Between August 1979 and September 1982, acute variceal haemorrhage has been managed in the University Department of Surgery, Glasgow Royal Infirmary by a policy of oesophageal tamponade and injection sclerotherapy. Haemorrhage was controlled in 90% of admissions with an admission mortality of 28%. Recurrent haemorrhage occurred in half the patients surviving their first admission to hospital despite entering a programme of elective sclerotherapy. The results of this management policy are reviewed and the means of selecting patients for more aggressive therapy discussed. The deficiencies of a modified Child's classification in selection of patients are highlighted and overcome by the development of a prognostic index obtained by regression analysis on data collected on patients managed over this 3 year period. The admission prognostic index clearly defines 'high' and 'low' risk groups and 'predicts' outcome following admission in 90% of patients. The use of this index is validated in a further group of patients managed by a similar policy. Further regression analysis is used to obtain a prognostic index for alcohol cirrhotic patients alone and to determine the factors associated with one year survival. These indices are used to audit the management policy. Prothrombin, creatinine and encephalopathy are shown to have a clear association with outcome when measured at the time of variceal haemorrhage whereas other factors such as albumin and haemoglobin emerge as having prognostic value when measured one month following the acute episode. The possible applications of these prognostic indices are investigated in a prospective two centre study assessing the efficacy of propranolol in preventing recurrent variceal haemorrhage. It is shown that they can be used to exclude patients from entry into a study assessing the longterm benefit of propranolol when the prospects of short-term survival are limited. Their value in auditing management and their possible use in withdrawing treatment are shown. The prognostic indices are used to compare results of treatment at the two hospitals and are shown to be of value in analysing the results of the trial. These prognostic indices provide an objective means of evaluating patient management and may allow selection of patients for consideration of other treatment options.