The pre-clinical validation of new joint prostheses is an important step in the development process of a new implant design. The high success rate of modern prostheses makes it very difficult to ensure that new designs will exhibit significant improvements combined with a risk of failure as low as that of the designs currently in use. This situation requires a systematic re-thinking of the pre-clinical validation protocols. We advocate that a careful combination of risk analysis, experimental and numerical modelling, and the use of statistical finite element analysis to generalize the results to a simulated population, can provide the level of confidence required by class 3 devices as the joint prostheses recently became. The proposed approach is illustrated using the validation of a cemented epiphyseal stem as an example.