PurposeTo model dose-response relationships for in vivo experiments with radiolabelled peptides enabling maximum therapeutic efficacy while limiting toxicity to kidney and bone marrow. MethodsA multiregional murine kidney phantom, with a kinetic model for cortex and outer medulla distribution, were used to predict renal toxicity.Maximum tolerated activities to avoid nephrotoxicity (at 40 Gy Biological Effective Dose BED) and hematologic toxicity (at 2 Gy) were compared.The therapeutic efficacy of 90Y, 161Tb, 177Lu and 213Bi was assessed at their respective maximum tolerated activities based on cellular-level dosimetry accounting for activity and tumor heterogeneity. These results were compared with average tumor-dosimetry-based predictions. ResultsThe kidney was found to be the dose-limiting organ for all radionuclides, limiting the administered activity to 44 MBq 177Lu, 34 MBq 161Tb, 19 MBq 90Y and 13 MBq 213Bi , respectively.The average S-values for the initial heterogeneous activity distribution in the tumor volume are not significantly different from the homogeneous ones. The in vivo tumor cell survivals predicted by assuming uniform dose rate-distributions are not significantly different from those for heterogeneous dose rate-based predictions.The lowest in vivo survival was found for 213Bi (2%) followed by 161Tb (30%), 177Lu (37%) and 90Y (60%). The minimal effective dose rate for cell kill is 13–14 mGy/h for β-emitters and 2.2 mGy/h for the α-particle emitter 213Bi, below these values proliferation takes over. ConclusionsRadionuclides emitting α-particles have the highest potential for improving therapeutic efficacy in tumors and metastases with uniform receptor expression, after careful evaluation of their burden to the healthy organs.