Most episodes of variceal haemorrhage will be successfully treated by resuscitation and injection sclerotherapy, and indeed the efficacy of this approach has now been confirmed by several well-controlled trials. Nevertheless, in a significant minority either acute bleeding will persist, or rebleeding will occur in the near or long term. In this situation there is little doubt that some form of portal–systemic decompression is the most successful treatment for bleeding.

However, this objective is achieved at a cost, because disabling portal–systemic encephalopathy and deterioration in liver function over subsequent years frequently develops in patients who have been treated by shunting. Also the operative mortality of emergency shunting procedures is high, even where hepatic reserve is good. Furthermore, in variceal bleeding from extrahepatic portal hypertension, 30 to 50 per cent of children and 40 to 50 per cent of adults will have extensive thrombosis of the splenic and mesenteric veins which makes decompressive procedures impossible. For all these reasons an alternative to portal decompression is required and clearly the repertoire of any surgeon involved in the treatment of variceal haemorrhage must include non-shunt procedures.

Many different non-shunt approaches to the treatment of bleeding oesophageal varices have been developed ranging from the direct attack by oesophagogastrectomy to variations on the theme of portoazygous disconnection and these are listed in Table 1 381. Indeed this problem has taxed the ingenuity of surgeons since the late nineteenth century when splenectomy and omentopexy were originally performed. The wide range of procedures devised since then reflects the generally unsatisfactory results achieved, particularly so in the presence of poor liver function. In recent years, the advent of stapling devices has reawakened interest in portoazygous disconnection as a suitable alternative to shunting. Currently, the importance of these procedures in comparison to injection sclerotherapy and portacaval shunting is under evaluation around the world.

In most centres the first line of management of variceal haemorrhage is by resuscitation, medical therapy with a variety of vaso-active agents, balloon tamponade as a holding procedure, and injection sclerotherapy of the varices. Operative treatment is generally reserved for continued bleeding, or where recurrent variceal haemorrhage is a problem (Table 2) 382. For example, it is now clearly established that if bleeding is not controlled after two sets of variceal injection, high rebleeding and mortality rates result from further attempts at injection; in this group emergency operative intervention is the best option.

Variceal bleeding in the presence of good hepatic function but a large spleen, for example in mansonic schistosomiasis, is a clinical scenario where a non-shunt procedure has many advantages over portal decompression. Usually massive splenomegaly and hypersplenism occur in a young patient with an excellent prospect of good life-long hepatocellular function. Shunting not only carries an immediate high incidence of portal–systemic encephalopathy but also in non-selective decompressive procedures an inexorable deterioration of liver function will take place over the years. This hepatocellular deterioration becomes highly relevant in patients with conditions such as schistosomiasis or extrahepatic portal hypertension where life expectancy and liver function are good. Oesophageal transection and oesophagogastric devascularization (Hassab and Sugiura procedures) offer low rates of rebleeding and advantages over shunting in terms of preserving liver function and avoiding portal–systemic encephalopathy. Unfortunately the dearth of properly constructed trials comparing shunt and non-shunt techniques makes it impossible to be more objective about these two surgical approaches. Obviously, when the splenic and superior mesenteric veins are thrombosed there is no alternative to using a non-shunt technique. Portosystemic encephalopathy is a devastating complication adversely affecting intellectual function.

Cirrhotic patients increasingly come to liver transplantation in the end-stages of their disease. Previous surgery, particularly portacaval shunting, will complicate the procedure but will also increase its morbidity and mortality, and adhesions around the left lobe of the liver and throughout the left upper quadrant from non-shunt procedures will undoubtedly increase the technical problems of the recipient hepatectomy. Arguably, however, dealing with these adhesions might be less difficult than the dissection around the portal structures with the added problems of sudden massive portal hypertension which occurs during the disconnection of patent portacaval shunt in a transplant. At present, where transplantation is not yet indicated, a mesocaval shunt is the operation most favoured by transplanters for control of variceal haemorrhage because of the minimal dissection around the liver. Simple oesophageal transection, however, will reliably control haemorrhage where injection sclerotherapy has failed, with much less adhesion development, and in this emergency situation must be the procedure of choice.

As in shunting, the more major devascularization procedures have a higher operative mortality in the emergency situation, particularly in the setting of poor hepatic reserve. Once again, an expeditiously performed simple oesophageal transection, using the stapler but without major devascularization, has been shown to be extremely effective in treatment of haemorrhage uncontrolled by medical therapy or injection.

Full investigation and assessment of liver disease and function is essential. Any clotting abnormality should be corrected, particularly in the presence of hypersplenism, and renal function assessed, monitored, and optimally maintained before, during and after the procedure. Preoperative angiography may be of value if devascularization is to be performed, providing a ‘road map’ of the collateral circulation around the oesophagogastric region. If splenectomy is to be performed, vaccination against pneumococcal infection is probably of value, as is maintenance low-dose penicillin cover for the perioperative period, and for at least 2 years afterwards.

Techniques such as transthoracic ligation of oesophageal varices, Tanner's high gastric disconnection, and splenic artery and coronary vein ligation have little to offer over current medical and surgical treatment and have been largely abandoned. However, these methods find occasional use in the emergency situation, particularly where equipment and expertise may be lacking, and where transfer of a desperately ill patient across large distances to an experienced centre is not feasible. In the vast majority of centres, the currently favoured non-shunting procedures are simple stapled oesophageal transection, oesophagogastric devascularization, or a combination of the two.

Interest in portoazygous disconnection was reawakened after Vanhemmel in 1974 first used a stapling device (the Russian SPTU stapling gun) in transection of the oesophagus for bleeding varices. With the advent of simpler more effective stapling guns (Autosuture and Ethicon) came a worldwide upsurge of interest in oesophageal transection, particularly in the emergency situation.

The procedure is performed via a left subcostal or upper midline incision. The left lobe of the liver is retracted and the oesophagogastric junction identified; retraction may be difficult in a rigid cirrhotic liver and in this case, mobilization of the left lobe may be necessary. The overlying peritoneal reflection is incised using diathermy to coagulate the multiple small vessels which develop in this layer as a result of portal hypertension. The oesophagus is mobilized with its surrounding para-oesophageal tissues, from the oesophagocardiac junction to the upper margins of the diaphragmatic crus, a length of approximately 5 cm. Difficulty may be experienced with this mobilization in patients who have previously undergone injection sclerotherapy where thickening of the para-oesophageal tissues often develops. Despite this handicap it is always possible to mobilize the oesophagus without complication. Ideally the vagus nerves should be mobilized, although no postvagotomy complications were found in a recent large series from the Royal Free Hospital (London) where no specific attempt to mobilize the vagi was made. A strong linen thread is passed around the mobilized distal oesophagus and a short anterior gastrotomy is made. After assessing the lumen of the oesophagus with passage of calibrated obturators, the appropriate size of staple gun is passed into the oesophagus. The head is then separated from the anvil, and the linen thread tied tightly on to the shaft of the gun just above the oesophagogastric junction. After removal of any sling or tape around the oesophagus, the anvil and staple cartridge are approximated and the gun fired. In one manoeuvre, the oesophagus is transected with removal of a 1-cm ring of oesophagus, and the ends reanastomosed with a secure double layer of steel clips (Fig. 1) 1308.

The gun is opened and the doughnut inspected, a complete ring indicating a technically satisfactory transection. A nasogastric tube is carefully guided through the anastomosis and its tip placed in the body of the stomach before closure of the gastrotomy. The abdomen is closed without drainage, the patient allowed to take small sips of water, and the nasogastric tube usually removed after 2 days before allowing return to normal oral intake.

This technically straightforward procedure can be performed within 1 to 2 h and usually with little blood loss. Varying degrees of devascularization can be added to the procedure with most authors advocating ligation of peri-oesophageal collaterals and the left gastric or coronary veins.

The principle behind this surgical approach is the disconnection of the venous circulation of the distal oesophagus and cardia from the hypertensive portal circulation by division of all the feeding vessels. The most frequently employed operations are the Hassab devascularization and Sugiura devascularization with oesophageal transection.

The Hassab procedure consists of devascularization of the upper half of the stomach and oesophagus (Fig. 2) 1309. The first step is usually splenic artery ligation followed by careful mobilization of the spleen. This mobilization as in all dissections in portal hypertension, requires patient ligation and coagulation of multiple collaterals within the peritoneal reflections, and after individual ligation and division of the short gastric vessels, the spleen is removed. The whole proximal stomach is then devascularized from the terminal two branches of the left gastric artery at the incisura angularis upwards by ligation and division of the lesser and greater omentum, and of the posterior gastric adhesions. After division of the oesophagogastric reflection of peritoneum and mobilization of the vagi, the distal 7 to 8 cm of oesophagus is mobilized and all feeding vessels are ligated and divided. Exposure in this part of the procedure is much facilitated by the use of costal margin retractors. The distal 3 cm of oesophagus and proximal 5 cm of stomach may then be opened longitudinally thus displaying the varices and allowing obliteration of each variceal column by undersewing from as high as possible within the oesophagus with an absorbable suture. After positioning of a nasogastric tube the oesophagogastrotomy is carefully closed by suturing or stapling. Some authors recommend closure by swinging a flap of stomach wall into the oesophageal defect, thus minimizing oesophageal stricturing.

The Sugiura operation is a much more radical development of the above method, classically performed in two staged procedures. At the first operation, via a left thoracotomy, the distal intrathoracic oesophagus is devascularized and an oesophageal transection performed. Six weeks later, via an upper abdominal midline incision, the intra-abdominal oesophagus and proximal stomach are devascularized by lesser and greater curve division and splenectomy. Vagotomy and pyloroplasty are then performed (Fig. 3) 1310. This massive procedure has been modified into a one-stage operation using a transabdominal approach facilitated by the use of costal margin and sternal retractors. After division of the crura of the diaphragm, 10 cm of oesophagus can be devascularized, a staple transection performed via a gastrotomy and the rest of the abdominal part of the operation completed.

Because of high long-term rebleeding rates of up to 50 per cent, simple oesophageal transection without some form of devascularization is rarely performed as a definitive, elective procedure. In the emergency situation however, transection alone has beenshown to be highly effective at controlling variceal haemorrhage and in a randomized controlled trial from the Royal Free Hospital was found to have morbidity and mortality similar to injection sclerotherapy across all grades of liver disease. Rebleeding does not occur before 3 months but when it does develop, is frequently not from varices at all but from ulceration at the transection line which is readily controlled by omeprazole. The incidence of portal-systemic encephalopathy is very low and dysphagia, although frequent, is usually transient but when persistent, is cured by dilatation.

The major complications of devascularization procedures are those of the oesophageal transection, possibly compounded by devascularization ischaemia, namely dysphagia, ulceration, and stricture formation but with an extremely low rate of rebleeding and recurrence of varices when compared with simple transection alone. All of these procedures are characterized by little or no increase in portal-systemic encephalopathy.

The majority of series of devascularization procedures report a significantly high risk of rebleeding although there is considerable variability in reported results. The exception to this is Sugiura's results with his two-stage operation wherein he reported an overall mortality of 4.9 per cent with a mortality of 13.3 per cent for emergency procedures. In a different series from Kuwait, Abouna reported a similar overall mortality of 7.7 per cent for his one-stage modification of the Sugiura technique. Both series had extremely low rebleeding rates of 1.5 and 3.4 per cent respectively with no increase in portal-systemic encephalopathy. These results are the best reported for any treatment for oesophageal varices, but unfortunately no Western group has been able to reproduce them, largely it is claimed because of the higher preponderance of alcoholic cirrhosis and hepatitis in the West. Indeed the Western experience of the Sugiura technique is characterized by high morbidity and mortality. Excellent results however, have been reported in the management of variceal bleeding in schistosomal portal hypertension where long-term liver function is good. Devascularization procedures, in particular the modified Sugiura, provide the treatment of choice for this condition.

Despite the success of injection sclerotherapy there remains a place for surgical treatment of bleeding oesophageal varices which is being slowly defined. In general, non-shunt procedures offer advantages over portosystemic shunting mainly in terms of lower incidence of portosystemic encephalopathy and maintenance of portal perfusion of the liver. Simple oesophageal transection is a proven, effective means of controlling acute variceal haemorrhage, particularly where injection sclerotherapy has failed. In this emergency situation this simple approach has many advantages and is probably the treatment of choice. However, this method is less valuable in the elective situation because of its high rebleeding rate, and thus the modified Sugiura procedure has much to offer although Western experience has been disappointing. Because properly conducted comparison is not available, the choice of approach, shunt or non-shunt, is determined by the local disease patterns, local surgical expertise, and management bias.

Abouna GM, Baissony H, Al-Nakib BM, Menkarios AT, Silva OSG. The place of Sugiura operation for portal hypertension and bleeding esophageal varices. Surgery, 1987; 101: 91–8.

Bornman PC, Terblanche J, Kahn D, Jonker MA, Kirsch RE. Limitations of multiple injection sclerotherapy sessions for acute variceal bleeding. S Afr Med J, 1986; 70: 34–6.

Burroughs AK, Hamilton G, Phillips A, Mezzanotte G, McIntyre N, Hobbs KEF. A comparison of sclerotherapy with staple transection of the esophagus for the emergency control of bleeding from oesophageal varices. N Engl J Med, 1989; 321: 857–62.

Gouge TH, Ranson JH. Oesophageal transection and paraoesophagogastric devascularization for bleeding oesophageal varices. Am J Surg, 1986; 151: 47–54.

Hassab MA. Gastro-oesophageal decongestion and splenectomy in the treatment of oesophageal varices in bilharzial cirrhosis: further studies with a report on 355 operations. Surgery, 1967; 61: 169–76.

Hosking SW, Johnson AG. What happens to oesophageal varices after transection and devascularization. Surgery, 1987; 101: 531–4.

Huizinga WK, Angorn PA, Baker LW. Esophageal transection versus injection sclerotherapy in the management of bleeding oesophageal varices in patients at high risk. Surg Gynecol Obstet, 1985; 160: 539–46.

Kaye GL, McCormick PA, Siringo S, Hobbs KEF, McIntyre N, Burroughs AK. Bleeding from staple line erosion after esophageal transection: effect of omeprazole. Hepatology, 1992; 15: 1031–5.

Matory WE Jr, Sedgwick CE, Rossi RL. Non-shunting procedures in management of bleeding esophageal varices. Surg Clin N Am, 1980; 60: 281–95.

McCormick PA, et al. Esophageal staple transection as a salvage procedure after failure of acute injection sclerotherapy. Hepatology, 1992; 15: 403–6.

Raia S, Mies S, Macedo AL. Surgical treatment of portal hypertension in schistosomiasis. World J Surg, 1984; 8: 738–52.

Sherlock S. The portal venous system and portal hypertension. In: Sherlock S, ed. Diseases of the liver and biliary system. 8th edn. Oxford: Blackwell Scientific Publications, 1989: 151–207.

Sugiura M, Futagawa S. Results of six hundred and thirty-six oesophageal transections with paraoesophagogastric devascularization in the treatment of oesophageal varices. J Vasc Surg, 1984; 1: 254–60.