In Reply: We reported “lack of evidence for a strong association between interferon beta treatment and disability progression. . . . ” We also raised the possibility that “a subgroup of patients [may] benefit from interferon beta treatment. . . . ”
Dr Goodin and colleagues elaborate on the issue of potential residual confounding by indication, also discussed at length in our article. While we agree that the probability of changing treatment status may not depend just on patients' characteristics at eligibility for interferon beta treatment (ie, the study baseline), showing that this results in substantial bias (for either the historical or contemporary control groups), or attempting to adjust for this possibility, is challenging, with few well-validated methods available. Ideally, every time a patient's treatment status changes, other important changes in his or her profile would be accounted for. Gaining this level of dynamic detail may not be possible. We are currently investigating several possible approaches, including marginal structural models.1 However, accurately estimating the various weights required by such models is challenging, and careful validation in MS is needed.