University of Washington School of Medicine Seattle, Washington 98195
The definition of hyperglycemia in the absence of diabetes (stress hyperglycemia) varies, but it is often defined as a plasma glucose level above 200 mg/dl (11.1 mmol/liter). The majority of studies examining the frequency of acutely ill patients reaching this glucose level have not been able to determine which of these patients had preexisting diabetes because hemoglobin A1c (HbA1c) levels were not measured. An exception to this was a report of 397 subjects with an acute myocardial infarction using HbA1c to diagnose preexisting diabetes (1). Undiagnosed diabetes was found in 4.3% of this population and contributed to 9.6% of hospital mortality. More recently, Levetan et al. (2) surveyed 1034 hospitalized patients and reported that over one third had one or more documented plasma glucose levels over 200 mg/dl (11.1 mmol/liter), with a mean of 299 mg/dl (16.6 mmol/liter). Although many of these patients likely had diabetes before their hospitalizations, only 7% had diabetes mentioned as a possible diagnosis in the daily progress notes.
Even when high glucose levels are noted, in my experience little effort is made to differentiate diabetes from stress hyperglycemia. Although many of us have been amazed at the apathy in treating people with known diabetes during an acute illness requiring hospitalization, treating stress hyperglycemia, even if profound, has not been a standard of care in this country.
The mechanisms for stress hyperglycemia are well known (3). Counterregulatory hormone and cytokine excess results in insulin resistance, and many hospitalized patients are insulin deficient for a variety of reasons (e.g. old age, pancreatitis, hypothermia, hypoxemia). Excess dextrose infusion is an often-overlooked contributor to hyperglycemia. In general, the sicker the patient, the more likely stress hyperglycemia will occur (4). Stress hyperglycemia also increases the risk of death, congestive heart failure, and cardiogenic shock after myocardial infarction (5) and increases in-hospital mortality after ischemic stroke (6).
There has been a growing interest in the effects of acute illness on increased cytokine production, particularly TNF-. This cytokine seems to influence muscle metabolism in three ways: promoting catabolism, inhibiting contraction, and modulating myogenesis (7). TNF-
has also been shown to have a role in the insulin resistance of obesity and type 2 diabetes in both liver and skeletal muscle, most likely through the modification of signaling properties of insulin receptor substrates (8). TNF-
is released early in the course of acute myocardial infarction and can directly decrease myocardial contractility in a dose-dependent fashion (9). This seems to occur by inducing myocardial apoptosis (10). Not surprisingly, heart failure results in high TNF-
levels (10). TNF-
also causes endothelial dysfunction (11), triggers procoagulant activity and fibrin deposition (12), and enhances nitric oxide (10).
Another cytokine released during acute illness is migration-inhibitory factor (MIF). This cytokine is secreted from macrophages in addition by a variety of tissues, including the pituitary, brain, kidney, lung, prostate, and testes. MIF has been shown to be increased in patients with sepsis and septic shock. MIF expression is increased in animals exposed to both Gram-negative and Gram-positive bacterial toxins, and neutralization of MIF or deletion of the MIF gene can protect mice from lethal endotoxemia or staphylococcal toxic shock. Interestingly, MIF knockout mice showed reduced circulating TNF-, suggesting that MIF regulates TNF-
synthesis.
As more is learned about the cellular mechanisms of acute illness, two recent studies deserve special mention. In this issue of JCEM, Umpierrez et al. (13) describe over 2000 patients admitted to a general medical ward. Similar problems with the classification of glycemia occurred: 7% of admission did not have even one blood glucose measurement. Of the rest, 26% had a previous history of diabetes and 12% had newly diagnosed hyperglycemia [admission or in-hospital fasting glucose >126 mg/dl (7.0 mmol/liter) or a random blood glucose >200 mg/dl (11.1 mmol/liter) on two or more occasions]. Those with new hyperglycemia had higher in-hospital mortality (16%) than those with a known history of diabetes (3%) and normoglycemia (1.7%), both P < 0.01. Multivariate odds ratios for mortality were calculated using logistic regression and compared with the normoglycemia group. After adjusting for eight factors, the new hyperglycemia group had an 18.3-fold increased mortality rate compared with a 2.7-fold increase in the known diabetes group. Interestingly, blood glucose levels assessed on admission and at random were lower in those patients with newly diagnosed hyperglycemia. The authors conclude that the reason for the higher mortality in the new hyperglycemia group is that they were more severely ill. Importantly, 42% of those with new hyperglycemia were prescribed insulin therapy with 35% receiving "sliding scale" insulin; however, we are given no details about the insulin therapy. For those with preexisting diabetes, 77% received insulin therapy.
Van Den Berghe et al. (14) reported on 1548 patients admitted to a surgical intensive care unit. Only 13% of patients had preexisting diabetes, although admission HbA1c levels were not measured. One group (intensive treatment group) received an insulin infusion to maintain normoglycemia with a final glucose concentration of 103 ± 19 mg/dl (5.7 mmol/liter). The other group (conventional therapy group) received insulin only if the blood glucose level exceeded 215 mg/dl (11.9 mmol/liter) with a goal of maintaining the glucose level between 180 and 200 mg/dl (10.011.1 mmol/liter). At the end of the study, the blood glucose concentration in this group was 153 ± 33 mg/dl (8.5 mmol/liter) because 61% did not require insulin therapy. During admission to the intensive care unit, intensive treatment with iv insulin reduced the risk of death by 42%. This finding was due exclusively to the sickest patients who required a stay in the intensive care unit of over 5 d (10.6% mortality in the intensive treatment group compared with 20.2% mortality in the conventional treatment group, P = 0.005). Other improved outcomes in the intensive treatment group included reductions in overall in-hospital mortality (34%), sepsis (46%), acute renal failure (41%), and the median number of red cell transfusions (50%).
Although this protocol can be criticized because it could not be blinded and it is possible there could have been other subtle differences in care with the intensive treatment group, the results are dramatic. This study adds to the growing literature that for patients without diabetes mild hyperglycemia increases both morbidity and mortality. One conclusion of this report is that aggressive attention to glycemia in a population of surgical patients in an intensive care unit can improve these important outcomes (15). But is this the correct conclusion?
Considering the increase in cytokine release during acute illness, is it possible that the insulin administration in both of these studies could have explained the outcomes? Studies from the 1980s clearly show that insulin inhibits TNF (also called cachectin) in rats (16, 17). Unfortunately, human data are lacking, but this hypothesis could explain why Van Den Berghe et al. (14) found such impressive results with almost all of the patients in the intensive therapy group receiving an iv insulin infusion. Furthermore, the ability of insulin to inhibit TNF may also explain why Umpierrez et al. (13) observed a higher mortality in the new hyperglycemia group even though this group had a lower mean blood glucose level than those with preexisting diabetes. Few patients with new hyperglycemia received insulin therapy in contrast to those with preexisting diabetes.
Of course, attributing the results of these two studies to the ability of insulin to inhibit TNF and possibly other cytokines is just a hypothesis and requires direct testing. Additional studies should measure these cytokines and determine the relationships between serum insulin levels, cytokine levels, and clinical outcomes. Is iv insulin superior to sc insulin? The above interpretation of the Umpierrez study (13) suggests sc insulin may be effective. If that is the case, would continuous sc insulin infusion be superior to intermittent insulin administration? How would insulin therapy affect other hospitalized populations? Patients with cancer and cystic fibrosis as well as those with sepsis would all be excellent models because they have demonstrated high levels of TNF. Finally, is cytokine inhibition by insulin one of the mechanisms responsible for the improved mortality seen with the glucose-insulin-potassium (GIK) infusion during acute myocardial infarction in patients with (18) and without (19, 20) diabetes? Certainly, the GIK infusion improves myocardial substrate metabolism (21) and seems to improve myocardial performance and results in a faster recovery after coronary artery bypass grafting (22). As noted, TNF- has many cardiotoxic effects during acute myocardial infarction. Is it possible the benefits of insulin administration during acute myocardial infarction are multifactorial with cytokine inhibition one of the mechanisms?
Although these hypotheses remain untested, what are we to do now? Do we continue to focus on glycemia of hospitalized patients (23), or should we instead focus on treating people more aggressively with insulin therapy? The first report of the GIK infusion improving mortality from acute myocardial infarction was in 1965 (24). At least in the United States, the GIK infusion is rarely performed for this indication. Apathy treatment of hospitalized patients with hyperglycemia, with or without a previous diagnosis of diabetes, is not going to change based on these two recent studies. This is due, in part at least, to the lack of experience using insulin therapy (25). For some physicians (and their patients), the primary goal of outpatient diabetes therapy is to avoid insulin treatment. Although the reasons for this insulin phobia are complex, the primary culprits are the physician and patient training and the absence of standardization for its use.
I suspect that in-patient diabetes management has progressed so little over the past 30 yr it is not realistic to expect physicians to leap to the type of care provided in the Van Den Berghe study (14). To be fair, it is premature to place all hospitalized patients on a GIK infusion. Nevertheless, it is my hope that the main effect of these studies will be to stimulate interest in treating hyperglycemia that otherwise would be ignored. Perhaps undiagnosed diabetes will be more readily noted and treated appropriately after discharge. Although guidelines for maintaining plasma glucose levels of patients with diabetes between 120 and 180 mg/dl (6.710.0 mmol/liter) are probably much too high, improving the miserable glycemic control currently accepted by physicians (2) by any amount would be an improvement of major proportions. Indeed, in-patient diabetes management has developed into an area of medicine that is less evidenced based and more of an ignorance-based culture with a core component of sliding scale insulin, a relic from generations past with no proven efficacy (26, 27). Although we now try to sort out the meanings of the studies from Umpierrez et al. (13) and Van Den Burghe et al. (14), it is my hope that the 60-yr-old patient admitted to the medicine service with pneumonia and a random admission plasma glucose of 230 mg/dl (12.8 mmol/liter) will not be metabolically ignored. Instead, a conscientious physician will provide an appropriate assessment and treat the patients hyperglycemia.
Acknowledgments
Footnotes
Address all correspondence and requests for reprints to: Irl B. Hirsch, M.D., University of Washington School of Medicine, 1959 NE Pacific Street, Box 356176, Seattle, Washington 98195-6176. E-mail: .
Abbreviations: GIK, Glucose-insulin-potassium; HbA1c, hemoglobin A1c; MIF, migration-inhibitory factor.
Received January 4, 2002.
Accepted January 4, 2002.
References