Classification and implementation of asthma phenotypes in elderly patients

Background No attempt has yet been made to classify asthma phenotypes in the elderly population. It is essential to clearly identify clinical phenotypes to achieve optimal treatment of elderly patients with asthma. Objectives To classify elderly patients with asthma by cluster analysis and developed a way to use the resulting cluster in practice. Methods We applied k-means cluster to 872 elderly patients with asthma (aged ≥65 years) in a prospective, observational, and multicentered cohort. Acute asthma exacerbation data collected during the prospective follow-up of 2 years was used to evaluate clinical trajectories of these clusters. Subsequently, a decision-tree algorithm was developed to facilitate implementation of these classifications. Results Four clusters of elderly patients with asthma were identified: (1) long symptom duration and marked airway obstruction, (2) female dominance and normal lung function, (3) smoking male dominance and reduced lung function, and (4) high body mass index and borderline lung function. Cluster grouping was strongly predictive of time to first acute asthma exacerbation (log-rank P = .01). The developed decision-tree algorithm included 2 variables (percentage of predicted forced expiratory volume in 1 second and smoking pack-years), and its efficiency in proper classification was confirmed in the secondary cohort of elderly patients with asthma. Conclusions We defined 4 elderly asthma phenotypic clusters with distinct probabilities of future acute exacerbation of asthma. Our simplified decision-tree algorithm can be easily administered in practice to better understand elderly asthma and to identify an exacerbation-prone subgroup of elderly patients with asthma.