Neuroendocrine tumours of the pancreas are being recognized with increasing frequency because of the development of radioimmunoassays for a growing number of circulating polypeptides. The clinical syndromes produced by excessive hormone secretion range from biochemical curiosities (PPoma) to life threatening derangements of homeostasis (insulinoma and VIPoma). Although characteristic groups of symptoms may trigger clinical suspicion, specific clinical diagnosis depends on measuring elevated levels of the circulating hormone. These tumours are generally slow growing and span the spectrum of behaviour from benign to malignant (Table 1) 394.

The islets of Langerhans contain at least five cell types: &bgr;-cells, which produce insulin, were the first to be identified. Subsequent immunocytochemistry has identified &agr;-cells, which produce glucagon, &ggr;-cells, which produce somatostatin, and F cells which produce pancreatic polypeptide. The fifth islet cell type is the enterochromaffin cell, which produces serotonin. All islet cells are of neuroectodermal origin, contain neuron-specific enolase, and are capable of amine precursor uptake and decarboxylation. Individual islet cells can produce more than one hormone. They are also capable of producing hormones not usually found in the pancreas, such as gastrin, adrenocorticotrophin, vasoactive intestinal peptide, and growth hormone. Although many pancreatic islet cell tumours are multihormonal, one peptide generally predominates and is responsible for producing the clinical syndrome. Approximately 10 per cent of pancreatic islet cell tumours arise in association with multiple endocrine neoplasia type I. Conversely, 85 per cent of patients with this syndrome have pancreatic neuroendocrine tumours or hyperplasia. Pancreatic neuroendocrine tumours associated with multiple endocrine neoplasia type I are non-&bgr;-cell tumours, and tend to be multifocal. Non-functioning islet cell tumours comprise about 20 per cent of all pancreatic neuroendocrine tumours. They may produce as yet unidentified peptides with effects too vague to produce a clearly defined syndrome, or they may have defective release mechanisms of established peptides.

Insulinomas are the most common pancreatic neuroendocrine tumours. The mean age of patients at presentation is 45 years, and the disease is rare in adolescence. Since the presentation may be insidious and the symptoms subtle, the diagnosis depends on carefully collected biochemical data, based on an understanding of normal insulin secretion and glucose homeostasis.

Since the measurement of insulin by radioimmunoassay is critical to the diagnosis of insulinoma, it is important to understand the mechanism of insulin synthesis, secretion, and release. Insulin is initially synthesized on the rough endoplasmic reticulum in a form known as pre-proinsulin. Following synthesis, a 24-amino acid peptide is cleaved to yield proinsulin, which is transported to the Golgi apparatus where it is packaged into secretory granules. Secretory granules contain proteases that cleave the C-peptide sequence of the molecule to yield biologically active insulin. Therefore, when insulin is secreted, equimolar amounts of C-peptide are also released; they can be detected in the plasma (Fig. 1) 1342.

Although some proinsulin is normally released from &bgr;-cells, it is usually less than 25 per cent of the amount of insulin. Serum levels of C-peptide and proinsulin are important in the differential diagnosis of hypoglycaemia (Fig. 2) 1343.

The clinical presentation of insulinomas is hypoglycaemia in virtually all cases. Symptoms caused by local effects of the tumour mass are rare: most commonly, patients present with neurological effects of hypoglycaemia, such as visual disturbances, confusion, lethargy, weakness, or transient motor deficits. Erroneous psychiatric or neurological diagnoses are common and insulinomas may go undiagnosed for as long as 3 years. Hypoglycaemia also causes catecholamine release with consequent autonomic symptoms such as sweating, anxiety, and palpitations. Symptoms are due to reactive hypoglycaemia (i.e. postprandial) or induced by fasting or exercise; the distinction is important. Symptoms from insulinomas tend to be progressive over time and to occur while fasting. To avoid symptoms, many patients learn to eat frequent meals and thus are overweight. When hyperinsulinism is severe, symptoms may develop as early as 2 to 4 h after meals. This phenomenon can be demonstrated in oral glucose tolerance tests; reactive hypoglycaemia is followed by a rebound to normal plasma glucose levels if the hypoglycaemia is functional. In insulinoma, however, blood glucose levels remain depressed.

In 1935, Whipple and Frantz proposed the diagnostic triad of symptoms of hypoglycaemia produced by fasting or exercise, blood glucose less than 45 (mmol/l), and symptoms relieved by administration of glucose. Advances in the ability to detect and measure insulin and its metabolites have allowed refinement of the diagnosis. The presence of hypoglycaemia in the face of inappropriately elevated levels of insulin is the key to diagnosis: this can be demonstrated by a prolonged fast in over 95 per cent of patients. Fifty-six per cent of patients with insulinomas will become hypoglycaemic within 14 h, 78 per cent by 18 h, and 87 per cent by 24 h. If the fast is extended to 72 h, over 95 per cent of patients will develop hypoglycaemia. If symptomatic or biochemical hypoglycaemia does not develop by 72 h, the patient should be exercised. This will provoke hypoglycaemia in virtually all insulinoma patients.

Absolute elevations of immunoreactive insulin concentrations are not found in all patients with insulinomas: the tumour may secrete in short bursts causing wide fluctuation in blood levels of immunoreactive insulin. While levels above 20 &mgr;u/ml are diagnostic, a normal level does not exclude the diagnosis. In normal subjects, blood immunoreactive insulin decreases with fasting to 3 to 7 &mgr;u/ml and changes in level parallel changes in glucose concentration. In insulinoma patients, the immunoreactive insulin levels may fluctuate, but they will not fall with fasting. The key in using measurements of immunoreactive insulin concentration for diagnosing insulinoma is the demonstration of inappropriate levels. After an overnight fast, the normal ratio of immunoreactive insulin (&mgr;u/ml)/glucose (mg/dl) is 0.3. States of insulin resistance such as obesity may cause an elevation of this ratio but these patients are not hypoglycaemic. In patients with organic hyperinsulinism, the ratio remains constant during fasting, but rises in the presence of insulinoma. This ratio should be studied with respect to its change over time during a fast, rather than as an absolute value. Another rule of thumb is that immunoreactive insulin levels above 6 &mgr;u/ml in the presence of a blood glucose concentration less than 40 mmol/l is diagnostic of insulinoma.

Proinsulin level is normally 10 to 20 per cent that of immunoreactive insulin but neoplastic tissue releases excess proinsulin. Up to 90 per cent of patients with insulinomas have proinsulin levels greater than 25 per cent that of immunoreactive insulin. Markedly elevated proinsulin levels serve as a marker for malignancy, although the overlap between benign and malignant disease is such that this finding is not an absolute discriminant. C-peptide levels should also be monitored: if factitious hypoglycaemia is produced by the administration of insulin appreciable amounts of this peptide will not be found. Sulphonylurea levels should also be measured. With the availability of radioimmunoassays, provocative tests such as the tolbutamide test are no longer employed.

Between 75 and 80 per cent of insulinomas are solitary, benign lesions. Approximately 10 per cent occur as multiple adenomas associated with multiple endocrine neoplasia type I, and another 10 per cent are malignant. Five per cent represent diffuse microadenomatosis, in which multiple small non-encapsulated lesions are distributed throughout the pancreas (nesidioblastosis). Although common in childhood this is rare in adults.

Most benign insulinomas are between 0.5 and 2 cm in diameter and are distributed evenly throughout the pancreas. Pathological analysis alone cannot distinguish between benign and malignant insulinomas, although malignant lesions tend to be larger (average size 6.2 cm) and one-half of malignant lesions have metastasized at the time of diagnosis. Malignant spread is generally confined to the peripancreatic region and liver. When insulinomas arise in patients with multiple endocrine neoplasia type I, multiple adenomas or a combination of micro- and macroadenomatosis is common, yet surprisingly good results are obtained with resection, perhaps because multiple lesions do not all secrete at the same time.

Once the diagnosis of insulinoma is made, the location of the tumour should be sought. While some experienced surgeons are able to identify 75 to 90 per cent of these tumours, the literature is replete with examples of missed lesions. Since most insulinomas are less than 2 cm in diameter, CT scan and ultrasound examinations fail to detect at least half of them. Many of these lesions are hypervascular and can be visualized by arteriography, with success rates of 15 to 85 per cent. Subselective injections are more likely to demonstrate the lesion than coeliac trunk or superior mesenteric artery injections. The false positive rate is as high as 10 per cent, presumably because of accessory spleens or overlapping shadows.

If the arteriogram fails to localize the adenoma, some authorities recommend proceeding to laparotomy, while others perform transhepatic portal venous sampling. This requires a highly experienced team of angiographers, and the best results are reported by groups with the largest experience. Although success rates of over 70 per cent are reported, the procedure is not without hazards, such as gallbladder perforation and haemobilia. Nonetheless, transhepatic portal venous sampling has theoretical attractions in being able to differentiate solitary adenomas from multiple lesions preoperatively, as well as directing the surgeon to one region of the pancreas if it is difficult to locate the lesion by palpation. At the time of surgery, intraoperative ultrasound may detect small lesions which are not readily palpable, particularly those in the head of the pancreas.

Once the diagnosis of insulinoma is made, surgery need not be urgent. Symptoms can often be controlled medically: euglycaemia may be maintained by frequent small meals. Diazoxide, 200 to 600 mg/day, suppresses insulin release through unknown mechanisms. Approximately 60 per cent of insulinomas respond to diazoxide. The long-acting somatostatin analogue, SMS201–995 (octreotide) also suppresses insulin secretion to variable degrees. Non-surgical control of symptoms preoperatively allows time for localization studies and evaluation for multiple endocrine neoplasia type I. In patients with hypercalcaemia, parathyroidectomy should be performed before pancreatic surgery. Elevation of serum pancreatic polypeptide is suggestive of a polyendocrine adenopathy.

At the time of laparotomy, the pancreas should be fully exposed to allow complete bimanual palpation. After dividing the gastrocolic ligament, a wide Kocher manoeuvre should be performed. Reflection of the right colon after division of the attachments to the lateral abdominal wall facilitates exposure and examination of the uncinate process. The lateral peritoneal attachments of the spleen should be incised to elevate the tail of the pancreas. The peritoneum along the inferior surface of the gland may then be incised to allow bimanual palpation of the body and tail of the gland.

In general, adenomas should be enucleated to preserve remaining pancreatic endocrine parenchyma. Between 20 and 30 per cent of patients with insulinomas will develop diabetes postoperatively. Lesions in the tail of the gland may be either enucleated or encompassed by a spleen-preserving distal pancreatectomy: the incidence of fistula formation following enucleation is equivalent to the morbidity of distal resection. If multiple adenomas are found, all gross disease should be resected. In patients with multiple endocrine neoplasia type I, a subtotal pancreatectomy may be performed, but total pancreatectomy is not indicated as an initial procedure, even when diffuse disease is present. When the lesion cannot be located intraoperatively and the diagnosis of insulinoma is biochemically certain, the presence of an ectopic insulinoma is excluded, and preoperative diazoxide has not controlled symptoms, blind distal resection of up to 85 per cent of the gland should be performed in two steps, with serial sectioning of the pancreas as it is removed. If the patient's glucose level has been controlled with diazoxide, however, the operation should be terminated and the patient sent to a centre capable of performing transhepatic portal venous sampling so as to direct subsequent resection (Fig. 3) 1344.

When malignant insulinomas are discovered, aggressive attempts to remove all gross disease, including hepatic metastases, should be undertaken. Hepatic lesions should be excised by wedge excision. Formal anatomical hepatic resections probably should not be undertaken unless preoperative chemotherapy has been given and has failed. Ligation of the hepatic artery should be avoided because regional infusion chemotherapy may be necessary in future management. Malignant insulinomas tend to be slow growing with a disease-free interval of up to 5 years following radical resection. Collective opinions hold that long-term remission following resection of hepatic metastases is rare and must be compared with regression of 50 per cent of tumours and complete remission in 20 per cent of patients obtained with streptozotocin treatment. Nonetheless, in competent hands, wedge resection of hepatic metastases is simple, carries a low morbidity, and may provide a long period of effective palliation. Chemotherapy of malignant insulinomas, like all islet cell tumours, is based primarily on streptozotocin. This will reduce tumour mass in 50 per cent of all patients, and 60 per cent will have a reduction in plasma insulin levels. Streptozotocin is fairly toxic: virtually all patients experience nausea and vomiting while two-thirds develop some renal toxicity and hepatotoxicity. The addition of 5-fluorouracil to streptozotocin markedly improves the response rate, with 33 per cent of patients showing a complete response and median survival improving from 16 months in patients treated with streptozotocin alone to 26 months in patients receiving combination chemotherapy.

The association of virulent peptic ulcer disease with a non-&bgr; pancreatic islet cell tumour was made in 1955 by Zollinger and Ellison. Subsequently the Zollinger–Ellison syndrome and gastrinoma have become synonymous. Gastrinomas are the second most common islet cell tumour and are found in patients of all ages without a significant predilection for sex, race, or socioeconomic status. Since the diagnosis rests on the radioimmunoassay of gastrin levels most reported cases are from cities with tertiary medical facilities.

Symptoms produced by gastrinomas are due to the effects of gastrin rather than to tumour mass. Gastric acid hypersecretion causes pain similar to typical peptic ulcer disease. One-third of patients, however, present with oesophagitis or diarrhoea due to hyperacidity. In the United States, 0.1 per cent of patients with duodenal ulcers have a gastrinoma. Patients diagnosed as having multiple peptic ulcers, ulcers in the postbulbar duodenum, ulcers resistant to medical therapy, or ulcers which relapse immediately upon cessation of medical therapy, should be suspected of having Zollinger–Ellison syndrome. Since 25 per cent of patients with gastrinomas have multiple endocrine neoplasia, a family history of endocrine disorders, hypercalcaemia, or ulcer diathesis should also raise suspicion that gastrinoma is present. Patients with multiple endocrine neoplasia type I generally present initially with hypercalcaemia.

The sine qua non in the diagnosis of gastrinoma is the demonstration of an elevated fasting gastrin level in the presence of acid hypersecretion. Prior to embarking on an extensive evaluation for gastrinoma, achlorhydria must be excluded. A basal acid output above 15 mEq/hour is highly suggestive of gastrinoma and basal acid output greater than 25 mEq/h is virtually diagnostic. Prior to the proliferation of radioimmunoassays, extensive acid studies were performed. Nonetheless, studies of acid secretion have a 10 per cent false-positive rate and a 15 per cent false-negative rate. Their value is to exclude achlorhydria and to monitor the efficacy of medical therapy.

Although a gastrin level greater than 500 pg/ml in a patient with hyperacidity is diagnostic of Zollinger–Ellison syndrome, up to 60 per cent of patients have gastrin levels in the intermediate range (150–500 pg/ml). Earlier detection of patients has made this scenario more common. To differentiate gastrinoma from other causes of hypergastrinaemia, a secretin test is performed. Secretin is a potent stimulant of gastrin release from gastrinomas, but secretin has little effect on serum gastrin levels in all other causes of hypergastrinaemia. A secretin stimulation test is performed by injecting 2 U/kg of secretin as an intravenous bolus and collecting blood samples at zero, 2, 5, 10, 15, and 30 min from the time of infusion for gastrin assay. An immediate rise of over 200 pg/ml is considered a positive test and is diagnostic for gastrinoma. Rarely, patients with Zollinger–Ellison syndrome will have a non-diagnostic secretin test (i.e. less than two-fold or 200 pg/ml increase in gastrin). In these patients, a combined secretin–calcium test may detect gastrinomas: a 1-min infusion of calcium gluconate together with a bolus of secretin will evoke the expected 200 pg/ml increase in gastrin concentration. These tests also serve to distinguish patients with a retained gastric antrum and G-cell hyperplasia from those with gastrinomas. Patients with G-cell hyperplasia have an exaggerated response to test meals (>100 per cent increase in serum gastrin) whereas patients with gastrinoma have only a moderate response to meals (Table 2) 395. Conversely, a test meal normally causes a two- to three-fold increase in plasma gastrin in normal patients, but there is only a minimal response in patients with gastrinoma because the high levels of circulating gastrin obscure the smaller physiologic response to the normal antral G-cells.

Gastrinomas occur in both sporadic and familial forms: about 25 per cent occur in patients with multiple endocrine neoplasia type I; tumours are usually multiple but only 30 per cent prove to be malignant. Eighty-eight per cent of these patients are hypercalcaemic. Gastrin production in many of these patients occurs in duodenal microadenomas and not in other areas of pancreatic islet cell hyperplasia. If these findings are confirmed, many patients with Zollinger–Ellison syndrome and multiple endocrine neoplasia type I, previously felt to be inoperable, might be helped by duodenal tumour resection.

When Zollinger–Ellison syndrome occurs in sporadic form tumours are unifocal, but over 60 per cent are malignant. Historically, 50 per cent of these patients have metastases at the time of diagnosis, but with the current widespread use of radioimmunoassays, small solitary tumours are being found at an earlier stage. Neither size nor histological appearance is capable of differentiating benign from malignant gastrinomas: only the presence of metastases can attest to their malignant potential. In contrast to insulinomas, gastrinomas are not evenly distributed throughout the pancreas; they are frequently extrapancreatic, 90 per cent being contained in an area known as the gastrinoma triangle (Fig. 4) 1345. This area is bounded by the junction of the cystic and common ducts superiorly, the junction of the second and third portion of the duodenum inferiorly, and the junction of the neck and body of the pancreas medially. Solitary extrapancreatic tumours in the duodenal mucosa or lymph nodes are found in up to 25 per cent of patients. Patients whose gastrinomas are found in lymph nodes, with no apparent primary tumour, are presumed to have occult lesions in the duodenal mucosa. This subgroup of patients has an excellent prognosis and can be cured when rendered eugastrinaemic by removal of affected lymph nodes.

Once the diagnosis of gastrinoma is made, acid secretion should be controlled with H&sub2;-receptor antagonists or omeprazole to allow time for localization studies. Localization should be undertaken in all patients, both to detect hepatic metastases and to aid the surgeon at laparotomy. Dynamic CT scanning is the single most accurate examination, detecting 75 to 80 per cent of all tumours. This result is comparable to those achieved using combined sonography, non-contrast enhanced CT, and selective arteriography. The sensitivity for detection of hepatic metastases is over 90 per cent. The success of dynamic CT scanning and selective arteriography is dependent to some extent on the experience of the radiological team.

In spite of these sophisticated imaging studies, a tumour will be found at laparotomy in nearly one-third of patients in whom all studies are negative. No preoperative test is superior to the ability of an experienced surgeon to find the offending gastrinoma. Intraoperative ultrasound may help to localize intrapancreatic tumours, but it is less helpful in identifying nodal gastrinomas and duodenal microadenomas. The value of transhepatic portal venous sampling is also debatable: it would seem to be of most use in patients in whom all of the localizing studies are negative. In such cases, the procedure should only be performed in centres with experience in the technique as the instance of false-positive results may be as high as 30 per cent.

Prior to the introduction of potent antisecretory drugs, surgical therapy was directed at preventing complications of ulcer disease by the performance of a total gastrectomy. The introduction of H&sub2;-receptor antagonists and, more recently, omeprazole initiated a wave of enthusiasm for non-surgical therapy in this disease. How[p094]ever, as patients have survived peptic complications only to succumb to metastatic disease, the importance of early tumour resection is now being re-emphasized. Zollinger reported a 76 per cent 5-year survival rate in patients with complete tumour resection, compared with a 21 per cent 5-year survival in those with incomplete resection. Since many tumours can now be detected earlier, curative resection may be possible in 40 per cent of patients, and all patients with apparent isolated gastrinomas and those in whom preoperative studies fail to demonstrate the site of a tumour should undergo surgical exploration. Extrapancreatic tumours, which frequently are not localized preoperatively, are the most curable; small intrapancreatic tumours may only be identified with careful surgical exploration and the use of intraoperative ultrasound. As with surgical exploration for insulinoma, the entire pancreas should be exposed and bimanually palpated. If no tumour is detected in the pancreas, all lymph nodes in the gastrinoma triangle should be examined and removed. If tumour is not detected by these manoeuvres, the duodenum should be opened and the mucosa everted and carefully inspected and palpated. Most microadenomas lie within 2 to 3 cm of the pylorus. Tumours in the body and head of the pancreas should be enucleated when possible. Distal pancreatectomy should be performed for lesions in the tail of the gland. Tumours located deep in the head or uncinate process may require a Whipple resection.

Occasionally no tumour is found at laparotomy: these patients may have either duodenal microadenomas, adenomatosis, or undiagnosed G-cell hyperplasia. The prognosis in these patients is excellent; indeed, they often outlive patients with resectable tumours. Patients with multiple endocrine neoplasia type I and no tumour demonstrable by preoperative studies often fall into this category; thus such patients should not undergo surgical exploration. However, patients with multiple endocrine neoplasia type I and a solitary lesion that can be localized preoperatively benefit from surgical excision of their gastrinomas.

The goal of medical therapy is to prevent acid-mediated complications. Because symptoms are notoriously unreliable indicators of the efficacy of acid suppression, therapy should be monitored with acid studies or periodic endoscopic examinations. Drug dosages should be adjusted to maintain the basal acid output at 10 mEq/h or less; high daily doses, such as 3.5 g of cimetidine or 1.2 g of ranitidine, are frequently required. Antiandrogenic side-effects of these drugs are common at such high doses and contribute to the 40 per cent failure rate of medical therapy. Omeprazole is the most efficacious agent for control of acid secretion, but since the long-term effects of achlorhydria are unknown, the dosage should be adjusted to maintain basal acid output above 0 mEq/h but below 10 mEq/h. Long acting somatostatin analogues, often used in conjunction with omeprazole also show promise. Thus, the pharmacological tools to control acid secretion are now available but co-operation of the patient and meticulous attention to his or her clinical and endoscopic status are necessary for successful treatment.

Malignant gastrinomas tend to be more indolent than other islet cell carcinomas, and the presence of metastatic disease at the time of initial diagnosis does not necessarily imply rapidly advancing disease. Five-year survival for patients with metastases ranges from 20 to 60 per cent. Metastatic gastrinomas, like other islet cell carcinomas, can secrete multiple hormones, and the surveillance of these patients should include awareness of complications induced by these other peptides. Surgical resection of isolated metastases may be beneficial if all tumour is removed. Most patients with metastatic gastrinoma, however, are candidates for chemotherapy. The best response rate is seen after the use of a combination of streptozotocin and 5-fluorouracil (as with insulinomas). Total gastrectomy should be reserved for patients whose tumour cannot be removed and whose hyperacidity cannot be controlled with medical therapy. With the advent of omeprazole treatment, this group of patients can be expected to become smaller. Nonetheless, total gastrectomy is extremely well tolerated in patients with Zollinger–Ellison syndrome, and can be performed with 0 to 5 per cent mortality under elective circumstances. Total gastrectomy remains the most certain way to eliminate all acid-related complications in Zollinger–Ellison syndrome.

Hypersecretion of glucagon from &agr;-cell neoplasms causes the ‘diabetes–dermatitis’ syndrome. While diabetes is mild, the rash, called necrotizing migratory erythema, is pathognomonic for glucagonoma. The rash involves the trunk, perineum, and thighs, and as the erythema spreads, central areas become necrotic and then heal. Other prominent features of this tumour, including weight loss, anaemia, and glossitis, are due to the catabolic effects of glucagon. Glucagon also has chronotropic activity; many patients manifest tachycardia, increased cardiac output, and constipation. The diagnosis, usually suspected on the basis of the rash, is confirmed by finding a plasma glucagon concentration greater than 50 pg/ml. Localization studies, as with insulinomas, are helpful in guiding surgical therapy and may be helpful in detecting metastases, as 50 per cent of tumours are malignant.

Preoperative correction of catabolism requires intravenous administration of amino acids, glucose, and supplemental insulin. Somatostatin analogues inhibit glucagon release and may palliate the rash. Prednisone is also beneficial. The surgical principles are similar to those described for insulinoma.

Vasoactive intestinal peptide is both an important neurotransmitter and hormone which is a potent stimulator of cAMP production by the gut, leading to massive secretion of water and electrolytes. The term ‘pancreatic cholera’ is an apt description of the fullblown effects of the tumour, with stool volumes sometimes being greater than 3 l a day. The diagnosis is made on the basis of watery diarrhoea, hypokalaemia, and achlorhydria in the presence of an elevated serum concentration of vasoactive intestinal peptide concentration. Flushing is also common because of the vasodilatory effects of the peptide. In the carcinoid syndrome, which has many similar symptoms, urinary levels of 5-hydroxyindole acetic acid are elevated and diarrhoea is less pronounced.

Ninety per cent of VIPomas lie within the pancreas; the remaining 10 per cent are found in neural tissue, including the adrenal medulla. Half of the tumours are malignant and 10 to 20 per cent are multifocal: localization studies are important to seek extrapancreatic tumours. Somatostatin analogues and indomethacin are useful in controlling diarrhoea, while dehydration and electrolyte abnormalities are corrected. Surgical principles are similar to those described for insulinomas.

Somatostatinomas are rare, less than 50 having been reported. Symptoms are vague, including diabetes, cholelithiasis, indigestion, steatorrhoea, and hypochlorhydria. These tumours are often found coincidentally at the time of cholecystectomy. Half of the reported tumours have been found in extrapancreatic locations such as the duodenum, jejunum, and biliary tree.

Non-functioning islet cell tumours present with symptoms caused by mass effect. If the tumour obstructs the pancreatic or bile duct, it may be detectable at a relatively early stage, but this is rare. The tumours are much larger than their functioning counterparts by the time they are diagnosed and thus they are more often malignant. In spite of their non-functional nature, immunoreactive cells are found in nearly 90 per cent of these tumours and 50 per cent contain multiple peptides. Their neuroendocrine origin is confirmed by finding neurone-specific enolase in the tumour, but there is little if any correlation between the immunohistochemical staining and clinical behaviour. All non-functioning islet cell tumours should be considered malignant because no gross or histological criteria, except the presence of metastases, can reliably establish the nature of the tumour. Aggressive surgical resection is indicated whenever this is feasible.

Delore R, Hermreck AS, Friesen SR. Selective surgical management of correctable hypergastrinema. Surgery 1989; 106: 1094–102.

Fajans SS, Vinik AI. Insulin producing islet cell tumors. Endocrinol Metab Clin N Am 1989; 18: 45–74.

Friesen SR. Update on the diagnosis and treatment of rare neuroendocrine tumors. Surg Clin N Am 1987; 67: 379–93.

Howard TJ, Zinner MJ, Stabile BE, Passaro E. Gastrinoma excision for cure: a prospective analysis. Ann Surg 1990; 211: 9–14.

Legaspi A, Brennan MF. Management of islet cell carcinoma. Surgery 1988; 104: 1018–22.

Pipeleers-Marichal M, et al. Gastrinomas in the duodenum of patients with multiple endocrine neoplasia Type 1 and the Zollinger Ellison syndrome. N Engl J Med 1990; 322: 723–7.

Proye C. Surgical strategy in insulinoma of adults. Acta Chir Scand 1987; 153: 481–91.

Zollinger RM, Ellison EH. Primary peptic ulcerations of the jejunum associated with islet cell tumours of the pancreas. Ann Surg 1955; 142: 709–28.