Dynamic prognostic nutritional index trajectories predict survival outcomes in nasopharyngeal carcinoma with persistently undetectable plasma Epstein–Barr virus DNA

Background This study aims to identify prognostic factors in nasopharyngeal carcinoma (NPC) patients with persistently undetectable plasma Epstein–Barr virus (EBV) DNA and to evaluate the association between longitudinal changes in the prognostic nutritional index (PNI) during treatment and survival outcomes. Methods We conducted a retrospective cohort study of 412 patients with persistently undetectable plasma EBV DNA NPC. Prognostic factors for progression-free survival (PFS) were identified via Cox regression analyses. A latent class growth mixture model (LCGMM) was performed on serial PNI measurements at baseline (T1), post-induction chemotherapy (T2), and post-radiotherapy (T3) in a subset of 252 patients receiving induction chemotherapy to characterize dynamic nutritional changes. Survival differences between PNI trajectory subgroups were assessed. Results Higher baseline PNI was independently associated with improved PFS (HR = 0.93, 95% CI: 0.87–0.98, p = 0.008) in patients with persistently undetectable plasma EBV DNA NPC. Longitudinal modeling revealed two distinct PNI trajectory classes in patients receiving induction chemotherapy: a high nutritional reserve group (27%) and a low nutritional reserve (73%) group. The high nutritional reserve type demonstrated superior PFS outcomes in both unadjusted (HR: 0.38, 95% CI: 0.16–0.91, p = 0.029) and adjusted models (HR: 0.37, 95% CI: 0.15–0.91, p = 0.031). Bootstrap validation supported model stability. No significant association was observed between PNI change rate and PFS ( p > 0.05). Conclusion Our findings demonstrate that baseline PNI is an independent prognostic factor for survival, and that the dynamic PNI trajectory during treatment provides significant prognostic information in NPC patients with persistently undetectable plasma EBV DNA. The identification of distinct nutritional reserve classes through longitudinal modeling provides a novel framework for real-time risk assessment and may help identify patients who could benefit from closer nutritional monitoring and supportive care in future prospective studies.