IMPACT, a 52-week, randomised, double-blind trial, assessed the efficacy and safety of fluticasone furoate/umeclidinium/vilanterol (FF/UMEC/VI) triple therapy versus FF/VI or UMEC/VI in patients with symptomatic COPD and a history of exacerbations. Subgroup analyses assessed whether the efficacy of FF/UMEC/VI versus FF/VI or UMEC/VI and UMEC/VI versus FF/VI varies according to prior exacerbation history, and the combined effects of exacerbation history and blood eosinophil counts. Three subgroups were defined: single moderate (1 moderate/no severe; n=3056 (30%)), frequent moderate (≧2 moderate/no severe; n=4628 (45%)) and severe (≧1 severe/any moderate; n=2671 (26%)). End-points included annual on-treatment moderate/severe exacerbation rate (pre-specified), lung function and health status (both post-hoc). Moderate/severe exacerbation rates (reduction % (95% CI)) were reduced in the FF/UMEC/VI group versus FF/VI (single moderate 20% (10–29), frequent moderate 11% (2–19), severe 17% (7–26)) and versus UMEC/VI (single moderate 18% (5–29), frequent moderate 29% (21–37), severe 26% (14–35)). Moderate/severe exacerbation rates were reduced in the FF/VI group versus UMEC/VI in the frequent moderate subgroup; a numerical reduction was observed in the severe subgroup (single moderate 2% (-12–18), frequent moderate 21% (11–29), severe 11% (-3–22)). Moderate/severe exacerbation rates were lower in the FF/VI group compared with UMEC/VI in patients with higher eosinophil counts. FF/UMEC/VI improved lung function and health status versus both dual therapies irrespective of exacerbation subgroup. UMEC/VI improved lung function versus FF/VI in all subgroups. Triple therapy was more effective than dual regardless of exacerbation history, consistent with results in the intent-to-treat population. Comparisons between dual therapies were influenced by prior exacerbation history and eosinophil counts.