Capecitabine is commonly used in treating breast cancer; however, therapeutic response varies among patients and there is no clinically validated model to predict individual outcomes. Here, we investigated whether drug sensitivity quantified in ex vivo patients' blood-derived cell lines can predict response to capecitabine in vivo. Lymphoblastoid cell lines (LCLs) were established from a cohort of metastatic breast cancer patients (n = 53) who were prospectively monitored during treatment with single agent capecitabine at 2000 mg/m2/day. LCLs were treated with increasing concentrations of 5'-DFUR, a major capecitabine metabolite, to assess patients' ex vivo sensitivity to this drug. Subsequently, ex vivo phenotype was compared to observed patient disease response and drug induced-toxicities. We acquired an independent cohort of breast cancer cell lines and LCLs derived from the same donors from ATCC, compared their sensitivity to 5'-DFUR. As seen in the patient population, we observed large inter-individual variability in response to 5'-DFUR treatment in patient-derived LCLs. Patients whose LCLs were more sensitive to 5'-DFUR had a significantly longer median progression free survival (9-month vs 6-month, log rank p-value = 0.017). In addition, this significant positive correlation for 5'-DFUR sensitivity was replicated in an independent cohort of 8 breast cancer cell lines and LCLs derived from the same donor. Our data suggests that at least a portion of the individual sensitivity to capecitabine is shared between germline tissue and tumor tissue. It also supports the utility of patient-derived LCLs as a predictive model for capecitabine treatment efficacy in breast cancer patients.