The most dangerous moment in the life of anyone with bleeding oesophageal varices caused by alcoholic cirrhosis is the instant of the first major haemorrhage (Fig. 1) 1300. The observed risk of death relative to baseline risk is increased 112-fold, and the overall mortality rate is increased by 35 to 70 per cent. More than 1 litre of blood can be lost in 1 h through a defect in the variceal wall only the size of a 22-gauge needle. Even though the chance of death decreases rapidly after the episode of bleeding subsides, rebleeding is common—perhaps even universal. In the evaluation of both clinical trials and the every day treatment of patients with variceal haemorrhage, control of the initial episode of bleeding and control of subsequent episodes of bleeding must be considered separately.

Randomized trials to assess the value of surgery, of obliteration of varices by sclerotherapy, and of division of the variceal columns hinge not only on the variables known by the investigators, but also upon how they are presented. The choice of the zero timepoint of observation in reference to the entire clinical course may bias results. For example, a classic paper extolling the superiority of surgical treatment used as its zero point the time of death of all patients who died in hospital. Once the dead were eliminated, the survival rate of the remainder, who had proved their stamina by living through a major operation and its sequelae, was superb.

My opinion has not wavered from that expressed in 1976: the complexities of portal hypertension confound the best intentioned clinical trials. Studies on the therapeutic value of shunts require uniform and unequivocal diagnosis of bleeding varices, but this never has been achieved. Prospective stratification by age and by integrity of hepatic function of patients to be operated upon and of controls, even in a single hospital, is rarely feasible because the number of patients is too small except in multicentre trials. Co-operative trials are, however, beset by problems ranging from the differing selectivity of investigators to the criteria applied for interpreting findings. Violations of protocol because of the practicalities of clinical care introduce undetectable bias, making comparisons difficult. The several studies done in the 1970s showing that a prophylactic portacaval shunt merely changes the cause of death from variceal bleeding to liver failure are all flawed in light of modern knowledge about statistical power and &bgr; error.

A derivative of these arguments is that any patient whose liver function is so bad that blood clotting is impaired cannot do well. Although that observation is undeserving of a controlled trial, it persists in being substantiated by much investigative effort. Likewise, one would expect patients whose clinical status by Child's criteria for assessment of hepatic reserve (Table 1) 375 puts them in class A to do best, in class B, less well, and in class C, dismally, as they do. A valid means of predicting the likelihood of postoperative death, after either portacaval or proximal splenorenal shunting is shown in Table 2 376 and is depicted in Fig. 2 1301.

The point of the foregoing paragraphs is to urge scepticism. Irrespective of what has been learned from clinical trials, treatment of bleeding varices is empirical. The fact that large varices are those most likely to bleed and that proper clotting is essential to survival are the unequivocal data to which one might repair.

No matter how much blood has been vomited by a patient with bleeding varices on admission to an emergency ward, the chances are better than even that bleeding will stop spontaneously. Letting the patient sit upright seems to help, but the patient who can tolerate sitting upright has to have a good intravascular volume and is thus a selected patient. Parenteral administration of vasopressin to constrict blood vessels, or of nitroprusside and other vasodilators is useless.

Insertion of a Sengstaken-Blakemore tube to tamponade oesophageal and gastric fundal varices is the first action. Table 3 377 outlines the criteria for safe use of the tube, as originally presented by Conn and as fitted to present-day criteria. The Minnesota modification of the tube refers to the presence of a fourth lumen, opening just above the inflated oesophageal balloon, which allows orotracheal secretions pooling above the balloon to be aspirated (Fig. 3) 1302. Formerly, a nasogastric tube for aspiration was tied to the Sengstaken tube above the oesophageal balloon. Since the Minnesota tube is now the only one being manufactured in the United States, there is no alternative. On the other hand, since the airway is protected by endotracheal or nasotracheal intubation with an inflated cuff on the tube to prevent aspiration of blood and secretion into the lungs, the fourth port of the Sengstaken tube is almost superfluous; the port seldom works as well as it should, anyway. The Idezuki tube, made of plastic components, rather than of rubber, is said to have more reliable and uniform characteristics of inflation.

Linton and Nachlas tubes, which have only a single large balloon, within the stomach, are used principally to tamponade gastric fundal varices (Fig. 4) 1303. Unlike the Sengstaken–Blakemore tube, traction must be applied to Linton and Nachlas tubes; hence, pressure necrosis of the oesophagogastric function can occur if they are left in place for more than 18 h.

If bleeding is controlled upon removal of a tamponading balloon tube, upper gastrointestinal endoscopy is performed; extensive ice-water gavage is required to remove pooled blood in the stomach. The objectives thereafter are to identify non-variceal sources of bleeding that can be treated specifically, as well as confirming the diagnosis of bleeding oesophageal varices and to display bleeding gastric varices. Gastric varices require surgical control since they are generally endoscopically inaccessible to sclerotherapy. Although the statistic ingrained in reports is that 30 per cent of bleeding is caused by non-variceal disease, this seems an overestimate. The Sengstaken tube itself cannot be counted upon for long-term control: there is a 60 per cent chance of rebleeding after its removal.

If bleeding is uncontrollable by tube tamponade, by sclerotherapy, by oesophageal stapling, or by banding the varices with Neoprene rings, an emergency portasystemic shunt may be required.

Child's class C patients are the most difficult to treat because of their poor liver function. Patients who need at least six blood transfusions have a mortality rate of about 50 per cent, irrespective of whether or not they are treated by sclerotherapy (Fig. 5) 1304 or by a portacaval shunt. The number of transfusions required is fewer after sclerotherapy, however. Costs of hospitalization are equal whether sclerotherapy or a portacaval shunt is used.

Initially, oesophageal varices that are demonstrably or putatively the cause of haemorrhage are equally well controlled by sclerotherapy with any of a variety of sclerosants (usually ethanolamine oleate, ethanol, or sodium tetradecyl sulphate) or by oesophageal transection and reanastomosis with a stapling device (Figs. 6, 7) 1305,1306. Neither sclerotherapy nor transection, however, prolongs life or prevents rebleeding. Sclerotherapy has the advantage of being a totally endoscopic technique. Its disadvantages are a rebleeding rate of 62 per cent after one session of therapy and of 82 per cent after three sessions. There are myriad complications, some common and potentially severe—such as ulceration, and some rare, but potentially lethal, such as thrombosis of the splenic vein, perforation of the oesophagus, pneumomediastinum, pneumothorax, and pneumoperitoneum. Banding of varices with Neoprene rings, as in a haemorrhoidectomy, is currently thought to be safer.

Stapling has the disadvantage that it requires an intra-abdominal operation, with its attendent complications. Its advantage is that only a single treatment is required and that stapling is at least as good as sclerotherapy over the long term. In the short term, stapling is 88 per cent effective against rebleeding within 5 days of transection, as opposed to a 62 per cent rate of protection after a single sclerotherapy treatment.

Although vasoconstrictive therapy with a vasopressin analogue, &bgr;-adrenergic blockage with propanolol, and vasodilation with glyceryl trinitrate (nitroglycerine) or nitroprusside have all been employed, these modalities are useless in the treatment of acute, massive variceal bleeding. Propanolol in a dose sufficient to reduce the heart rate of cirrhotics by 25 per cent may delay or prevent the initial haemorrhage, but it has no effect on the mortality rate.

Recurrent massive variceal bleeding in a patient who has an open portal vein is better and more economically treated by an emergency portasystemic shunt than by sclerotherapy, if liver function and general health are adequate for surgery. Although one prospective, but uncontrolled trial detected little evidence of portasystemic encephalopathy after an emergency portasystemic shunt, encephalography occurs frequently. The best predictor of postoperative encephalopathy is the presence of preoperative encephalopathy.

In Japan, venous devascularization of the oesophagus and stomach from the level of the inferior pulmonary vein to the left phrenic and the short gastric veins, plus transection of the oesophageal mucosa and oversewing the mucosa, is a common form of treatment. In the West it has never been widely successful as a form of emergency therapy. A rate of recurrent bleeding after devascularization estimated as only 6.6 per cent by some authorities seems lower than that experienced by others.

Andreani T, et al. Preventive therapy of first gastrointestinal bleeding in patients with cirrhosis: results of a controlled trial comparing propanolol, endoscopic sclerotherapy and placebo. Hepatology, 1990; 12: 1413–9.

Branicki FJ, et al. Emergency surgical treatment for nonvariceal bleeding of the upper part of the gastrointestinal tract. Surg Gynecol Obstet, 1991; 172: 113–120.

Brooks, WS Jr. Sclerotherapy: analysis and review. Endosc Rev 1985; May/June; 11–53.

Burnett DA, Rikkers LF. Nonoperative emergency treatment of variceal hemorrhage. Surg Clin N Am, 1990; 70: 291–306.

Burroughs AK, et al. A comparison of sclerotherapy with staple transection of the esophagus for the emergency control of bleeding from esophageal varices. N Engl J Med, 1989; 321: 857–62.

Cello JP, et al. Endoscopic sclerotherapy versus portacaval shunt in patients with severe cirrhosis and acute variceal hemorrhage. N Engl J Med, 1987; 316: 11–15.Gitlin N. Treatment options in the management of varices. Curr Opin Gastroenterol, 1991; 7: 357–63.

Grace ND. Prevention of recurrent variceal bleeding—is surgical rescue the answer? Ann Intern Med, 1990; 112: 242–4.

Henderson JM, et al. Endoscopic variceal sclerosis compared with distal splenorenal shunt to prevent recurrent variceal bleeding in cirrhosis. Ann Intern Med, 1989; 112: 262–9.

Idezuki Y, et al. Current strategy for esophageal varices in Japan. Am J Surg, 1990; 160: 98–104.

Jansen PLM. Current management of esophageal varices. Curr Opin Gastroenterol, 1990; 6: 597–602.

Lacaine F, LaMuraglia GM, Malt RA. Prognostic factors in survival after portasystemic shunts: multivariate analysis. Ann Surg, 1985; 202: 729–34.

Malt RA. Portasystemic venous shunts. N Engl J Med, 1976; 295: 24–29, 80–86.

O'Connor KW, et al. Comparison of three nonsurgical treatments for bleeding esophageal varices. Gastroenterology, 1989; 96: 899–906.

Orloff MJ, Bell RH. Long-term survival after emergency portacaval shunting for bleeding varices in patients with alcoholic cirrhosis. Am J Surg, 1986; 151: 176–83.

Ottinger LW, Moncure AC. Transthoracic ligation of bleeding esophageal varices in patients with intrahepatic portal obstruction. Ann Surg, 1974; 179: 35–8.

Planas R, et al. Portacaval shunt versus endoscopic sclerotherapy in the elective treatment of variceal hemorrhage. Gastroenterology, 1991; 100: 1078–86.

Smith JL, Graham DY. The variceal hemorrhage: a critical evaluation of survival analysis. Gastroenterology, 1982; 82: 968–73.

Sugiura M, Futagawa S. Esophageal transection with paraesophagogastric devascularizations (the Sugiura procedure) in the treatment of esophageal varices. World J Surg, 1984; 8: 673–82.

Terblanche J, Burroughs AK, Hobbs KEF. Controversies in the management of bleeding esophageal varices. N Engl J Med, 1989; 320: 1393–8, 1469– 74.