Three-dimensional (3D) imaging with high-resolution peripheral quantitative computed tomography (HR-pQCT) and micro-finite element (FE) analysis provides important insight into bone health. Longitudinal analyses of bone morphology maximize precision by using 2D slicematching registration (SM) or 3D rigid-body registration (3DR) to account for repositioning error between scans, however, the compatibility of these techniques with FE for longitudinal bone strength estimates is limited. This work developed and validated a FE approach for longitudinal HR-pQCT studies using 3DR to maximize reproducibility by fully accounting for misalignment between images. Using a standard imaging protocol, ex vivo (N=10) and in vivo (N=40) distal radius HR-pQCT images were acquired to estimate the efficacy of 3DR to reduce longitudinal variability due to repositioning error and assess the sensitivity of this method to detect true changes in bone strength. In our proposed approach, the full common bone volume defined by 3DR for serial scans was used for FE. Standard FE parameters were estimated by no registration (NR), SM, and 3DR. Ex vivo reproducibility was estimated by the least significant change (LSC) in each parameter with a ground truth of zero change in longitudinal estimates. In vivo reproducibility was estimated by the standard deviation of the rate of change (σ) with an ideal value that was minimized to define true changes in longitudinal estimates. Group-wise comparisons of ex vivo and in vivo reproducibility found that FE reproducibility was improved by both SM (CVRMS<0.80%) and 3DR (CVRMS<0.62%) compared to NR (CVRMS~2%), and 3DR was advantageous as repositioning error increased. Although 3D registration did not negate motion artifacts, it played an important role in detecting and reducing variability in FE estimates for longitudinal study designs. Therefore, 3D registration is ideally suited for estimating in vivo effects of interventions in longitudinal studies of bone strength.