Chapter 78. Diabetic Ketoacidosis, Hyperglycemic Hyperosmolar Nonketotic Coma, and Hypoglycemia

• Diabetic ketoacidosis (DKA), hyperglycemic hyperosmolar nonketotic coma (HHNC), and hypoglycemia are life-threatening disorders of glucose metabolism.
• Altered glucose metabolism should be considered in the differential diagnosis of all patients with mental status changes, neurologic deficits, and severe illness.
• Therapy of DKA and HHNC requires replacement of deficits of fluids, electrolytes, and insulin.
• Mixed-anion-gap acidosis and hyperosmolarity are encountered often in the same patient.
• Continuous insulin therapy is essential in DKA.
• In DKA and HHNC, careful physical examination and laboratory follow-up with a flow sheet make it possible to prevent the disastrous consequences of aggressive therapy—cerebral edema, pulmonary edema, hypoglycemia, hypokalemia, and hyperchloremic metabolic acidosis.
• A search for the underlying cause of metabolic decompensation is required.
• Therapy of hypoglycemia requires immediate establishment and maintenance of modest hyperglycemia (glucose concentration above 100 mg/dL or 5.5 mmol/L).

Diabetes mellitus and its major life-threatening complications—diabetic ketoacidosis (DKA), hyperglycemic hyperosmolar nonketotic coma (HHNC), and hypoglycemia—may precipitate or complicate critical illness. This chapter will focus on the management of DKA because it has been studied more exhaustively and because the same principles are applicable to HHNC. A brief review of the physiology and therapy of hypoglycemia will follow. These conditions have been the subject of other reviews.^{ch78ref1–9a},9b,9c

Currently, most patients with diabetes can be classified as having either type 1 diabetes (previously known as juvenile-onset, ketosis-prone, insulin-dependent, or brittle diabetes) or type 2 diabetes (previously, adult-onset, non-ketosis-prone, non-insulin-dependent, or obesity-related diabetes). Type 1 diabetes results from the autoimmune destruction of the insulin-secreting β cells in the pancreas, with consequent insulin deficiency. Most persons with type 1 diabetes present in childhood or early adulthood, although well-documented cases also have been described in the geriatric population. Hyperglycemia is usually not present until approximately 90% of insulin secretory capacity is lost. In times of stress (either physiologic or emotional), however, insulin requirements increase and often precipitate metabolic decompensation because of inadequate insulin secretory reserve. Patients with insulin deficiency require insulin therapy for survival. The half-life of insulin in the circulation is 7 minutes, and therapy for insulin deficiency requires continuous delivery of insulin to the circulation by infusion or by diffusion of insulin from depot injections.

The pathophysiology of type 2 diabetes is more heterogeneous. At presentation, persons with type 2 diabetes exhibit both tissue resistance to the action of insulin and a relative deficiency of insulin secretion. Insulin resistance can be overcome only by (1) increasing circulating insulin concentrations, with insulin (often at high doses) or sulfonylureas, (2) imposing calorie restriction, (3) instituting exercise, (4) administering certain antidiabetic medications (metformin and thiazolidine diones), or (5) treating reversible causes of the disease (e.g., obesity; hyperglycemia; sepsis; uremia; catecholamine; excess levels of glucocorticoids, growth hormone, or thyroxine; and antibodies to insulin or to the insulin receptor). Most persons with type 2 diabetes present in middle age and are overweight and sedentary. There is a growing prevalence of the disorder in children and adolescents, as well as a less common familial adolescent form of the disease termed maturity-onset diabetes of youth (MODY).

DKA is the life-threatening metabolic consequence of insulin deficiency. Insulin deficiency, in concert with excess secretion of glucagon (primarily), catecholamines, glucocorticoids, and growth hormone, produces hyperglycemia by stimulating glycogenolysis and gluconeogenesis and impairing glucose disposal. As has been elegantly demonstrated by Foster and McGarry,10 this hormonal milieu also results in lipolysis and unrestrained fatty acid oxidation, producing acetone, β-hydroxybutyrate, and acetoacetate—and thus ketoacidosis.

HHNC is the most severe presentation of a syndrome in which insulin resistance and relative insulin deficiency produce hyperglycemia. The glucose remains largely in the extracellular space and shifts water osmotically from the intracellular compartment. Glucose, water, and salts are filtered at the glomerulus; because the reabsorptive capacity of the renal tubules for glucose has a maximum of approximately 200 mg/min, an osmotic diuresis occurs, with water lost in excess of salts. A vicious cycle of cellular dehydration and diuresis is produced that can only be compensated by adequate fluid intake. With inadequate fluid intake, hyperosmolarity and hypovolemia develop, further increasing insulin resistance and hyperglycemia.

Patients rarely fit purely insulin-deficient or insulin-resistant physiology, particularly when in extremis. Patients with type 1 diabetes whose disease is not well controlled or who are under stress are insulin-resistant. Patients with poorly controlled type 2 diabetes may become insulin-deficient through a process of β-cell exhaustion and present with DKA. In fact, when careful studies of insulin levels, secretory dynamics, and tissue sensitivity are performed, there is considerable overlap between patients with DKA and patients with HHNC. One of the keys to the management of these patients is recognizing the contributions of insulin deficiency and resistance and optimizing each patient's individual therapy.

Early in their course, patients with uncontrolled diabetes present with nonspecific complaints. If the disease follows an indolent course over months to years (more common for type 2 than for type 1 diabetes), patients can present with profound wasting, cachexia, and prostration similar in degree to that of patients with long-standing malignancy or chronic infection. With significant physical or emotional stress, sudden metabolic decompensation can occur.10a The cases of DKA and HHNC that are missed usually occur in patients with new-onset diabetes. All patients with nonspecific complaints should be questioned about more specific symptoms of diabetes. Polyuria (or at least nocturia) and weight loss are almost always present, although often not reported by the patient. Any patient with severe illness (acute or chronic) or neurologic changes should have glucose and electrolyte levels measured.

History and Physical Examination

DKA and HHNC are easily recognized, when considered (Table 78-1). In DKA, metabolic decompensation develops over a period of hours to a few days. Patients with severe DKA classically present with lethargy and a characteristic hyperventilation pattern with deep, slow breaths (Kussmaul respiration) associated with the fruity odor of acetone. They often complain of nausea and vomiting; abdominal pain is somewhat less frequent. The abdominal pain can be quite severe and may be associated with distention, ileus, and tenderness without rebound, but it usually resolves relatively quickly with therapy unless there is underlying abdominal pathology. Most patients are normotensive, tachycardic, and tachypneic, with signs of mild to moderate volume depletion. Hypothermia has been described in DKA, and patients with underlying infection may not manifest fever. Cerebral edema does occur.

Patients with HHNC are usually stuporous, with obvious profound dehydration, and may demonstrate focal neurologic deficits, such as Babinski reflexes, asymmetric reflexes, cranial nerve findings, paresis, fasciculations, and aphasia. The syndrome evolves over days to weeks with a progressive decrease in fluid intake and decline in mental status; usually there is no prior history of diabetes. These patients are hypotensive and exhibit normal to depressed ventilation. Hypothermia, again, is common, although cerebral edema is rare.

Laboratory Tests and Differential Diagnosis

Once DKA or HHNC is considered, the diagnosis can be made quickly with routine laboratory tests. DKA and HHNC need to be differentiated from each other and from the other causes of ketosis and metabolic acidosis. This is often quite difficult; in fact, these disorders often coexist. Table 78-2 and the following discussion are provided to help in making the distinction. Blood and urine glucose and ketone levels can be obtained in minutes using glucose oxidase–impregnated strips and the nitroprusside reaction, respectively.

The sine qua non of HHNC is hyperosmolarity. The osmolarity can be measured by freezing-point depression, or it can be estimated using the following formula:

where the concentrations of glucose, blood urea nitrogen (BUN), and ethanol are in milligrams per deciliter, and the concentration of sodium is in milliequivalents per liter. In HHNC, the osmolarity generally is greater than 350 mOsm/L and can exceed 400 mOsm/L. The serum sodium and potassium levels can be high, normal, or low and do not reflect total body levels, which are uniformly depleted. The glucose concentration is usually greater than 600 mg/dL (33.3 mmol/L), and levels over 1000 mg/dL (55.6 mmol/L) are quite common. In pure HHNC, there is usually no significant metabolic acidosis or anion gap. However, severe anion-gap metabolic acidosis of uncertain etiology has been reported in HHNC.11

The sine qua non of DKA is ketoacidosis. As mentioned earlier, two ketoacids are produced in DKA—β-hydroxybutyrate and acetoacetate—as well as the neutral ketone acetone. The nitroprusside reaction commonly used to detect ketone bodies detects acetoacetate much better than acetone and does not react with β-hydroxybutyrate at all. Particularly in severe DKA, β-hydroxybutyrate is the predominant ketone, and it is possible, though unusual, to have a negative serum nitroprusside reaction in the face of severe ketosis. β-Hydroxybutyrate levels can be measured at many centers and commercially, but this service usually is not readily available. Fortunately, there is a readily available index for unmeasured anions in the blood—namely, the anion gap (normally less than 14 mEq/L):

Most persons with DKA present with an anion gap greater than 20 mEq/L, and some have an anion gap greater than 40 mEq/L. However, some patients have a hyperchloremic metabolic acidosis without a significant anion gap.12 Patients with DKA almost invariably have large amounts of ketones in their urine. The serum glucose level in DKA is usually about 500 mg/dL (27.8 mmol/L). However, an entity known as euglycemic DKA has been described, particularly in the presence of decreased oral intake or pregnancy, where the serum glucose level is normal or near normal, but the patient requires insulin therapy for the clearance of ketoacidosis.13 The arterial pH is commonly less than 7.3 and can be as low as 6.5. There is partial respiratory compensation with hypocapnia. Patients are often mildly hyperosmolar, although osmolalities greater than 330 mOsm/kg are unusual without mental status changes.

Not all patients with hyperglycemia and an anion gap metabolic acidosis will have DKA. Lactic acidosis is the most common cause of metabolic acidosis in hospitalized patients and can be seen in uncomplicated diabetes, as well as in DKA and HHNC. Lactic acidosis usually occurs in the setting of decreased tissue oxygen delivery, resulting in the nonoxidative metabolism of glucose to lactic acid. Lactic acidosis complicates other primary metabolic acidoses as a consequence of dehydration or shock. Thus, assessing its relative contribution can be difficult. The presentation is identical to that seen in DKA. In pure lactic acidosis, the serum glucose and ketone levels should be normal, and the serum lactate concentration should be greater than 5 mM. The therapy of lactic acidosis is directed at the underlying cause and at optimizing tissue perfusion.14

Starvation ketosis results from inadequate carbohydrate availability, resulting in physiologically appropriate lipolysis and ketone production to provide fuel substrates for muscle. The blood glucose level usually is normal. Although the urine can have large amounts of ketones, the blood rarely does. Usually the arterial pH is normal, and the anion gap is at most mildly elevated.

Alcoholic ketoacidosis is a more severe form of starvation ketosis wherein the appropriate ketogenic response to poor carbohydrate intake is increased through as yet poorly defined effects of alcohol on the liver. Classically, these patients are long-standing alcoholics for whom ethanol has been the main caloric source for days to weeks. The ketoacidosis manifests itself when, for whatever reason, the intake of alcohol and calories decreases. In isolated alcoholic ketoacidosis, the metabolic acidosis is usually mild to moderate in severity. The anion gap is elevated. Serum and urine ketones are always present. However, alcoholic ketoacidosis produces an even higher ratio of β-hydroxybutyrate to acetoacetate than does DKA, and negative or weakly positive serum nitroprusside reactions are common. Respiratory alkalosis associated with delirium tremens, agitation, or pulmonary processes often normalizes the pH but the underlying acidosis should be evident with careful analysis of acid-base status. Usually the patient is normoglycemic or hypoglycemic; sometimes mild hyperglycemia is present. Patients who are significantly hyperglycemic should be treated as if they had DKA. The therapy of alcoholic ketoacidosis consists of administration of thiamine, carbohydrates, fluids, and electrolytes, with special attention to the more severe consequences of alcohol toxicity, alcohol withdrawal, and chronic malnutrition. In more severely ill patients in whom alcoholic ketoacidosis is considered a possibility, there is usually another underlying illness, such as pancreatitis, gastrointestinal bleeding, hepatic encephalopathy, delirium tremens, or infection complicated by coexisting lactic acidosis.15

Uremic acidosis is characterized by very large elevations in the BUN level and the creatinine level with normoglycemia. The pH and anion gap are usually only mildly abnormal. The treatment is supportive, with careful attention to fluid and electrolytes until dialysis can be performed. Rhabdomyolysis is a cause of renal failure in which the anion gap can be elevated significantly and acidosis can be severe. There should be marked elevation of creatine phosphokinase (CPK) and myoglobin levels. It should be noted that mild rhabdomyolysis is not uncommon in DKA, but the presence of hyperglycemia and ketonemia leaves no doubt as to the primary etiology of the acidosis.16

Toxic ingestions can be differentiated from DKA by history and laboratory investigation. Salicylate intoxication can produce an anion-gap metabolic acidosis, usually with a respiratory alkalosis. The plasma glucose level is normal or low and the osmolality normal, and salicylates are detected in the urine or blood. Salicylate uncouples oxidative phosphorylation; consequently, ketonemia and lactic acidosis develop. It should be noted that salicylates can cause a false-positive glucose determination when the cupric sulfate method is used and a false-negative result when the glucose oxidase reaction is used. Methanol and ethylene glycol also produce an anion gap metabolic acidosis without hyperglycemia or ketones; they need to be kept in mind primarily because they produce an increase in the measured serum osmolality but not in the calculated serum osmolality—an osmolar gap. Their serum levels also can be measured. Isopropyl alcohol does not cause a metabolic acidosis, but it needs to be borne in mind because it is metabolized to acetone, which can produce a positive result on the nitroprusside reaction commonly used for the detection of ketoacids. Therapy of these intoxications is discussed elsewhere17–19 (see Chap. 102). Rare cases of anion gap acidoses have been reported with other ingestions, including toluene, iron, hydrogen sulfide, nalidixic acid, papaverine, paraldehyde, strychnine, and isoniazid.

It should be remembered that patients often present with combinations of the preceding conditions. As already noted, DKA not uncommonly appears with hyperosmolarity and coma. HHNC can have mild to moderate ketonemia and acidosis. Alcoholic ketoacidosis can contribute to either DKA or HHNC, and lactic acidosis is common in severe DKA and HHNC. We believe that any patient with hyperglycemia above 250 mg/dL (13.9 mmol/L) and an anion gap metabolic acidosis should be treated according to the general principles outlined below, with special consideration for possible other contributing metabolic acidoses (see Chaps. 77 and 102).

The optimal management of DKA and HHNC has been the object of considerable controversy over the past half century. The guidelines we propose rely heavily on prospective studies of DKA treatment by Kitabchi and coworkers1 and are readily applicable to HHNC. The general approach (Table 78-3) is (1) to provide necessary fluids to restore the circulation, (2) to treat insulin deficiency with continuous insulin, (3) to treat electrolyte disturbances, (4) to follow the patient closely and carefully, and (5) to search for underlying causes of metabolic decompensation.

Fluids

Volume contraction is one of the hallmarks of DKA and HHNC. It can contribute to acidosis via lactic acid production, as well as decreased renal clearance of organic and inorganic acids. It contributes to hyperglycemia by decreasing renal clearance of glucose. If decreased tissue perfusion is significant, it causes insulin resistance by decreasing the delivery of insulin to the sites of insulin-mediated glucose disposal—muscle and adipose tissue—as well as by stimulating catecholamine and glucocorticoid secretion. Fluid deficits on the order of 5 to 10 L are common in DKA and HHNC. It should be remembered that the urine produced during the osmotic diuresis of hyperglycemia is approximately half normal with respect to sodium content. Therefore, water deficits are in excess of sodium deficits. Historically, large quantities of isotonic intravenous fluids have been administered rapidly to patients in DKA or HHNC. We tend to be more cautious; for patients with a history of congestive heart failure, chronic or acute renal failure, severe hypotension, or significant pulmonary disease, we give early and frequent consideration to the use of invasive hemodynamic montoring.

When there is physical evidence of dehydration—hypotension, decreased skin turgor, or dry mucous membranes—we generally administer 1 L of normal saline solution over the first few minutes and 200 to 500 mL/h in subsequent hours, until hypotension resolves and an adequate circulation is maintained. If hypotension is severe, with clinical evidence of hypoperfusion, and does not respond to crystalloid, then therapy with colloid is considered, often in combination with invasive hemodynamic monitoring. If there is no hypotension and no concern of renal failure, we administer 1 L of half-normal saline solution over the first hour and 100 to 500 mL/h in subsequent hours.

During that first hour, the results of initial laboratory tests usually are obtained and can be quite helpful in planning further therapy. Despite the excess of water losses over sodium, the measured sodium level is usually low because of the osmotic effects of glucose. These osmotic effects can be corrected using the following simple formula:

Severe hypertriglyceridemia, which is common in severe diabetes, can cause a false decrease in the serum sodium concentration by approximately 1.0 mEq/L at a serum lipid concentration of 460 mg/dL.20 An estimated water deficit21 can be calculated using the corrected sodium concentration:

Using these formulas, a 70-kg patient with a measured sodium level of 140 mEq/L and a glucose level of 1000 mg/dL would have a calculated water deficit of 4.3 L. If the patient is normotensive after the first liter of fluids, we provide generous maintenance fluids while the patient is NPO, replace urinary losses with one-half normal saline solution, and in addition, provide approximately one-half the calculated water deficit as 5% dextrose over the first 12 to 24 hours (using the preceding example, the amount would be 2 L) and the remainder over the subsequent 24 hours. The plan for fluid therapy should be reevaluated continuously in light of the clinical and laboratory response of the patient. Once the serum glucose level reaches 250 mg/dL (13.9 mmol/L), fluids should contain 5% dextrose, and therapy should be aimed at maintaining the serum glucose concentration in the 150 to 250 mg/dL (8.3–13.9 mmol/L) range for 24 hours to allow slow equilibration of osmotically active substances across cell membranes. At least 100 g of dextrose per 24 hours (at least 80 mL/h of 5% dextrose if the patient is NPO) must be given to prevent ketogenesis. In late pregnancy, approximately 100% more carbohydrate may be needed, so 10% dextrose should be used.

The primary goal of fluid therapy is to maintain adequate circulation, and the secondary goal is to maintain brisk diuresis. Beyond that, pulmonary edema, hyperchloremic metabolic acidosis, and a rapid fall in the serum osmolality should be avoided by frequent monitoring of the patient, the glucose level, and the electrolyte levels. It has been demonstrated that fluid administration and subsequent continued osmotic diuresis are responsible for a large portion of the initial decline in glucose during therapy. In general, persons with HHNC present with more profound dehydration and require more intravenous fluids than persons with DKA.

Insulin

Insulin is the mainstay of therapy of DKA, which is essentially an insulin-deficient state. In the past, high doses of insulin (upwards of 50 units/h) were favored. In more recent studies, low-dose insulin therapy (0.1 unit/kg per hour) has been shown to be equally effective in producing a decrease in serum glucose concentration and effecting clearance of ketones. Furthermore, low-dose therapy results in a dramatic reduction in the major morbidity of intensive insulin therapy—hypoglycemia and hypokalemia.

Studies have also shown that intravenous (IV) insulin is significantly more effective than intramuscular (IM) or subcutaneous (SC) insulin in lowering ketone body concentration over the first 2 hours of therapy. The SC route is inappropriate for critically ill patients because of the possibility of tissue hypoperfusion and the slower kinetics of absorption. Numerous studies attest to the efficacy of IM therapy in severe DKA. In cases where there is insufficient nursing monitoring or IV access to allow for safe IV administration, hourly IM injections of regular insulin is an acceptable substitute. It is important that a needle of adequate length—at least 1.5 inches for adults—be used.

Last, it has been shown that the administration of a 10-unit IV priming dose of insulin, when a patient is being started on insulin therapy (whether IV or IM), significantly improves the glycemic response to the first hour of therapy. The rationale is to fully saturate insulin receptors before continuous therapy is started and thus to avoid the lag time necessary to achieve steady-state insulin levels. It seems to be possible to mix insulin into normal saline solution without adding albumin to prevent adsorption of insulin onto the infusion set. However, the IV tubing should be flushed with the insulin infusate prior to use.

In those rare instances where the glucose level does not decrease by at least 10%, or 50 mg/dL (2.8 mmol/L), in 1 hour, the insulin infusion rate should be increased, and a second bolus of IV insulin should be administered. As the glucose level decreases, it is usually necessary to decrease the rate of infusion. After the glucose level reaches approximately 250 mg/dL (13.9 mmol/L), it is prudent to decrease the insulin infusion rate and to administer dextrose as described under “Fluids” above. It is a common mistake to discontinue IV insulin in response to a falling glucose level. Because of its short half-life, IV insulin should not be discontinued until after SC insulin is initiated. With resolution of ketosis, the rate of insulin administration approaches the physiologic range of 0.5 to 1.0 units/kg per day.

If the patient is not nauseated, early feeding should be considered because it allows more carbohydrates and insulin to be administered, which accelerates the clearance of ketones. When the decision is made to feed the patient, the patient should be switched from IV or IM therapy to SC therapy. SC insulin should be administered before a meal, and the insulin drip should be discontinued approximately 30 minutes later if regular insulin is used, approximately 15 minutes later if a rapid-acting insulin analogue is used, and in 2 hours if only intermediate or long-acting insulin is used. The glucose level should be checked in 2 hours and at least every 4 hours subsequently, until a relatively stable insulin regimen is determined. Early conversion to oral feeding and SC insulin therapy has been associated with shorter length of hospital stay. A proportion of patients with HHNC do not require therapy for their diabetes after hospitalization, and most do not require insulin.

Potassium

Potassium losses during the development of DKA and HHNC are usually quite high (3 to 10 mEq/kg) and are mediated by protein catabolism compounded by hyperaldosteronism and osmotic diuresis. Most persons with DKA or HHNC have normal or even high serum potassium levels at presentation, but the initial therapy with fluids and insulin will cause the serum potassium level to fall. Our approach has been to monitor the electrocardiogram (ECG) for signs of hyperkalemia (peaked T wave, QRS widening) initially and to administer potassium if these signs are absent and the serum potassium level is less than 5.5 mEq/L. If the patient is oliguric, we do not administer potassium unless the serum concentration is less than 4 mEq/L or if there are ECG signs of hypokalemia (U wave)—and even then, only with extreme caution. With therapy of DKA, the potassium level always falls, usually reaching a nadir after several hours. We usually replace potassium at 10 to 20 mEq/h, two-thirds as potassium chloride and one-third as potassium phosphate, and monitor serum levels at least every 2 hours initially, as well as following ECG morphology. Some patients with DKA who have had a protracted course including vomiting present with hypokalemia and acidosis; they may require 20 to 40 mEq/h or more by central line to avoid further decreases in the serum potassium concentration; consideration in these rare cases of delaying the initiation of insulin infusion until potassium concentrations are greater than 3.3 mEq/L may be prudent to avoid precipitating a further decline in serum potassium concentration.

Phosphate

Like potassium, phosphate is depleted in patients with DKA and HHNC. Although patients usually present with an elevated serum phosphate level, the serum level declines with therapy. Although no well-documented clinical benefit of phosphate administration has been demonstrated, most authorities recommend phosphate therapy as above and monitoring for its possible complications—hypocalcemia and hypomagnesemia.

Bicarbonate

The serum bicarbonate level is always low in DKA, but no true deficiency is present because the ketoacid and lactate anions are metabolized to bicarbonate during therapy. The use of bicarbonate in the therapy of DKA is highly controversial; no benefit of bicarbonate therapy has been demonstrated in clinical trials. In fact, in two trials, hypokalemia was more common in bicarbonate-treated patients. There are theoretical considerations against the use of bicarbonate. Cellular levels of 2,3-diphosphoglycerate are depleted in DKA, causing a leftward shift in the oxyhemoglobin dissociation curve and thus impairing tissue oxygen extraction. Acidemia has the opposite effect; therefore, acute reversal of acidosis might decrease tissue oxygen extraction. In addition, there are in vitro data to suggest that pH is a regulator of cellular lactate metabolism and that correction of acidosis could increase lactate production. Most important, bicarbonate administration acutely elevates PCO2, which may lower intracellular pH. We do not administer bicarbonate routinely, but we understand that other authors advocate its use. Bicarbonate therapy should be restricted to (1) patients with severe acidosis (pH below 6.9), (2) patients with hemodynamic instability in whom the pH is less than 7.1, and (3) patients with hyperkalemia and ECG findings. When bicarbonate is used, it should be used sparingly as a temporizing measure while definitive therapy with insulin and fluids are under way. Approximately 1 mEq/kg of bicarbonate is administered as a rapid infusion over 10 to 15 minutes, with further therapy based on repeat measurement of arterial blood gases every 30 to 120 minutes. Potassium therapy should be considered prior to treatment with bicarbonate because transient hypokalemia is a not uncommon complication of the administration of alkali.

Monitoring

It is possible to manage many cases of mild DKA without ICU admission, depending on staff availability. We routinely admit patients with DKA to the ICU if they have a pH below 7.2. If mental status is compromised, then prophylactic intubation is considered, and nasogastric suctioning is always performed because of frequent ileus and danger of aspiration. Only if the patient cannot void at will is bladder catheterization necessary to follow urine output adequately. ECG monitoring is continuous, with hourly documentation of QRS intervals as well as T-wave morphology. Initially, the levels of serum glucose, electrolytes, BUN, creatinine, calcium, magnesium, phosphate, ketone, lactate, and CPK are measured; liver function tests, urinalysis, ECG, upright chest x-ray, and a complete blood count (CBC) are performed; and arterial blood gases are measured. If there is any concern about possible toxic ingestions, toxicology screening is also performed. Subsequently, glucose and electrolyte levels are measured at least hourly; calcium, magnesium, and phosphate levels are measured every 2 hours; and the levels of BUN, creatinine, and ketones are measured every 6 to 24 hours. This frequency of laboratory analysis generally can be tapered as trends in improvement become clear; glucose monitoring using bedside point-of-care systems can be substituted for laboratory analysis of glucose if appropriate quality assurance and documentation programs are in place once glucose levels are in the 400 to 600 mg/dL (22–33 mmol/L) range. The urine ketone concentration is a more sensitive indicator of ongoing or recurrent hepatic ketogenesis than is the level of serum ketones. It is usually not necessary to routinely monitor arterial blood gases because bicarbonate and the anion gap are relatively good indices of the response to therapy. Monitoring venous pH also has been shown to adequately reflect acidemia and response to therapy. In severely ill patients with obvious underlying disease, the course is often more protracted; when venous access is a problem, early consideration should be given to placement of an arterial line. A flow sheet tabulating these findings, as well as mental status, vital signs, insulin dose, fluid and electrolytes administered, and urine output, allows easy analysis of response to therapy. Once the acidosis begins to resolve and the response to therapy becomes predictable, it is reasonable to curtail laboratory use. The goal should be to rapidly achieve hemodynamic stability and to fully correct DKA in 12 to 36 hours.

The Search for Underlying Causes

After the patient is stabilized, a careful history and physical examination and a diagnostic strategy should be aimed at determining the precipitating event. In our practice, the most common cause of DKA is poor compliance with insulin therapy or inadequate outpatient insulin regimens; these causes usually are treated easily. The second most common cause is infection, with urinary tract infection, pelvic inflammatory disease, and respiratory tract infections predominating. It is often difficult to determine initially if the patient is infected. Fever is absent in a significant proportion of patients with diabetic emergencies. The white blood cell count (WBC) is not uncommonly elevated to the range of 20,000/μL or higher, even in the absence of infection.21 We therefore perform cultures on most patients with DKA or HHNC, and we often treat empirically with antibiotics while awaiting the results of bacterial cultures. Meningitis should be considered in any patient with altered mental status. In this regard, our approach has been to perform lumbar punctures on all patients with meningismus and on patients in whom the change in mental status does not appear to be related to the metabolic insult. The cerebrospinal fluid (CSF) glucose level is not particularly useful in the diagnosis of meningitis in the setting of DKA or HHNC, but a CSF glucose level of less than 100 mg/dL (5.5 mmol/L) is unusual when the serum glucose level is greater than 250 mg/dL (13.9 mmol/L).22 The relative frequency of sinus infection (the most serious being mucormycosis), foot infection, bacterial arthritis, cholecystitis, cellulitis, and necrotizing fasciitis should be considered. Pneumonia can be difficult to diagnose in patients with dehydration because the alveolar edema fluid, which causes infiltrates on chest x-ray, often is not present but develops along with progressive hypoxia during hydration. To avoid this occurrence, we are particularly judicious with fluid administration in patients we suspect of having pneumonia. Pancreatitis and pregnancy are common precipitants and need to be specially considered in the assessment of abdominal pain, which is extremely common at presentation. Abdominal guarding and tenderness associated with vomiting are seen often. Rebound is present only rarely. These symptoms and findings usually resolve quickly with therapy in the absence of intraabdominal pathology. The serum amylase level is often elevated without pathologic significance, but a high lipase level is usually more specific for a diagnosis of pancreatitis.23 Acute myocardial infarction and stroke, as well as thromboembolic phenomena, are frequent precipitants and complications of DKA.

Precipitating factors are similar in HHNC, but they are found more commonly and are often more severe. Common associated illnesses are chronic renal insufficiency, pneumonia, gastrointestinal hemorrhage, sepsis (particularly with gram-negative organisms), myocardial infarction, pulmonary embolism, and intracranial events. Burns, hyperalimentation, dialysis, heat stroke, thyrotoxicosis, acromegaly, pancreatitis, and a variety of drugs (mostly diuretics, steroids, and β-blockers) also have been associated.

The more insulin-resistant the patient, the more likely it is that a precipitating cause is present. If a precipitating cause is found, treatment is essential if adequate metabolic control is to be achieved.

Complications and Prognosis

It should be possible to treat almost all cases of DKA successfully. The most troublesome complication is cerebral edema. It is particularly common in children and can be fatal. In most series, specific causes cannot be assigned, although the use of bicarbonate, more severe DKA, and aggressive hydration, particularly with hypotonic fluids, may contribute.24 In 50% of patients who subsequently have a respiratory arrest, there are premonitory symptoms; despite early intervention, only half of such patients avoid severe or fatal brain damage. Cerebral edema is much less common in HHNC but may be associated with decreases in plasma glucose concentration below 250 mg/dL (13.9 mmol/L) during the first day of therapy.25 Other complications of life-threatening severity that have been reported include the acute respiratory distress syndrome (ARDS) and bronchial mucus plugging.26–28

The most common complication of HHNC is death. Female gender, lack of prior knowledge of diabetes, and acute infections are risk factors for the development of HHNC. Despite aggressive therapy, mortality is in the range of 20% to 50% in most recent case series. Prognosis is inversely related to serum osmolality, BUN level, and sodium level, although large differences in values are not apparent between survivors and nonsurvivors.29 The mortality seems to be in large part a function of the underlying illness and the duration of illness before appropriate medical therapy is instituted. Although complete recovery of neurologic function in survivors is the rule, persistent coma is not infrequent. Arterial and venous thromboembolic events are quite common. Standard prophylactic low-dose heparin administration is certainly a reasonable precaution for patients with HHNC, but currently no indication exists for full anticoagulation.

Recurrent admissions for DKA occur in up to one-third of patients. Ensuring adequate and timely outpatient follow-up and diabetes education will reduce relapses. Patients with frequent DKA are rare and require special consideration of neuropsychological contributors.30

Although not a complication of diabetes per se, severe hypoglycemia is a not uncommon complication of its therapy. It also may be a major complication of critical illness—particularly hepatic failure, drug overdose, and sepsis. The serum glucose concentration is very strictly regulated in normal persons, being maintained between approximately 60 and 140 mg/dL (3.3–7.8 mmol/L). The differential diagnosis of hypoglycemia can be divided into fasting and postprandial forms (Table 78-4). Fasting hypoglycemia can result from endocrine disorders, hepatic dysfunction, antibodies to insulin or the insulin receptor, or substrate deficiency. The major endocrine causes are excess secretion of insulin or insulin-like growth factor and deficiency of growth hormone or cortisol. Insulin hypersecretion can be from an insulinoma or from normal islet tissue under pharmacologic stimulation. Certain malignancies, particularly sarcomas, hepatomas, and tumors of the biliary tract, can cause elevated levels of free insulin-like growth factors and can present with hypoglycemia, often in the presence of cachexia. Any cause of severe hepatic dysfunction—toxins, viruses, congestive heart failure, or shock—and enzymatic defects causing decreased gluconeogenesis or glycogen storage diseases can cause hypoglycemia. Hypoglycemia can occur with fasting in alcoholism, pregnancy, or uremia due to substrate unavailability, but otherwise, it occurs only with end-stage protein-calorie malnutrition. Postprandial hypoglycemia can be caused by rapid gastric emptying owing to postoperative changes, or it can be idiopathic. It has been associated with early type 2 diabetes. Drugs that can cause hypoglycemia include alcohol, insulin, sulfonylureas, salicylates, β-blockers, pentamidine, some antiarrhythmic agents, akee fruit, and the rat poison Vacor.

The symptoms of hypoglycemia initially result from sympathoadrenal responses and consist of palpitations, anxiety, tremulousness, sweating, and hunger. If hypoglycemia progresses, neuroglycopenic symptoms develop. These include headache, nightmares, fatigue, confusion, bizarre behavior, hallucinations, seizures, coma, and focal neurologic deficits. The absolute glucose concentration at which these symptoms develop is highly variable; in part it depends on the rate of fall in glucose concentration. In critically ill patients, symptoms may be obscured by concurrent problems until profound hypoglycemia, with seizures or coma, evolves.

Patients who present with signs and symptoms compatible with hypoglycemia should have their blood glucose concentration measured and glucose administered—by mouth if conscious, intravenously otherwise. The disappearance of symptoms over a period of minutes suggests that hypoglycemia is the likely etiology. IV administration of 50 mL of 50% dextrose will awaken more than 80% of hypoglycemic patients destined for neurologic recovery.

Patients with hypoglycemia should be fully evaluated to determine etiology. A complete history, physical examination, and routine laboratory tests will result in a likely diagnosis in most cases. The various studies that can be used in less obvious cases should be considered only with the help of a specialist and are beyond the scope of this review. Glucose tolerance testing is usually nondiagnostic. A 72-hour fast will produce a diagnosis in the vast majority of patients with fasting hypoglycemia. At a minimum, glucose, insulin, C-peptide, proinsulin, and urine or serum drug screens for oral hypoglycemic agents should be obtained in both the normal and the hypoglycemic state.

Treatment

The most critical step in the therapy of hypoglycemia is recognition of its cause, which indicates its likely duration and severity (Table 78-5). If a patient took his or her insulin, forgot to eat breakfast, and subsequently lost consciousness, an ampule of 50% dextrose administered in the field and followed by breakfast almost certainly would be curative. However, if a patient drank a quart of liquor and then made a suicide gesture by taking long-acting oral hypoglycemic agents, aspirin, and β-blockers, then the hypoglycemia is likely to be severe and long-lasting and will require intensive therapy for days. Similarly, patients with hepatic failure will require continuous therapy with hypertonic glucose once hypoglycemia develops and until hepatic function returns. The goal of therapy is to establish and maintain modest hyperglycemia immediately and to monitor blood glucose concentration intensively using hand-held monitors. Seltzer9 has described a protocol for use in severe hypoglycemia as follows: In the presence of documented hypoglycemia, whether or not the patient responds to 50 mL of 50% dextrose, a continuous infusion of 10% glucose at a rate sufficient to keep the blood glucose level above 100 mg/dL (5.5 mmol/L) should be started. If more than 200 mL/h is necessary to maintain hyperglycemia, then 100 mg hydrocortisone and 1 mg glucagon are added to each liter of 10% dextrose for as long as necessary. Intravenous diazoxide administration to decrease insulin secretion also can be considered. It is important to remember that much more carbohydrate can be delivered enterally than intravenously, and so in extreme cases, jejunal feeding should be considered early. If the patient awakes, a high-carbohydrate diet should be initiated. Progressive hyperglycemia is the signal that therapy can be tapered. With early aggressive therapy, usually coma is reversed and focal deficits resolve over time.