In the experimental and contact models, the absolute values of displacement under loading were different, but aspects of the displacement under loading conditions caused by differences in the loading points were similar and showed similar tendencies. The correlation coefficient of the two was 0.925, representing a significant and strong correlation (p < 0.01). This shows that the behavioral tendencies of the contact model are reflective of those in the experimental model and that the results obtained had high validity. The CV of the displacement under loading conditions was calculated as about 10% in some areas of the FEA models, but the mean was about 5% for all three models, representing a relatively low value. This suggests that all models had highly reproducible displacements under loading and that the results obtained had high validity. Although there are limitations to the reproducible range, it appears possible to infer phenomena to an extent if the properties are understood and the limitations are known. Analysis by three-dimensional finite element models has been shown to be an effective means for studying the behavioral tendencies of implants under loading conditions.

Hotta et al. [30] measured the amount of displacement under the loading of implants placed in human mandibles. When a load is applied at a location that deviates from the long-axial direction of implant more buccally and lingually, the forces from rotation and inclination are propagated to the implant as an eccentric axial load. The implant displacement of eccentric axial loading has been reported to be larger than that during long-axial direction loading. Awazawa et al. [31] measured displacements under the loading of implants placed in canine mandibles and reported no substantial difference in displacement based on whether the loading direction was towards the buccal or lingual side during buccolingual loading. These reports are consistent with the results of x-axis displacement in this study and support the clinical validity of the constructed FEA models.