Non-rigid organ misregistration is an important problem for patients undergoing sequential quantitative SPECT for 3D dosimetry for targeted radionuclide therapy (TRT) treatment planning. This study aims to evaluate effects of these misregistrations on the accuracy of 3D dosimetry. We used 3 anatomical variations and 3 respective In-Ill Zevalin distributions of the digital 4D Extended Cardiac Torso (XCAT) phantom to model the deformation in different organs such as liver, kidneys, spleen and stomach. We simulated SPECT scans acquired at 5 time points, i.e., 1, 12, 24, 72 and 144 hrs post-injection of In-111 Zevalin. Organs with uniform activity concentrations were randomly translated and rotated within 5 pixels/degrees, while the change of the total volume of each organ was within 5% except for the stomach. The 24-hr scan served as a reference. An analytical projector modeling attenuation, scatter and the geometric collimator-detector-response of a medium energy collimator was used to generate noisy projections representing a realistic count level for 128 views over 360 degrees. Reconstructed images were obtained using OS-EM with attenuation, scatter and geometric collimator-detector-response compensation. Voxel-by-voxel integration over different time points followed by convolution with a Y-90 dose kernel was used to generate 3D dose distribution images. For each phantom, we compared the organ dose and its dose volume histogram (DVH) for (i) no organ deformation and (ii) organs with deformation. The mean difference of organ doses between two sets of images were 3.88%, -6.73%, -7.32% and -14.42% for lung, liver, kidneys and spleen respectively. However, even for the organs with dose errors <5%, the associated normalized absolute errors in DVH were >10%. We conclude that organ misregistration and deformation are important factors in limiting accuracy of 3D dosimetric quantities and whole body non-rigid registration of sequential quantitative SPECT is essential for accurate TRT treatment planning.