The data on the importance of controlling glucose in hospital settings spans diverse disciplines of medicine. Studies in the areas of stroke, myocardial infarction (MI), bypass surgery, and wound and nosocomial infections all point to the tremendous potential to reduce morbidity and mortality among hospitalized patients with hyperglycemia. It is essential to identify hyperglycemia at the time of hospital admission and to implement therapy to achieve and maintain glucose levels as close to normal as possible, regardless of a patient's primary reason for admission or previous diabetes status.

In the United States, there are more than 4.2 million hospitalizations annually among people with diabetes.¹ Additionally, as many as 1.5 million hospitalized individuals have significant hyperglycemia but no history of diabetes.² Identification of and therapeutic interventions to treat hyperglycemia must be initiated in tandem with treatment of the presenting medical problem rather than days after admission when many of the acute issues have been addressed. The data presented strongly suggest that an early and aggressive approach to the management of hyperglycemia may reduce mortality, morbidity, excessive hospital stays, and added costs.

Why is hyperglycemia so easy to overlook?
Hyperglycemia is often overlooked when a hospitalized patient is acutely ill and facing a life-threatening illness. Commonly, a patient is admitted with an acute MI or cerebrovascular accident and is coincidentally found to have an elevated admission glucose value. Often, such patients have no history of glucose intolerance.

Physicians routinely discontinue their patients' outpatient diabetes care regimen and initiate sliding scale insulin coverage.³ Many physicians erroneously consider this practice the standard of care. Concerns about precipitating hypoglycemia may limit more aggressive strategies for managing hyperglycemia, particularly when patients are not tolerating regular meals or when their intake is being limited for any of a variety reasons, including pending surgical procedures or diagnostic tests.

Despite the ease and high frequency of use, sliding scale insulin coverage often results in a deterioration rather than an improvement in glycemic control.³ Concerns about hypoglycemia are warranted. However, hyperglycemia, regardless of whether a previous diabetes diagnosis has been made, may pose even greater risks by reducing hospital survival rates among patients admitted with stroke or MI.^4-6

Is hyperglycemia caused by stress or diabetes? It doesn't matter; treat it.
There are no unique diagnostic criteria that account for stress in acutely ill patients, nor are there recommendations for making a definitive diagnosis of diabetes in the hospital. There is a well-defined pathophysiology of the effects of intercurrent illness and surgery on carbohydrate metabolism, which may lead to hyperglycemia. However, many studies have also demonstrated that stress may result in diminished glucose values.^7-20 This reduction in glucose levels during times of stress is frequently attributed to insufficient administration of exogenous glucose.

Among hospitalized patients with acute MI, an admission glucose value of >180 mg/dl predicted undiagnosed diabetes rather than stress hyperglycemia in a study in which newly recognized hyperglycemic individuals had subsequent glucose tolerance testing performed 2 months after hospital discharge.^4,5 The Diabetes Insulin-Glucose in Acute Myocardial Infarction (DIGAMI) trial demonstrated significant reductions in mortality when an intensive insulin regimen was administered to hyperglycemic patients hospitalized with acute MI.^4,5 Enrollment in the DIGAMI study included all patients with glucose values >198 mg/dl without regard to previous diabetes status. Nearly 15% of the study population did not have a history of glucose intolerance.

Is the hospital really the time to consider diabetes? Yes, if the diagnosis is missed, the risk for death increases.
Mortality rates among patients with diabetes are known to be significantly higher than those of nondiabetic individuals.²¹ The Whitehall Study reported a mortality rate of 12 per 1,000 person-years for nondiabetic individuals, 40 per 1,000 person-years for those with undiagnosed diabetes, and 27 per 1,000 person years for people with diagnosed diabetes.²² The Paris Prospective Study demonstrated comparable findings when comparing the mortality rates among patients with undiagnosed diabetes (23%) with those with a definitive diagnosis of diabetes (20%), while patients without diabetes had a 9% mortality rate.²³ In both studies, 60­70% of the deaths were attributed to cardiovascular disease (Figure 1).²¹

Where is the best place to find unrecognized diabetes? In the hospital.
Approximately 5.3 million Americans have undiagnosed diabetes. As many as one-third of the patients who have significant hyperglycemia during their hospital admission do not have any diabetes history.² Given the 7- to 10-year delay between the onset of diabetes and the time of diagnosis, there is a high likelihood that patients exhibiting hyperglycemia without a history of glucose intolerance may have unrecognized diabetes.²⁰ The beneficial impact of interventions that improve hyperglycemia is based on the presence of diabetes, regardless of whether a physician has made the diabetes diagnosis.

When high glucose levels are present, assume that the patient has diabetes, and initiate treatment to control the glucose levels to as close to normal as possible. Further evaluation of the patient's diabetes status can occur following hospitalization. However, failure to address and treat the hyperglycemia represents a missed opportunity not only to reduce hospital morbidity and mortality, but also to initiate interventions that may delay the long-term complications of this disease.

What if the hyperglycemia is caused by medication? Treat it like diabetes.
Often, medications such as steroids and thiazide diuretics, which have been associated with worsening glycemic control, are required for patients with diabetes. Dozens of medications have been described as inducing diabetes, yet there has been little documented about the glycemic status of these patients before the initiation of the offending drugs.

Decades ago, corticosteroids were studied as a means of unmasking impaired glucose tolerance.²⁴ Only 3% of normal control subjects had positive glucose tolerance tests when pretreated with corticosteroids. Other studies among corticosteroid-treated individuals have found that <20% of steroid-treated individuals develop diabetes.

This indicates that hyperglycemia in hospital settings should be assumed to be diabetes.^25,26 When medications that potentially produce hyperglycemia are required to treat another medical problem, treatment of hyperglycemia should be initiated.

Is there a diagnostic role for hemoglobin A_1c? Maybe.
Many clinicians have advocated the use of hemoglobin A_1c (HbA_1c) in the diagnosis of diabetes.^27-30 Although an elevated HbA_1c collected when hyperglycemia is first noted can help determine that the hyperglycemia existed before hospitalization, a normal HbA_1c value does not preclude the diagnosis of diabetes. Normal HbA_1c values measured during hospital admission did not exclude the diagnosis of diabetes when glucose tolerance testing was performed following discharge.³¹

As more laboratories begin using the standardized methodologies for performing HbA_1c testing as established by the National Glycohemoglobin Standardization Program (NGSP), it will be easier to establish norms and ranges for HbA_1c and enhance the ability to develop guidelines for the diagnosis of diabetes based on elevated HbA_1c values. Laboratories certified by the NGSP all have equivalent HbA_1c assays and can be compared nationally regardless of the location or the specific laboratory performing the test. Normal HbA_1c values in the hospital will not preclude the diagnosis of diabetes, and such patients will need further follow-up to evaluate diabetes status.

Diabetes isn't the reason for admission, so why treat? Hyperglycemia does affect the primary medical problem.
Ninety percent of hospitalizations among patients with a known diagnosis of diabetes are for reasons other than diabetes. Approximately 75% of hospital admissions for people with diabetes are attributable to cardiovascular disease.³² When hyperglycemia is not addressed until after the presenting medical problems have been treated, the length of hospital stay can be significantly longer than that of nondiabetic patients with similar diagnoses. When the diabetes is addressed along with other acute problems, not only are there improved outcomes, but also potentially shorter hospital stays.³³

Do transient rises in glucose really matter? Modest glucose elevations can worsen inpatient outcomes.
For the past 50 years, researchers have been evaluating the impact of glucose on immune biology. In 1964, Bybee and Roberts reported impairment of phagocytic function in white blood cells from patients with diabetes.³⁴

For several decades, the cellular mechanisms through which glucose affects immune function and vascular biology have been explored. Short and transient periods of hyperglycemia have resulted in abnormalities in granulocyte adherence, chemotaxis, phagocytosis, and bacterocidal function.^35-41 These impairments in immune function have been associated with an increased risk for wound and nosocomial infections. Acute hyperglycemia also results in adverse consequences on volume status and electrolyte imbalances, which further potentiate poor wound healing and can increase the risk for infections.^35-41

The well-documented studies demonstrating the deleterious short-term effects of hyperglycemia ultimately led to two landmark studies: the Diabetes Control and Complications Trial (DCCT) and the United Kingdom Prospective Diabetes Study (UKPDS), both of which demonstrated the long-term benefits of glycemic control. The DCCT found that a 1% reduction in HbA_1c could result in as much as a 45% risk reduction in the development of serious retinopathy. Comparable reductions in the risk were demonstrated for nephropathy and neuropathy. The UKPDS reported that a 1% reduction in HbA_1c resulted in a 25% reduction in all-cause diabetes-related mortality among a large population of newly diagnosed type 2 patients.

The Inpatient Data
Inpatient studies have been designed to evaluate the impact of potentially correcting the pathophysiological changes that accompany hyperglycemia. Studies among patients with stroke, MI, and those undergoing bypass surgery suggest that treatment of hyperglycemia can potentially affect morbidity and mortality, and similar benefits may apply to controlling hyperglycemia regardless of whether patients have a prior diagnosis of diabetes (Table 1).

Myocardial infarction
Diabetes is an independent risk marker for morbidity and mortality among patients who have suffered an MI.^4,5,42-45 Soler and Frank observed that patients with the highest glucose values after an acute MI also had the highest mortality rates.⁴²

The DIGAMI study underscores the importance of early and aggressive interventions designed to bring glucose levels into the normal range regardless of a patient's prior diabetes status.^4,5 This large, randomized prospective trial enrolled 660 patients with admission glucose values of >198 mg/dl. Hyperglycemic individuals were randomized to receive either conventional diabetes care or intravenous insulin followed by four insulin injections of daily.

One year after admission, there was a 30% reduction in mortality among the intervention patients. The greatest benefits from intensive insulin therapy were seen in the subgroup of patients without a history of diabetes, who had a 58% risk reduction in hospital mortality and a 52% risk reduction in mortality when followed for 1 year compared to conventionally treated patients.

When outcomes were tracked for a mean of 3.4 years, intervention patients had 25% lower death rates. The DIGAMI study demonstrated that for every nine patients receiving intensive glucose control, one life was saved.

Stroke
Epidemiological data have demonstrated an increased risk of stroke among patients with diabetes.^46-59 The Framingham Study found that the incidence of stroke was 2.5 times higher in diabetic men and 3.6 times higher in diabetic women than in those without diabetes.⁶ Several studies have suggested that hyperglycemia is an independent risk factor influencing stroke severity. Hyperglycemia in nondiabetic individuals with stroke has also been associated with increased mortality.⁶⁰

Although many studies suggest that normalizing glucose may improve stroke outcomes among patients with diabetes and hyperglycemia, randomized trials such as DIGAMI have not been conducted among stroke patients. In a prospective analysis, Pulsinelli et al. reported that both patients with diabetes and hyperglycemic patients without an established diagnosis of diabetes had worse neurological outcomes than those who were normoglycemic.⁵³ Stroke-related deficits were more severe when the admission glucose values were >120 mg/dl (6.7 mmol/l). Only 43% of the patients with an admission glucose value of >120 mg/dl were able to return to work, whereas 76% of patients with lower glucose values regained employment.

Jorgenson et al. evaluated the relationship between plasma glucose and mortality rates among 233 diabetic patients hospitalized with an acute stroke. Diabetes was an independent risk factor that doubled the mortality rate.⁶ Among 623 hyperglycemic patients who did not have a history of diabetes, mortality and initial stroke severity increased linearly with increasing admission glucose values >108 mg/dl (Figure 2).

Weir and colleagues found that a plasma glucose >144 mg/dl within the first 24 hours of admission was a risk factor independent of age, stroke type, and stroke severity and predicted double the mortality risk.⁵⁶ Similar findings by Asplund et al. demonstrated that the presence of diabetes adversely affected short- and long-term neurological outcomes among patients with both ischemic and hemorrhagic strokes.⁶⁰

Other studies have found that admission glucose levels and/or HbA_1c values correlate to stroke size, clinical severity, and prognosis, and a similar relationship between glucose levels and outcomes applies to hyperglycemic, nondiabetic individuals with hyperglycemia.^48-50

Coronary Artery Bypass and General Surgery
Several studies and data from the National Cardiac Surgery Database (NCSD) have documented an increased operative mortality and postoperative mortality among patients with diabetes.^61-68 Herlitz et al. reported that the diabetic mortality rate within the 30 days following bypass surgery was 6.7% compared to 3.0% in nondiabetic subjects.⁶¹ Weintraub et al. reported an in-hospital mortality rate of 5.4% among 279 diabetic patients undergoing angioplasty compared to no deaths among the 610 nondiabetic patients undergoing angioplasty during the same time period.⁶² The Bypass Angioplasty Revascularization Investigation (BARI) reported that surgically treated diabetic patients had lower 5-year survival rates compared to nondiabetic patients (80.6% vs. 91.4%).⁶³

Multiple factors have been implicated as possible mechanisms contributing to poorer outcomes, including suboptimal perioperative glucose control, which increases fatty acid metabolism, compromises glycolysis in muscle, and impairs platelet and fibrinolytic function. These factors predispose hyperglycemic patients to infection and impaired wound healing. Therapies to improve glycemic control have demonstrated the potential benefits in reducing morbidity and mortality among patients undergoing cardiac bypass surgery.

Bypass surgery: perioperative control counts. A provocative study by Kalin and colleagues evaluated the impact of intravenous insulin administered to patients with diabetes before and throughout bypass surgery.⁶⁹ Outcomes of 400 patients with diabetes were compared to a population of 876 nondiabetic patients who underwent bypass and to national data from the NCSD during the same time period. Hospital mortality was almost identical between the diabetic and nondiabetic patients (1.75% vs. 1.71%). During this same time period, the NCSD reported nearly a 50% higher mortality (RR 1.46) for patients with diabetes compared to nondiabetic patients undergoing bypass. The use of intravenous insulin initiated before surgery potentially reduced the excess hospital mortality that was observed nationally among patients with diabetes undergoing bypass surgery.

In a smaller study among bypass patients, Marcus et al. optimized glucose control through the use of intravenous insulin among diabetic patients with glucose levels >120 mg/dl.⁷⁰ Medical interventions were made to correct other metabolic abnormalities, including elevated triglycerides, uric acid, and urine microalbumin. Patients receiving these metabolic interventions inclusive of intravenous insulin had a significantly lower combined incidence of operative death, MI, cerebrovascular accidents, renal failure, and the need for intraoperative balloon pump (18% vs. 57%, P < 0.004).

Perioperative glucose control reduces infections and death. Diabetes has been established as an independent risk marker for postoperative surgical wound infection and carries with it a 2-5 times higher infection rate than nondiabetic populations.^71,72 Additionally, medianstinitis following bypass surgery occurs significantly more often among patients with diabetes, and hyperglycemia in the immediate postoperative period is an independent risk marker for deep sternal wound infections.

Furnary and colleagues prospectively compared the outcomes of patients receiving subcutaneous insulin injections to those receiving intravenous insulin initiated perioperatively.⁷³ The incidence of deep sternal wounds was 0.8% in the group receiving intravenous insulin compared with 2.0% in patients receiving intermittent subcutaneous insulin. Additionally, Furnary reported a fivefold increased risk for death (3.8% vs. 19%) among the patients who developed deep sternal wounds as compared to those who did not develop wounds postoperatively. Zerr and colleagues reported that the use of an intravenous insulin protocol begun immediately postoperatively resulted in a 60% reduction in deep sternal wound infections. Wound infections were best correlated to an average blood glucose level of >200 mg/dl on postoperative day 1⁷⁴ (Figure 3).

Figure 3. Postoperative blood glucose at 48 hours predicts wound infection (P = 0.002). From ref. 74. Reprinted with permission.

Golden and colleagues evaluated 411 patients with diabetes who underwent coronary artery bypass surgery and tracked the glucose values during the 36-hour period following surgery.⁷⁵ Postoperative hyperglycemia predicted risk for short-term infections and was a risk factor independent of age, sex, race, underlying co-morbidity, or severity of the underlying illness. Patients with a mean glucose of >200 mg/dl were more likely to develop infections than those with values <200 mg/dl. (See related article on p. 44.)

In a randomized study by Rassias et al. testing the effect of an insulin infusion on perioperative neutrophil function in diabetic patients scheduled for coronary artery bypass surgery, participants (n = 26) were randomly assigned to receive either aggressive insulin therapy or standard insulin therapy during surgery.⁷¹ Blood was drawn for neutrophil testing before surgery, 1 hour after the completion of cardiopulmonary bypass, and on the first postoperative day. Neutrophil phagocytic activity decreased to 75% of baseline activity in the aggressive therapy group and to 47% of baseline activity in the standard group (P < 0.05). The authors concluded that continuous insulin infusion improves the ability of white cells to function in diabetic patients and may increase resistance to infection after surgery.

General surgery and medical patients. Pomposelli et al. found that among 97 patients with diabetes undergoing surgery, a postoperative day 1 glucose of >220 mg/dl was a sensitive predictor of nosocomial infections.³⁵ Patients with elevated glucose values developed infections at a rate 2.7 times higher than did patients with glucose values <220 mg/dl. Many of these infections seen in the diabetic population were quite severe and included sepsis, pneumonia, and wound infections. A separate evaluation that removed patients with less severe urinary tract infections from the analysis found that the risk for severe infection during the hospitalization rose 5.9 times higher for patients with glucose values of >220 mg/dl on postoperative day 1.

Similarly, Korytkowski et al. found that the degree of hyperglycemia on admission contributed significantly to length of hospital stay among patients admitted with acute medical problems, including pneumonia and acute MI.⁷⁶

A New Hospital Paradigm for Glycemic Control Is Needed
Korytkowski reported that, despite significantly elevated glucose levels in the majority of hospitalized patients with diabetes, there is no standardized method for controlling glucose.⁷⁶ There is a strong need to standardize the monitoring and intervention practices for hospitalized patients who have diabetes as a secondary medical problem. The deleterious consequences of suboptimal glycemic control among inpatients have been clearly documented, but few studies have evaluated how to implement strategies for achieving better glycemic control.

Sliding scale coverage does not work.
Sliding scale regimens have often been considered the standard of care, yet few studies have evaluated outcomes between sliding scale insulin coverage and other treatment regimens. Queale et al. found that sliding scale insulin therapy provided no benefit as compared to a standing dose of intermediate-acting insulin, and patients who received sliding scale regimens were 3 times more likely to have hyperglycemic episodes.³

No data have been published demonstrating that the use of sliding scale insulin benefits hospitalized patients. Very few studies have directly evaluated which is the best modality for treating inpatients with hyperglycemia, but numerous papers have pointed out the potential deleterious consequences of using sliding scale coverage rather than definitive therapy.³

Intravenous insulin makes physicians uncomfortable.
Studies demonstrate the importance of controlling glucose and the advantages of intravenous insulin over subcutaneous insulin. Many physicians are uncomfortable treating diabetes, and the use of intravenous insulin therapy has been limited to intensive care units or special care units in which the nursing staff has expertise using intravenous insulin infusions.⁷⁷

Despite the trend toward replacing specialty care with primary care, patients may fare better when subspecialists are involved in the care of acutely ill hospitalized patients with hyperglycemia.^78-80 Not only are there improved outcomes resulting from consultations by diabetes specialty teams, but studies have also demonstrated the beneficial impact of cluster units in which patients with both primary and secondary diagnoses of diabetes benefit from care by nurses and other medical staff with specialized training in diabetes. Reduced lengths of hospital stay have been reported when such units are available to both medical and surgical patients.⁸¹

New technology is near.
As continuous subcutaneous glucose monitoring is now becoming available, monitoring of glucose and treatment of hyperglycemia may become much easier. This type of monitoring device measures interstitial glucose every 5 minutes. It may be an extremely important tool for physicians treating hospitalized patients.

Frequent monitoring can allow for more accurate dosing of insulin and can dramatically improve care. The tremendously successful insulin pump therapy, which delivers subcutaneous insulin in easily adjustable dosages, could potentially be utilized in conjunction with the new glucose monitoring technologies and could revolutionize the treatment hyperglycemia in the hospital.

Conclusion
Today, there are neither systems for the identification of patients with hyperglycemia nor standardized protocols for the monitoring and treatment of hyperglycemia in hospitals. Subcutaneous sliding scale insulin coverage has been a hospital standard despite the fact that there is no evidence demonstrating its efficacy, and the medical literature suggests that subcutaneous coverage alone may actually be detrimental to glycemic control.

Newer methodologies for earlier identification of patients with hyperglycemia coupled with innovative systems for monitoring glucose and delivering insulin will help reduce morbidity and mortality among the nearly 6 million annual hospitalizations that are accompanied by hyperglycemia in the United States.

The hospital setting provides a unique opportunity to identify patients who may have unrecognized diabetes and may be the easiest and best place to evaluate and address a patient's diabetes status and develop a treatment plan for subsequent outpatient diabetes follow-up and care.

Resa Levetan, MD, is director of diabetes education at MedStar Research Institute in Washington, D.C. She is an associate director of Clinical Diabetes.