Data-driven asthma phenotypes fail to accommodate personalized follow-up strategies in primary care

Introduction Asthma is a common and heterogeneous disease in primary care. Asthma phenotypes are recognisable clusters of for example clinical characteristics. Current asthma symptoms and previous exacerbations are used to assess the level of asthma control. Asthma control is used clinically to plan follow-up strategies. Aim The aim of this study is to examine if an data-driven algorithm based on sex and age of onset can categorize an asthma population at a primary care center into three phenotypes with different risk of disease. To investigate if the results can be generalized by comparing to an epidemiological survey in Sweden. Secondary aims are to investigate if these phenotypes predict the level of follow-up and which factors influence asthma control. Methods In this cross-sectional study, 335 participants from one primary care site and 1442 participants from an epidemiological study were compared on sex, age, medical treatment, respiratory allergy, smoking, asthma symptoms and exacerbations. Logistic regression analyses focusing on factors affecting asthma control were performed in a consolidated dataset. Results An adult asthma population can easily be categorized according to the data-driven algorithm. However, these phenotypes do not predict follow-up strategies. Clinical follow-up based on level of asthma control, did not differ between the phenotypes. There were statistically significant differences between the phenotypes regarding respiratory allergy and smoking. In the logistic regression, smoking has the highest odds for poor asthma control. Conclusion The clinical use of the data-driven phenotypes were limited. Follow-up strategies are probably best based on traditional clinical outcomes like asthma control