The ancient Egyptian papyrus Ebers describes a patient with dysphagia in whom the stomach had the appearance of a shrivelled fetal face. This may well represent the first description by Egyptian physicians of a gastric carcinoma, a disease that is now a major health care and surgical problem.

As is the case with many neoplasms there is a widespread variation in the incidence of gastric carcinoma around the world. Gastric cancer occurs in almost epidemic proportions in Japan with an incidence 70 per 100000 males and a cumulative risk of developing the disease by the age of 75 of 11 per cent. It is said to be the ‘national disease’, accounting for nearly 52 per cent of deaths from malignancy in men and 38 per cent in women in 1960. Other countries with a high incidence of the disease (over 30 per 100000 males) include Chile, Costa Rica, Hungary, Poland, Portugal, Iceland, Rumania, Indonesia, and Italy. There is a particularly low incidence in some central African countries (Uganda).

Marked regional variations have been reported. In Columbia, an incidence of 50 per 1000000 males occurs in the inland city of Cali, compared with 12 per 100000 males on the coast at Cartagena. The rural population in Poland is at greater risk than the urban population, in contrast to the situation in other cancers.

In the United States and the United Kingdom the overall incidence has fallen over the last 50 years. In the United States the age adjusted mortality (per 100000) for men fell from 28 in 1935 to 9.7 in 1967 and to 8.2 in 1986. This decline has occurred in all racial groups and in both sexes, and reflects the change in incidence and not earlier diagnosis, better treatment, or changes in definition. Although a falling incidence has been noted in most countries, there has been no decline in Poland or amongst Japanese men. There has been an increase in the number of deaths reported in both men and women in Portugal.

Carcinoma of the stomach is closely correlated with age, occurring predominantly between the fifth and seventh decade and in people in the lower socioeconomic groups (social class III and IV). In countries with a high incidence of disease the peak incidence tends to occur at an earlier age than is seen in low-risk areas. It is universally more common in men than women. In the United Kingdom the male : female ratio is 1:1 for young adults, rising to 2:1 in the sixth decade. Variations in geographical incidence are not incompatible with environmental factors playing an important role: first-generation migrants sustain the risk rate of their country of origin, but the mortality rate in subsequent generations of Polish and Japanese immigrants to the United Stares falls to an intermediate level.

As well as a fall in the incidence of gastric carcinoma in the United States and United Kingdom there has been an impression that the tumour is moving proximally, with a rising incidence of adenocarcinoma of the cardia. In one series from the United Kingdom the number of carcinomas reported in the upper third of the stomach rose from 17 per cent (1951–55) to 39 per cent (1981–85). This was accompanied by a reduction in tumours of the middle third (31 per cent to 19 per cent) and lower third of the stomach (51 per cent to 47 per cent). Similarly in Japan, the proportion of proximally located cancer has increased from 10 per cent to 40 per cent of diagnosed tumours. However, antral and prepyloric tumours remain the most common, followed by those in the body and fundus. There is some evidence that tumours in the fundus are more aggressive, with a greater tendency to submucosal invasion regardless of the histological type. The gross morphological difference in the fundus that may explain this is the thinner muscularis mucosa and the presence of tightly packed glands that might block lateral growth. In addition signet ring carcinoma is more common in the fundus. Proximally placed tumours generally have a worse prognosis than those placed distally. This phenomenon may be a result of the features described above or because they tend to be at a more advanced stage at presentation.

There is clear evidence of familial clustering of carcinoma of the stomach, the most famous being that of the Bonapartes. Napoleon, his father, his grandfather, brother, and three sisters died of gastric carcinoma. However, only 4 per cent of patients have a family history of the disease, and there is slightly greater correlation between identical rather than fraternal twins.

Aird, in 1953, described an association between blood group A and gastric carcinoma: the relative risk over patients with blood group O is 1.2 times. This difference has been related to the nature of mucopolysaccharide secretion in the stomachs of group A patients, and greater susceptibility to ingested carcinogens. The cancer that develops is of the diffuse type (Lauren classification), there being no increased incidence of the intestinal type.

Although genetic factors may be important, environmental and dietary agents are of far greater importance. Various foodstuffs have been implicated in the aetiology of the disease, predominantly low quality diets poor in milk, animal protein, and vitamins but rich in starch. Heavily salted pickles favoured in Japan (tsukemono) have been implicated, as have the smoked fish and meats eaten in Iceland and Scandinavia: these smoked meats contain polycyclic hydrocarbons (such as benzopyrene) which are probable carcinogens. The decline in the incidence of the tumour in the United States has been attributed to the widespread use of refrigeration with a decline in food preservation by smoking or salting.

Much interest has been devoted to the role of nitrosamines, which have been shown to be carcinogenic to the gastric mucosa in experimental situations. It is suggested that ingested nitrates and nitrites present in some high protein diets, as food preservatives, or in water and soil may be converted to nitrosamines (N-nitroso compounds) by the action of gastrointestinal bacteria. The presence of atrophic gastritis and associated achlorhydric stomach may predispose to the production of these N-nitroso carcinogens. High gastric pH encourages bacterial overgrowth in the stomach; these organisms are able to reduce dietary nitrates and convert the nitrites to carcinogenic nitrosamines in the presence of dietary protein. This sequence may only be one factor and defects in the mucosal barrier associated with diseased mucosa in patients with atrophic gastritis is likely to facilitate carcinogen penetration.

Other ingested agents implicated are neat spirits favoured by some Scandinavian countries, Japanese saki, and contaminated whisky. Cigarette smoking and elevated levels of zinc and lead in drinking water have been implicated, as has talc and asbestos in the atmosphere. Groups of workers which have been shown to be at high risk include metal industry workers, painters, printers, fishermen, and ceramic and clay workers.

This group of conditions includes those that, if untreated, become malignant, as well as disorders of the stomach that may predispose to gastric cancer. In particular patients with hypogammaglogulinaemia (50-fold excess) and pernicious anaemia (three-fold excess) are at high risk.

Patients with hypogammaglobulinaemia and pernicious anaemia have chronic atrophic gastritis and the incidence of gastric cancer is said to be higher in patients with this condition. Certainly achlorhydria results from the chronic atrophic gastritis and 75 per cent of patients with gastric carcinoma are achlorhydric. Strickland has divided chronic atrophic gastritis into two subgroups: type A, which is associated with pernicious anaemia, predominantly affects the fundus and body and is autoimmune in origin, while type B gastritis affects the antrum and is related to environmental factors. This type is also found in the stomach some years after gastrectomy for benign peptic ulcer disease and may be regarded as a failure of the gastric mucosa to respond to repeated injury. Both types predispose to cancer.

Intestinal metaplasia of the gastric mucosa is commonly found in association with gastric cancers, and epidemiological studies have confirmed that populations with a high incidence of carcinoma of the stomach also have a high incidence of intestinal metaplasia. However, mucosal atrophy and intestinal metaplasia are common phenomena, their incidence increasing with increasing age, and they are particularly common in elderly populations. Some suggest that the intestinal type of gastric cancer results from gastric mucosa that has undergone a sequence of mutations and defined histopathological changes that may start in the first decade of life. The first lesion is atrophic gastritis followed by progressive intestinalization of the mucosa to intestinal metaplasia, then dysplasia and finally carcinoma. The finding of some of these precursor conditions alone cannot at present be regarded as definitely premalignant unless dysplastic mucosa is found.

Recently, gastric cancer has been associated with Helicobacter pylori infection. This organism is associated with antral inflammation and gastritis. It is proposed that infection and inflammation may result in the production of an epidermal growth factor which may have an oncogenic action on gastric mucosa.

The relationship between benign gastric ulcer and gastric cancer is controversial. There is debate over whether benign chronic gastric ulcers have malignant potential; the suggestion is that the regenerating mucosa around an ulcer is prone to become malignant. The issue is further clouded since some ulcerating gastric cancers can mimic benign gastric ulcers closely, sometimes healing in response to medical treatment. Approximately 4 to 10 per cent of all gastric cancers behave in this way, and 70 per cent of early gastric cancers may heal as part of the ‘lifecycle of the malignant ulcer’. Improvements in endoscopy with biopsy have improved detection of both benign and malignant disease, and it appears that the development of cancer in the edge of a benign gastric ulcer is rare. There is no clear description of a proven benign ulcer turning malignant. The stable incidence of gastric ulcer with a decline in the incidence of gastric carcinoma supports the view that gastric ulcer is not a premalignant condition. Similarly the location of benign and malignant disease is different, with most benign gastric ulcers (50–70 per cent) occurring on the lesser curve. It is now accepted that the most important question on finding a gastric ulcer is to decide whether it is benign or malignant from the outset.

Gastric polyps are found in 0.5 per cent of individuals undergoing autopsy. Most (65–90 per cent) are hyperplastic polyps that are regenerative, non-neoplastic lesions and usually smaller than 2 cm. Only two patients have been reported in whom a carcinoma was found in association with a hyperplastic polyp. There is no strong relationship with gastric carcinoma and these polyps should not be regarded as premalignant.

In contrast adenomatous polyps are truly premalignant. They are often larger (80 per cent greater than 2cm) and are tubulovillous or villous on microscopic examination. The frequency of malignant change increases with increasing size. In a large series, 38 per cent of patients with gastric adenomatous polyps had gastric carcinoma. Similarly 34 per cent of post-gastrectomy specimens removed for gastric cancer contain adenomatous polyps, and severe dysplasia with carcinoma in situ has been found in over 20 per cent of removed polyps.

As early as 1922 Balfour reported a gastric cancer occurring in the residual stomach after surgery for benign peptic ulcer disease. The term ‘stump cancer’ was soon used since carcinoma seemed to occur more frequently after Billroth I and II gastrectomy than after vagotomy with pyloroplasty or gastroenterostomy. A large postmortem study demonstrated that gastric cancer was less frequent in patients who had undergone gastric surgery 15 years prior to death, but six times greater in those who had surgery 25 years earlier. In a historical prospective cohort study in Denmark from 1955 to 1982, the risk of gastric cancer immediately, and for 15 years after, peptic ulcer surgery was less than expected. This was attributed to patients having less gastric mucosa exposed to carcinogen following gastrectomy. However, the risk of cancer was 2.1 times greater than the general population 25 years after surgery. The greatest risk (3.2 times) was in male patients who had had a Billroth II gastrectomy. Patients who had simple suture of a perforated ulcer had no increased risk, indicating that peptic ulcer disease was not a risk factor. The pathogenesis of gastric stump cancer has been shown experimentally to be related to operations that promote duodenogastric bile reflux, achlorhydria, and atrophic gastritis. Overall, the risk of developing gastric cancer following gastrectomy has been reported as between 3 and 10 per cent. The decline in surgery for peptic ulcer disease means that there is every prospect that in 20 or 30 years stump cancer may become a rare phenomenon.

Gastric carcinoma has been described as a complication of Ménétrier's disease, but the magnitude of risk is unknown. In Ménétrier's disease, there is giant hyperplasia of the gastric mucosal folds and the condition can be difficult to distinguish from gastric polyposis or lymphoma (see Section 15.6) 104. The mucosal abnormality associated with Ménétrier's disease results in hyperplasia of mucus glands, whereas the parietal cell mass falls. Thus gastric secretion is rich in protein and mucus but is often hypochlorhydric. Hypersecretory conditions, including Zollinger–Ellison syndrome, may be associated with hyperplastic rugal folds and excessive acid secretion without increased risk of gastric cancer.

Marked variations in the gross appearance of operative and excised specimens have been described in the Western literature and by the Japanese, using the endoscopic descriptions of predominantly early gastric cancers. Approximately 10 per cent of these tumours are polypoid fungating tumours, with a nodular polypoid surface with superficial ulceration (Fig. 1) 935. This group of cancers has a relatively good prognosis after aggressive surgical management, many being well-differentiated adenocarcinomas. Ulcerating or penetrating cancers, which occur in more than 50 per cent of patients (Fig. 2) 936, are sessile and may have the appearance of a benign gastric ulcer. Superficial spreading carcinoma is more unusual (6 per cent of tumours). This tumour is diffusely infiltrative over a wide area, although it is predominantly confined to the mucosa and submucosa; 50 per cent have metastasized to the perigastric lymph nodes at the time of presentation and operation. A further subgroup of diffusely infiltrative cancer is the linitis plastica (leather bottle) carcinoma (Fig. 3) 937, which is characterized by extensive infiltration of the submucosa and muscular layers with a marked fibroblastic/desmoplastic reaction around columns of malignant cells. Endoscopic recognition of this neoplasm may be difficult and superficial biopsy of mucosa alone may prove negative. Extension into the oesophagus and mesentery may occur and advanced tumours may involve the entire stomach. Superficial erosions are common but deep ulceration is unusual. Although spread of gastric cancer beyond the pyloric ring is generally considered to be rare, up to 25 per cent of diffusely infiltrative tumours may involve the first part of the duodenum. This tumour carries a particularly poor prognosis.

The morphology of advanced gastric cancer can be divided into three strict morphovolumetric types. The ratio of the amount of muscle invasion to mucosa affected defines a funnel type (mucosal involvement greater than muscle: ratio less than 0.75), column type (equal involvement: ratio 0.75 to 1.25), and the mountain type (muscle greater than mucosal involvement: ratio over 1.25). The metastatic characteristics of these tumours are variable: the funnel type carry the best prognosis, with 62 per cent lymph node involvement, followed by the column type (80 per cent lymph node metastasis), and finally the mountain type (lymph node metastasis in 85 per cent).

The great interest in the detection of early cancer by endoscopy has produced a separate classification based on endoscopic appearance. The term early gastric cancer is used to describe tumours confined to the gastric mucosa and submucosa, irrespective of nodal status: there is a close correlation between the depth of invasion at microscopic examination and postoperative survival. Early carcinoma of the stomach is divided into three main groups, type I—protruding, type II—superficial, and type III—excavated (Fig. 4) 938. Type II is divided into three subgroups, elevated (a), flat (b), and depressed (c). There is undoubted overlap between the protruded (I) and superficial elevated (IIa) types, and also between excavated (III) and superficial depressed types (IIc). As well as these basic types there are three combined types which exhibit features of two different types, type I and IIa (III), type IIa and IIc, and type IIc and III.

The precise classification of early gastric cancer by the Japanese and the use of vigorous investigation has increased its detection rate from 5 to 40 per cent of all gastric cancers over the past 20 years. In the West early gastric cancer accounts for 9 per cent of all cancers detected and 19 per cent of resected tumours; the rate of detection does not appear to be increasing. Comparison of the clinical, morphological, and histological features of early gastric cancer reveals remarkable similarity in the disease between Japan and the United Kingdom.

There are few data on the natural history of early gastric cancer, since it is usually actively treated. One study has shown that 50 per cent of cases progress to advanced cancer within 3 years from the time of endoscopic diagnosis, and almost all patients develop advanced disease within 5 years. Recently, the definition of early gastric cancer has been challenged, since the 5-year survival is 99 per cent if there are no lymph node metastases, but falls to 73 per cent if lymph nodes are involved.

Multiple synchronous gastric cancer was described by Moertel, who defined strict criteria for diagnosis. Each lesion must be proved histopathologically to be malignant, all lesions must be separated by normal gastric wall, and the possibility that the lesions represent a local extension or metastasis must be ruled out beyond reasonable doubt. It is sometimes difficult to verify this last criterion, and attention is paid to the extent of venous and lymphatic invasion surrounding each tumour. If there is extensive invasion a diagnosis of multiple gastric cancer is inappropriate. The incidence was originally thought to be 2 per cent of all gastric cancers, but with advances in diagnostic techniques it appears to be between 6 and 9 per cent.

The great majority of gastric cancers are adenocarcinoma; other tumours are rare. Classification is difficult, primarily because different histological features may coexist in one tumour mass. The WHO classification based on morphology divides gastric cancer into five types: adenocarcinoma, adenosquamous carcinoma, squamous cell carcinoma, undifferentiated carcinoma, and unclassified carcinoma. The adenocarcinomas are divided into four patterns; papillary, tubular, mucinous, and signet ring, which may be divided by degrees of differentiation. In tumours presenting a mixed picture with varying degrees of differentiation in the same tumour, classification is based on the predominant type.

A simpler classification by Lauren recognized an intestinal type similar to colon adenocarcinoma and a diffuse type that did not form glands. The intestinal type is defined by the cellular architecture, and the diffuse by its pattern of growth: polypoid and superficial spreading carcinomas are of the intestinal type, whereas linitus plastica is diffuse. Ulcerative tumours fall into either group. The intestinal type carries the more favourable prognosis and is more common in areas with a high incidence of the disease. In general, intestinal cancer is said to arise within intestinalized mucosa and the diffuse tumour is thought to originate from normal epithelium without evidence of a preceding lesion. Synchronous cancers are more commonly of the intestinal type of tumour. Some carcinomas fall into a mixed or unclassified category.

A similar classification for advanced carcinoma invading beyond the muscularis propria was proposed by Ming in 1977. Cancers were divided into an expanding type that produces nodules that compress adjacent tissue (similar to Lauren's intestinal type), and the infiltrative type that does not form masses (Lauren's diffuse type). The best concordance for tumour classification by different pathologists examining the same specimens is with Lauren's classification and this is the most widely used.

Considerable interest surrounds the prognostic significance of histological evidence of host defence against the tumour. Infiltration of the tumour by macrophages, polymorph leucocytes, lymphocytes, and plasma cells is a favourable sign, as is evidence of a host response in draining lymph nodes, characterized by sinus histiocytosis and follicular hyperplasia.

The microscopic appearance of early gastric cancer is strongly correlated with the macroscopic appearance (Lauren classification). Type I and IIa are of the intestinal type whereas type IIc and III are all diffuse type: this is not surprising since intestinal type carcinomas have an expansile growth pattern leading to polypoid growth. The prognosis of early gastric cancer is predominantly related to the depth of invasion and this greatly over-rides any histological classification.

One of the characteristics of gastric cancer is the tendency of the tumour to spread intramurally via lymphatic channels. Thus the surgical resection margins must be at some distance from the palpable edge of the tumour. The extent of intramural spread is related to the depth of invasion of the tumour. Lateral extension of tumour confined to the muscularis propria lateral does not extend beyond 3 cm, but significant numbers of tumours penetrating this layer show intramural spread beyond this distance, although never beyond 6cm.

In Western series between 40 and 60 per cent of patients have obviously incurable and often disseminated disease at presentation. In Japan the number who can be offered curative resection with excision of all macroscopic tumour is much higher at 75 to 80 per cent. The poorly differentiated mucinous and signet ring tumours are more invasive, although in series from the United Kingdom most tumours are large and advanced at presentation and have spread, and the degree of differentiation is of little relevance. In Japan up to 40 per cent of patients have early gastric cancer with tumours confined to the mucosa and submucosa: those with poorly differentiated carcinomas are more subject to metastasis and have decreased survival.

The depth of invasion of the cancer through the gastric wall has a marked effect on the prognosis. If the muscularis propria and then the serosa are breached the prognosis following treatment becomes considerably worse. Lymph node metastasis occurs in 18 per cent of patients in whom the serosa is not penetrated and over 50 per cent survive for 5 years following resection. If the serosa is penetrated 80 per cent of patients have lymph node metastases, with a correspondingly poorer prognosis. Once the serosa of the stomach has been breached the tumour is likely to spread by transcoelomic implantation of shed cells. If the area of serosa involved is less than 10 cm², free viable intraperitoneal malignant cells can be found in 22 per cent of patients, compared with 72 per cent if the area is greater than 20 cm². Implantation of these cells characteristically occurs in the ovary, producing Krukenberg tumours (bilateral ovarian tumours from a signet ring carcinoma), or in the pelvis to produce a shelf which is palpable on rectal examination (Blummer's shelf). It must be emphasized that penetration of the gastric serosa is the most important prognostic indicator.

Gastric tumours may spread by direct extension and invasion of adjacent structures including liver, pancreas, and spleen. In post-mortem studies from the United States the peritoneum, mesentery, and omentum are invaded in over 40 per cent of patients.

The involvement of lymph nodes and the pattern of spread is particularly important for gastric cancers. In Western series some form of lymph node involvement is present in 70 per cent of patients undergoing resection of gastric tumours. Even in Japanese series, mucosal cancer is associated with a 3 per cent incidence of lymph node involvement; this rises to 15 per cent if the submucosa is involved. The anatomical classification of lymph nodes varies in the West and Japan, and confuses the interpretation of the results of surgery. In the Western literature four zones of lymph drainage corresponding to the blood supply are identified. Zone 1 is located in the gastrocolic omentum along the right gastroepiploic vessels, draining the pyloric portion of the greater curve and draining to the pylorus and then to coeliac and aortic nodes. Zone 2 is in the gastrocolic and gastrosplenic omentum around the left gastroepiploic vessels draining from the upper half of the greater curve. From here drainage is to the pancreaticosplenic lymph nodes and to the aortic nodes. Zone 3 has efferent channels from the proximal two-thirds of the stomach and the upper lesser curve and surrounds the left gastric artery. Some lymph from this area drains into perioesophageal lymph nodes. The distal portion of the lesser curve and pylorus drains to Zone 4, situated above the pylorus and going to the hepatic artery and para-aortic lymph nodes. However, according to Western data, the lymphatic drainage of the stomach is unpredictable and radical resection must include lymph nodes from all zones.

The Japanese have adjusted and stratified the classification of lymph drainage (Fig. 5) 939. Group I (N1) are perigastric lymph nodes, group II (N2) nodes along and at the roots of the major vessels, group III (N3) are lymph node at the root of the superior mesenteric artery, in the hepaticoduodenal ligament, and behind the pancreas. Group IV (N4) are distant lymph nodes. It is clear that this definition may not be adequate for planning surgical resection along strict oncological principles, since a perigastric node (N1) at the antrum is a distant node (N4) if the cancer is in the cardia. Thus the lymph nodes can be given station numbers ( Table 1 305 and Fig. 5 939) and the node status and surgical excision planned according to the node station in relation to the site of the primary tumour (Table 2) 306. The Japanese have found that there is an orderly spread to regional lymph nodes which is clearly related to the position of the tumour. These data have not been confirmed in Western studies. The difference may, in part, relate to the difference between advanced gastric cancer in the West and the greater incidence of potentially curable patients identified in Japan. However, skip lesions to distant lymph nodes do undoubtedly occasionally occur; N2 nodes are involved without N1 (perigastric) nodal involvement in 11 per cent of patients.

Carcinoma at the cardia presents a distinct problem, for it may involve lymph nodes in the mediastinum. In a review of over 400 patients neoplastic involvement of paracardial nodes occurred in 48 per cent and para-oesophageal nodes were affected in 37 per cent of patients.

In general, any involvement of the lymph nodes is a poor prognostic indicator, as is the number of nodes involved, invasion of four or more lymph nodes being unfavourable compared with fewer being involved. This statement implies that an adequate lymphadenectomy has been performed to sample these nodes. The number of lymph nodes to be found at each station is very variable. If stations 1 to 11 are removed (R2 gastrectomy) an average of 27 nodes should be removed. If stations 1 to 16 are cleared (R3 gastrectomy) the mean number rises to 43.

The most common sites for distant metastatic spread are the liver (49 per cent), lung (33 per cent), ovary (14 per cent), bones (11 per cent), and cervical and supraclavicular (Virchow) nodes (8 per cent). Patients with distant metastases have a poor prognosis: 95 per cent of patients with liver involvement will die within 1 year, irrespective of the primary cancer.

As with all neoplasms a uniform staging system is required to allow the results of treatment to be compared. The TNM system of staging can be applied to gastric cancer and is summarized in Table 3 307. Another widely used staging system based on the results following excision and pathological examination is somewhat simpler (Table 4) 308. For cancers that are locally resectable the most important prognostic sign that can be assessed at laparotomy is penetration of the wall of the stomach by the carcinoma.

The symptoms of gastric carcinoma are often vague, non-specific, and attributed to non-specific dietary indiscretion or indigestion. By the time the diagnosis is made, the tumour is often incurable or non-resectable. This was elegantly illustrated by Theodor Storm's (1888) poem ‘Beginn des Endes’ describing his death from gastric carcinoma. ‘'Tis a prick, 'tis scarce a pain . . . indeed 'tis naught; . . . it is too late’. Unfortunately 100 years later this experience is still common. Definite symptoms do not usually occur until the tumour is large enough to obstruct the lumen or cause disordered gastric function by invading a large segment of the wall, or until it bleeds. Over 70 per cent of patients have had some symptoms for greater than 6 months prior to seeking advice, the most common of which are vague indigestion or upper abdominal/epigastric pain, followed by weight loss, nausea and vomiting, haematemesis and malaena, profound anorexia, early satiety, and flatulence. The pain may mimic angina pectoris, or may exhibit periodicity and be relieved by food, mimicking benign gastric disease. Interestingly true postprandial pain is relatively unusual in patients with gastric carcinoma. If the tumour is at the cardia, over 60 per cent of patients will present with dysphagia, indicating a greater than 80 per cent obstruction of the oesophageal/gastric cardia lumen. Weight loss may be quite insidious and although most patients have lost significant amounts few complain of this directly.

Early gastric cancer presents with symptoms similar to those of dyspepsia. Purely mucosal gastric cancers are symptomatic in 50 per cent of patients, and early endoscopy is advised for patients over the age of 40 years with persistent dyspeptic symptoms. If these patients have dysplastic changes then regular endoscopy is necessary. Some recommend resection for severe dysplastic changes in young, otherwise healthy, patients. The duration of symptoms prior to surgery in those with early gastric cancer ranges from 3 to 72 months; surprisingly, this is little different from those who present with advanced cancers. There are generally few physical signs in patients with early gastric cancer, although epigastric tenderness is present in 10 per cent, the remainder having a normal physical examination. It is not possible to be highly specific about important symptoms and no symptom correlates with extent of disease. The presence of an epigastric mass is a poor prognostic sign, but is not in itself an indication of inoperable disease.

Approximately 10 per cent of patients in Western series present with palpable cervical lymph nodes, ascites, jaundice, a palpable abdominal, or a pelvic mass. Sister Joseph's nodule, named after a nurse at the Mayo Clinic who discovered the phenomenon, a visible and palpable secondary deposit at the umbilicus due to spread along the lymphatics around the falciform ligament, is not an uncommon presentation of advanced disease. In patients presenting with this sign, the primary tumour is most commonly in the ovary, followed by the stomach and colon. It is a poor prognostic sign and the mean survival of patients in whom the primary lesion is gastric is only 3.5 months. Troisier's sign, an enlarged lymph node in the left supraclavicular fossa (Virchow node), indicates lymphatic spread via the thoracic duct. Armand Trousseau (1801–1867), a Paris physician, first suspected he had a gastric cancer when he developed superficial thrombophlebitis on the legs (Trousseau's sign), but Trousseau's sign is also associated with pancreatic cancer.

Routine biochemical and haematological tests may disclose anaemia which is common in these patients. This is usually of an iron deficiency type, being microcytic and hypochromic reflecting blood loss and is not only a feature of advanced disease, since 20 per cent of patients with early gastric cancer will be anaemic. Testing of the stools for occult blood may well be positive. Approximately two-thirds of patients with gastric cancer have achlorhydria but this is of little diagnostic relevance. The level of carcinoembryonic antigen in the serum is elevated in 30 per cent of patients with advanced tumours but fails to detect early disease. Similarly oncofetal antigen is elevated in some benign inflammatory gastric disorders as well as in malignancy.

The mainstay of diagnosis for many years was barium contrast studies of the upper gastrointestinal tract. In Japan the double contrast air/barium study has been perfected and used for mass population screening. Air and barium are introduced together to coat the mucosa with a thin layer of barium and enhance mucosal detail. The technique can be further refined by using high density barium, carbon dioxide, simethicone for gas dispersion, and glucagon to induce gastroparesis. In Western countries endoscopy is more sensitive than radiography, but some Japanese studies using modern photographic techniques have found the diagnostic accuracy of these methods to be similar. Computerized tomography, ultrasonography, and magnetic resonance are most helpful in the diagnosis of metastatic disease. These methods identify enlarged lymph nodes and tumour deposits but do not differentiate tumour from benign change. In early gastric cancer, metastases are usually confined to the perigastric lymph nodes, which are difficult to detect. Attempts to improve the accuracy of preoperative lymph node involvement include localization with monoclonal targeted isotopes, endoscopic lymphography, endoluminal ultrasound, and dynamic CT.

Endoscopic ultrasonography discloses perigastric lymph nodes greater than 3 mm in diameter in 70 per cent of patients. If the examination is preceded by the oral administration of an oil-in-water emulsion uninvolved nodes can be identified by echo enhancement at their margin, whilst metastatic nodes show no enhancement. The sensitivity of this test in Japanese hands reaches 92 per cent with a specificity of 100 per cent for nodes greater than 3 mm. Endoscopic lymphography following submucosal injection of contrast can identify metastatic filling defects in involved lymph nodes. Dynamic CT has only slightly improved the diagnostic accuracy of conventional CT and some lymph nodes regions cannot be clearly seen.

The development of fibreoptic, flexible forward viewing endoscopes with a controllable tip has been a major advance, allowing direct visualization of the stomach and accurate biopsy of any lesion which is identified. Endoscopy is observer dependent and visual endoscopic interpretation, even in experienced hands, is often incorrect in deciding on the nature of a lesion in over 10 per cent of cases. If up to 10 biopsies are taken from each lesion the diagnostic accuracy reaches 100 per cent. The accuracy of biopsy is now clearly been shown to be related to the number of biopsies taken. The working recommendation is that four to six biopsies should be taken from each lesion, and from the inner border of the edge of any ulcer.

Improvements in endoscopic techniques include dye-spraying, fluorescence endoscopy, magnified endoscopy, electronic endoscopy, and endoscopic ultrasonography. This last technique can differentiate early gastric cancer from advanced tumours in 80 per cent of patients, but has so far failed to allow preoperative differentiation of mucosal from submucosal cancer because of the problems of fibrosis in the muscular propria and submucosa that may surround ulcerated early gastric cancers.

Cytological study of gastric aspirate for exfoliated malignant cells has been reported for patients with advanced disease, but has a variable accuracy of between 40 and 90 per cent. The cytological specimen may be collected by gastric washing, washing with addition of a mucolytic agent, and the passage of a balloon. Cytological analysis of brushings of suspicious lesions collected directly has an accuracy of 81 per cent; combination with biopsy improves the accuracy to 91 per cent. The principal limitation of both cytology and biopsy is failure to obtain an adequate sample. Interpretation problems may be posed by atypical cells associated with regeneration around an ulcer. Immunocytochemical stains such as fetal sulphaglycoprotein may be useful for cytological preparations.

Laparoscopy may be valuable as an initial operative assessment to exclude extensive disease, since hepatic, peritoneal, ovarian, or extensive local disease may be assessed. Laparotomy is necessary in all patients with local disease. The presence of extensive nodal disease excludes patients from curative surgery. Some advocate biopsy and frozen section of a lymph node from the infracoeliac periaortic region prior to commencing radical resection. In Japan, peritoneal washings are taken at initial laparotomy for cytological examination, and lymph node sampling with frozen section is performed prior to deciding whether a curative resection is feasible. Locally invasive disease may be excised curatively only if it is minimal.

Government-subsidized mass screening has been introduced in Japan, using predominantly radiological methods, with photofluorographic radiology performed at designated centres or by mobile units. Approximately 5 million people are screened every year. After 25 years of screening, the rate of gastric cancer had decreased from causing over 50 per cent of all deaths from malignant disease in men to 33 per cent. Similarly, deaths in women have been reduced from 38 per cent to 28 per cent. The incidence/detection of early gastric cancer has risen from under 2 per cent in 1945 to 63 per cent in some centres. The overall 5-year survival from gastric cancer has risen to over 50 per cent. There are clear shortcomings in using survival rate in evaluating the effectiveness of screening. In particular there is concern at the definition of early gastric cancer and its distinction from dysplasia, or ‘worrying mucosal appearance’. Also most studies report 5-year survival and there may well be a lead time bias. Recently, a detailed follow-up has demonstrated that two-thirds of patients with cancer detected by screening in Osaka were cured of their disease 15 years after resection.

Screening of asymptomatic patients is not considered feasible except in Japan, where 25 per cent of gastric cancers are identified by screening programmes. Despite the wider use of the flexible endoscope in the West, the detection of early stage I cancers has not increased. Screening of some high-risk groups is feasible: certain centres have introduced programmes to allow early endoscopy of dyspeptic patients over the age of 40 years, and serial endoscopy until healing of all patients with gastric ulcer. Patients with atrophic gastritis, dysplasia, or adenomatous polyps are offered endoscopy on an annual basis. Among 2659 patients examined in this scheme, 57 cancers have been identified, 20 per cent of which were early gastric cancers, but only 60 per cent of these were suitable for attempted curative surgery.

The excellent results of surgery obtained in Japan have fuelled debate as to whether the disease there is different from that in the West. In all countries Stage I disease is associated with a relatively good prognosis. Yet the overall 5-year survival in Europe, for this stage of disease, was reported as 70 per cent compared with 98 per cent in Japan (Table 5) 309, suggesting differences in biological behaviour of the tumour. A detailed review of tumour classification has clarified this issue. The British Society of Gastroenterology reviewed the histology of 319 patients from 41 hospitals with histopathological findings of dysplasia and early gastric cancer. Although there was good agreement between pathologists on the difference between dysplasia and cancer, over one-third of patients thought to have early gastric cancer were found to have more advanced disease on reassessment. The true 5-year survival rate of the group redefined as having early gastric cancers was 90 per cent, compared with 75 per cent of the group initially defined as early gastric cancer. Thus European survival rates for the disease approach those in the East if strict Japanese criteria are used. Unfortunately, the issue cannot yet be regarded as completely resolved. In Japan there is a higher incidence of better prognosis tumours and some tumours do behave differently in the East and West; however, when direct comparisons are made between like tumours the results appear similar.

Historical references to what appears to be gastric carcinoma can be found in the writings of Hippocrates, Galen, and Avicenna. However, it was not until the nineteenth century that Jean Cruveilhier attempted to distinguish between benign and malignant gastric lesions, and gastric cancer was initially referred to as ‘la maladie de Cruveilhier’ by French physicians. A treatise in 1839 by Bayle clearly described the detailed pathology of gastric cancer and suggested treatments. Resection was attempted by Jules Péan (France) in 1879 and 1 year later by L. Rydygier (Poland). Reconstruction was undertaken using catgut to form a gastroduodenostomy. Both operations were unsuccessful, the patients dying shortly after the procedure as a result of peritonitis from gastric leakage. It appears that an inadequate number of sutures were used. It has been reported that a French surgeon in Arras attempted gastric resection 1 year prior to Péan but details are not available. Theodor Billroth (1829–1894) believed that a safe technique could be developed for the excision of gastric cancer and dispatched Carl Gussenbauer and Alexander Winwater to the postmortem room and the laboratory to investigate the possibility. They found after examining the records of the Vienna Pathological Institute that over 40 per cent of pyloric tumours were amenable to surgical resection and did not appear to be associated with distant metastases. They developed the procedure of the Billroth I gastrectomy in the dog: two of seven dogs treated survived for a prolonged period and two others died of anastomotic dehisence.

Billroth's first gastrectomy for gastric cancer was performed on Frau Therese Heller in 1881, under chloroform anaesthesia, using antiseptic precautions. The gastroduodenal anastomosis was performed with carbolized silk sutures. Unfortunately, a few hazelnut sized mesenteric lymph nodes were found and histological examination confirmed tumour involvement. The patient survived for 4 months, and died of recurrent disease.

Immediately after the initial success of the Billroth I operation, Wölfler (Billroth's assistant) was instructed by his superior to devise a palliative operation to relieve gastric outlet obstruction for unresectable tumours. Thus the technique of bypass gastroenterostomy was devised in 1881, and used clinically later that year. In January 1885, Billroth performed a resection of the distal stomach, closing the transected duodenum and stomach and restoring continuity by a gastroenterostomy to the posterior wall of the stomach. The patient survived for 18 months and the Billroth II operation was born (Fig. 6) 940.

By 1890, 41 resections for gastric cancer had been performed at the Billroth clinic with a successful outcome in 19; by 1894 the overall mortality was reported as 55 per cent. These results prompted Welch in 1885 to write that no instances of prolonged recovery or cure followed operation for gastric cancer despite the ‘great sensation’ produced by Billroth's operative achievements. Extended surgery was explored as a means of improving survival: Carl Schlatter (1897) performed total gastrectomy with end-to-side oesophagojejunal anastomosis. Brigham (1898) also excised the entire stomach in a patient who survived for 18 years and died of other causes. The fact that this patient survived longer than 3 years prior to the identification of gastric intrinsic factor (in 1929 by W.B. Castle) and when vitamin B&sub1;&sub2; supplements were not available may indicate that the gastrectomy was not total.

Billroth's initial gastroduodenal anastomosis with partial closure of the gastric lesser curve was soon found to be associated with leaks at the junction of the lesser curve and the duodenal anastomosis. This area was called the ‘jammer ecke’ (crying corner) and a further inverting triangular suture was subsequently inserted in this area. It is appropriate to acknowledge the work of the Budapest surgeon Eugen Polya (1876–1944), which was first recognized by William Mayo. He devised the retrocolic anastomosis of the entire width of the gastric segment to the jejunum after gastrectomy (Fig. 6) 940.

There are now many distinguished names associated with various forms of gastrectomy. There was an extended period of debate in the 1940s as to whether the jejunal anastomosis should be isoperistaltic, retroperistaltic, antecolic, retrocolic, with a long or short afferent loop, to the entire gastric stoma, or whether a valve was necessary. Various modifications allowed the attachment of an eponym to each modification of gastrectomy, often with a strongly held anecdotal belief of its superior merits. At present most surgeons favour partial closure of the gastric stoma as described by Hofmeister, with an antecolic anastomosis for reconstruction and a short afferent loop after a cancer resection. The precise form of reconstruction is of little relevance.

Since Billroth's first resection, gastrectomy has been the mainstay of therapy. However, the optimal surgical management is still a subject of intense debate. The controversy lies between the predominantly Western belief that the pathological stage of disease is the crucial prognostic factor and the widespread Eastern opinion that the extent and nature of the resection, combined with adjuvant therapy, are major factors in improving the results. Predetermined factors such as location of the tumour and the extent of nodal and, in particular, serosal involvement undoubtedly affect survival, but the manner in which the resection is performed can have a significant impact on the prognosis. Thus one view must not be exclusive of the other. The importance of early detection is now universally accepted and in the absence of distant metastasis, aggressive resection of the carcinoma is justified.

Poor prognostic factors in relation to the presentation of the tumour are serosal invasion, presence of lymph node metastasis, presence of free carcinoma cells in the peritoneum, Lauren's classification (intestinal better than diffuse), cardial tumour, histological invasion of lymph vessels, tumour stage, tumour depth and size, and the patient's age. The poor prognostic variables that relate to surgery are positive resection margins, inadequate lymphadenectomy, and the need for an associated splenectomy. The most important single prognostic factor is serosal invasion. It is now clearly demonstrated that the number of lymph nodes affected or the extent of resection does not affect the prognosis if the serosa is involved.

The basic oncological approach for resection of mucosal cancers is wide excision of the primary tumour with en-bloc removal of the lymphatic drainage and network of lymph nodes. This approach has undergone re-evaluation for some tumours, with lesser resection in combination with other means of treatment providing a more conservative and equally effective option. At present our understanding of the biological behaviour of gastric carcinoma, in particular with regard to the very limited effectiveness of radiotherapy and chemotherapy, means that radical excisional surgery probably offers the best chance of cure.

The problems of definition of radical excision in treatment of gastric cancer confuse the interpretation of the results of surgery. The main areas that must be addressed are the extent of gastric resection, and the extent of lymph node resection.

Adequate gastrectomy implies surgical margins in the stomach free of tumour. Gastrectomy may therefore be partial or subtotal if the tumour is distal or total if it is more proximal. Adequate resection margins in the stomach are defined as an 8- to 10-cm proximal and distal clearance in the unstretched stomach. Failure to resect the stomach widely with microscopic clear margins is highly detrimental to survival. If the resection margin is not confirmed to be free of microscopic diseasee, the prognosis of a Stage II tumour falls to that of a Stage IV tumour. The specific operation for a gastric carcinoma depends somewhat on its site. There is a widespread belief that subtotal gastrectomy is associated with poor local control. However, total gastrectomy appears to have no advantage over adequate gastrectomy for local tumours. A prospective controlled study comparing total and subtotal gastrectomy (with the same node dissection) for carcinoma of the gastric antrum demonstrated no survival difference between the two groups. At Memorial Sloan-Kettering Cancer Center total gastrectomy was found to be detrimental to survival, not because of the extent of resection, but predominantly because splenectomy was associated with this operation. It is recommended that an exclusive policy of elective total gastrectomy when the tumour can be widely resected by a subtotal gastrectomy is incorrect. For patients with synchronous multiple gastric cancers, logic suggests that a total gastrectomy is essential. Paradoxically the survival after total gastrectomy does not exceed that after partial gastrectomy, and it has been suggested that small synchronous lesions may regress after resection of the main tumour, although there is no evidence for this. In general, multiple cancers should be treated by total gastrectomy where possible.

Lesions in the body and fundus of the stomach present different problems. Total gastrectomy is often advocated in patients with such lesions because the small amount of stomach that remains after more conservative surgery has little reservoir capacity. Both total and proximal gastric resection require the creation of an oesophagoenteric anastomosis and the operative mortality is 5 per cent. Theoretically, total gastrectomy carries a lower risk of recurrence or a second gastric cancer in long-term survivors. Total gastrectomy with a Roux-en-Y reconstruction is also superior to proximal gastrectomy since these patients are less subject to alkaline reflux. These tumours are often far advanced at presentation because they reach considerable size before producing symptoms. They have a less favourable prognosis than tumours in the antrum.

The preferred method of reconstruction after total gastrectomy is as a Roux-en-Y with a 60-cm Roux to prevent bile reflux. The creation of a pouch to act as a reservoir to prevent early satiety has been advocated. The Hunt-Lawrence pouch (Fig. 7) 941 is created from a long Roux loop with an enteroenteric anastomosis at 80 cm. It is simply folded on itself and a 10-cm side-to-side anastomosis created below the oesophageal anastomosis. A partial gastrectomy is best reconstructed as an antecolic polya type of gastrectomy (Fig. 6) 940.

Gastric cancer at the cardia represents a special management problem. At presentation only 10 per cent of patients have disease restricted to the stomach, and only 2 per cent are early cancers. The tumour may infiltrate the lower oesophagus, and a 10-cm oesophageal clearance is advised to be certain of clear resection margins. Thus surgery involves principles of gastric as well as oesophageal surgery, and in certain respects it should be regarded as a separate entity.

The benefit of lymphadenectomy in the treatment of gastric cancer has been emphasized by the Japanese. This concept is not new in oncology. Wide lymphadenectomy encompassing one or more echelons of uninvolved lymph nodes improves survival in the small number of patients with limited nodal involvement seen in Western series. The Japanese Research Society for gastric cancer defines a curative resection as one in which patients without peritoneal, serosal, or hepatic involvement have a gastric resection with a lymph node dissection one level beyond that of pathological lymph node involvement.

The Japanese classification of the type of gastrectomy corresponds to how radical the operation is in terms of removing the groups of lymph nodes, according to the Japanese classification (N0–N4). R0 gastrectomy does not remove any lymph node group, R1 gastrectomy removes those nodes in group I (N1), which are predominantly perigastric lymph nodes, but leaves a large portion of the greater omentum. No formal lymphadenectomy is performed during this resection. A hybrid form of this resection, a traditional radical gastrectomy including omentectomy and partial lymphadenectomy is predominantly performed in the West. The Japanese results suggest that in most circumstances the radical R2 gastrectomy gives the best chance of prolonged survival, and this operation is predominantly performed in that country.

R2 gastrectomy carries the same criteria for adequate gastric removal but includes lymphadenectomy to remove Group II (N2) nodes en bloc with stomach. The precise tissues removed depends on the site of the cancer. In general the entire greater omentum is removed, with the superior leaf of the transverse mesocolon, pancreatic capsule, and lesser omentum. Lymph node dissection starts by removing the nodes along the gastroduodenal artery to its origin at the hepatic and is continued laterally to the porta hepatitis along the common hepatic artery. The nodes are cleared medially along the common hepatic artery to the coeliac axis which is cleared and continued along the splenic artery to the hilum of the spleen. Clearance in this area is facilitated by mobilization of the distal pancreas. R3 gastrectomy attempts to remove nodes in Group III (N3) and involves pancreatic and splenic resection. The overall Japanese results demonstrate a corrected 5-year survival following R0, R1, R2, and R3 resections of 26 per cent, 42 per cent, 50 per cent, and 40 per cent. Thus the Japanese have adopted R2 gastrectomy as the usual form of treatment for operable early gastric and advanced gastric cancer. The justification for the use of this radical operation in early tumours is the finding that N1 lymph nodes are involved in 1 to 5 per cent of mucosal and 11 to 19 per cent of submucosal cancers. The rates of metastasis in N2 lymph nodes is 0 to 2 per cent for mucosal and 2 to 9 per cent for submucosal cancers. Metastasis to Group N3 and N4 nodes is unusual. A selective policy could be adopted if the depth of the primary could be accurately assessed preoperatively. Indeed there is no difference in survival after R1 and R2 gastrectomy for mucosal cancer without lymph node metastasis, but a significant survival advantage in patients undergoing R2 resection if lymph nodes are involved. R2 and R3 gastrectomy are performed remarkably safely by Japanese surgeons, with an operative mortality of 1 to 3 per cent.

Resistance to, and failure to adopt the R2 gastrectomy in the West, has occurred because a clear advantage over the traditional Western radical resection has not been demonstrated, predominantly because of varying experience. In the West, the R2 operation is associated with a longer duration of surgery, greater blood loss, longer postoperative hospital stay, and greater morbidity. The Japanese surgeons have demonstrated no difference between postoperative mortality and the level of nodes dissected.

Lesions at the gastric cardia tend to metastasize to all the regional lymph nodes and extended total gastrectomy (R3) is required to encompass all the lymph nodes, in addition to mediastinal dissection. This carcinoma has such a dismal prognosis that almost any surgical procedure is palliative and often a more simple proximal gastric resection is most appropriate.

The treatment of stump cancer should follow the same principles for radical excision. The pattern of lymph node spread is different. The previous resection has often left lymph channels along the left gastric vessels, and lymph drainage from the greater curve and into the jejunal mesentery are more important. Radical excision usually implies total gastrectomy and excision to the origin of the jejunal artery supplying the gastroenterostomy.

The mortality and morbidity of the operation is related to the nature of resection, with distal partial gastrectomy being safest, followed by subtotal gastrectomy, and then total gastrectomy. Factors that are significantly related to postoperative complications, apart from operative procedure, are age, sex, and splenectomy. Men and patients of more advanced age are more prone to complications, as are those who have a splenectomy. In general the mortality of surgical resection of gastric cancer in Western countries is falling, from less than 10 per cent in the 1970s to 2 per cent in the 1990s, and is approaching the results that are obtained in Japan.

The results of surgery are summarized in Table 5 309, which compares series from Japan and the United Kingdom. Overall the survival results are substantially better in Japan than in the United Kingdom.

In the West very few patients present with early gastric cancer; in one large series only 90 of 13000 patients with gastric cancer fulfilled the criteria. Approximately 55 per cent of patients with gastric cancer are suitable for laparotomy, with curative resection in only 21 per cent. The remainder will have no procedure (15 per cent), palliative resection (10 per cent), or bypass (9 per cent). Unfortunately, laparotomy and palliative surgery carry a significant mortality (10–30 per cent) and morbidity, and before embarking on palliative excision it is important to define the goals of this form of surgery. Symptoms due to obstruction, such as dysphagia, vomiting, and obstructive pain may be relieved by surgical resection or bypass (Fig. 8) 942. Bleeding can be relieved by resection. If at laparotomy curative resection is impossible, the alternatives are a palliative resection, a bypass, or no procedure. The consensus is that palliative non-radical resection may provide the best chance of relief of symptoms, and some patients (6 per cent) show prolonged survival after this procedure. Palliative bypass procedures do not increase survival times and are only necessary if there is obstruction. The benefit on the quality of life is highly questionable, and the mean survival for patients is 5 months. There are various methods of non-operative palliation, including laser therapy and intubation for dysphagia, and interstitial laser therapy for bleeding gastric cancers.

Local and regional recurrence occurs in 80 per cent of patients with distant metastases, and at death 16 to 22 per cent have locally recurrent disease only. ‘Second look’ laparotomy in asymptomatic patients without evidence of distant metastasis reveals local recurrence in 69 per cent. These findings have been used to suggest a role for adjuvant radiotherapy. Gastric adenocarcinoma has generally been regarded as radioresistant and, although it is less sensitive than squamous cell carcinoma, useful response and tumour shrinkage has been achieved in patients given palliative radiotherapy for malignant dysphagia. The major limitation to radiotherapy has been the problem of achieving a dose that will spare adjacent normal tissue, including the liver and the small intestine. Intraoperative radiotherapy may overcome some of these problems. In some Japanese centres, after attempted curative resection, intraoperative radiotherapy is administered to the coeliac axis and the tumour bed. There is some evidence to suggest that patients with stage III and IV disease undergoing this therapy have a prolonged survival compared with historical controls. Radiotherapy delivered postoperatively to patients following palliative and curative resection seems to have little influence on survival.

The benefits of radiotherapy are best if bulk disease is controlled by other means. Unfortunately, patients with gastric cancer may have large and unresectable tumours at presentation. It has been suggested that preoperative radiotherapy may be used to shrink the tumour and allow subsequent resection. Unfortunately accurate radiotherapy planning is difficult preoperatively, and there is little evidence that preoperative radiotherapy is beneficial.

The rationale for the use of adjuvant chemotherapy in gastric cancer is clear, since surgery is only potentially curative, and 70 per cent of patients with cancer confined to the gastric wall on pathological examination will die of the disease. Although the disease is apparently localized it has, therefore, in fact extended beyond surgical resection or is already disseminated and systemic therapy is required for any attempt to eradicate the disease. Gastrointestinal cancers are generally unresponsive to chemotherapy, but gastric cancers appear to be more sensitive than most and in particular respond better than colorectal cancer. Despite evidence to suggest that combined chemotherapy is better, most studies have been performed with single agents. Mitomycin C, doxorubicin, 5-fluorouracil, and the nitrosoureas will produce tumour shrinkage in up to 30 per cent of patients with advanced disease. The combination of 5-fluorouracil, doxorubicin, and mitomycin C at present represents the most effective regimen for advanced gastric cancer, with 40 per cent of patients showing a partial response, although only 5 per cent will have a complete response. The translation of a response in advanced cancer into an effective adjuvant regimen in less advanced disease has proved difficult. Again there is differing experience between Western series and the positive Japanese studies. In the extensive European and American trials, no survival benefit has been demonstrated with combinations of fluorouracil, mitomycin, and methyl CCNU, and with 5-fluorouracil, doxorubicin, and mitomycin C administered after potentially curative surgery. The Japanese have demonstrated some survival advantage in patients treated with mitomycin C, alone or in combination, when given intraoperatively or in the very early postoperative period. The trials have been criticized because the survival benefit has been demonstrated by the retrospective formation of subgroups. Overall there is no convincing evidence that cytotoxic chemotherapy is of use after resection for gastric cancer. General application of adjuvant chemotherapy cannot be recommended. The role of preoperative (neoadjuvant) chemotherapy has to be established, but may increase the resectability rate. The combination of postoperative chemotherapy with radiotherapy has also failed to improve survival in this disease.

Since malignant cells are shed into the peritoneum by the tumour, various local methods have been developed to eradicate localized intraperitoneal disease. Local instillation of cytotoxics, or hyperthermic peritoneal perfusion with cytotoxic agents initially appeared effective but no survival benefit was demonstrated in a randomized clinical trial. Palliation of advanced tumours with endoscopic regression and some long-term survival can be achieved by local arterial infusion of cytotoxics, but this remains inappropriate for most patients.

Abe S, Shiraishi M, Nagaoka S, Yoshimura H, Dhar DK, Nakamura T. Serosal invasion as the single prognostic indicator in stage IIA (T3N1M0) gastric cancer. Surgery 1991; 109: 582–8.

Adam YG, Efron G. Trends and controversies in the management of carcinoma of the stomach. Surg Gynecol Obstet 1990; 169: 371–85.

Alcobendas F, Milla A, Estape J, Curto J, Pera C. Mitomycin C as an adjuvant in resected gastric cancer. Ann Surg 1983; 198: 13–17.

Allum H, Powell J, McConkey C, Fielding WL. Gastric cancer: a 25 year review. Br J Surg 1989; 76: 535–40.

Boku T, et al. Prognostic significance of serosal invasion and free intraperitoneal cancer cells in gastric cancer. Br J Surg 1990; 77: 436–9.

Craanen ME, Dekker W, Ferwerda J, Blok P, Tytgat NJ. Early gastric cancer: a clinicopathologic study. J Clin Gastroenterol 1991; 13: 274–83.

De Dombal FT, et al. The British Society of Gastroenterology early gastric cancer/dysplasia survey: an interim report. Gut 1990; 31: 115–20.

Dent DM, Madden MV, Price SK. Randomized comparison of R1 and R2 gastrectomy for gastric carcinoma. Br J Surg 1988; 75: 110–12.

Fielding JWL. Gastric cancer: different diseases. Br J Surg 1989; 76: 1227.

Hioki K, Nakane Y, Yamamoto M. Surgical strategy for early gastric cancer. Br J Surg 1990; 77: 1330–4.

Husemann B. Cardia carcinoma considered as a distinct clinical entity. Br J Surg 1989; 76: 136–9.

Kampschoer GHM, Nakajima T, van de Velde CJH. Changing patterns in gastric cancer. Br J Surg 1989; 76: 914–16.

Lauren P. The two main types of gastric carcinoma: diffuse and so-called intestinal-type carcinoma. Acta Pathol Microbiol Scand 1965; 64: 31–49.

Lawrence M, Shiu MH. Early gastric cancer: twenty-eight year experience. Ann Surg 1991; 213: 327–34.

Maruyama K, Gunven P, Okabayashi K, Sasako M, Kinoshita T. Lymph node metastases of gastric cancer. General pattern in 1931 patients. Ann Surg 1959; 210: 596–602.

Mitsudomi I, Wantanabe A, Matsusaka T, Fujinaga Y, Fuchigami T, Iwashita A. A clinicopathological study of synchronous multiple gastric cancer. Br J Surg 1989; 76: 237–40.

Oota K, Sobin LH. Histological typing of gastric and oesophageal tumors. International Histological Classification of Tumors. 18. Geneva: WHO, 1977.

Paterson IM, Easton DF, Corbishley CM, Gazet J-C. Changing distribution of adenocarcinoma of the stomach. Br J Surg 1987; 74: 481–2.

Schlag P. Adjuvant chemotherapy in gastric cancer. World J Surg 1987; 11: 473–7.

Shiu MH, et al. Influence of the extent of resection on survival after curative treatment of gastric carcinoma. Arch Surg 1987; 122: 1347–51.

Toftgaard C. Gastric cancer after peptic ulcer surgery. A historic prospective cohort investigation. Ann Surg 1989; 210: 159–64.

Yasuna O, Hayashi S. Factors influencing the postoperative course of 113 patients with early gastric cancer. Jpn J Clin Oncol 1986; 16: 325–34.