Normal liver complications have not been observed in Y-90 microsphere therapy of hepatic tumors [selective internal radiation (SIR)], despite clinical studies reporting estimated absorbed doses to normal liver between 100 and 150 Gy. The purpose of the study was to see whether predictions of normal tissue complication probability (NTCP) models for liver based on clinical data from external beam therapy are consistent with clinical results of SIR. Liver NTCP was calculated using a parallel architecture model and normal liver dose-volume histograms that have been proposed for SIR. A parallel model including internal functional subunit structure is also proposed. Dose rate effects are incorporated. A criterion for comparing model calculations with clinical data is presented. For the parallel architecture model, the predicted NTCP is sensitive to the dose distribution in normal liver and to the model parameters, particularly the repair time. With reasonable assumptions about the microsphere distribution, the parallel model with parameters deduced from external beam therapy outcome analysis is consistent with the observed lack of liver complications. Inclusion of FSU structure widens the range of assumptions under which consistency is found. The parallel model can be consistent with the clinically observed lack of liver complications in SIR. More information about the activity distribution and the radiobiology of normal liver under conditions typical of microsphere therapy should be sought.