This study was performed to investigate whether long-term monitoring of dynamic changes in plasma Epstein-Barr virus (EBV) DNA could improve prognosis prediction of nasopharyngeal carcinoma (NPC). About 1077 nonmetastatic NPC patients were recruited to retrospectively analyze the prognostic value of plasma EBV DNA load pretreatment and 3, 12, 24, and 36 months posttreatment. We also examined the prognostic value of dynamic changes in plasma EBV DNA at various time points. Patients with plasma EBV DNA load above optimal pre- and posttreatment cut-offs had significantly worse five-year progression-free survival, distant metastasis-free survival, locoregional relapse-free survival, and overall survival (OS) at all-time points, excluding only OS at 36 months posttreatment due to limited mortalities. Patients with persistently undetectable plasma EBV DNA at the first four time points had the best prognosis, followed by those with positive detection pretreatment and consistently negative detection posttreatment, those with negative detection pretreatment and positive detection at one time point posttreatment, and those with positive detection pretreatment and at one time point posttreatment, whereas patients with positive detection at ≥2 time points posttreatment had the worst prognosis. Cox proportional hazard models identified the dynamic change pattern as an independent prognostic factor, and receiver operating characteristic curve analysis demonstrated that the dynamic change at four time point was more valuable than any single time point for predicting disease progression, distant metastasis, locoregional relapse, and mortality. Dynamic changes in plasma EBV DNA pre- and posttreatment could predict the long-term survival outcome and provide accurate risk stratification in NPC. © 2020 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.