Nasopharyngeal carcinoma (NPC) is a malignant tumor with a high incidence in South China. The majority of patients are defined as locoregionally advanced disease (LA-NPC, stages III-IVa) (1, 2). Concurrent chemoradiation (CCRT) is regarded as a standard treatment modality for patients with LA-NPC (3). In recent years, the addition of induction chemotherapy (IC) to the previously established regimen has received considerable attention. A previous randomized phase 2 study reported that the application of IC to CCRT was superior to CCRT alone for 3-year overall survival (OS, 94.1% vs. 67.7%, P = 0.012), and also led to a trend to improve progression-free survival (4). Later a phase 3 randomized controlled trial provided evidence that cisplatin/fluorouracil/docetaxel (TPF) induction chemotherapy (IC) followed by CCRT provided 5-year overall survival (OS: 85.6% vs. 77.7%, p = 0.042) benefits compared with CCRT alone in LA-NPC (5), and also improved long-term failure-free survival (FFS, 77.4% vs. 66.4%, p = 0.019), distant failure-free survival (DMFS, 88% vs. 79.8%, p = 0.030), and locoregional failure-free survival (FFS, 90.7% vs. 83.8%, p = 0.044). Other studies also indicated that TPF, cisplatin/fluorouracil (PF), or gemcitabine/cisplatin (GP) IC plus CCRT significantly improved tumor control and survival than CCRT alone (6–9). IC has since played an important role of the treatment regimen for LA-NPC. Although IC produces a survival benefit, 20% to 30% of patients continue to have local recurrence or distant metastasis after standard CCRT (10). Hence, developing a prognostic model for predicting survival outcome and early progression to guide risk stratification and treatment regimen modification is urgent.

Currently, the Union Internationale Contre le Cancer/American Joint Committee Cancer (UICC/AJCC) TNM staging system is widely applied for survival outcome prediction and treatment decision guidance in NPC. Nevertheless, the prognosis differs among patients with the same stage (11). The probable reason for this phenomenon is that the TNM staging system is mainly based on anatomic location, which is unable to reflect the biological variability of the tumor itself. Thus, we need a new effective prognostic model to refine the current TNM staging system and accurately predict which patients would benefit from more intensive treatment.

In this study, we used Cox regression analysis to identify important predictors and to develop a risk model for the prediction of 3-year OS and PFS probabilities and risk stratification in NPC patients.

In this study, a nomogram was developed and validated by a retrospective analysis of patients with locoregionally advanced NPC to predict OS and PFS individually and showed good prognostic value in clinical decision making. The results showed that tumor volume and EBV DNA load after induction chemotherapy (IC) are independent predictors for survival. Our nomogram was established by the addition of post-IC tumor volume and EBV DNA load to the eighth UICC/AJCC TNM staging system and improved prognostic accuracy, which would assist clinicians in guiding further risk stratification and early treatment modification.

In endemic areas, NPC is an Epstein-Barr virus (EBV)-driven malignancy. Previous studies have reported that pretreatment EBV DNA load is considered to be an indicator of tumor load, and is correlated with risk stratification and prognosis prediction (22–25). In addition, Li’s (19) and Yang’s (18) studies both developed a nomogram for survival prediction using pretreatment EBV-DNA load. However, the clinical prognostic models containing EBV DNA load after IC is limited. Some studies demonstrated that the post-DNA is a powerful and early prognostic factor in LA-NPC patients (26, 27). Our results agree with the findings of these studies, which found that detectable post-DNA is an effective factor in predicting treatment failure. The mechanism behind this phenomenon may be related to the relatively unsatisfactory sensitivity of chemotherapy in patients with detectable post-DNA, and it reflects the existence of residual tumor cells. More specifically, EBV DNA may originate from necrotic or apoptotic cells and is representative of tumor DNA levels (28).

In clinical practice, treatment decisions mainly depend on the current TNM staging system, which does not accurately reflect the tumor burden. Tumor volume represents the tumor burden and can be easily calculated from the radiotherapy planning system. The Response Evaluation Criteria in Solid Tumors criteria mainly measure the reduction or enlargement in the maximum diameter of the tumor on the two-dimensional imaging (29). However, the value of using traditional response evaluation criteria to evaluate the response of NPC patients to IC is limited in clinical practice, as the primary tumor of LA-NPC patients is irregular in shape and tends to invade the bone structure at the base of the skull. Therefore, it may not be appropriate to evaluate the response based on the conventional criteria for LA-NPC. In contrast, the tumor volume provides three-dimensional information that more intuitively and accurately reflects the response to treatment, thus reflecting the patient’s sensitivity to chemotherapy (30). Multiple studies have suggested that tumor volume is significantly associated with survival outcome (31–33). However, a major number of studies have focused on the prognostic significance of pretreatment tumor volume (34, 35), and the potential of post-IC tumor volume for the prediction of prognosis and guidance of treatment regimens remains unclear. Chen et al. (36) found that post-IC GTVnx and GTVnd are considered indicators of survival prognosis and have important significance in guiding treatment decisions. Intriguingly, according to our findings, we found that a larger post-IC GTVnx or GTVnd was related to a significantly worse clinical outcome compared than a smaller post-IC GTVnx or GTVnd for 3-year OS and PFS. Hence, post-IC tumor volume can not only reflect tumor load but also represent the chemotherapeutic sensitivity of IC to a certain extent and can guide the risk stratification of LA-NPC.

Based on the conclusions of these studies, the levels of post-DNA, GTVnx, and GTVnd were evaluated in this study and used as independent prognostic variables to establish the nomograms. The TNM staging system is based primarily on anatomical information, and it lacks precision in stratifying patients at risk and individual therapy. The prognostic nomograms of OS and PFS were established in combination with the important prognostic factors in this study, and the C index and AUC were higher than those of the eighth TNM staging system, which indicated that our models were better than the current staging system in terms of prognostic efficiency. Recently, the net reclassification improvement (NRI) has been recognized as a new indicator that can be used to compare the accuracy of the predictive power of the two models. We used the NRI to quantify the improvement in survival prediction from the addition of some new markers compared to the traditional staging. Adding post-IC EBV and tumor volume on the basis of the traditional TNM staging system yielded NRI of 0.098 for OS and 0.064 for PFS. In addition, patients can be divided into different risk subgroups according to the total score produced by the nomogram, in which the patients in the high-risk group had a poorer prognosis. Our results demonstrated that the new model could improve the predictive power and the accuracy of individual risk stratification of the eighth TNM staging system.

In the era of rapid development of modern treatment, individualized treatment becomes particularly important. Patients in the high-risk group are likely to have a higher tumor burden. Thus, identifying high-risk patients as early as possible may be the key to improve survival outcomes and will assist in guiding treatment decisions prior to CCRT. The nomogram in our study shows great potential for application in clinical practice. Clinicians could evaluate the condition of LA-NPC patients before the implement of CCRT using the model and select high-risk patients who may benefit from intensive treatment. Specifically intensive treatment after IC included the following: [1] increase the cumulative cisplatin dose during CCRT. Many patients who receive IC before CCRT experience grade 3 or 4 adverse events, which would reduce their tolerance to concurrent chemotherapy (7). Therefore, it is important to screen patients with LA-NPC who may benefit from a higher cumulative cisplatin dose (CCD). Wen et al. demonstrated that high-risk patients with advanced T or N stage, higher pre-DNA or larger post-IC tumor volume benefited from CCD ≥ 200 mg/m² for PFS and DMFS (37). A previous study also showed that patients with detectable post-DNA showed significantly improved 3-year PFS and LRFS by receiving ≥ 160 mg/m² CCD (38). [2] Add targeted therapy to the treatment regimen. Epidermal growth factor receptor (EGFR) is also highly expressed in NPC (39) and a previous study has shown that the application of cetuximab (CTX) or nimotuzumab (NTZ) (anti-EGFR monoclonal antibodies) was a significant protective factor for OS, DFS, and DMFS in patients treated with CCRT (40). [3] Add in adjuvant chemotherapy to treatment. Previous studies have demonstrated a survival benefit from adjuvant chemotherapy in high-risk subgroups (such as advanced T and N stage, higher NLR and LDH) by stratifying patients for risk (11, 41, 42). However, it has also been shown that adding adjuvant chemotherapy to CCRT increases acute toxicity in patients with LA-NPC (43, 44). Therefore, the application of traditional intravenous adjuvant chemotherapy regimens in NPC is controversial. Previous retrospective studies have shown that oral fluorouracil chemotherapeutic agents (such as capecitabine) are well tolerated in adjuvant chemotherapy, can be protective factors for survival and may be more suitable for patients with radical radiotherapy (42, 45, 46). [4] Increase the use of immunotherapy. Some features of NPC justify the use of immunotherapy, such as the association with EBV infection, upregulation of PD-L1 expression, and a high number of tumor-infiltrating lymphocytes (47). For example, the discovery of the co-inhibitory molecules PD-1 and PD-L1 were groundbreaking event for immunotherapy in this era of personalized management. For patients with PD-L1-positive NPC, a 22% response rate was observed after anti-PD-1 treatment (48). [5] Increase the radiotherapy dose. Some studies have reported that the minimum point dose (Dmin) escalation to the primary gross tumor volume may lead to better local control (10, 49). Our findings indicated that although treatment failure in LA-NPC is not uncommon, low-risk patients may have a better prognosis with a relatively lower tumor burden. Intensive treatments, such as targeted therapy, immunotherapy, and adjuvant therapy, may not produce survival benefits but increase the risk of toxicity and the economic burden. Moreover, a previous study has confirmed that a higher cumulative cisplatin dose during CCRT was not superior to a lower one in terms of survival for LA-NPC patients in the low-risk group with a smaller post-IC tumor volume or undetectable post-IC EBV DNA (37, 38). As a result, for these patients, screening needs to be early and the intensity of treatment can be reduced to avoid unnecessary toxicities and costs.

The advantages of this study are as follows. First, this is the first study ever reported to establish a model that features independent variables after IC in LA-NPC patients. The results of this study have certain educational and practical relevance. Second, in most healthcare institutions, the variables involved in the nomogram are readily available, so the model has broad clinical applicability. Finally, the nomogram is a visually predictive tool that can be used in clinical practice to help doctors quickly assess the prognosis of patients through simple calculations, and to classify patients with different severities to help determine individualized treatment.

Our study had some limitations. First, our follow-up time was not long enough and the number of patients was relatively small because of the key inclusion criteria. There should be a longer follow-up time and a larger study population if possible. Second, studies from a single institution do not provide robust evidence; we need prospective studies with large cohorts to validate the conclusion. Moreover, radiomics is a new research hot topic that has been confirmed by an increasing number of research studies to have strong prognostic potential (18, 50). As the technology to acquire radiomic features of medical imaging has not been widely developed in our center, this study failed to discuss it in depth. The further study of radiomics is needed. Lastly, similar to other retrospective studies, the selection bias was unavoidable. However, this retrospective study has clinical significance because it provides a reference for the planning of some prospective studies.

In summary, our study developed and validated a novel model for predicting OS and PFS rates after induction chemotherapy for LA-NPC patients in an endemic area that involved the eighth TNM staging system, post-IC EBV DNA and tumor volume. The predictive value of the nomogram was obviously better than that of the current eighth TNM staging system. In the future, prospective studies are required to generalize the clinical usefulness of our model.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Affiliated Cancer Hospital of Guangxi Medical University. The ethics committee waived the requirement of written informed consent for participation. Written informed consent was not obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Study concepts: X-DZ and Y-TJ. Study design: X-DZ, Y-TJ, and K-HC. Data acquisition: Y-TJ, K-HC, JY, and Z-GL. Quality control of data and algorithms: Y-TJ and K-HC. Data analysis and interpretation: Y-TJ, K-HC, JY, and Z-GL. Statistical analysis: Y-TJ, K-HC, SQ, and LL. Manuscript preparation: Y-TJ, K-HC, JY, and Z-GL. Manuscript editing: Y-TJ, K-HC, SQ, and LL. Manuscript review: SQ, LL, and X-DZ. All authors read and X-DZ approved the final manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by the Key Research and Development Program Project of Guangxi Zhuang Autonomous Region (grant GuikeAB18221007) and the Independent Project of Key Laboratory of Early Prevention & Treatment for Regional High-Incidence Tumor (grant GKE-ZZ202014).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.