**6. Asthma phenotypes based on cluster analyses**

In general, milder asthma phenotypes respond well to standard therapy with corticosteroids (with or without long-acting beta2-agonists), while those with more severe disease urged the development of new therapeutic modalities. To enable the development of effective (targeted) therapies, it is crucial to understand the pathophysiological mechanisms driving these subsets of asthmatic patients. Haldar et al. performed a cluster analysis on baseline data of 184 patients with mild to moderate asthma coming from different general practitioners (GP) and baseline data of 187 patients with refractory disease from specialist settings [3]. Additionally, a third dataset comprised baseline and longitudinal data of 68 patients with refractory disease followed for 12 months. Hierarchical cluster analysis revealed five different clusters, with some overlapping features between patients from GP and specialist origins. Most importantly, patients with concordant symptoms and (eosinophilic) inflammation (based on sputum analysis) were mostly coming from GP and were characterised by overall milder, often atopic, well-controlled disease, with a benign disease course. Alternatively, patients with uncontrol‐ led disease, characterised by either discordant symptoms (i.e. many symptoms, little airway eosinophilia or non-eosinophilic inflammation) or discordant inflammation (few symptoms, prominent airway eosinophilia) mostly originated from the specialist settings. Commonly found confounders consisted of obesity and non-compliance. Overall, these findings support‐ ed a symptom-guided management for mild-moderate "concordant"-type asthma, while "discordant"-type refractory asthmatics might benefit from inflammation-guided therapy [78].

Using unsupervised hierarchical cluster analysis in a group of 726 patients from the Severe Asthma Research Program (SARP) revealed five distinct clinical subphenotypes within this population [4], showing some overlap with the findings by Haldar et al. [3]. The results of both cluster analysis studies underscore disease heterogeneity, even in subsets of patients with similar clinical characteristics, with potentially different pathophysiological and immunolog‐ ical mechanisms, requiring different therapeutic approaches.

Further analysis into the molecular mechanisms underlying different asthma phenotypes revealed at least two distinct subsets with a "Th2-high" and a "Th2-low" profile, respectively [5], based on the expression of IL-13 inducible airway epithelial genes (POSTN (periostin), CLCA1 (chloride channel regulator 1) and SERPINB2 (serpin peptidase inhibitor clade B, member2)) as previously described by this research group. Not unexpectedly, patients with Th2-driven asthma responded well to inhaled corticosteroids while those with a "Th2-low" profile did not. Hence, there is an urgent need for effective therapeutic options for "Th2-low" asthmatic patients that appeared to comprise approximately 50% of the study population, and hence, in reality may be larger than originally thought.

Additionally, these findings urged phenotyping of patients (i.e. including an adequate target population) and/or using an appropriate disease model [8], for adequate interpretation of effectiveness data in targeted intervention studies. So far, several applicable (surrogate) biomarkers have been validated to phenotype potential responders and to monitor the effects of currently available (or under development) targeted therapies, i.e. anti-IgE, and Th2 pathway targeted therapies (anti-IL-5, anti-IL-4 and anti-IL-13) [86]. Presently, biomarkers including blood eosinophils, FeNO and serum periostin thus moved the first steps to person‐ alised medicine [87]. Further insight into the heterogeneity of Th2-driven/type 2 asthma, "Th2 low" subsets, as well as further refinement of sensitive (composite) biomarkers should be considered the next steps in this promising direction to optimise and personalise asthma management.
