Normally intelligent and pragmatic surgeons spend too much time trying to decide which portasystemic shunt is best. They have done so for 40 years—a strange state of affairs given that only three impregnable concepts about surgical relief of portal hypertension have ever been enunciated. One of these concepts is that decompressing a high-pressure splanchnic venous circulation into a low-pressure systemic venous circulation reduces or eliminates the risk of bleeding from oesophagogastric varices, though at the risk of precipitating portasystemic encephalopathy. All other arguments are biased and in the absence of a unifying concept of the aetiology and treatment of bleeding varices, will continue to be.

While there can be no doubt that the coronary–caval (Fig. 1) 1314 (left gastric) shunt is best, because it drains the specific venous compartment at risk immediately into the systemic circulation without diverting the flow of portal blood from the liver, this shunt is seldom feasible in Caucasian patients, as opposed to the frequency with which it can be usefully performed in Japanese patients. Although, over a period of 20 years we have found only five patients suitable for this operation, they have had no encephalopathy. The anatomy of the patient or the skill of the operator, or both, must be determinants of the result. In any event, the coronary–caval shunt is becoming increasingly feasible because of surgeons' ability to handle difficult venovenous anastomoses in patients with portal hypertension.

A key variable in almost every study of elective shunts is the extent to which the procedure being examined prevents or eliminates portasystemic encephalopathy. Divergent contentions about the incidence of portasystemic encephalopathy can be ignored, inasmuch as estimates of the incidence and prevalence of portasystemic encephalopathy are inaccurate. Despite hopes that computed electroencephalography would provide an objective assessment of the degree of encephalopathy, there is no unequivocal method of assessing the likelihood and the degree of portasystemic encephalopathy. Thus, stratification of patients and of results by the apparent degree of portasystemic encephalopathy is fraught with error and bias.

One should not be influenced by arguments masquerading as facts. The classical statement about the value of Warren's distal splenorenal shunt says that the incidence of encephalopathy after this procedure is about 5 per cent. Yet, in Grace's randomized prospective trial of the Warren shunt, as opposed to ‘total’ shunts, the chance of portasystemic encephalopathy was 51 per cent, granted that the skills of several operators in a multicentre trial might not equal the special talents of the originator of the procedure.

The surgeon should perform the kind of shunt he or she knows best how to do, without worrying about putative fine points of physiology.

One reason for inconstant results of any operation is that the splanchnic system itself is not freely interconnected. The flexible veins of the splanchnic circulation do not follow Poiseuille's law concerning the flow of liquids in rigid tubes. Each compartment is self-contained: the splanchnic, the splenic, the gastroduodenal, and the oesophageal (Fig. 2) 1315. Some compartments are freely interconnected; some are not. For example, when the splenic vein is clotted as a result of pancreatitis, variceal bleeding is cured by splenectomy, not by a shunt.

Conn once said, ‘We must either learn to select better the patients we shunt or to shunt better the patients we select.’

Although the enthusiasts state that a central splenorenal shunt diverts less portal vein blood from the splanchnic circulation than does an end-to-side portacaval shunt, Fig. 3 1316 shows what splenorenal shunting actually accomplishes. In fact, appreciable portal vein blood is diverted from the liver into the systemic circulation, and the functional anatomy is that of a side-to-side portacaval shunt. Nonetheless, the classical proximal (central) splenorenal shunt is as good as almost any other shunt, with the exception of a coronary–caval shunt. It is valuable for treating patients with severe hypersplenism and thrombocytopenia because the spleen is removed at the onset of the operation, facilitating thrombocytosis and coagulation. Extrahepatic portal hypertension in children, produced by neonatal thrombosis of the portal vein and causing oesophageal variceal haemorrhage at any age from birth to adolescence (Fig. 4) 1317, can be relieved by a central splenorenal venous anastomosis, sometimes as small as 4 mm in diameter, with only an 8 per cent chance of failure. The only issue is that children virtually never die from variceal bleeding, because their compensatory systems are so resilient. Today, they are mostly at risk from the consequences of infection by HIV and other contaminants in transfused blood.

Certain cases of schistosomiasis are also reasons to consider this procedure; however, the distal splenorenal shunt is probably superior for treatment of portal hypertension from bilharzial disease (schistosomiasis) because of its apparent lower incidence of portasystemic encephalopathy.

Despite the merits of a distal splenorenal shunt, it is not uniformly suitable for the emergency control of variceal bleeding, because it may decompress the splanchnic circulation too slowly to control bleeding. The proximal splenorenal shunt can often be used for emergency decompression, although it is less reliable than a portacaval shunt. However, neither the proximal nor the distal splenorenal shunt interferes with a subsequent orthotopic liver transplant, but a portacaval shunt does. The incidence of recurrent bleeding after a successful distal splenorenal shunt is about 29 per cent, as compared with 14 per cent after a portacaval shunt. Like a proximal splenorenal shunt, the distal splenorenal shunt is, or becomes, essentially a side-to-side portacaval shunt as collateral veins envelope the pancreas (Fig. 5) 1318.

Over the years argument has raged about the merits of the proximal splenorenal shunt as opposed to those of the portacaval shunt. Although at one time the results of the proximal splenorenal shunt in terms of the incidence of portasystemic encephalopathy were clearly better than were those of the portacaval shunt, the reason for the seeming advantage was that the patients who had had a proximal splenorenal shunt were those who had survived a rigorous selection process. That is, a transthoracic transoesophageal ligation of bleeding varices was performed as a preliminary to a proximal splenorenal shunt. Obviously, anyone who had survived the first operation was preselected as an excellent candidate for any kind of surgery.

A postoperative mortality rate of about 20 per cent is common to both a proximal splenorenal shunt and a portacaval shunt. The 5-year survival rate of 42 &plusmin; 7.4 per cent (SE) for the proximal splenorenal shunt is not statistically different from the 29 &plusmin; 7.5 per cent survival rate after a portacaval shunt. Unlike the other portasystemic venous shunts, an end-to-side portacaval shunt, constructed with a patent inferior vena cava and a patent portal vein, almost never fails to cure bleeding oesophageal varices and to lower portal pressure, unless a neophyte surgeon puckers or occludes the anastomosis.

Table 1 385 shows a simple schema for predicting the likelihood of survival from data relating to the last blood samples drawn before either emergency or elective surgery. The bilirubin level is a parsimonious predictor of liver function and of the amount of blood products received. The presence of ascites reflects the serum albumin level (hence, the synthetic function of the liver) and the degree of portal hypertension. An emergency portasystemic shunt is an independent, ominous predictor. Obviously, if a patient's blood will not clot before surgery, it will be unable to clot afterwards. Overzealous administration of an electrolyte solution by the anaesthetist is a common cause of death from dilutional coagulopathy. In Fig. 2 1309 of Section 24.5 the operative mortality after shunting based on the Massachusetts General Hospital score is shown during two eras of management.

A transabdominal exposure is the easiest for the patient to tolerate (Fig 6) 1319. The need to use a transthoracic approach seldom arises, unless one can divine that the spleen is stuck tight to the diaphragm. Patients undergoing a splenorenal shunt for portal hypertension due to cystic fibrosis should always have an intra-abdominal operation, to avoid the added pulmonary hazards of a thoracoabdominal procedure.

The splenic vein is isolated from the splenic artery and surrounding structures (Fig. 7) 1320. It is hydrostatically dilated to increase its lumen (Fig. 8) 1321. The most difficult part of the operation is securing the small veins that drain from the pancreas to the splenic vein. These veins are fragile beyond belief, and they emit a torrent of blood out of proportion to their size.

The most common technical faults are:

1. Not freeing the splenic vein well enough from the pancreas;

2. Freeing the vein too well, so that the vein buckles when the organs shift;

3. Sewing the splenorenal union poorly. Very often, the anastomosis is done poorly as a result of lack of room to move, ordinarily as a result of having too many clamps in the field.

The only clamps required are three gentle bulldog clamps, which should be placed to occlude completely the medial and lateral aspects of the left renal vein and distal end of the splenic vein. Within broad limits, one does not need to be concerned about the effects of absent renal venous outflow from the left kidney, because the collateral veins (such as the adrenal, gonadal, and retroperitoneal veins) in a patient with portal hypertension divert enough blood to prevent venous infarction of the left kidney. Nevertheless, to avoid a source of potential anxiety, one should avoid taking an excretory urogram of the patient for a month after surgery. A recent study of the effects of ligating the left renal vein during aortic aneurysmectomy suggests that the consequences may be more important than have been supposed.

The freed splenic vein should be brought to the renal vein in as near a streamline as is possible. If the renal vein approaches the splenic vein from cephalad, the anastomosis ought to be constructed on the cephalad surface of the renal vein (Fig. 9) 1322. A small lenticular segment of the renal vein is cut from its surface. The anastomosis is carried out by stitching the veins together with fine polypropylene suture while the maximal circumference of the vein is maintained by the weight of fine haemostats at the end of stitches of diametrically opposite sutures. The stitches in the heel of the splenic vein, closest to the aorta, are continuous and encompass about 90° of the graft (Fig. 9) 1322. The rest are interrupted to permit the onlaid vein to billow. The veins are flushed free of clots (Fig. 10) 1323.

To guarantee patency of the anastomosis, both the renal vein and the splenic vein must be full of blood when the anastomosis is complete, and both must be capable of being emptied quickly by compressing them. Otherwise, splanchnic pressure should be measured, either by needle-puncture manometry within the veins or by Doppler scanning to ascertain flow. (Portacaval shunts do not require pressure measurements because the volume of blood is so great that one can easily feel and see the currents.) Persistent oozing of blood or lymph, or both, normally stops as soon as a patent shunt is opened.

Alvarez F, et al. Portal obstruction in children: II. Results of surgical portosystemic shunts. J Pediatr 1983; 103: 703–7.

Bismuth H, Franco D, Alagille D. Portal diversion for portal hypertension in children: The first ninety patients. Ann Surg 1980; 192: 18–24.

Calligaro KD, Savarese RP, McCombs P R, DeLauentis D A. Division of the left renal vein during aortic surgery. Am J Surg 1990; 160: 192–6.

Collini FJ, Brener B. Portal hypertension. Surg Gynecol Obset 1990; 170: 177–92.

Conn HO. To shunt or not to shunt. Gastroenterology 1974; 67: 1065–71.

Ezzat FA, et al. Selective shunt versus nonshunt surgery for management of both schistosomal and nonschistosomal variceal bleeders. Ann Surg 1990; 21: 97–108.

Fischer JE, et al. Comparison of distal and proximal splenorenal shunts: a randomized prospective trial. Ann Surg 1981; 194: 531–44.

Grace ND. A hepatologist's view of variceal bleeding. Am J Surg 1990; 160: 26–31.

Grace ND, Conn HO. Comparing nonselective and selective shunts (letter). Hepatology 1990; 12: 377–8.

Grace ND, et al. Distal splenorenal vs. portal-systemic shunts after hemorrhage from varices: a randomized controlled trial. Hepatology 1988; 8: 1475–81.

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

Langer B, Taylor BR, Greig PD. Selective or total shunts for variceal bleeding. Am J Surg 1990; 160: 75–9.

Linton RR. Atlas of vascular surgery. Philadelphia: WB Saunders, 1973: 158–205.

Malt RA. Portasystemic venous shunts. N Engl J Med 1976; 296: 24–9 & 80–6.

Malt RB, Malt RA. Tests and management affecting survival after portacaval and splenorenal shunts. Surg Gynecol Obstet 1979; 149: 220–4.

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

Rosemurgy AS, McAllister EW, Kearney RE. Prospective study of a prosthetic H-graft portacaval shunt. Am J Surg 1991; 161: 159–64.

Rössle M, Gerok W. Comparing nonselective and selective shunts. (letter). Hepatology 1990; 2: 377.

Salam AA, Ezzat FA, Abu-Elmagd KM. Selective shunt in schistosomiasis in Egypt. Am J Surg 1990; 160: 90–2.

Sarfeh IJ, et al. Serial measurement of portal hemodynamics after partial portal decompression. Surgery 1986; 100: 52–7.

Smith GW, Cameron JL, Malt RA, Turcotte JG. Total portosystemic shunts. In: Rutherford RB, ed. Vascular surgery. Philadelphia: W. B. Saunders, 1989: 1155–72.

Waddell WG, et al. Functional relations of the proximal components of the portal system: a preliminary report. J Surg Res 1972; 12: 281–9.

Whipple AO. The problem of portal hypertension in relation to the hepatosplenopathies. Ann Surg 1945; 122: 449–75.