Managing blood glucose in critically ill patients with or without diabetes

Relationships between blood glucose concentration and outcome and also the optimum management of blood glucose remain highly contentious issues for critical care practitioners. Among the many controversies is whether critically ill patients with diabetes should be treated differently from those without diabetes. Krinsley and colleagues assembled a large observational database that sheds further light on the relationships of hyperglycemia, hypoglycemia and glucose variability and risk of death in critically ill patients with and without diabetes. Defining the optimum treatment and whether this should differ by diabetic status requires high-quality primary interventional research.

Th e best management of stress-induced hyperglycemia in critically ill patients remains unknown and among the many controversies is whether patients with diabetes should be managed diff erently from those without diabetes. In a recent edition of Critical Care, Krinsley and colleagues report on associations between the 'three domains of glycemic control' (hyperglycemia, hypoglycemia and glucose variability) in patients with or without a diagnosis of diabetes [1]. Th ey assembled and analyzed a large database containing data from patients admitted to 23 ICUs in nine countries. Th eir purpose was to determine how diabetic status aff ected the relationships between glycemic control and outcome.
Krinsley and colleagues report some interesting fi ndings. For critically ill patients in general, the relationship between mean blood glucose concentration [BG] and outcome is known to be U-shaped, with increased mortality observed with both high and low mean [BG] [2]. When examining those patients with and without diabetes separately, Krinsley and colleagues report that the nadir of the U-shaped curve (the [BG] associated with lowest mortality) occurs at a diff erent [BG] in those with diabetes. In patients with diabetes, an associated increase in mortality does not occur until a higher [BG]. Moderate and severe hypoglycemia (defi ned as minimum [BG] 40 to 69 mg/dl (2.2 to 3.8 mmol/l) and <40 mg/dl (<2.2 mmol/l), respectively) were associated with increased mortality regardless of diabetic status, whereas increased [BG] variability defi ned by the coeffi cient of variation was only associated with increased risk of death in patients without diabetes.
How should we interpret these fi ndings? Do they contradict or support what we already know about patients with diabetes admitted to ICUs, and should these results infl uence how we treat patients with diabetes?
Firstly, we should recognize both the strengths and the weaknesses of the study. Th e strength of the study is mainly its size and that the authors drew data from diff erent ICUs in diff erent countries. Th e weaknesses are the observa tional nature of the study, which limits the inferences that can be drawn, and the lack of data on the actual management of [BG]. Additionally, the authors did not have access to measures of glycemic control prior to admission to the ICU and so their presumption that patients with diabetes may behave diff erently due to adaptive mechanisms developed owing to chronic hyperglycemia is based on supposition rather than on objective measurements such as glycosylated hemoglobin.
What do we already know about patients with diabetes treated in ICUs? Patients with known diabetes have typically constituted around 15 to 20% of patients enrolled in trials of intensive versus conventional glucose control in the ICU [3,4], and perhaps around 19% of patients admitted to ICUs overall [5]. With the notable exception of patients undergoing cardiac surgery, a diagnosis of diabetes is not associated with an increased risk of death following admission to the ICU [5]. Th e reasons for this somewhat surprising fi nding have been

Abstract
Relationships between blood glucose concentration and outcome and also the optimum management of blood glucose remain highly contentious issues for critical care practitioners. Among the many controversies is whether critically ill patients with diabetes should be treated diff erently from those without diabetes. Krinsley and colleagues assembled a large observational database that sheds further light on the relationships of hyperglycemia, hypoglycemia and glucose variability and risk of death in critically ill patients with and without diabetes. Defi ning the optimum treatment and whether this should diff er by diabetic status requires high-quality primary interventional research. discussed, including that patients with diabetes may be receiving medications that increase their chances of surviving critical illness [6]. An alternative explanation is that the apparent protective eff ect of diabetes is due to other confounding infl uences -for example, less sick patients with diabetes being admitted to the ICU because they are incorrectly perceived to be at increased risk of death. Data from randomized trials also confi rm that patients with diabetes do not benefi t from being treated with insulin with the goal of achieving normoglycemia [3,4], although whether they diff er from patients without diabetes in this regard remains unclear and cannot be determined using observational data.
While researchers have studied the eff ects of diff erent blood glucose targets on the long-term outcomes of ambulant patients with diabetes [7], there has been no study specifi cally examining the eff ects of diff erent blood glucose targets in critically ill patients with diabetes. As the optimum [BG] target for critically ill patients overall is not known, it would be vastly premature to call for future studies to be limited to patients with diabetes. Th is question can be addressed in an appropriately designed study recruiting ICU all-comers as long as patients with type I and type II diabetes are identifi ed at baseline. Stratifying randomization by a diagnosis of diabetes would further increase the robustness of the results.
In conclusion, Krinsley and colleagues report that the associations between [BG] and outcome are diff erent in patients with and without diabetes although appropriate targets for management of blood glucose concentration can only be defi ned by future randomized trials. Such trials can and should be designed to determine whether treatment eff ects diff er in patients who do or do not have diabetes at the time of enrolment.