External validity of a prognostic nomogram for locoregionally advanced nasopharyngeal carcinoma based on the 8th edition of the AJCC/UICC staging system: A retrospective cohort study

Abstract

Background: The tumor-node-metastasis (TNM) staging system does not perform well for guiding individualized induction or adjuvant chemotherapy for patients with locoregionally advanced nasopharyngeal carcinoma (NPC). We attempted to externally validate the Pan's nomogram, developed based on the 8th edition of the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) staging system, for patients with locoregionally advanced disease. In addition, we investigated the reliability of Pan's nomogram for selection of participants in future clinical trials. Methods: This study included 535 patients with locoregionally advanced NPC who were treated between March 2007 and January 2012. The 5-year overall survival (OS) rates were calculated using the Kaplan-Meier method and compared with predicted outcomes. The calibration was tested using calibration plots and the Hosmer-Lemeshow test. Discrimination ability, which was assessed using the concordance index, as compared with other predictors. Results: Pan's nomogram was observed to underestimate the 5-year OS of the entire cohort by 8.65% [95% confidence interval (CI) - 9.70 to - 7.60%, P < 0.001] and underestimated the 5-year OS of each risk group. The differences between the predicted and observed 5-year OS rates were smallest among low-risk patients (< 135 points calculated using Pan's nomogram; which predicted minus observed OS, - 6.41%, 95% CI - 6.75 to - 6.07%, P < 0.001) and were largest among high-risk patients (≥ 160 points) (- 13.56%, 95% CI - 15.48 to - 11.63%, P < 0.001). The Hosmer-Lemeshow test suggested that the predicted and observed 5-year OS rates had no ideal relationship (P < 0.001). Pan's nomogram had better discriminatory ability compared with the levels of Epstein-Barr virus DNA acid (EBV DNA) and the 7th or 8th AJCC/UICC staging system, although not better compared with the combination of EBV DNA and the 8th staging system. Additionally, Pan's nomogram was marginally inferior to our predictive model, which included the 8th AJCC/UICC N-classification, age, gross primary tumor volume, lactate dehydrogenase, and body mass index. Conclusions: Pan's nomogram underestimated the 5-year OS of patients with locoregionally advanced NPC at our cancer center, and may not be a precise tool for selecting participants for clinical trials.