Du är nu på väg att lämna en GSK-hemsida

Denna länk leder till en hemsida som inte tillhör GSK. GSK tar inget ansvar för innehållet på tredje parts hemsidor.



Why phenotype in severe asthma?

Phenotyping in the clinic

Asthma phenotyping is an evolving area of research, and a general consensus of what defines each phenotype has still to be developed. Phenotyping should evolve into endotyping, which links to clinical features with identifiable molecular pathways. 1

Ultimately, the evolution occurring in patient phenotyping is set to optimise therapy by enabling targeted and personalised biomarker-driven treatment for severe asthma. 1 2

Using biomarkers to establish phenotypes for treatment

Phenotyping methods 

Current understanding of phenotypes in asthma

In severe asthma, identification of phenotypes has commonly been approached in two ways:3

  1. Using existing definitions of a phenotype based on clinical characteristics of subjects
  2. Analysing pathobiologic differences in sputum or bronchoscopy specimens

Both of these approaches have been used to further clarify severe asthma phenotypes, via a method known as ‘cluster analysis’. 4

What is cluster analysis?

Cluster analysis is a multivariate mathematical method. It uses specific variables to quantify the similarity between individuals and group them into clusters.5

Cluster analysis is an evolving area in the asthma field. Individual studies have already described a number of phenotypes, which significant overlap with already established phenotypes such as late-onset eosinophilic and early-onset allergic asthma.2 5-6

Theoretical grouping of asthma phenotypes

Schematic showing patients divided by period of onset and by eosinophilic versus non-eosinophilic airway inflammation.

Schematic showing adult patients, divided by period of onset and by eosinophilic versus non-eosinophilic airway inflammation.
Adapted from Hekking and Bel, 2014.

Despite overlap, the approaches used to date have not yet led to a definitive suite of asthma phenotypes.2 5-6The studies published to date have shown differences and similarities across the phenotypes identified, likely due to demographic and clinical differences in the populations studied and variations in statistical methods. 2The cluster analysis process is inherently biased, because variables must be selected by the triallists and as studied populations are limited to those recruited based on the overarching study criteria. Some studies, for example, did not include typical lung function measures, whilst others did not include inflammatory markers. 25However, these differences support the concept of disease heterogeneity in asthma and suggest differences in pathophysiologic mechanisms between the clusters, leaving room for evolution into endotyping. 1 2 5-6

The studies described below are examples of cluster analyses that have identified particular asthma phenotypes.

  • First cluster analysis to phenotype asthma

    Haldar P, et al. Am J Respir Crit Care Med 2008;178:218–24. 4

    This study was the first to apply the principles of cluster analysis to distinguishing asthma phenotypes. It highlighted the discordance between symptoms and inflammation in severe/refractory asthma.

    Datasets from three discrete populations with asthma were compared: a mild-to-moderate, primary care-managed group (n=184) and a secondary care-managed group with refractory asthma (n=187).

    The study identified two clusters common to both the primary care and secondary care refractory asthma population datasets:

    • Early-onset allergic: concordant airway dysfunction, symptoms and eosinophilic airway inflammation; lower clinical adherence
    • Obese, noneosinophilic: later-onset, predominantly female, highly symptomatic without eosinophilic inflammation, high BMI, steroid resistance

    A third patient dataset with predominantly refractory asthma (n=68) was collected from a prospective clinical study. The data were analysed for treatment outcomes and compared with eosinophilic inflammation levels. Clusters identified were comparable to the clusters in the secondary care refractory dataset.

    Patients with refractory asthma tended to be either highly symptomatic with low-level inflammation, or have high-level inflammation but few symptoms. These groups were the most difficult to treat, whereas the groups whose symptoms correlated with their inflammation levels were well managed in primary care.

    Click here for the full article.

  • Severe Asthma Research Programme (SARP)

    Moore WC, et al. Am J Respir Crit Care Med 2010;181:315–23. 3

    The SARP group analysed a dataset from a population of 726 non-smoking subjects aged 12 years and older, who met the ATS definition of severe asthma. Five distinct clinical asthma phenotypes were identified. All clusters contained patients with severe asthma, which supports the concept of variability in clinical asthma.

    Patients were analysed according to five clinical variables: lung function, onset and duration of disease, sex, SABA usage and ICS/OCS usage. Five clusters were identified:

    • Mild allergic asthma: early onset; allergic; normal lung function; low medication and healthcare usage; minimal sputum eosinophilia
    • Mild to moderate allergic asthma: most common cluster; early onset; allergic; borderline FEV1, but reverse to normal; low medication and healthcare usage; minimal sputum eosinophilia
    • Severe older‐onset asthma: older; very late onset; obesity prevalent; less allergic; slightly decreased FEV1 with some reversibility; high medication requirement including OCS; sputum eosinophilia
    • Severe variable allergic asthma: early onset; allergic; severely decreased, reversible FEV1; highly symptomatic; high medication requirement; sputum eosinophilia
    • Severe fixed‐airflow asthma: older; longest duration; less allergic; severely decreased FEV1 with low reversibility; highly symptomatic; high medication usage including OCS; comorbidities prevalent; sputum eosinophilia and neutrophilia

    Clusters were validated using pre- and post-bronchodilator FEV1 and age of onset. 80% of patients were assigned to the correct cluster, which suggests a potentially simple method for phenotyping asthma.

    Click here for the full article.

  • Adult asthma population-based study

    Siroux V, et al. Eur Respir J 2011;38:310–7. 7

    This was the first study to use cluster analysis in an adult asthma population-based study. The study’s approach allowed it to recruit and distinguish a wide range of phenotypes and supports the use of statistical methods such as cluster analysis for phenotype identification. In particular, the results indicate treatment is an important factor when classifying patients with asthma.

    The analysis was conducted in two large adult European asthma patient populations and was analysed independently. Clusters were based on age, sex, asthma symptoms over the previous year, allergic characteristics, lung function, airway hyperresponsiveness and asthma treatment.

    Similar results were observed in both populations. Two clinically relevant phenotypes stood out in particular as being present in both groups:

    • Active-treated allergic childhood-onset asthma: younger subjects, early disease onset, allergy, active disease at time of examination
    • Active-treated adult-onset asthma: older subjects, predominantly female, late disease onset, active disease at time of examination, highly symptomatic, frequent exacerbations

    The other two phenotypes identified were characterised by inactive or mild, untreated asthma and differed by allergy and age of onset.

    Click here for the full article

  • First clustering in an Asian population

    Kim T-B, et al. Eur Respir J 2013;41:1308–14. 8

    This is the first study to assess asthma clustering in an Asian population. There were no significant correlations between asthma severity and BMI or allergy, which differs from similar analyses conducted in Western populations.

    Kim, et al. analysed two adult Korean patient cohorts with either active, bronchodilator-responsive allergic asthma, or who had been diagnosed and treated according to GINA guidelines (N=2567). Patients were clustered based on FEV1, BMI, age at disease onset, allergic status, smoking history and history of exacerbation-related hospital use.

    The analysis identified four asthma clusters:

    1. Smoking asthma: predominantly male, late onset, high healthcare usage, relatively preserved FEV1
    2. Severe obstructive: allergy predominant, reversible FEV1 limitations, high bronchoconstriction
    3. Early-onset allergic: the largest cluster, young age of onset, high levels of allergy, low smoking levels
    4. Late-onset mild: highest lung function, late onset

    Click here for the full article

  • Unbiased Biomarkers for the Prediction of Respiratory Disease outcomes (U-BIOPRED)

    Shaw DJ, et al. Eur Respir J 2015; Epub ahead of print. 9

    The U-BIOPRED study aimed to improve the understanding of asthma using a systems biology approach. It was a prospective multicentre study that included patients with severe asthma, mild/moderate asthma and healthy controls.

    Analysis of data from the adult participants found a discord between medication use and airway inflammation. Additionally, smoking status did not necessarily correlate with asthma severity or symptoms, although mean age of asthma onset was later in smokers and ex-smokers than non-smokers.

    Patients were grouped at screening according to asthma severity and smoking status, for comparison with a cohort of healthy non-smoking controls (N=611). They were analysed by symptoms, quality of life, asthma control, allergic status and genotype over 12–24 months.

    The following characteristics were associated with each predefined adult asthma group:

    1. Severe non-smoking: predominantly female, higher ICU admission rates vs other groups, high BMI, nasal polyps and GERD prevalence, low mucolytic usage
    2. Smoking and ex-smoking with severe asthma: late-onset, high BMI, nasal polyp and GERD prevalence, low SABA and antihistamine usage
    3. Mild/moderate non-smoking: high positive atopy testing, asthma generally well controlled and low comorbidity prevalence

    Click here for the full article


  1. Chung KF, et al. Eur Respir J 2014;43:343–73.
  2. Wenzel SE. Nat Med 2012;18:716–25.
  3. Moore WC, et al. Am J Respir Crit Care Med 2010;181:315–23.
  4. Haldar P, et al. Am J Respir Crit Care Med 2008;178:218–24.
  5. Xie M, Wenzel SE. Chin Med J (Engl) 2013;126:166–74.
  6. Hekking PPW, Bel EH. J Allergy Clin Immunol Pract 2014;2:671–80.
  7. Siroux V, et al. Eur Respir J 2011;38:310–7.
  8. Kim T-B, et al. Eur Respir J 2013;41:1308–14.
  9. Shaw DJ, et al. Eur Respir J 2015;46:1308–21.