The incidence of pancreatic cancer in the United States was 9.2 per 100 000 in 1987: this represents a significant increase over that found in 1940, but has remained essentially unchanged for the last 15 years. Other developed countries have observed similar trends in the incidence of this disease. Pancreatic cancer is the second most common gastrointestinal malignancy after colorectal cancer, which affects five times more people. However, while colorectal cancer has a 5-year survival rate of over 50 per cent, that of pancreatic cancer is only 3.1 per cent: it is, in fact, the malignancy with the lowest 5-year survival.

Malignant neoplasms of the pancreas can occur at any age, but they are rare before the age of 40: the mean age of diagnosis is 64 years, after which the incidence increases rapidly.

The aetiology of pancreatic cancer is largely unknown; however, epidemiological studies have shown that several environmental factors are likely to be implicated. Cigarette smoking is the best established risk factor: smokers are two to three times more likely to develop pancreatic cancer than non-smokers, and they develop the disease at a median age 15 years younger. There is a direct relationship between the number of cigarettes smoked and the increase in risk.

The next most important risk factor is probably diet, which could also account in part for the wide variations in incidence found in different geographical regions, as well as for the altered risk that has been observed in certain migrant populations. A direct correlation between dietary fat intake and occurrence of pancreatic cancer has consistently been found in different populations, and the addition of unsaturated fats to the diet of azaserine-treated rats promotes development of pancreatic tumours. On the other hand, a high intake of fresh fruits and vegetables is associated with a significantly decreased incidence of pancreatic cancer in man. This protective effect has also been found with consumption of proteins of vegetable origin, such as beans and lentils, which contain high levels of protease inhibitors.

Exposure to several chemical agents, including naphthylamine, benzidine, and petrol has also been linked to pancreatic cancer. Workers from industries that produce or use these chemicals have been found to have up to a fivefold increase in mortality from this disease. Pancreatic cancer can be induced in rats with azaserine (O-diazoacetyl- l-serine) and in Syrian hamsters with nitrosamines.

Diabetes mellitus and chronic pancreatitis are both associated with pancreatic cancer, although the nature or even the direction of the cause–effect relationship has not been established. At least 15 per cent of patients with pancreatic cancer are discovered to have diabetes or have had diabetes diagnosed within the year prior to diagnosis of the cancer. It is unlikely that such new diabetes represents destruction of the pancreas by the tumour sufficient to cause endocrine insufficiency. A true association between pancreatic cancer and chronic pancreatitis is probably confined to familial and tropical forms of chronic pancreatitis.

Throughout the world, pancreatic cancer affects more men than women, with a male to female ratio of between 1.5 and 2. In experimental carcinogenesis, growth of azaserine-induced tumours is stimulated by androgens and is inhibited by antiandrogens or castration. Male patients with pancreatic cancer have low serum testosterone and high androstenedione levels and also show an altered testosterone:dihydrotestosterone ratio. These findings seem to be a consequence rather than the cause of the neoplastic disease: normal values are restored after resection of the tumour.

Much public and scientific controversy has been generated around the possible link between coffee consumption and pancreatic cancer. The first study to propose such an association appeared in 1981, and since then, more than 25 studies have attempted to answer this issue, with conflicting results. The majority of studies do not support a causal relationship between coffee and pancreatic cancer, but suggest rather that cigarette smoking, dietary habits, and other confounding factors were responsible for the association found in the earlier studies.

Prior surgery in the alimentary tract has also been implicated as a causative factor. Patients with a history of gastrectomy have at least a threefold greater risk of dying from pancreatic cancer. A relationship with cholecystectomy has also been proposed. The basis for these observations is not known.

The increased incidence of pancreatic cancer in certain populations such as Polynesians and the American black population suggests that genetic factors are also involved in its genesis.

Malignant tumours of the pancreas can occur in either the exocrine parenchyma or in the endocrine cells of the islets of Langerhans (Table 1) 392, but exocrine neoplasms are far more common. Adenocarcinoma accounts for around 80 per cent of pancreatic neoplasms, and is thought to be of ductal origin. It is twice as frequent in the head of the pancreas as in the body or tail, and at the time of diagnosis more than 85 per cent of these tumours have extended beyond the limits of the organ. Perineural invasion within and beyond the gland is particularly prominent in this type of cancer, although lymphatic spread also leads to early metastasis to adjacent and distant lymph nodes. The most common sites of extralymphatic involvement are the liver and peritoneum. The lungs are the most frequently affected of the extra-abdominal organs. Rare carcinomas of the pancreas include giant cell, adenosquamous, and acinar cell varieties, all of which have a similar or worse prognosis than does ductal adenocarcinoma.

Malignant or potentially malignant cystic tumours of the exocrine pancreas, which include the mucinous cystadenoma–cystadenocarcinoma and the papillary–cystic tumour, behave quite differently. Most patients are women; in particular, papillary–cystic tumours almost exclusively affect young women. Both of these tumours tend to be large (>6 cm diameter on average) at the moment of diagnosis, and this fact, together with their cystic appearance on scan and at operation, usually allows them to be easily differentiated from ductal adenocarcinoma. The prognosis for these cystic neoplasms is very good if they are completely resected. The 5-year survival rate for all cystadenocarcinomas is 50 per cent, but reaches 75 per cent for the two-thirds of patients whose tumour is completely removed. The great majority of papillary–cystic tumours (up to 95 per cent) are cured.

Another pancreatic malignancy with a relatively favourable prognosis is lymphoma, which accounts for up to 5 per cent of all malignant pancreatic tumours in selected series and which is amenable to chemotherapy and irradiation.

The pancreas can also be the site of metastasis from malignant neoplasms of other organs. In autopsy studies, for every primary cancer of the pancreas there are three cases of metastatic carcinoma, breast and lung being the most common sites of origin.

At the time of presentation, most patients with pancreatic cancer have had symptoms for less than 4 months. Their principal complaints are pain, jaundice, and weight loss, usually developing insidiously. Pain is most commonly located in the epigastrium or left upper quadrant of the abdomen; it has a dull, aching nature and can radiate to the back. The patient may experience some relief when lying or sitting in a flexed or curled position. The pain becomes more severe as the disease advances and can become troublesome and difficult to control.

About 50 per cent of patients present with jaundice due to common bile duct obstruction. The jaundice is progressive and rarely regresses, is painless in up to one-third of cases, and is likely to be accompanied by pruritus.

Weight loss can be prominent; it is a result of decreased appetite, malabsorption, and other unknown factors which cause cancer cachexia. In advanced stages duodenal obstruction and perhaps disturbances of gastrointestinal motility from splanchnic nerve invasion can also be contributory.

Other symptoms include psychiatric disturbances, particularly depression, which are seen in up to 75 per cent of patients. Diabetes mellitus is present in 15 per cent, and is of recent onset in over 50 per cent of cases in which both entities are associated. An episode of acute pancreatitis is occasionally the first indication of the disease, but there is some sign of pancreatic inflammation in up to 14 per cent of cases. Gastrointestinal bleeding may also occur; this is most commonly secondary to gastric or duodenal invasion by the tumour, but may also originate from varices, which are the consequence of portal or splenic vein occlusion. Other rare manifestations include migratory thrombophlebitis and the appearance of tender subcutaneous nodules or polyarthritis secondary to metastatic fat necrosis. The last two have been particularly associated with acinar cell carcinomas.

In the early stages of pancreatic cancer, physical signs include jaundice, evidence of weight loss, and hepatomegaly, which reflects bile duct obstruction. A non-tender gallbladder can be palpated in 50 per cent of jaundiced patients (Courvoisier's sign). In advanced disease, ascites and a palpable mass are indicative of an unresectable tumour.

The two methods most frequently used to confirm a clinical suspicion of pancreatic cancer are ultrasonography and computerized tomography (CT). Both of these can demonstrate pancreatic masses, dilation of the pancreatic duct as well as of the bile duct and gallbladder, hepatic metastasis, and extrapancreatic spread. An ultrasound scan usually shows pancreatic tumours to be less echogenic than the surrounding parenchyma, and accompanied by changes in the contour of the gland. Overall, the diagnostic sensitivity of this diagnostic method is about 70 per cent, and its specificity 95 per cent. Ultrasonography has the inconvenience that it relies heavily on the experience of the examiner and that a satisfactory scan cannot be obtained in about 20 per cent of patients because of obesity or the presence of intestinal gas.

Computerized tomography has a higher sensitivity than that of ultrasound, with a similar specificity; it should probably be used in all patients in whom a strong suspicion of pancreatic cancer exists. Diagnosis is also based on the presence of focal enlargement of the gland (Fig. 1) 1332. Additional signs, such as dilation of the pancreatic duct in the absence of intraductal or parenchymal calcification, can be highly suggestive of a pancreatic or ampullary malignancy. Neither CT nor ultrasonography are capable of identifying tumours less than 2 cm in diameter.

Magnetic resonance imaging has no advantage over CT. Recent reports have also described encouraging results with the use of endoscopic ultrasonography, especially for tumours located in the head of the pancreas.

Endoscopic retrograde cholangiopancreatography is a valuable tool in the differential diagnosis of the cause of obstructive jaundice when pancreatic or other periampullary cancer is suspected. In certain cases, when the initial manifestation of the carcinoma is an episode of acute pancreatitis, this technique may demonstrate small tumours that could not otherwise be detected by CT. Virtually all pancreatic cancers show abnormalities in the pancreatogram, these consisting mainly of stenosis (Fig. 2) 1333 or obstruction of the pancreatic duct. In some patients, these findings may be difficult to differentiate from those of chronic pancreatitis. The study also permits visualization of the bile duct and placement of stents for the relief of jaundice. In selected cases, especially those in which the main suspicion is a proximal bile duct carcinoma, rather than a pancreatic or periampullary neoplasm, percutaneous transhepatic cholangiography may be preferred. This study is also used when endoscopic retrograde cholangiopancreatography fails or is not practicable due to the presence of anatomic barriers arising from prior gastric surgery.

The differential diagnosis of malignancies around the pancreatobiliary junction includes tumours of the head of the pancreas, ampulla, lower third of the bile duct, and duodenum. Pancreatic cancer is by far the most common of these, accounting for about 80 per cent of cases, followed by ampullary cancer (10 per cent). Certain clinical characteristics may aid in the differential diagnosis, such as the high frequency of intestinal bleeding in patients with ampullary and duodenal tumours, or the fluctuation of jaundice seen in some patients with ampullary carcinoma. Endoscopy and retrograde cholangiopancreatography are the mainstays in this differentiation, although in some patients definite diagnosis is not obtained until the tumour is resected.

Chronic pancreatitis can present a clinical and radiological picture indistinguishable from that of pancreatic cancer, and is the main benign condition from which this neoplasm has to be differentiated. Calcifications are uncommon in pancreatic carcinoma; their presence favours the diagnosis of chronic pancreatitis. However, more than 50 per cent of patients with chronic pancreatitis have no calcifications, and a diagnosis of malignancy cannot be definitely excluded. Although extremely rare, tuberculosis and sarcoidosis can also present as a pancreatic mass associated with pain or jaundice.

Several studies have been directed at the detection of serum markers to aid in this differential diagnosis of pancreatic cancer, as well as its diagnosis at subclinical stage. Such markers include tumour-associated antigens, enzymes and endocrine markers. The most extensively studied of these is the tumour-associated antigen CA19–9, the presence of which has a reported sensitivity for pancreatic cancer of over 90 per cent, but a relatively low specificity of about 75 per cent. Most of the false positive results occur in patients with other gastrointestinal malignancies, particularly ampullary and biliary carcinomas. One of the major drawbacks of this serum marker in common with most others, is that it is most likely to be positive in the advanced stages of the disease and most frequently negative in patients with the small tumours which are the easiest to cure. Two exceptions to this may be elevated levels of immunoreactive elastase and a testosterone: dihydrotestosterone ratio of 5 or less. These have been found to be positive in a high proportion (albeit or a small number) of stage I pancreatic cancers. A testosterone: dihydrotestosterone ratio of 5 or less is highly specific for pancreatic carcinoma and could therefore be useful in its differential diagnosis from chronic pancreatitis.

A definite diagnosis of pancreatic carcinoma can only be made following histological examination. Material obtained by percutaneous fine-needle aspiration of the tumour for cytological examination is invaluable, particularly for patients with advanced stage pancreatic cancer, who otherwise would require surgical exploration. The procedure is usually performed under direct guidance by ultrasonography or CT. A positive result reliably confirms the diagnosis of malignancy, but a negative one does not exclude it since there is a 10 to 20 per cent false negative rate. Percutaneous biopsy should not be used in patients with potentially resectable tumours since seeding of cancer cells along the needle track has been documented. There is also evidence that intraperitoneal spread may occur following this procedure.

With few exceptions, only patients presenting with jaundice ever have a cancer which is potentially curable by resection; in fact, only 10 to 20 per cent of jaundiced patients have a resectable cancer. In the great majority of patients only a palliative procedure such as a biliary or duodenal bypass is feasible. Traditionally, most patients with pancreatic cancer underwent surgery to obtain tissue for diagnosis and to assess resectability. In the past few years this approach has been changing, as new techniques for percutaneous biopsy and preoperative evaluation have evolved, and preoperative staging is becoming one of the most important aspects in the management of patients with this neoplasm.

The goal of preoperative staging of pancreatic cancer is to ascertain the optimal treatment option for each patient. Specifically, the aim is to delineate which patients have tumours that are potentially resectable, which still have localized cancer, and which already have distant metastases. These facts in turn allow effective triage of patients, as well as a choice of therapeutic approach. Pancreatoduodenectomy in inexperienced hands can cause considerable morbidity and up to 40 per cent mortality, although many specialized centres are currently reporting vastly improved results with less than 5 per cent mortality. It may therefore be appropriate to stage patients locally but to consider transfer to a regional centre if resection seems appropriate. On the other hand, patients who have a demonstrably unresectable cancer have a median survival of about 6 months. Many of these patients should be spared surgery. A tissue diagnosis can be obtained by percutaneous needle aspiration. Stents can be endoscopically or percutaneously placed to relieve biliary obstruction with efficacy equal to surgical bypass. Only patients with duodenal obstruction invariably require surgical intervention.

The staging classification for pancreatic cancer proposed by the American Joint Committee for Cancer is shown in Table 2 393. In some patients, the ultrasound or CT scan that was used to confirm the clinical diagnosis will already have demonstrated the presence of liver metastases or a large tumour with major vascular encasement or occlusion, thus effectively ending staging once tissue diagnosis is obtained percutaneously. If an initial ultrasound screening reveals no metastasis or major vascular invasion a CT scan should be performed: this gives a better outline of the pancreatic boundaries and is more sensitive in detecting liver and large nodal metastasis.

In our practice, the next step in staging is laparoscopy (Fig. 3) 1334. Laparoscopy will show evidence of hepatic or peritoneal spread in more than 40 per cent of patients, even when a CT scan is negative. Typically, the lesions seen at laparoscopy in these patients are 1- to 2-mm nodules on the surface of the liver, parietal peritoneum, or omentum—too small to be detected by any other means. Laparoscopy also allows for cytological examination of peritoneal washings, which, if positive (Fig. 4) 1335, indicates probable unresectability and a poor prognosis.

If laparoscopy is negative, the next study to be performed is angiography of the coeliac trunk and superior mesenteric artery, which will detect extrapancreatic arterial and especially venous encasement. Obstruction or narrowing of the portal (Fig. 5) 1336, splenic or mesenteric veins or of the coeliac, hepatic, or superior mesenteric arteries is usually an indication of unresectability, although some groups excise and reconstruct these vessels. Angiography also gives valuable information on the vascular anatomy of the patient. An anomalous origin of the hepatic arteries is seen in about one-third of patients, and preoperative knowledge of this allows operative injury to these structures to be avoided.

CT scan, laparoscopy, and angiography give different indications of tumour resectability and are complementary. In our experience, 78 per cent of pancreatic head tumours can be resected in patients in whom all three tests are negative; if one or more of the three is positive, the resectability rate is only 5 per cent. In the future endoscopic ultrasonography may be refined enough to add another dimension to preoperative detection of lymph node metastases and extrapancreatic extension, thereby improving staging accuracy even further. It is also possible that future studies will establish the finding of a positive peritoneal cytology as a definite index of disseminated disease, even in the absence of obvious macroscopic metastasis, thereby precluding further consideration of resection.

While the cure rate for pancreatic cancer is admittedly low, resection offers the only possibility of obtaining disease-free long-term survival. Radiotherapy, chemotherapy, and hormone therapy are sometimes used as adjuvants to resection, but they have practically no impact on survival when used alone. Pancreatoduodenectomy has been the standard operation for carcinoma of the pancreatic head since its demonstration by Whipple in 1935. Variations on this procedure include total, subtotal, and radical pancreatectomy, as well as the pylorus-preserving pancreatoduodenectomy. Carcinomas of the body and tail of the pancreas are somewhat less common; they are usually discovered at an advanced stage when they are not amenable to resection. Rarely, a small resectable tumour may be found in this location, usually by chance. Tumours to the left of the portal vein can theoretically be treated by distal pancreatectomy.

Many patients will undergo surgery before a histological diagnosis has been established. Traditionally, one of the first steps has been to establish the diagnosis of carcinoma, commonly by transduodenal needle biopsy of the tumour: this may be time consuming as well as a potential source of complications. Now that the morbidity and mortality of pancreatoduodenectomy are so low preoperative confirmation of malignancy is no longer mandatory. A biopsy is needed only if the surgical procedure is likely to differ significantly according to the diagnosis. However, the most commonly encountered diagnostic difficulty is differentiating between chronic pancreatitis and carcinoma, both of which will be well served by a pancreatoduodenectomy. Resection of the pancreatic head is therefore the procedure of choice both for cancer and for situations of diagnostic uncertainty.

If a diagnosis is required during surgery, tissue can be obtained by either incisional or needle biopsy of metastases or of the primary tumour mass. In some centres fine-needle aspiration has replaced needle-core biopsies for obtaining material for cytological examination because of the possibility of obtaining multiple samples with a lower complication rate. Whenever feasible, transduodenal needling is preferred to direct sampling.

The first step of the procedure is assessment of resectability. This entails a thorough exploration of all of the abdominal organs and peritoneal surfaces, extensive mobilization of the duodenum and head of the pancreas (Kocher's manoeuvre), and separation of the posterior neck of the pancreas from the superior mesenteric vein. Enlarged peripancreatic, periportal, and coeliac lymph nodes that might harbour metastases are biopsied. If there is no evidence of metastases, gross extrapancreatic invasion of the retroperitoneal structures or attachment of the cancer to mesenteric vessels, resection is started. The gallbladder is removed and the elements of the hepatic hilum are dissected. The common hepatic duct is divided above the cystic duct entry and its distal stump ligated. The fatty and lymphatic tissue anterior and lateral to the portal vein, as well as that surrounding the hepatic artery are dissected down towards the pancreas; both vessels are skeletonized. Care must be taken to preserve the right hepatic artery, which courses separately and to the right of the portal vein in 15 per cent of individuals. The gastroduodenal artery is then divided and ligated at its origin from the hepatic artery. This preparation allows further exposure of the portal vein to the point where it emerges above the pancreatic neck, thus completing the dissection behind the pancreatic neck to the superior mesenteric vein. Resection is continued by dividing the gastrohepatic omentum and reflecting this tissue down to the gastric antrum, dividing the right gastric and distal left gastric arteries. The gastrocolic omentum is opened and the lesser sac developed. After the gastroepiploic vessels have been ligated at the greater curvature the stomach is divided at the junction of the corpus and the antrum. The pancreas is then transected anteriorly or immediately to the left of the portal vein, and haemostasis of the distal stump carefully obtained with cautery and sutures. The proximal jejunum is divided, as is its mesentery down toward the ligament of Treitz, which is lysed. The proximal end of the jejunum is passed to the right under the mesenteric vessels, and the final steps of resection are then accomplished. The uncinate process of the pancreas must be dissected from the portal and mesenteric veins, carefully isolating and ligating the thin walled venous communications (Fig. 6) 1337. Finally, the retroperitoneal tissue and arterial branches to the uncinate from the superior mesenteric artery must be severed and ligated. After removal of the specimen, the peritoneum is closed at the ligament of Treitz.

Reconstruction of the alimentary tract is achieved by passing the distal jejunum through the right side of the transverse mesocolon and anastomosing the pancreas, bile duct, and stomach to the intestine, usually in that order (Fig. 7) 1338. It is thought that the alkaline pH of the pancreatic and biliary secretions are protective against marginal ulceration at the gastroenteric anastomosis, and also, if a troublesome leak develops from the pancreatojejunal anastomosis, the pancreatic remnant can be excised without disturbing the other connections. Pancreatojejunostomy can be carried out by invaginating the pancreatic remnant into the jejunum, or, as we prefer, by performing an end-to-side anastomosis approximating the mucosa of the pancreatic duct to that of the intestine. Interrupted fine polyglactin sutures are used for the mucosal anastomosis, which is routinely stented with a 5F paediatric feeding tube, brought out through the side wall of the jejunum and through the abdominal wall. An outer layer of interrupted silk sutures is also placed between the pancreas and the serosa of the jejunum. The biliary anastomosis is constructed in a similar fashion, but stents are generally unnecessary for the dilated bile duct. An antecolic gastrojejunostomy completes the reconstruction. Closed system soft silicone suction drains are left in the vicinity of the pancreatic and biliary anastomosis.

Complications of pancreatoduodenectomy are common: the most frequent are pancreatic (10–20 per cent) and biliary (5–10 per cent) fistulae, wound haemorrhage, cardiopulmonary events, and infection. Pancreatic fistulae are the principal source of postoperative morbidity and mortality, but the technique of stented mucosa–mucosa anastomosis has lowered our incidence of pancreatic fistula to 3 per cent (four of 140 cases), and has been innocuous in all but one case. Biliary fistulae are equally rare and will usually close spontaneously. Persistent fistulae may heal with the help of a percutaneous transhepatic stent bridging the anastomosis. Late postoperative haemorrhage is usually the consequence of vascular erosion associated with intrabdominal sepsis, most commonly related to a pancreatojejunal or other anastomotic leak. Other complications include wound abscess (which has become uncommon with routine use of antibiotic prophylaxis) and urinary and respiratory infections.

Mortality rates following the Whipple procedure have decreased markedly over the past 30 years. Prior to 1980 the average mortality in over 4000 patients from different sources was 17 per cent; however, in the last 5 years, at least six major institutions in the United States and Europe have reported substantial experiences with mortality rates of less than 5 per cent.

Survival rates 5 years after Whipple resection for pancreatic cancer have significantly improved: while such survival was rare 25 years ago, current reports quote figures between 5 and 18 per cent. In a series from Japan, the 5-year survival rate for patients with tumours less than 2 cm in diameter was 30 per cent, and 20 per cent of all those who underwent resection survived this long. The median survival of resected patients is about 18 months, which is significantly better than the figure of 6 months quoted for unresected patients, including those in whom the tumour was small and apparently confined locally. These observations suggest that resection of the tumour not only gives reasonable (and the only) hope of cure, but also the best palliation.

This modification of the Whipple procedure was first proposed by Traverso and Longmire in 1978. The antrum, pylorus, and first 2 cm of duodenum are preserved (Fig. 8) 1339 with the intention of decreasing the frequency of marginal ulceration and improving postoperative nutritional status by avoiding postgastrectomy symptoms. The concern that this procedure might compromise the resection margins and the removal of potentially malignant lymph nodes, and therefore lessen the effectiveness of the resection, has not been borne out. Long-term survival has been equivalent to that of the conventional Whipple procedure. We currently use this operation in the treatment of carcinoma of the pancreatic head, provided that the first portion of the duodenum is not affected by the tumour and that the patient has no history of duodenal ulcer.

In an effort to improve the results of the Whipple procedure in the treatment of carcinoma of the pancreatic head, several groups started to perform total pancreatectomy in the 1970s. The arguments for this modification were that tumour was found either in the resection margin or distal to it in a multicentric fashion in up to one-third of patients, and that in extending the pancreatectomy a more complete lymphadenectomy could be achieved. It was also hoped that postoperative morbidity and mortality would be diminished by eliminating the pancreatic anastomosis. In a review of more than 350 patients reported by different surgical groups, postoperative mortality was 17 per cent, mean survival 18 months, and 5-year survival 6.3 per cent. Thus, no advantage has been found to this approach, which also causes labile diabetes because of the absence of both insulin and glucagon. In hopes of avoiding this difficult diabetic state, some groups in Europe leave a small fragment of the pancreatic tail which is oversewn. The frequency of postoperative diabetes is less than that after total pancreatectomy and is easier to manage, but long-term results are not yet available. Total pancreatectomy should probably be reserved only for those resectable cases in which there is obvious involvement of the rest of the organ or in which tumour is found in the frozen section of the margin of the resected pancreas.

In this operation, the pancreas and surrounding tissue are resected en bloc, along with a segment of the portal vein and, occasionally, segments of the hepatic or superior mesenteric arteries. The procedure entails either total or subtotal pancreatectomy. Its main advocate is Fortner, who described it in 1973. His operative mortality is now down to 8 per cent, but his long-term results are no different to those obtained with conventional pancreatoduodenectomy.

At the time of diagnosis curative surgical resection is already impossible for 90 per cent of patients with pancreatic cancer. Although this can usually be established by preoperative staging, unresectability may not be established before surgical exploration. Decompression of the biliary duct is indicated for relief of symptoms due to obstruction. Either cholecystojejunostomy or choledochojejunostomy may be suitable, but the latter is associated with a lower incidence of cholangitis and recurrent jaundice due to obstruction of the cystic duct by the tumour growing along the common duct. Anastomosis of the bile duct to the duodenum rather than the jejunum also works well; it is rarely prevented by cancer involving the proximal duodenum. For jaundiced patients in whom an unresectable or metastatic tumour is diagnosed preoperatively, and in those in whom tumour in the porta prevents access to the common duct, endoscopic stenting (Fig. 9) 1340 is the best option, especially now that large diameter tubes (10–12 Fr) can be used. Survival following stent compression is equal to that after surgical bypass, while cost and morbidity are reduced.

Biliary drainage achieved by a percutaneous stent is an option in patients for whom an endoprosthesis cannot be placed endoscopically. The time required to achieve placement of percutaneous endoprosthesis is greater than that for endoscopic placement, and complications are more common.

Obstruction of duodenal or gastric outlets can be another source of considerable distress for patients with inoperable pancreatic cancer and gastrojejunostomy is required. It has been stated that these patients have associated duodenal motility disorders which complicate the early postoperative course. For this reason, some surgeons insert a gastrostomy drainage tube at the time of gastrojejunostomy. Between 8 and 30 per cent of patients treated for biliary obstruction alone subsequently develop duodenal obstruction, perhaps justifying prophylactic gastrojejunostomy at the time of surgical biliary bypass.

In an effort to ameliorate pain from retroperitoneal nerve invasion, some surgeons perform chemical splanchnicectomy (chemoneurolysis) by infiltrating the coeliac plexus with phenol or 50 per cent ethyl alcohol at the time of palliative surgery. This may be of benefit for several months. Nerve blockade can also be achieved percutaneously, although the results may not be as durable.

External beam radiation therapy with 40 to 60 Gy may produce some palliation of pain for patients with unresectable pancreatic cancer, but survival is not significantly prolonged. The combination of radiotherapy with 5-fluorouracil was evaluated by the Gastrointestinal Tumor Study Group. The combined therapy produced a median survival of 10 months, patients who received only external beam radiation therapy survived 5.5 months. This combined modality was also tested in patients who underwent resective surgery with a curative intent, and the treated group had a significantly increased median survival of 21 months compared with 11 months in the untreated group. Although the number of patients treated was small, the findings suggest that combined adjuvant radiation and 5-fluorouracil can be of benefit after curative resection.

Radiation can be delivered to a pancreatic tumour in greater effective doses and with less injury to neighbouring viscera if it is performed during surgery. This objective has been accomplished with conventional orthovoltage, electron beams, and implantation of radioactive iodine. Only the electron beam therapy is said to prolong survival, and this treatment also appears to give substantial and lasting relief of pain in at least one-half of treated patients. Late duodenal ulceration and stenosis may result from radiation injury, and for this reason all patients treated with intraoperative electron beam also receive a gastrojejunostomy.

Several chemotherapeutic agents have been used in the treatment of pancreatic cancer, but only 5-fluorouracil, mitomycin C, streptozotocin, and ifosfamide have shown response rates above 20 per cent. The most extensively studied drug is 5-fluorouracil, which has a response rate of 28 per cent. However, the duration and value of the ‘response’ are minimal, and the end results of both single and combination drug treatment have been uniformly disappointing.

Experimental pancreatic carcinoma is responsive to hormone manipulation, and human tumours bear sex steroid receptors. This, coupled with the androgen derangement (low testosterone, high androstenedione, and altered testosterone to dihydrotesterone ratio) seen in patients with this neoplasm, has prompted several groups to investigate endocrine treatment. Two phase II studies of tamoxifen treatment in patients with unresectable pancreatic cancer demonstrated an increase in median survival, as did a study with the use of an analogue of luteinizing hormone-releasing hormone. Phase III trials are currently in progress.

Cameron JL, et al. Factors influencing survival following pancreaticoduodenectomy for pancreatic cancer. Am J Surg 1991; 161: 120–25.

Grace PA, Pitt HA, Longmire WP. Pylorus preserving pancreatoduodenectomy: an overview. Br J Surg, 1990; 77: 968–74.

Greenway BA. Carcinoma of the exocrine pancreas: a sex hormone responsive tumour? Br J Surg 1987; 74: 441–2.

Gudjonsson B. Cancer of the pancreas; 50 years of surgery. Cancer 1987; 60: 2284–303.

Fontham ETH, Correa P. Epidemiology of pancreatic cancer. Surg Clin North Am 1989; 69: 551–67.

Warshaw AL. Implications of peritoneal cytology for staging of early pancreatic cancer. Am J Surg 1991; 161: 26–30.

Warshaw AL, Swanson RS. Pancreatic cancer in 1988. Possibilities and probabilities. Ann Surg 1988; 208: 541–53.

Warshaw AL, Zhuo-yun G, Wittenberg J, Waltman AC. Preoperative staging and assessment of resectability of pancreatic cancer. Arch Surg 1990; 125: 230–3.