Study design Retrospective study. Summary of background data Short-segment pedicle instrumentation (SSPI) is widely used to treat thoracolumbar junction fracture. Implant failure is the most common complication of SSPI and often necessitates revision surgery. The stand-alone anterior technique for failed SSPI provides excellent decompression and anterior column reconstruction, but it is incapable of restoring normal stability. High rate of complications is solely attributable to the anterior approach. Thus, the reconstruction of the anterior column with posterior compression instrumentation is the strategy of choice. In this study, we use a modified pedicle subtraction osteotomy (PSO) technique through a single posterior approach as the salvage method for the failed SSPI. Materials and methods Thirteen patients with failed SSPI after thoracolumbar fracture were included and followed up at regular intervals. Revision indications include intractable pain, deteriorating neurological deficits, and progressive deformity. The modified PSO was performed. After osteotomy, the anterior cortex was thinned and the anterior longitudinal ligament was also preserved to maintain stability during correction. Autograft was inserted into the osteotomy gap to increase stability and fusion rate. The correction had been achieved with closure. We evaluated the patients' clinical symptoms, segmental kyphosis correction, bony fusion time, and complications. Results Segmental kyphosis correction was from preoperative average 20.9 degrees (range, 9.5-38.5 degrees) to 3.0 degrees (range, 1-5.5 degrees) immediately after operation and 6.1 degrees (range, 3-8 degrees at the last follow-up). Bony fusion was confirmed on radiographs in all patients at an average of 9.9 months (range, 7.5-12 mo) after revision surgery. There was no implant failure in any of the 13 patients. Average preoperative visual analog scale was 6.5 (range, 5-9) and reduced to 3.7 (range, 2-5) at the last follow-up. There was also a significant decrease in mean preoperative Oswestry Disability Index from 55.2 (range, 38-76) to 32.8 (range, 16-56). No patients suffered any neurological deterioration related to revision surgery. Complications were encountered in 6 patients (46.1%), including 4 with cerebrospinal fluid leak and 2 with superficial wound infection. All these complications were managed conservatively and none of them underwent reoperation. Conclusion We conclude that modified PSO possesses the advantages of excellent kyphosis correction, a safe and reliable salvage alternative for the revision of SSPI failure.