Paging goldilocks: How much glycemic control is just right?

T here is no doubt that hyperglycemia among hospitalized patients correlates with worse prognosis. Further, there are well-documented mechanisms by which poor glycemic control may directly impact outcomes. For example, hyperglycemia and insulin deficiency can impair neutrophil function, exacerbate inflammation, and impair endothelium-mediated dilatation, whereas hypoglycemia increases sympathetic tone. And both severe hyperglycemia and hypoglycemia, of course, can precipitate altered mental status. But certainly not all of the morbid outcomes associated with poor glycemic control in the hospital— including infection, cardiac events and death—are caused by poor glycemic control in the hospital. Elevated glucose levels in the hospital are often seen in sicker patients with raging stress hormones and in brittle diabetics with a present-on-admission condition that has been ravaging their vasculature for years. This means that virtually all observational studies demonstrating worse outcomes in the setting of poor glucose control in the hospital will be severely confounded by comorbid illness, and much confounding will remain even after multivariate adjustment. Nonetheless, high-quality randomized controlled trials that have focused on critically ill patients, rather than general medical patients, have generated intense interest and fostered the belief that controlling the glucose level of all hospitalized patients is probably a good idea. (Although, more recently, even the data supporting glycemic control in the critically ill have been challenged.) Enthusiasm for implementing aggressive glycemic control protocols outside of the intensive care unit (ICU) appears widespread, as is evident in this issue of JHM. In this issue, two articles detail the challenges of implementing glycemia control protocols. The research teams employed different protocols and used different metrics, but there are common themes: (1) The process was iterative. Interventions were piloted, then rolled out, and substantial effort was needed to foster continued attention to the interventions. (2) The process was multidisciplinary. Buy-in and input were needed not only from physicians, but also from nurses, pharmacists, dieticians, clinical data system experts, and probably patients. (3) Impacting process measures was easier than impacting surrogate outcome measures. Specifically, despite dramatic changes in the use of carefully vetted order sets and protocols, the impact on glycemia was modest and sometimes inconsistent. These studies illustrate that implementing protocols to control glycemia is neither easy, nor consistently associated with improved glycemic control—let alone improved major clinical outcomes. Three complementary observational studies further illustrate how hard it is to optimize glycemic control in the hospital setting. Together, the observational and interventional