Prediction of the prognosis for metastatic breast cancer patients depends on molecular subtypes similar to those found in patients with primary breast cancer. Several studies have shown that estrogen receptor (ER) and progesterone receptor (PR) status determine the course of the disease and the prognosis. As Ki-67 helps to differentiate molecular subtypes in patients with primary breast cancer, the aim of this study was to assess the prognostic relevance of Ki-67 in the primary tumor in relation to its prognostic relevance for patients with metastatic breast cancer. A total of 467 patients with invasive breast cancer were identified in the database of a single breast cancer center, in whom Ki-67 had been assessed in tumor material from the breast at the time of the primary diagnosis and who had developed a metastasis at any time during the subsequent course. For these patients, tumor and patient characteristics were used to determine prognostic factors relative to overall survival after the diagnosis of distant metastases. Ki-67 was added to this model to investigate whether this might improve the prediction of overall survival. In the multivariate Cox model, age at diagnosis, body mass index, nodal status, tumor size, ER and PR status, and time from diagnosis to metastasis were identified as relevant prognostic factors. Adding Ki-67 to the model improved the prediction of overall survival. There was also a significant and relevant interaction with the PR status. In patients with a low-proliferation primary tumor, a high level of PR expression would indicate an extraordinarily good prognosis (HR 0.39; 95 % CI, 0.23-0.66). In patients with higher-proliferation primary tumors, PR status was not capable of differentiating prognostic groups. Ki-67 is useful in addition to known prognostic factors for breast cancer. It is able to indicate a group of women with a poorer prognosis, specifically in the group of patients with PR-positive breast cancer.