Shoulder musculoskeletal modelling is a computational technique that can revolutionise orthopaedics, by allowing precise estimation of the muscle and joint forces implied while moving. Nonetheless, musculoskeletal predictions are highly sensitive to modelling inputs and parameters, as well as their inherent errors. Consequently, confidence must first be obtained before affirming clinical relevance. This thesis aimed to study how the computations are influenced by errors in the modelled shoulder anatomy and in the experimental motion data. Based on the acquired confidence, the final aim of the thesis was to investigate how joint stability is regained after a clinical Latarjet surgery.