Introduction and objectives Blood eosinophils are a potentially useful biomarker to guide choice of maintenance treatment in COPD. We aimed to explore rates of blood eosinophil testing, blood eosinophil counts, and their relationship to patient characteristics and disease outcomes in routine primary care.

Methods We used routinely collected data from UK primary care, in the Clinical Practice Research Datalink, linked with Hospital Episode Statistics. Eligible patients were ≥40 years with COPD, history of smoking and diagnostic spirometry, not previously treated with inhaled corticosteroids, starting a new inhaled maintenance medication between 2005 and 2015. We explored the relationship between the most recent blood eosinophil count in the two years before index date, divided into high (≥150/µL) and low (<150/µL) groups, with patient and disease characteristics; Cox regression compared eosinophil groups for time-to-first-exacerbation after maintenance treatment initiation.

Results 30 384 patients fulfilled inclusion criteria, of whom 18 462 (60.8%) had a valid eosinophil count. Males, current smokers, those with asthma, those with more severe COPD and with more frequent exacerbations were tested less often. In those tested, median eosinophil count was 200 cells/µL (interquartile range 100–300/µL), with 31.8% in the low and 68.2% in the high eosinophil group. Higher eosinophil counts were significantly associated with male sex, younger age, ex-smokers, lower severity classification, atopy, asthma and higher baseline exacerbations. Median time-to-first-exacerbation was 524 (95%CI 510–540) days, with no difference between the eosinophil count groups (adjusted hazard ratio 0.98 (0.93–1.02, p=0.26)), and this did not change when eosinophil threshold was increased to ≥300/µL. A high most recent eosinophil count was 98.4% predictive of a high mean eosinophil count over the previous two years.

Conclusions In a primary care population commencing a new maintenance treatment, recent blood eosinophil count is not a marker of prognosis. However, those with eosinophils tested found in routinely collected data may not be representative of the whole COPD population. Most recent eosinophil count may be a suitable surrogate for multiple values over time.