OBJECTIVE: In this study, we retrospectively evaluate the surgical outcome of the cervical fracture-dislocation in order to define the criteria for the operative approach selection based on fracture characteristics. METHODS: Thirty one consecutive patients (29 males and 2 females) who underwent operation for the cervical fracture-dislocation between 1997 and 2001 at a single institute were included in this study. Plain X-ray, computed tomography, and magnetic resonance imaging studies were performed in all patients. Injuries were characterized using Denis's three-column plain X-ray model, Daffer's computed tomography, and Oner's magnetic resonance imaging classifications. The Frankel classification was used for neurological deficits. The mean postoperative follow-up period was 11.93 months (range 1-61 months). RESULTS: The anterior approach was performed in 14 and the posterior approach in 10 of the 31 patients, respectively. Both anterior and posterior fixation was performed electively in 7 of the 31 patients. Daffer and Oner's classification was found to be able to predict surgical failures by either the anterior or the posterior approach. Three-column injury was misinterpreted as two-column injury only by plain radiography. No differences in neurological outcome, pain relief, or bone fusion rate were observed between the anterior approach and the posterior approach. However, the posterior approach was associated more frequently with postoperative kyphosis and instability. CONCLUSION: Surgical approach is usually determined on the basis of whether the compression is ventral or dorsal. Anterior fixation only may be an alternative to both anterior and posterior fixation in three-column injury, but posterior fixation alone is not.