Key Clinical Questions
Why control diabetes and hyperglycemia in the hospital?
What are the goals of glycemic management in critical care and noncritical care settings?
When should patients be treated with oral agents, with subcutaneous insulin, and with continuous insulin infusions?
Is it safe to prescribe subcutaneous insulin for the first time in a hospitalized patient?
How should insulin orders be adjusted in response to poor inpatient glucose control?
What should be considered when discharging a patient with diabetes or inpatient hyperglycemia?
How may the quality of glucose management be improved in hospital settings?
Introduction
Diabetes mellitus is common in hospitalized patients. In 2007, diabetes was listed as a diagnosis in approximately 19% of all hospital discharges in the United States. Because discharge diagnoses may not capture undiagnosed diabetes or hospital-related hyperglycemia, the true prevalence of diabetes or hyperglycemia in hospitalized patients is likely much higher. For example, in one study of 1886 general medicine and surgery admissions, hyperglycemia (defined as a fasting plasma glucose of ≥ 126 mg/dL or a random plasma glucose of ≥ 200 mg/dL) was found in 38% of patients. Of these, 68% had known diabetes, but the other 32% had no previous diagnosis and consisted of patients with undiagnosed diabetes, “prediabetes” unmasked by the stress of illness, and pure “stress hyperglycemia.”
Hyperglycemia is associated with worse outcomes among hospitalized patients, including infections, increased length of stay, decreased independent living after discharge, and increased mortality. Hyperglycemia among known diabetics in the hospital carries a 2.7 relative risk of in-hospital death; even more dramatic is the 18-fold increased risk of death associated with hyperglycemia among those without a previous diagnosis of diabetes. Similar results have been seen in patients with pneumonia, myocardial infarction, chronic obstructive pulmonary disease, stroke, and those receiving total parenteral nutrition. Some of this excess mortality reflects the role of hyperglycemia as a marker for physiologic stress and severity of underlying disease. However, even when rigorous adjustment for severity of illness is conducted, excess mortality remains.
Rationale for Inpatient Glucose Control
Does correcting inpatient hyperglycemia improve outcomes? In the noncritical care setting, this question has remained unanswered until recently.
RABBIT Surgery was a dual-site randomized controlled trial of 211 patients with type 2 DM admitted to general surgery services. The intervention group received a basal-bolus regimen with glargine once daily and glulisine before meals with a goal pre-meal glucose of 100–140 mg/dL. The control group received sliding scale regular insulin 4 times daily for glucose > 140 mg/dL. Mean glucose for the hospitalization was 157 vs. 176 mg/dL in favor of the intervention. Proportion of patients with any glucose < 40 mg/dL was 3.8% in the intervention group and 0% in the control group. Most importantly, the primary outcome, a composite of hospital complications including postoperative wound infection, pneumonia, respiratory failure, acute renal failure, and bacteremia, occurred in 24% in the control group and only 9% in the intervention group (p = .003; number needed to treat of 7 to prevent one complication). In the critical care setting, several studies of tight glycemic control have been performed, with mixed results. Early studies, such as the Leuven study conducted by van den Berghe, et al, in surgical intensive care unit (ICU) patients, were clearly positive. In Leuven I, patients were randomized to an insulin infusion protocol triggered when the blood glucose was > 100 mg/dL or a protocol that was triggered when glucose was > 215 mg/dL. The mean achieved glucose was 103 mg/dL vs. 153 mg/dL. Outcomes in the intervention arm included a 46% reduction in sepsis, 41% reduction in the need for dialysis, 50% reduction in blood transfusions, 44% reduction in polyneuropathy, and a 34% reduction in inpatient mortality. However, subsequent studies have not been so favorable. A follow-up study conducted by the same group on medical ICU patients was equivocal: mortality did not improve except in a post hoc analysis of patients with an ICU length of stay of three days or more, although morbidity was lessened (faster weaning from mechanical ventilation, discharge from the ICU, and discharge from the hospital). Of note, separation of glucose control was less pronounced than in the SICU study (mean glucose 153 mg/dL vs. 111 mg/dL), and rates of severe hypoglycemia were much higher (19% vs. 3% of patients with at least one glucose < 40 mg/dL). More recently, the NICE-SUGAR study was even less favorable to tight glucose control. In this open-label study of 6104 critically ill patients (medical and surgical), the intervention arm had a slightly increased 90-day mortality rate (27.5% vs. 24.9%), and a much higher rate of severe hypoglycemia (6.8% vs. 0.5%).
Some providers have interpreted NICE-SUGAR as justifying lax glucose control in critically ill patients. There are many reasons to avoid this approach:
In the NICE-SUGAR study, “control patients” actually had reasonably good glucose control: 69% were on an insulin infusion, and the median glucose was 141 mg/dL.
The high rates of hypoglycemia observed may be a specific problem with the Leuven insulin infusion protocol. Lower rates of hypoglycemia are seen with the use of other insulin infusion protocols, such as Yale and Glucommander, which are easier to administer and take better account of trends in glucose control over time (Table 149-1).
Septic patients may be particularly prone to hypoglycemia (medical ICU patients in these studies always have higher rates of hypoglycemia than surgical ICU patients; Table 149-1). The rational clinical response should be higher glucose targets in these patients, not lack of any target at all.
Insulin is one of the highest risk medications used in the hospital, and yet its inpatient use has been historically haphazard and irrational. A standardized approach to insulin use can prevent both severe hyperglycemia and hypoglycemia, both known to be hazardous to patients.
Several quality improvement studies, including several conducted by hospitalists, have shown the ability to institute processes that reduce hyperglycemia, while maintaining or even reducing the rate of hypoglycemia.
Study | Patient Population | Hypoglycemia Rate* |
---|---|---|
Leuven Protocol | ||
Leuven 1 (N Engl J Med. 2001;345:1359–1367) | Surgical | 5.1% |
Leuven 2 (N Engl J Med. 2006;354:449–461) | Medical | 19% |
Glucontrol (Intensive Care Med. 2009;35:738–748) | Medical /Surgical | 8.6% |
VISEP (N Engl J Med. 2008;358:125–139) | Medical | 17% |
NICE-SUGAR (N Engl J Med. 2009;360:1283–1297) | Medical /Surgical | 6.8% |
Yale Protocol | ||
Yale 1 (J Cardiothorac Vasc Anesth. 2004;18:690–697) | Surgical | 0% |
Yale 2 (Diabetes Care. 2004;27:461–467) | Medical | 4.3% |
Glucommander Protocol | ||
Glucommander 1 (Diabetes Care. 2005;28:2418–2423) | Surgical | 2.6% |
In response to recent studies, the American Association of Clinical Endocrinologists (AACE) and the American Diabetes Association (ADA) have updated their guidelines. In critically ill patients, extrapolating from NICE-SUGAR’s control group, they recommend starting insulin infusions at a threshold no higher than 180 mg/dL, and aiming for a target range of 140–180 mg/dL. Recommendations for noncritically ill patients remain essentially unchanged: goals of premeal glucose readings < 140 mg/dL and random glucose readings < 180 mg/dL, with more or less stringent goals depending on previously achieved control and medical comorbidities (note that these goals might change in response to RABBIT Surgery, published after the release of these guidelines).
We agree with the approach of the National Quality Forum, whose Recommended Safe Practice advocates taking “actions to improve glycemic control by implementing evidence-based intervention practices that prevent hypoglycemia and optimize the care of patients with hyperglycemia and diabetes.” Future studies will hopefully provide further guidance on which goals apply to which patient populations and which protocols are most safe and effective.
Pathophysiology
Even short-term hyperglycemia may be hazardous in the context of medical illness. Hyperglycemia is associated with immune dysfunction, specifically neutrophil impairment. The secondary metabolic perturbations associated with hyperglycemia, such as overproduction of free fatty acids, reactive oxygen species, ketones, and lactic acid, may lead to cellular injury, inflammation, and defective wound repair. These adverse effects are likely to be most evident in patients with infections (eg, pneumonia), wounds (eg, sternotomy), and infarction or ischemia (eg, myocardial infarction, stroke), exactly the populations in which observational data linking hyperglycemia to increased mortality are the strongest. Hyperglycemia may also lead to fluid and electrolyte imbalances once thresholds for glucosuria are reached (generally around 200 mg/dL). These can be potentially catastrophic in patients already suffering from acute renal failure, hypernatremia, and lactic acidosis. In addition preventing to these short-term problems, good inpatient glucose management may lead to better glucose control after discharge, for example, triggering a change to insulin use in a patient poorly controlled on maximal doses of oral agents. Measures to improve outpatient glucose control can have sustained effects on long-term diabetic complications such as neuropathy, nephropathy, retinopathy, and cardiovascular disease.
Diagnosis
Hyperglcyemia is not usually the primary reason for hospitalization in diabetic patients. When it is, it is usually clinically obvious, as in diabetic ketoacidosis (DKA) in patients with type 1 diabetes, and hyperglycemic hyperosmolar state (HHS, formerly known as hyperosmolar nonketotic coma) in patients with type 2 diabetes. Diagnostic errors that may occur in the management of patients with diabetes and hyperglycemia include the following.
Failure to recognize hyperglycemia in a patient without a previous diagnosis of diabetes (eg, failure to notice a routine laboratory glucose of 200 mg/dL as abnormal). Any routine glucose reading over 180 mg/dL should trigger further evaluation and monitoring, including at least 24 hours of point-of-care (POC) glucose testing before meals and a hemoglobin A1c test. The hemoglobin A1c may be useful in the diagnosis of type 2 diabetes and helps assess the prior 90 days of glucose control, enabling the provider to distinguish diabetes from stress hyperglycemia. It is also extremely useful for discharge planning.
Failure to recognize type 1 diabetes, as opposed to type 2 diabetes. While the diagnosis may be obvious in a young, thin patient who presents for the first time in DKA, it should also be considered in the patient with chronic pancreatitis and pancreatic insufficiency, the patient who has had complete or partial pancreatectomy for malignancy, and other situations in which beta cell failure may be the predominant problem. These patients must receive insulin at all times, are more insulin sensitive, and typically have more labile glucose control due to the lack of pancreatic reserve. Many of these patients would benefit from endocrinology input for both diagnosis and management.
Glucose monitoring should be obtained in any patient with a history of diabetes, inpatient hyperglycemia, or receiving treatment that might cause hyperglycemia (such as systemic corticosteroids). In the setting of corticosteorids, monitoring can be stopped after 24 hours if hyperglycemia is not observed and increases in glucocorticoid doses are not planned. Patients eating discrete meals should generally be tested before each meal and before bedtime, while patients who are receiving nothing by mouth (NPO), receiving continuous enteral nutrition (tube feeds), or total parenteral nutrition (TPN) should be tested every six hours. Additional testing is optional depending on the circumstance. For example, testing at 3 AM