To determine if breast tumour characteristics varied by antibiotic use prior to diagnosis in 2,266 women with primary, invasive breast cancer. Subjects were women enrolled at Group Health Cooperative, a health plan in western Washington state, for at least 1 year and diagnosed with breast cancer between 1 January 1993 and 30 June 2001. Case status, tumour features, and patient characteristics were ascertained from the Surveillance, Epidemiology, and End Results cancer registry and Group Health Cooperative electronic files. Prescription information was obtained from electronic pharmacy records. Compared to non-use, antibiotic use prior to breast cancer diagnosis was not associated with a less favourable tumour profile (as measured by cancer stage, grade, and size), oestrogen receptor negative tumours, or lobular histology, after controlling for age and length of enrollment. Nonetheless, while not achieving statistical significance, our results suggest that antibiotic use may be associated with less favourable breast tumour features. Overall, we found no association between antibiotic use and breast tumour features and no dose-response gradient. However, the results are consistent with the possibility that antibiotic use may increase the risk of less favourable tumours. Larger studies are required to further investigate this hypothesis.