To create an individualized predictive tool for the risk of malignancy in solid breast masses, based on echographic and clinical characteristics. Research Ethics Committee approval and informed consent were obtained. This multi-center study included 1,403 solid breast masses prospectively. Each ultrasound feature was analyzed and compared with the definitive diagnosis. The ultrasound results, women's ages and family histories of breast cancer were included in a multivariate logistic regression model. Among the 1,403 lesions included in the study, 1,390 (99.1%) had a conclusive diagnosis: 343 malignant tumors (24.7%), and 1,047 benign masses (75.3%). The odds ratio (and confidence interval) for breast malignancy for each variable included in the model, as calculated by multivariate analysis, were as follows: irregular shape/noncircumscribed margins, 16.02 (7.75-33.09); heterogeneous echo texture, 4.50 (2.42-8.23); vertical orientation (not parallel to the skin), 2.23 (1.04-4.75); anterior echogenic rim, 2.62 (1.09-6.31); posterior shadowing, 2.38 (1.23-4.62); age more than 40 years, 2.19 (1.26-3.81); positive first-degree family history (mother, sister or daughter), 7.50 (2.65-21.18). There was no advantage in including the presence of internal vascularity, presence of thickened Cooper's ligaments or size of the mass, in the model. The predictive tool was named SONOBREAST and it is freely available for medical purposes on the internet site: http://www.sonobreast.com. The probability of malignancy in breast masses can be specified based on their ultrasound features, the woman's age and the family history of breast cancer.
© 2011 Wiley Periodicals, Inc.