Background Glenohumeral osteoarthritis is common degenerative disease within the elderly population, which causes pain and reduced mobility. In advanced cases, total shoulder arthroplasty (TSA) is usually performed. Although TSA is an established procedure, prosthesis failure rate is still quite significant. A new prosthesis model is being investigated at the Laboratory of Biomechanical Orthopedics with the intent to reduce this failure rate. For such investigation, a finite element patient-specific model is being used. Due to their current research applications and future clinical applications, accuracy of patient-specific models needs to be ensured. This project studies to sensitivity of the FE patient specific model to uncertainties related to the model construction process Methods The present work studies the effect of mesh characteristics, the CT systematic error, material law and homogeneity in the FE model outputs. To reduce uncertainties due to mesh characteristics, a mesh convergence study is performed. The effect of CT systematic error, material law and homogeneity is studied through a factorial analysis. Results Variations in the systematic error, material law and homogeneity choice caused a difference in strain outputs in the bone to differ up to a value of two. On the other hand, stress and strain values in the cement underwent no significant variations. Homogeneous models followed the same patterns. However, the output values yielded by homogeneous models were smaller than those yielded by heterogeneous models. Conclusion Variations in the material law have much larger influence in the results than variations in the homogeneity or the CT systematic error. The cement layer attaching the glenoid implant to the scapula is not affected significantly by the material law or homogeneity of the bone. Homogeneous bone models tend to underestimate peak strain values, and thus, they should be avoided when studying critical performance conditions.