The presence of axillary lymph node metastases (ALNMs) is the most important prognostic factor in breast carcinoma. If ALNMs were predictable without performing axillary lymph node dissection (ALND), this procedure would not be necessary in selected patients. Using a combination of some of the new biological markers with the classical ones, our objective was I) to identify the best set of predictors of ALNMs, and II) to define predictive models with either high or low probability of ALNMs. We studied 102 patients with invasive breast carcinoma. All patients underwent ALND, and at least 10 axillary lymph nodes per case were obtained. In the primary tumour we evaluated size, histological subtype and grade, lymphatic/vascular invasion and margin. Hormone receptor status, MIB1 index, microvessel density, c-erbB-2 and cathepsin D expression were assessed by immunohistochemistry, and DNA ploidy and S-phase by flow cytometry. Risk factors for ALNMs were estimated by nonlinear logistic regression analysis. The best predictors of ALNMs were: tumour size > 2 cm [OR 6.45, 95% confidence interval (CI) 21.74 to 1.91], presence of lymphatic/vascular invasion [OR 4.95, CI (14.50 to 1.69)], infiltrative margin [OR 9.87 CI (37.44 to 2.60)] and high MIB-1 index [OR 8.39, CI (33.47 to 2.10)]. Two subsets had a very high risk of ALNMs: I) tumour size > 2 cm, with lymphatic/vascular invasion and infiltrative margin; 26 (89.66%) of 29 patients of this subgroup had ALNMs, and (II) tumour size > 2 cm, with lymphatic/vascular and high MIB1 index.; eight of the nine (89%) patients of this subgroup had ALNMs. We could also identify a two-variable model with a very low risk of ALNMs constituted by tumour with circumscribed margin and low MIB-1 index. Of the 19 patients showing these features, only 1 (5.26%) had ALNMs. Therefore, pathological features of the primary tumour can help to assess the risk for ALNM in invasive breast carcinoma. Such risk assessment might avoid regional surgical overtreatment.