A LARGE registry study has identified key biological and clinical factors that directly and indirectly drive the risk of severe asthma exacerbations, offering new insights into risk prediction and patient management.
Accurate forecasting of severe exacerbations remains a cornerstone of care but has proven challenging due to the interplay of multiple predictors. To address this, researchers analysed data from 6,814 biologic-naive adults enrolled in the Severe Asthma Registry between 2017–2021. Predictors included demographics, lung function, biomarkers of inflammation, health care use, medication history, prior exacerbations, and comorbidities. A Bayesian network model was developed by integrating expert clinical input with machine learning algorithms, and the model was validated internally. An influence diagram incorporating decision and utility nodes was also generated to support clinical application.
The analysis revealed that blood eosinophil count, fractional exhaled nitric oxide (FeNO), and FEV₁ were the strongest direct predictors of the transition from prior to future severe exacerbations. Chronic rhinosinusitis emerged as an important indirect contributor, influencing all three immediate predictors: blood eosinophils, FeNO, and % predicted FEV₁. In addition, macrolide use was independently associated with a history of exacerbations and exerted a downstream effect on the likelihood of future episodes. Model performance showed moderate discrimination in both cross-validation approaches and strong calibration in train-test evaluation, supporting its robustness.
These findings highlight a prediction pathway centred on inflammatory biomarkers and lung function, modified by comorbid chronic rhinosinusitis and medication use. The results provide a framework to enhance risk stratification and support shared decision-making in the management of severe asthma.
Reference
Yadav CP et al. Prediction pathway for severe asthma exacerbations: a Bayesian network analysis. Chest. 2025;168(2):301-16.
