This study analyzes the interobserver variability in interpreting atypia on breast core needle biopsies and in each category of atypia calculates the upgrade risk of carcinoma in the subsequent surgical excision. We identified 51 cases of atypia on breast core needle biopsies performed at our institution from January 2003 to August 2006. The atypia was classified into 4 categories: atypical ductal hyperplasia (ADH), atypical lobular hyperplasia (ALH), flat epithelial atypia (FEA), and atypia of undetermined significance (AUS). After a tutorial session, these cases were independently reviewed by four pathologists, whose overall multi-rater kappa value for agreement on different categories of atypia was 0.79 (95% CI, 0.69-0.89), which is within the substantial agreement range. The upgrade risk in each category of atypia was as follows: ADH 20% (p = 0.04); ALH 10% (p = 0.6); FEA 16.6% (p = 0.23), and AUS 100% (p = 0.96). Based on our findings, we conclude that follow-up excision should be performed after a diagnosis of ADH. The upgrade risk did not reach statistical significance in ALH or FEA. Although follow-up excision cannot be strongly recommended in ALH and FEA, it should be considered since the upgrade risk is not negligible. Strict adherence to the diagnostic criteria and tutorial sessions can help pathologists to achieve substantial agreement in interpreting atypia on breast core needle biopsies.