Objectives: To demonstrate the problem of "confounding by severity" using the example of intravenous (IV) versus oral corticosteroids for the treatment of acute asthma. Double-blind, randomized trials have clearly demonstrated that IV and oral corticosteroids have comparable efficacy.
Methods: Using a standardized protocol, 64 emergency departments enrolled 1,847 patients, aged 18-54 years, with acute asthma. Because route of corticosteroid therapy was not randomized, potential confounders of the association between corticosteroid route and hospital admission were controlled for by multivariate logistic regression and stratification.
Results: Among the 1,193 patients, 383 (32%) received IV corticosteroids and 810 (68%) received oral corticosteroids. The two groups differed markedly at baseline, with patients receiving IV corticosteroids having more severe asthma. Overall, patients receiving IV corticosteroids were more likely to be admitted or experience a relapse event within 48 hours (51% vs. 19%; p < 0.001). On multivariate analysis, patients receiving IV corticosteroids remained more likely to be admitted or experience a relapse event within 48 hours (odds ratio = 2.6; 95% confidence interval = 1.2 to 6.0).
Conclusions: In this observational study, patients with worse asthma exacerbations were more likely to receive IV corticosteroids as compared with oral corticosteroids. Although we controlled for many markers of asthma severity, we were unable to completely control for baseline differences between the IV and oral corticosteroid groups. Observational research continues to serve as an important tool for describing problems and for understanding many exposure-disease associations. For examining the impact of treatments on adverse outcomes, randomized trials are often required to avoid intractable confounding by severity.