A prospective study in 35 osteoporotic patients with 120 multiple-level vertebral compression fractures (VCFs) assessed the use of magnetic resonance imaging (MRI) to determine painful vertebrae for treatment with kyphoplasty (KP). A total of 51 vertebrae were identified as painful and selected for KP based on changes in MRI signal intensity between T1-weighted, T2-weighted and short tau inversion recovery MRI. Efficacy was assessed by the mean change in anterior/middle vertebral body height, Cobb's angle, a visual analogue pain scale and the Oswestry Disability Index at pre-operative, post-operative and final follow-up assessments. Significant improvements in all efficacy measures were observed at the postoperative versus pre-operative assessments; no significant differences were observed between post-operative and final follow-up assessments. It is concluded that painful vertebrae can be determined by MRI signal intensity changes and their selection for KP can improve outcomes in patients with multiple-level VCFs.