Purpose: To evaluate the diagnostic utility of second-look ultrasonography (US) in the assessment of lesions identified at breast magnetic resonance (MR) imaging.
Materials and methods: A systematic review of the PubMed database for articles published up to January 6, 2013, was performed by using predefined search terms applied in a standardized manner. Second-look US studies for the assessment of breast lesions identified at MR imaging were eligible for this meta-analysis. Two independent reviewers performed the literature review and data extraction. Eligible studies presented data on the number of lesions examined and the number of lesions detected at second-look US. The reference standard for lesion diagnosis was either histopathologic or follow-up examination. Sources of bias were assessed by using the Quality Assessment of Diagnostic Accuracy Studies 2, or QUADAS-2 Quality Assessment of Diagnostic Accuracy Studies 2 , tool. Statistical analysis included data pooling, heterogeneity testing, and meta-regression.
Results: Seventeen studies that included benign and malignant lesions met the inclusion criteria. The general lesion detection rate at second-look US was very heterogeneous and ranged between 22.6% and 82.1% (pooled rate, 57.5% [1266 of 2201]; 95% confidence interval [ CI confidence interval ]: 50.0%, 64.1% [random-effects model]; I(2) = 90.9%; P < .0001). The highest second-look US detection rates were observed for mass lesions (as opposed to nonmass lesions) and malignant (vs benign) lesions (P < .001 for both). Pooled positive and negative predictive values (positive or negative second-look US correlates of MR imaging-detected malignant or benign lesions) were calculated as 30.7% (95% CI confidence interval : 25.3%, 36.4%; I(2) = 75.4%; P < .0001) and 87.8% (95% CI confidence interval : 82.0%, 92.7%; I(2) = 82.1%; P < .0001), respectively, by using random-effects models.
Conclusion: The results of this study demonstrated variable utility of second-look US in MR imaging-detected lesions, as lesion detection rates were very heterogeneous. Subgroup analysis showed that malignant and mass lesions were more likely to be detected at second-look US. Furthermore, malignancy was not excluded if a lesion was not detected at second-look US.