WEUSKOP6069: Exploratory Cluster Analysis: Relovair COPD exacerbation studies - which patients benefit most from ICS?

Trial description: Rationale and background: Chronic obstructive pulmonary disease (COPD) exacerbations are significant events in the natural history of COPD. Identifying which patients may benefit most from inhaled corticosteroids (ICS) is important for optimal disease management.Research question and objectives: To identify patient groups that have the greatest response to ICS in terms of exacerbation reduction when comparing fluticasone furoate/vilanterol (FF/VI) to vilanterol (VI) using data from two completed clinical trials. To validate prior patient subgroups found previously in studies of fluticasone propionate / salmeterol fixed dose combination (FSC) and salmeterol (SAL) [DiSantostefano, 2013]. Study Design: This exploratory retrospective analysis of two clinical trials will be performed using a modified recursive partitioning algorithm, where clusters of patients with differential responses to treatment FF/VI vs. VI will be identified. The recursive partitioning approach will maximize treatment differences (mean exacerbation rates between FF/VI and VI groups) among subgroups. Clusters will be characterized based on the tree and by descriptive statistics.Data source & Setting: HZC102871/HZC102970- Two 52-week efficacy and safety studies that compared the effect of three dosage strengths of FF/VI (50 µg FF/ 25 µg VI, 100 µg FF/ 25 µg VI, and 200 µg FF/ 25 µg VI) inhalation powder with VI (25 µg VI) on the annual rate of exacerbations in subjects with COPD. Doses of FF/VI will be combined within each cluster.Population: The primary analysis population is the Intent-to-Treat (ITT) population, including subjects who had been randomized to and received at least one dose of randomized treatment.Variables: Baseline and screening characteristics from HZC102871/HZC102970. The primary outcome/efficacy variable will be the annual rate of moderate/severe COPD exacerbations as defined in the two clinical trials used as the data source. Exposure will be based on the randomized treatment assignment.Data analysis: The analysis will consist of two approaches. First, we will attempt to validate prior findings using two year-long exacerbation studies of FSC vs. SAL. Using the FF/VI data from the two year-long exacerbation studies, the same negative binomial model and final clusters from the recursive partitioning tree for FSC/SAL [DiSantostefano, 2013] will be evaluated to determine if ≥12% reversibility and/or diuretic use are associated with a reduction in exacerbation for the FF/VI vs. VI treatment group. Briefly, if the resulting rate ratios from these clusters are in the same direction and meet the p-value thresholds specified in the analysis plan, we will conclude that the study results have been mathematically validated.Second, a data-driven modified recursive partitioning technique will also be employed using the primary exacerbation model for the FF/VI vs. VI studies in a similar manner as done previously for FSC vs. SAL [DiSantostefano, 2013]. The negative binomial model of moderate to severe exacerbation will be adjusted for smoking status at screening, geographical region (random effect), percentage predicted FEV1 at randomization, and study 1 vs. study 2 indicator variable, with log time on treatment per patient as an offset [Dransfield, 2013]. It is expected that a different set of patient characteristics and/or different cutpoints could be found than reported previously. Generalized linear models using a negative binomial function will be used to compare the likelihood of having an exacerbation, looking at treatment by subgroup interaction. The spilt of the tree resulting from recursive partitioning will be determined by maximizing subgroups by treatment interaction effect (mean exacerbation rates between FF/VI vs. VI groups) among subgroups. Rate ratios and 95% confidence intervals will be used to estimate the differences in annual mean exacerbations rates for each cluster. Subgroup membership will then be assigned to each patient based on the selected tree.Ad-hoc pruning may be applied to the split tree that optimized the modified recursive partitioning algorithm, to ensure that subgroups chosen are consistent with a clinical interpretation. Resources: (1) DiSantostefano RL, Li H, Rubin DB, Stempel DA. Which patients with chronic obstructive pulmonary disease benefit from the addition of an inhaled corticosteroid to their bronchodilator? A cluster analysis. BMJ Open. 2013 Apr 22;3(4).(2) Dransfield MT, Bourbeau J, Jones PW, Hanania NA, Mahler DA, Vestbo J, Wachtel A, Martinez FJ, Barnhart F, Sanford L, Lettis S, Crim C, Calverley PM. Once-daily inhaled fluticasone furoate and vilanterol versus vilanterol only for prevention of exacerbations of COPD: two replicate double-blind, parallel-group, randomized controlled trials. Lancet Respir Med. 2013 May;1(3):210-23. Erratum in: Lancet Respir Med. 2013 May;1(3):186.RELVAR is a trademark of the GSK group of companies.