Treatment options for metastatic breast cancer (MBC) are complex, and some patients experience early discontinuation or switching of treatment (ETDS). We examined the relationship between ETDS and patient-reported symptom burden among patients receiving first-line treatment of MBC in community oncology settings. This retrospective observational study used the ACORN Data Warehouse, a comprehensive community oncology repository of medical records and patient-reported outcomes. Patients with first-line treatment for MBC who had Patient Care Monitor (PCM) surveys were eligible. ETDS was defined as: record stating ETDS, treatment duration < planned, and planned therapy <6 weeks. Symptom burden was measured by two PCM composite scores [continuous (0-22) and categorical (absent, mild, moderate, and severe)] computed from 22 PCM items with varying cut points to assess symptom burden over time. Cox regression with time-varying covariates was used to assess risk for ETDS controlling for patient characteristics and treatment type: chemo (chemotherapy without targeted therapy (±hormone therapy); targeted (chemotherapy plus targeted therapy (±hormone therapy); and hormone (hormone therapy only). Overall, 197 (24.7 %) of a total sample of 797 patients had an ETDS event, of which 109 (55.3 %) were switches rather than early discontinuation. ETDS rate was nominally lower in the hormone group (11.1 %) versus chemo (27.6 %) or targeted (26.1 %). PCM continuous composite score predicted ETDS, controlling for other variables (HR = 1.132, p < 0.0001). ETDS was predicted by moderate and severe levels of PCM categorical composite score (HR = 4.135, and HR = 5.287 vs. absent, respectively, both p < 0.0001), with the pattern suggesting a threshold effect. Moderate or severe levels of a wide range of patient-reported symptoms and the accumulation of symptoms over time significantly predicted ETDS. Providers may better maintain patients on planned therapy if they attend to overall symptom burden patients experience over time.