In a prospective study, we attempted to predict axillary metastases in 135 breast cancer patients by a preoperative multiparameter evaluation including axillary ultrasound scanning (US). After surgery, factors associated with lymph node metastases by univariate analysis were included in a multivariate model. By multivariate analysis, the stronger independent predictors of lymph node metastases were suspicious axillary US (p<0.001), tumor location in the outer quadrants (p=0.001) and high Ki-67 index (>10%) (p=0.002). A predictive model based on these variables, identified a high-risk group (20.0%) represented by women with suspicious axillary US, tumor in the outer quadrants and high Ki-67 index, with axillary metastases in 100%, whereas all patients with opposite features (8.1%) had uninvolved axillary lymph nodes. This multiparameter evaluation including axillary US may be used to optimize the selection of breast cancer patients candidate to sentinel lymph node biopsy or axillary lymph node dissection. The accuracy of this predictive model still requires prospective validation in a larger sample of women.