Conceived and designed the experiments: ERS EG TSK EL DYM. Performed the experiments: ERS EG ADS LPJ. Analyzed the data: ERS TSK EL. Contributed reagents/materials/analysis tools: ERS DYML. Wrote the paper: ERS EG TSK JVF DYML. Substantial contributions to analysis and interpretation of data: ERS EG ARS JVF DYML.
Dr. Sutherland has read the journal’s policy and has the following conflicts: Consultant: Forest Laboratories, GlaxoSmithKline, Merck, Novartis, Dey. Grants unrelated to the current study: Boehringer Ingelheim, Novartis. Educational presentation: Genentech. There are no patents, products in development or marketed products to declare. This does not alter the authors’ adherence to all the PLoS ONE policies on sharing data and materials, as detailed online in the guide for authors.
Asthma is a heterogeneous disease with variability among patients in characteristics such as lung function, symptoms and control, body weight, markers of inflammation, and responsiveness to glucocorticoids (GC). Cluster analysis of well-characterized cohorts can advance understanding of disease subgroups in asthma and point to unsuspected disease mechanisms. We utilized an hypothesis-free cluster analytical approach to define the contribution of obesity and related variables to asthma phenotype.
In a cohort of clinical trial participants (n = 250), minimum-variance hierarchical clustering was used to identify clinical and inflammatory biomarkers important in determining disease cluster membership in mild and moderate persistent asthmatics. In a subset of participants, GC sensitivity was assessed via expression of GC receptor alpha (GCRα) and induction of MAP kinase phosphatase-1 (MKP-1) expression by dexamethasone. Four asthma clusters were identified, with body mass index (BMI, kg/m2) and severity of asthma symptoms (AEQ score) the most significant determinants of cluster membership (F = 57.1, p<0.0001 and F = 44.8, p<0.0001, respectively). Two clusters were composed of predominantly obese individuals; these two obese asthma clusters differed from one another with regard to age of asthma onset, measures of asthma symptoms (AEQ) and control (ACQ), exhaled nitric oxide concentration (FENO) and airway hyperresponsiveness (methacholine PC20) but were similar with regard to measures of lung function (FEV1 (%) and FEV1/FVC), airway eosinophilia, IgE, leptin, adiponectin and C-reactive protein (hsCRP). Members of obese clusters demonstrated evidence of reduced expression of GCRα, a finding which was correlated with a reduced induction of MKP-1 expression by dexamethasone
Obesity is an important determinant of asthma phenotype in adults. There is heterogeneity in expression of clinical and inflammatory biomarkers of asthma across obese individuals. Reduced expression of the dominant functional isoform of the GCR may mediate GC insensitivity in obese asthmatics.
Cluster analyses of cross-sectional data from clinical populations have identified phenotypic subsets of patients with asthma, and the assessment of BMI in recent asthma cluster analyses has allowed assessment of the relationship of BMI to clinical features of asthma. Haldar and colleagues reported that obesity was associated with increased symptom expression, reduced eosinophilic airway inflammation, adult age of onset, and female sex, while also being associated with reduced clinical responsiveness to inhaled corticosteroids (ICS)
These reports notwithstanding, other analyses using standard comparative analytical approaches between asthmatics categorized by BMI have suggested that there is phenotypic heterogeneity among obese asthmatics, with some studies suggesting that asthma is more severe in obese asthmatics
To define the contribution of obesity and related variables to asthma phenotype.
Data from adults with persistent asthma participating in the common run-in period of the TALC
During the common run-in period, all participants received hydrofluoroalkane beclomethasone dipropionate (HFA-BDP) at a dose of 80 mcg (2 puffs of 40 mcg) twice daily for a 4-week period and were provided an albuterol metered-dose inhaled for rescue use. Clinical and inflammatory parameters were assessed as reported previously
All participants provided written informed consent. The protocol was reviewed and approved at each institutional IRB listed in the
Ward’s minimum-variance hierarchical clustering method
Differences between clusters were evaluated using analysis of variance or Student’s t-test for normally-distributed continuous variables. Chi-square analysis was used for categorical measures. Non-normally distributed data were log-transformed for analysis. Unadjusted analyses correlating continuous variables were performed using simple linear regression, with least-squares regression was used to perform adjusted analyses. Numeric data are presented as mean (standard deviation), except in the case of geometric mean (coefficient of variation) for log-transformed data.
Data from 250 participants were analyzed (
Measured at study initiation | |
n, subjects | 250 |
Sex (% male) | 32 |
Race (% white) | 59 |
Age (years) | 37.6 (12.5) |
Age of asthma onset (years) | 15.4 (14.7) |
Asthma duration (years) | 22.2 (12.2) |
BMI (kg/m2) | 29.9 (8.3) |
FEV1 (L) | 2.8 (0.8) |
FVC (L) | 3.9 (1.1) |
FEV1/FVC (%) | 71.8 (8.7) |
FEV1 (% predicted) | 82.2 (13.8) |
PC20 (mg/mL)† | 1.2 (1.2) |
Asthma Evaluation Questionnaire Score | 0.7 (0.8) |
Measured after 2 weeks HFA-BDP | |
Asthma Control Questionnaire Score | 1.0 (0.8) |
IgE (IU/mL) † | 105.4 (1.6) |
hsCRP (mg/L )† | 1.8 (1.4) |
Interleukin-6 (pg/mL)† | 1.4 (0.9) |
TNFα(pg/mL)† | 1.7 (0.8) |
Adiponectin (mcg/mL)† | 7.0 (0.7) |
Leptin (ng/mL)† | 10.8 (1.3) |
Measured after 4 weeks HFA-BDP | |
FENO (ppb) † | 19.9 (0.6) |
Sputum eosinophils (%) † | 0.8 (1.0) |
Asthma Evaluation Questionnaire Score | 0.6 (0.7) |
Asthma Control Questionnaire Score | 0.9 (0.8) |
Numeric data presented as mean (standard deviation), except †geometric mean (coefficient of variation), log-transformed for analysis.
Discriminant analysis revealed that 16 variables (
Variable | Partial R-Square | F | p |
BMI | 0.4105 | 57.1 | <.0001 |
AEQ (symptoms) | 0.3542 | 44.8 | <.0001 |
ACQ (control) | 0.1339 | 12.5 | <.0001 |
Race | 0.1039 | 9.4 | <.0001 |
Change in AEQ after 4 weeks of HFA-BDP | 0.1021 | 9.1 | <.0001 |
Age of asthma onset | 0.0991 | 8.8 | <.0001 |
FENO | 0.0845 | 7.4 | <.0001 |
Asthma controller type | 0.0724 | 6.2 | 0.0005 |
FEV1% predicted | 0.0696 | 5.9 | 0.0007 |
Leptin | 0.0651 | 5.5 | 0.0012 |
Asthma duration | 0.0630 | 5.2 | 0.0017 |
Adiponectin | 0.0601 | 5.0 | 0.0022 |
TNFα | 0.0587 | 4.9 | 0.0027 |
PC20 | 0.0474 | 3.9 | 0.0100 |
IgE | 0.0385 | 3.1 | 0.0282 |
FVC | 0.0358 | 2.9 | 0.0372 |
Analysis revealed four unique clusters of asthma patients, with characteristics as reported in
Nonobese female asthmatics | Nonobese male asthmatics | Obese uncontrolled asthma | Obese well-controlled asthma | p | |
Cluster number | 1 | 2 | 3 | 4 | - |
n | 114 | 52 | 30 | 54 | - |
Sex (% male) | 18 | 83 | 17 | 24 | <0.01 |
Race (% white) | 77 | 67 | 37 | 26 | <0.01 |
Age at onset (years) | 19.1 (16.1) | 9.8 (11.8) | 10.0 (10.8) |
16.1 (13.9) |
<0.01 |
Asthma duration (years) | 18.3 (11.3) | 26.2 (11.5) | 25.9 (12.0) | 24.6 (12.9) | <0.01 |
BMI (kg/m2) | 25.8 (5.0) | 26.9 (4.4) | 34.7 (8.0) | 38.5 (9.2) | <0.01 |
FVC (L) | 3.8 (0.7) | 4.9 (1.3) | 3.2 (0.9) | 3.3 (0.9) | <0.01 |
FEV1 (% predicted) | 87.7 (12.1) | 82.3 (16.4) | 73.5 (9.0) | 75.5 (11.1) | <0.01 |
FEV1/FVC (%) | 74.1 (8.7) | 68.5 (8.7) | 71.5 (8.0) | 69.7 (8.0) | <0.01 |
PC20, mg/mL |
1.2 (1.2) | 1.6 (1.3) | 0.7 (1.2) |
1.5 (0.9) |
0.02 |
ACQ Score | 0.8 (0.7) | 0.8 (0.6) | 1.8 (1.0) |
0.9 (0.9) |
<0.01 |
AEQ Score | 0.5 (0.6) | 0.4 (0.5) | 1.3 (0.9) |
0.7 (0.8) |
<0.01 |
FENO (ppb) |
20.8 (0.6) | 21.6 (0.6) | 24.8 (0.7) |
14.9 (0.7) |
<0.01 |
Eosinophils (%) |
0.8 (0.9) | 0.9 (1.0) | 0.8 (1.1) | 0.7 (0.9) | 0.44 |
IgE (IU/mL) |
78.1 (1.7) | 99.8 (1.3) | 201.9 (1.5) | 146.1 (1.4) | <0.01 |
hsCRP (mg/L ) |
1.3 (1.3) | 0.8 (1.1) | 4.2 (1.2) | 4.5 (1.1) | <0.01 |
Interleukin-6 (pg/mL) |
1.2 (1.0) | 0.9 (0.6) | 1.9 (0.7) | 2.1 (0.7) | <0.01 |
TNFα(pg/mL) |
2.0 (1.0) | 1.4 (0.4) | 1.4 (0.6) | 1.5 (0.7) | 0.03 |
Adiponectin (mcg/mL) |
10.2 (0.6) | 4.8 (0.6) | 6.3 (0.7) | 4.9 (0.7) | <0.01 |
Leptin (ng/mL) |
9.3 (1.0) | 3.4 (1.3) | 23.1 (0.9) | 29.3 (0.8) | <0.01 |
Use of medium/high-dose ICS (%) | 26 | 21 | 37 | 43 | 0.06 |
Table p values from Pearson chi-square test (Exact or CMH test) or analysis of variance comparing all 4 clusters.
indicates p<0.05 for comparison of clusters 3 and 4.
Numeric data presented as Mean (Standard Deviation), except.
Geometric Mean (Coefficient of Variation), log-transformed for analysis.
ACQ: asthma control questionnaire score after 4 weeks of HFA-BDP, AEQ: asthma evaluation questionnaire score after 4 weeks of HFA-BDP.
As reported in
Non-obese clusters 1 and 2 differed from each other with regard to baseline lung function, with FEV1% predicted of 87.7% (12.1) in cluster 1 and 82.3% (16.4) in cluster 2 (p = 0.02). A similar trend was seen with FEV1/FVC ratio, which was 74.1 (8.7)% in cluster 1 and 68.5 (8.7)% in cluster 2 (p<0.01). These two clusters also differed with regard to the percent of subjects who were male, at 18 vs. 83% (p<0.01) and age at asthma onset, at 19.1 (16.1) vs. 9.8 (11.8) years (p<0.01). Asthma symptom expression (AEQ scores of 0.5 (0.6) and 0.4 (0.5), p = 0.66) and degree of asthma control (ACQ scores of 0.8 (0.7) and 0.8 (0.6), p = 0.81) were similar between the two clusters, and these clusters were also similar with regard to biomarkers of inflammation (FENO, IgE and hsCRP), indicating that the observed differences between clusters in lung function, sex, and age at disease onset were not linked with a distinct inflammatory phenotype (
Markers of
When we analyzed the correlation between log-transformed GCRα expression in all 49 participants, we observed an inverse correlation (
Finally, to determine if reduced GCRαexpression might be one factor leading to reduced
The application of an hypothesis-free cluster analytical approach to a well-characterized cohort of adults with mild-to-moderate persistent asthma demonstrates that obesity is a determinant of clinical phenotype in asthma, playing a more significant role than other commonly-assessed clinical, physiologic or inflammatory variables. Of the four distinct clusters of asthma revealed, two had BMI in the obese range and two did not. There was heterogeneity of airway inflammation, symptoms and control in the obese clusters, suggesting that asthma phenotype is not uniform in obese individuals. In the two non-obese clusters, sex emerged as an important determinant of cluster membership; one cluster had a predominance of males the other a predominance of females, with comparatively earlier age of onset and lower lung function (as reflected by FEV1% predicted) in the male-predominant cluster. Additionally, we have demonstrated that
Our findings also suggest that the mechanisms which underlie clinical response to GC in obese asthmatics are complex and likely involve an interaction between alterations in GC-mediated anti-inflammatory processes and both systemic and airway inflammation. This conclusion is based on our observation that while evidence of
Potential limitations of our must be considered: first, our analytical approach is hypothesis-independent. While this provides the opportunity to identify new associations that one might not be able (on the basis of current knowledge) to prespecify, it runs the risk of returning results that are counterintuitive or which differ from current hypothetical constructs of disease. Second, as with any meta-analytical technique, the results are entirely dependent on the data available for entry into the analysis. Thus, while we have attempted to include all clinically-relevant data, the derivation of our data from a clinical trial dataset limits the availability of certain data (e.g. socioeconomic or environmental status) and may introduce issues of generalizability given the highly-selected nature of clinical trial participants. Next, as with any cross-sectional data, we are unable to comment on causation,
Our analytical approach and validation of clinical phenotypes with studies of the molecular mechanisms of GC insensitivity in asthma strengthen the assertion that patients with asthma, both adult and pediatric
Clinicians frequently encounter obese asthmatics who do not respond optimally to therapy, but no specific guidance currently exists in national and international guidelines as to the optimal therapeutic approach to the obese asthmatic
The following Asthma Clinical Research Network sites and investigators participated in the parent clinical trials which obtained the clinical data analyzed in this study.
(DOCX)