To examine the effectiveness of liver transplantation (LTx) for the treatment of primary biliary cirrhosis (PBC) the actual survival of 30 PBC patients who received liver grafts was compared with predictions of what survival would have been without transplantation. Three models, based on Cox' regression analysis, were used. Two models were derived from survival of PBC patients in drug trials and the third from cirrhotic patients who did not receive transplants. Observed and expected survival were compared for a follow-up time of 7 years. After 1 year the difference in favour of LTx was small, but after 5 years survival with LTx exceeded all predicted survival probabilities without LTx. After 3 years every year of follow-up added about 0.3 years to expected survival gain per transplanted patient, resulting in 1.5 to 2.3 life-years gained at 7 years' follow-up, depending on the model used. The benefit was greatest for patients in Child-Pugh classes B and C. The consistency between the three models in their predictions supports the validity of the use of predictive models in the indirect assessment of LTx.