A major problem threatening the long-term integrity of total hip replacement is the loss of proximal bone often found around noncemented stems in the long term. It is generally accepted that 'stress shielding' is the cause for this problem: after implantation of the prosthesis the surrounding bone is partially 'shielding' from load carrying and starts to resorb. One of the proposed answers to this problem is the application of press-fitted stems. These smooth-surfaced implants are thought to provoke higher proximal bone loading, and, hence, less stress shielding than bonded implants, because they are wedged into the femur every time when loaded. However, in a two-year experiment in dogs, similar amounts of resorption of the proximal cortex were found around press-fitted and bonded implants. The question arises how similar resorption patterns can develop under completely different stress conditions, and whether this phenomenon can be explained by adaptive bone remodeling theories based on Wolff's law. In the present study an answer was sought for this question. An advanced iterative computer simulation model was used to analyze the remodeling process in the animal experiment. Three-dimensional finite element models were constructed from the animal experimental configuration, in which smooth, press-fitted stems were applied unilaterally in the canine. The FE model was integrated with iterative remodeling procedures, validated in earlier studies. In the model an appropriate non-linear representation of the loose bone-implant interface was realized, also capable of simulating the proximal interface gap that was found around the uncoated implants. The simulation models predicted similar amounts of proximal bone loss and distal bone densification as found in the animal model. Hence, the cortical bone loss could indeed be predicted by the strain-adaptive bone remodeling theory. By unraveling the simulation process, the question stated above could be answered. Densification of the distal bone bed during the initial remodeling process was found to cause reduced axial stem displacement (elastic subsidence), decreasing the wedging effect of the stem and, hence, decreasing the loading of the proximal bone, resulting in proximal bone loss. Hence, whereas in the case of bonded stems the proximal resorption process develops monotonously to a new equilibrium, the process around smooth, press-fitted stems develops nonmonotonously. This is due primarily to the unbonded interface conditions and the development of a proximal fibrous membrane. The remodeling process then gradually causes the stem to be jammed in the distal diaphyses (proximal 'stress bypass').