Some patients with breast cancer currently undergo bone marrow biopsy to make clinical decisions regarding therapy; however, lobular carcinoma can be difficult to detect in routine histologic sections. The authors reviewed retrospectively all available bone marrow biopsies from patients with lobular carcinoma diagnosed between January, 1, 1989, and September, 25, 1997, at the City of Hope National Medical Center to identify useful morphologic features and to determine the utility of pan-keratin immunohistochemical (IHC) staining. A total of 65 biopsies from 54 patients were reviewed. Thirteen of the 65 biopsies were classified initially as containing metastatic tumor based on histologic features alone. With the addition of keratin IHC, seven additional cases of metastatic disease were detected. Forty of the 54 patients received stem cell replacement or autologous bone marrow transplantation. Disease-free survival after high-dose chemotherapy with stem cell replacement or autologous bone marrow transplantation was stratified into three groups based on hematoxylin and eosin (H&E) staining and IHC results. Two-year disease-free survival was 33% for the H&E−/IHC+ group versus 90% for the H&E−/IHC− group (p = 0.005) among patients clinically free of disease at the time of stem cell replacement or autologous bone marrow transplantation. Two-year disease-free survival was 0% in the H&E+/IHC+ group (p = 0.04, compared with the H&E−/IHC+ group). The authors conclude that routine morphologic examination without the aid of keratin IHC is unreliable in detecting clinically relevant metastatic lobular carcinoma in bone marrow biopsies. These findings suggest that pan-keratin immunostaining may be indicated on bone marrow biopsy specimens from lobular carcinoma patients if the biopsy appears histologically negative for metastatic tumor on H&E sections.