Abstract

Peter J. Pronovost, MD, PhD*† Perioperative glycemic control is related to patient outcome. Although guidelines for glucose management in hospitalized patients have undergone dramatic changes over the last 5 years, most hospital-based physicians, including anesthesiologists, have not changed their approach to glucose management (1). In this editorial, we discuss the article by Duncan et al. (2) appearing in this issue of Anesthesia & Analgesia, review recent recommendations for perioperative glucose management, and highlight areas of uncertainty. We endeavor to improve perioperative glucose management for patients. There is considerable uncertainty regarding how to manage surgical patients who are taking metformin, an oral hypoglycemic drug. The potential for postoperative lactic acidosis in patients taking this drug has prompted some clinicians and health care systems to routinely cancel surgical procedures if metformin is taken within 48 h of surgery. Other clinicians or health care systems continue metformin, both before and after the surgical procedure. Duncan et al. (2) conducted a retrospective review of a large cohort of diabetic patients admitted for cardiac surgery at their institution who recently took an oral hypoglycemic drug. They compared outcomes among patients who took metformin versus patients taking a non-metformin oral hypoglycemic drug. The authors found that patients taking metformin had lower risks for several complications, and concluded that metformin appeared to be safe for use in the perioperative period. The article raises several points worthy of reflection. First, the authors used a propensity score to try to account for differences among patients who were taking, versus not taking, metformin. This statistical method used logistic regression, in which metformin becomes the dependent variable. Predictor variables are included in the model (in this case 54 variables) to identify those patient characteristics associated with metformin treatment. The output of this analysis includes a C statistic that is interpreted to mean that, for any given pair of patients, in which one patient received metformin and one did not, how often did the model identify the one receiving metformin? In this analysis, the C statistic was 0.68, suggesting that more than 3 of 10 patients were not correctly classified. Thus, there is the potential that unmeasured differences between groups, and not metformin treatment or nontreatment, could influence the results. Other evidence suggests that propensity scores are no better than standard regressions at controlling for selection bias (3). Second, the authors evaluated a variety of outcomes, though many were uncommon and did not include explicit definitions, and concluded that metformin was safe. It would be helpful to consider the precise safety of these estimates (3). For example, even if none of the 523 metformin patients had an adverse event in this study, the upper limit of that confidence interval is seven events in 1000 patients. This rate would likely warrant concern from clinicians and lead to alterations of their practice. Rather than thinking of drug safety as a dichotomous variable, (i.e., safe versus unsafe) it may be helpful to think of safety as a continuous variable to help regulators, clinicians, and consumers make a more informed risk/benefit assessment. From the Departments of *Anesthesiology and Critical Care Medicine and †Surgery, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland. Accepted for publication October 26, 2006. Conflicts of Interest: Dr. Martinez has agreed to be a consultant for Bay City Capital LLC, but has not received any compensation as of this submission. The authors have no other potential conflicts of interest to disclose. Reprints will not be available from the authors. Address correspondence to Elizabeth A. Martinez, MD, MHS, ACCM/Adult Critical Care Division, 600 N. Wolfe St., Meyer 296, Baltimore, MD 21287. Address e-mail to emartine@jhmi.edu. Copyright © 2006 International Anesthesia Research Society DOI: 10.1213/01.ane.0000252348.81206.7f

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call