Background: The safety and the effectiveness of pedicle-screw instrumentation in the spine have been questioned despite its use worldwide to enhance stabilization of the spine. This review was performed to answer questions about the technique of insertion and the nature and etiology of complications directly attributable to the screws.Methods: We performed a retrospective review of all of the pedicle-screw procedures that were done by us from January 1, 1984, to December 31, 1993. We inserted 4790 screws during 915 operative procedures on 875 patients; 668 (76.3 percent) of the patients had a lumbosacral arthrodesis. The mean duration of follow-up was three years (range, two to five years). The accuracy of screw placement was assessed on intraoperative, immediate postoperative, and follow-up radiographs with use of a technique that was developed by one of us (F. D.); this technique has yet to be validated to determine the prevalence of various types of error.Results: Of the 4790 screws, 4548 (94.9 percent) had been inserted within the pedicle and the vertebral body. One hundred and thirty-four (2.8 percent) of the screws had perforated the anterior cortex, and this was the most common type of perforation. One hundred and fifteen (2.4 percent) of the screws were associated with complications that could be ascribed to the use of pedicle screws. The most common problem was late-onset discomfort or pain related to a pseudarthrosis or perhaps to the screws; this problem was associated with 1102 (23.0 percent) of the screws, used in 222 (24.3 percent) of the procedures. The symptoms necessitated removal of the instrumentation with or without repair of the pseudarthrosis. A pseudarthrosis was found during forty-six (20.7 percent) of the 222 procedures.Irritation of a nerve root occurred after nine procedures (1.0 percent) and was caused by eleven screws (0.2 percent); it was more commonly caused by medially placed screws. Three patients had residual neurological weakness despite removal of the screws. Twenty-five screws (0.5 percent), used in twenty procedures (2.2 percent), broke. The screws that broke were of an early design. A pseudarthrosis was found in thirteen of twenty patients who had broken screws. Sixteen of the twenty patients had an exploration; three of them were found to have a solid fusion, and thirteen were found to have a pseudarthrosis. The remaining four patients had evidence of a solid fusion on radiographs and had no pain.Conclusions: There are few problems associated with the insertion of screws, provided that the surgeon is experienced and adheres to the principles and details of the operative technique. Our review revealed a low rate of postoperative complications related to pedicle screws. The problem of late-onset pain may be related to the implants or to the stiffness of the construct; however, it is difficult to accurately identify its exact etiology.