A multi-centre random trial of 57 cases of alcoholic cirrhosis with refractory ascites was carried out; 29 patients received a LeVeen shunt and 28 were treated by conventional medical therapy. The effectiveness of the LeVeen shunt in reducing ascites was good in the first month, but was not different from conventional medical therapy by the end of one year. Complications were significantly more frequent in the surgical group. Of the 29 patients fitted with a LeVeen shunt, 25 developed one or more complications. Of the 28 patients in the medical control group, only 8 developed complications. The mortality rate of the two groups also differed significantly. Twelve patients in the surgical group (41%) died in the course of the first month against only 5 (18%) in the medical control group. By the end of one year, the mortality rate of the two groups was almost identical: 23 (79%) and 21 (75%) respectively. These observations confirm the poor prognosis for refractory ascites in cases of alcoholic cirrhosis and the inadvisability of attempting to treat it by implanting a LeVeen shunt.