This paper proposes a locally modified phantom model to numerically assess the worst-case configuration of orthopedic implants under magnetic resonance imaging (MRI). The proposed model is developed based on the standard American Society for Testing and Materials (ASTM) phantom and bone models with cancellous or cortical materials. Three orthopedic implant families, metallic rods, a nail and screw system, and a plate and screw system, are studied. The worst-case configurations of orthopedic implants are identified inside the proposed model and ASTM phantom. These worst-case heating configurations are then implanted in a human body model to evaluate the RF-induced heating in terms of peak SAR1g. For the orthopedic implants fully inside the bone, like the rod and the nail and screw systems, the peak SAR1g values of worst-case configurations obtained from the proposed phantom model are higher than those obtained inside the ASTM phantom. For the orthopedic implants that are mainly outside the bone, such as the plate and screw system, similar worst-case configurations lead to a negligible variation of peak SAR1g inside the human body model.Clinical Relevance- The new phantom model leads to more accurate predictions of the worst-case configuration of orthopedic implants for MR conditional labeling.