Background: The peripheral blood eosinophil count might help identify those patients with chronic obstructive pulmonary disease (COPD) who will experience fewer exacerbations when taking inhaled corticosteroids (ICS). Previous post-hoc analyses have proposed eosinophil cutoffs that are both arbitrary and limited in evaluating complex interactions of treatment response. We modelled eosinophil count as a continuous variable to determine the characteristics that determine both exacerbation risk and clinical response to ICS in patients with COPD.

Methods: We analysed data from three AstraZeneca randomised controlled trials of budesonide-formoterol in patients with COPD with a history of exacerbations and available blood eosinophil counts. Patients with any history of asthma were excluded. Negative binomial regression analysis was done using splines for modelling of continuous variables to study the primary outcome of annual exacerbation rate adjusted for exposure time and study design. The trials are registered with ClinicalTrials.gov, NCT00206167, NCT00206154, and NCT00419744.

Findings: 4528 patients were studied. A non-linear increase in exacerbations occurred with increasing eosinophil count in patients who received formoterol alone. At eosinophil counts of 0·10 × 10⁹ cells per L or more, a significant treatment effect was recorded for exacerbation reduction with budesonide-formoterol compared with formoterol alone (rate ratio 0·75, 95% CI 0·57-0·99; p_interaction=0·015). Interactions were observed between eosinophil count and the treatment effects of budesonide-formoterol over formoterol on St George's Respiratory Questionnaire (p_interaction=0·0043) and pre-bronchodilator FEV₁ (linear effect p<0·0001, p_interaction=0·067). Only eosinophil count and smoking history were independent predictors of response to budesonide-formoterol in reducing exacerbations (eosinophil count, p_interaction=0·013; smoking history, p_interaction=0·015).

Interpretation: In patients with COPD treated with formoterol, blood eosinophil count predicts exacerbation risk and the clinical response to ICS.

Funding: AstraZeneca.