Patient derived xenografts (PDX) are valuable, clinically-relevant models of cancer. Their close genomic, phenotypic, and temporal association with patient tumors makes them well-suited for pre-clinical and co-clinical studies that assess the potential of new therapeutics. Here we report extension of this technology to inform patient prognosis in an IRB-approved study of precision oncology. Based on earlier data indicating that PDX engraftment correlates with poor outcomes in breast cancer, we tested the prospective validity of this information by conducting a blinded non-interventional trial with 80 patients. We found that PDX engraftment is a strong, independent prognostic factor for early relapse and death from breast cancer in triple negative and hormone receptor-low, HER2-negative breast cancers, and is more accurate and specific than standard assessments for recurrence risk. This study confirms that tumor engraftment assays can measure the aggressiveness of tumors, identifying patients who are at highest risk for poor outcome and may benefit from additional treatment. Future directions are to validate the findings with longer follow-up time, and to develop a clinically feasible surrogate biomarker for engraftment as a future prognostic test.