Real-world clusters of severe asthma from the Italian Registry on Severe Asthma (IRSA): a longitudinal Latent Class Analysis

Biomarkers in asthma are commonly used to categorize patients with a priori cut-off values, mostly based on criteria for prescription of biologics. The purpose of this study was to identify real-world clusters, using an unsupervised machine-learning technique (Latent Class Analysis) on patients with severe asthma from the Italian Registry on Severe Asthma (IRSA), and to assess their changes after one year of follow-up. From 632 patients, 3 real-world classes were identified (Figure). Class B, with the highest value of eosinophils, FeNO, and exacerbations. Class B, with the lowest frequency of atopy, the worst control of asthma and FEV1/FVC ratio, and the higher frequency of treatment with anti-IL5(R). Class C, with a significant dissociation between eosinophils and FeNO. At follow-up, Class B achieved the best control of symptoms, reduction of exacerbations, and improvement in lung function. The 3 clusters exhibit significantly different baseline characteristics, treatments’ choices, and improvements after one year of follow-up. These differences might reflect distinctive new real-world phenotypes with potential impact on asthma management and recognition of particularly “complex” and hidden traits. <fig><object-id>erj;64/suppl_68/PA4803/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig>