In the haemodynamic analysis of the splanchnic venous system, complex interplay among flow, resistance, shunting, and ignorance exceed the bounds of simple models and heuristic arguments prevail over logic. But one over-riding fact is that the principles of newtonian flow in rigid tubes do not hold when applied to the flexible, elastic veins. Moreover, the splanchnic venous system is not freely intercommunicating. Functional or anatomic blocks may isolate one part so that decompression of another part has no effect on it. For example, isolated splenic vein thrombosis causes left upper quandrant varices and is curable only by a splenectomy. To the contrary, the huge retroperitoneal collateral veins sometimes connecting the splenic vein to the renal vein in portal hypertension are ineffective decompression.

A ‘total’ shunt in a normal person deprives the hepatocytes of all splanchnic venous blood and leads to portasystemic encephalopathy. Modern research into the problems actually began with just such a preparation in a patient whose portal vein was purposely resected in 1952 during a pancreatectomy for cancer, followed by an end-to-side anastomosis of the superior mesenteric vein to the inferior vena cava. By and large, however, total shunts are infrequent nowadays, as compared with 1977, a decade ago, and patients with a normal liver rarely undergo a total shunt except in cases of portal-vein injury.

In patients with long-standing portal hypertension inferences that some shunts are total shunts because of apparent flow patterns of radiographic contrast media are heavily influenced by artefacts and are thus inexact. Doppler duplex ultrasonography seems more valid; it shows that in a ‘total’ side-to-side portacaval shunt total loss of splanchnic flow does not occur, an observation confirmed by other means. In almost every case, compensatory mechanisms increase hepatic arterial flow and collateral splanchnic venous flow so that perfusion of the hepatocytes continues—sometimes to a nearly normal level, sometimes very little; sometimes with considerable splanchnic venous blood, sometimes with almost none. Moreover, the relationship even between the classic Eck's fistula and ‘meat intoxication’ in dogs may have to be revised in light of studies showing that adequate nutrition prevents hepatic coma, but not hepatic atrophy.

By design, the end-to-side portacaval shunt removes residual direct splanchnic venous perfusion of the hepatocytes and should thus be ‘total’. In addition, the side-to-side portacaval shunts, the proximal splenorenal shunts (with splenectomy), the mesocaval shunts, and the renoportal shunts also have the potentiality for being nominally total. Whether or not any one of them actually is depends on the balance among inflow pressures, intrahepatic and intravascular resistances, shunts between the hepatic arterial system and the portal venous system, and sites of decompression by collateral veins.

Only the selective distal splenorenal shunt is, indeed, selective in the sense of not depriving the liver of portal venous circulation in the early postoperative phase. But the choice of patients for a distal splenorenal shunt requires the preselection of those with ‘prograde’ (centripetal) flow of portal-vein blood, thus coincidentally selecting the best-risk patients—those whose cirrhosis is not so bad as to create a high resistance to splanchnic venous perfusion or to cause hepatic arterial shunting into the portal circulation. With time, even selective shunts become non-selective and act as functional side-to-side shunts. The pancreas may become overgrown with collateral veins, requiring a hazardous devascularization to correct it.

The end-to-side portacaval shunt is appropriate for both the emergency arrest of otherwise uncontrollable bleeding from oesophageal varices and the elective control of varices in patients who have bled. Results following its use may be as good as those following use of any of the more complicated shunts. The difficulty is in predicting which patients will do so well that their lives are almost normal, except for the limits imposed by the primary disease, and which will do badly from hepatic failure and portasystemic encephalopathy.

Prophylactic shunts should not be done for patients whose varices have never bled because they only change the mode of death from bleeding to hepatic encephalopathy and failure. For the patient whose varices have bled, conventional wisdom derived from results of randomized trials also says there is no value to shunting in prolonging life. Yet, analysis of the statistics from the studies on which this conclusion was reached indicates to the surgeon potential flaws in the arguments. Most patients entered into the trials were good risks, stratification was imprecise, the populations examined were usually small, varying skills of different surgeons and hospitals could not be controlled, and the &bgr;-error and power of negative observations was never calculated; the issue of statistical power was never calculated; the issue of power was considered only in studies done during the early 1980s. Thus, studies describing a 20 per cent increase in survival rate at 3 years after an end-to-side shunt may be more ‘significant’ than has been credited. An increase in the number of patients studied might have yielded generally accepted statistical significance.

Once it is granted that a shunt is appropriate treatment for a patient with varices that have bled—lesser measures having failed or being inappropriate—an end-to-side shunt is not only the easiest to do but the one most nearly certain to control variceal bleeding. An end-to-side shunt can even be done without preliminary imaging studies. It can be done with considerable confidence that the portal vein will be patent, otherwise suitable for use, and nearly certain to work. Doppler duplex ultrasonography promises to add certainty and reduce the need for preliminary angiography.

Although the techniques of end-to-side shunting are described elsewhere and will not be considered here, a few principles should be mentioned.

(a)The focus of dissection should be limited to the relevant portion of the inferior vena cava and the nearby portal vein.

(b)Initial exposure is facilitated by reflecting the duodenum with Kocher's manoeuvre, normally an easy dissection.

(c)Other structures in the portal triad should be retracted from the portal vein, not dissected individually.

(d)Only the exposure of the medial side of the portal vein entails a dangerous dissection.

(e)Clamping the hepatic end of the vein does not require good visualization of it, if an appropriate method is used.

(f)The vein must run in a good haemodynamic line to the medial aspect of the vena cava. The presence of a replaced hepatic artery running along the lateral aspect of the triad may require ingenious rerouting of portal vein; an impeding lip of pancreas or a wad of dense fat may require division.

(g)Caution should be exercised in using the vena cava as a conduit for decompression if there is angiographic or haemodynamic evidence of its compression by an engorged liver (as in Budd-Chiari syndrome) or a nodular caudate ‘lobe’; nonetheless, most ostensible compression disappears when the patient is erect and any ascites present are drained. Actual anatomic obstructions in the vena cava (webs, for instance) are, of course, another story.

Failure of an end-to-side portacaval shunt to control oesophageal varices usually means that two walls of the vessels have been stitched together or that a clot has been allowed to form in the portal vein. The presence of postoperative ascites does not mean the shunt is thrombosed, internists' views to the contrary.

Aside from its utility in relieving portal hypertension, the end-to-side portacaval shunt can be used to correct metabolic disease by eliminating hepatic transformation of substances in the splanchnic venous circulation. It (and its modifications) can ameliorate hypoglycaemia and promote glycogenolysis in patients with some forms of glycogen storage disease, and can reduce synthesis of cholesterol and mobilize lipoprotein deposits in patients with homozygous familial hypercholesterolaemia.

Not only does the side-to-side portacaval shunt have the propensity for decompressing the distal portal vein and its tributaries, it can decompress the hepatic end of the portal vein, reducing intrahepatic portal venous pressure. While on the one hand this decompression may be desirable to remedy effects of a Budd-Chiari syndrome or to treat the now rare case of intractable ascites, it has the potential liability of depriving the liver of even more portal circulation than an end-to-side shunt, i.e., in cirrhotic patients the characteristic flow of hepatic arterial blood through arteriovenous shunts at the sinusoidal level could be diverted down the cephalic end of the side-to-side decompression, depriving the liver further of its vascular supply. As stated above, however, these arguments may not be valid. But because of this uncertainty and although it is widely used, the side-to-side shunt for emergency decompression of the splanchnic circulation is not considered appropriate by many authorities. It may, however, be useful in elective operations. Recent evidence suggests that a controlled 8-mm diameter side-to-side portacaval shunt has the potentiality for decompressing the portal circulation enough to prevent variceal bleeding, while permitting enough splanchnic hypertension to avoid the consequences of encephalopathy after total decompression.

The direct side-to-side shunt is more difficult to perform than the end-to-side portacaval shunt because the portal vein and the inferior vein may be too far apart to be joined easily or may have their potential route of connections blocked by an engorged caudate ‘lobe’ of the liver. Generalized engorgement of the liver may compress the intrahepatic vena cava, raising its pressure considerably and vitiating use of the vena cava as a low-pressure avenue for decompression. Narrowing of a side-to-side shunt from tension on the anastomosis caused by pulling two distant veins together predisposes to thrombosis. Indeed, the very disease for which this operation is best (Budd-Chiari syndrome) is the one most likely to give rise to all these problems. Thus, if division of several lumbar veins and dissection of the portal vein do not produce enough mobility for easy approximation of the portal vein and the inferior vena cava, an interposition (‘H’) graft of a vascular prosthesis or of autogenous vein is required. A double-barrel end-to-side portacaval shunt is particularly efficacious when circumstances allow its construction.

The principle of a mesocaval shunt was originally proposed for decompression of portal hypertension in childhood, when the splenic vein and other peripheral splanchnic veins were considered too small for use.

Although one might anticipate this shunt would function like a mere side-arm tap off the portal vein—that is, like a side-to-side shunt—relationships of flow and resistance are such that it, too, can divert too much portal-vein blood into the systemic circulation. In terms of function, it has no advantage over a side-to-side portacaval shunt.

Nonetheless, the mesocaval shunt is undoubtedly valuable in decompressing the portal vein when the portal vein is occluded, but the superior mesenteric vein is open. It is an alternative to the side-to-side portacaval shunt for treatment of the Budd-Chiari syndrome. Extreme obesity and infrahepatic scarring may be relative indications for its use, as compared with a direct venovenous shunt. For glycogen storage disease and homozygous familial hypercholesterolaemia, the mesocaval shunt should work like the end-to-side portacaval shunt.

In terms of ease of use, those well-versed say it is a simpler operation than the portacaval shunt. Even if this assertion is correct, the H-graft mesocaval shunt must be considered less desirable than a portacaval shunt because it introduces problems of kinking and occlusion of the superior mesenteric vein, of clotting in the prosthesis, of infection, and of erosion into the duodenum. The international incidence of thrombosis and other complications is certainly 20 per cent and may actually be over 30 per cent. Use of autologous jugular vein as the H-graft offers no advantage in adult cirrhotics. In children with extrahepatic splanchnic venous occlusion or hypertension second to biliary atresia, cystic fibrosis, or congenital hepatic fibrosis, however, a high success rate may be possible, even if the vein has a diameter of only 5 mm.

For emergency shunting a small and imperfect randomized trial of the mesocaval shunt concluded it had no advantages over the end-to-side portacaval shunt for surgeons reared in the tradition of direct shunting. Instead, the facility of working with autogenous veins made the direct end-to-side portacaval shunt more appropriate.

Haemodynamics of this shunt are doubtless the same as those of the H-graft, and its range of applications is the same. Its overwhelming utility is that it is unlikely either to erode into the duodenum or another structure, or to kink the superior mesenteric vein, because lines of stress on the vein wall are better. Too few operations have been done for the likelihood of thrombosis overall to be accurately assessed. In expert hands, a prevalence of patency over 90 per cent is possible.

Obliteration of the portal vein and of its intrahepatic branches from neonatal omphalitis and thrombophlebitis often spares the superior mesenteric vein. Until recently, the best shunt for children with portal hypertension from any extrahepatic block of this kind was a Marion–Clatworthy–Valdoni operation, which entails dividing the inferior vena cava, turning it up, and anastomosing it to the mesenteric vein for decompression. Nowadays the procedure is obsolescent because good results are possible by direct splenorenal anastomosis, even in small children, unless the splenic vein is also thrombosed (see below). The side-to-end shunt should never be used in adults because the frequent massive oedema of the lower extremities is too great a price to pay, as long as alternative forms of decompression exist.

Because it is harder to perform than any of the shunts previously discussed, twice as liable to clot than the end-to-side portacaval shunt (29 versus 14 per cent worldwide), introduces the risk of overwhelming postsplenectomy sepsis in children, and is not suitable for emergency use because it decompresses the portal system too slowly, unique applications must be found if the proximal splenorenal shunt is to have a therapeutic role. The main acceptable indication is extrahepatic portal hypertension in children who have a patent splenic vein.

Figure 1 1311 shows that decompression with the expectation of long-term patency is possible in young children. Results such as these and the 94 per cent patency rate in an earlier series are putting to rest the old saw that bleeding varices in childhood should be treated non-operatively, waiting until the child reaches his or her teens before attempting a shunt. While the traditional advice is true in the sense that children almost never die from variceal haemorrhage because their cardiovascular system is resilient, and some children stop bleeding as they age, the risks of hepatitis, cytomegalovirus infection, and acquired immune deficiency syndrome (AIDS) from blood transfusions may now be greater than those of surgery.

Relief of splanchnic venous hypertension and hypersplenism in patients with schistosomiasis and other forms of presinusoidal hypertension is also feasible with the proximal splenorenal shunt, but with the risk of a 31 per cent incidence of encephalopathy over the short term. However, the alternative operation, a selective distal splenorenal shunt, gives results at least as good in terms of control of varices and preservation of the spleen, and is associated with only a 13 per cent incidence of encephalopathy. (Oesophagogastric disconnection may in fact be superior to either.) The larger splenic vein in these diseases facilitates either type of anastomosis.

Considering the number of cirrhotic American patients only, the great question is where the proximal splenorenal shunt with splenectomy stands in the priority list of operations for patients with bleeding varices. There is no question that some patients do superbly after this operation. Unfortunately, as with all portasystemic shunts, matching the type of shunt to the patient remains the dilemma.

As a result of an unpublished trial of 30 patients randomized between a proximal shunt and a distal shunt at our hospital, only two conclusions are justified: (a) It was impossible to predict from any preoperative study whether the proximal shunt or the distal shunt would be easier to do. Anatomic situations that looked difficult beforehand for one kind of operation turned out to be easy, and vice versa. (b) The patient with the worst encephalopathy had a selective distal splenorenal shunt.

Analysis of the results of shunting 120 patients in the 8 years from 1966 to 1973 (portacaval shunt, 57 per cent of total; proximal splenorenal shunt, 43 per cent) showed no differences in survival rates or encephalopathy rates between patients who underwent a portacaval shunt or a splenorenal shunt. These data were compared with those of 141 patients in the 8 years from 1974 to 1981 (portacaval shunt, 23 per cent of total; proximal splenorenal shunt, 58 per cent, mesocaval shunt, 12 per cent, distal splenorenal shunt, 6 per cent, coronary-caval shunt, 1 per cent).

Although the Child-Turcotte criteria (Table 1) 383 were useful in predicting the mortality rate of operations, the criterion of ‘encephalopathy’ was not a specific one because it predicted the likelihood of future encephalopathy (83 per cent accuracy) rather than the probability of survival. Nutritional status was rarely assessed accurately.

Considering the 1974–1981 patients, the validity of a simple six-point scale to predict the likelihood of a postoperative death was confirmed ( Table 2 384, Fig. 2 1312) using the last data collected before an operation. An equation derived by logistic regression identified independent prognostic significance for an emergency operation, serum albumin and bilirubin concentrations, age, and sex (men, worse). With these data the cutpoint probability of 0.75 separated patients who had an 84 per cent chance of survival (above 0.75) from those below, who had at least a 77 per cent chance of death (Fig. 3) 1313.

A Cox regression model for long-term survival after an emergency operation defined male sex and a prolonged partial thromboplastin time as poor prognostic factors, while after an elective operation only the serum albumin level was prognostic. Once a patient had survived the emergency operation (46 per cent mortality rate versus 9 per cent elective), his or her likelihood of 5-year survival (30 per cent) was as good as that of survivors of elective operations (37 per cent).

The rate of encephalopathy was 35 per cent among all patients and over all time, if encephalopathy was defined by the most sensitive criterion (any report of encephalopathy by any physician); the severity of encephalopathy was not assessable. The frequency of encephalopathy after end-to-side portacaval shunts in a randomized trial has been examined in only one study. That study showed an incidence in all of encephalopathy control (non-shunted) patients rising from 18 to 38 per cent during 4 years of observation, compared with 20 per cent, rising to 53 per cent, in shunted patients. The difference between 38 and 53 per cent was not statistically significant. The rate of severe encephalopathy, however, was 3 per cent in control subjects and 20 per cent after portacaval shunts. Contemporary repetition of such a study with all modern controls and means of estimating encephalopathy would be highly desirable, especially if extended to comparisons with proximal and the distal splenorenal shunts.

Alvarez F, Bernard O, Brunelle, F, Hadchouel P, Odievre M, Alagille D. Portal obstruction in children. II. Results of surgical portasystemic shunts. J Pediatr 1983; 5: 703–7.

Barsoum MS, Rizk-Allah MA, El-Said Khedr M, Khattar NY. A new posterior exposure of the splenic vein for an H-graft splenorenal shunt. Br J Surg 1982; 69: 376–9.

Bismuth H. End-to-side splenorenal anastomosis. In: Malt RA, ed. Surgical Techniques Illustrated. Philadelphia: WB Saunders, 1985: 465–73.

Bismuth H, Franco D. Portal diversion for portal hypertension in early childhood. Ann Surg 1976; 183: 439–46.

Callow, AD. Portacaval shunts. World J Surg 1984; 8: 688–97.

Cello JP, et al. Factors influencing survival after therapeutic shunts: results of a discriminant function and linear logic regressions analysis. Am J Surg 1981; 141: 257–65.

Child CG III, Turcotte JG. Surgery and portal hypertension. In: Child CG III, ed. The Liver and Portal Hypertension, Vol I. Dunphy JE, ed. Major Problems in Clinical Surgery. Philadelphia: WB Saunders, 1964:1–85.

Conn HO, Lindenmuth WW, May CJ, Ramsby GR. Prophylactic portacaval anastomosis. A tale of two studies. Medicine 1972; 51: 27–40.

Donovan AJ. Surgical treatment of portal hypertension: a historical perspective. World J Surg 1984; 8: 626–45.

Gliedman ML. Mesocaval shunt. In: Malt, RA, ed. Surgical Techniques Illustrated. Philadelphia: WB Saunders, 1985: 442–8.

Inokuchi K. Present status of surgical treatment of oesophageal varices in Japan: a nationwide survey of 3,588 patients. World J Surg 1985; 9: 171–80.

Jackson FC, Perrin EB, Felix WR, Smith AG. A clinical investigation of the portacaval shunt: V. Survival analysis of the therapeutic operation. Ann Surg 1971; 174: 672–701.

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

McDermott WV Jr. The treatment of cirrhotic ascites by combined hepatic and portal decompression. N Engl J Med 1958; 259: 897–901.

McDermott WV Jr. Surgery of the Liver and Portal Hypertension. Philadelphia: Lea & Febiger, 1974: 117–23.

McDermott WV Jr, Adams RD. Episodic stupor associated with an Eck fistula in the human with particular reference to the metabolism of ammonia. J Clin Invest 1954; 33: 1–9.

McNamara DJ, et al. Treatment of familial hypercholesterolemia by portacaval anastomosis: effect on cholesterol metabolism and pool sizes. Proc Natl Acad Sci USA 1983; 80: 564–8.

Malt RA. Emergency and elective operations for bleeding oesophageal varices. Surg Clin N Am 1974; 54: 561–71.

Malt RA. Portasystemic venous shunts. N Engl J Med 1976; 295: 24–9, 80–6.

Malt RA. Proximal splenorenal venous shunts. In: Rutherford RB, ed. Vascular Surgery. 2nd edn. Philadelphia: WB Saunders, 1984: 1021–8.

Malt, RA, Abbott, WM, Warshaw AL, Vander Salm TJ, Smead WL. Randomized trial of emergency mesocaval and portacaval shunts for bleeding oesophageal varices. Am J Surg 1978; 135: 584–8.

Mikkelsen WP. End-to-side portacaval anastomosis. In: Malt RA, ed. Surgical Techniques Illustrated. Philadelphia: WB Saunders, 1985: 436–41.

Ottinger LW. The Linton splenorenal shunt in the management of the bleeding complications of portal hypertension. Ann Surg 1982; 196: 664–8.

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

Resnick RH, Iber GL, Ishihara AM, Chalmers TC, Zimmerman H. A controlled study of the therapeutic portacaval shunt. Gastroenterology 1974; 67: 843–57.

Reynolds TB, Donovan AJ, Mikkelsen WP, Redeker AG, Turrill FL, Weiner JM. Results of a 12-year randomized trial of portacaval shunt in patients with alcoholic liver disease and bleeding varices. Gastroenterology 1981; 80: 1005–11.

Sanfey H, Cameron JL. Mesocaval shunts. In: Rutherford RB, ed. Vascular Surgery. 2nd edn. Philadelphia: WB Saunders, 1984: 1029–51.

Sarfeh IJ, Rypins EB, Mason GR. A systematic appraisal of portacaval H-graft diameters: clinical and haemodynamic perspectives. Ann Surg 1986; 204: 356–63.

Sarr MG, Herlong HF, Cameron JL. Long-term patency of the mesocaval C shunt. Am J Surg 1986; 151: 98–103.

Stipa S, Ziparo V. Mesentericocaval shunt (MCS) with autologous jugular vein. World J Surg 1984; 8: 702–5.

Thompson JS, Schafer DF, Haun J, Schafer GJ. Adequate diet prevents hepatic coma in dogs with Eck fistulas. Surg Gynecol Obstet 1986; 162: 126–30.

Turcotte JG, Erlandson EE. Portacaval shunts. In: Rutherford RB, ed. Vascular Surgery. 2nd edn. Philadelphia: WB Saunders, 1984: 1013–20.

Valayer J, Hay J-M, Gauthier F, Broto J. Shunt surgery for treatment of portal hypertension in children. World J Surg 1985; 9: 258– 58.

Voorhees AB Jr, Price JB. An update of portal systemic shunting and its complications. World J Surg 1984; 8: 698–701.

Warren WD, et al. Splenopancreatic disconnection: Improved selectivity of distal splenorenal shunt. Ann Surg 1986; 204: 346–55.