Many diseases of the liver or associated structures obstruct portal blood flow and result in portal hypertension. Such obstruction is accompanied by increased mesenteric blood flow, the so-called hyperaemia of portal hypertension. These phenomena are generally considered to be the two principal factors in the pathophysiology of portal hypertension.

Obstruction may occur in the hepatic veins, within the liver parenchyma, or in the portal vein itself. The most common cause of portal hypertension in Europe and North America is cirrhosis, which increases intrahepatic vascular resistance by fibrosis, thrombosis, and nodular regeneration. Portal vein obstruction by thrombosis or extrinsic compression is the second most common cause. Hepatic vein occlusion is the third most common, but much less frequent, cause.

A more precise classification of portal hypertension is by site of obstruction at the presinusoidal, sinusoidal, and postsinusoidal levels (Table 1) 372. Intrinsic liver disease is usually absent in patients with presinusoidal obstruction: the natural history and prognosis of portal hypertension is therefore relatively good. Post-sinusoidal obstruction can result in hepatocellular damage as a secondary congestive phenomenon. In the acute phase of the illness this may result in death from liver failure rather than from variceal haemorrhage. In patients with sinusoidal block such as occurs in cirrhosis, the degree of hepatocellular damage is the most important factor in the natural history of the disease. Death may result not only from variceal haemorrhage but also from liver failure, malnutrition, or infection.

Irrespective of cause, portal hypertension results in the development of collateral channels between the portal and systemic venous circulations, of which the most important clinically are those that develop in the oesophagus. These varices are the source of spectacular haemorrhage and thus from a clinical viewpoint, the most important complication of portal hypertension.

The liver has a dual supply from both the hepatic artery and the portal vein. The superior mesenteric and splenic veins drain the splanchnic and splenic beds and join to form the portal vein. Entering into the origin of the portal vein is the coronary or left gastric vein which drains the lesser curve of the stomach and the lower oesophagus. The inferior mesentric vein drains the left hemicolon and most of the rectum, and joins the splenic vein just before its confluence with the superior mesenteric vein (Fig. 1) 1282. After division into left and right branches, the hepatic artery parallels the portal system, and subsequently divides into branches corresponding to the segmental anatomy of the liver.

The portal venous system is entirely devoid of valves. Numerous small tributaries connect the portal and systemic venous systems, and these can evolve into major collateral channels when portal hypertension supervenes. Formation of such collaterals is triggered when portal pressure rises above the normal level of 5 to 10 mmHg. This absence of valves is a feature of practical importance, in that it allows portal venous pressure to be measured during surgical procedures by cannulation of any small mesenteric or omental vein.

The most important of these portal-systemic channels is the left gastric or coronary vein, which connects the oesophagocardiac venous plexus with the splenic or portal vein; the short gastric and left gastroepiploic veins, which connect the oesophageal and gastric plexus with the splenic vein; numerous retroperitoneal portal radicles, which connect to the left renal vein via the left adrenal vein; the umbilical and periumbilical veins connecting to the left portal vein; and the inferior mesenteric vein connecting via the superior haemorrhoidal vein to the middle and inferior haemorrhoidal veins of the systemic circulation. The last are often responsible for the formation of large hypertensive haemorrhoids. Collaterals may also form across the diaphragm, within adhesions from previous abdominal surgery or from inflammatory bowel disease, or at mucocutaneous junctions, such as those that occur at ileostomies or colostomies (Fig. 2) 1283. In addition to these channels, intrahepatic shunts develop through which a significant proportion of portal venous flow can pass. These routes allow up to 80 per cent of the liver's blood supply to bypass the sinusoidal circulation.

The portal vein normally carries 75 per cent of the blood supply of the liver, with an average flow of 1200 ml/min. The hepatic artery supplies the remainder, at an average flow of 400 ml/min; it also provides 30 to 40 per cent of the liver's normal oxygen requirement since portal blood is already partly deoxygenated. When flow to the sinusoids is significantly reduced, a compensatory increase in hepatic arterial inflow takes place to provide the liver's oxygen requirements. However, intrahepatic shunting may cause up to 30 per cent of hepatic arterial flow to bypass the sinusoidal bed. Despite the compensatory increase in arterial flow, therefore, an overall significant reduction in hepatic tissue perfusion is highly likely to occur.

An appreciation of the basic haemodynamic principles governing flow within blood vessels allows a better understanding of the pathophysiology of portal hypertension. As in all blood vessels, pressure within the portal system is determined by the interaction of flow and vascular resistance. As these parameters change, so does portal pressure. This relationship is expressed by Ohm's law, Equation 29

where D&subp; is change in pressure, Q is flow, and R is resistance in a vessel.

where n is the coefficient of viscosity, L the vessel length, and r the vessel radius. Within the liver, viscosity and length of vessel are relatively constant and thus changes in vascular resistance are mainly due to changes in radius. Because the relationship is to the fourth power of the radius, small changes in portal vessel diameter will have profound effects on vascular resistance.

The normal liver has a very low resistance, but this increases dramatically in disease. As well as structural disturbances associated with cirrhosis, inflammatory changes in the hepatic venous tree and deposition of fibrous tissue around the terminal hepatic venules and adjacent sinusoids have been described. In cirrhotic livers, intrahepatic resistance may also be affected by proliferation of myofibroblasts around the sinusoids and terminal hepatic venules, resulting in increased contractility which may raise resistance and contribute to portal hypertension. Sinusoids may be compressed by hepatocyte enlargement in regenerating parts of the liver. This can occur as a result of several toxic, infectious, or metabolic insults to the parenchyma and may explain the portal hypertension seen in non-cirrhotic conditions such as alcoholic hepatitis.

Despite the decompressive effects of collaterals and intrahepatic shunts, with up to 80 per cent of the portal blood flow bypassing the liver, portal pressure remains elevated. Two major hypotheses have been advanced to explain this phenomenon.

The ‘backward’ theory postulates that portal hypertension is due to increased hepatic vascular resistance which develops as a specific response so that in the presence of normal flow, pressure must increase. Hypertrophy of myofibroblasts within the sinusoidal bed is one possible mechanism by which this could be achieved. Ohm's law suggests that portal hypertension would develop only if normal liver blood flow was maintained, but logically, in the presence of extensive low-resistance collaterals, blood flow should be diverted away and thus portal pressure should decrease. In reality, however, the portal and splanchnic circulation is not only markedly increased but hyperdynamic, thus negating the likelihood that this theory can explain portal hypertension.

The ‘forward’ theory was initially postulated by Banti in 1883. He suggested that splenomegaly, portal hypertension, and cirrhosis were the result of increased splenic arterial blood flow. The subsequent confirmation of increased splanchnic blood flow as a major feature of portal hypertension has led to further development of this theory. This now suggests that increased and hyperdynamic flow, together with decreased splanchnic precapillary resistance, maintain portal hypertension even in the face of extensive portal-systemic shunting. Since Banti first postulated his theory, opinions on the relative contribution of increased splanchnic blood flow to the development of portal hypertension have varied. More recently Groszmann and other workers have increasingly implicated both theories as being involved in the pathophysiology of this condition.

It is currently believed that the principal and initial abnormality is increased vascular resistance to portal flow and that portal hypertension is then maintained by increased blood flow into the portal circulation, a phenomenon which has been confirmed conclusively both experimentally and clinically. Blood flow to the stomach, spleen, and intestines is increased by 50 per cent in portal hypertension: this is achieved largely by splanchnic vasodilatation and raised cardiac output.

A hyperdynamic systemic circulation is frequently found in patients with portal hypertension. The intensity of this phenomenon was previously related to the degree of underlying hepatocellular dysfunction. However, this state is also seen in patients with extrahepatic causes of portal hypertension, in whom hepatic function is usually normal. It is, therefore, the extent of portal-systemic shunting which is the major determinant in the pathogenesis of this clinical feature. Various vasoactive substances produced locally in the splanchnic and portal circulations to modulate and increase blood flow are then transported via collaterals into the systemic side, where their vasoactive effects cause a generalized hyperdynamic circulation.

Several substances have been implicated as hormonal factors which act on both the splanchnic and systemic circulations to produce hyperaemia. The most important of these are bile acids, serotonin, glucagon, and prostaglandins, in particular prostacyclin.

Elevated levels of bile acids have been demonstrated in patients with liver disease and implicated as promoters of splanchnic hyperaemia. In the experimental situation however, lowering of serum bile acid levels to normal has no effect on the increased splanchnic blood flow, and their role remains uncertain.

Serotonin has a powerful vasoconstrictor effect on all vascular smooth muscle and is normally released into the portal circulation by the enterochromaffin cells of the gastrointestinal tract. In experimental portal hypertension, selective blockade of serotonin receptors reduces portal pressure, thus supporting the role of serotonergic mechanisms in the pathogenesis of portal hypertension.

Glucagon, a potent splanchnic vasodilator, is elevated in patients with portal hypertension, and probably accounts for 30 per cent of the increased splanchnic blood flow. The effect of somatostatin reducing portal pressure and blood flow may in part result from its inhibitory effect on the release of glucagon, in addition to its direct vasoactive properties.

Prostacyclin is another naturally occurring powerful vasodilator, production of which is elevated in the portal vein endothelium of rats with portal hypertension. Its production is directly related to the portal pressure. Prostacyclin production is also elevated in experimental arterial hypertension: this may represent a compensatory physiological response of endothelial cells to reduce hypertension. Cirrhotic patients have extremely high levels of urinary 6-ketoprostaglandin F&sub1;&subagr;, a stable metabolite of prostacyclin, and pharmacological inhibition of prostacyclin production reduces portal pressure in these patients.

Thus several important vasoactive substances are delivered directly into the systemic circulation via portal-systemic collaterals which carry 80 per cent of the portal blood flow. These substances will remain vasoactive since they have avoided deactivation by the hepatocytes. They provide a likely explanation for the systemic hyperaemia of portal hypertension.

Cirrhosis of the liver accounts for approximately 90 per cent of all cases of portal hypertension presenting in the West. In Eastern and tropical countries non-cirrhotic causes predominate. Distinct geographical distributions are found for non-cirrhotic portal fibrosis, schistosomiasis, and idiopathic portal hypertension.

In India, idiopathic portal fibrosis causes 20 to 30 per cent of all cases of portal hypertension; a further 20 to 30 per cent are due to portal vein thrombosis secondary to dehydration and portal infection in the neonatal period. In Japan approximately 10 per cent of cases of portal hypertension are idiopathic, as opposed to 1 to 6 per cent in the West. This condition remains most prevalent in the tropics.

Schistosomiasis affects more than 200 million people world-wide: Schistosoma mansoni is particularly prevalent in the Middle East, Africa, and South America, while S. japonicum is the causative agent in the Far East. Hepatic schistosomiasis is particularly common in Egypt, where oesophageal variceal bleeding secondary to this disease is the most common cause of upper gastrointestinal haemorrhage.

In children, extrahepatic portal vein obstruction is the main cause of upper gastrointestinal bleeding: this accounts for 40 to 50 per cent of all cases of portal hypertension in those under 17 years old. It is much more common in developing countries.

The natural history and prognosis of portal hypertension depends primarily on the underlying hepatic functional reserve. Conditions causing presinusoidal block but in which hepatocellular function remains good carry the best prognosis: the major lethal complication is haemorrhage from oesophageal varices. The mortality for each bleeding episode is about 5 per cent, and where facilities exist for prompt blood transfusion and injection sclerotherapy the long-term outlook is good. In postsinusoidal block (Budd–Chiari syndrome or veno-occlusive disease) severe secondary hepatocellular damage in the acute phase can result in death from liver failure and massive ascites rather than from haemorrhage.

In patients with cirrhosis the overall mortality rate from oesophageal variceal bleeding is about 40 per cent. If the patient recovers from the acute bleed the risk of recurrent haemorrhage during the same hospital admission is 60 per cent; this increases to more than 80 per cent at 2 years. The long-term survival of cirrhotics following variceal bleeding is poor, ranging from 6 to 35 per cent at 5 years. It is important to emphasize, however, that only about 30 per cent of cirrhotic patients will ever experience such bleeding. The remainder die of liver failure, cachexia, and infection.

The outlook in cirrhosis is generally determined by the severity of associated hepatocellular disease. In 1964 Child introduced a classification of severity of liver disease based on clinical findings and liver function tests. Classification into grades A, B, or C (Table 2) 373 allows long-term survival and operative risk to be predicted for each patient. Groups A (good liver function) and B (moderate impairment of liver function) have a 70 to 80 per cent survival rate at 1 year while only 30 per cent of patients in group C (poor liver function) survive for 1 year. Child originally reported operative mortality rates of 5 per cent in groups A and B, and of 53 per cent in patients of group C undergoing portocaval shunting procedures. Using this classification, similar predictive patterns of mortality have been reported consistently throughout the literature. Patients with alcoholic liver disease or chronic active hepatitis also have particularly poor prognosis; patients with primary biliary cirrhosis generally have a better outlook.

Acute hepatic decompensation may occur during a bleeding episode. This may result in deterioration such that a patient initially in a good prognostic group enters a lower Child's grade. This phenomenon is of crucial importance in timing of intervention and most particularly, in analysis by Child's grading of the efficacy of specific interventional procedures.

Rupture of oesophageal varices (and, less commonly, of gastric varices) is the most dramatic complication of portal hypertension and that which is most likely to require surgical management. Recent elegant anatomical studies have revealed in detail the complex and peculiar portal-systemic connection which occurs in the distal 2 to 5 cm of the oesophagus, precisely the zone where varices develop (Fig. 3) 1284. Four distinct layers of veins have been found in this zone, extending from the luminal surface to the adventitial layer. Intraepithelial veins or vascular epithelial channels drain into a superficial venous plexus, which lies just below the oesophageal epithelium. This plexus is in turn connected to a layer of deep intrinsic veins just outside the muscularis mucosae. The deep intrinsic veins connect with the outermost, external venous plexus coursing within the oesophageal adventitia via the perforating veins, which traverse the muscular layers of the oesophagus. Thus in this distal portion of the oesophagus, venous channels are concentrated largely in the mucosa and run in longitudinal pallisades. These connect with the more deeply located submucous plexus in the cephalad oesophagus and stomach respectively. This arrangement of veins probably underlies one of the intrinsic mechanisms of the physiological oesophageal sphincter, the so-called mucosal rosette, which is of importance in prevention of reflux.

In patients with portal hypertension this venous complex, situated at the watershed between the portal and systemic circulations, dilates dramatically and forms varices, classically running in three to five distinct trunks. The intraepithelial channels probably form the cherry red spots which can be seen endoscopically and are recognized as predictors of impending rupture. They are also found in histological specimens of oesophageal transection rings.

Studies using the oesophageal Doppler ultrasound probe have shown that blood flow and velocity within the palisade zone is complex and considerable. While flow is mainly cephalad, as would be expected, bidirectional flow has also been demonstrated, particularly where perforating veins from the deep adventitial plexus enter the varices. Turbulence of flow has also been demonstrated in these perforating veins. The palisade and perforating venous systems are therefore critical to the development of varices. Less severe variceal bleeds probably result from rupture of the intraepithelial channels, while torrential haemorrhage results from rupture of the deeper, high-flow intrinsic venous channels.

Gastric varices are a much less common source of haemorrhage, probably because of their deeper submucosal situation. These varices are mainly supplied by the short gastric veins and course into the deep intrinsic veins of the distal oesophagus. Congestive gastropathy, which is common in portal hypertension, is associated with mucosal congestion, resulting from increased submucosal arteriovenous communications which develop between the muscularis mucosae and dilated arterioles and venules. Such congested mucosa is particularly susceptible to gastritis and haemorrhage.

Significant rupture occurs in only 30 per cent of patients, and the causes remain poorly understood. Two main theories for variceal rupture are commonly advanced. The erosive theory postulates that mucosal damage due to reflux oesophagitis causes erosion into the varix and thus haemorrhage. Reflux oesophagitis is not common, however, either at endoscopy or at operative inspection of the distal oesophagus. Inflammation is rarely found in specimens of oesophageal rings removed at transection, and controlled trials have found that cimetidine has no advantage over placebo in preventing recurrent variceal bleeding. Erosion of the oesophageal mucosa as a cause of rupture therefore seems unlikely.

The eruptive theory proposes that variceal rupture is related to pressure where the varix is in close proximity to the oesophageal lumen: thus it commonly occurs in the distal 2 to 5 cm of the oesophagus. Although increased portal pressure (>12 mmHg) is required for variceal development, a direct correlation between portal pressure and risk of variceal bleeding has not been found. There is more circumspect evidence to support a relationship between the severity of portal hypertension and risk of variceal haemorrhage. Portal pressure in alcoholic patients following a variceal bleed was higher in those who did not survive. In addition, angiography and blood volume expansion, both of which may increase portal pressure, occasionally precipitate haemorrhage. Variceal haemorrhage has also been reported to occur after deep inspiration, coughing, and following the Valsalva manoeuvre, all of which cause an increased pressure differential between oesophageal varix and lumen. Local factors such as variceal wall thickness and that of the overlying mucosa are obviously also of importance.

Variceal size has been reported to be a risk factor for haemorrhage, the larger varices being more likely to bleed. The lack of relationship between pressure and variceal size reinforces the importance of local factors in varix development, however.

Groszmann has proposed that variceal wall tension may be the unifying predictor of rupture. Tension in a variceal wall can be derived from a modification of Laplace's law Equation 31

where T = tension, TP = transluminal pressure, r = vessel radius, w = wall thickness.

In oesophageal varices, transmural tension is the difference in pressure between the oesophageal lumen and that of the varix. Thus tension is directly related not only to this pressure difference, but also to the varix radius, and is inversely related to variceal wall thickness. These three variables correspond in the clinical situation to variceal size, thickness of epithelium overlying the varix, and degree of portal hypertension, all of which have been described as independent predictors of bleeding. When portal pressure is increased the more superficial protruding varix will have less supporting connective tissue and a larger radius than a deeper varix embedded in supporting tissue which is under identical pressure. Wall tension will therefore be higher in these superficial varices, and rupture will result when this tension is no longer in equilibrium with the outwardly directed expansile force of portal pressure. When the elastic limit of the vessel has been reached, small changes in volume or radius will result in large changes in wall tension and imminent rupture (Fig. 4) 1285.

The parameters considered in the above equation cannot be measured clinically. However endoscopic classification of varices according to size and presence of cherry red spots and red wall markings, both of which are related to thinness of the overlying epithelium, has markedly increased the clinician's ability to predict impending variceal haemorrhage.

Reduction of variceal pressure by shunting, disconnection, or by pharmacological means decreases transluminal pressure, vessel radius, and wall tension. Sclerotherapy, by causing dense perivariceal scarring, will increase the surrounding supporting tissue, strengthening the variceal wall and thus decreasing tension.

Once rupture has occurred the severity of haemorrhage will be affected by haemodynamic factors and by the disordered haemostasis that frequently accompanies liver disease. The relevant haemodynamic factors have been expressed by Groszmann in the following equation. Equation 32

It follows that large holes under greater pressure will bleed more severely than small holes under lower pressure. Blood viscosity is directly related to the haematocrit: reduction of haematocrit after haemorrhage, in anaemia, and after volume replacement with crystalloids, may increase the severity of bleeding.

Patients with liver disease may show multiple disorders of haemostasis, including deficient production by the liver of the coagulation proteins, thrombocytopenia secondary to hypersplenism, impaired platelet function, and increased fibrinolytic activity. The relative importance of all these factors discussed in the clinical situation remains unclear. Haemorrhage generally responds most readily to transfusion of whole, preferably fresh, blood, and clotting factors such as fresh frozen plasma and platelets. Empirically, haemodynamic factors may be most important in the development of rupture, while coagulatory factors may be of greater importance once profuse haemorrhage is established.

Portal hypertension per se has little in the way of clinical features, and usually presents with complications such as variceal haemorrhage, ascites, or hypersplenism. Haemorrhage occurs in only 30 per cent of these patients: the remainder may present with a clinical history and findings indicative of liver disease, particularly cirrhosis. The clinical history is therefore of major importance when portal hypertension is suspected.

Careful questioning of the patient about alcoholism, past jaundice or hepatitis, exposure to hepatotoxins, and a past history of blood transfusion or drug abuse should be routine. A history of neonatal or intra-abdominal sepsis, or of a myeloproliferative disorder is suggestive of an extrahepatic portal vein block. A history of myeloproliferative disorder may indicate the Budd–Chiari syndrome. The use of oral contraceptives or other sex hormone-containing medications is also significant.

The most common clinical presentation of portal hypertension is profuse gastrointestinal haemorrhage: the severity, frequency, and dates of these episodes must be determined. Elucidation of symptoms of hepatic decompensation, in particular ascites and portosystemic encephalopathy as manifested by intellectual dulling, confusion, or coma, is important. Haematemesis is the most common presentation, but episodes of melaena without haematemesis may also occur. A history of dyspepsia or past peptic ulceration suggests bleeding from one of these sources and the results of any previous barium studies or endoscopies may be informative.

Careful physical examination may reveal the stigmata of liver disease, but in the absence of hepatocellular decompensation the single most important finding is an enlarged spleen. If the spleen cannot be felt, or if splenomegaly is not confirmed on imaging, portal hypertension is unlikely to be present. Although enlargement of the spleen is progressive, the degree of hyperplenism is not related to the severity of portal hypertension. Abdominal wall veins from portal-systemic shunts may be visible, although this sign is uncommon. The caput medusae, in which several veins can be seen radiating out across the abdomen from the umbilicus is the most striking sign. In patients with obstruction of the inferior vena cava in the Budd–Chiari syndrome or as a result of severe ascites, abnormal veins course upwards across the abdomen from the iliac fossae to cross the costal margin and drain into superior vena cava territory.

Rarely, a murmur or venous hum may be heard in the region of the xiphoid process, arising from turbulent flow in a large umbilical collateral. A systolic murmur overlying the liver may indicate alcoholic hepatitis or a primary hepatocarcinoma. The liver may be enlarged, particularly in alcoholics, or shrunken: the latter can be confirmed by careful percussion. A shrunken liver is particularly common in portal hypertension and is reported to be associated with higher portal pressures. The consistency of the liver should also be assessed: a firm nodular liver indicates cirrhosis, while a smooth soft liver is suggestive of extrahepatic portal obstruction.

Ascites usually develops in hepatocellular decompensation and is rarely due to portal hypertension alone. Peripheral oedema is also frequently present. Further signs of liver failure include palmar erythema, spider naevi, white nails, jaundice, and portal-systemic encephalopathy. Rectal varices are common in portal hypertension but cannot be detected by digital examination alone. Bleeding from rectal varices is uncommon, rarely severe, and must obviously be distinguished from that associated with simple haemorrhoids.

Haemorrhage from oesophageal varices remains the most common presentation of portal hypertension: if these varices did not form, portal hypertension would be of virtually no clinical significance. Bleeding is typically dramatic and catastrophic, with massive haematemesis and circulatory collapse. Bleeding per rectum may also be profuse and fresh. A slow oozing of blood from varices may occasionally result in true melaena only. Variceal bleeding in patients with cirrhosis may cause marked deterioration of liver cell function. This, together with increased protein absorption from the blood-laden gut, may lead to portal-systemic encephalopathy and coma.

Non-variceal gastrointestinal bleeding is also common in alcoholic patients who may suffer from peptic ulceration, gastric erosions, and the Mallory–Weiss syndrome.

Clinical evaluation of the patient is of utmost importance, particularly when surgical intervention is being contemplated. Operative procedures in patients with compromised liver function carry significantly increased risk. Evaluation of the patient's hepatocellular reserve is therefore of crucial importance when selecting treatment for bleeding varices. Such assessment may be best made in conjunction with a hepatologist or physician with a special interest in liver disease. Recommended laboratory investigations are listed in Table 3 374. Once the underlying liver disease has been diagnosed and liver function assessed, treatment should be selected on the basis of liver reserve, as classified by Child. The next stage in investigation is imaging of the varices and the portal circulation.

Endoscopy allows rapid and safe confirmation of the source of bleeding from the upper gastrointestinal tract. It is of vital importance in the exclusion of other causes of bleeding, such as peptic ulceration, gastritis, duodenitis, or an oesophageal tear. During active haemorrhage the source of bleeding may be obscured, particularly if it is from gastric varices or gastritis. Repeat endoscopy after 3 to 4 h will often find the stomach emptied of clot or blood, allowing visualization of the source. Once the diagnosis of variceal haemorrhage is confirmed, a skilled endoscopist can undertake immediate treatment by injection sclerotherapy.

Endoscopy also allows the size, distribution, and colour of varices to be determined. Varices are normally white and opaque: red coloration and the presence of cherry red spots are accurate predictors of variceal bleeding. Larger varices are also more likely to bleed. Endoscopic assessment of scarring or ulceration following injection sclerotherapy is of major importance, particularly when bleeding continues or recurs and procedures such as oesophageal transection are being considered. Surgical intervention should be avoided in patients with injection site ulceration: this is a frequent cause of haemorrhage which usually responds to medical treatment.

Plain radiographs of the abdomen and chest are of limited value in the management of portal hypertension, but they may yield useful information. In an abdominal radiograph liver and spleen size may be assessed and rarely, gas shadows in the portal circulation may be detected in patients with enterocolitis, intestinal infection, or disseminated intravascular coagulation syndromes (Figs. 5, 6) 1286,1287. Widening of the left paravertebral shadow due to lateral displacement of the pleural reflection between aorta and vertebral column by a dilated hemiazygos vein may also be seen. Rarely, extensive para-oesophageal collaterals may appear as a posterior mediastinal mass (Fig. 7) 1288.

Where endoscopic facilities are not available, barium studies can be useful in demonstrating oesophageal and gastric varices (Fig. 8) 1289. The typical appearances are of a dilated oesophagus containing multiple irregular filling defects, usually in the distal third but occasionally running throughout the entire oesophagus. Barium studies are also useful in diagnosing other causes of bleeding, such as peptic ulceration. For this reason a full upper gastrointestinal barium study is preferable to a barium swallow alone.

Once the diagnosis is made and operative intervention is being considered, the anatomical details of the portal circulation are needed. Classically this is obtained by angiography, but less invasive techniques such as duplex ultrasonography, contrast-enhanced CT scanning, and magnetic resonance imaging are being increasingly employed.

Angiography plays a major role in the investigation of portal hypertension. Hepatic blood flow, free and wedged hepatic pressures, and inferior vena cava pressures can all be measured during this procedure. The main indication, however, is visualization of the portal system, particularly for identification of major portosystemic collaterals and provision of a map to allow planning of surgical intervention. It is also particularly useful in the diagnosis and assessment of non-cirrhotic portal hypertension, such as that due to portal vein thrombosis, and when there is reason to suspect occlusion of a portal-systemic shunt. The portal circulation can be demonstrated by several angiographic techniques.

Coeliac and superior mesenteric angiography is the safest, most commonly used, but indirect means of visualizing the portal venous circulation (Fig. 9) 1290. Using the Seldinger technique under local anaesthesia, the coeliac trunk and superior mesenteric artery are selectively and individually cannulated using a small bore arterial catheter (French 5) and 50 to 60 ml of contrast medium. The venous circulation is visualized awaiting the venous return of contrast after its arterial injection: this technique requires careful timing by the radiologist, and has been greatly simplified by the advent of digital subtraction angiography. Detailed imaging of the entire portal circulation with identification of major collaterals and shunts can be obtained routinely. Additional information such as the presence of hepatofugal flow, may influence the choice of surgical procedure.

Injection of contrast into the coeliac trunk allow the hepatic and splenic arterial circulation to be visualized. The intrahepatic circulation is frequently abnormal in cirrhosis, showing tortuosity or corkscrewing of the hepatic arterial branches. Haemangiomata, tumour circulation, hepatic artery-to-portal vein shunting, aneurysms, or anatomical variations of the major arteries can be readily and accurately demonstrated. Injection of contrast into the splenic artery allows evaluation of the splenic vein and coronary vein, and demonstration of oesophagogastric varices. Superior mesenteric angiography is used to delineate the superior mesenteric and portal veins and their collaterals.

Indirect visualization of the portal tree suffers from a degree of loss of detail, but this is rarely severe enough to warrant direct venography. Digital subtraction angiography is being used increasingly often with good results, but resolution is often poorer than that of conventional angiography.

The best imaging of the splenic and portal veins, and collaterals is obtained using this technique, but at the cost of increased risk. Splenic venography is of particular value in the investigation of extrahepatic portal vein obstruction or when cavernous transformation of a thrombosed portal vein is suspected (Figs. 10 and 11) 1291,1292. After infiltration of local anaesthetic into tissues overlying the spleen, a fine bore catheter is passed into the spleen and splenic pulp pressure is measured. Venography is then performed by injection of contrast into the splenic pulp. Once imaging is completed, the catheter tract is sealed by injection of pledgets of gelatine foam. Providing any coagulopathy is corrected beforehand, this procedure is rarely complicated by significant haemorrhage, but extravasation of contrast around the spleen may cause pain, both local and referred to the left shoulder. The risk of complication remains higher than with angiography and splenic venography should therefore be reserved for patients in whom imaging by angiography is inadequate.

This technique is performed by percutaneous puncture of the liver, introduction of a fine bore catheter into an intrahepatic radicle, and injection of contrast. Excellent imaging of the splenic and portal vein is obtained (Fig. 12) 1293 and varices can be embolized via the coronary or larger collateral veins. Initial success in stopping variceal haemorrhage is high (80–90 per cent), but there is a 25 to 30 per cent rebleeding rate within a few days. This, together with a high complication rate and the technically demanding nature of the procedure, has led to its virtual abandonment.

In the Budd–Chiari syndrome the inferior vena cava is often severely stenosed or entirely occluded. The degree to which this vessel is compressed by the hypertrophied caudate lobe can be assesed both by visualization and measurement of the pressure gradient between the suprahepatic and infrahepatic inferior vena cavae (Fig. 13) 1294: a high gradient mitigates against treatment by portacaval shunt. Hepatic venography is of value in diagnosing hepatic vein thrombosis in patients with veno-occlusive disease associated with myeloproliferative disorders. Its most common use, however, is in measurement of free and wedged hepatic pressures in the assessment of cirrhosis and portal hypertension. These measurements are of particular value as a research tool in evaluation of the haemodynamic effects of drugs on the portal circulation.

Real-time ultrasound scanning can provide much detailed and accurate information by totally non-invasive means. It is of particular value for scanning the hepatic and splenic parenchyma for haemangioma, tumour, and cirrhosis. Thrombosis or occlusion of the portal, superior mesenteric, and splenic vein can be accurately detected. Large collaterals can be visualized, helping to confirm the diagnosis of portal hypertension. Although the results are often compromised by duodenal or intestinal gas, its absolute safety is making ultrasound scanning the first line of investigation in portal hypertension. Recent advances in duplex ultrasound technology, in particular colour-flow coding, allow faster and more accurate assessment of direction and flow in the portal circulation (Fig. 14) 1295,1296. The accuracy is such that flow through portosystemic shunts can be measured, and duplex scanning is superseding angiography as the investigation of choice when graft occlusion is suspected. Duplex ultrasonography has become an important diagnostic and research tool with a promising future role in the investigation of portal hypertension.

The major advantage of CT scanning lies in the detection of focal liver disease and hepatosplenomegaly (Fig. 15) 1297. Used with contrast enhancement and dynamic scanning, portal vein patency and para-oesophageal and retroperitoneal portosystemic collaterals can be demonstrated readily. This is rarely the diagnostic procedure of choice, however, since ultrasound, and increasingly magnetic resonance imaging, delineate these lesions more easily and accurately. CT assessment of liver parenchymal volume is used in some centres as part of the assessment and selection of candidates for liver transplantation.

This imaging modality allows excellent parenchymal visualization, imaging in any plane, and better blood vessel imaging than CT scanning. MRI angiography is currently limited by poor resolution and long imaging times. However development of this modality is progressing rapidly and it is very likely to supersede CT scanning and even conventional angiography in the future (Figs. 16 and 17) 1298,1299.

Dick R. Angiography. In: Millward-Sadler GH, Wright R, Arthur MJP, eds. Wright's Liver and Biliary Disease. London: WB Saunders, 1992: 582–94.

Karin SK. Progress report. Non-cirrhotic portal fibrosis. Gut 1989; 30: 406–15.

Kitano S, Terblanche J, Kahn D, Bornman PC. Venous anatomy of the lower oesophagus in portal hypertension: practical implications. Br J Surg 1986; 73: 525–31.

Mahl TC, Groszmann RJ. Pathophysiology of portal hypertension and variceal bleeding. Surg Clin N Am 1990; 70: 251–66.

Pilar Pizcueta M, Garcia-Pagan JC, Fernandez M, Casamitjana R, Bosch J, Rodès J. Glucagon hinders the effects of somatostatin on portal hypertension. A study in rats with portal vein ligation. Gastroenterology 1991; 101: 1710–15.

Polio J, Groszmann RJ. Haemodynamic factors involved in the development and rupture of oesophageal varices: a pathophysiologic approach to treatment. Semin Liver Dis 1986; 6: 318–31.

Sherlock S. Disease of the Liver and Biliary System, 8th edn. Oxford: Blackwell, 1991.

Sitzman JV, Bulkley GB, Mitchell MC, Campbell K. Role of prostacyclin in the splanchnic hyperaemia contributing to portal hypertension. Ann Surg 1989; 209: 322–7.

Sitzman JV, Li S-S, Adkinson F. Evidence for role of prostacyclin as a systemic hormone in portal hypertension. Surgery 1991; 109: 149–53.

Spence RAJ, Terblanche J. Venous anatomy of the lower oesophagus: a new perspective on varices. Br J Surg 1987; 74: 659–60.