BackgroundRed bone marrow (RBM) toxicity is dose-limiting in (pretargeted) radioimmunotherapy (RIT). Previous blood-based and two-dimensional (2D) image-based methods have failed to show a clear dose-response relationship. We developed a three-dimensional (3D) image-based RBM dosimetry approach using the Monte Carlo-based 3D radiobiological dosimetry (3D-RD) software and determined its additional value for predicting RBM toxicity.MethodsRBM doses were calculated for 13 colorectal cancer patients after pretargeted RIT with the two-step administration of an anti-CEA × anti-HSG bispecific monoclonal antibody and a 177Lu-labeled di-HSG-peptide. 3D-RD RBM dosimetry was based on the lumbar vertebrae, delineated on single photon emission computed tomography (SPECT) scans acquired directly, 3, 24, and 72 h after 177Lu administration. RBM doses were correlated to hematologic effects, according to NCI-CTC v3 and compared with conventional 2D cranium-based and blood-based dosimetry results. Tumor doses were calculated with 3D-RD, which has not been possible with 2D dosimetry. Tumor-to-RBM dose ratios were calculated and compared for 177Lu-based pretargeted RIT and simulated pretargeted RIT with 90Y.Results3D-RD RBM doses of all seven patients who developed thrombocytopenia were higher (range 0.43 to 0.97 Gy) than that of the six patients without thrombocytopenia (range 0.12 to 0.39 Gy), except in one patient (0.47 Gy) without thrombocytopenia but with grade 2 leucopenia. Blood and 2D image-based RBM doses for patients with grade 1 to 2 thrombocytopenia were in the same range as in patients without thrombocytopenia (0.14 to 0.29 and 0.11 to 0.26 Gy, respectively). Blood-based RBM doses for two grade 3 to 4 patients were higher (0.66 and 0.51 Gy, respectively) than the others, and the cranium-based dose of only the grade 4 patient was higher (0.34 Gy). Tumor-to-RBM dose ratios would increase by 25% on average when treating with 90Y instead of 177Lu.Conclusions3D dosimetry identifies patients at risk of developing any grade of RBM toxicity more accurately than blood- or 2D image-based methods. It has the added value to enable calculation of tumor-to-RBM dose ratios.