Angiogenesis is a key tumor microenvironment (TME) event underpinning tumor growth and metastasis. Nevertheless, the relatively poor performance of anti-angiogenic therapies in clinical trials compared to pre-clinical studies implies that classical subcutaneous xenograft models have limited predictive potential in this setting. To address this issue, we established orthotopic surgical resection models of breast cancer, which replicate the phenotype of clinical post-resection micro-metastasis. To demonstrate the power and precision of these models, we recapitulated the BETH adjuvant trial (NCT00625898) where the addition of bevacizumab (BVZ) to chemotherapy plus trastuzumab (Trast) failed to provide additional benefit. SCID mice were orthotopically implanted with bioluminescent Her2+ MDA-MB-231 or HCC1954 cells and tumors resected c.5 weeks later. Following resection, mice were treated with 10 mg/kg Trast +5 mg/kg paclitaxel (PAC) IP once weekly for 6 cycles +/- weekly BVZ (5 mg/kg IP). Metastasis was monitored by imaging. Using these models our data confirms that the addition of the anti-angiogenic antibody BVZ to adjuvant Trast + chemotherapy provides no additional benefit compared with Trast + chemotherapy alone. Previous studies using non-resection subcutaneously engrafted xenografts failed to predict this outcome. Our results provide compelling evidence for the utility of cell line xenograft resection models to predict clinical outcome for TME targeting agents.