The C-peptide suppression test by insulin-induced hypoglycaemia may give results that do not clearly distinguish all patients from controls, and its interpretation may be difficult, since the results depend on several parameters including gender and body mass index (Service et al. 1992). A more sophisticated C-peptide suppression test, combined with the euglycaemic clamp, was successfully employed for the diagnosis of insulinoma, but it cannot be recommended for clinical practice (Gin et al. 1987).

\* Regarding the diagnostic criteria for inappropriate insulin secretion, previously employed ratios (insulin/glucose ratio, Turner's index) are no more in use and do not bring greater diagnostic accuracy than the presently-used criteria (Vezzosi et al. 2007) (Service 1995). Though insulinoma secrete a high proportion of proinsulin, the proinsulin/insulin ratio

Diagnosis and Treatment of Insulinomas in the Adults 145

nutrient intake into three meals and two or three snacks. Repeated symptoms suggesting

Actual reactive insulin-mediated hypoglycaemia is observed in patients with dumping syndrome (see below, causes of hypoglycaemia differing from insulinomas), and this must

The first step of any evaluation is to check whether some of the medications ordered to the patient can cause hypoglycaemia. Insulin, insulin analogues or insulin secretagogues (sulfonylureas, glinides) are the most common causes of hypoglycaemia. Some patients may be mistaken and accidentally take inappropriate doses of medications, or undergo concurrent disorders that increase the bioavailability of the drug. Though metformin cannot induce hypoglycaemia, as confirmed by the expert consensus (Cryer et al. 2009), it has been reported to cause severe hypoglycaemia in the 1995 UK prospective diabetes study (Cryer et al. 2009) and in an elderly patient (Zitzmann et al. 2002). Thiazolidine-diones, alphaglucosidase inhibitors, GLP1-receptor agonists and dipeptidyl peptidase IV inhibitors do not induce hypoglycaemia themselves but increase the risk of hypoglycaemia when combined with an insulin secretagogue or insulin. Other medications are known to be liable to induce hypoglycaemia, mostly in aged patients and patients with associated diseases, or combination of treatments, or renal failure. Miscellaneous mechanisms may be involved: increased insulin release, reduced insulin clearance, interference with glucose metabolism,

A list of the most commonly employed medications liable to induce hypoglycaemia has been given in the 2009 consensus (Cryer et al. 2009): some of them are associated with a risk of hypoglycaemia with moderate quality of evidence: cibenzoline, gatifloxacin, pentamidine, quinine, indomethacin. Other medications are involved with a lesser quality of evidence. The most commonly reported offending drugs appeared to be quinolones, pentamidine, quinine, beta blockers, angiotensin-converting enzyme agents, and IGF (Murad et al. 2009). Salicylates in large doses (4-6 g/day) may produce rarely hypoglycaemia in adults, and more often in children, by reducing basal glucose production and possibly increasing sensitivity to insulin (Cryer 2008). Disopyramide (Marks & Teale 1999), non-steroid anti-inflammatory drugs, analgesics, antidepressants (Ben Salem et al. 2011) and antifungal agents (Lionakis et al. 2008) have been reported to induce hypoglycaemia. However hypoglycaemia appears to be rare with such medications, unless employed in combination with other medications or in patients with underlying diseases. Recently tyrosine kinase inhibitors were found to decrease plasma glucose concentrations (Agostino et al. 2010). Patients with huge non functional endocrine tumours and multiple liver metastases may be at risk of developing hypoglycaemia when treated by long-acting somatostatin analogues, as a result of reduced glucagon and GH secretion, impairment of hepatic glucose output and delay in intestinal absorption of

In such patients the diagnosis is often clinically obvious. Concurrent intake of hypoglycaemic agents must always be screened for, and in patients without any treatment

hypoglycaemia despite correct nutrition warrant further investigation.

be distinguished from "reactive functional hypoglycaemia".

**Medications known to be used by the patient** 

liver or renal toxicity (Ben Salem et al. 2011).

carbohydrates (Unek et al. 2009).

**Patient with known or obvious acute or chronic disorder** 

**2.2.3 Causes of hypoglycaemia differing from insulinomas** 

does not bring more valuable information than proinsulin levels for the diagnosis of insulinoma (Vezzosi et al. 2007).

#### **2.2 Differential diagnosis 2.2.1 Evaluation of the patient**

In an adult patient clinically otherwise healthy who presents with severe hypoglycemia, the diagnosis of insulinoma must be considered as very likely, provided that the patient does not use insulin or other hypoglycaemic agents.

Apart from "reactive functional hypoglycaemia"(see below), which is a controversial and heterogeneous medical entity without actual hypoglycaemia below 0.55 g/L, other actual causes of hypoglycaemia are often observed in patients presenting with other clinical symptoms related to their disease: alcohol ingestion, critically ill patients or septicemia, severe prolonged undernutrition, liver or renal deficiency, adrenal insufficiency, hypopituitarism, IGF-II or pro-IGF-II secreting tumours, or very rare patients with other tumour secretion (IGF-I, GLP1 and somatostatin).

Thus the first step is to record carefully the history of the patient including exposure to any medication, and the possible access to hypoglycaemic agents, and to perform a clinical physical examination.

Then the causes of hypoglycaemia are identified mainly on the basis of insulin, C-peptide and proinsulin levels at the time of hypoglycaemia. Additional biological evaluation comprises measurement of oral hypoglycaemic agents (at the time of hypoglycaemia) and anti-insulin antibodies (at any time) in plasma samples in patients who are found to have insulin-mediated hypoglycaemia. Other plasma measurements will be made if there is clinical presumption of a specific disease, for instance diseases known to lead to hypoglycaemia mediated by IGF-II or incompletely processed IGF-II ("big IGF-II"), exceptionally IGF-I. Finally, according to the 2009 expert consensus (Cryer et al. 2009), "a test of adrenocortical function is reasonable".

#### **2.2.2 Reactive "functional hypoglycaemia"**

Reactive "functional hypoglycaemia" is a controversial entity, and its existence is still debated. The first definition of "reactive functional hypoglycaemia" had been based on the belief that hypoglycaemic disorders could be divided into disorders responsible for fasting hypoglycaemia (mainly insulinoma) and disorders responsible for reactive hypoglycaemia (mainly "functional hypoglycaemia"). It is now known that many organic disorders may result both in fasting and/or reactive hypoglycaemia. However, many patients present with dizziness or minor autonomic symptoms within the hours following a meal, and most of them do not meet the criteria for actual hypoglycaemia, i.e. they do not have actually low plasma glucose levels below 0.55 g/L, and do not necessarily fulfil Whipple's triad. The socalled "reactive functional hypoglycaemia" may be the only likely diagnosis for some of these patients, but glucose levels do not drop below 0.55 g/L, and it cannot be responsible for symptoms of neuroglycopenia. It could reflect various mechanisms: 1) rapid changes in plasma glucose concentrations after ingestion of nutrients with a high glycaemic index; 2) a plasma glucose threshold for autonomic symptoms higher than that of other subjects 3) various reactions after ingestion of nutrients, including mechanisms differing from changes in plasma glucose levels. The patients are asked to reduce their intake of simple carbohydrates while increasing that of complex carbohydrates, and to divide their daily

does not bring more valuable information than proinsulin levels for the diagnosis of

In an adult patient clinically otherwise healthy who presents with severe hypoglycemia, the diagnosis of insulinoma must be considered as very likely, provided that the patient does

Apart from "reactive functional hypoglycaemia"(see below), which is a controversial and heterogeneous medical entity without actual hypoglycaemia below 0.55 g/L, other actual causes of hypoglycaemia are often observed in patients presenting with other clinical symptoms related to their disease: alcohol ingestion, critically ill patients or septicemia, severe prolonged undernutrition, liver or renal deficiency, adrenal insufficiency, hypopituitarism, IGF-II or pro-IGF-II secreting tumours, or very rare patients with other

Thus the first step is to record carefully the history of the patient including exposure to any medication, and the possible access to hypoglycaemic agents, and to perform a clinical

Then the causes of hypoglycaemia are identified mainly on the basis of insulin, C-peptide and proinsulin levels at the time of hypoglycaemia. Additional biological evaluation comprises measurement of oral hypoglycaemic agents (at the time of hypoglycaemia) and anti-insulin antibodies (at any time) in plasma samples in patients who are found to have insulin-mediated hypoglycaemia. Other plasma measurements will be made if there is clinical presumption of a specific disease, for instance diseases known to lead to hypoglycaemia mediated by IGF-II or incompletely processed IGF-II ("big IGF-II"), exceptionally IGF-I. Finally, according to the 2009 expert consensus (Cryer et al. 2009), "a

Reactive "functional hypoglycaemia" is a controversial entity, and its existence is still debated. The first definition of "reactive functional hypoglycaemia" had been based on the belief that hypoglycaemic disorders could be divided into disorders responsible for fasting hypoglycaemia (mainly insulinoma) and disorders responsible for reactive hypoglycaemia (mainly "functional hypoglycaemia"). It is now known that many organic disorders may result both in fasting and/or reactive hypoglycaemia. However, many patients present with dizziness or minor autonomic symptoms within the hours following a meal, and most of them do not meet the criteria for actual hypoglycaemia, i.e. they do not have actually low plasma glucose levels below 0.55 g/L, and do not necessarily fulfil Whipple's triad. The socalled "reactive functional hypoglycaemia" may be the only likely diagnosis for some of these patients, but glucose levels do not drop below 0.55 g/L, and it cannot be responsible for symptoms of neuroglycopenia. It could reflect various mechanisms: 1) rapid changes in plasma glucose concentrations after ingestion of nutrients with a high glycaemic index; 2) a plasma glucose threshold for autonomic symptoms higher than that of other subjects 3) various reactions after ingestion of nutrients, including mechanisms differing from changes in plasma glucose levels. The patients are asked to reduce their intake of simple carbohydrates while increasing that of complex carbohydrates, and to divide their daily

insulinoma (Vezzosi et al. 2007).

not use insulin or other hypoglycaemic agents.

tumour secretion (IGF-I, GLP1 and somatostatin).

test of adrenocortical function is reasonable".

**2.2.2 Reactive "functional hypoglycaemia"** 

**2.2 Differential diagnosis 2.2.1 Evaluation of the patient** 

physical examination.

nutrient intake into three meals and two or three snacks. Repeated symptoms suggesting hypoglycaemia despite correct nutrition warrant further investigation.

Actual reactive insulin-mediated hypoglycaemia is observed in patients with dumping syndrome (see below, causes of hypoglycaemia differing from insulinomas), and this must be distinguished from "reactive functional hypoglycaemia".

#### **2.2.3 Causes of hypoglycaemia differing from insulinomas**

#### **Medications known to be used by the patient**

The first step of any evaluation is to check whether some of the medications ordered to the patient can cause hypoglycaemia. Insulin, insulin analogues or insulin secretagogues (sulfonylureas, glinides) are the most common causes of hypoglycaemia. Some patients may be mistaken and accidentally take inappropriate doses of medications, or undergo concurrent disorders that increase the bioavailability of the drug. Though metformin cannot induce hypoglycaemia, as confirmed by the expert consensus (Cryer et al. 2009), it has been reported to cause severe hypoglycaemia in the 1995 UK prospective diabetes study (Cryer et al. 2009) and in an elderly patient (Zitzmann et al. 2002). Thiazolidine-diones, alphaglucosidase inhibitors, GLP1-receptor agonists and dipeptidyl peptidase IV inhibitors do not induce hypoglycaemia themselves but increase the risk of hypoglycaemia when combined with an insulin secretagogue or insulin. Other medications are known to be liable to induce hypoglycaemia, mostly in aged patients and patients with associated diseases, or combination of treatments, or renal failure. Miscellaneous mechanisms may be involved: increased insulin release, reduced insulin clearance, interference with glucose metabolism, liver or renal toxicity (Ben Salem et al. 2011).

A list of the most commonly employed medications liable to induce hypoglycaemia has been given in the 2009 consensus (Cryer et al. 2009): some of them are associated with a risk of hypoglycaemia with moderate quality of evidence: cibenzoline, gatifloxacin, pentamidine, quinine, indomethacin. Other medications are involved with a lesser quality of evidence. The most commonly reported offending drugs appeared to be quinolones, pentamidine, quinine, beta blockers, angiotensin-converting enzyme agents, and IGF (Murad et al. 2009). Salicylates in large doses (4-6 g/day) may produce rarely hypoglycaemia in adults, and more often in children, by reducing basal glucose production and possibly increasing sensitivity to insulin (Cryer 2008). Disopyramide (Marks & Teale 1999), non-steroid anti-inflammatory drugs, analgesics, antidepressants (Ben Salem et al. 2011) and antifungal agents (Lionakis et al. 2008) have been reported to induce hypoglycaemia. However hypoglycaemia appears to be rare with such medications, unless employed in combination with other medications or in patients with underlying diseases. Recently tyrosine kinase inhibitors were found to decrease plasma glucose concentrations (Agostino et al. 2010). Patients with huge non functional endocrine tumours and multiple liver metastases may be at risk of developing hypoglycaemia when treated by long-acting somatostatin analogues, as a result of reduced glucagon and GH secretion, impairment of hepatic glucose output and delay in intestinal absorption of carbohydrates (Unek et al. 2009).

#### **Patient with known or obvious acute or chronic disorder**

In such patients the diagnosis is often clinically obvious. Concurrent intake of hypoglycaemic agents must always be screened for, and in patients without any treatment

Diagnosis and Treatment of Insulinomas in the Adults 147

Hypoglycaemia has been occasionally reported in patients with congestive heart failure. Its pathogenesis is unknown, probably multifactorial, and may involve inhibited

Inanition related hypoglycaemia, with low insulin, C-peptide and proinsulin levels has been found in some patients with very severe malnutrition, such as patients with very severe chronic anorexia nervosa (Rich et al. 1990; Yanai et al. 2008). The occurrence of inanitionrelated severe hypoglycaemia in a patient with anorexia nervosa is thought to imply a grave

Adrenal and/or pituitary failure does not result in plasma glucose concentrations below 0.55 g/L unless associated with other disorders or hypoglycaemic medications. Plasma glucose concentrations of 0.6 g/L may be found in such patients, and are often associated with hyponatremia and/or typical clinical symptoms of the disease. As stated in the 2009 expert consensus (Cryer et al. 2009), adrenocortical failure without associated conditions is unlikely to induce hypoglycaemia below 0.55 g/L, and a low cortisol level at the time of hypoglycaemia may be the consequence of a lower glycaemic threshold to stimulate cortisol

in patients with recurrent hypoglycaemia without adrenal failure.

\* Hypoglycaemia related to non-insulin secreting tumours : IGF-II and precursors

If insulin, C-peptide and proinsulin levels are low at the time of hypoglycemia, the diagnosis of non pancreatic IGF-II (or pro-IGF-II) –secreting tumour must be considered. However, one must bear in mind that rare insulin-specific assays may not be able to identify insulin analogues (see below, "factitious hypoglycaemia"). Patients with IGF-II secreting tumours have low IGF-I levels and increased IGF-II/IGF-I ratios, which are pathognomonic of the diagnosis. IGF-II precursors ("big IGF-IIs") do not complex normally with IGFBPs, thus can reach more readily target tissues, which is thought to be the mechanism of hypoglycaemia in most patients. Enhanced mRNA expression of IGF-II and defective expression of prohormone convertase 4, a potential protease responsible for IGF-II precursor processing, were found in a patient with pleural solitary fibrous tumour (Tani et al. 2008). Glucose consumption is also thought to be a mechanism of hypoglycaemia in such tumours, and could be the major mechanism in some patients without evidence for IGF-II secretion, in whom PET-scan with 18-Fluoro-deoxy-glucose showed preferential tumour glucose uptake (Habra et al. 2010). Most IGF-II secreting tumours are large tumours that can be found on clinical examination or chest X-ray. IGF-II related hypoglycaemia is well known in patients with pleural solitary fibrous tumour (Doege-Potter syndrome). Solitary fibrous tumour is a rare mesenchymal neoplasm composed of CD34-positive fibroblastic cells; it can be either pleural or extrapleural, with abdominal locations, of whom one was found to grow in the bladder (Bruzzone et al. 2010). IGF-II secreting mesenchymal neoplasms are numerous, and also comprise synovial sarcomas, myxoid liposarcomas, GISTs, malignant peripheral nerve sheath tumours, chondrosarcomas, undifferentiated pleomorphic sarcomas, Ewing's

**Low insulin, C-peptide and proinsulin levels** 

**Heart failure** 

**Inanition** 

**Other causes** 

("big IGF-IIs")

gluconeogenesis (Cryer 2008).

prognosis (Rich et al. 1990). **Adrenal or pituitary failure** 

with insulin secretagogues, measurement of insulin, C-peptide, proinsulin and (whenever possible) beta-hydroxy-butyrate levels at the time of hypoglycaemia below 0.55 g/L will generally rule out insulin-mediated hypoglycaemia, unless the underlying disorder (especially in renal failure and acute liver failure) is liable to alter insulin, C-peptide and proinsulin metabolic processing and clearance.

## **Alcohol**

Alcohol ingestion is known to cause hypoglycaemia. Serum insulin and C-peptide are appropriately low during episodes of alcohol-induced hypoglycaemia (Marks & Teale 1999). Ethanol inhibits gluconeogenesis (Cryer 2008) and hormonal counterregulation to hypoglycaemia (Marks & Teale 1999; Cryer 2008), and can exert influences on pancreatic microcirculation, resulting in a massive redistribution of blood flow from the exocrine into the endocrine part of the pancreas, augmenting late-phase insulin secretion (Huang & Sjoholm 2008). In patients presenting in coma from alcohol-induced hypoglycaemia, recovery after glucose injection is immediate, with no relapse if there is no concurrent cause of hypoglycaemia; recovery is delayed when brain swelling has occurred, which requires specific treatment. Glucagon is not recommended (Marks & Teale 1999). Alcoholinduced hypoglycaemia typically occurs after a binge of alcohol consumption within 6 to 36 hours generally in a patient who presents glycogen depletion, and ethanol is measurable in blood but poorly correlates with glucose levels; it may also be a late feature of alcoholic keto-acidosis and ethanol may be undetectable in blood samples (Marks & Teale 1999).

#### **Acute hepatic failure**

Acute hepatic failure, with massive and rapid liver destruction, rather than common forms of cirrhosis and hepatitis, can lead to severe hypoglycaemia. Massive destruction of the liver tissue and defective glucose storage in extrahepatic organs are thought to be the main mechanisms, but non-insulin hypoglycaemic factors secreted by the damaged liver might be involved, since reversal of fulminant hepatitis-associated hypoglycaemia was observed at the anhepatic stage of liver transplantation (Ilan et al. 1996).

#### **Renal failure**

Hypoglycaemia is not rare in renal failure and its pathogenesis involves generally several mechanisms (Arem 1989; Basu et al. 2002). Mechanisms directly related to renal failure are involved: reduced renal gluconeogenesis, impaired renal insulin degradation and clearance. Additional mechanisms may be involved, such as medications, sepsis, poor nutrition with deficiency of precursors of gluconeogenesis, impaired glycogenolysis, defective counterregulation with frequent autonomic dysfunction, concurrent disorders such as concomitant liver disease, congestive heart failure, or an associated endocrine deficiency. In patients undergoing haemodialysis a particular mechanism is postdialysis glucose-induced hyperinsulinemia, caused by a high glucose content in the dialysate (Arem 1989; Basu et al. 2002). On the other hand dialysis with glucose-free solution results in frequent hypoglycaemia, whereas using glucose-added dialysis solution at 90 mg/dL was found to reduce the number and severity of hypoglycaemic episodes (Burmeister et al. 2007). Since clearance of insulin, C-peptide and proinsulin is impaired in renal failure, beta-hydroxybutyrate during a fast test and glucose response to glucagon were reported to be the best biological parameters for the differential diagnosis of insulinoma in patients with renal failure (Basu et al. 2002).

#### **Heart failure**

146 Basic and Clinical Endocrinology Up-to-Date

with insulin secretagogues, measurement of insulin, C-peptide, proinsulin and (whenever possible) beta-hydroxy-butyrate levels at the time of hypoglycaemia below 0.55 g/L will generally rule out insulin-mediated hypoglycaemia, unless the underlying disorder (especially in renal failure and acute liver failure) is liable to alter insulin, C-peptide and

Alcohol ingestion is known to cause hypoglycaemia. Serum insulin and C-peptide are appropriately low during episodes of alcohol-induced hypoglycaemia (Marks & Teale 1999). Ethanol inhibits gluconeogenesis (Cryer 2008) and hormonal counterregulation to hypoglycaemia (Marks & Teale 1999; Cryer 2008), and can exert influences on pancreatic microcirculation, resulting in a massive redistribution of blood flow from the exocrine into the endocrine part of the pancreas, augmenting late-phase insulin secretion (Huang & Sjoholm 2008). In patients presenting in coma from alcohol-induced hypoglycaemia, recovery after glucose injection is immediate, with no relapse if there is no concurrent cause of hypoglycaemia; recovery is delayed when brain swelling has occurred, which requires specific treatment. Glucagon is not recommended (Marks & Teale 1999). Alcoholinduced hypoglycaemia typically occurs after a binge of alcohol consumption within 6 to 36 hours generally in a patient who presents glycogen depletion, and ethanol is measurable in blood but poorly correlates with glucose levels; it may also be a late feature of alcoholic keto-acidosis and ethanol may be undetectable in blood samples (Marks &

Acute hepatic failure, with massive and rapid liver destruction, rather than common forms of cirrhosis and hepatitis, can lead to severe hypoglycaemia. Massive destruction of the liver tissue and defective glucose storage in extrahepatic organs are thought to be the main mechanisms, but non-insulin hypoglycaemic factors secreted by the damaged liver might be involved, since reversal of fulminant hepatitis-associated hypoglycaemia was observed at

Hypoglycaemia is not rare in renal failure and its pathogenesis involves generally several mechanisms (Arem 1989; Basu et al. 2002). Mechanisms directly related to renal failure are involved: reduced renal gluconeogenesis, impaired renal insulin degradation and clearance. Additional mechanisms may be involved, such as medications, sepsis, poor nutrition with deficiency of precursors of gluconeogenesis, impaired glycogenolysis, defective counterregulation with frequent autonomic dysfunction, concurrent disorders such as concomitant liver disease, congestive heart failure, or an associated endocrine deficiency. In patients undergoing haemodialysis a particular mechanism is postdialysis glucose-induced hyperinsulinemia, caused by a high glucose content in the dialysate (Arem 1989; Basu et al. 2002). On the other hand dialysis with glucose-free solution results in frequent hypoglycaemia, whereas using glucose-added dialysis solution at 90 mg/dL was found to reduce the number and severity of hypoglycaemic episodes (Burmeister et al. 2007). Since clearance of insulin, C-peptide and proinsulin is impaired in renal failure, beta-hydroxybutyrate during a fast test and glucose response to glucagon were reported to be the best biological parameters for the differential diagnosis of insulinoma in patients with renal

the anhepatic stage of liver transplantation (Ilan et al. 1996).

proinsulin metabolic processing and clearance.

**Alcohol** 

Teale 1999).

**Renal failure** 

failure (Basu et al. 2002).

**Acute hepatic failure** 

Hypoglycaemia has been occasionally reported in patients with congestive heart failure. Its pathogenesis is unknown, probably multifactorial, and may involve inhibited gluconeogenesis (Cryer 2008).

#### **Inanition**

Inanition related hypoglycaemia, with low insulin, C-peptide and proinsulin levels has been found in some patients with very severe malnutrition, such as patients with very severe chronic anorexia nervosa (Rich et al. 1990; Yanai et al. 2008). The occurrence of inanitionrelated severe hypoglycaemia in a patient with anorexia nervosa is thought to imply a grave prognosis (Rich et al. 1990).

#### **Adrenal or pituitary failure**

Adrenal and/or pituitary failure does not result in plasma glucose concentrations below 0.55 g/L unless associated with other disorders or hypoglycaemic medications. Plasma glucose concentrations of 0.6 g/L may be found in such patients, and are often associated with hyponatremia and/or typical clinical symptoms of the disease. As stated in the 2009 expert consensus (Cryer et al. 2009), adrenocortical failure without associated conditions is unlikely to induce hypoglycaemia below 0.55 g/L, and a low cortisol level at the time of hypoglycaemia may be the consequence of a lower glycaemic threshold to stimulate cortisol in patients with recurrent hypoglycaemia without adrenal failure.

#### **Other causes**

#### **Low insulin, C-peptide and proinsulin levels**

\* Hypoglycaemia related to non-insulin secreting tumours : IGF-II and precursors ("big IGF-IIs")

If insulin, C-peptide and proinsulin levels are low at the time of hypoglycemia, the diagnosis of non pancreatic IGF-II (or pro-IGF-II) –secreting tumour must be considered. However, one must bear in mind that rare insulin-specific assays may not be able to identify insulin analogues (see below, "factitious hypoglycaemia"). Patients with IGF-II secreting tumours have low IGF-I levels and increased IGF-II/IGF-I ratios, which are pathognomonic of the diagnosis. IGF-II precursors ("big IGF-IIs") do not complex normally with IGFBPs, thus can reach more readily target tissues, which is thought to be the mechanism of hypoglycaemia in most patients. Enhanced mRNA expression of IGF-II and defective expression of prohormone convertase 4, a potential protease responsible for IGF-II precursor processing, were found in a patient with pleural solitary fibrous tumour (Tani et al. 2008). Glucose consumption is also thought to be a mechanism of hypoglycaemia in such tumours, and could be the major mechanism in some patients without evidence for IGF-II secretion, in whom PET-scan with 18-Fluoro-deoxy-glucose showed preferential tumour glucose uptake (Habra et al. 2010). Most IGF-II secreting tumours are large tumours that can be found on clinical examination or chest X-ray. IGF-II related hypoglycaemia is well known in patients with pleural solitary fibrous tumour (Doege-Potter syndrome). Solitary fibrous tumour is a rare mesenchymal neoplasm composed of CD34-positive fibroblastic cells; it can be either pleural or extrapleural, with abdominal locations, of whom one was found to grow in the bladder (Bruzzone et al. 2010). IGF-II secreting mesenchymal neoplasms are numerous, and also comprise synovial sarcomas, myxoid liposarcomas, GISTs, malignant peripheral nerve sheath tumours, chondrosarcomas, undifferentiated pleomorphic sarcomas, Ewing's

Diagnosis and Treatment of Insulinomas in the Adults 149

If insulin, C-peptide and proinsulin levels are elevated at the time of hypoglycemia, insulinoma is the most likely diagnosis. One must remind that some insulinoma patients do not have elevated insulin levels when insulin is measured with insulin-specific assays (Vezzosi et al. 2003). When an insulinoma is not found by imaging techniques, two very different diagnoses must be suspected: factitious hypoglycemia related to sulfonylurea or glinide, or nesidioblastosis, which is rare in adult patients. Exceptionally, anti-insulin antibodies and extra-pancreatic non-islet cell insulin-secreting tumours may also be found; sometimes a genetic metabolic disorder which has been overlooked at a younger age may be diagnosed in an adult patient. Finally, a particular condition is reactive insulin-mediated hypoglycaemia in patients with dumping syndrome; some of these patients were reported

A plasma sample must be collected to measure oral hypoglycaemic agents (ideally all available sulfonylureas and glinides) with specific methods at the time of symptomatic hypoglycaemia. This is the only way to distinguish factitious (or accidental) hypoglycaemia related to insulin secretagogues from the endogenous abnormal insulin secretion observed in patients with insulinomas and nesidioblastosis. The diagnosis is generally suspected in patients whose relatives are diabetic or health professionals, but it may be very difficult

Anti-insulin antibodies responsible for hyperinsulinaemic hypoglycaemia have been rarely found (Taylor et al. 1989; Redmon & Nuttall 1999) generally in patients with autoimmune disorders who developed such antibodies spontaneously or a few weeks after specific treatments for autoimmune diseases, especially medications with a sulphydryl group, or in very rare cases of myeloma (Halsall et al. 2007). Anti-insulin antibodies must be measured in patients with hypoglycaemia related to endogenous hyperinsulinism, not necessarily during hypoglycaemia (Cryer et al. 2009). Insulin levels are often very high, partly as a consequence of an interference in the insulin assay by the anti-insulin antibody. Most patients have reactive hypoglycaemia, which could be a consequence of meal-induced exaggerated insulin secretion, overcoming the buffering effect of the antibody (Taylor et al. 1989); however some patients also have fasting hypoglycaemia (Basu et al. 2005). The natural course is not predictable, some patients may be controlled by frequent meals and many experienced remission some months after the withdrawal of the drug responsible for the onset of the

Non islet cell insulin-secreting tumours are quite exceptional. They must be distinguished from "ectopic insulinomas" which are extrapancreatic islet-cell tumours located in the proximity of the pancreas. Insulin-mediated hypoglycaemia has been reported in the following tumours : a small cell carcinoma of the cervix (Seckl et al. 1999), a neuroectodermal brain tumour (Nakamura et al. 2001), and a neuroendocrine carcinoma of the gallbladder (Ahn et

Todd et al reported a GLP-1 and somatostatin-secreting ovarian tumour (Todd et al. 2003). The patient had alternate episodes of hyper- and hypoglycaemia, with a negative fast test.

**High insulin, C-peptide and proinsulin levels** 

to have post-gastric bypass nesidioblastosis.

(Hirshberg et al. 2001). \* Anti-insulin antibodies

al. 2007).

\* Factitious hypoglycaemia related to sulfonylureas or glinides

disorder, but others required glucocorticoids or even plasmapheresis. \* Non islet cell insulin –secreting tumours and GLP-1 secreting tumour

sarcomas and tenosynovial giant cell tumours (Steigen et al. 2009) . We observed hypoglycaemia with suppressed insulin, C-peptide and proinsulin levels and high pro-IGF-II levels in a patient with a malignant meningioma and lung and pleural metastases (unpublished observation). Rare cases of adrenocortical carcinomas have also been reported to be responsible for hypoglycaemia (Cryer 2008). One exceptional patient with an islet-cell tumour of the pancreas was reported to have IGF-II-related hypoglycaemia (Chung et al. 2008).

\* Other tumours (non-insulin, non-IGF-II secreting tumours)

Exceptionally, tumour-induced hypoglycaemia was found to be related to hormone secretions that differed from insulin or IGF-II: IGF-I, in a metastasizing undifferentiated large-cell carcinoma of the lung (Nauck et al. 2007), somatostatin in an ovarian teratoma (Gregersen et al. 2002). The latter patient had alternate episodes of hyper- and hypoglycaemia, with only reactive post-prandial hypoglycaemia and a negative fast test; insulin levels were low at the time of hyperglycaemia, and were not reported at the time of hypoglycaemia. It was hypothesized that somatostatin led to suppression of glucagon secretion to a greater extent than that of insulin secretion in the post-prandial period, thereby leading to reactive hypoglycaemia. Another ovarian tumour was reported to induce hypoglycaemia by secreting GLP-1 and somatostatin (see below).

## **High insulin levels with low C-peptide and proinsulin levels**

\* Factitious hypoglycaemia related to insulin and insulin analogues

In clinical practice, the finding of high insulin levels with concomitant low C-peptide and proinsulin levels is almost specific of the diagnosis of surreptitious insulin injections (Grunberger et al. 1988). It must be known that some insulin-specific assays may not be able to identify insulin analogues (Sapin 2003; Neal & Han 2008); insulin assays that can identify the presence of insulin analogues are to be preferred.

\* Anti-insulin receptor antibodies

Such antibodies were found mostly in patients with immunity disorders, either before or after the use of medications like methimazole (Redmon & Nuttall 1999). Many of these patients have type B insulin resistance and diabetes mellitus, but some of them develop hypoglycaemia. Anti-insulin receptor antibodies bind to the insulin receptor, may mimic insulin action and cause fasting hypoglycaemia; also, they may inhibit insulin binding thereby inhibiting insulin clearance and increasing insulin levels, while the resulting hypoglycaemia suppresses beta-cell secretion and C-peptide levels (Taylor et al. 1989). Spontaneous remissions and improvement after glucocorticoid treatment have been reported (Taylor et al. 1989).

\* Insulin receptor mutation

Exceptionally patients may present an insulin receptor mutation, which is transmitted as an autosomal dominant trait, and delays insulin clearance, resulting in hypoglycaemia with inappropriate insulin levels after an overnight fast, and low C-peptide levels, with increased insulin/C-peptide ratio (Hojlund et al. 2004), or reactive insulin-mediated hypoglycaemia; in such patients the fast test is positive for hypoglycaemia (< 0.4 g/L), but insulin, C-peptide and proinsulin are suppressed during a prolonged fast. This disease can be considered to be one of the established genetic causes of nesidioblastosis (see below).

sarcomas and tenosynovial giant cell tumours (Steigen et al. 2009) . We observed hypoglycaemia with suppressed insulin, C-peptide and proinsulin levels and high pro-IGF-II levels in a patient with a malignant meningioma and lung and pleural metastases (unpublished observation). Rare cases of adrenocortical carcinomas have also been reported to be responsible for hypoglycaemia (Cryer 2008). One exceptional patient with an islet-cell tumour of the pancreas was reported to have IGF-II-related hypoglycaemia (Chung et al.

Exceptionally, tumour-induced hypoglycaemia was found to be related to hormone secretions that differed from insulin or IGF-II: IGF-I, in a metastasizing undifferentiated large-cell carcinoma of the lung (Nauck et al. 2007), somatostatin in an ovarian teratoma (Gregersen et al. 2002). The latter patient had alternate episodes of hyper- and hypoglycaemia, with only reactive post-prandial hypoglycaemia and a negative fast test; insulin levels were low at the time of hyperglycaemia, and were not reported at the time of hypoglycaemia. It was hypothesized that somatostatin led to suppression of glucagon secretion to a greater extent than that of insulin secretion in the post-prandial period, thereby leading to reactive hypoglycaemia. Another ovarian tumour was reported to induce

In clinical practice, the finding of high insulin levels with concomitant low C-peptide and proinsulin levels is almost specific of the diagnosis of surreptitious insulin injections (Grunberger et al. 1988). It must be known that some insulin-specific assays may not be able to identify insulin analogues (Sapin 2003; Neal & Han 2008); insulin assays that can identify

Such antibodies were found mostly in patients with immunity disorders, either before or after the use of medications like methimazole (Redmon & Nuttall 1999). Many of these patients have type B insulin resistance and diabetes mellitus, but some of them develop hypoglycaemia. Anti-insulin receptor antibodies bind to the insulin receptor, may mimic insulin action and cause fasting hypoglycaemia; also, they may inhibit insulin binding thereby inhibiting insulin clearance and increasing insulin levels, while the resulting hypoglycaemia suppresses beta-cell secretion and C-peptide levels (Taylor et al. 1989). Spontaneous remissions and improvement after glucocorticoid treatment have been

Exceptionally patients may present an insulin receptor mutation, which is transmitted as an autosomal dominant trait, and delays insulin clearance, resulting in hypoglycaemia with inappropriate insulin levels after an overnight fast, and low C-peptide levels, with increased insulin/C-peptide ratio (Hojlund et al. 2004), or reactive insulin-mediated hypoglycaemia; in such patients the fast test is positive for hypoglycaemia (< 0.4 g/L), but insulin, C-peptide and proinsulin are suppressed during a prolonged fast. This disease can be considered to be one of the established genetic causes of nesidioblastosis (see

\* Other tumours (non-insulin, non-IGF-II secreting tumours)

hypoglycaemia by secreting GLP-1 and somatostatin (see below). **High insulin levels with low C-peptide and proinsulin levels**  \* Factitious hypoglycaemia related to insulin and insulin analogues

the presence of insulin analogues are to be preferred.

\* Anti-insulin receptor antibodies

reported (Taylor et al. 1989). \* Insulin receptor mutation

below).

2008).

#### **High insulin, C-peptide and proinsulin levels**

If insulin, C-peptide and proinsulin levels are elevated at the time of hypoglycemia, insulinoma is the most likely diagnosis. One must remind that some insulinoma patients do not have elevated insulin levels when insulin is measured with insulin-specific assays (Vezzosi et al. 2003). When an insulinoma is not found by imaging techniques, two very different diagnoses must be suspected: factitious hypoglycemia related to sulfonylurea or glinide, or nesidioblastosis, which is rare in adult patients. Exceptionally, anti-insulin antibodies and extra-pancreatic non-islet cell insulin-secreting tumours may also be found; sometimes a genetic metabolic disorder which has been overlooked at a younger age may be diagnosed in an adult patient. Finally, a particular condition is reactive insulin-mediated hypoglycaemia in patients with dumping syndrome; some of these patients were reported to have post-gastric bypass nesidioblastosis.

\* Factitious hypoglycaemia related to sulfonylureas or glinides

A plasma sample must be collected to measure oral hypoglycaemic agents (ideally all available sulfonylureas and glinides) with specific methods at the time of symptomatic hypoglycaemia. This is the only way to distinguish factitious (or accidental) hypoglycaemia related to insulin secretagogues from the endogenous abnormal insulin secretion observed in patients with insulinomas and nesidioblastosis. The diagnosis is generally suspected in patients whose relatives are diabetic or health professionals, but it may be very difficult (Hirshberg et al. 2001).

#### \* Anti-insulin antibodies

Anti-insulin antibodies responsible for hyperinsulinaemic hypoglycaemia have been rarely found (Taylor et al. 1989; Redmon & Nuttall 1999) generally in patients with autoimmune disorders who developed such antibodies spontaneously or a few weeks after specific treatments for autoimmune diseases, especially medications with a sulphydryl group, or in very rare cases of myeloma (Halsall et al. 2007). Anti-insulin antibodies must be measured in patients with hypoglycaemia related to endogenous hyperinsulinism, not necessarily during hypoglycaemia (Cryer et al. 2009). Insulin levels are often very high, partly as a consequence of an interference in the insulin assay by the anti-insulin antibody. Most patients have reactive hypoglycaemia, which could be a consequence of meal-induced exaggerated insulin secretion, overcoming the buffering effect of the antibody (Taylor et al. 1989); however some patients also have fasting hypoglycaemia (Basu et al. 2005). The natural course is not predictable, some patients may be controlled by frequent meals and many experienced remission some months after the withdrawal of the drug responsible for the onset of the disorder, but others required glucocorticoids or even plasmapheresis.

\* Non islet cell insulin –secreting tumours and GLP-1 secreting tumour

Non islet cell insulin-secreting tumours are quite exceptional. They must be distinguished from "ectopic insulinomas" which are extrapancreatic islet-cell tumours located in the proximity of the pancreas. Insulin-mediated hypoglycaemia has been reported in the following tumours : a small cell carcinoma of the cervix (Seckl et al. 1999), a neuroectodermal brain tumour (Nakamura et al. 2001), and a neuroendocrine carcinoma of the gallbladder (Ahn et al. 2007).

Todd et al reported a GLP-1 and somatostatin-secreting ovarian tumour (Todd et al. 2003). The patient had alternate episodes of hyper- and hypoglycaemia, with a negative fast test.

Diagnosis and Treatment of Insulinomas in the Adults 151

hypoglycaemia (Goldfine et al. 2007). GLP-1 is thought to be unable to induce hypoglycaemia in patients with normal beta cells, but high GLP-1 levels might override glucose-dependency and cause hypoglycaemia (Toft-Nielsen et al. 1998). The degree of postprandial GLP-1-induced glucagon suppression (which is also dependent on glucose, GLP-2 and GIP levels) might play a role in the resulting post-prandial glucose levels. Insulinmediated hypoglycaemia after gastric bypass surgery is not associated with overexpression of GLP-1 receptor in islets (Reubi et al. 2010). McLaughlin et al reported immediate reversal of reactive hypoglycaemia by feeding a patient with Roux-en-Y gastric bypass through a gastrostomy tube, which argued against a structural pancreatic abnormality in this patient and provided major evidence for a functional dysregulation of post-prandial insulin secretion in patients with gastric bypass-induced hypoglycaemia (McLaughlin et al. 2010). Nevertheless such patients often require insulin-suppressing treatments (diazoxide, calcium channel blockers, somatostatin agonists) (Tack et al. 2009) (Vella & Service 2007). Acarbose, which delays absorption of simple carbohydrates and may reduce GLP-1 secretion, was successfully employed (in combination with dietary advices) (Tack et al. 2009; Hanaire et al. 2010) but it was found relatively ineffective by others (Vella & Service 2007). A patient who had dumping syndrome after partial gastric resection was finally admitted in our department, 20 years after gastric surgery, for severe reactive insulin-mediated hypoglycaemia, with a negative fast test, and normal pancreas on CT-scan; hypoglycaemia was controlled by diazoxide, not by octreotide. Partial (distal) pancreatectomy has been performed to control hypoglycaemia in some of these patients, as in patients with nesidioblastosis. Since insulinomas may present with GIP and/or GLP-1 receptors, occurrence of insulin-mediated hypoglycaemia after bypass surgery or gastrectomy can be caused by an insulinoma. Radiological evaluation is of course

The topographic assessment of an insulinoma must be performed only after the biological diagnosis of hypoglycemia related to inappropriate insulin secretion has been confirmed. Insulinomas are generally thought to be difficult to localize due to their small size. Intraoperative palpation and ultrasound examination of the pancreas had been thought to be the best methods to detect insulinomas. Using such methods intra-operatively remains mandatory, but to date careful radiological examination must be performed to localize insulinomas before surgery in order to avoid failure of surgery to cure the disease and reoperations. The imaging techniques now in use have greatly improved, so that most

High definition multidetector helical CT-scans have become widely available since the early 2000s, so that CT-scan is now again the first examination that must be performed to localize insulinomas. Multidetector CT-scan allows optimization of the scan protocols (with very rapid scan times reducing movement artefacts), and arterial and portal venous phase images on thin-sections. A meticulous technique is mandatory and detailed recommendations have been made regarding the examination procedure (Rockall & Reznek 2007) (Sundin et al. 2009). Most insulinomas are small, isodense with the pancreas on pre-contrast images, then hypervascular on arterial phase images, but sometimes are more easily detected on portal venous phase images. More rarely, the tumours are hyperdense to the pancreas, or have nodular calcification, or appear hypovascular, cystic or hypodense after injection of contrast medium. Exceptionally, ectopic insulinomas are located in the proximity of the pancreas or

mandatory in such patients.

**2.3 Topographic assessment** 

insulinomas are detected preoperatively.

Insulin-mediated hypoglycaemia was thought to be a consequence of GLP-1 and somatostatin secretion; since the effects of GLP-1 on insulin secretion are dependent on glucose levels, the inhibitory effects of somatostatin on glucagon secretion may have played a role in inducing reactive hypoglycaemia.

\* Genetic metabolic disorders

Some patients with hypoglycaemia related to metabolic genetic disorders may occasionally be diagnosed to have such disorders when they are adults. Some of these disorders can be considered as possible causes of nesidioblastosis (Christesen et al. 2008; Palladino & Stanley 2011) (Flanagan et al. 2011b) (Flanagan et al. 2011a) (see last section of this chapter): 1) physical exercise-induced hypoglycaemia caused by failed silencing of monocarboxylate transporter 1 in pancreatic beta cells; 2) patients with mostly post-prandial (but also fasting) symptoms of hypoglycaemia in mutations of insulin receptor (see above), activating mutation of glucokinase with some patients diagnosed only as adults (Christesen et al. 2008), and hyperinsulinism-hyperammonemia syndrome due to activating mutations of glutamate dehydrogenase; 3) patients with mostly fasting (and also post-prandial) hypoglycaemia due to mutations in the genes of the subunits of K-ATP channel (ABCC8, KCNJ11) and mutations of 3 hydroxy-acyl-CoA dehydrogenase (HADH) gene. Among 79 patients admitted in our department between 1990 and 2010 for organic causes of hypoglycaemia (69 of them with insulinomas) only one had hyperinsulinism-hyperammonemia syndrome (diagnosed at the age of 30), with an activating mutation of glutamate dehydrogenase, and another one had two heterozygous mutations of the ABCC8 gene (diagnosed at the age of 47). Both patients had been treated since early childhood for epilepsy.

Other patients may present with other metabolic disorders such as disorders of the glycogen metabolic pathway. Some patients with almost asymptomatic forms of glycogen storage disease type Ia may develop hepatocellular adenomas and carcinomas as adults (Cassiman et al. 2010).

A detailed review of these genetic disorders in childhood has been made by Palladino (Palladino et al. 2008).

\* Reactive hypoglycaemia in patients with dumping syndrome

Postprandial hypoglycaemia caused by rapid emptying of the gastric remnant has long been known to be a possible complication of gastric resection; it is associated with various symptoms, including symptoms unrelated to hypoglycaemia, which are termed dumping syndrome. Many, but not all patients with post-gastric bypass were found to have reactive severe insulin-mediated hypoglycaemia a few months or years after bypass surgery. Continuous glucose monitoring can be a useful tool to identify hypoglycaemia in patients after bariatric surgery (Hanaire et al. 2010). The abnormal insulin responses to intraarterial calcium injection and the histopathological findings in 6 patients, compared to pancreata from patients with pancreatic cancer, led to the conclusion than such patients might have gastric bypass-induced nesidioblastosis (Service & O'Brien 2005) (see below). This conclusion was challenged, since Meier JJ et al. did not confirm the histopathological diagnosis of nesidioblastosis in the 6 patients, when compared to pancreata collected at autopsy; they only observed a greater beta-cell nuclear diameter than in controls, and this nuclear diameter was positively correlated with BMI, if the BMI used for the patients in the correlation was the BMI observed prior to bypass surgery (Meier et al. 2006). Goldfine AB et al reported increased GLP-1 levels after a liquid mixed-meal in patients who had postgastric bypass surgery hypoglycaemia, in comparison with those who did not have

Insulin-mediated hypoglycaemia was thought to be a consequence of GLP-1 and somatostatin secretion; since the effects of GLP-1 on insulin secretion are dependent on glucose levels, the inhibitory effects of somatostatin on glucagon secretion may have played

Some patients with hypoglycaemia related to metabolic genetic disorders may occasionally be diagnosed to have such disorders when they are adults. Some of these disorders can be considered as possible causes of nesidioblastosis (Christesen et al. 2008; Palladino & Stanley 2011) (Flanagan et al. 2011b) (Flanagan et al. 2011a) (see last section of this chapter): 1) physical exercise-induced hypoglycaemia caused by failed silencing of monocarboxylate transporter 1 in pancreatic beta cells; 2) patients with mostly post-prandial (but also fasting) symptoms of hypoglycaemia in mutations of insulin receptor (see above), activating mutation of glucokinase with some patients diagnosed only as adults (Christesen et al. 2008), and hyperinsulinism-hyperammonemia syndrome due to activating mutations of glutamate dehydrogenase; 3) patients with mostly fasting (and also post-prandial) hypoglycaemia due to mutations in the genes of the subunits of K-ATP channel (ABCC8, KCNJ11) and mutations of 3 hydroxy-acyl-CoA dehydrogenase (HADH) gene. Among 79 patients admitted in our department between 1990 and 2010 for organic causes of hypoglycaemia (69 of them with insulinomas) only one had hyperinsulinism-hyperammonemia syndrome (diagnosed at the age of 30), with an activating mutation of glutamate dehydrogenase, and another one had two heterozygous mutations of the ABCC8 gene (diagnosed at the age of 47). Both patients had

Other patients may present with other metabolic disorders such as disorders of the glycogen metabolic pathway. Some patients with almost asymptomatic forms of glycogen storage disease type Ia may develop hepatocellular adenomas and carcinomas as adults (Cassiman

A detailed review of these genetic disorders in childhood has been made by Palladino

Postprandial hypoglycaemia caused by rapid emptying of the gastric remnant has long been known to be a possible complication of gastric resection; it is associated with various symptoms, including symptoms unrelated to hypoglycaemia, which are termed dumping syndrome. Many, but not all patients with post-gastric bypass were found to have reactive severe insulin-mediated hypoglycaemia a few months or years after bypass surgery. Continuous glucose monitoring can be a useful tool to identify hypoglycaemia in patients after bariatric surgery (Hanaire et al. 2010). The abnormal insulin responses to intraarterial calcium injection and the histopathological findings in 6 patients, compared to pancreata from patients with pancreatic cancer, led to the conclusion than such patients might have gastric bypass-induced nesidioblastosis (Service & O'Brien 2005) (see below). This conclusion was challenged, since Meier JJ et al. did not confirm the histopathological diagnosis of nesidioblastosis in the 6 patients, when compared to pancreata collected at autopsy; they only observed a greater beta-cell nuclear diameter than in controls, and this nuclear diameter was positively correlated with BMI, if the BMI used for the patients in the correlation was the BMI observed prior to bypass surgery (Meier et al. 2006). Goldfine AB et al reported increased GLP-1 levels after a liquid mixed-meal in patients who had postgastric bypass surgery hypoglycaemia, in comparison with those who did not have

a role in inducing reactive hypoglycaemia.

been treated since early childhood for epilepsy.

\* Reactive hypoglycaemia in patients with dumping syndrome

et al. 2010).

(Palladino et al. 2008).

\* Genetic metabolic disorders

hypoglycaemia (Goldfine et al. 2007). GLP-1 is thought to be unable to induce hypoglycaemia in patients with normal beta cells, but high GLP-1 levels might override glucose-dependency and cause hypoglycaemia (Toft-Nielsen et al. 1998). The degree of postprandial GLP-1-induced glucagon suppression (which is also dependent on glucose, GLP-2 and GIP levels) might play a role in the resulting post-prandial glucose levels. Insulinmediated hypoglycaemia after gastric bypass surgery is not associated with overexpression of GLP-1 receptor in islets (Reubi et al. 2010). McLaughlin et al reported immediate reversal of reactive hypoglycaemia by feeding a patient with Roux-en-Y gastric bypass through a gastrostomy tube, which argued against a structural pancreatic abnormality in this patient and provided major evidence for a functional dysregulation of post-prandial insulin secretion in patients with gastric bypass-induced hypoglycaemia (McLaughlin et al. 2010). Nevertheless such patients often require insulin-suppressing treatments (diazoxide, calcium channel blockers, somatostatin agonists) (Tack et al. 2009) (Vella & Service 2007). Acarbose, which delays absorption of simple carbohydrates and may reduce GLP-1 secretion, was successfully employed (in combination with dietary advices) (Tack et al. 2009; Hanaire et al. 2010) but it was found relatively ineffective by others (Vella & Service 2007). A patient who had dumping syndrome after partial gastric resection was finally admitted in our department, 20 years after gastric surgery, for severe reactive insulin-mediated hypoglycaemia, with a negative fast test, and normal pancreas on CT-scan; hypoglycaemia was controlled by diazoxide, not by octreotide. Partial (distal) pancreatectomy has been performed to control hypoglycaemia in some of these patients, as in patients with nesidioblastosis. Since insulinomas may present with GIP and/or GLP-1 receptors, occurrence of insulin-mediated hypoglycaemia after bypass surgery or gastrectomy can be caused by an insulinoma. Radiological evaluation is of course mandatory in such patients.

#### **2.3 Topographic assessment**

The topographic assessment of an insulinoma must be performed only after the biological diagnosis of hypoglycemia related to inappropriate insulin secretion has been confirmed. Insulinomas are generally thought to be difficult to localize due to their small size. Intraoperative palpation and ultrasound examination of the pancreas had been thought to be the best methods to detect insulinomas. Using such methods intra-operatively remains mandatory, but to date careful radiological examination must be performed to localize insulinomas before surgery in order to avoid failure of surgery to cure the disease and reoperations. The imaging techniques now in use have greatly improved, so that most insulinomas are detected preoperatively.

High definition multidetector helical CT-scans have become widely available since the early 2000s, so that CT-scan is now again the first examination that must be performed to localize insulinomas. Multidetector CT-scan allows optimization of the scan protocols (with very rapid scan times reducing movement artefacts), and arterial and portal venous phase images on thin-sections. A meticulous technique is mandatory and detailed recommendations have been made regarding the examination procedure (Rockall & Reznek 2007) (Sundin et al. 2009). Most insulinomas are small, isodense with the pancreas on pre-contrast images, then hypervascular on arterial phase images, but sometimes are more easily detected on portal venous phase images. More rarely, the tumours are hyperdense to the pancreas, or have nodular calcification, or appear hypovascular, cystic or hypodense after injection of contrast medium. Exceptionally, ectopic insulinomas are located in the proximity of the pancreas or

Diagnosis and Treatment of Insulinomas in the Adults 153

pancreas, such tumours being impalpable at surgery. A lower sensitivity for pancreatic tail lesions (37-50%) has been reported (Ardengh et al. 2000). Pancreatic nodularities may be mistaken for insulinomas (Kann et al. 2003). Endoscopic ultrasound is an invasive method, which requires sedation, and its results are operator-dependent. Therefore it is now a second-line examination after conventional imaging. Even if recent CT-scan techniques lead to localization of most insulinomas, endoscopic ultrasound still brings the advantage of making a better evaluation of the lesions (especially if they are multiple, as often in MEN-1) and of their proximity with pancreatic ducts and vessels, and it also enables the operator to

Intra-operative ultrasound is mandatory during surgery of insulinomas in order to localize impalpable tumours. High-frequency intra-operative ultrasound has been shown in expert hands to improve detection of small tumours in the pancreatic head and multiple pancreatic lesions to up to 97% (Grant 1988) (Grant 1999). Intra-operative ultrasound examination of the pancreas is also mandatory (using a specific material) during laparoscopic surgery of insulinomas. Despite the high sensitivity of intra-operative ultrasound, a thorough preoperative examination is necessary to localize insulinomas, in order to maximize the chance for a successful initial resection, minimize the risk of re-operation with its possible morbidity, help in the choice of the surgical technique and make laparoscopic surgery easier

Nuclear imaging of insulinomas is also possible. One of the first functional imaging method was based on the fact that most endocrine tumours display somatostatin receptors, which led to the use of Octreoscan\* scintigraphy (111In-DTPA-octreotide). On the other hand, sst2A receptors were found in only about 66% of insulinomas (Reubi & Waser 2003). Recommendations have been made regarding the protocol (Balon et al. 2001; Bombardieri et al. 2003). Because of competition of the somatostatin analogues at the receptor site, somatostatin analogue treatment should be withdrawn before performing Octreoscan scintigraphy, though there are contradictory data (Dörr et al. 1993). The sensitivity for tumour detection is poor in patients with insulinomas (20-50%) (Krenning et al. 1993) in comparison with other endocrine tumours. However it was 80% in a series of 14 insulinoma patients with the use of SPECT (single-photon emission computed tomography) and an injected activity of 250 MBq of Octreoscan instead of the commonly recommended activity of 150-200 MBq (Schillaci et al. 2000). We only found a 24% sensitivity and neither pre-treatment with octreotide nor the lack of expression of sst2A receptors could account for all the negative scans (Vezzosi et al. 2005). In our experience, Octreoscan scintigraphy proved to be useful in a patient with an ectopic insulinoma, which was located about 2 cm below the body of the pancreas. Using the same technique for Octreoscan scintigraphy, 89.1% of non-insulinoma well-differentiated digestive endocrine tumours were detected in our centre. Small tumours cannot be detected by Octreoscan scintigraphy (Dromain et al. 2005). Octreoscan displays a lower affinity for sst2 receptors than octreotide (Reubi et al. 2000). Specific problems of imaging techniques may play a role in this low detection of insulinomas, even in sst2-positive insulinomas (Vezzosi et al. 2005). Octreoscan scintigraphy may help in localizing metastases, but MRI is better than Octreoscan scintigraphy for detection of liver metastases of endocrine tumours (Dromain et al. 2005). To date, the systematic use of Octreoscan scintigraphy in

Positron emission tomography (PET) with 18-F-labelled deoxy-glucose (FDG) has not proven advantageous for imaging endocrine tumours with the exception of poorly

when it appears to be the possible choice to remove the insulinoma.

insulinoma patients is questionable.

make biopsies.

of the liver. While the sensitivity of non helical CT had been 29%, with multidectector helical CT-scan a meticulous technique can result in 94 % sensitivity (Rockall & Reznek 2007). CTscan can show metastases (see below, malignant insulinomas)

MRI is often complementary to CT, in order to confirm a suspected lesion on CT, or to search for a tumour that CT-scan has not been able to localize; combining both methods (CT and MRI) improves the accuracy for detection of insulinomas. Detailed technical information has been provided (Rockall & Reznek 2007). Temporal resolution and spatial resolution are greater with CT than with MRI, but MRI allows high resolution contrastenhanced imaging. Pancreatic endocrine tumours generally appear hypo-intense on T1 weighted sequences and hyperintense on T2-weighted sequences. On pre-contrast imaging, fat suppressed T1 and fat saturation T2 may improve visualization of the tumours, which appear of low intensity on T1 and have a bright T2 signal. There is a marked homogeneous enhancement after injection of gadolinium, which may in some cases render the lowintensity tumour isointense and less detectable (Rockall & Reznek 2007). The sensitivity of MRI for detection of pancreatic endocrine tumours was first reported to be less than that of CT, but is now similar to that of CT: a sensitivity of up to 94% has now been reported (Thoeni et al. 2000). MRI has also proven very helpful in the diagnosis of hepatic metastases of malignant insulinomas (see below, malignant insulinomas).

Trans-abdominal ultrasound examination is traditionally thought to be of low sensitivity to detect pancreatic endocrine tumours like insulinomas: the sensitivity was as low as 20 % in the series reported in the 1990s, with a maximum of 66-80% (Grant 1999). Insulinomas generally appear as small hypoechoic solid masses. The limitations of ultrasound examination are related to body habitus and tumour size; also, the presence of bowel obscures the left upper quadrant, and insulinomas located in the body and the tail of the pancreas may be difficult to visualize. Water intake by the patient allows to use the stomach as an acoustic window and may improve the detection. A sensitivity of up to 79.3% has been reported with recent material for transabdominal ultrasound (Oshikawa et al. 2002) (Rockall & Reznek 2007).

More recently, contrast-enhanced ultrasonography has been reported to detect insulinomas in a high number of patients. Very promising results were found by An L et al (An et al. 2010), who reported the results obtained with a second-generation sonographic contrast agent (SonoVue) in 31 patients with hypoglycaemia related to endogenous hyperinsulinism, among whom 27 had a solitary insulinoma, 3 had multiple insulinomas, and 1 had no definite tumour during laparoscopic evaluation. The results of contrast-enhanced ultrasonography were in agreement with surgical findings except in three patients: one solitary tumour had been localized in the head of the pancreas and was found in the duodenum ligament; one tumour had been visualized whereas no tumour was found during surgery; in the patient with 6 insulinomas, only 2 tumours had been visualized. Thus the diagnostic sensitivity and localization specificity were 33 (89.2%) and 32 (86.5%) of 37 surgically verified insulinomas, respectively. The enhancement pattern of insulinoma was fast wash-in and slow wash-out. The insulinomas displayed homogeneous hypervascularity in the earlier arterial phase and had still a hyperenhancing pattern in the late phase. This ultrasound non invasive examination could probably be developed in the next years.

Endoscopic ultrasound has been reported to be the reference method to localize insulinomas in the 1990s: it allows to use a high frequency US probe in close proximity to the pancreas, and in expert hands its sensitivity has been shown to be 79-100% (Rosch et al. 1992) (McLean & Fairclough 2005). It may show very small tumours (about 3 mm in size) in the head of the

of the liver. While the sensitivity of non helical CT had been 29%, with multidectector helical CT-scan a meticulous technique can result in 94 % sensitivity (Rockall & Reznek 2007). CT-

MRI is often complementary to CT, in order to confirm a suspected lesion on CT, or to search for a tumour that CT-scan has not been able to localize; combining both methods (CT and MRI) improves the accuracy for detection of insulinomas. Detailed technical information has been provided (Rockall & Reznek 2007). Temporal resolution and spatial resolution are greater with CT than with MRI, but MRI allows high resolution contrastenhanced imaging. Pancreatic endocrine tumours generally appear hypo-intense on T1 weighted sequences and hyperintense on T2-weighted sequences. On pre-contrast imaging, fat suppressed T1 and fat saturation T2 may improve visualization of the tumours, which appear of low intensity on T1 and have a bright T2 signal. There is a marked homogeneous enhancement after injection of gadolinium, which may in some cases render the lowintensity tumour isointense and less detectable (Rockall & Reznek 2007). The sensitivity of MRI for detection of pancreatic endocrine tumours was first reported to be less than that of CT, but is now similar to that of CT: a sensitivity of up to 94% has now been reported (Thoeni et al. 2000). MRI has also proven very helpful in the diagnosis of hepatic metastases

Trans-abdominal ultrasound examination is traditionally thought to be of low sensitivity to detect pancreatic endocrine tumours like insulinomas: the sensitivity was as low as 20 % in the series reported in the 1990s, with a maximum of 66-80% (Grant 1999). Insulinomas generally appear as small hypoechoic solid masses. The limitations of ultrasound examination are related to body habitus and tumour size; also, the presence of bowel obscures the left upper quadrant, and insulinomas located in the body and the tail of the pancreas may be difficult to visualize. Water intake by the patient allows to use the stomach as an acoustic window and may improve the detection. A sensitivity of up to 79.3% has been reported with recent material for transabdominal ultrasound (Oshikawa et al. 2002) (Rockall

More recently, contrast-enhanced ultrasonography has been reported to detect insulinomas in a high number of patients. Very promising results were found by An L et al (An et al. 2010), who reported the results obtained with a second-generation sonographic contrast agent (SonoVue) in 31 patients with hypoglycaemia related to endogenous hyperinsulinism, among whom 27 had a solitary insulinoma, 3 had multiple insulinomas, and 1 had no definite tumour during laparoscopic evaluation. The results of contrast-enhanced ultrasonography were in agreement with surgical findings except in three patients: one solitary tumour had been localized in the head of the pancreas and was found in the duodenum ligament; one tumour had been visualized whereas no tumour was found during surgery; in the patient with 6 insulinomas, only 2 tumours had been visualized. Thus the diagnostic sensitivity and localization specificity were 33 (89.2%) and 32 (86.5%) of 37 surgically verified insulinomas, respectively. The enhancement pattern of insulinoma was fast wash-in and slow wash-out. The insulinomas displayed homogeneous hypervascularity in the earlier arterial phase and had still a hyperenhancing pattern in the late phase. This ultrasound non invasive examination could probably be developed in the next years. Endoscopic ultrasound has been reported to be the reference method to localize insulinomas in the 1990s: it allows to use a high frequency US probe in close proximity to the pancreas, and in expert hands its sensitivity has been shown to be 79-100% (Rosch et al. 1992) (McLean & Fairclough 2005). It may show very small tumours (about 3 mm in size) in the head of the

scan can show metastases (see below, malignant insulinomas)

of malignant insulinomas (see below, malignant insulinomas).

& Reznek 2007).

pancreas, such tumours being impalpable at surgery. A lower sensitivity for pancreatic tail lesions (37-50%) has been reported (Ardengh et al. 2000). Pancreatic nodularities may be mistaken for insulinomas (Kann et al. 2003). Endoscopic ultrasound is an invasive method, which requires sedation, and its results are operator-dependent. Therefore it is now a second-line examination after conventional imaging. Even if recent CT-scan techniques lead to localization of most insulinomas, endoscopic ultrasound still brings the advantage of making a better evaluation of the lesions (especially if they are multiple, as often in MEN-1) and of their proximity with pancreatic ducts and vessels, and it also enables the operator to make biopsies.

Intra-operative ultrasound is mandatory during surgery of insulinomas in order to localize impalpable tumours. High-frequency intra-operative ultrasound has been shown in expert hands to improve detection of small tumours in the pancreatic head and multiple pancreatic lesions to up to 97% (Grant 1988) (Grant 1999). Intra-operative ultrasound examination of the pancreas is also mandatory (using a specific material) during laparoscopic surgery of insulinomas. Despite the high sensitivity of intra-operative ultrasound, a thorough preoperative examination is necessary to localize insulinomas, in order to maximize the chance for a successful initial resection, minimize the risk of re-operation with its possible morbidity, help in the choice of the surgical technique and make laparoscopic surgery easier when it appears to be the possible choice to remove the insulinoma.

Nuclear imaging of insulinomas is also possible. One of the first functional imaging method was based on the fact that most endocrine tumours display somatostatin receptors, which led to the use of Octreoscan\* scintigraphy (111In-DTPA-octreotide). On the other hand, sst2A receptors were found in only about 66% of insulinomas (Reubi & Waser 2003). Recommendations have been made regarding the protocol (Balon et al. 2001; Bombardieri et al. 2003). Because of competition of the somatostatin analogues at the receptor site, somatostatin analogue treatment should be withdrawn before performing Octreoscan scintigraphy, though there are contradictory data (Dörr et al. 1993). The sensitivity for tumour detection is poor in patients with insulinomas (20-50%) (Krenning et al. 1993) in comparison with other endocrine tumours. However it was 80% in a series of 14 insulinoma patients with the use of SPECT (single-photon emission computed tomography) and an injected activity of 250 MBq of Octreoscan instead of the commonly recommended activity of 150-200 MBq (Schillaci et al. 2000). We only found a 24% sensitivity and neither pre-treatment with octreotide nor the lack of expression of sst2A receptors could account for all the negative scans (Vezzosi et al. 2005). In our experience, Octreoscan scintigraphy proved to be useful in a patient with an ectopic insulinoma, which was located about 2 cm below the body of the pancreas. Using the same technique for Octreoscan scintigraphy, 89.1% of non-insulinoma well-differentiated digestive endocrine tumours were detected in our centre. Small tumours cannot be detected by Octreoscan scintigraphy (Dromain et al. 2005). Octreoscan displays a lower affinity for sst2 receptors than octreotide (Reubi et al. 2000). Specific problems of imaging techniques may play a role in this low detection of insulinomas, even in sst2-positive insulinomas (Vezzosi et al. 2005). Octreoscan scintigraphy may help in localizing metastases, but MRI is better than Octreoscan scintigraphy for detection of liver metastases of endocrine tumours (Dromain et al. 2005). To date, the systematic use of Octreoscan scintigraphy in insulinoma patients is questionable.

Positron emission tomography (PET) with 18-F-labelled deoxy-glucose (FDG) has not proven advantageous for imaging endocrine tumours with the exception of poorly

Diagnosis and Treatment of Insulinomas in the Adults 155

(Guettier et al. 2009), there are several caveats, because this technique is based on the following 3 assumptions: 1) a 2-fold insulin increase is pathological; 2) normal beta cells do not show a positive response to calcium injection; 3) insulinomas always respond to calcium injection. All 3 assumptions can be challenged on the basis of the available data. According to Guettier's report, there was no explanation for 3 false-negative cases of insulinomas. A 2 fold increase in insulin levels was found in two or more vessels in 24 of 45 cases, which could suggest that some normal pancreatic areas respond to calcium injection, as already reported by Wiesli (Wiesli et al. 2004a), who recommended that only a 5-fold increase should be considered as a positive response if insulin levels are measured with an insulinspecific assay. This invasive test should probably be a last-line investigation in patients with hypoglycaemia related to endogenous hyperinsulinism that cannot be controlled easily by a medical treatment. Within the last 10 years, we used this method according to the modified protocol described by O'Shea (O'Shea et al. 1996) in only two patients, who were finally found to have nesidioblastosis (with two heterozygous mutations in the ABCC8 gene in one

A recent study (Druce et al. 2010) compared the diagnostic accuracy of most methods employed to localize insulinomas in the years 1990-2009 in 36 patients, among whom 30 were treated by surgery. Among these 30 patients, 25 had CT and 28 had MRI with successful localization in 16 (64%) and 21 (75%) by CT and MRI, respectively. Together this led to 80% of successful localizations. Radiolabelled octreotide scanning was positive in 10 out of 20 cases (50%). Endoscopic ultrasound identified 17 tumours in 26 patients (65.4%). 27 patients had selective intra-arterial calcium stimulation: 6 (22%) did not localize, 17 (63%) were correctly localized, and 4 (15%) gave discordant or confusing results. Thus combination of CT and MRI predicts tumour localization with high accuracy, Octreoscan and endoscopic ultrasound are less helpful but could be valuable in selected cases, and

In most patients, by now, multidetector CT-scan (with trans-abdominal ultrasound and/or MRI as complementary techniques) employed by an experienced radiologist will localize insulinomas, and must be considered as a first-line examination, while contrast-enhanced

Surgery remains the only curative treatment of insulinomas. Long-term remission can be achieved by surgery in 95% of patients according to a recent study (Zhao et al. 2011). Two different types of surgery can be performed : minimal resection i.e. either tumour enucleation whenever it is possible or central pancreatectomy, or a more extended resection, i.e. left-sided pancreatectomy or pancreatico-duodenectomy. The type of surgery depends on the size and the location of the tumour and of its proximity with specific anatomical structures (pancreatic duct, vessels, adjacent organs). The improvement in the pre-operative imaging techniques enables the surgeon to have an accurate pre-operative topographic assessment and to decide the surgical approach pre-operatively. However intra-operative

Whenever it is possible, tumour enucleation is to be chosen (Crippa et al. 2007). It allows to cure the patient in most cases with minimized risks of post-operative pancreatic exocrine deficiency and diabetes mellitus. It must be performed only to remove small tumours on the surface of the pancreas, with a distance of more than 2-3 mm between the

patient) with false positive "pseudo-insulinoma" results.

ultrasound needs further evaluation.

**2.4 Treatment 2.4.1 Surgery** 

calcium stimulation may provide an additional functional perspective.

bidigital palpation and ultrasound remain valuable (Fendrich et al. 2009).

differentiated, highly proliferative tumours (Pasquali et al. 1998; Sundin et al. 2009). The unstable glucose levels in insulinoma patients may also increase the difficulties.

Since pancreatic endocrine tumours are part of the APUDomas because of their ability for amine precursor uptake and decarboxylation through the action of aromatic amino acid decarboxylase, 18F-labelled-DOPA was also employed as a tracer. The first results in 10 patients with hyperinsulinemic hypoglycaemia were very encouraging. The pancreatic lesion was localized in 9/10 patients with subsequent confirmation by histological analysis (Kauhanen et al. 2007). The following studies did not confirm these results. A 27% sensitivity was reported for detection of pancreatic endocrine tumours (Montravers et al. 2009). In six patients with hyperinsulinemic hypoglycaemia (4 solitary insulinomas, 1 diffuse beta-cell hyperplasia, 1 malignant insulinoma), F-DOPA-PET was positive in only one case and it underestimated the extent of the disease in the malignant insulinoma with liver metastases (Tessonnier et al. 2010). More data are necessary to conclude, but to date F-DOPA PET does not appear to be a major help in managing pancreatic endocrine tumours.

The first results achieved with a radiolabelled GLP-1 analogue seem promising. Insulinomas were reported to be characterized by a very high incidence of GLP-1 receptors, these receptors being expressed in a particularly high density (Reubi & Waser 2003). A radiolabelled GLP-1 analogue (111In-DOTA-exendin-4) was employed in 6 patients (5 with pancreatic insulinomas, 1 with an extrapancreatic insulinoma). Scintigraphy with GLP-1 analogue correctly localized the insulinoma in all 6 patients, while CT-scan was positive in 1, MRI-scan in 1 and endoscopic ultrasound in 4. Four patients had an intense uptake 4 hours after injection, and the other 2 patients showed demarcation between tumour and kidney only at late scans 3-7 days after injection (Christ et al. 2009). Interestingly, while exendin-4 does not induce hypoglycaemia in normal subjects, the insulinoma patients experienced a decrease in plasma glucose levels, with a nadir 40 min after injection, and an exogenous glucose infusion was necessary in 3 patients (for a maximum of 120 min), so that scintigraphy with radiolabelled exendin-4 appears to be also a provocative test in some insulinoma patients, and plasma glucose levels must be monitored after injection.

Finally trans-hepatic pancreatic venous sampling, to screen for an insulin concentration gradient, or calcium injection after selective catheterization of the arteries supplying the pancreas combined with measurement of insulin levels in the right hepatic vein can be used in difficult cases. The latter method was first described by Doppman et al (Doppman et al. 1993). A recent report by an expert team has evaluated the diagnostic accuracy of this technique (Guettier et al. 2009). A 2-fold or greater step-up in the right hepatic vein insulin concentration from baseline at times 20, 40 or 60 seconds after injection was considered to be positive. When a positive response was found at more than one injection site, the dominant site was used to predict tumour localization. In the absence of anatomical variants, a positive response to calcium injection in the gastroduodenal artery or superior mesenteric artery predicts a head-neck tumour, a positive response to the injection into the proximal splenic or the midsplenic artery predicts a body-tail tumour, and a response to calcium injection into the proper hepatic artery represents liver metastases. In 45 patients with surgically proven insulinomas, 38 (84%) localized to the correct region; in 5/45 (11%) the result was falsely negative, and 2/45 (4%) had false-positive localizations. Thus in expert hands selective intra-arterial calcium stimulation is a valuable tool to localize insulinomas when conventional imaging is negative. However, as pointed out by the expert group

differentiated, highly proliferative tumours (Pasquali et al. 1998; Sundin et al. 2009). The

Since pancreatic endocrine tumours are part of the APUDomas because of their ability for amine precursor uptake and decarboxylation through the action of aromatic amino acid decarboxylase, 18F-labelled-DOPA was also employed as a tracer. The first results in 10 patients with hyperinsulinemic hypoglycaemia were very encouraging. The pancreatic lesion was localized in 9/10 patients with subsequent confirmation by histological analysis (Kauhanen et al. 2007). The following studies did not confirm these results. A 27% sensitivity was reported for detection of pancreatic endocrine tumours (Montravers et al. 2009). In six patients with hyperinsulinemic hypoglycaemia (4 solitary insulinomas, 1 diffuse beta-cell hyperplasia, 1 malignant insulinoma), F-DOPA-PET was positive in only one case and it underestimated the extent of the disease in the malignant insulinoma with liver metastases (Tessonnier et al. 2010). More data are necessary to conclude, but to date F-DOPA PET does not appear to be a major help in managing pancreatic endocrine

The first results achieved with a radiolabelled GLP-1 analogue seem promising. Insulinomas were reported to be characterized by a very high incidence of GLP-1 receptors, these receptors being expressed in a particularly high density (Reubi & Waser 2003). A radiolabelled GLP-1 analogue (111In-DOTA-exendin-4) was employed in 6 patients (5 with pancreatic insulinomas, 1 with an extrapancreatic insulinoma). Scintigraphy with GLP-1 analogue correctly localized the insulinoma in all 6 patients, while CT-scan was positive in 1, MRI-scan in 1 and endoscopic ultrasound in 4. Four patients had an intense uptake 4 hours after injection, and the other 2 patients showed demarcation between tumour and kidney only at late scans 3-7 days after injection (Christ et al. 2009). Interestingly, while exendin-4 does not induce hypoglycaemia in normal subjects, the insulinoma patients experienced a decrease in plasma glucose levels, with a nadir 40 min after injection, and an exogenous glucose infusion was necessary in 3 patients (for a maximum of 120 min), so that scintigraphy with radiolabelled exendin-4 appears to be also a provocative test in some insulinoma patients, and plasma glucose

Finally trans-hepatic pancreatic venous sampling, to screen for an insulin concentration gradient, or calcium injection after selective catheterization of the arteries supplying the pancreas combined with measurement of insulin levels in the right hepatic vein can be used in difficult cases. The latter method was first described by Doppman et al (Doppman et al. 1993). A recent report by an expert team has evaluated the diagnostic accuracy of this technique (Guettier et al. 2009). A 2-fold or greater step-up in the right hepatic vein insulin concentration from baseline at times 20, 40 or 60 seconds after injection was considered to be positive. When a positive response was found at more than one injection site, the dominant site was used to predict tumour localization. In the absence of anatomical variants, a positive response to calcium injection in the gastroduodenal artery or superior mesenteric artery predicts a head-neck tumour, a positive response to the injection into the proximal splenic or the midsplenic artery predicts a body-tail tumour, and a response to calcium injection into the proper hepatic artery represents liver metastases. In 45 patients with surgically proven insulinomas, 38 (84%) localized to the correct region; in 5/45 (11%) the result was falsely negative, and 2/45 (4%) had false-positive localizations. Thus in expert hands selective intra-arterial calcium stimulation is a valuable tool to localize insulinomas when conventional imaging is negative. However, as pointed out by the expert group

unstable glucose levels in insulinoma patients may also increase the difficulties.

tumours.

levels must be monitored after injection.

(Guettier et al. 2009), there are several caveats, because this technique is based on the following 3 assumptions: 1) a 2-fold insulin increase is pathological; 2) normal beta cells do not show a positive response to calcium injection; 3) insulinomas always respond to calcium injection. All 3 assumptions can be challenged on the basis of the available data. According to Guettier's report, there was no explanation for 3 false-negative cases of insulinomas. A 2 fold increase in insulin levels was found in two or more vessels in 24 of 45 cases, which could suggest that some normal pancreatic areas respond to calcium injection, as already reported by Wiesli (Wiesli et al. 2004a), who recommended that only a 5-fold increase should be considered as a positive response if insulin levels are measured with an insulinspecific assay. This invasive test should probably be a last-line investigation in patients with hypoglycaemia related to endogenous hyperinsulinism that cannot be controlled easily by a medical treatment. Within the last 10 years, we used this method according to the modified protocol described by O'Shea (O'Shea et al. 1996) in only two patients, who were finally found to have nesidioblastosis (with two heterozygous mutations in the ABCC8 gene in one patient) with false positive "pseudo-insulinoma" results.

A recent study (Druce et al. 2010) compared the diagnostic accuracy of most methods employed to localize insulinomas in the years 1990-2009 in 36 patients, among whom 30 were treated by surgery. Among these 30 patients, 25 had CT and 28 had MRI with successful localization in 16 (64%) and 21 (75%) by CT and MRI, respectively. Together this led to 80% of successful localizations. Radiolabelled octreotide scanning was positive in 10 out of 20 cases (50%). Endoscopic ultrasound identified 17 tumours in 26 patients (65.4%). 27 patients had selective intra-arterial calcium stimulation: 6 (22%) did not localize, 17 (63%) were correctly localized, and 4 (15%) gave discordant or confusing results. Thus combination of CT and MRI predicts tumour localization with high accuracy, Octreoscan and endoscopic ultrasound are less helpful but could be valuable in selected cases, and calcium stimulation may provide an additional functional perspective.

In most patients, by now, multidetector CT-scan (with trans-abdominal ultrasound and/or MRI as complementary techniques) employed by an experienced radiologist will localize insulinomas, and must be considered as a first-line examination, while contrast-enhanced ultrasound needs further evaluation.
