Simple SummaryThe majority of breast cancer deaths are caused by the spread of the disease to distant locations. The biological processes and molecular characteristics that eventually transform breast cancer into a life-threatening metastatic disease are not fully understood. The molecular subtyping of breast cancer into four tumor subtypes—namely luminal A, luminal B, human epidermal growth factor receptor 2-enriched, and basal-like subtypes—has been implemented for therapeutic guidance in patients with early breast cancer. It is not settled whether molecular subtypes in metastatic tissue can guide the choice of systemic therapy and how these subtypes may change throughout tumor progression. In this study, breast cancer subtypes at different stages of the disease were investigated, and we found changes to more unfavorable subtypes to be common throughout the progression of the disease. These findings suggests that molecular subtyping in metastatic disease could add important prognostic and predictive information to complement information from the primary tumor.Background: PAM50 breast cancer intrinsic subtyping adds prognostic information in early breast cancer; however, the role in metastatic disease is unclear. We aimed to identify PAM50 subtypes in primary tumors (PTs) and metastases to outline subtype changes and their prognostic role. Methods: RNA was isolated from PTs, lymph node metastases (LNMs), and distant metastases (DMs) in metastatic breast cancer patients (n = 140) included in a prospective study (NCT01322893). Gene expression analyses were performed using the Breast Cancer 360 (BC360) assay from Nano-String. The subtype shifts were evaluated using McNemar and symmetry tests, and clinical outcomes were evaluated with log-rank tests and Cox regression. Results: The PAM50 subtype changed in 25/59 of paired samples between PTs and LNMs (Psymmetry = 0.002), in 31/61 between PTs and DMs (Psymmetry < 0.001), and in 16/38 between LNMs and DMs (Psymmetry = 0.004). Shifts toward subtypes with worse outcomes were the most common. Patients with shifts from the luminal PT to non-luminal DM subtypes had worse progression-free survival compared to patients with a stable subtype (hazard ratio (HR): 2.3; 95% confidence interval (CI): 1.14–4.68, p = 0.02). Conclusion: Strong evidence of PAM50 subtype shifts toward unfavorable subtypes were seen between PTs and metastatic samples. For patients with a shift in subtype from luminal PT to non-luminal DM, a worse prognosis was noted.