Research synthesis: What is the diagnostic performance of magnetic resonance imaging to discriminate benign from malignant vertebral compression fractures? systematic review and meta-analysis

Study Design. This study is a research synthesis of the published literature evaluating the performance of magnetic resonance imaging (MRI) for differentiation of malignant from benign vertebral compression fractures (VCFs). Objective. Perform a systematic review and meta-analysis to summarize and combine the published data on MRI for discriminating malignant from benign VCFs. Summary of Background Data. The differentiation between benign and malignant VCFs in the spine is a challenging problem confronting spine practitioners. Methods. MEDLINE, EMBASE, and other databases were searched by 2 independent reviewers to identify studies that reported the performance of MRI for discriminating malignant from benign VCF. Included studies were assessed for described MRI features and study quality. The sensitivity, specificity, and diagnostic odds ratio (OR) of each feature were pooled with a random-effects model weighted by the inverse of the variance of each individual estimate. Results. A total of 31 studies with 1685 subjects met the selection criteria. All the studies focused on describing specific features rather than overall diagnostic performance. Signal intensity ratio on opposed phase (chemical shift) imaging 0.8 or more (OR = 164), apparent diffusion coefficient on echo planar diffusion-weighted images 1.5 × 10^-3 mm/s or less with b value 500 s/mm² (OR = 130), presence of other noncharacteristic vertebral lesions (OR = 55), presence of paraspinal mass (OR = 33), involvement of posterior element (OR = 28), involvement of pedicle (OR = 24), complete replacement of normal bone marrow in VCF (OR = 19), presence of epidural mass (OR = 13), and diffuse convexity of posterior vertebral border (OR = 10) were associated with malignant VCFs, whereas coexisting healed benign VCF (OR = 0.006), presence of "fluid sign" (OR = 0.08), presence of focal posterior vertebral border convexity/retropulsion (OR = 0.08), and band-like shape of abnormal signal (OR = 0.07) were associated with benign VCFs. Conclusion. Several specific MRI features using signal intensity characteristics, morphological characteristics, quantitative techniques, and findings at other levels can be useful for distinguishing benign from malignant VCFs and can serve as inputs for a prediction model. Observer performance reliability has not been adequately assessed.