Introduction Much progress has been made in implementing selective internal radiation therapy (SIRT) as a viable treatment option for hepatic malignancies. However, there is still much need for improved options for calculating the amount of activity to be administered. To make advances towards this goal, this study examines the relationship between predicted biological outcomes of liver tumors via tumor control probabilities (TCP) and parenchyma via normal tissue complication probabilities (NTCP) given variations in absorbed dose prescription methodologies. Methods Thirty-nine glass microsphere treatments in 35 patients with hepatocellular carcinoma or metastatic liver disease were analyzed using 99mTc-MAA SPECT/CT and 90Y PET/CT scans. Predicted biological outcomes corresponding to the single compartment (standard) model and multi-compartment (partition) dosimetry model were compared using our previously derived TCP dose-response curves over a range of 80–150 Gy prescribed absorbed dose to the perfused volume, recommended in the package insert for glass microspheres. Retrospective planning dosimetry was performed on the MAA SPECT/CT; changes from the planned infused activity due to selection of absorbed dose level and dosimetry model (standard or partition) were used to scale absorbed doses reported from 90Y PET/CT including liver parenchyma and lesions (N = 120) > 2 ml. A parameterized charting system was developed across all potential prescription options to enable a clear relationship between standard prescription vs. the partition model-based prescription. Using a previously proposed NTCP model, the change in prescribed dose from a standard model prescription of 120 Gy to the perfused volume to a 15% NTCP prescription to the normal liver was explored. Results Average TCP predictions for the partition model compared with the standard model varied from a 13% decrease to a 32% increase when the prescribed dose was varied across the range of 80–150 Gy. In the parametrized chart comparing absorbed dose prescription ranges across the standard model and partition models, a line of equivalent absorbed dose to a tumor was identified. TCP predictions on a per lesion basis varied between a 26% decrease and a 81% increase for the most commonly chosen prescription options when comparing the partition model with the standard model. NTCP model was only applicable to a subset of patients because of the small volume fraction of the liver that was targeted in most cases. Conclusion Our retrospective analysis of patient imaging data shows that the choice of prescribed dose and which model to prescribe potentially contribute to a wide variation in average tumor efficacy. Biological response data should be included as one factor when looking to improve patient care in the clinic. The use of parameterized charting, such as presented here, will help direct physicians when transitioning to newer prescription methods.