Circulating Plasma Epstein-Barr Virus DNA Load During the Follow-up Periods Predicts Recurrence and Metastasis in Nasopharyngeal Carcinoma

Background Epstein-Barr virus DNA (EBV DNA) load has been identified as a prognostic factor in nasopharyngeal carcinoma (NPC), whereas the dynamic changes in the long period have not been explored. In this study, we evaluated EBV DNA kinetics and its role in the survival. Methods We conducted a retrospective review of 900 NPC patients. Plasma EBV DNA levels were measured at various time points after treatment. The correlations of EBV kinetics with recurrence and metastasis were analyzed. After stratifying patients according to the EBV results, survival was compared using Kaplan-Meier estimates. Twelve- and 24-month landmark analyses for overall survival (OS) data were performed according to the EBV groups. Results Patients with post-EBV of less than 2500 copies/mL achieved better survival than did those with higher ones. Furthermore, patients with continuously elevated EBV DNA expressed significantly poorer OS (hazard ratio [HR], 2.542, 95% confidence interval [CI], 2.077–3.111; P < 0.001), distant metastasis-free survival (HR, 2.970; 95% CI, 2.392–3.687; P < 0.001), locoregional-free survival (HR, 1.699; 95% CI, 1.072–2.692; P = 0.013), and progression-free survival (HR, 2.535; 95% CI, 1.987–3.233; P < 0.001) than did patients with continuously normal EBV or those with elevated levels at any time point. The 5-year OS with elevated EBV was lower than that of the remission group by using the 12- and 24-month landmark analysis. Conclusions Elevated EBV DNA after treatment was a better predictive indicator of survival than the baseline concentrations. Furthermore, continuously elevated EBV DNA after treatment indicated recurrence, metastasis, and unfavorable prognosis for NPC. In addition, there were consistent patterns of EBV DNA kinetics during long-term follow-up, which warrant further study.

T he incidence of nasopharyngeal carcinoma (NPC) has remained consistently high in endemic regions such as Southern China and Southeast Asia, with a peak incidence of 50 cases per 100,000 individuals. 1,2 According to the National Comprehensive Cancer Network guidelines, 3 radiotherapy is the primary treatment for NPC, combined with different chemotherapeutic strategies according to the disease staging. Concurrent chemoradiotherapy is the standard option for locoregionally advanced NPC. With the improved local control and survival resulting from more precise imaging and radiotherapy, distant metastasis has become the main cause of failure of this mode of treatment. 4 Positive rescue treatment is necessary for locoregional recurrence and distant metastasis of NPC and can achieve some level of curative success. In Sun Yat-Sen University Cancer Center (SYSUCC), treatment of 132 patients with recurrent NPC with intensity-modulated radiotherapy (IMRT) resulted in a 3-year local progression-free survival (PFS) and overall survival (OS) rates of 85.3% and 41.6%, respectively. 5 Cervical residual lesions can be treated effectively by surgical resection. Improvements in chemotherapy and the application of targeted therapy for distant metastasis of NPC have greatly improved the efficacy of rescue treatment and further improved patients' survival. 6 Early diagnosis of recurrence and metastasis is the key to improve the efficacy of rescue treatment. 7 For all patients, regular review is conducted every 4 to 6 months after 3 to 5 years of treatment, and imaging is the main monitoring method. The scattered nature of recurrent tumor cells after radiotherapy renders biopsy unreliable for diagnosis, leading to a delay in rescue treatment of more than 10 weeks. 8 Fludeoxyglucose-positron emission tomography provides higher sensitivity and specificity for the diagnosis of recurrence and metastasis than does computed tomography (CT) or magnetic resonance imaging, although the examination is expensive. 9,10 Therefore, a sensitive, cost-effective, and practical diagnostic index is required.
We have previously reported the potential benefits of adjuvant treatment for patients with detectable Epstein-Barr virus DNA (EBV DNA) load at the end of treatment, 11 in addition to the prognostic value of EBV DNA concentrations during treatment. 12 Tumor response and patient survival can also be predicted based on the clearance rates of plasma EBV DNA during the first month of chemotherapy. Therefore, early changes in the chemotherapy regimen may be considered for patients with a slow plasma EBV DNA clearance rate. 13  clinical remission ( P < 0.01), EBV DNA is implicated as a sensitive tumor marker for monitoring tumor recurrence and metastasis after radiotherapy in NPC. [14][15][16] The mechanism underlying the role of EBV DNA in NPC remains unclear, and no risk classification has been effectively demonstrated; however, previous studies have focused on the dynamic changes in viral load during the follow-up. Thus, in this study, we hypothesized that the dynamic changes in plasma EBV DNA after treatment can be used to predict recurrence or metastasis of NPC. To better assess the correlation of survival with EBV DNA during the follow-up, we performed 12-and 24-month landmark analyses.

Patients
From January 2010 to September 2013, 900 consecutive patients who had confirmed diagnosis of nonmetastatic NPC and received radical radiotherapy in SYSUCC were enrolled for initial screening. The selection criteria were as follows: (1) good performance status (Karnofsky Performance Status score ≥80); (2) normal renal, cardiac, and liver function; (3) complete EBV DNA record during treatment and 3 years after treatment; and (4) complete follow-up data.

Treatments
The therapeutic protocol was based strictly on the National Comprehensive Cancer Network guidelines. All patients received the following treatment: stage I, IMRT alone; stage II, IMRT with concurrent chemotherapy; and stages III and IVA-B, concurrent chemoradiotherapy with or without neoadjuvant or adjuvant chemotherapy. All patients underwent targeted radiotherapy using a thermoplastic mask based on CT simulation. Radiation Therapy Oncology Group guidelines were applied for target delineation. Planning target volume (PTV) was defined as the primary tumor (PTVp) and positive lymph node (PTVn), high-risk PTV (PTV-HR) was defined as the PTVp with an appropriate margin, and lowrisk PTV (PTV-LR) was defined as PTV-HR with an appropriate margin plus the elective lymph node region. The treatment consisted of definitive radiotherapy with a total dose of 68 to 72 Gy to PTVp, 60 to 68 Gy to PTVn, 60 to 62 Gy to PTV-HR, and 50 to 54 Gy to PTV-LR. Neoadjuvant or adjuvant chemotherapy consisted of cisplatin with 5-fluorouracil or cisplatin with taxanes, or all 3 together, administered triweekly for 2 or 3 cycles. Concurrent chemotherapy consisted of cisplatin/nedaplatin administered triweekly or weekly until the end of radiotherapy.

Quantification of Plasma EBV DNA
Samples of peripheral venous blood (5 mL) were obtained from all patients. For isolation of plasma and peripheral blood cells, the blood samples were placed in EDTA-containing tubes and centrifuged at 1600g for 15 minutes. Plasma DNA was extracted using a QIAamp Blood Kit (Qiagen, Hilden, Germany). Epstein-Barr virus DNA concentrations were detected using the BamHI-W region of the EBV genome and a real-time polymerase chain reaction assay according to a previously described method. 17 The sequences of the forward and reverse primers used in the assay were as follows: 5′-GCCAG AGGTA AGTGG ACTTT-3′ and 5′-TACCA CCTCC TCTTC TTGCT-3′. The dual fluorescently-labeled oligomer of 5′-(FAM) CACAC CCAGG CACAC ACTAC ACAT (TAMRA)-3′ served as the probe.

Follow-up
The endpoints for this study were OS, measured from the date of first NPC diagnosis to either the date of death or loss to follow-up, distant metastasis-free survival (DMFS, to distant relapse or patient censorship or death at last follow-up), locoregional-free survival (LRFS, to tumor relapse or patient censorship or death at last follow-up), and PFS (to locoregional failure, distant failure, or death from any cause, whichever occurred first). After treatment, patients were scheduled for regular follow-up visits every 3 months during the first 2 years and then every 6 months until the fifth year and yearly thereafter. Evaluations during each outpatient visit included physical examination, hematology and biochemistry profiles, magnetic resonance imaging scan, chest radiography, abdominal sonography, and a whole-body bone scan. The surveillance workup included EBV DNA load.

Statistical Analysis
Cutoff values for pretreatment plasma EBV DNA load was established using receiver operating characteristic analysis. A χ 2 test was used to determine the correlation between plasma EBV DNA and treatment outcomes. A univariate logistic regression model was used to examine the association between clinical factors, EBV status, and survival outcomes. Factors with a P ≤ 0.05 in the univariate analysis were entered into the multivariate Cox regression model. Survival outcomes were estimated using the Kaplan-Meier method and compared with the log-rank test. All tests were 2-sided, and a P < 0.05 was considered statistically significant. Statistical analysis was performed using IBM SPSS software version 22.0 (IBM Corp., Armonk, NY).
The inherent statistical weakness is guarantee-time bias. As recommended by statisticians, in this study, we investigated EBV NDA as a predictor of survival by performing time-dependent response analyses using landmark and risk-of-death methods, thereby minimizing guarantee-time bias.

ETHICS STATEMENT
The study was approved by the ethics committee of SYSUCC, and the informed consent for collection and publication of medical information was obtained from all study patients at their first visit to our hospital.

Dynamic Change Patterns of EBV DNA
A total of 900 NPC patients treated at SYSUCC between January 2010 and September 2013 were retrospectively enrolled. Median age was 44 years (range, 11-77 years) and 723 of 900 patients (80.3%) were male. All patients had nonkeratinizing undifferentiated squamous cell carcinoma. According to the clinical examination during the follow-up (median, 55.78 months), there were 21 cases of nasopharyngeal locoregional recurrence, and 97 cases developed metastasis (25 cases of bone metastasis, 31 lung, 26 liver, 2 lymph node, 2 bone and lung, 2 liver and lung metastasis, 2 bone and liver, and 7 multiple sites). The baseline clinical features of the patients are shown in Table 1.
Patients were then divided into 3 groups according to the follow-up results. Sustained remission consisted of 778 patients without recurrence or metastasis at the end of the follow-up, locoregional recurrence consisted of 21 patients with recurrence at any time point, and metastasis consisted of 97 patients with metastasis at any time point.
We then evaluated the dynamic changes in EBV DNA load at each time point. At the end of therapy, the EBV DNA load decreased rapidly in the sustained remission and metastasis groups (median, 0 copies/mL). Subsequently, there was no marked change of the EBV DNA load in sustained remission group (<4000 copies/mL), although the EBV DNA load in the metastasis group increased significantly ( P = 0.025), reaching a peak at 3 months followed by a decrease to 1 year, and no marked change to 3 years (<5000 copies/mL). The viral load in the recurrence group had increased rapidly 3 months after treatment and was lower than that in the metastasis group 1 year after treatment. This was followed by a gradual increase after 1 year, with a higher peak viral load in the recurrence group than that in the metastasis group, reaching 26,000 copies/mL at 12 months, 168,000 copies/mL at 24 months, and 375,000 copies/mL at 36 months, until the highest level at the time of relapse was 24,100,000 copies/mL (Fig. 1).
A total of 473 patients had pretreatment EBV DNA loads below the cutoff value of 4000 copies/mL. Among them, 21 patients had posttreatment EBV DNA loads greater than 2500 copies/mL, there were 18 cases with residual tumor, 32 developed metastasis, and 45 died.
A total of 427 patients had pretreatment EBV DNA loads greater than 4000 copies/mL. Among them, 369 patients had posttreatment EBV DNA loads of less than 2500 copies/mL, 46 patients developed metastasis, eight had recurrence, and 41 died. Among the 58 of 427 patients with posttreatment EBV DNA greater than 2500 copies/mL, 19 patients developed metastasis, 4 locoregional recurrence, and 22 died, all with a higher proportion than the lower EBV ones. Thus, these data indicate that posttreatment EBV DNA load had a much more marked influence on the prognosis of patients with NPC than pretreatment EBV DNA load.
The patients were then divided into 2 groups according to the posttreatment EBV DNA cutoff values. The prognosis in the high posttreatment EBV DNA group was poorer than that of the lower EBV DNA group. Patients with posttreatment EBV DNA of less than 2500 copies/mL achieved better OS ( P < 0.001), DMFS ( P = 0.040), LRFS ( P = 0.004), and PFS ( P = 0.001) than did patients with higher posttreatment EBV DNA.
Other risk factors, including smoking, advanced tumor stage, and elevated C-reactive protein before treatment and therapy, were correlated with poor OS, DMFS, and PFS. Multivariate analyses were also performed, incorporating patient factors (age, sex, smoking), tumor factors (TNM stage), treatment methods (radiotherapy and chemotherapy), and posttreatment EBV DNA load as covariates. As shown in Table 2, posttreatment EBV DNA load was confirmed as an independent prognostic factor for OS (

Landmark Analyses
Of 900 patients in the whole group, a total of 887 patients were alive and had follow-up for at least 12 months, and therefore, they were included in the 12-month landmark analysis for survival (Fig. 5A). The 5-year OS of the 746 patients whose EBV DNA load sustained remission was 91.3%, and the elevated group was 74.4% (HR, 3.182; 95% CI, 2.114-4.788; P < 0.001).
Besides, a total of 854 patients were alive or had follow-up for at least 24 months and therefore were included in the 24-month landmark analysis for survival (Fig. 5B). The 5-year OS of the 732 patients whose EBV DNA load sustained remission was 95.0%, and the elevated group was 62.9% (HR, 9.619; 95% CI, 6.027-15.353; P < 0.001).

DISCUSSION
Many nonspecific markers have been demonstrated to play important roles in predicting tumor progression and prognosis. Epstein-Barr virus DNA load is the most representative marker in NPC 18 and is widely used in screening, monitoring, and prediction of relapse in nonmetastatic NPC. 19 Circulating EBV DNA concentrations correlate positively with disease stage, as well as exhibit prognostic importance in NPC. 12 Based on the great variety of published studies, 13,20,21 EBV DNA concentrations during treatment and plasma EBV DNA clearance rates have been identified as emerging biomarkers of survival. 21,22 Furthermore, some studies have stratified the role of EBV DNA in predicting metastasis  19 A prospective study demonstrated that EBV DNA load was a predictor of distant metastasis. 23 In patients with recurrent NPC requiring salvage nasopharyngectomy, preoperative plasma EBV DNA identifies patients at high risk of subsequent distant failure after surgery. Serial measurements of plasma EBV DNA after surgery, especially for those with high preoperative levels, are crucial to allow early detection of local distant failure. 24 In contrast, there were few reports describing the role of posttreatment EBV DNA in NPC, especially in-depth analysis of the dynamic changes and common variation tendencies. Li et al. 25 revealed that 6.5% of patients (19/292) with undetectable EBV  DNA and 57.0% of patients (53/93) with detectable EBV DNA during posttreatment follow-up experienced tumor recurrence. Plasma EBV DNA level during posttreatment follow-up is a good marker for predicting distant metastasis, but not locoregional recurrence.
The dynamic changes of EBV DNA in the recurrence and metastasis groups can be summarized by the common rule that the EBV DNA decreased at the end of treatment, approaching 0 copies/mL, although the low concentration was maintained for a short time (approximately 3 months). After 3 months of treatment, the EBV DNA increased gradually until the peak value was reached at the point of clinical diagnosis of metastasis or recurrence. In the recurrence group, EBV DNA increased slowly within 1 year after treatment and increased rapidly after 1 year, with no obvious peak reached. In the metastasis group, EBV DNA began to rise after treatment, which was earlier than that observed in the recurrence group. Although the reason why the patterns of EBV DNA dynamics were different was still undefined, we can make a hypothesis that the patients who suffered recurrence expressed the bigger primary tumor load, which may be more EBV-related than the distant metastatic sites.
The mechanisms underlying the impact of the dynamic changes in EBV DNA load on NPC recurrence and metastasis remain to be clarified, and no risk classification has been effectively demonstrated. Thus, in this study, we stratified the kinetic changes in EBV DNA loads at 0 month, 3 months, 1 year, 2 years, and 3 years after treatment and revealed that different combinations of posttreatment EBV DNA groups may better predict OS, DMFS, LRFS, and PFS in NPC patients. Furthermore, we characterized the patterns of dynamic changes in EBV DNA load in the sustained remission, recurrence, and metastasis groups.
Even if the pretreatment EBV DNA load is higher than the cutoff value of (4000 copies/mL, the prognosis remains good if the fluctuation range of EBV DNA in patients after treatment is lower than the cutoff value. However, the prognosis of patients with EBV DNA loads that continue to exceed the cutoff value after treatment is poor, and the probability of clinical events is high. In such cases, pretreatment EBV DNA load has no prognostic value. Thus, the influence of EBV DNA on survival after treatment is stronger than that before treatment.
In addition, there were significant differences in OS by 12and 24-month landmark analyses in patients with different EBV DNA load groups. Thus, analysis of this marker should be added to the imaging examination to provide a complete posttreatment follow-up.
Moreover, the present study has several limitations. First, biases due to the retrospective nature of the analyses are unavoidable when the follow-up time is inadequate, and it is a single-center study. Second, not all of the patients in that period have detected the EBV DNA completely and regularly, we have selected patients with complete data. Despite these limitations, the highlights of this study are that it is a large-scale study that evaluated the dynamics of EBV DNA during follow-up and its value for predicting survival. In summary, future prospective, multicenter clinical studies should be warranted to verify the results of this current study.

CONCLUSIONS
Epstein-Barr virus DNA kinetics reflects disease progression. Our results showed that in patients with NPC, elevated EBV DNA during follow-up can reflect the possibility of recurrence and metastasis, as well as the unfavorable prognosis. Also, there were significant differences in OS by landmark analysis. These findings indirectly verify the significance of EBV DNA during the follow-up time in NPC. We believe that a better understanding of EBV DNA kinetics would be of paramount importance for the rational use of this new class of tumor marker. Thus, we can use EBV DNA load to intensify the examination or treatment regimens accordingly.