Spinal pedicle fixation: Reliability and validity of roentgenogram-based assessment and surgical factors on successful screw placement

The increased popularity of pedicle fixation prompted research to address two issues: the reliability and validity of roentgenograms as a technique for evaluating the success of pedicle fixation, and the effects of surgical factors on successful fixation. Thus, does approach the point and angle of screw insertion, surgeon experience, practice, level of the spine involved, and screw Size effect success of pedicle fixation? Eight fresh thoracolumbar spines were harvested and cleaned of all soft tissues. Two surgeons, one more experienced in pedicle fixation than the other, used two pedicle fixation approaches (Weinstein and Roy-Camllle) on both the left and right sides at levels T11-S1 of each specimen. All screws were placed under anteroposterior (AP) and lateral c-arm control. For specimens 1 to 3, 5.5 mm screws were used at T11-L1, and 7.0 mm screws at L2-S1. Unacceptable failure rates at L2 and L3 for the first three specimens resulted in a change of instrumentation for the remaining specimens, with 5.5 mm screws used at T11-L3 and 7.0 mm screws at L4-S1. When surgeons completed the fixations for a specimen, AP and lateral roentgenograms were taken and both surgeons independently evaluated the films to assess the success of each fixation. Failure was defined as evidence of any cortical perforation on any side of the pedicle in or outside of the spinal canal. After completing the roentgenogram evaluation, the specimen was transected in the midljne, and the success of each pedicle fixation was evaluated by visual/tactile Inspection. There were no disagreements between surgeons on the visual/tactile evaluations of the specimens. In contrast, inter-rater adjusted percent agreement for roentgenograms ratings was 74, judged to be less than satisfactory when considering that the ratings were on a 2-point scale (S or F). The overall failure rate was approximately 21% (26/124). Of the 26 failures, 92% represented cortical perforations within the spinal canal. Discrepancies between visual/tactile and roentgenogram-based evaluations were not encouraging. Occurrence of false-positive results was at rates of 8.1% and 6.5%, and false-negative results occurred at a rate of 14.5% amd 12.9% for the more and !ess experienced surgeons, respectively. In general, success rate was independent of surgical factors. Success was not significantly related to approach, surgeon experience, screw size, or spine level. There was, however, an appreciable practice effect, x² = 8.84, P < 0.003. Failure rates were 26.4% compared to 6.4% for specimens 1-4 and 5-8, respectively. Also of interest was a trend for success to be related to approach, depending on the region of the spine involved, F_1,100 = 3.38, P < 0.07. Examination of relative success rates indicated no significant differences between the Weinstein and Roy-Camille approach in the upper lumbar spine (T11-L2), but a trend toward greater success with the Weinstein approach in the lower lumbar spine (L3-S1); a 93.1% success rate for the Weinstein, compared with 78.6% for the Roy-Camille approach. Roentgenograms were found to produce unacCeptably high rates of false-positive and false-negative evaluations. The lack of differences in success rates between the surgeons with different levels of experience in conjunction with a strong relationship between success rate and practice is consistent with poor roentgenogram evaluation. Surgeons cannot be expected to improve with experience when their tools (roentgenograms) do not allow accurate evaluation of their performance. However, significant improvement in success rate can be expected when accurate evaluation is provided. Unfortunately, this research exposes the inadequacies of current evaluation procedures without providing viable alternatives. The trend toward superior success rates with the Weinstein approach in the lower lumbar spine is particularly important since the approach is believed to have the added advantage of providing less interference with uninvolved adjacent motion segments. Such ah advantage would be of little value if the success rate was not at least as good as demonstrated for other approaches.