Biomechanical in vitro tests analysing screw loosening often include high standard deviations caused by high variabilities in bone mineral density and pedicle geometry, whereas standardized mechanical models made of PU foam often do not integrate anatomical or physiological boundary conditions. The purpose of this study was to develop a most realistic mechanical model for the standardized and reproducible testing of pedicle screws regarding the resistance against screw loosening and the holding force as well as to validate this model by in vitro experiments. The novel mechanical testing model represents all anatomical structures of a human vertebra and is consisting of PU foam to simulate cancellous bone, as well as a novel pedicle model made of short carbon fibre filled epoxy. Six monoaxial cannulated pedicle screws (Ø6.5 × 45mm) were tested using the mechanical testing model as well as human vertebra specimens by applying complex physiological cyclic loading (shear, tension, and bending; 5Hz testing frequency; sinusoidal pulsating forces) in a dynamic materials testing machine with stepwise increasing load after each 50.000 cycles (100.0N shear force + 20.0N per step, 51.0N tension force + 10.2N per step, 4.2Nm bending moment + 0.8Nm per step) until screw loosening was detected. The pedicle screw head was fixed on a firmly clamped rod while the load was applied in the vertebral body. For the in vitro experiments, six human lumbar vertebrae (L1-3, BMD 75.4 ± 4.0mg/cc HA, pedicle width 9.8 ± 0.6mm) were tested after implanting pedicle screws under X-ray control. Relative motions of pedicle screw, specimen fixture, and rod fixture were detected using an optical motion tracking system. Translational motions of the mechanical testing model experiments in the point of load introduction (0.9-2.2mm at 240N shear force) were reproducible within the variation range of the in vitro experiments (0.6-3.5mm at 240N shear force). Screw loosening occurred continuously in each case between 140N and 280N, while abrupt failures of the specimen were observed only in vitro. In the mechanical testing model, no translational motion was detected in the screw entry point, while in vitro, translational motions of up to 2.5mm in inferior direction were found, leading to a slight shift of the centre of rotation towards the screw tip. Translational motions of the screw tip of about 5mm in superior direction were observed both in vitro and in the mechanical testing model, while they were continuous in the mechanical testing model and rapidly increasing after screw loosening initiation in vitro. The overall pedicle screw loosening characteristics were qualitatively and quantitatively similar between the mechanical testing model and the human vertebral specimens as long as there was no translation of the screw at the screw entrance point. Therefore, the novel mechanical testing model represents a promising method for the standardized testing of pedicle screws regarding screw loosening for cases where the screw rotates around a point close to the screw entry point.