Two studies recently published in Clinical Chemistry explore novel computer simulation studies of glycemic control in patients hospitalized in intensive care units (ICUs). The studies take “alternative and differing modeling approaches on the basis of real patient data,” explain James C. Boyd, MD, and David E. Bruns, MD, in an accompanying editorial. Boyd is a deputy editor of Clinical Chemistry; Bruns served as the journal’s editor from 1990 to 2007.

“These studies may help to answer questions regarding what analytical quality is required of glucose assays, and what impact more frequent measurement has on patient outcomes and/or the analytical performance requirements of glucose measurements,” wrote Boyd and Bruns. “Although earlier modeling studies of these questions have used established physiologic models of glucose homeostasis, what makes the current modeling studies unique and so important is that the models in these studies are based on data from actual [intensive care unit] ICU patients.”

One study was done using 56 virtual patients who were modeled individually “by fitting the data from an actual patient episode in the ICU to a complex, physiologically based compartmental model that uses differential equations,” explained Boyd and Bruns. The second study used a different approach, instead recording interventions taken in ICU patients based on glucose measurements done with blood gas analyzers. These investigators then compared that data with measurements taken with simulated glucose sensors that were less accurate.

Each study used different outcome measures. The one involving virtual patients used a traditional approach, basing outcomes on the rates of hypoglycemia, hyperglycemia, time range, and glycemic variability. The second study’s outcomes were based on the probabilities of an error occurring in each category. “Despite these differences in outcome measures monitored, results of both studies suggested that CGM [continuous glucose monitoring] glucose sensors with mean absolute relative difference (MARD) scores <11% gave the best results and lowest frequencies of hypoglycemia,” wrote Boyd and Bruns. MARD scores are calculated based on the “mean percentage difference of the absolute concentration measured by the evaluated glucose sensor from a reference glucose concentration,” they explained.

Regardless of approach, the “models agree that control of glycemia is intimately related to the analytical performance of the glucose measurement system, the frequency of the measurements, and the protocol by which a measured glucose is translated into an intervention,” according to the editorial. The study authors should be “congratulated for the valuable contributions their modeling studies have made in suggesting potential new criteria for CGM performance that not only are in agreement, but are based on real patient data,” concluded Boyd and Bruns.