Purpose: Although, small (<= cm) node negative estrogen receptor positive breast cancer has a very good prognosis, treatment is controversial, with oncologists selecting tamoxifen alone (T) or chemotherapy plus tamoxifen (CT). A 21 gene assay has been developed to separate risk groups with the aim of selective use of the more toxic and costly CT. Methods: A Markov cost-effectiveness model was created comparing T vs. CT vs. Gene assay followed by T for low-risk patients and CT for intermediate or high risk patients. Because only 53 similar patients (<= cm) were in the initial validation study for the gene assay, we created several additional Markov models to explore a) varying combinations of test sensitivity and specificity b) variable oncologists' risk thresholds for chemotherapy, and c) simulation of a current multi-group trial (TAILORx) applied to larger (<5 cm) tumors Results: For the base case scenario, the gene assay is more effective and less expensive than CT, and has an incremental cost-effectiveness averaging $20K vs. T alone. This result is robust through a variety of sensitivity analysis, including the cost of the gene assay and chemotherapy. However, CT is preferred if sensitivity for detecting low risk is <0.85. Conversely, the specificity of the test has little effect on the preferred strategy. When extrapolated to larger tumors (>1 cm, < 5cm), the preferred strategy varies based on oncologists' reliance on test results, test characteristics, and tumor size. The model simulating TAILORx (T for low-risk, TC for high risk, and randomization for intermediate risk, suggests that T will be the preferred strategy for patients with recurrence scores <20, and therefore the use of the gene assay is cost-effective when applied to this larger subset of patients. Conclusions: Targeted treatment based on gene assay is a cost-effective approach to treatment for ER+, node negative patients. Before the test is considered standard practice, additional validation studies should be performed to ascertain the true characteristics of the test and compared with a simpler model based on parameters that are routinely utilized. The test has little value when used by oncologists with a very low threshold for chemotherapy. The current randomized trial should be useful as a validation study and to evaluate extrapolation to intermediate risk patients.