Early distant relapse occurs in a minority of node-negative breast cancer patients. Whether this poor prognosis can be predicted by the features of the primary tumour, or by the presence of occult metastases in the "negative" lymph nodes (LNs), remains a matter of debate. One hundred and four T(1-2)N(0)M(0) breast carcinoma patients were divided into two groups: group 1 (44%) showing early distant relapse with a median disease-free survival of 25 months, and group 2 (56%) showing no evidence of disease after a median follow-up of 91.5 months. All patients had received locoregional treatment only. All tumours were evaluated for medial/lateral location, histological type, size, grade, mitotic activity, fibrotic focus, necrosis, angiogenesis, growth pattern, and lymphatic vessel permeation. The haematoxylin and eosin-stained slides of all axillary LNs were revised and two additional levels were cut from each paraffin block for cytokeratin immunohistochemistry. In 24 patients (23%), occult metastases were found. These consisted of single cells or small clusters (SCs) in the marginal sinus in 17 patients (16%) and of larger colonies of cells in seven patients (7%). All detected metastases were smaller than 2 mm in diameter (micrometastases). There was no significant correlation between the presence of occult LN metastases (SCs or colonies) and the prognostically important features of the primary tumour. Early metastatic disease was significantly correlated with larger tumour size (p=0.02), higher histological grade (p=0.0008), mitotic activity (p<0.0001), presence of necrosis (p=0.0004), presence of fibrotic foci (p=0.0005), angiogenesis (p=0.0009), and lymphatic vessel permeation (p=0.018). Multiple logistic regression analysis showed that histological grade and the presence of a fibrotic focus were the only independent prognostic factors and that the presence of occult LN metastases was inversely correlated with early distant relapse. Prospective prognostic studies of occult LN metastases should consider the features of the primary tumour in a multivariate analysis.