Purpose: The aims of the study was to analyze the features of cancers missed as tumor on ultrasonography (US), to determine the predictive power of US tumor descriptors in the differentiation of benign and malignant breast tumors, to evaluate US as adjunct to mammography, and to assess the validity and reliability of mammographic, US, and combined interpretation of breast tumors.
Material and methods: Prospectively recorded mammographic and US findings of 355 malignant breast tumors among 2,985 consecutive patients who underwent breast US were compared with clinical findings and pathologic subtypes of the tumors. In addition, a 10-year material of 95 invasive lobular carcinomas (ILCs) were investigated. Three retrospective studies assessed the validity and reliability of mammographic, US, and combined interpretation of 200 palpable mammographically noncalcified breast masses by four radiologists.
Results: A total of 97.5% of the palpable and 67.9% of the nonpalpable malignant neoplasms were detected as tumor on US. Most carcinomas missed as tumor on US were ductal carcinomas in situ (DCISs) and microinvasive ductal cancers dominated by DCIS manifesting with suspicious calcifications on mammography. Irregular shape, irregular contour, extensively hypoechogenicity, hyperechoic rim (halo), and distortion of the surrounding tissue were the US features with the highest odds of predicting carcinomas. A pseudocapsule is the strongest predictor of a benign tumor, the odds of cancer being 0.03 in nonpalpable and 0.08 in palpable breast tumors. A negative predictive value of 100% in palpable and 96% in nonpalpable tumors was achieved using strict US criteria. In patients with ILC, US measurements predicted tumor size more accurately than mammography. US as adjunct to mammography correctly diagnosed ("upgraded") 9.5% of tumors with benign or indeterminate mammographic diagnoses. Excluding mammographically conclusive malignant tumors and carcinomas presenting with microcalcifications. US correctly upgraded 42% of the palpable and 44% of the nonpalpable cancers. Combined mammographic-US interpretation offers the highest diagnostic performance in noncalcified breast tumors. The lowest interobserver agreement was found in US interpretation.
Conclusion: The impact of US in mixed cancer populations is limited. US is, however, a valuable adjunct to mammography in patients with nonconclusive mammographic findings. Negative predictive values on US approaching 100% may be achieved using strict criteria for a benign diagnosis. A considerable interobserver variation in the US interpretation is a limiting factor for the potential of breast US in the differentiation of benign and malignant breast tumors.