The value of early sclerotherapy for variceal haemorrhage remains unsettled, possibly because the treatment may be beneficial to some patients and harmful to others. On the basis of a randomized clinical trial of sclerotherapy in 187 patients presenting with their first variceal haemorrhage, we examined the relationship between clinical, endoscopic and biochemical characteristics at admission and the treatment effect on mortality. As previously published, sclerotherapy had no overall effect on the very high mortality during the first 6 weeks (47%), but thereafter the mortality and risk of rebleeding were reduced. The analysis showed that in the 48% of the patients with disturbed consciousness and/or elevated plasma creatinine, sclerotherapy considerably increased short-term mortality, and this was not compensated for by increased long-term survival. Among patients without these characteristics, sclerotherapy reduced mortality in the 25% with ascites, but did not affect short-term mortality in the 27% without. Sclerotherapy significantly improved the long-term survival of these patients. The results suggest that sclerotherapy should not be used in patients with disturbed cerebral or renal function, whereas it may be beneficial in patients without these characteristics.