Cluster analysis has greatly expanded our investigation of asthma phenotypes. These multivariate approaches explore the heterogeneity of asthma in a way hypothesis-driven univariate analyses cannot. Nearly all cluster analyses currently published, however, are cross-sectional glimpses of groups of patients as they look today without a preview of how they will look in the future (1–6). What is the natural disease progression of these new asthma phenotypes?
A good analogy is to conceptualize asthma severity as a three-dimensional labyrinth. Cross-sectional “cluster” phenotypes are different entry portals into the maze. Within the labyrinth, there are ladders that allow you to ascend to the upper level (improving asthma severity). Likewise, there are chutes that lead to the dungeon (worsening asthma severity). Many participants, however, stay on the same level, managing to sidestep the ladders and chutes (stable asthma severity). The question is whether the initial “entry” phenotype determines the journey through the labyrinth. Can a multivariate “cluster” phenotype warn of potential chutes or identify potential ladders? The article in this issue of the Journal by Boudier and colleagues (pp. 550–560) describes asthma phenotypes by cluster analysis at two time points 10 years apart (7). It is a glimpse into the natural history of complex asthma phenotypes and variability over time.
The most important elements in cluster analysis are the strength of the cohort and the phenotypic characteristics included in the analysis. The ideal cohort has a large number of participants with a spectrum of asthma severity (heterogeneity). Ideally, these participants would have undergone extensive clinical and biologic characterization. In general, it is difficult to have both of these ideals. Thus, current asthma cluster analyses fall into two groups, large population studies with fewer variables or smaller cohorts with extensive variables (1–6). The two approaches are complementary and each has merits as we explore asthma heterogeneity.
The cohort used in the study by Boudier and colleagues is of the former group, a combination of three very large epidemiologic populations with a limited number of variables (7). Participants with asthma were identified by self-report on the basis of answers to questionnaires, and there are inherent issues with misdiagnosis of disease, especially in cohorts that include smokers who could have other etiologies for respiratory symptoms. Only 12% of the asthma subjects in these combined cohorts had an FEV1 < 80% predicted, suggesting that most of the participants had milder disease. Thus, the spectrum of asthma severity is narrow, and this impacts the ability of the analysis to divide subjects into distinctive clinically relevant clusters.
The seven asthma phenotypes described are most markedly distinguished by allergic sensitization as measured by skin prick testing or specific IgE levels; four of the phenotypes are atopic, and three are clearly not. Nearly all subjects in phenotypes A, D, E, and F are atopic. It is reassuring that two-thirds of these subjects remain in the same phenotypic group over the 10 years. There is some movement between phenotypes A and F; these groups have normal lung function, and many do not have positive bronchoprovocation tests, begging the question as to whether they are primarily patients with allergic rhinitis, some of which may have airway hyperresponsiveness and asthma-like symptoms when provoked. Phenotypes D and E represent the spectrum of traditional allergic asthma with positive methacholine challenges; 25% of subjects had an FEV1 < 80% predicted, consistent with chronic airflow obstruction and more severe disease. The movement of 38% of patients with phenotype D (more severe) to phenotype E (less severe) with only 11% worsening from E to D is inspiring to clinicians. This finding suggests that most of the patients with mild-to-moderate persistent asthma were stable or improved in the “real world” over a 10-year period.
The three nonallergic asthma phenotypes are much more difficult to interpret and not as satisfying. Many of the patients in these clusters do not have airway hyperresponsiveness to methacholine, again bringing up the possibility of a misdiagnosis of asthma. In other cluster analyses, less allergic phenotypes are most often highly symptomatic, older asthma subjects with later-onset disease, higher body mass indices, and multiple comorbidities (1, 3, 5, 8). We are not given any of these clinical characteristics for subjects in phenotypes B, C, and G, but the large odds ratios associated with adult onset asthma and phenotypes C and G suggests that these phenotypes may be similar to those described in prior cluster analyses. There is much greater movement of patients between phenotypes in the nonallergic groups; 20–45% of patients change phenotypes at the longitudinal assessment. These findings suggest that there may be a continuum of patients with nonallergic asthma rather than discrete nonallergic asthma phenotypes whose progression can be followed over time.
The study by Boudier and colleagues is a provocative look at disease progression in multivariate asthma phenotypes identified from a large cohort. It offers optimism that mild-to-moderate allergic asthma phenotypes may not progress to more severe disease phenotypes over time, implying that there are more ladders than chutes for these patients. The picture is less clear for the nonallergic phenotypes, groups that are confounded by the comorbidities common in later-onset asthma. It is important to recognize that there were very few patients with severe asthma in these combined cohorts, and thus, the temporal stability of cluster phenotypes in this subgroup of patients cannot be inferred. Severe asthma patients have the most to gain with longitudinal multivariate phenotyping approaches, since they likely have more chutes to avoid (recurrent exacerbations), and are in need of new ladders to climb (novel therapeutic approaches) (9, 10).
The next step is to perform similar longitudinal phenotypic analyses in smaller comprehensively characterized asthma cohorts that include extended physiologic measures, validated questionnaires, and biologic specimens for omics investigations. The National Heart, Lung, and Blood Institute’s Severe Asthma Research Program and the European U-BIOPRED (Unbiased Biomarkers in Prediction of Respiratory Disease Outcomes) networks are both currently performing comprehensive evaluations in longitudinal asthma protocols. Both networks are recruiting participants with milder and more severe asthma, enriching their cohorts for the latter. As such, both networks will have a broad spectrum of asthma severity in which to investigate disease heterogeneity. These groups should extend what we know about the stability of multivariate asthma phenotypes over time and identify risk factors to predict chutes and ladders in patients with mild, moderate, and severe asthma phenotypes.
Footnotes
Author disclosures are available with the text of this article at www.atsjournals.org.
References
- 1.Haldar P, Pavord ID, Shaw DE, Berry MA, Thomas M, Brightling CE, Wardlaw AJ, Green RH. Cluster analysis and clinical asthma phenotypes. Am J Respir Crit Care Med. 2008;178:218–224. doi: 10.1164/rccm.200711-1754OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Weatherall M, Travers J, Shirtcliffe PM, Marsh SE, Williams MV, Nowitz MR, Aldington S, Beasley R. Distinct clinical phenotypes of airways disease defined by cluster analysis. Eur Respir J. 2009;34:812–818. doi: 10.1183/09031936.00174408. [DOI] [PubMed] [Google Scholar]
- 3.Moore WC, Meyers DA, Wenzel SE, Teague WG, Li H, Li X, D’Agostino R, Jr, Castro M, Curran-Everett D, Fitzpatrick AM, et al. National Heart, Lung, and Blood Institute’s Severe Asthma Research Program. Identification of asthma phenotypes using cluster analysis in the Severe Asthma Research Program. Am J Respir Crit Care Med. 2010;181:315–323. doi: 10.1164/rccm.200906-0896OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Siroux V, Basagaña X, Boudier A, Pin I, Garcia-Aymerich J, Vesin A, Slama R, Jarvis D, Anto JM, Kauffmann F, et al. Identifying adult asthma phenotypes using a clustering approach. Eur Respir J. 2011;38:310–317. doi: 10.1183/09031936.00120810. [DOI] [PubMed] [Google Scholar]
- 5.Sutherland ER, Goleva E, King TS, Lehman E, Stevens AD, Jackson LP, Stream AR, Fahy JV, Leung DY Asthma Clinical Research Network. Cluster analysis of obesity and asthma phenotypes. PLoS ONE. 2012;7:e36631. doi: 10.1371/journal.pone.0036631. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kim T-B, Jang A-S, Kwon H-S, Park J-S, Chang Y-S, Cho S-H, Choi BW, Park J-W, Nam D-H, Yoon H-J, et al. COREA Study Group. Identification of asthma clusters in two independent Korean adult asthma cohorts. Eur Respir J. 2013;41:1308–1314. doi: 10.1183/09031936.00100811. [DOI] [PubMed] [Google Scholar]
- 7.Boudier A, Curjuric I, Basagana X, Hazgui H, Anto JM, Bousquet J, Bridevaux PO, Dupuis-Lozeron E, Garcia-Aymerich J, Heinrich J, et al. Ten-year follow-up of cluster-based asthma phenotypes in adults: a pooled analysis of three cohorts. Am J Respir Crit Care Med. 2013;188:550–560. doi: 10.1164/rccm.201301-0156OC. [DOI] [PubMed] [Google Scholar]
- 8.Amelink M, de Nijs SB, de Groot JC, van Tilburg PMB, van Spiegel PI, Krouwels FH, Lutter R, Zwinderman AH, Weersink EJ, ten Brinke A, et al. Three phenotypes of adult-onset asthma. Allergy. 2013;68:674–680. doi: 10.1111/all.12136. [DOI] [PubMed] [Google Scholar]
- 9.Bai TR, Vonk JM, Postma DS, Boezen HM. Severe exacerbations predict excess lung function decline in asthma. Eur Respir J. 2007;30:452–456. doi: 10.1183/09031936.00165106. [DOI] [PubMed] [Google Scholar]
- 10.O’Byrne PM, Naji N, Gauvreau GM. Severe asthma: future treatments. Clin Exp Allergy. 2012;42:706–711. doi: 10.1111/j.1365-2222.2012.03965.x. [DOI] [PubMed] [Google Scholar]