Breast cancer is noted for long periods of tumor dormancy and metastases can occur many years after treatment. Adjuvant chemotherapy is used to prevent metastatic recurrence but is not always successful. As a model for studying mechanisms of dormancy, we have used two murine mammary carcinoma cell lines: D2.0R/R cells, which are poorly metastatic but form metastases in some mice after long latency times, and D2A1/R cells, which form more numerous metastases much earlier. Previously we identified a surprisingly large population of dormant but viable solitary cells, which persisted in an undivided state for up to 11 weeks after injection of D2.0R/R cells. Dormant cells were also detected for D2A1/R cells, in a background of growing metastases. Here we used this model to test the hypothesis that dormant tumor cells would not be killed by cytotoxic chemotherapy that targets actively dividing cells, and that the late development of metastases from D2.0R/R cells would not be inhibited by chemotherapy that effectively inhibited D2A1/R metastases. We injected mice with D2A1/R or D2.0R/R cells via a mesenteric vein to target liver. We developed a doxorubicin (DXR) treatment protocol that effectively reduced the metastatic tumor burden from D2A1/R cells at 3 weeks. However, this treatment did not reduce the numbers of solitary dormant cells in mice injected with either D2A1/R or D2.0R/R cells. Furthermore, DXR did not reduce the metastatic tumor burden after an 11-week latency period in mice injected with D2.0R/R cells. Thus, apparently effective chemotherapy may spare non-dividing cancer cells, and these cells may give rise to metastases at a later date. This study has important clinical implications for patients being treated with cytotoxic chemotherapy.