Purpose: Previous conflicting results about the prognostic significance of estrogen receptor (ER)-h in breast cancer may be explained by contribution of isoforms, of which five exist. Our aim was to elucidate the prognostic significance of ERh1, ERh2, and ERh5 by immunohistochemistry in a large cohort of breast carcinomas with long-term follow-up. Experimental Design: Tissue microarrays were stained with ERh1, ERh2, and ERh5 antibodies and scored as percentage of positive tumor cells and using the Allred system. Nuclear and cytoplasmic staining was evaluated and correlated with histopathologic characteristics, overall survival (OS), and disease-free survival (DFS). Results: Nuclear ERh2 and ERh5, but not ERh1, significantly correlated with OS (P = 0.006, P = 0.039, and P = 0.099, respectively), and ERh2 additionally with DFS (P = 0.013). ERh2 also predicted response to endocrine therapy (P = 0.036); correlated positively with ERa, progesterone receptor, androgen receptor, and BRCA1; and correlated inversely with metastasis and vascular invasion. Tumors coexpressing ERh2 and ERa had better OS and DFS. Cytoplasmic ERh2 expression, alone or combined with nuclear staining, predicted significantly worse OS. Notably, patients with only cytoplasmic ERh2 expression had significantly worse outcome (P = 0.0014). Conclusions: This is the first study elucidating the prognostic role of ERh1, ERh2, and ERh5 in a large breast cancer series. ERh2 is a powerful prognostic indicator in breast cancer, but nuclear and cytoplasmic expression differentially affect outcome. Measuring these in clinical breast cancer could provide a more comprehensive picture of patient outcome, complementing ERa.