Prognostic models for prediction of overall survival after first-line platinum-based chemotherapy for primary metastatic nasopharyngeal carcinoma.

Abstract

e17501 Background: Platinum-based chemotherapy remains the main treatment strategy for patients with primary metastatic nasopharyngeal carcinoma (mNPC). However, survival outcomes are heterogeneous. The present study sought to develop clinical predictive models for overall survival (OS) in mNPC. Methods: Institutional database was queried for patients with mNPC who received first-line platinum-based chemotherapy between 2008 and 2015. Eligible patients (n = 460) were split into a training (2008-2013, n = 296) and an internal validation cohort (2014-2015, n = 164) according to the period of diagnosis. Potential prognostic factors for OS were evaluated using univariate and multivariate analyses. The performance of prognostic models was assessed by discrimination and calibration. The clinical usefulness of prognostic models was evaluated by decision curve analysis. Results: We established two nomogram-based prognostic models to predict OS probability, the first (baseline model) based on baseline factors, the second (composite model) incorporating posttreatment plasma Epstein-Barr virus DNA (EBV DNA) and objective response (OR) to first-line chemotherapy. The incorporation of posttreatment plasma EBV DNA and OR to the model improved the predictive accuracy. The C-index of composite model for predicting OS was 0.767 (P < .001), which was statistically higher than the C-index values for baseline model (0.686). The results were confirmed in the internal validation cohort. Conclusions: We developed and validated two prognostic models that can be used before and after completion of first-line chemotherapy, respectively. The proposed models can be used for prognosis prediction and better selection of patient into clinical trials.