PURPOSE: To identify MR predictors of bone cement leakage in patients receiving percutaneous vertebroplasty. MATERIALS AND METHODS: Percutaneous vertebroplasties were performed in 45 vertebras (T7; one, T8; two, T10; two, T11; two, T12; eight, L1; fifteen, L2; eight, L3; five, L4; two) in 35 patients (age 52-83). The procedure was performed using an 11 G Jamshidi needle, which was inserted into the target by the bipedicular approach. Kyphoplasty, unilateral pedicular approach and extrapedicular approach cases were excluded. Shortly after the procedure, all patients underwent a noncontrast CT covering the vertebroplasty sites. A retrospective study was performed to determine whether cement leakage is related to any of following MR findings: presence of cortical disruption of the vertebral body, severity of body compression (proportion of abnormal to normal vertebral body volumes), bone cement amount, bone cement amount/severity of body compression ratio, proportion of low-signal area in a vertebral body on T1 weighted image, presence of either vacuum or cystic portion below a linear dark signal in a fractured vertebra, and the location of dark signal intensity line in a vertebral body. Logistic discrimination model stepwise method was used in the statistical analysis. RESULTS: On post-vertebroplasty CT scan, bone cement leakage was detected in or around 29 vertebrae (64%), including 11 vertebrae (24%) where leakage was found in the epidural space or radial vein. No patients displayed any neurological symptoms or signs. The most frequent site of bone cement leakage was the anterior external vertebral venous plexus (49%). Endplate cortical bone disruption was related to an increased risk of intervertebral bone cement leakage (p<0.05). Bone cement leakage tended to occur less frequently when there is a vacuum or cystic change below the dark linear signal intensity in a fractured vertebra (p<0.05). No other MR findings showed a statistically significant correlation with bone cement leakage. CONCLUSION: On pre-vertebroplasty MR imaging, vertebral endplate cortical bone disruption and vacuum or cystic change below dark linear signal intensity in fractured vertebra showed a significant correlation with bone cement leakage.