The purpose of this study was to determine factors associated with the incidence of axillary lymph node metastases (ALNM) in T1 tumors and cases in which axillary dissection could be omitted. Data from 195 patients with T1 primary invasive breast cancer (size < or = 2 cm) who underwent either mastectomy or wide local excision of the tumor and axillary dissection were reviewed. ALNM was found in 59 of 195 patients with T1 tumors (30.3%). Tumor size was found to be the only independent predictor of ALNM, having a directly analogous relationship with the probability of invaded nodes: T1a (< or = 5 mm) tumors had 0 per cent ALNM, whereas T1b (5 mm < T1b < or = 10 mm) and T1c (10 mm < T1c < or = 20 mm) tumors had 25.7 per cent and 33.8 per cent ALNM respectively. Among the other factors studied (patient age, tumor site, hormone receptor status, histologic type, and grade of the tumor) only the histologic grade of the tumor cells appeared to correlate with the incidence of lymph node involvement, but this was not statistically significant. In conclusion only tumor size has statistically significant correlation with the incidence of ALNM. Routine axillary dissection could be omitted only in patients at minimal risk of ALNM (ductal carcinoma in situ and T1a) and when treatment decisions were not influenced by lymph node status (e.g., elderly patients with clinically negative axilla). Axillary dissection (at least levels I and II) should be performed in all cases with primary invasive breast cancer with tumor size > 5 mm.