Insulin receptor substrate-1 (IRS-1) is a cytoplasmic scaffolding protein that is phosphorylated by insulin-like growth factor-I receptor and recruits downstream effectors. Recent evidence suggests that IRS-1 has a nuclear localization and function. Here we investigated whether nuclear and cytoplasmic IRS-1 levels are associated with clinico-pathological characteristics and clinical outcome in breast cancer patients. Tissue microarrays from 1,097 patients with stage I-II breast cancer were stained by immunohistochemistry for IRS-1. Nuclear and cytoplasmic IRS-1 were scored separately according to the Allred score. Nuclear IRS-1 showed a positive association with estrogen receptor (ER) (r = 0.09, P = 0.003) and progesterone receptor (PR) (r = 0.08, P = 0.008) status and a negative correlation with lymph node involvement (r = -0.10, P = 0.001). Cytoplasmic IRS-1 did not correlate with ER or PR but showed a positive correlation with tumor size (r = 0.10, P = 0.001) and S-phase fraction (r = 0.16, P < 0.001). In univariate analysis, tamoxifen-treated patients with tumors showing positive nuclear IRS-1 had a better recurrence-free survival (RFS) (P = 0.009) and overall survival (OS) (P = 0.0007), while no association was shown between cytoplasmic IRS-1 and RFS or OS in the same group of patients. In multivariate analysis of patients receiving tamoxifen, negative nuclear IRS-1 showed a significantly reduced RFS (P = 0.046) and OS (P = 0.018). Combining both PR and nuclear IRS-1, tamoxifen-treated patients with PR+/IRS-1+ tumors had a better RFS (P = 0.0003) and OS (P < 0.0001) when compared with patients with PR-/IRS-1- tumors. In conclusion, nuclear IRS-1 may be a useful marker to predict tamoxifen response in patients with early breast cancer, particularly when assessed in combination with PR.