In Reply We thank Drs van Dijk and Käyser for their thoughtful comments on our study. They argue that our study design may be prone to residual confounding, despite our use of sophisticated methods to control for this. We agree and said so in our article.1 However, randomized trials are also prone to methodological limitations, particularly when attempting to assess harms of therapy, which observational studies can overcome. Specifically, randomized trials often exclude participants with comorbidities or demographic characteristics that would place them at increased risk of adverse events, even when such characteristics are common among patients who in clinical practice will go on to receive the treatments being evaluated. Also, as serious adverse events are typically infrequent and often unexpected, randomized trials commonly lack sufficient power to identify them adequately.
Stanbrook MB, Campitelli M, Gershon A. Bronchodilator Safety in Chronic Obstructive Pulmonary Disease—Reply. JAMA Intern Med. 2014;174(4):648. doi:10.1001/jamainternmed.2013.12684