Three-dimensional (3D) finite element (FE) models are commonly used to analyze the mechanical behavior of the bone under different conditions (i.e., before and after arthroplasty). They can provide detailed information but they are numerically expensive and this limits their use in cases where large or numerous simulations are required. On the other hand, 2D models show less computational cost, but the precision of results depends on the approach used for the simplification. Two main questions arise: Are the 3D results adequately represented by a 2D section of the model? Which approach should be used to build a 2D model that provides reliable results compared to the 3D model? In this paper, we first evaluate if the stem symmetry plane used for generating the 2D models of bone-implant systems adequately represents the results of the full 3D model for stair climbing activity. Then, we explore three different approaches that have been used in the past for creating 2D models: (1) without side-plate (WOSP), (2) with variable thickness side-plate and constant cortical thickness (SPCT), and (3) with variable thickness side-plate and variable cortical thickness (SPVT). From the different approaches investigated, a 2D model including a side-plate best represents the results obtained with the full 3D model with much less computational cost. The side-plate needs to have variable thickness, while the cortical bone thickness can be kept constant.