Development and validation of a risk prediction model for overall survival in patients with nasopharyngeal carcinoma: a prospective cohort study in China

Nasopharyngeal carcinoma (NPC) is an epithelial carcinoma originating from the nasopharyngeal mucosal tissue, with the characteristic of distinct geographical distribution of occurrence [1, 2]. NPC occurred highly in East and Southeast Asia [1], and it particularly prevalent in Guangdong and Guangxi, the regions of southern China [3]. In 2019 the number of NPC deaths in China reached 28,659, accounting for 40% of NPC deaths worldwide [4]. China accounts for a significant proportion of mortality of NPC over the world, especially in southern China [5, 6].

Epstein-Barr virus (EBV) is one of the most common causative agents, and it can be detected in all types of NPC [7]. Radiotherapy is the primary treatment choice for NPC treatment due to the radiosensitive characteristic of NPC tumor [8]. And precise staging is crucial for reducing mortality in patients with NPC. However, heterogeneities of clinical outcomes of the NPC patients with the same clinical stage and degree of EBV were reported in considerable recent research. Those findings indicated that it is not enough to refine the prediction of outcomes for NPC patients only considering single factors. Recently, a major current focus in the area of prognostic of NPC is to find more risk factors to get a more accurate predictive model. Numerous studies have demonstrated that hematological biomarkers were associated with survival outcomes of NPC patients, such as lymphocyte-albumin ratio (LAR), neutrophil–lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) [9‐11]. However, the literature related to the survival outcomes of the cohort among NPC patients that take various factors into account in model design are limited. Therefore, in this study, our objective was to develop a clinically useful prognostic model in which demographic variables, hematological biomarkers, and oncogenic pathogens were considered to predict overall survival among NPC patients in the region.

In the present study, we used the follow-up database from the Chongqing University Cancer Hospital to establish a novel nomogram prognostic model of NPC and complete internal verification, by incorporating demographics, hematological biomarkers, and oncogenic pathogens. And a user-friendly online calculator was developed to help clinicians in treatment decision making.

Several previous studies have been published using the nomogram to predict the OS of NPC patients. In 2018, Wu and colleagues evaluated a nomogram for predicting long-term OS for patients wish NPC using demographic variables and TNM stage [14]. And Huang etc. developed a prognostic nomogram to reveal the relationship between EBV and NPC in 2021 [15]. Although the western region is not a high incidence area of nasopharyngeal carcinoma, the survival prediction of its patients should not be ignored. Inflammatory markers have been widely used in various cancers, but rarely in NPC, thus we established a prognostic model considered several systemic inflammation parameters and EBV.

Several of our findings are worth highlighting. First, age, stage, radio therapy, EBV, LAR, NLR and PLR were recognized as independent prognostic parameters based on the univariate and multivariate cox regression analysis, and the conclusion was in general agreement with previous reports [9, 16‐18]. EBV infection is the most common causal agent [19] and a useful prognostic factor of NPC [20], and has been used to assess the disease progression and population screening [21]. LAR is a novel independent prognostic risk factor [9] and have a strong survival predictive power for OS in NPC [18]. Li etc. concluded that NLR could be an attractive indicator for evaluating the 5-year OS in NPC patients with stage III [22]. High PLR was associated with poor OS in NPC patients [23]. Notably, chemical therapy was suggested that it did not reach statistical significance to be a prognostic factor in our study, and it was unlike some previous results [24]. Considering the chemotherapy sensitizing the tumor to the toxic effects of the radiotherapy [25] and the choice to chemotherapy depending on clinical risk, for example, the results obtained in the study is reasonable. Radiotherapy is the primary curative treatment of NPC, and combing chemotherapy with radiotherapy is a rational option in the treatment of locoregionally advanced NPC [26]. Therefore, to avoid missing important factors and based on the clinical features, we conduct model incorporating chemotherapy. In addition, it should be noted that through the univariate model, the correlation between pathological and NPC was of no significance, which was contradictory to other researches.

In our study, the calibration curve pointed optimal accordance between predicted survival probability and actual value, which indicated good repeatability and reliability of the model. And the C-index presented the same performance of our model, in the range of 0.72–0.82 (in training and validation cohorts). In addition, the DCA curves illustrated a better performance of survival predictions of nomogram than the models with stage, EBV, and stage + EBV. In conclusion, the results were suggested that our nomogram was a reliable and precise prognostic tool to predict OS in NPC patients.

Our study is not devoid of limitations. First, there may exist a potential source for selection bias based on the serious inclusion and exclusion criteria. Second, our samples were collected from a single center from a non-endemic region in China and lack of external verification. It is necessary for our study, aimed at exploring the performance of combination of OS and disease-free survival, to design a multicenter randomized controlled study in the next step.

Patients with NPC have heterogeneous survival outcomes, which can be predicted using our novel prognostic model. And it can support help in clinicians deciding treatment and patient counseling. Furthermore, the nomogram was deployed into a website server for use.