Chapter 32. Glucose Management in Critical Care

The American Diabetes Association (ADA) defines inpatient hyperglycemia as a fasting blood glucose (BG) >126 mg/dL or a random BG >200 mg/dL that reverts to normal after discharge.1 The prevalence of hyperglycemia in the acutely ill patient in the intensive care unit (ICU) has been shown to be as high as 83%.2 Hyperglycemia in critical illness may occur due to stress-related surges in counterregulatory hormones, preexisting diabetes, impaired glucose tolerance, and insulin resistance. Whether it is a condition necessitating intervention or a marker of disease severity, hyperglycemia has been shown to be an independent risk factor for increased mortality in the ICU.3 Despite this association, tight glycemic control (TGC) has not been shown to consistently improve patient outcomes and surprisingly may, in some subgroups, cause more harm than good. This chapter examines the historical background, essential pathophysiology, associations, key clinical studies, current protocols, and recommendations regarding hyperglycemia in the critically ill.

Hyperglycemia was first detected as glucosuria in ether-anesthetized patients 150 years ago, and in 1877, Bernard described hyperglycemia in a canine model of hemorrhagic shock.4 For many years, hyperglycemia in the critically ill was considered an adaptation to stress and was not treated. In fact, some early ICU practitioners recognized insulin resistance and believed that elevated glucose levels (160–200 mg/dL) would promote cellular glucose uptake. In 2001, Van den Berghe demonstrated a statistically significant mortality benefit with TGC in surgical ICU patients. Subsequently, many professional societies including the Surviving Sepsis Campaign (SSC) in 2004 endorsed TGC.5 The Leuven (Van den Berghe et al) medical trial in 2006, Efficacy of Volume Substitution and Insulin Therapy in Severe Sepsis (VISEP, Brunkhorst et al) trial in 2008, and Normoglycemia in Intensive Care Evaluation—Survival Using Glucose Algorithm Regulation (NICE-SUGAR, Finfer et al) and Glucontrol (Preiser et al) trials published in 2009 have contributed most to the continuously evolving issue of glucose management in the critically ill patient.

Risk factors for the development of hyperglycemia include preexisting diabetes mellitus, advanced age, infusion of catecholamine pressors, glucocorticoids, obesity, excessive dextrose resuscitation, sepsis, hypothermia, hypoxia, uremia, and cirrhosis.6 These proven risk factors highlight the multifactorial pathophysiologic mechanisms underlying ICU hyperglycemia.

In the critically ill patient, hyperglycemia can be explained by increased glucose production (glycogenolysis and gluconeogenesis) and decreased peripheral uptake (insulin resistance) (Figure 32-1):

• Increased glucose production: Counterregulatory hormones and catecholamines such as glucagon, growth hormone, cortisol, and epinephrine increase adipose tissue lipolysis and skeletal muscle proteolysis. The end products from this process (glycerol, alanine, and lactate) then fuel hepatic gluconeogenesis. By directly enhancing hepatic glycogenolysis, the above hormones simultaneously further raise glucose levels. Impairment of cellular glycogen synthesis is another important pathway leading to increased glucose levels.
• Decreased peripheral uptake: In a healthy subject, insulin binds to its receptor triggering a signaling pathway that ultimately leads to the translocation of the intracellular Glut4 protein to the plasma membrane where it is responsible for glucose uptake. Although not well understood, it has been postulated that critical illness inhibits Glut4 translocation resulting in hyperglycemia. Counterregulatory hormones and cytokines are believed to play an important role in this process.

Insulin resistance, defined as ongoing gluconeogenesis, glycogenolysis, lipolysis, and proteolysis despite normal or elevated insulin levels, is directly or indirectly (via counterregulatory hormones) modulated by proinflammatory cytokines such as tumor necrosis factor (TNF)-a, interleukin (IL)-1, and IL-6.

Even prior to the seminal randomized control trial (RCT) by Van den Berghe et al in 2001 that showed a statistically significant 30% excess mortality, there have been numerous retrospective studies showing a strong association between poor ICU-related outcomes in hyperglycemic patients.7 For instance, Sung et al demonstrated that admission hyperglycemia in trauma patients was an independent risk factor for increased mortality, ICU length of stay, and infection.8 In traumatic brain injury patients, Young et al showed significantly worse 3-month and 1-year outcomes if the BG levels were >200 mg/dL.9 In patients with ischemic and hemorrhagic stroke, Weir et al demonstrated that a plasma glucose concentration above 144 mg/dL predicted poorer chances of survival and functional independence, even after adjusting for age and stroke severity.10

Similar results have been found in retrospective studies, which focused on a heterogeneous population of ICU patients. Most notably, Krinsley studied 1,826 consecutive ICU patients (roughly 80% medical and 20% surgical) and found that hospital mortality increased progressively as glucose values increased, reaching 43% among patients with mean glucose values exceeding 300 mg/dL.3

Among surgical patients, one of the key mechanisms by which insulin may improve outcomes is through reduction in infections. Studies supporting such a connection have demonstrated a 3-fold increased risk of postoperative wound infections and a 4-fold increased incidence of intravascular infections in hyperglycemic surgical ICU patients.5,6

Initial enthusiasm for intensive insulin therapy (IIT) stimulated by the initial Leuven surgical trial was dampened by the subsequent four RCTs that not only failed to show clear mortality benefits but also highlighted the possible dangerous consequences of IIT.

Subgroup analysis of the initial Leuven surgical trial showed that the greatest mortality benefit was achieved in patients who had a prolonged ICU stay of >5 days.11 Consequently, the Leuven medical trial specifically targeted patients with prolonged (>3 days) ICU stay. Although demonstrating a mortality benefit in that subgroup once again, the intention-to-treat design failed to demonstrate a mortality benefit of IIT in overall patients and is widely considered a negative trial despite the listed morbidity benefits in all subgroups.12

The purpose of the VISEP trial was to determine if the benefit of strict glucose control applied to patients with severe sepsis and septic shock. Due to unacceptably high hypoglycemia rates, the VISEP trial was prematurely terminated and did not reach the recruitment goal. The trial's subsequent lack of power coupled with potential confounding agents inherent in the four-arm design may explain why this trial also failed to show an IIT benefit for both mortality and morbidity.13

A noteworthy point of the negative Glucontrol trial was the lower BG target levels (140–180 mg/dL vs. 180–200 mg/dL) in the conventional arm. Although terminated early for protocol violations and failure to reach target BG levels, those IIT patients who did reach target levels still did not have a mortality benefit when compared with the conventional arm.14

Finally, the largest and most definitive trial to date, NICE-SUGAR, showed a significantly increased mortality rate in the IIT group at 90 days and no positive effect on morbidity as well. For reasons that are not yet clear, excess deaths were predominantly due to cardiovascular causes. Interestingly, despite worse mortality, there were no observable differences in ICU or hospital length of stay, new organ failure rates, ventilator days, bacteremia, or transfusion requirements between the two groups. Also noteworthy is that roughly 33% of the NICE-SUGAR patients were surgical and, unlike the original Leuven surgical trial, there was no mortality benefit in this subgroup.15

A concise summary of the key trials highlighting the study population, endpoints, adverse effects, and criticisms can be found in Table 32-1.

In addition to the aforementioned key RCTs, TGC has also been studied in patients with acute myocardial infarctions (MI), coronary artery bypass grafts (CABG), and cerebral vascular accidents (CVA).

MI and Post-CABG Patients

Hyperglycemia has been shown to be a risk factor for mortality in patients with acute MI. In a study of 16,781 patients with acute MI, the mortality rate increased incrementally with each 10 mg/dL rise in glucose over 120 mg/dL.16 Studies have shown that hyperglycemia in the presence of ischemia is associated with decreased collateral circulation, increased infarct size, and prolonged QT interval.17 Since hyperglycemia has been clearly linked to less “Thrombolysis in Myocardial Infarction (TIMI)” flow before primary percutaneous coronary intervention (PCI), it has been hypothesized to be a strong prothrombotic stimulus initiating procoagulant factors and inhibiting thrombolysis.18 In addition to counteracting these deleterious procoagulant effects, insulin blocks the accumulation of free fatty acids generated from ischemia-induced myocardial anaerobic metabolism that otherwise would promote further oxygen debt and arrhythmias.19

Initial enthusiasm was sparked by the Diabetes and Insulin–Glucose Infusion in Acute Myocardial Infarction (DIGAMI) trial, an RCT which studied diabetics with acute MI. An impressive 30% mortality reduction was shown in the glucose, insulin, and potassium (GIK) infusion group.20 However, the larger and more recent 2005 follow-up study, DIGAMI-2, did not replicate such findings.21 Similarly, the CREATE-ECLA trial showed no difference in mortality, cardiac arrest, or cardiogenic shock in ST-elevation MI (STEMI) patients randomized to a GIK infusion.22 Although the subsequent HI-5 study in 2006 also did not show a mortality benefit in diabetic acute MI patients randomized to GIK, there was a significantly lower incidence of heart failure and reinfarction in that group that has revived the pro-IIT debate.23

It is worth noting that unlike the TGC studies such as NICE-SUGAR and the Leuven trials, the GIK infusion acute MI studies such as the aforementioned CREATE-ECLA and DIGAMI trials did not really achieve TGC. Ostensibly these trials focused more on insulin therapy than on the control of hyperglycemia.

Hyperglycemia is a known risk factor for mortality, deep sternal wound infections, and increased length of stay in patients undergoing CABG. The ongoing large Portland Diabetic Project that is a prospective, nonrandomized, observational study including 5,510 diabetic patients has shown a dramatic and significant 60–77% decrease in mortality and infection risk when employing a continuous insulin infusion.24

CVA Patients

Patients with acute stroke (CVA) have been shown to have worse outcomes when hyperglycemic. The detrimental effects of hyperglycemia may include increasing tissue acidosis secondary to anaerobic glycolysis, lactic acidosis, and free radical production along with a possible contribution to cerebral edema via an effect on the blood-brain barrier.25 There is a recognized 3-fold increase in hemorrhagic transformation in tissue plasminogen activator (tPA)-treated hyperglycemic post-CVA patients as well. A 2001 comprehensive cohort review demonstrated a 3-fold increase in 30-day mortality among hyperglycemic post-CVA patients.26 While admission BG >140 mg/dL has significant long-term mortality association, Baird et al found that persistent hyperglycemia (BG > 200 mg/dL) during the first 24 hours after stroke independently predicted expansion of the volume of ischemic stroke and poor neurologic outcomes.27 A 2009 large observational Glycemia in Acute Stroke (GLIAS) study showed that a BG >155 mg/dL or higher at any time within the first 48 hours from stroke onset was associated with poor outcome independently of stroke severity, infarct volume, diabetes, or age.28

To date, few RCTs studying BG control in the post–CVA patient have been published. One such RCT, the GIST-UK trial in 2008, did not show a mortality or morbidity effect of IIT in acute CVA patients, but was underpowered and the length of treatment (24 hours) was minimal.29 Additionally, the IIT group in the GIST-UK trial only had a mean BG level of 10.3 mg/dL lower than the control group. Since it was suggested that greater reductions in BG levels might be needed to show a clinical benefit, two recent RCTs focused on the clinical feasibility and safety of more aggressive intensive insulin protocols. These studies, which focused on patients with preexisting diabetes, suggested clinical benefit, but were unable to demonstrate definitive improvement.30,31 Of importance, the aforementioned prospective RCTs excluded nondiabetics because they tend to self-correct and enter the target range without intervention.

Given the paucity of RCTs, the guidelines for BG control in CVA patients vary and are constantly evolving. The European Stroke Organization guidelines recommend starting insulin therapy when the BG is >10 mmol/L (181 mg/dL).32 The Stroke Council of the American Stroke Association, which in 2003 initially recommended BG control only when the BG is >300 mg/dL, has recently changed the BG target to 140–185 mg/dL (7.7–10.2 mmol/L).33

For a detailed summary of the MI and CVA studies, see Table 32-2.

The Leuven surgical and medical, NICE-SUGAR, VISEP, and Glucontrol trials, as well as the aforementioned RCTs in selected subgroup populations, have all shown a risk of hypoglycemia in the intensive glucose control groups. Symptoms of hypoglycemia such as headache, fatigue, confusion, and dysarthria are often masked in an ICU patient and may not become evident until the BG is <40 mg/dL. Complications of such severe hypoglycemia include coma, seizures, and even cardiac arrest. In 2007, Krinsley and Grover identified severe hypoglycemia in ICU patients as an independent risk factor for mortality. In fact, a single episode of severe hypoglycemia was shown to significantly increase the risk of mortality over 2-fold. The populations at greatest risk of mortality secondary to hypoglycemia were the patients with preexisting diabetes, on mechanical ventilation, with an admitting diagnosis of septic shock, and with a very high Acute Physiology and Chronic Health Evaluation (APACHE) score.34 It has been suggested that the mortality benefit provided with IIT can perhaps be offset if IIT is too intense and thus finding the right balance is of utmost importance.

In review, although it is clear that severe hyperglycemia is associated with poor outcomes, there are no convincing data to suggest that the tightest glucose control (80–110 mg/dL) provides mortality benefits. An early single-center surgical study showing morbidity and mortality improvements rushed in an era of intensive insulin protocols around the world, but subsequent studies including medical patients and a more heterogeneous group of ICU patients have failed to replicate this finding.

The updated SSC guidelines recommended that patients with severe sepsis and hyperglycemia should be placed on intravenous (IV) insulin infusion to reduce BG levels with a goal of <150 mg/dL.35 In June 2009, an addendum published by the SSC glucose control subgroup in response to the NICE-SUGAR publication recommended against IIT aimed at BG 80–110 mg/dL in patients with severe sepsis. They did however recommend considering glucose control when levels exceed 180 mg/dL, with a goal BG approximating 150 mg/dL.36 In 2009 the ADA and American Association of Clinical Endocrinologists (AACE) guidelines recently advocated a BG target of 140–180 mg/dL with initiation of insulin therapy when BG is >180 mg/dL.37

Glycemic control in the ICU can be obtained by either IV or subcutaneous (SQ) insulin. In a systematic literature review by Meijering et al in 2005, the IV route achieved a target BG level in a higher number of patients than the SQ route alone.38 The standard of care in the ICU is to use IV infusions when feasible and not SQ and IV together.

There are at least 18 current insulin protocols available, but the concept is the same: achieve the best BG control while minimizing the risk of hypoglycemia.39 A typical “sliding” protocol, as seen in the Leuven surgical trial, sets a predetermined amount of insulin according to the range in which the last BG value fell. On the other hand, “dynamic” protocols, such as Goldberg's Yale protocol, make adjustments based on glycemic levels, as well as the rate of change and the degree of insulin resistance.

In transitioning from the ICU to general or intermediate care within the hospital, BG control is still targeted, albeit less tightly. The use of a basal–bolus regimen (i.e., adding basal coverage in addition to the sliding scale dose) has been shown to be almost two times more effective at achieving target BG levels in non-ICU patients than the traditional sliding scale protocols alone.40

Current and future studies focus on incorporating some new approaches evaluating metrics of hyperglycemia beyond average values, such as glycemic variability. Interesting ongoing studies with the goal of avoiding or reducing insulin use focus on utilizing insulin-like growth factor-1 (IGF-1, a signal in the insulin pathway) and glucagon-like peptide-1 (GLP-1, a glucose-lowering incretin) or even simply restricting carbohydrates.41 Providing better and safer target BG levels, improving protocol adherence via online and computerized tools, and incorporating advanced glucose monitoring devices (such as continuous glucose sensors) are also exciting subjects for future investigation, and are likely to contribute to the evolution of glycemic control in the ICU population.