Purpose/objective(s): Patient-derived xenografts (PDX) can help identify oral cavity squamous cell carcinoma (OSCC) patients at risk for disease recurrence and optimize clinical decision-making. In this study, we develop and validate a prediction score for locoregional failure (LRF) and distant metastases (DM) in OSCC that incorporates PDX engraftment in addition to known clinicopathological risk factors.Materials/methods: PDX models were generated from OSCC patients. Patients were scored as a ''non-engrafter" if PDX formation did not occur within 6 months. Multivariable analysis (MVA) was used to identify predictors of LRF and DM. Factors retained in the final MVA were used to construct a prediction score and classify patients into risk groups using a 10-fold cross-validation approach.Results: Overall 288 OSCC patients were analyzed. MVA identified pT3-4, pENE, and engraftment as predictors of LRF and DM. Patients whose tumors engrafted (n = 198) were more likely to develop LRF (HR 1.98, 95% CI: 1.24-3.18, P < 0.01), and DM (HR 2.64, 95% CI 1.21-5.75, P < 0.01) compared to non-engrafters. A prediction score based on the aforementioned variables identified patients at high-risk (defined as having at least two of the three high risk features i.e., engraftment, pT3-4, pENE) and low-risk for LRF (43.5% vs 26.5% at 5-years, P < 0.001), DM (38.2% vs 8.4% at 5-years, P < 0.001), and poorer 5-year OS (34% vs 66%, P < 0.001). The prediction model that included engraftment had the highest discriminatory capacity in the cross-validation analysis (AUC: 67.8 [63.5-72.9]), while removal of engraftment as a predictor resulted in a lower c-index (AUC: 62.7 [57.0-68.4]). In patients classified based on a clinical score only (i.e., presence or absence of pT3-4 and pENE), engraftment remained useful in identifying those with worse outcomes. Compared to non-engrafters, engraftment was associated with higher rates of DM (15.8% vs. 5.4%, P < 0.05) in clinically "high risk" patients as well as higher rates of LRF (31.9% vs. 13.8%, P < 0.05) in clinically "low risk" patients at 5-years. Finally, engraftment was associated with poorer 5-year OS in both clinically "high risk" (36% vs. 65%, P < 0.05), and "low risk" patients (57% vs. 78%, P < 0.01).Conclusion: A prediction score utilizing OSCC PDX engraftment, in conjunction with pT3-4 and pENE, improves the prognostic utility of existing clinical models and predicts patients at risk for LRF, DM and poor survival.