A consecutive series of 14 patients with hepatic malignant disease treated by palliative hepatic arterial embolisation was reviewed. Twelve patients had hepatic pain from their tumour and two were suffering from the carcinoid syndrome. Six patients died within four weeks of the procedure (group 1) and eight survived for at least 10 weeks (group 2). Factors were sought that might permit prediction of a high risk of early death (group 1). The pre-embolisation angiograms reviewed by a "blind" observer showed no differences in vascularity or tumour size between the groups and no difference in the extent of arterial occlusion after embolisation. The portal vein was patent in all patients. No significant difference was seen between the groups in the pre-embolisation biochemical values, with the exception of lower serum albumin concentrations and higher alkaline phosphatase activities in group 1. All those who died early had serum alkaline phosphatase activities of 45 KAU or above, while six of the eight who survived longer had activities below this value (p less than 0.02). These findings suggest that serum alkaline phosphatase activity of 45 KAU or more (normal range 3-13) might alone be a useful predictor of early death. Stepwise discriminant analysis using a weighted combination of serum alkaline phosphatase activity and albumin concentration predicted the outcome in all but one of the patients studied (p less than 0.002).