Objective: Despite general agreement on the goals of surgical treatment in thoracolumbar burst fractures, considerable controversy exists regarding the choice of operative techniques. This study is to evaluate the efficacy of short-segment fixation for thoracolumbar burst fractures after long-term follow-up and to analyze the causes of treatment failures. Methods: 48 out of 60 patients who underwent short-segment fixation for thoracolumbar burst fractures between January 1999 and October 2002 were enrolled in this study. Their neurological status, radiological images, and hospital records were retrospectively reviewed. Simple radiographs were evaluated to calculate kyphotic angles and percentages of anterior body compression (%ABC). Results: The average kyphotic angles were $20.0^{circ}$ preoperatively, $9.6^{circ}$ postoperatively, and $13.1^{circ}$ at the latest follow-up. The average %ABC were 47.3% preoperatively, 31.2% postoperatively, and 33.3% at the latest follow-up. The treatment failure, defined as correction loss by $10^{circ}$ or more or implant failure, was detected in 6 patients (12.5%). 5 out of 6 patients had implant failures. 2 out of 5 patients were related with osteoporosis, and the other 2 were related with poor compliance of spinal bracing. 3 patients with poor initial postoperative alignment had implant failure. 4 patients with screws only on the adjacent vertebrae and not on the injured vertebra itself showed poor initial and overall correction. Conclusion: With proper patient selection, adequate intraoperative reduction with screw fixation involving the injured vertebra, and strict postoperative spinal bracing, the short-segment fixation is an efficient and safe method in the treatment of thoracolumbar burst fracture.