The small number of nodes harvested with lymphatic mapping and sentinel lymph node (SLN) biopsy has allowed a more detailed pathologic examination of those nodes. Immunohistochemical stains for cytokeratin (CK-IHC) have been used in an attempt to minimize the false negative rate for SLN mapping. This study examines the value of CK-IHC positivity in predicting further lymph node involvement in the axillary basin. From April 1998 through May 1999, 519 lymphatic mappings and SLN biopsies were performed for invasive breast cancer. SLNs were examined by imprint cytology, hematoxylin and eosin (H&E), and CK-IHC. Patients with evidence of metastatic disease by any of the above techniques were eligible for complete axillary node dissection (CAND). The frequency with which these modalities predicted further lymph node involvement in the axillary basin was compared. Of the 519 lymphatic mappings, 39 patients (7.5%) had a CK-IHC-positive-only SLN. Five (12.8%) of these 39 patients had at least 2 SLNs positive by CK-IHC. Twenty-six of the CK-IHC-positive-only patients underwent CAND. Three of these 26 patients (11.5%) had additional metastases identified after CAND. The sensitivity levels with which each modality detected further axillary lymph node involvement were as follows: CK-IHC, 98 per cent; H&E, 94 per cent; and imprint cytology, 87 per cent. A logistic regression to compare the prognostic value of the three modalities was performed. All were significant, with odds ratios of 19.1 for CK-IHC (P = 0.015), 5.3 for H&E (P = 0.033), and 3.86 for imprint cytology (P = 0.0059). These data validate the enhanced detection of CK-IHC for the evaluation of SLNs. Detection of CK-IHC-positive SLNs appears to warrant CAND in patients with invasive breast cancer. However, the therapeutic value of CAND or adjuvant therapies based on CK-IHC-positive SLNs would be best answered by prospective randomized trials.