In Reply Dr Roumie and colleagues have clarified the interpretation of their analysis of unmeasured confounder imbalance in their study.1 Because their analysis reported absolute differences, as they point out in their letter, the prevalence of unmeasured confounders would need to be quite high in the insulin-treated group with little or no prevalence in the sulfonylurea-treated group to explain the findings of their study.
Safford MM. Insulin vs Sulfonylureas for Second-Line Diabetes Treatment—Reply. JAMA. 2014;312(16):1695-1696. doi:10.1001/jama.2014.11129