The purpose of this study was to investigate the effects of the design of the screw, the depth of insertion, the vertebral level, and the quality of the host bone on the pull-out resistance of screws used in the lateral masses. The study included twelve fresh cervical spines from human cadavera. Radiographs were made of each specimen to ensure the absence of defects, and then the cancellous-bone density of the vertebral bodies was measured at each level with quantitative computed tomography scanning. Six commercially available screws of various diameters and thread configurations (2.7, 3.2, 3.5, and 4.5-millimeter cortical-bone screws; a 3.5-millimeter cancellous-bone screw; and a 3.5-millimeter self-tapping screw) that are currently used for fixation of the cervical lateral masses were tested for axial load to failure. A twelve-by-twelve Latin square design was used to randomize the screws with regard to level (second through seventh cervical vertebrae), side (right and left), and depth of insertion (unicortical or bicortical purchase). Each screw was then subjected to uniaxial load to failure. The data were analyzed to determine if the diameter of the screw, the thread configuration, the number of cortices engaged, the cervical level, or the bone density was associated with the load to failure.Three major subgroups (greatest, intermediate, and lowest pull-out resistance) were identified. The subgroup with the greatest pull-out resistance included only screws with bicortical purchase; the 3.2, 3.5, and 4.5-millimeter cortical-bone screws and the 3.5-millimeter cancellous-bone screw were in this subgroup. Regardless of the thread configuration, no screw with unicortical purchase was in the group with the greatest pull-out resistance. Two of the three values in the subgroup with the lowest pull-out resistance were for the 3.5-millimeter self-tapping screw (with unicortical or bicortical purchase). The cancellous-bone density of the vertebral body was not associated with pull-out resistance and it did not vary significantly according to the cervical level, with the numbers available. However, the pull-out resistance of the screws varied significantly (p = 0.004) by level: it was the greatest at the fourth cervical level, decreasing cephalad and caudad to that level.CLINICAL RELEVANCE: Posterior plate fixation of the cervical spine with screws inserted into the lateral masses may be appropriate in certain circumstances. Previous studies on cadavera and clinical experience have attested to the potential pitfalls of fixation with screws in the lateral masses. Our data suggest that the surgeon should consider not only the type and size of the screws but also whether unicortical or bicortical purchase should be achieved. Bicortical purchase engenders a greater risk of injury to local anatomical structures, but this may be an acceptable compromise at the cephalad and caudad regions of the cervical spine, where the purchase of screws is relatively weaker.