RT Journal Article
SR Electronic
T1 Statistical cluster analysis on the BTS refractory asthma cohort
JF European Respiratory Journal
JO Eur Respir J
FD European Respiratory Society
SP p4163
VO 38
IS Suppl 55
A1 Chris Newby
A1 Liam Heaney
A1 Andrew Menzies-Gow
A1 Rob Niven
A1 Chris Brightling
YR 2011
UL http://erj.ersjournals.com/content/38/Suppl_55/p4163.abstract
AB Background: Severe asthma is no longer believed to be a single homogeneous condition but rather a heterogeneous disease possibly containing subsets of patients. A number of clustering algorithms have been carried out on various datasets of asthmatics of differing severities but few have been confirmed by follow-up studies.Aims and objectives: To further understand these clusters, statistical analysis was carried out to determine the patterns of variation seen in a large cohort with well characterised refractory asthma (Heaney. L et al,2010).Methods: Independent structure was determined using Bayesian factor analysis followed by statistical cluster analysis. The analysis revealed clinically relevant factors and statistical significant clusters.Results: 4 factors were obtained from the Bayesian factor analysis. Factor 1 described lung function, factor 2 atopy, factor 3 BMI and factor 4 Inflammation. 4 clusters were found. These clusters were significant at the 0.05 level for over 26 asthma variables e.g. Pre bronchodilator FEV1 (p<0.001), age of asthma onset (p<0.001), gender (p=0.041), atopy (p=0.004). Notably,the proportion of each cluster was differentially represented across different Centres.Conclusions: Cluster 1, described an older group of patients with the longest disease duration. Cluster 2, a mainly female group that had late onset and high BMI. Cluster 3 described an atopic group with high blood eosinophils and low age of onset. Cluster 4 were mainly male, had good lung function and were also the youngest. These clusters have been differentially represented in previous work in asthmatic populations of variable severity. Further analysis will be carried out longitudinally to determine outcomes for the clusters.