Sacral screw fixation is frequently used for fusion of the lower lumbar spine, but sacral screws appear to offer less secure fixation than lumbar pedicle screws, and failure due to loosening under fatigue loading is common. The aim of this study was to examine in vitro the stability of medial and lateral bicortical and unicortical sacral screw fixation under a physiologically relevant fatigue-loading pattern. Bone mineral density, screw insertion torque, and screw-fixation stiffness were measured prior to cyclic loading between 40 and 400 N compression at 2 Hz for 20,000 cycles. The screw-fixation stiffness was measured every 500 cycles, and the axial pullout strength of the screws was recorded following loading. All of the lateral insertions loosened under the applied loading, but some of the medial insertions remained stable. Medial insertions proved stiffer and stronger than lateral insertions, and bicortical fixations were stronger than unicortical fixations. Bone mineral density and insertion torque were correlated with screw stiffness and pullout strength, although better correlation was found for insertion torque than bone mineral density. Bone mineral density is a good preoperative indicator of sacral screw-fixation strength, and insertion torque is a good intraoperative indicator. An insertion torque greater than 1.5 Nm is suggested as an indicative value for a stable medial unicortical insertion, whereas an insertion torque greater than 2 Nm suggests a stable medial bicortical insertion. It appears that, apart from the choice of technique (screw orientation and depth), minimizing the load on the screws during the initial part of the fusion process is also critical to maintain stability of the fused section and to obtain a solid fusion mass.