The relationships between portal pressure, liver function and clinical variables on one hand and development of variceal hemorrhage and death on the other were investigated in 58 men with newly diagnosed alcoholic cirrhosis. Portal pressure was determined during hepatic vein catheterization as wedged minus free hepatic vein pressure, and median pressure was 14 mm Hg (range = 3 to 26 mm Hg). Fourteen of 31 patients (45%) had esophageal varices at upper gastrointestinal endoscopy (the size being considered large in nine patients). During follow-up (median = 31 months; range = 2 to 51 months), 12 patients (21%) developed variceal hemorrhage. Applying Cox's regression analysis, information about previous variceal bleeding (p = 0.0046), large varices at endoscopy (p = 0.012), hepatic vein pressure gradient (p = 0.0056) and indocyanine green clearance (p = 0.038) all contained significant prognostic information regarding development of variceal hemorrhage, even when easily obtained variables with known prognostic information were included [modified Child-Turcotte's criteria and incapacitation index (a weighted sum of days without normal health)]. During follow-up, 17 patients (29%) died. Applying Cox's regression analysis, large varices at endoscopy (p = 0.012) and hepatic vein pressure gradient (p = 0.019) contained significant prognostic information regarding death, in addition to the information contained in the modified Child-Turcotte's criteria and incapacitation index. In conclusion, prediction of prognosis in alcoholic cirrhotic men may be significantly improved by information about size of esophageal varices and level of portal pressure.