Radiation doses to the tumour and non-tumorous liver compartments from yttrium-90 microspheres in the treatment of hepatic cancer, as estimated by a partition model, have been verified by correlation with the actual doses measured with a beta probe at open surgery. The validity of the doses to the lungs, the tumour and non-tumorous liver compartment as estimated by the partition model was further evaluated in clinical settings. On the basis of the observation that one of three patients who received more than 30 Gy from a single treatment and one of two patients who received more than 50 Gy from multiple treatments developed radiation pneumonitis, it was deduced that an estimated lung dose < 30 Gy from a single treatment and a cumulative lung dose < 50 Gy from multiple treatments were probably the tolerance limits of the lungs. Three of five patients who received lung doses > 30 Gy as estimated by the partition model and were predicted to develop radiation pneumonitis, did so despite the use of partial hepatic embolization to reduce the degree of lung shunting. Furthermore, a higher radiological response rate and prolonged survival were found in the group of patients who received higher tumour doses, as estimated by the partition model, than in the group with lower estimated tumour doses. Thus the radiation doses estimated by the partition model can be used to predict (a) complication rate, (b) response rate and (c) duration of survival in the same manner as the actual radiation doses measured with a beta probe at open surgery. The partition model has made selective internal radiation therapy using 90Y microspheres safe and repeatable without laparotomy.