage of black/African Americans, and cluster 3 had the lowest percentage.

Figure 3.

Distinguishing features of clusters 1 and 2. (A–D) At baseline, clusters 1 and 2 had opposite sputum neutrophil (A) and macrophage (C) percentages and reciprocal changes after triamcinolone treatment (B and D). (E and F) They also had the highest number of specific IgE (E) and the highest percentage of patients with perennial allergy (F).

Figure 4.

Distinguishing features of cluster 3. (A, C, and E) At baseline, cluster 3 had the highest sputum eosinophil percentages (A) and lowest lung function (C and E). (B, D, and F) After triamcinolone treatment, cluster 3 had the greatest decrease in sputum eosinophil percentages (B) and the largest improvement in lung function (D and F). The box plots in this figure are shown without outliers. See Figure E5 for box plots showing all data points for the same variables. BD = bronchodilator.

Figure 5.

Distinguishing features of cluster 4. (A–E) At baseline, cluster 4 had the worst activity limitation Asthma Quality of Life Questionnaire (AQLQ) (A); highest pulse (B), total white blood cell (WBC) count (C), and healthcare use (D); and largest increase in FVC after albuterol treatment (E). (F) After triamcinolone treatment, the increase in FVC after albuterol treatment decreased the most among cluster 4 patients. The box plots in this figure are shown without outliers. See Figure E6 for box plots showing all data points for the same variables. ER = emergency room.