The present study is a numerical comparison using finite-element analysis (FEA) of two different concepts of spinal fixation devices when implanted. These implants are 1) the Easy®, “rigid” Screw/Rod (o 6mm) system; 2) the Twinflex®, “dynamic” system (o 2 X 2.5 mm ELF). A parameterised 3D FEA model of an L3-sacrum segment, developed by Lavaste, Skalli & Robin, was used. Geometric and mechanical models of each implant were then constructed, before being inserted in the spinal segment model. Then, for model validation, these two L3-S2 instrumented segmental models were submitted to similar boundary conditions as used in a previous in vitro comparison of the same implants. Flexion loaddisplacement curves were then controlled using experimental results. Loads acting on screws and longitudinal elements were calculated and analysed for a better understanding of the intrinsic differences between both constructs. Load-displacement responses of both constructs were quite similar (L3 sagittal rotation at 10 N.m = ~1.5°), while loads in the implant were not. For example, the axial push-in forces at the S1 screws were equal to 30 N for the Twinflex®, and 150 N for the Easy® Screw/Rod system. The pull-out forces at the S2 screws were respectively 100N and 200 N for the Twinflex and Screw/Rod Concept. At other levels, axial forces were all lower than 60 N, the Twinflex®, values being higher than the Easy® ones. Bending moments along screws were respectively 0.7 N.m and 1.4 N.m at the L3 level for the Twinflex® and the Easy® systems. At lower levels, values were all below 0.6 N.m, again with a reversed proportion. Bending moments calculated along longitudinal elements were always lower than 0.3 Nm for the Twinflex®, and up to 2 N.m for the Easy® system. Axial forces in the Twinflex® longitudinal elements were about 160 N, and about 100 N in the Easy® rods. Although the numerical approach mainly provides tendencies, it clearly seems that reducing flexural stiffness of lumbar fixation induces more homogeneous load transmission along the construct, and greatly reduces axial push-in/pull-out forces at the S1/S2 levels, and all this without reducing the rigidity of the whole construct. Conversely, it has been shown that “rigid” longitudinal elements may concentrate stresses at the construct extremities, relieving loads at intermediate levels at the same time, which may be the sign of a stress-shielding-like phenomenon. These differences arise from a fundamental difference between both kinds of longitudinal elements in the way they transmit loads. The o 6 mm rods mainly oppose a bending reaction torque to the applied flexion moment, whereas the Twinflex® construct mainly balances the applied flexion torque by an anterior compression of the anterior column, combined with posterior traction on its longitudinal elements (ELF).