OBJECTIVES: The goal of personalized medicine is to identify the right treatment for the right patient at the right time. Prior to treating metastatic breast cancer (MBC) patients with Herceptin, tumors are tested for overexpression of HER2.  Still for many HER2 + patients (treated with herceptin), disease progression continues. It has previously been shown that for MBC changes in patients’ serum HER-2 levels during treatment is predictive of their eventual response to therapy. Through modeling and simulation we examine the potential impact of ceasing Herceptin therapy prior to the end of the planned treatment cycle for patients whose serum HER-2 levels predict an eventual lack of therapeutic response. METHODS: Markov Cycle Tree models were constructed to simulate disease progression and therapy for MBC patients using our custom simulation software, ProFound. The progression of disease was dependent on the patient’s current therapy: Herceptin, Tykerb, or Paclitaxel, and model parameters were based on meta-analysis of clinical trials. Patients are stratified into three sets: Serum HER-2 Increasing (>20%), Not Changing, and Decreasing (>20%). We compared the following alternative treatment strategies: Paclitaxel, Herceptin, and initial Herceptin with change in therapy after 1 month based on changes in serum HER2. Patients removed from Herceptin either simply discontinued therapy or were switched to Tykerb. RESULTS: Compared with continued treatment with Herceptin, moving patients with >20% increase in Serum HER-2 to Tykerb resulted in an additional one life-month gained. All other strategies performed worse than continued Herceptin therapy, including moving patients whose Serum HER2 levels are not decreasing to Tykerb. This highlights the importance of identifying the right subset of patients who will benefit from a change in therapy. CONCLUSIONS: A biomarker that can predict therapy failure prior to the end of treatment as part of the treatment decision-making process may extend the lives of patients.