INTRODUCTION — Insulinomas are rare tumors that may present a diagnostic dilemma for the clinician. The classic diagnostic test for an insulinoma has been the 72-hour fast. There are two reasons to perform a prolonged supervised fast:
●To confirm the presence of a hypoglycemic disorder (by demonstrating low venous glucose concentrations at the time of symptoms and resolution of these symptoms with reversal of hypoglycemia). In this setting, measurement of serum insulin, C-peptide, and proinsulin helps establish the mechanism of hypoglycemia [1,2].
●To determine the mechanism of hypoglycemia when the presence of a hypoglycemic disorder has already been established.
This distinction is important because the criteria for ending a fast will differ. If the presence of a hypoglycemic disorder is already established, a fast can be stopped when glucose is ≤55 mg/dL (<3 mmol/L) even in the absence of hypoglycemic symptoms. This is because at these concentrations, endogenous insulin secretion should be suppressed [3].
If a patient has a spontaneous episode of symptomatic hypoglycemia that is fortuitously observed and confirmed with laboratory testing, a 72-hour fast is not needed. The 72-hour fast is also unnecessary in the rare patient with insulinoma who has exclusively postprandial symptoms; in such cases, hypoglycemia is evaluated during a mixed-meal test.
The presence of inappropriately high serum insulin, C-peptide, and proinsulin concentrations at the time of symptomatic and confirmed hypoglycemia in a patient who has a negative test for insulin secretagogues (sulfonylureas or meglitinides) establishes the diagnosis of insulinoma.
We consider the following values (measured in highly sensitive assays) as inappropriately high in a symptomatic patient with serum glucose concentration ≤55 mg/dL (≤3 mmol/L) (table 1). (See "Hypoglycemia in adults without diabetes mellitus: Diagnostic approach".)
●Serum insulin – ≥3 microU/mL (immunochemiluminometric assay [ICMA])
●Serum C-peptide – ≥200 pmol/L
●Serum proinsulin – ≥5 pmol/L
When the serum insulin and C-peptide values are lower, determination of markers of insulin action such as serum beta-hydroxybutyrate and the serum glucose response to intravenous glucagon at the end of the fast may provide important diagnostic information [1]. In normal subjects who fast for 72 hours, serum beta-hydroxybutyrate concentrations rise, and liver glycogen stores are mobilized. In contrast, in patients with an insulinoma who continue to secrete insulin, serum beta-hydroxybutyrate concentrations remain low (2.7 mmol/L or less) and serum glucose concentrations increase by more than 25 mg/dL (1.4 mmol/L) 30 minutes after the intravenous administration of 1 mg glucagon. (See "Hypoglycemia in adults without diabetes mellitus: Diagnostic approach".)
The cases described below highlight some of the problems that may be encountered during the evaluation of patients suspected of having an insulinoma or in the management of those with a biochemically confirmed insulinoma.
CASE 1 — A 41-year-old woman was referred for evaluation of repeated episodes of sweating, slurred speech, tremulousness, and confusion during the last nine months that could be aborted by eating. On one occasion, she was unresponsive to questions and thrashing about in bed. Capillary glucose measured by emergency personnel was 31 mg/dL (1.7 mmol/L), and she improved after intravenous glucose administration.
After fasting for 24 hours, she became diaphoretic and confused. Serum values at that time were as follows:
●Glucose – 32 mg/dL (1.8 mmol/L)
●Insulin – 6.5 microU/mL (39 pmol/L)
●C-peptide – 533 pmol/L
●Proinsulin – 84 pmol/L
●Beta-hydroxybutyrate – 0.1 mmol/L
●Glucose increase after glucagon – 64 mg/dL (3.6 mmol/L)
●Sulfonylurea – Negative
Ultrasonography and spiral computed tomography (CT) of the abdomen showed a 1.5 cm lesion in the head of the pancreas. The location of the tumor was confirmed by intraoperative ultrasonography (see "Insulinoma"). The patient underwent enucleation of the insulinoma; pathologic examination confirmed the preoperative diagnosis of insulinoma.
Comment — This is a classic case of insulinoma. The patient was healthy but had episodes of neuroglycopenia. Whipple's triad (symptoms of hypoglycemia, low serum glucose concentrations at the same time, and relief of symptoms by glucose administration) was satisfied. That the hypoglycemia was caused by endogenous insulin was confirmed by the high serum insulin, C-peptide, and proinsulin concentrations and supported by the low serum beta-hydroxybutyrate concentration and the small rise in serum glucose after intravenous glucagon administration.
The tumor was detected by our preferred techniques: ultrasonography and spiral CT of the abdomen (see "Classification, epidemiology, clinical presentation, localization, and staging of pancreatic neuroendocrine neoplasms"). Confirmation of the location was established by intraoperative ultrasonography. In this instance, enucleation was the surgical procedure of choice.
CASE 2 — A 29-year-old man with a past history of primary hyperparathyroidism (single parathyroid adenoma) was referred for evaluation of hypoglycemia. During the last year, he had repeated episodes of tremor, sweating, and confusion, usually after physical exertion; the symptoms were relieved by eating. The patient's father and three paternal aunts had primary hyperparathyroidism; in addition, a distant cousin had hyperparathyroidism and a benign pancreatic tumor.
During his initial evaluation, he had symptoms of hypoglycemia, at which time the following serum values were obtained:
●Glucose – 25 mg/dL (1.4 mmol/L)
●Insulin – 100 microU/mL (600 pmol/L)
●C-peptide – 2800 pmol/L
●Proinsulin – 44 pmol/L
●Beta-hydroxybutyrate – 0.1 mmol/L
●Glucose increase after glucagon – 29 mg/dL (1.6 mmol/L)
●Sulfonylurea – Negative
Ultrasonography of the abdomen was uninterpretable. A spiral computed tomography (CT) of the abdomen showed a 3 cm mass in the head of the pancreas consistent with an insulinoma. Intraoperative ultrasonography confirmed the presence of a mass in this location and revealed multiple small nodules in the body and tail of the pancreas. He underwent enucleation of the dominant mass in the head of the pancreas and distal subtotal pancreatectomy. Pathologic examination of the resected pancreatic tissue revealed multiple islet cell adenomas.
Comment — This is a case of insulinoma in a patient with multiple endocrine neoplasia type 1 (MEN1). Although the majority of patients with insulinoma have a solitary adenoma, multicentric islet cell adenomas are the rule in patients with MEN1 [4,5]. (See "Multiple endocrine neoplasia type 1: Clinical manifestations and diagnosis".)
Because neither ultrasonography nor palpation at surgery may be sufficiently sensitive to detect all small tumors, resection of most of the distal pancreas with enucleation of any tumors elsewhere in the pancreas is often indicated [6]. Alcohol ablation may be reasonable for small or unresectable tumors and can ameliorate symptoms in selected cases [7]. (See "Insulinoma", section on 'Treatment'.)
MEN1 should be considered in all patients with a personal or family history of hyperparathyroidism (or pituitary tumor) who have an insulinoma. These patients are at risk for multiple insulinomas; as a result, they usually require extensive pancreatic surgery to extirpate all the hyperfunctioning islet tissue.
CASE 3 — A 44-year-old woman was referred for evaluation of recurrent syncope for five years. These episodes were characterized by premonitory symptoms of headache and slurred speech followed by suddenly finding herself on the floor. A "full-blown" episode could be aborted by eating. She had no family history of diabetes, but her boyfriend was taking a sulfonylurea drug for type 2 diabetes mellitus.
Three years ago, she had an 80 percent pancreatectomy, but no insulinoma was found. She was asymptomatic for four months after the operation, but her symptoms then recurred. She was treated with prednisone but became hyperglycemic, at which time glipizide and metformin were added. The prednisone (10 mg every other day) was discontinued three days and the glipizide (5 mg/day) and metformin (500 mg twice daily) two days before she came to our institution seeking another pancreatic operation. At that time, her serum glucose concentration was 50 mg/dL (2.8 mmol/L) 12 hours after her last meal.
The results of an outpatient fasting study were:
●Glucose – 38 mg/dL (2.1 mmol/L)
●Insulin – 9 microU/mL (54 pmol/L)
●C-peptide – 490 pmol/L
●Proinsulin – 33 pmol/L
●Beta-hydroxybutyrate – 0.2 mmol/L
●Glucose increase after glucagon – 32 mg/dL (1.8 mmol/L)
When the fast was ended, she had no symptoms or signs of hypoglycemia. Ultrasonography and spiral computed tomography (CT) of the abdomen were normal. Serum samples sent for analysis of sulfonylurea revealed the presence of glipizide, the concentrations being 473 and 853 ng/mL, respectively, when the serum glucose concentrations were 50 and 38 mg/dL (2.8 and 2.1 mmol/L); no other sulfonylureas were detected.
The patient did not acknowledge that she had continued to take glipizide. We therefore considered it possible, though unlikely, that her hypoglycemia was not due to glipizide. She then underwent an inpatient 72-hour supervised fast that was completely normal. The serum values were:
●Glucose – 50 mg/dL (2.8 mmol/L)
●Insulin – Undetectable
●C-peptide – 159 pmol/L
●Proinsulin – 3 pmol/L
●Beta-hydroxybutyrate – 4.5 mmol/L
●Glucose increase after glucagon – 8 mg/dL (0.4 mmol/L)
●Sulfonylurea – Negative
Comment — The initial biochemical data in this patient were consistent with endogenous hyperinsulinemia but were undoubtedly due to glipizide (see "Factitious hypoglycemia"). It is unlikely that glipizide would be measurable in high and increasing concentrations 48 hours or more after the last dose. The serum insulin, C-peptide, proinsulin, and beta-hydroxybutyrate concentrations in patients with hypoglycemia due to a sulfonylurea drug ingestion may be very similar to those in patients with an insulinoma. It is therefore essential to measure sulfonylureas in the same blood sample drawn for the other measurements. Liquid chromatography tandem mass spectrometry can detect both first- and second-generation agents and repaglinide at low dose.
Another noteworthy point in this case is that the initial fasting study was ended despite the patient being asymptomatic. When a fast is ended on the basis of a low serum glucose concentration in the absence of symptoms or signs of hypoglycemia, the results should be interpreted with caution because, in some normal young women, serum glucose concentrations may fall to 30 to 40 mg/dL (1.7 to 2.2 mmol/L) with no symptoms [8,9]. Furthermore, some patients with an insulinoma may appear normal when their serum glucose concentrations are low; when the serum glucose is then raised, it may be apparent that the patient did not have normal mental function when hypoglycemic. The decision to end the fast can, therefore, be difficult. A Mini-Mental State Examination may help establish a cognitive baseline to compare with repeat examination during a supervised fast [10].
Although the patient continued to deny surreptitious administration of glipizide, a diagnosis of sulfonylurea-induced hypoglycemia was suspected. The basis for this diagnosis was the detection of glipizide in serum, a negative supervised 72-hour fast when sulfonylurea was undetectable in serum, and the patient's access to a sulfonylurea drug.
CASE 4 — A 27-year-old man was referred by his local clinician for evaluation of hypoglycemia found incidentally during an assessment for peptic ulcer disease. During the last four months, he had several episodes of weakness and feeling "shaky inside" late in the evening. During the night, he would periodically drink soda. When symptomatic, reflectance meter blood glucose values measured by the patient using equipment purchased for his seven-year-old daughter (diagnosed with type 1 diabetes one year earlier) had been in the range of 40 to 50 mg/dL (2.2 to 2.8 mmol/L). Serum values after an overnight fast were:
●Glucose – 36 mg/dL (2.0 mmol/L)
●Insulin – 140 microU/mL (840 pmol/L)
●C-peptide – <33 pmol/L
●Proinsulin – 0.9 pmol/L
The low serum C-peptide and proinsulin values indicate that the hyperinsulinemia (140 microU/mL [840 pmol/L]) was due to exogenous insulin administration. The patient's daughter was being treated with a split-mixed insulin program; her mother administered the morning dose and her father the evening dose. When confronted with the evidence for factitial hypoglycemia due to insulin self-administration, the patient adamantly denied he was injecting himself.
Comment — This is a case of surreptitious insulin injection, which almost always occurs in patients with ready access to insulin. The diagnosis of factitial hypoglycemia due to insulin administration should be suspected if the serum insulin concentration exceeds 100 microU/mL (600 pmol/L) when the patient is hypoglycemic. It is confirmed if the serum C-peptide and proinsulin concentrations are low. It is important to appreciate that insulin analogs may not always be detected by a given insulin immunoassay, and exogenous insulin use should be suspected when the patient has access to insulin and C-peptide is suppressed even when insulin is not markedly elevated. It is important for the clinician to know which insulin assay is used and whether it measures human insulin, insulin analogs, or both. (See 'Case 8' below.)
A serum insulin/C-peptide molar ratio of more than 1.0 has been suggested as identifying factitial hypoglycemia due to insulin administration [11]. This is in contrast to ratios of 0.2 at the end of a supervised fast in patients with an insulinoma and in normal subjects [12]. In most patients with factitial hypoglycemia due to insulin administration, the discrepancy between the serum insulin (as long as the assay can detect the insulin or insulin analog used) and C-peptide concentrations is so great that determination of the molar ratio is superfluous. (See "Factitious hypoglycemia".)
CASE 5 — A 75-year-old woman was referred for the evaluation of episodes of weakness, tremor, and confusion that were relieved by eating. She was usually weak in the morning and felt better after breakfast. On one occasion, she felt weak and dizzy while walking and had to hold onto a tree to prevent herself from falling. On another occasion, she became confused while driving and mistakenly parked her car halfway across a sidewalk.
Her serum glucose, insulin, and C-peptide concentrations during an inpatient 72-hour fast are shown in the table (table 2). The fast was continued for the full 72 hours despite hypoglycemia at 48 hours and later because the patient was asymptomatic. At the end of the fast, her serum beta-hydroxybutyrate concentration was 4.9 mmol/L and her serum glucose response to intravenous glucagon was an increase of 12 mg/dL (0.7 mmol/L). On closer questioning later, the patient acknowledged that she had blurred vision and weakness during the last 24 hours of the fast. In retrospect, the fast was positive and should have been terminated earlier. Ultrasonography of the abdomen revealed a tumor in the head of the pancreas. The tumor was enucleated; pathologic examination revealed an insulinoma.
Comment — This case illustrates the importance of carefully questioning and examining patients for subtle symptoms and signs of hypoglycemia throughout the duration of a prolonged fast. Despite the degree and duration of hypoglycemia in this patient, the fast was not terminated, because she was judged to be asymptomatic.
Should a fast be terminated solely on the basis of the serum glucose concentration? This is problematic because there is no value that clearly distinguishes a patient with an insulinoma from a normal subject. Thus, there is no substitute for repeated assessment of cognitive function of any patient undergoing a fast to help to decide when to end the fast. However, for the patient who has already had Whipple's triad confirmed on another occasion, the purpose of doing a 72-hour fast is to assess the role of beta cell polypeptides in the genesis of hypoglycemia. In such a case, the fast can be terminated when the glucose is <55 mg/dL (3.0 mmol/L) since beta cell polypeptides should be suppressed at that glucose range.
The serum beta-hydroxybutyrate concentration and the serum glucose response to intravenous glucagon at the end of the fast were not diagnostic for insulinoma. It is possible that the counterregulatory forces operative during prolonged hypoglycemia predominated over the effects of insulin. We chose to discount these results because of the inappropriately high serum insulin, C-peptide, and proinsulin concentrations. The presence of a pancreatic mass provided additional evidence that the clinical assessment was correct. Markers of insulin action are therefore most useful when serum insulin, C-peptide, and proinsulin values are borderline.
CASE 6 — A 68-year-old woman had a six-year history of episodic confusion associated with unsteadiness and slurred speech. These symptoms tended to occur in the early morning hours and were relieved within five minutes after eating. Shortly after admission to the hospital and 12 hours after her last meal, she had the following serum values:
●Glucose – 31 mg/dL (1.7 mmol/L)
●Insulin – 43 microU/mL (258 pmol/L)
●C-peptide – 770 pmol/L
●Proinsulin – 240 pmol/L
●Beta-hydroxybutyrate – 0.4 mmol/L
●Glucose increase after glucagon – 45 mg/dL (2.5 mmol/L)
●Sulfonylurea – Negative
A diagnosis of insulinoma was made. Ultrasonography and spiral computed tomography (CT) of the abdomen were negative. The patient underwent celiac axis arteriography and a selective arterial calcium stimulation study in which calcium gluconate is injected into the superior mesenteric, gastroduodenal, or splenic arteries and blood is collected from the right hepatic vein for measurement of insulin levels [13]. A positive result is defined as at least a tripling of the basal hepatic venous serum insulin concentration. In one analysis, a fivefold or greater rise in insulin is more specific for a diagnosis of insulinoma [14]. In this patient, there was a more than twofold increase in serum insulin concentration after calcium was injected into the superior mesenteric artery, a 1.4-fold increase after calcium was injected into the splenic artery, and little increase after calcium was injected into the gastroduodenal artery (table 3). (See "Insulinoma", section on 'Diagnosis and staging'.)
A small vascular mass was seen in the mid-body of the pancreas on angiography; this is the typical appearance of an insulinoma on angiography. The blood supply to this mass arose from both the superior mesenteric and splenic arteries, as expected from the arterial stimulation test. The patient underwent successful enucleation of an insulinoma from the mid-body of the pancreas.
Comment — This case illustrates the utility and potential pitfalls of selective calcium stimulation for localization of a small insulinoma [15]. Ultrasonography and spiral CT of the abdomen detect only approximately 60 percent of insulinomas even in the most centers [16]. Accurate preoperative localization of an insulinoma is desirable because 10 percent of tumors may not be palpable at the time of surgery [17]. Intraoperative ultrasonography and palpation of the pancreas at surgery have a sensitivity approaching 100 percent in centers with considerable experience with this tumor [18].
Selective intra-arterial injection of calcium with sampling of hepatic venous insulin for the regionalization of insulinoma has been reported from two centers. This test is based upon the observation that calcium stimulates release of insulin from insulinomas but not normal beta cells [13]. In this patient, the arteriographic findings were subtle, and the presence of a positive calcium stimulation test in the same arterial territory as the vascular abnormality helped to confirm the presence of a tumor. The surgeon was therefore able to plan the operation in advance.
CASE 7 — A 76-year-old woman was referred for the evaluation of postprandial adrenergic symptoms with occasional visual changes. There was one episode of confusion while on a telephone call to her daughter. During an episode of lightheadedness, sweating, weakness, and irritability two hours after breakfast (which occurred while under observation), serum values were as follows:
●Glucose – 51 mg/dL (2.8 mmol/L)
●Insulin – 6.4 microU/mL (45.9 pmol/L)
●C-peptide – 858 pmol/L
●Proinsulin – 33 pmol/L
●Beta-hydroxybutyrate – 0.1 mmol/L
●Glucose increase after glucagon – 46 mg/dL (2.6 mmol/L)
●Sulfonylurea – Negative
A mixed-meal test was performed because of the presence of postprandial symptoms accompanied by biochemical evidence of insulin-mediated hypoglycemia. Biochemical testing 180 minutes after a mixed meal revealed the following:
●Glucose – 43 mg/dL (2.4 mmol/L)
●Insulin – 22.0 microU/mL (157.8 pmol/L)
●C-peptide – 1551 pmol/L
The biochemical tests confirmed postprandial hypoglycemia. However, some caution is needed as to the mechanism. In the postprandial period, there are no clear normative values for beta-cell polypeptides, although in the absence of upper gastrointestinal tract surgery, a diagnosis of endogenous hyperinsulinemic hypoglycemia is likely. The differential diagnosis included noninsulinoma pancreatogenous hypoglycemia (which is associated with postprandial hypoglycemia), insulin autoimmune hypoglycemia (postprandial or fasting hypoglycemia), or insulinoma, which more commonly, but not always, presents as fasting hypoglycemia [5,19]. (See "Noninsulinoma pancreatogenous hypoglycemia syndrome" and "Insulinoma".)
Insulin antibodies were negative, effectively ruling out insulin autoimmune hypoglycemia. Imaging studies were performed to localize possible insulinoma. However, ultrasound and computed tomography (CT) of the pancreas were negative. Therefore, a selective arterial calcium stimulation test to distinguish between a diffuse process (islet cell hypertrophy/nesidioblastosis) or a focal abnormality (insulinoma) was performed. This test involves selective injection of calcium gluconate into the gastroduodenal, splenic, and superior mesenteric arteries with subsequent sampling of the hepatic venous effluent for insulin. A positive result is a doubling or tripling of basal insulin concentrations.
In patients with insulinoma, the increase in insulin occurs in samples from the artery supplying the tumor, which facilitates operative localization. In contrast, patients with noninsulinoma pancreatogenous hypoglycemia have variable responses. There may be a positive result after injection of the splenic artery alone, splenic and gastroduodenal arteries, and occasionally after injection into all arteries supplying the pancreas. (See "Insulinoma", section on 'Diagnosis and staging'.)
The hepatic vein insulin concentrations after calcium gluconate injection into the gastroduodenal, superior mesenteric, and splenic arteries are shown in the table (table 4). The test was not positive, because it did not show a doubling or tripling of basal insulin concentrations after calcium gluconate injection into any of the arteries (see 'Case 6' above). However, the test results were unusual in that there was a decrease in insulin concentration in the superior mesenteric artery (SMA) at 20 minutes and then an increase. Although a technical error in the performance of the selective arterial calcium stimulation test or in the assay for insulin could not be established, the test was considered uninterpretable and, therefore, was repeated.
The repeat stimulation test was highly positive following the SMA injection and negative for the other two arteries, consistent with a focal abnormality (insulinoma) in the region of the SMA. The hepatic vein insulin concentrations after SMA injection are shown in the table (table 5).
The focality (and magnitude) of the response to the repeat selective arterial calcium stimulation test suggested a discrete lesion. Since the transabdominal ultrasound and abdominal CT were negative, an endoscopic ultrasound was performed for tumor localization. Endoscopic ultrasonography revealed a 1.3 cm insulinoma in the uncinate process that was later enucleated at pancreatic exploration.
Comment — This case illustrates the importance of distinguishing between a negative test result and one that is uninterpretable. It is unclear why the initial calcium stimulation test was uninterpretable, but the clinicians recognized that it was unusual. Selective arterial calcium stimulation tests are performed at few centers and are reserved for complex cases. They should be performed in centers where clinicians regularly perform and interpret the results.
Another important feature of this case is that the patient presented with postprandial hypoglycemia. Although insulinoma more commonly presents with symptoms of fasting hypoglycemia, it can also cause exclusively postprandial hypoglycemia.
CASE 8 — A 54-year-old woman presented for evaluation of postprandial hypoglycemia. Ten years previously she was diagnosed with necrotizing fasciitis and uncontrolled diabetes. After the fasciitis resolved, she was maintained on insulin therapy for approximately four years. She discontinued insulin approximately six years ago after losing weight. Hypoglycemia was documented in the postprandial period on at least two prior occasions. She had blood drawn at the endocrine testing center during a typical episode. Serum values were as follows:
●Glucose – 51 mg/dL (2.8 mmol/L)
●Insulin – 2.2 microU/mL (13.2 pmol/L)
●C-peptide – 700 pmol/L
●Proinsulin – 10 pmol/L
●Beta-hydroxybutyrate – 0.1 mmol/L
●Glucose increase after glucagon – 60 mg/dL (3.3 mmol/L)
●Sulfonylurea – Negative
Due to clinical suspicion of exogenous insulin use, the laboratory used a mass spectrometric analysis to demonstrate the presence of aspart insulin (442 pmol/L) present at the time of hypoglycemia.
Comment — Many insulin immunoassays do not reliably detect insulin analogs. Therefore, in situations where both C-peptide and insulin are suppressed, the use of insulin analogs should be suspected. This is especially true in patients with a prior history of insulin use or patients who have other access to insulin.
SUMMARY — The evaluation of patients suspected of having an insulinoma traditionally involves a two-step approach that requires biochemical confirmation of a low serum glucose in the presence of endogenous hyperinsulinemia (high serum insulin and C-peptide concentrations) in the absence of any evidence of hypoglycemic drug ingestion, followed by an attempt to localize the tumor preoperatively.
The first of these objectives can be best achieved by studying the patient during a spontaneous episode of hypoglycemia. Failing that, have the patient with food-deprived symptoms fast for up to 72 hours and the rare patient with postprandial symptoms undergo a mixed-meal test. The results are unequivocal in most patients with insulinoma. Some patients, however, have serum insulin and C-peptide concentrations that are borderline or normal and, therefore, indirect assessment of hyperinsulinemia by measuring serum beta-hydroxybutyrate and the serum glucose response to glucagon at the end of the fast may be helpful. Ultrasonography and spiral computed tomography (CT) of the abdomen should be performed once biochemical evidence of insulinoma is established. When these imaging techniques do not disclose the site of the tumor, endoscopic ultrasonography or arteriography and selective arterial calcium stimulation should be considered.
ACKNOWLEDGMENTS — The UpToDate editorial staff acknowledges Neena Natt, MD, who contributed to an earlier version of this topic review.
The UpToDate editorial staff also acknowledges F John Service, MD, PhD, now deceased, who contributed to an earlier version of this topic.
