Aims: To compare the efþcacy and possible complications of two approaches in the management of unstable burst fractures of the thoracolumbar spine. Methods: retrospective review of two consecutive series of patients with a minimum 2Y follow-up. Group 1, 22 patients operated up to 1998: isolated posterior approach using a rod-screw instrumentation, posterior grafting and correction of the kyphosis by in situ contouring of the rods. Group 2, 15 patients operated from 1998: isolated approach for strut grafting with rib, iliac crest or þbular fragments and a Z-Plate instrumentation. Results: preop, postop and 2Y FU kyphosis averaged respectively 11û/11.6û, 2.4û/-2û, 6.5û/5û in group 1 and 2. Initial correction was better with the posterior in-situ contouring of the rods but there was more loss of correction at þnal follow-up (7û). With the anterior approach, initial correction was more difþcult to obtain but loss of correction was less (4.1û). Though more invasive, the thoracic, thoracolumbar or lumbar anterior approaches did not lead to complications in our series. In 1/3 of the patients operated by anterior approach, a two level fusion could be performed instead of a three level fusion. Conclusions: the two procedures gave similar þnal results but an early surgery was necessary in the case of a posterior approach whereas correction remained possible after a greater delay with the anterior procedure. The anterior approach allowed a shorter fusion in cases where the inferior part of the vertebral body was not severely damaged.