OBJECTIVE. We aimed to systematically examine the reliability and validity of different MRI sequences in differentiating benign and malignant vertebral fractures, appropriately select the best MRI sequence to improve the diagnostic accuracy, and compare the diagnostic accuracy of MRI sequences in the context of different study designs or publication date. MATERIALS AND METHODS. Computer and manual retrieval were conducted on studies published between January 1, 2000, and September 31, 2016. Studies relevant to the differential diagnosis of benign and malignant vertebral fractures by MRI and reference standard (histopathologic diagnosis or clinical follow-up examination) were analyzed. RESULTS. Eighteen articles were included. Neither threshold (p = 0.86) nor nonthreshold (p = 0.06) effects were present. The pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were 89% (95% CI, 86-92%), 88% (95% CI, 85-91%), 6.54 (95% CI, 4.44-9.65), 0.14 (95% CI, 0.09-0.21), and 55.76 (95% CI, 37.06-83.89), respectively. The AUC was 0.95. The risk of publication bias was negligible (p = 0.33). CONCLUSION. MRI sequences could provide appreciable diagnostic performance in differentiating benign and malignant vertebral fractures. However, our pooled estimates do not support the superiority of one set of sequences over another, and there is not sufficient evidence to show that prospective or recent studies are obviously better than retrospective or older studies.
Keywords: MRI; benign vertebral fracture; diagnostic accuracy; malignant vertebral fracture; meta-analysis.