A computational model of the human torso has been developed to study the stability of implanted leads that are part of a sacral nerve stimulation system. The model was built using presegmented anatomies that were themselves built from imaging of human patients. The sacral leads are represented using beam elements, and their interaction with the tissue is defined using a function that relates frictional force to the amount of slip between the lead and tissue. Displacements to the skin in the sacral region are applied to simulate activities of daily living, and the resulting displacement of the tip of the lead is indicative of its tendency to dislodge in real patients. Validation of the model was performed using experimental results collected in human cadavers. In these experiments, analogous displacements of the skin were applied after implantation of the leads per normal implant procedures. The displacement of the distal tip of the lead was measured using computed tomography (CT) imaging, allowing direct comparison to the predictions of the model. Recognizing that many model inputs were informed by sparse literature values, a novel application of uncertainty quantification methodology was developed wherein all model inputs were treated as uncertain intervals. This allowed an optimization approach to be used for estimating the uncertain interval for the model outputs. The computational model and cadaver results were used to study the performance of a new sacral lead design, relative to a predicate product. The results showed that the reduction in lead axial stiffness in the new design leads to less lead tip displacement, such that the lead is more likely to remain near the therapeutic target in patients.