Treatment response-adapted risk index model for survival prediction and adjuvant chemotherapy selection in nonmetastatic nasopharyngeal carcinoma

Dynamic response to therapy is strongly associated with cancer outcomes. We aim to develop the response-adapted individualized risk index (RAIRI) as an individual prognostic approach and predictive biomarker for adjuvant chemotherapy (AC) benefit in nasopharyngeal carcinoma (NPC) based on pretreatment clinical characteristics, longitudinal cell-free Epstein-Barr virus DNA, and MRI-based tumor regression measurements collected during treatment. Using Bayesian joint model, we developed and validated RAIRI, a dynamic and multidimensional model, with 2148 patients in training, internal validation, external validation, and RCT cohorts (ClinicalTrials.gov NCT02958111 2016-11-04 and NCT02143388 2014-05-18). RAIRI predictions were refined over time using serially collected longitudinal data. RAIRI demonstrated accurate calibration and high prognostic accuracy, superior to conventional models. In RCT cohort, RAIRI identified approximately 70% of low-risk patients who did not benefit from AC, whereas the high-risks experienced substantial benefits from AC. Therefore, RAIRI could provide real-time updated quantitative survival estimates for individuals and facilitate personalized AC selection.