Aspirin Use and Respiratory Morbidity in COPD A Propensity Score-Matched Analysis in Subpopulations and Intermediate Outcome Measures in COPD Study

BACKGROUND: Aspirin use in COPD has been associated with reduced all-cause mortality in meta-regression analysis with few equivocal studies. However, the effect of aspirin on COPD morbidity is unknown. METHODS: Self-reported daily aspirin use was obtained at baseline from SPIROMICS participants with COPD (FEV1/FVC < 70%). Acute exacerbations of COPD (AECOPD) were prospectively ascertained through quarterly structured telephone questionnaires up to 3 years and categorized as moderate (symptoms treated with antibiotics or oral corticosteroids) or severe (requiring ED visit or hospitalization). Aspirin users were matched one-to-one with nonusers, based on propensity score. The association of aspirin use with total, moderate, and severe AECOPD was investigated using zero-inflated negative binomial models. Linear or logistic regression was used to investigate the association with baseline respiratory symptoms, quality of life, and exercise tolerance. RESULT Among 1,698 participants, 45% reported daily aspirin use at baseline. Propensity score matching resulted in 503 participant pairs. Aspirin users had a lower incidence rate of total AECOPD (adjusted incidence rate ratio [IRR], 0.78; 95% CI, 0.65-0.94), with similar effect for moderate but not severe AECOPD (IRR, 0.86; 95% CI, 0.63-1.18). Aspirin use was associated with lower total St. George's Respiratory Questionnaire score (beta, -2.2; 95% CI, -4.1 to -0.4), reduced odds of moderate-severe dyspnea (modified Medical Research Council questionnaire score >= 2; adjusted odds ratio, 0.69; 95% CI, 0.51-0.93), and COPD Assessment Test score (beta, -1.1; 95% CI, -1.9 to -0.2) but not 6-min walk distance (beta, 0.7 m; 95% CI, -14.3 to 15.6). CONCLUSION: Daily aspirin use is associated with reduced rate of COPD exacerbations, less dyspnea, and better quality of life. Randomized clinical trials of aspirin use in COPD are warranted to account for unmeasured and residual confounding.