Failure of an orthopaedic fixation due to stress shielding and consequent screw loosening is a major concern among surgeons: the loosened screws could not only interfere with the healing process but also endanger adjacent anatomical structures. In this study, the effect of the screw's engineering design (dimensions, profile shape and material properties) on the load sharing with adjacent bone and consequent bone resorption was tested, using a set of two-dimensional computational (finite element) models. An algorithm simulating local bone adaptation to strain energy density (SED) mechanical stimuli was developed and used to evaluate the biomechanical performances of different commercial screws. Two new designs, a 'graded-stiffness' composite screw, with a reduced-stiffness titanium core and outer polymeric threads, and an active-compression hollow screw that generates compressive stresses on the surrounding bone, were also evaluated. A dimensionless set of stress transfer parameters (STPs) were utilised for ranking the performances of the different screws according to the expected screw-bone load sharing and its evolution with adaptation of the surrounding tissue. The results indicated that commercial wide (6 mm thread diameter) trapezoidal and rectangular screw profiles have superior biomechanical compatibility with bone (i.e. predicted to be stable after 2 years). The graded-stiffness and active-compression screws provided the best biomechanical performances: bone loading around them was predicted to decrease by no more than 15% after 3 years, compared with a decrease of 55-70% in bone loading around commercially available screws. Computer simulations of bone adaptation around orthopaedic screws are demonstrated to be effective means for objective and quantitative evaluation of the biomechanical aspects of implant-tissue compatibility.