We would like to thank Gaster and Hirsch for their comments regarding our article and the accompanying editorial by Sawin. We would like to address the specific issues they raised.
First, we agree that observational studies have an inherent bias when evaluating treatments. Unlike a randomized controlled trial, treatment groups in an observational study are likely to differ in other ways besides treatment itself. Therefore, in observational studies, one must make adjustments for group differences using statistical techniques. The groups may differ in respects that cannot be identified by available data. We were able to adjust for a wide range of potentially confounding factors, including underlying comorbidity, glucose levels on admission, severity of diabetes complications, and severity of illness. Despite adjusting for these variables there was still an impressive 3-fold increased risk of hyperglycemia with the use of sliding scale insulin alone.
Queale W, Brancati FL. Sliding Scale Insulin Use and Rates of Hyperglycemia—Reply. Arch Intern Med. 1998;158(1):95. doi: