Can diffusion-weighted imaging be used to differentiate benign from pathologic fractures? A meta-analysis

Abstract

Objective: Conventional MR sequences are sometimes nonspecific in differentiating benign from pathologic fractures. To address this difficulty, diffusion-weighted images were conjectured to aid in this discrimination with variable results. As each of these studies contained somewhat small numbers of patients, we performed a meta-analysis to determine if this sequence may be used for this important diagnostic problem.

Materials and methods: We reviewed and statistically analyzed the results of eight studies, performed between 1998 and 2003, comparing diffusion-weighted magnetic resonance signal intensity characteristics of benign and pathologic vertebral body fractures. Diffusion-weighted imaging (DWI) signal characteristics and apparent diffusion coefficient (ADC) values of 104 benign fractures and 161 combined malignant vertebral body lesions and pathologic fractures were statistically evaluated in terms of mean ADC, as well as percentage classified as either hypointense or isointense. The meta-analysis to compare benign fractures with the combined pathologic fractures and metastatic lesions in terms of mean ADC used Hedge's g statistic with a small sample bias adjustment; the comparison of the percentage hypo- or isointense used the Mantel-Haenszel method to calculate a weighted summary odds ratio. All summary effect sizes were computed under a random effects model to account for study heterogeneity.

Results: The mean ADC was significantly higher (p < 0.01) among benign fractures, with a standardized mean difference (SMD) of 2.8 and a 95% confidence interval (CI) for the SMD of 2.1 to 3.5. Lesions classified as hypointense were significantly more likely to be benign (p < 0.01), based on a summary odds ratio (OR) of 24.5 and 95% confidence that the OR exceeds 1.7. Lesions classified as isointense were not significantly more likely to be benign or malignant (p > 0.1), based on a summary OR of 3.6 and a 95% CI for the OR of 0.35 to 36.6.

Conclusion: Even though the literature has been inconsistent, ADC maps appear to be a reliable method to differentiate benign from malignant fractures.