UK guidelines on the management of variceal haemorrhage in cirrhotic patients

Management of Variceal Haemorrhage

1.0 Introduction

These guidelines on the management of variceal haemorrhage were commissioned by the British Society of Gastroenterology under the auspices of the Liver Section. They were written in June 1998 and have been corrected and agreed upon by the members of the Liver Section. The nature of variceal haemorrhage in cirrhotic patients with its complex range of complications makes rigid guidelines inappropriate.

Over the past few years there have been numerous advances in the management of variceal haemorrhage in patients with cirrhosis. These have included better endoscopic techniques with the widespread availability of video endoscopy, establishment of variceal band ligation, availability of newer drugs such as somatostatin and vasopressin analogues, better surgical techniques, and finally the availability of transjugular intrahepatic portosystemic stent shunt (TIPSS).

These guidelines deal specifically with the management of varices in patients with cirrhosis and are not designed to address: (1) the management of the underlying liver disease; (2) the management of variceal haemorrhage in children; or (3) variceal haemorrhage from other aetiological conditions.

2.0 Validity and grading of recommendations

These guidelines have been produced to conform to the system proposed by the North of England evidence based guidelines development project.1 2


These are graded as follows:

Grade Ia: evidence obtained from meta-analysis of randomised trials.

Grade Ib: evidence obtained from at least one randomised trial.

Grade IIa: evidence obtained from at least one well designed controlled study without randomisation.

Grade IIb: evidence obtained from at least one other type of well designed quasi experimental study.

Grade III: evidence obtained from well designed non-experimental descriptive studies such as comparative studies, correlation studies, and case studies.

Grade IV: evidence obtained from expert committee reports, or opinions or clinical experiences of respected authorities.

The evidence category is indicated after the citations in the reference section.


The strength of each recommendation is dependent upon the category of the evidence supporting it.


Variceal haemorrhage is defined as bleeding from an oesophageal or gastric varix at the time of endoscopy or the presence of large oesophageal varices with blood in the stomach and no other recognisable cause of bleeding. An episode of bleeding is clinically significant when there is a transfusion requirement of 2 units of blood or more within 24 hours of the time zero, together with a systolic blood pressure of less than 100 mm Hg or a postural change of greater than 20 mm Hg and/or pulse rate greater than 100 beat/min at time zero (time zero is the time of admission to the first hospital the patient is taken to).


The acute bleeding episode is represented by an interval of 48 hours from time zero with no evidence of clinically significant bleeding between 24 and 48 hours. Evidence of any bleeding after 48 hours is the first rebleeding episode.


Variceal rebleeding is defined as the occurrence of new haematemesis or malena after a period of 24 hours or more from the 24 hour point of stable vital signs and haematocrit/haemoglobin following an episode of acute bleeding. All bleeding episodes regardless of severity should be counted in evaluating rebleeding.


The definition of failure to control active bleeding is divided into two time frames:

(i) Failure to control bleeding acute bleeding within six hours:

Transfusion requirement of 4 units or more and inability to achieve an increase in systolic blood pressure by 20 mm Hg or to 70 mm Hg or more, and/or inability to achieve a pulse rate reduction to less than 100 beat/min or a reduction of 20 beat/min from baseline pulse rate.

(ii) Failure to control bleeding after six hours: any of the following factors

Occurrence of haematemesis from the six hour point.

Reduction in blood pressure of more than 20 mm Hg from the six hour point and/or increase in pulse rate of more than 20 beat/min from the six hour point on two consecutive readings an hour apart, transfusion of 2 units of blood or more (over and above the previous transfusions) required to increase the haematocrit to above 27%, or haemoglobin to above 9 g/l.


Death within six weeks of the initial episode of bleeding.

4.0 Natural history of varices in cirrhosis


The rise in portal pressure is associated with the development of collateral circulation which allows the portal blood to be diverted into the systemic circulation. These spontaneous shunts occur: (a) at the cardia through the intrinsic and extrinsic gastro-oesophageal veins; (b) in the anal canal where the superior haemorrhoidal vein belonging to the portal system anastomoses with the middle and inferior haemorrhoidal veins which belong to the caval system; (c) in the falciform ligament of the liver through the para-umbilical veins which are the remains of the umbilical circulation of the fetus; (d) in the abdominal wall and the retroperitoneal tissues, from the liver to the diaphragm, veins in the lienorenal ligament, in the omentum and lumbar veins; and (e) blood diversion from the diaphragm, gastric, pancreatic, splenic, and adrenal which may drain into the left renal vein.

Numerous lines of evidence suggest that varices develop and enlarge with time. Christensen and colleagues4 followed a cohort of 532 patients with cirrhosis and showed that the cumulative incidence of patients with varices increased from 12% to 90% over 12 years. In a study involving 80 patients followed for 16 months, Cales and Pascal5 showed that 20% of patients who did not have varices developed new varices and 42% of patients with small varices showed definite enlargement. Czaja and colleagues6 also showed that the prevalence of varices increased from 8% to 13% over five years in a cohort of patients with chronic active hepatitis even though they were treated with prednisolone.

The two factors that appear to determine the development of varices are continued hepatic injury and the degree of portosystemic shunting. Evidence for the former is derived from studies in which varices were shown to regress with time. Baker and colleagues7 followed a cohort of 112 patients with oesophageal varices and showed that varices had disappeared in nine patients, regressed in seven, and remained unchanged in six. They concluded that the disappearance and regression of varices may be related to abstinence from alcohol. This observation was confirmed in a study by Dagradi and colleagues8 who followed a cohort of patients with alcoholic cirrhosis over three years and showed a reduction in variceal size in 12 of the 15 patients with alcoholic cirrhosis who stopped drinking and an enlargement in variceal size in 17 patients who continued to drink. On the other hand, Cales and Pascal5 showed that regression of varices occurred in 16% of patients with alcoholic cirrhosis who continued to imbibe alcohol. This may be related to the development of large portosystemic collaterals which decompress the portal system.


The factors that predispose to and precipitate variceal haemorrhage are still not clear. The suggestion that oesophagitis may precipitate variceal haemorrhage has been discarded.9Presently, the most important factors that have been held responsible include: (i) pressure within the varix, (ii) variceal size, (iii) tension on the variceal wall, and (iv) severity of the liver disease.

4.2.1 Portal pressure

In most cases, portal pressure reflects intravariceal pressure10 and a hepatic venous pressure gradient greater than 12 mm Hg is necessary for the development of and bleeding from oesophageal varices but there is no linear relationship between the severity of portal hypertension and the risk of variceal haemorrhage.11 12 However, the hepatic venous pressure gradient (HVPG) tends to be higher in bleeders as well as in patients with larger varices. In a prospective study comparing propranolol with placebo for the prevention of first variceal haemorrhage, Groszmann and colleagues13 showed that bleeding from varices did not occur if the portal pressure gradient could be reduced to less than 12 mm Hg. This pressure has since been accepted as the aim of pharmacological therapy of portal hypertension.

4.2.2 Variceal size

This is best assessed endoscopically. Variable results in the literature are because of the lack of a definition regarding the distinction between large and small varices. Numerous studies11 14have shown that the risk of variceal haemorrhage increases with the size of varices.15

4.2.3 Variceal wall and tension

Polio and Groszmann16 using an in vitro model showed that rupture of varices was related to the tension on the variceal wall. The tension depends on the radius of the varix. In this model, increasing the size of the varix and decreasing the thickness of the variceal wall caused variceal rupture.

Endoscopic features such as “red spots” and “wale” markings were first described by Dagradi.8They have been described as being important in the prediction of variceal haemorrhage. These features represent changes in variceal wall structure and tension associated with the development of microtelangiectasias. In a retrospective study by the Japanese Research Society for Portal Hypertension, Beppu and colleagues17 showed that 80% of patients who had blue varices or cherry red spots bled from varices, suggesting that this was an important predictor of variceal haemorrhage in cirrhosis.

4.2.4 Severity of liver disease and bleeding indices

Two independent groups prospectively assessed factors predicting first variceal haemorrhage in cirrhosis prospectively. The North Italian Endoscopic Club (NIEC)18 reported their findings in 1988, followed in 1990 by data from the Japanese.19Both of these studies showed that the risk of bleeding was based on three factors: severity of liver disease as measured by Child class, variceal size, and red wale markings. The NIEC study showed a wide range for the risk of bleeding of 6–76% depending on the presence or absence of the different factors. This index was prospectively validated in a study by Prada and colleagues.20 Using the same variables the NIEC index was simplified by De Franchis and colleagues21 and shown to correlate with the original index. Further studies showed that the HVPG and intravariceal pressure were also independent predictors of first variceal haemorrhage when analysed in conjunction with the NIEC index.22 23

In summary, the two most important factors that determine the risk of variceal haemorrhage are the severity of liver disease and the size of varices. Measurement of HVPG is a useful guide for selection of patients for treatment and their response to therapy.


Data describing the overall risk of bleeding from varices must be viewed with caution and have some pitfalls in interpretation. The natural history of patients who have varices that are diagnosed as part of their work up is different from patients who have complications of liver disease. Patients in trials may represent a different population to patients who have had documented varices and not bled during follow up. Most studies do not comment on either the severity of liver disease or whether patients with alcoholic cirrhosis are continuing to drink. Both these factors have a significant effect on the risk of variceal haemorrhage.

Most studies report bleeding from varices in about 20–50% of patients with cirrhosis during the period of follow up. Baker and colleagues7 reported variceal bleeding in 33 of 115 patients that they followed for a mean of 3.3 years, with a mortality of 48% from first variceal haemorrhage. These data were confirmed by Christensen and colleagues.4 About 70% of episodes of bleeding occur within two years of diagnosis.

Analysis of the non-active treatment arms in the primary prophylaxis trials comparing propranolol with placebo show results similar to those of the primary prophylaxis shunt trials, with most of the episodes of bleeding occurring within the first two years of follow up. In these studies the rate of first variceal haemorrhage ranged from 22% to 61%.24-28 This large difference in the rate of first bleed relates almost certainly to the number of patients with severe liver disease included in the study (Pascal, Child C—46%, bleeding—61%; IMPP, Child C—6%, bleeding—32%; Conn, Child C—6%, bleeding—22%). Mortality varied from 24% to 49% over two years (Pascal, mortality—49%; IMPP, mortality—24%; Conn, mortality—24%).


The average mortality of the first episode of variceal bleeding in most studies is 50%. As discussed, this mortality from variceal haemorrhage is related closely to the severity of liver disease.29-33 Over a mean follow up of one year, the average mortality from subsequent variceal haemorrhage is 5% in Child class A patients, 25% in Child class B patients, and 50% in Child class C patients. Although serum creatinine has been shown in some studies to predict overall survival,34 35 Child class is superior to any other predictive factor in determining mortality within six weeks or 30 days of the initial haemorrhage (see box 1).

Vinel and colleagues14 showed that HVPG was predictive of survival when this was measured at two weeks after the acute bleed. However, it is unclear if this was independent of the severity of liver disease. Whether active bleeding at the time of endoscopy predicts mortality is not clear. Although Cardin and colleagues36found that this was an important factor, Balanzo and colleagues37 could not confirm this finding. Active bleeding at the time of endoscopy does however predict early rebleeding.38 Risk of death decreases quickly after admission such that the risk of death becomes virtually constant about six weeks after bleeding.39 40


Since 30–50% of patients with portal hypertension will bleed from varices and about 50% will die from the effects of the first bleed, it seems rational to develop prophylactic regimens to prevent the development of, and bleeding from, these varices. However, most of the published trials do not have sufficient power to identify favourable treatment effects. Based on the expected bleeding and death rates in the control group, the minimum number of patients needed to detect a 50% reduction in bleeding would be 270 patients, and 850 patients in each arm to detect the same reduction in mortality. A proposed algorithm for surveillance and prophylaxis of varices is shown in fig 1.



Although numerous methods have been described for grading varices, the simplest method is to divide them into three grades:

Grade 1: varices that collapse to inflation of the oesophagus with air.

Grade 2: varices between grades 1 and 3.

Grade 3: varices which are large enough to occlude the lumen. (Strength of recommendation grade CII.)

4.5.1 Surgery

Portacaval shunts.

There are four trials in the literature which have randomised a total of 302 patients42-44 either to prophylactic shunt surgery or to non-active treatment. A meta-analysis of these studies showed a significant benefit in the reduction of variceal bleeding (odds ratio (OR) 0.31, 95% confidence interval (CI) 0.17–0.56) but also a significantly greater risk of hepatic encephalopathy (OR 2, 95% CI 1.2–3.1) and of mortality (OR 1.6, 95% CI 1.02–2.57) in patients treated with shunt surgery.45

Devascularisation procedures.

Inokuchi and colleagues19 showed that there was a significant reduction in variceal bleeding and in mortality in patients treated with a variety of devascularisation procedures. There are, however, numerous problems with the interpretation of this study because of the use of different procedures in each of the 22 centres. These results require confirmation.

4.5.2 Pharmacological therapy