The association between tumor cyclooxygenase 2 (COX-2) expression and breast cancer prognosis has been inconsistent. The purpose of this study was to evaluate the prognostic significance of COX-2 tumor expression according to adjuvant treatment, and potential effect modifications of non-steroid anti-inflammatory drug (NSAID) use, and other tumor and lifestyle factors. A prospective cohort of 1,116 patients with primary breast cancer in Lund, Sweden, included 2002-2012 was followed until June 2014 (median 5 years). Tumor-specific COX-2 expression was evaluated on tissue microarrays using immunohistochemistry. Associations between COX-2 intensity (negative, weak-moderate, high) and patient and tumor characteristics as well as prognosis were analyzed. Tumor-specific COX-2 expression was available for 911 patients and was significantly associated with higher age at diagnosis and less aggressive tumor characteristics. Higher COX-2 expression was associated with lower risk for breast cancer events during the first five years of follow-up, adj HR 0.60 (95%CI: 0.37-0.97), per category. The association between COX-2 expression and prognosis was significantly modified by oral contraceptive (OC) use (Pinteraction ?=?0.048), preoperative NSAID use (Pinteraction ?=?0.009), and tumor size (Pinteraction ?=?0.039). COX-2 negativity was associated with increased risk for events during the first five years in ever OC users, adj HR 1.94 (1.01-3.72) and during the 11-year follow-up in preoperative NSAID users, adj HR 4.51 (1.18-11.44) as well as in patients with large tumors, adj HR 2.57 (1.28-5.15). In conclusion, this study, one of the largest evaluating COX-2 expression in breast cancer, indicates that the prognostic impact of COX-2 expression depends on host factors and tumor characteristics.