To measure the failure rate of percutaneous iliosacral screw fixation of vertically unstable pelvic fractures and particularly to test the hypothesis that fixations in which the posterior injury is a vertical fracture of the sacrum are more likely to fail than fixations with dislocations or fracture-dislocations of the sacroiliac joint. Retrospective review. Level 1 trauma center. All patients with pelvic fractures admitted between January 1, 1993, and December 31, 1998, were identified from the trauma registry. Hospital records were used to identify patients treated with iliosacral screws. Radiologic studies were examined to identify patients who had unequivocally vertically unstable pelvic fractures. Immediate postoperative and follow-up anteroposterior, inlet, and outlet radiographs from a minimum of 12 months postinjury were examined. Position, length, and numbers of iliosacral screws and any evidence of screw failure (eg, bending or breakage) were recorded. Residual postoperative displacement and late displacement of the posterior pelvis were measured. The main outcome measure was failure, defined as at least 1cm of combined vertical displacement of the posterior pelvis compared with immediate postoperative position. The main analysis was for association between fracture pattern and failure. Patient demographic data, iliosacral screw position, and anterior pelvic fixation method also were studied. The study group comprised 62 patients with unequivocally vertically unstable pelvic fractures in whom the posterior injury was treated with closed reduction and percutaneous iliosacral screw fixation. Of patients, 32 had dislocations or fracture-dislocations of the sacroiliac joint, and 30 had vertical fractures of the sacrum. Fixation failed in four patients, all with vertical sacral fractures and all within the first 3 weeks after surgery. These four patients required revision fixation. In two further cases with vertical sacral fractures, there was evidence that the fracture had only barely been held by the fixation, but these fractures healed, and follow-up radiographs did not meet the displacement criteria for failure. A vertical sacral fracture pattern was associated significantly with failure (Fisher exact test, P = 0.04); the excess risk of failure compared with sacroiliac joint injury was 13% (95% confidence interval 1% to 25%). There was no significant association between failure and anterior fixation method, iliosacral screw arrangement or length, or any demographic or injury variable. Percutaneous iliosacral screw fixation is a useful technique in the management of vertically unstable pelvic fractures, but a vertical sacral fracture should make the surgeon more wary of fixation failure and loss of reduction.