Open Access Research BMJ Open: first published as 10.1136/bmjopen-2012-001838 on 22 April 2013. Downloaded from Which patients with chronic obstructive pulmonary disease benefit from the addition of an inhaled corticosteroid to their bronchodilator? A cluster analysis Rachael L DiSantostefano,1 Hao Li,1 David B Rubin,2 David A Stempel2 To cite: DiSantostefano RL, ABSTRACT et al ARTICLE SUMMARY Li H, Rubin DB, . Which Objective: To identify subsets of chronic obstructive patients with chronic pulmonary disease (COPD) patients who are more obstructive pulmonary Article focus protected from exacerbations with the use of an disease benefit from the ▪ This paper describes a cluster analysis of a β addition of an inhaled inhaled corticosteroid/long-acting 2 agonist (ICS/ pooled cohort of chronic obstructive pulmonary corticosteroid to their LABA) combination, compared with the use of LABA disease (COPD) patients receiving salmeterol bronchodilator? A cluster monotherapy. (SAL) alone or in combination with fluticasone analysis. BMJ Open 2013;3: Design: Post hoc cluster analysis of patients from two propionate (SFC) for 1 year. The analysis sought e001838. doi:10.1136/ randomised clinical trials of salmeterol/fluticasone to identify clusters of patients who could benefit bmjopen-2012-001838 propionate (SFC) and salmeterol (SAL) that had primary most from the addition of fluticasone propionate endpoints of moderate/severe exacerbation rates. to their long-acting bronchodilator therapy based ▸ Prepublication history and Setting: Centres in North America. on the annual rates of moderate/severe additional material for this Participants: 1543 COPD patients were studied. exacerbations. paper are available online. To Interventions: SFC 50/250 µg or SAL 50 µg, twice daily. view these files please visit Key message the journal online Primary and secondary outcome measures: The ▪ Three clusters were identified. Two clusters, (http://dx.doi.org/10.1136/ analysis identified clusters of COPD patients more including patients receiving diuretics and those bmjopen-2012-001838). responsive to SFC versus SAL with respect to the annual not receiving diuretics but with baseline bron- http://bmjopen.bmj.com/ rate of moderate/severe exacerbations and compared their chodilator reversibility of ≥12%, exhibited a sig- Received 19 July 2012 baseline clinical characteristics. nificantly greater reduction in exacerbations Revised 8 March 2013 Results: Overall, SFC significantly reduced the annual rate Accepted 19 March 2013 when treated with SFC versus SAL. No difference of moderate/severe exacerbations as compared with SAL was seen between treatments in the third patient alone (rate ratio (RR)=0.701, p<0.001). Three-patient This final article is available cluster—persons without bronchodilator revers- for use under the terms of clusters were identified: COPD patients receiving diuretics ibility and not receiving diuretics. These analyses the Creative Commons (RR=0.56, p<0.001); patients not receiving diuretics but highlight two strata of COPD patients who may Attribution Non-Commercial with forced expiratory volume in 1 s (FEV1) reversibility be more likely to benefit from inhaled cortico- 2.0 Licence; see ≥12% (RR=0.67, p<0.001) exhibited a substantial on September 30, 2021 by guest. Protected copyright. steroid therapy combined with a long-acting β2 http://bmjopen.bmj.com reduction in the annual rate of moderate/severe agonist bronchodilator. exacerbations relative to SAL. A third cluster, consisting of patients not receiving diuretics and without FEV1 Strengths and limitations of this study reversibility, demonstrated no difference for SFC versus ▪ Pooled and systematically collected data from SAL. Patients receiving diuretics had a significantly higher >1500 well-characterised patients from two ran- prevalence of comorbid cardiovascular disease. domised controlled trials were used in the ana- Conclusions: COPD patients receiving diuretics and lysis, which was validated using half of the study those not receiving diuretics but with FEV1 reversibility population. The conclusions are limited by the 1GlaxoSmithKline, Research >12% at baseline were significantly more likely to uncertainty of extrapolating results derived from Triangle Park, Durham, experience a reduction in COPD-associated exacerbations participants enrolled in a randomised clinical trial North Carolina, USA with SFC versus SAL alone. in which exacerbation in the previous year was 2Respiratory Clinical Trial registration: NCT00115492, NCT00144911 an entry requirement for COPD patients in the Development, general population. GlaxoSmithKline, Research Triangle Park, Durham, North Carolina, USA fi in 1 s (FEV1) alone is an insuf cient measure INTRODUCTION of disease severity.1 Importantly, the revised Correspondence to Rachael L DiSantostefano; The Global Strategy for the Diagnosis, Management GOLD strategy document also recommends rachael.l.disantostefano@gsk. and Prevention of COPD (GOLD) was revised in therapy with an inhaled corticosteroid/long- fl β com 2011 to re ect that Forced Expiratory Volume acting 2 agonist (ICS/LABA) combination, or DiSantostefano RL, Li H, Rubin DB, et al. BMJ Open 2013;3:e001838. doi:10.1136/bmjopen-2012-001838 1 Cluster analysis of response to ICS/LABA therapy BMJ Open: first published as 10.1136/bmjopen-2012-001838 on 22 April 2013. Downloaded from a long-acting muscarinic antagonist, for patients at risk of least one moderate or severe COPD exacerbation in the two or more exacerbations per year, even in the presence year prior to screening. A moderate exacerbation was of mild airflow limitation. This recommendation reflects defined as one requiring outpatient antibiotic and/or oral the established associations between frequent exacerba- corticosteroid use, and a severe exacerbation was defined as tions, more rapid decline in lung function2 and greater one requiring hospitalisation. Current and former smokers impairment of health status.34 were included. The key exclusion criteria were a current Chronic obstructive pulmonary disease (COPD) is a diagnosis of asthma based on the American Thoracic complex and heterogeneous disease with pulmonary and Society standards for diagnosis,16 other active chronic extra-pulmonary manifestations.5 Significant inroads have respiratory disorders apart from COPD, a moderate/severe recently been made in understanding clinical subtypes and exacerbation that had not resolved prior to visit 1, or con- their pathophysiology,6 and how they may contribute to the current use of anticholinergics, theophyllines and leuko- development of a customised approach to therapeutic inter- triene modifiers, or a history or current significant health vention based on the patient’s individual COPD pheno- conditions that could affect subject safety or effectiveness type.7 Han et al7 have advocated the following process for evaluation if the condition exacerbates during the study. selection of a COPD phenotype: identify a candidate Participants with a history of or current clinically significant phenotype, determine its relevance to clinical outcomes cardiac arrhythmias, uncontrolled/unstable congestive and then validate it with longitudinal data collection in care- heart failure, uncontrolled hypertension or unstable angina fully characterised patient groups. An example of such a were excluded from the study. Participants (n=36) with phenotype established through this process is that of the protocol violations or missing data required for the primary ‘frequent exacerbator’ identified in the ECLIPSE cohort. In model were excluded (n=1543 analysed vs n=1579). that analysis, the presence of two or more exacerbations in the previous year was shown to strongly predict the occur- Cluster analysis methodology renceofanexacerbationinthecomingyear.8 Theannualmoderate/severeexacerbationratewasentered Statistical techniques may assist in the identification of into a cluster analysis using an interaction tree algorithm to COPD phenotypes, with cluster analysis being the most maximise the identification of subgroups showing differ- – commonly used approach.9 12 Cluster analysis uses algo- ences in their response to SFC and SAL treatment.13 rithms to group a patient population, without an a The cluster analysis aimed to find subgroups in the priori hypothesis, into cohorts where those in the same study subjects that had similar baseline characteristics group are more similar to each other than they are to and with maximum treatment differences for mean those in other groups. This is in contrast to subgroup yearly moderate/severe exacerbation rate ratio (RR). analysis, where populations are predefined and statistical Subjects included in the cluster analysis were required 13 http://bmjopen.bmj.com/ testing is applied to identify differences. to have the following baseline variables: FEV1%pre- ≥ In the present study, cluster analysis was conducted using dicted, FEV1 reversibility stratum (yes/no for 12% data pooled from two clinical trials that studied differences improvement and ≥200 ml), time on treatment and geo- in exacerbation rates in COPD patients randomly assigned graphical region. Reversibility following the administra- either to LABA (salmeterol (SAL)) or to ICS/LABA (sal- tion of four puffs of albuterol was determined prior to meterol/fluticasone propionate (SFC)).14 15 The objective randomisation to treatment, following completion of the of this cluster analysis was to identify patients who benefit 4-week FSC 250/50 run-in period. Missing values for the most from the