Fulvestrant use in pretreated metastatic breast cancer patients is associated with variable response rates. This study aimed to characterize these responses and to develop a prediction model to identify those patients who could potentially derive the most clinical benefit. A nationwide review of patients enrolled in a Canadian compassionate use program from 1999 to 2006 was performed. Prior therapy with tamoxifen, steroidal, and nonsteroidal aromatase inhibitors was mandatory. The dependent variable in the analysis was the proportion of patients requiring chemotherapy at 3 months following the start of fulvestrant. General Linear Mixed modeling was used to identify factors significantly associated with this dependant variable and to subsequently develop the prediction model. Three hundred and five women received at least one dose of fulvestrant; 207 went on to receive chemotherapy (68%). Median duration of fulvestrant treatment was 4.1 months (range 0.8-63.1). Factors predictive of being chemotherapy free at 3 months included older age, no prior adjuvant hormonal therapy, and the absence of lung or brain metastases at the start of therapy. A receiver operating characteristic (ROC) curve analysis had an area under the curve of 0.70 (95% CI 0.60-0.80). This model was able to identify risk information that could be helpful in assessing which patients would most likely benefit from fulvestrant as an intervention with the objective being a delay in chemotherapy.