Background: Inhaled corticosteroids are important in the management of chronic obstructive pulmonary disease (COPD), but can slightly increase the risk of pneumonia in patients with moderate-to-severe COPD. Patients with circulating eosinophil counts of 2% or more of blood leucocytes respond better to inhaled corticosteroids than do those with counts of less than 2% and it was therefore postulated that blood eosinophil count might also have an effect on the risk of pneumonia in patients with COPD. In this post-hoc meta-analysis, we investigate whether a 2% threshold can identify patients who differ in their risk of pneumonia, irrespective of inhaled corticosteroid treatment.

Methods: From the GlaxoSmithKline trial registry, we selected randomised, double-blind, clinical trials of patients with COPD that had: inhaled corticosteroid arms (fluticasone propionate and salmeterol or fluticasone furoate and vilanterol); a control arm (not given inhaled fluticasone); and pre-randomisation measurements of blood eosinophil counts and were of at least 24 weeks in duration. With use of specified terms from the Medical Dictionary for Regulatory Activities we identified pneumonia adverse events in patient-level data. We calculated number of patients with pneumonia events, stratified by baseline blood eosinophil count (<2% vs ≥2% of blood leucocytes) and whether or not patients had received inhaled corticosteroids.

Findings: We identified ten trials (conducted between 1998 and 2011), with eosinophil count data available for 10 861 patients with COPD. 4043 patients had baseline blood eosinophil counts of less than 2% and 6818 patients had baseline blood eosinophil counts of 2% or more. 149 (3·7%) patients with counts less than 2% had one or more pneumonia adverse events compared with 215 (3·2%) with counts of 2% or more (hazard ratio [HR] 1·31; 95% CI 1·06-1·62). In patients not treated with inhaled corticosteroids, 40 (3·8%) patients with less than 2% blood eosinophil counts had a pneumonia event versus 48 (2·4%) with 2% or more blood eosinophils (HR 1·53; 95% CI 1·01-2·31). In patients treated with inhaled corticosteroids, events occurred in 107 (4·5%) versus 164 (3·9%; HR 1·25; 95% CI 0·98-1·60), respectively.

Interpretation: Using 2% baseline eosinophil count as a threshold, patients with COPD with lower blood eosinophil counts had more pneumonia events than did those with higher counts. The magnitude of this increased risk was small and should be further explored in large, prospective studies. These data should be considered when making treatment decisions, alongside existing evidence that patients with COPD and baseline blood eosinophil counts less than 2% have a poorer response to inhaled corticosteroids.

Funding: GlaxoSmithKline.