UNLABELLED The aim of this study was to find both predictors of axillary lymph node involvement and predictors of level II or III involvement. With these predictors patients should be selected who can be spared axillary lymph node dissection or who qualify for removal of only first-level lymph nodes. 239 consecutive patients with invasive breast cancer stage I-III treated with total axillary dissection were evaluated. In multiple logistic regression analysis 17 clinical and histopathological variables were included. We found 4 multivariate significant predictors for metastatic axillary involvement: Clinically positive axilla, peritumoural lymphatic vessel invasion, multicentric or multifocal tumours and a large number of past pregnancies. Multivariate significant predictors of involvement of level II or III were peritumoural lymphatic vessel invasion, younger age, larger tumour size, multicentric or multifocal tumours and postmenopausal patient. The metastatic involvement of more than 3 nodes was associated with young age and clinically positive axilla. With these easily available predictors a reliable assessment of risk of metastatic involvement of the upper levels of the axillary nodes could be defined. The binary logistic regression model to axillary node involvement had a rather unsatisfactory predictive value. CONCLUSIONS The risk calculation of level II or III involvement with these predictors may help the surgeon to find the adequate extent of axillary clearance needed in the individual patient. Multiparity proved to be an independent predictor of metastatic axillary node involvement.