Subdivision of de novo metastatic nasopharyngeal carcinoma based on tumor burden and pretreatment Epstein-Barr virus DNA for therapeutic guidance of primary tumor radiotherapy

Purpose: To improve individualized treatment of de novo metastatic nasopharyngeal carcinoma (dmNPC) patients by investigating prognostic factors and identifying patients who achieved better survival outcomes after locoregional radiotherapy (LRRT). Materials and methods: Our study included a cohort of 498 dmNPC patients. Overall survival (OS) was the primary endpoint. We analyzed the correlation of all potential prognostic factors and survival outcomes by Kaplan-Meier survival curves using log-rank test and Cox proportional hazards regression model. Results: Multivariate analysis identied three independent prognostic factors: Epstein-Barr virus (EBV) DNA, number of metastatic lesions, and number of metastatic organs. Through these factors, we successfully divided all patients into 3 subgroups: low-risk (single metastatic organ, EBV DNA ≤ 25,000 copies/ml, and ≤ 5 metastatic lesions), intermediate-risk (single metastatic organ, EBV DNA > 25,000 copies/ml, and ≤ 5 metastatic lesions), and high-risk (multiple metastatic organs or > 5 metastatic lesions or both). By comparing LRRT and non-RT groups, we found statistical differences in OS in the low-risk and intermediate-risk subgroups (p = 0.039 and p = 0.010, respectively) but no signicant difference in OS in the high-risk subgroup (p = 0.076). Further multivariate analysis of different risk stratications revealed that LRRT was a protective factor only for the low- and intermediate-risk subgroups. Conclusions: The risk stratication of dmNPC may be used as a new prognostic factor to help clinicians organize individualized LRRT treatment to improve the survival outcomes of dmNPC patients.


Introduction
Nasopharyngeal carcinoma (NPC) is a malignancy predominantly associated with infection by the Epstein-Barr virus (EBV) (1). It originates in the nasopharyngeal epithelium, which is found on the nasopharyngeal posterior wall. Approximately 12,900 new cases of NPC occur each year, with more than 70% of cases occurring in the east and southeast Asia (1,2). Differently from ordinary head and neck squamous cell carcinoma, NPC is highly sensitive to radiotherapy (RT) due to its particular biological properties. With the development of RT technology, the 3-year local control rate and survival rates of non-metastatic NPC reached 80-90% in the intensitymodulated RT (IMRT) era (3). Notably, among new cases of NPC, 6-15% of patients are diagnosed with distant metastatic NPC (de novo mNPC, dmNPC) (4,5). Whether these patients should receive locoregional RT (LRRT) needs to be clari ed.
According to a previous study, dmNPC patients that underwent LRRT plus palliative chemotherapy (PCT) achieved greater overall survival (OS) than those that received PCT alone. However, it was not clear whether all patients bene ted from the primary tumor treatment (6). In 2020, the National Comprehensive Cancer Network (NCCN) guidelines recommended LRRT alone following systemic chemotherapy for patients with oligometastatic disease (7). Nevertheless, the numbers of metastatic sites and organs that should be considered as "localized" or "widespread" have not been reported. Additionally, the aforementioned study did not analyze pretreatment plasma EBV DNA concentration, which is regarded as a prognosis indicator for NPC, to select candidates for LRRT (6).
Therefore, we conducted this study to investigate prognostic factors for dmNPC patients and to identify patients who achieved improved OS after LRRT by taking the prognostic factors into consideration. With our results, we aim to provide important information for the individualized treatment of these patients.

Methods
All methods were carried out in accordance with relevant guidelines and regulations as the National Comprehensive Cancer Network (NCCN) guidelines and eighth edition of the American Joint Committee on Cancer staging system. All patients in our study were staged by eighth edition of the American Joint Committee on Cancer staging system and all patients were treated according to the NCCN guidelines.

Patients
In total, 11,235 patients were newly diagnosed with NPC in the Sun Yat-sen University Cancer Center in China from November 2006 to October 2016. Of these, 498 patients were enrolled in this retrospective study ( Fig. 1) according to the following inclusion criteria: (I) dmNPC patients diagnosed histologically or radiologically; (II) age between 18 and 70 years; (III) no history of malignancy or synchronous cancer; (IV) treatment with cisplatin-based chemotherapy regimen; (V) normal hematopoietic function: white blood cell count ≥ 4×10 9 /L, platelets ≥ 100×10 9 /L, hemoglobin ≥ 90 g/L, and neutrophil granulocytes > 2.0×10/L; (VI) normal liver function test: aspartate aminotransferase and alanine aminotransferase < 2.5-fold of upper limit of normal (ULN), and total bilirubin < 2.0×ULN; (VII) normal renal function test: creatinine clearance ≥ 60 ml/min or creatinine ≤ 1.5×ULN; and (VIII) male or non-pregnant female. All patients were restaged based on the 8th edition of the American Joint Committee on Cancer/International Union Against Cancer staging system. This study was approved by the clinical research ethics committee of Sun-Yat sen university cancer center, and written informed consent was obtained from each patient.
During enrollment, general evaluation tests mainly included physical examination of the head and neck region (including nasopharynx and cervical lymph nodes), physical examination of the nervous system, EBV serologic tests, EBV DNA quantitative determination, nasal endoscopy, lesion biopsy, head and neck magnetic resonance imaging (MRI) scan, emission compared tomography (ECT) scan, and chest and abdominal CT scan. Positron emission tomography (PET-CT) was considered an optional evaluation test based on the patient's nancial burden.
We evaluated metastatic lesions and organs based on radiological criteria. The number of metastatic bone lesions and metastatic lesions (excluding bones) were assessed by ECT (or PET-CT) and chest and abdominal CT (or PET-CT) scans, respectively. The number of metastatic organs was evaluated by both ECT and chest and abdominal CT (or PET-CT) scans. included EBV DNA copy detection, nasopharyngoscopy, head and neck MRI scan, chest and abdominal CT scan, and ECT or PET/CT scans. The primary endpoint of this study was OS, which was measured from initial diagnosis to death from any cause or loss to follow-up.

Statistical analysis
The clinical characteristics of patients from different treatment groups were compared using the Pearson χ 2 test. The number of metastatic lesions and EBV DNA copies were transformed into dichotomous variables based on cutoff values de ned by the receiveroperating characteristic (ROC) analysis. The survival outcomes of patients from different subgroups were analyzed using Kaplan-Meier curves and the log-rank test. The independent prognosis predictors were evaluated using the Cox proportional hazards regression model. All data analyses were performed using the Statistical Package for Social Sciences (SPSS for macOS, version 21.0, IBM Corp., Armonk, NY). A two-tailed p < 0.05 was considered statistically signi cant.
Regarding metastatic lesions, 338 (67.9%) and 160 (32.1%) patients had ve or less and more than ve lesions, respectively. Based on ROC analysis, the EBV DNA copies cutoff value was set at 25,000 copies/ml and 284 (57.0%) patients had levels that surpassed this value. As shown in Table 1, we found statistical differences in the number of metastatic organs, number of metastatic lesions, and pretreatment EBV DNA copies between the different treatment groups. 2. Analysis of clinical characteristics' in uences on the prognosis of dmNPC patients All factors that may in uence prognosis were included in the Cox proportional hazards regression model. As shown in The Kaplan-Meier survival curves also showed an association between RT and improved OS (3-year OS, 27% vs. 13%; p < 0.001) ( Fig. 2A). As expected, patients with the aforementioned risk factors had shorter OS than other patients (p < 0.001 for all) (Fig. 2B-D). According to the risk factors de ned in the previous subsection, we divided patients into eight subgroups: group A, single organ metastasis, EBV DNA concentration ≤ 25,000 copies/ml, and 5 or fewer metastatic lesions; group B, single organ metastasis, EBV DNA concentration > 25,000 copies/ml, and 5 or fewer metastatic lesions; group C, multiple organs metastasis, EBV DNA concentration ≤ 25,000 copies/ml, and 5 or fewer metastatic lesions; group D, multiple organs metastasis, EBV DNA concentration > 25,000 copies/ml, and 5 or fewer metastatic lesions; group E, single organ metastasis, EBV DNA concentration ≤ 25,000 copies/ml, and more than 5 metastatic lesions; group F, single organ metastasis, EBV DNA concentration > 25,000 copies/ml, and more than 5 metastatic lesions; group G, multiple organs metastasis, EBV DNA concentration ≤ 25,000 copies/ml, and more than 5 metastatic lesions; and group H, multiple organs metastasis, EBV DNA concentration > 25,000 copies/ml, and more than metastatic lesions.
The Kaplan-Meier survival curves showed that patients in groups C-H had shorter OS than those in groups A-B; moreover, the OS of group A was signi cantly longer than that of group B (p < 0.05 for all). However, further paired comparisons revealed no signi cant differences in OS among groups C-H (p > 0.05 for all) (Fig. 3A). Subsequently, we classi ed group A as a low-risk subgroup (single organ metastasis, EBV DNA concentration ≤ 25,000 copies/ml, and 5 or fewer metastatic lesions), group B as an intermediate-risk subgroup (single organ metastasis, EBV DNA concentration > 25,000 copies/ml, and 5 or fewer metastatic lesions), and groups C-H as a high-risk subgroup (multiple organs metastasis or more than 5 metastatic lesions or both). The survival curves of patients in different risk strata are displayed in Fig. 3B. According to the Pearson χ 2 test, the subgroups only differed in chemotherapy regimens (p < 0.001, p = 0.004 in low-risk and high-risk subgroups respectively) ( Table 3) and no signi cant difference was found in other clinical characteristics. We further investigated the differences in OS between patients that did and did not receive RT in each classi cation of risk. Interestingly, we found that not all patients bene ted from RT. We found statistical differences in OS among patients in the low-risk and intermediaterisk subgroups (p = 0.039 and p = 0.010, respectively), whereas no signi cant difference was found in the high-risk subgroup (p = 0.076) (Fig. 4). Subsequently, we performed the Cox proportional hazards regression model for all subgroups (

Discussion
In this study, we strati ed dmNPC patients into different risk levels based on the number of metastatic lesions, number of metastatic organs, and level of pretreatment EBV DNA. In exploring the role of LRRT, we found that patients with a single metastatic organ and no more than ve metastases bene ted the most from the therapy, and among which, the patients with EBV DNA concentration ≤ 25,000 copies/ml have better OS than those with EBV DNA concentration > 25,000 copies/ml, which provides important information for individual treatment management in clinical practice.
Distant metastasis has become the main cause of death for NPC patients (3,10). Among all patients with distant metastasis, some had it detected at initial admission, which is de ned as dmNPC (5). Unlike patients with metastasis after treatment, these patients had no previous RT for the primary tumor. Whether the use of LRRT is necessary has become a concern of clinicians. Recently, two studies have shown that the addition of LRRT to PCT is associated with a longer survival time for dmNPC (6,11). However, another question needs to be clari ed: can all these patients bene t from RT? By analyzing the role of LRRT in dmNPC, You et al. found that patients with liver metastasis did not bene t from the primary tumor treatment while patients with other metastasis did (12). But the M1 stage subdivisions of the study did not take tumor burden and pre-treatment EBV DNA copies into considerations. Similarly, NCCN guidelines recommend chemotherapy combined with LRRT only for patients with limited metastasis sites or a low tumor burden, but the standard for "low tumor burden" of dmNPC is not de ned (7).
In this study, we con rmed that local RT was a protective factor for dmNPC patients. Consistently with previous studies, more than ve metastases and multiple metastatic organs were identi ed as independent risk factors (13,14). The cutoff value for the number of metastases was based on the de nition of 'oligo metastasis' used in clinical trials (15). Moreover, it has been demonstrated that pretreatment EBV DNA is closely associated with prognosis in locally advanced NPC (16,17). Similar to non-metastatic patients, our results suggested that high levels of EBV DNA may also be associated with worse prognosis in dmNPC patients. Based on the three identi ed prognostic factors, we divided all patients into eight groups then further classi ed these groups into three risk levels according to the statistical differences in survival among the eight groups. Due to limitations of the TNM staging system for metastatic patients, the prognosis of patients in the M1 stage could not be further classi ed. Based on our results, we strati ed these patients into different risk levels to facilitate a general prognosis assessment according to their baseline data.
More importantly, our team was the rst to identify the optimal candidates for LRRT based on a biomarker and tumor burden. In exploring the role of LRRT, we surprisingly found that only low-and moderate-risk patients (single metastatic organ and no more than ve metastases) bene ted from primary tumor treatment. This phenomenon may be explained by the following: as the illness of patients with oligo metastases is more likely to be controlled by PCT, they should be responsive to LRRT, which is a good way to prevent further disease progression from the primary tumor. However, primary RT may not improve the survival of patients with more than ve metastatic lesions or multiple metastatic organs. Considering the high cost and serious treatment-related toxicity, the administration of LRRT should be treated with caution. The main treatment goals for high-risk patients should include long-term survival with tumor and improvement of life quality (18). Therefore, systemic chemotherapy and symptomatic treatment may be preferred treatment strategies for these patients.
The continuous progress of medical science has improved the OS of dmNPC patients remarkably. However, as shown in this study, the survival condition of high-risk patients is still unsatisfactory, with a 3-year survival rate of 35%. Therefore, new therapeutic methods need to be developed, such as epidermal growth factor receptor (EGFR) targeted drugs. Unfortunately, although EGFR overexpression was detected in NPC, a retrospective study showed that the use of anti-EGFR drugs did not further improve the survival of dmNPC patients (19,20). Immunotherapy, represented by PD-1 antibodies treatment, is another recent promising research path (21). To explore the treatment e cacy of PD-1 antibody in metastatic NPC, our group launched a global multicenter, double-blind, randomized controlled phase III clinical trial. At present, patients' enrollment has been completed and we are looking forward to the results of long-term followup.
Our study has the following limitations: our study design was retrospective and selective bias was unavoidable. Furthermore, we conducted a single center study and most cases came from epidemic areas. Therefore, our conclusions need to be con rmed by multicenter prospective clinical trials.

Declarations
All methods were carried out in accordance with relevant guidelines and regulations as the National Comprehensive Cancer Network (NCCN) guidelines and eighth edition of the American Joint Committee on Cancer staging system.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. If anyone wants to request the data from the study, please contact maihq@sysucc.org.cn.