The results of a 3-year, placebo-controlled trial of prednisolone treatment in primary biliary cirrhosis (PBC) are presented. The active (n = 19) and placebo (n = 17) arms were initially well matched for age, menopausal status and disease severity. At 3 years hepatic symptoms were relatively improved in the prednisolone group. Hepatic mortality was 3/19 (prednisolone), 5/17 (placebo) (p = n.s.). For all liver blood tests the trend favoured prednisolone treatment, though the differences were only significant for alkaline phosphatase and protein. All immunoglobulins fell significantly. Quantitative ELISA determination of antimitochondrial antibody showed a significant fall in the prednisolone group compared with placebo (p less than 0.001 at 1 year, p less than 0.05 at 3 years). Deterioration in histology (appearance of cirrhosis) was more common in the placebo group. Overall hepatic function (hepatic mortality, doubling in bilirubin, 6 milligrams fall in albumin, de novo appearance of cirrhosis or symptoms of portal hypertension) was significantly worse in the placebo group (p less than 0.01). After 3 years no significant differences could be detected in bone mineral content (single photon absorptiometry of radius and femur) between the two groups or in comparison with other PBC patients. Thus, after 3 years, prednisolone treatment was associated with a better overall hepatic outcome and little evidence of increased bone loss.