B44-23 Recounting the Risk: Serial Eosinophils and Exacerbations in COPD

Abstract Background Type 2 (T2) inflammation, identified by peripheral blood eosinophils, is a recognized therapeutic target in COPD. Pivotal trials (BOREAS, NOTUS, MATINEE) commonly apply ≥ 300 cells/µL to define an eosinophil-high phenotype. However, real-world evidence combining serial eosinophil assessment with explicit asthma exclusion remains limited. Objective To determine whether any-time eosinophilia ≥ 300 cells/µL (vs always 300 cells/µL) is independently associated with recent exacerbation burden. Methods Retrospective, single-center cohort (São Paulo, Brazil; June 2022-June 2025) based on outpatient electronic medical records. Peripheral blood eosinophils were extracted immediately after each dated entry, and patients were classified as any-time ≥300 or always 300. Individuals with any asthma-suggestive features (family/personal history, nocturnal worsening, seasonal pattern) were excluded. Exacerbations were counted only for 2023-2025. We performed (i) a negative binomial (NB) model without covariates to obtain the crude IRR (displayed in the figure), and (ii) a multivariable NB model (age, FEV1% predicted, inhaled corticosteroid ICS use, LAMA/LABA use) to estimate the adjusted IRR (ICS effect not interpreted due to indication bias). Complete-case analysis was used. Results N = 99 patients were included; median of 3 eosinophil measurements per patient (IQR 2-4). Any-time eosinophilia was present in 33/99 (33%). Unadjusted NB: eosinophils ≥300 were associated with higher exacerbation counts, IRR 1.64 (95% CI 1.02-2.63), p = 0.042. Adjusted NB: the association attenuated and was not statistically significant, IRR 1.31 (95% CI 0.87-1.96), p = 0.190. FEV1% predicted remained independently protective, IRR 0.98 per percentage point (95% CI 0.967-0.994), p = 0.005. Conclusions In an asthma-excluded Brazilian cohort with serial eosinophil measurements, a binary, trial-aligned definition of eosinophilia (≥300 cells/µL at any time) showed a significant crude association with exacerbation burden, which attenuated after adjustment; FEV1% was a robust inverse correlate. These data complement BOREAS, NOTUS and MATINEE, underscoring that serial phenotyping adds operational clarity for patient selection, while emphasizing the need for prospective studies with uniform follow-up and richer covariate control (e.g., smoking status, sex, recent systemic corticosteroids) to refine the exacerbation signal in T2-leaning COPD. This abstract is funded by: None