Dynamic Changes of Post-Radiotherapy Plasma Epstein–Barr Virus DNA in a Randomized Trial of Adjuvant Chemotherapy Versus Observation in Nasopharyngeal Cancer

After radiotherapy or chemoradiation, about 10%–30% of patients with nasopharyngeal cancer (NPC) will still harbor residual plasma Epstein–Barr virus (pEBV) DNA in the circulation (1–3). These patients were reported to be at higher risk of treatment failure on long-term follow-up (4). The absolute level of post-radiotherapy pEBV DNA correlated significantly with the risk of locoregional failure, distant metastasis, and death. For each 10-log increase in pEBV DNA concentration, the HR was 1.96 [95% confidence interval (CI), 1.77–2.16] for locoregional failure, 2.14 (95% CI, 1.93–2.39) for distant metastasis, and 2.12 (95% CI, 1.89–2.37) for death (5). Post-radiotherapy pEBV DNA has been used to stratify patients and guide adjuvant treatment in completed (NCT00370890) or ongoing clinical trials of adjuvant therapy in NPC (NCT02135042, NCT02363400, NCT02874651, and NCT03544099; refs. 6, 7).

It was well described that the post-radiotherapy elevated pEBV DNA could be transient with spontaneous loss of detectable pEBV DNA in the blood during follow-up, but it remains uncertain whether those patients with subsequent clearance of pEBV DNA are still at higher risk of disease progression (8, 9). In a large-scale big-data intelligence platform-based analysis of 3,269 patients with NPC who had pEBV DNA measured at the end of therapy (±1 week), 238 (7.0%) patients had detectable pEBV DNA at the end of therapy. Of these 238 patients with residual pEBV DNA, 192 had pEBV DNA measured 3 months after and spontaneous clearance of pEBV DNA occurred in 72.4% (139/192) patients. However, these patients still had a poor 3-year disease-free survival (55.1% vs. 89.8%) and overall survival (OS; 79.1% vs. 96.2%) than patients with initially undetectable pEBV DNA at end of therapy (all P < 0.001), and patients with persistent detectable post-therapy pEBV DNA had the worst outcome (10). In another institutional big-data research platform study of 1,984 patients with NPC, blood samples were collected within 3 months post-radiotherapy and every 3–12 months thereafter. During follow-up, a total of 767 patients (38.7%) had detectable pEBV DNA. The recurrence rate among these patients were 63.8% (489/767 patients), which was significantly higher than that in patients with undetectable pEBV DNA (8.6%; 105/1,217 patients). In addition, of the 278 pEBV DNA-positive patients who did not develop disease recurrence, 227 (81.7%) had transiently positive pEBV DNA that fell to undetectable levels during long-term monitoring (11).

However, retrospective studies were often limited by case selection bias because patients who had clinical suspicion were more likely to have undergone pEBV DNA surveillance during follow-up (12). In several prospective studies, pEBV DNA clearance was established as a significant predictor of favorable treatment response to chemotherapy and survival in both advanced, recurrent, or metastatic NPC (13–16).

In the multi-center 0502 pEBV DNA screening trial, we prospectively enrolled 789 patients with NPC and randomly assigned patients with detectable post-radiotherapy pEBV DNA to adjuvant chemotherapy versus clinical observation. We have previously reported that adjuvant chemotherapy with cisplatin and gemcitabine did not improve the primary endpoint of progression-free survival (PFS) in patients with high-risk NPC identified by detectable post-radiotherapy pEBV DNA (5). In this preplanned biomarker analysis of the randomized controlled trial, we assessed the performance of pEBV DNA for predicting disease progression and evaluated pEBV DNA clearance as a potential surrogate endpoint of long-term survival and therapeutic efficacy of adjuvant chemotherapy.

We have previously reported on the primary analysis of the prospective pEBV DNA screening conducted at six public oncology centers in Hong Kong (5, 17). This analysis focused on the dynamic changes of plasma EBV DNA in the companion biomarker study. Participants enrolled in the screening trial had histologic diagnosis of NPC, Union for International Cancer Control (UICC; 6th edition) stage IIB to IVB, no clinical evidence of residual locoregional disease or distant metastasis after completion of primary radiotherapy/chemoradiation with or without preceding neoadjuvant chemotherapy, age ≥ 18 years old, Eastern Collaborative Oncology Group (ECOG) performance status of 0 to 1, adequate bone marrow and organ function. Exclusion criteria included second primary cancer, more than 12 weeks after completion of radiotherapy, and peripheral neuropathy or ototoxicity above grade 2.

Eligible patients were consented for pEBV DNA screening at 6 weeks post-radiotherapy. Patients with undetectable pEBV DNA (defined as below the detection limit or 0 copy/mL, see below) were categorized as low-risk group and they underwent clinical observation and routine follow-up. Patients with detectable pEBV DNA (defined as any detectable signal by real-time PCR, i.e., >0 copy/mL) were categorized as high-risk group and were offered a second consent for randomization to adjuvant chemotherapy versus clinical observation. After restaging, patients who fulfilled all eligibility criteria were randomly assigned in 1:1 to adjuvant chemotherapy or clinical observation.

This study was conducted in compliance with the principles of the Declaration of Helsinki and International Conference on Harmonization of Good Clinical Practice (ClinicalTrials.gov identifier: NCT00370890). Institutional Review Board at each participating center approved the study protocol. All patients provided written informed consent prior to study procedures.

We followed the REMARK (REporting recommendations for tumour MARKer prognostic studies) recommendation for analysis of pEBV DNA (18). We collected 10 mL peripheral blood into EDTA tubes, centrifuged within 6 hours to separate plasma which was transported to Department of Chemical Pathology at The Chinese University of Hong Kong (Shatin, Hong Kong). We quantified pEBV DNA by real-time PCR that targeted the BamHI-W fragment of the EBV genome as reported previously (5, 19, 20). To determine the ability of the assay for detecting low levels of EBV DNA, we analyzed 20 replicates of negative control samples with no template input. All the negative samples did not show any detectable signal. Hence, any detectable signal in a test plasma sample was defined as positive for plasma EBV DNA. Through analysis of replicates with low EBV DNA concentrations, we further showed that the assay could consistently detect plasma EBV DNA at a concentration of two EBV genomes per PCR reaction which is equivalent to 20 copies/mL plasma. This sensitive detection was achieved through targeting the repeat elements BamHI-W fragment of the EBV genome. We defined post-radiotherapy pEBV DNA clearance as undetectable second post-radiotherapy pEBV DNA (= 0 copy/mL) in patients who had detectable first post-radiotherapy pEBV DNA (>0 copy/mL).

Patients with detectable pEBV DNA at 6 weeks post-radiotherapy and no evidence of locoregional disease or distant metastasis by conventional staging procedures (including nasopharyngeal endoscopy and biopsy, chest radiograph, ultrasound or CT scan of abdomen, and bone scan) were eligible for randomization to adjuvant chemotherapy versus clinical observation. Patients in both study arms had pEBV DNA checked at 6 weeks post-radiotherapy and then at the 6 months post-randomization. The result of the first post-radiotherapy pEBV DNA was reported as negative (undetectable) or positive (detectable) for the purpose of randomization. Patients randomized to adjuvant chemotherapy had a baseline whole-body ¹⁸FDG-PET CT scan (PET-CT; refs. 14, 21) performed after registration and before the start of adjuvant chemotherapy, and then at 6 months post-randomization. Patients randomized to clinical observation had the baseline PET-CT performed after registration and then at 6 months post-randomization. The value of performing a PET-CT in patients with NPC after completion of primary radiotherapy was undefined when the study was opened in 2006. Therefore, the baseline PET-CT was blinded to both investigators and patients; it was only reported when patients completed the second PET-CT. This blinding was part of study design and was included in the patient's consent. In the current analysis, lesions on the second PET-CT were considered positive when the SUV_max was >2.5 in the local or nodal bed or at distant sites without evidence of infection.

The primary endpoint of the randomized clinical trial was PFS, defined as the duration from the date of randomization to the date of first NPC progression (local, regional, or distant) or death from any cause, which ever occurred first, or censored at the date of last follow-up. Secondary endpoints included OS, local regional failure-free survival (LRFS), distant metastasis-free survival (DMFS), and correlation of post-radiotherapy pEBV DNA and PET-CT with clinical outcome. OS was defined as the duration from the date of randomization to the date of death from any cause or censored at the date of last follow-up. LRFS was defined as the duration from the date of randomization to the date of first local or regional (neck) recurrence or death from any cause or censored at the date of last follow-up. DMFS was defined as the duration from the date of randomization to the date of first distant metastasis or death from any cause or censored at the date of last follow-up. In the analysis of patient population that underwent pEBV DNA screening with undetectable first pEBV DNA, and patients with detectable first pEBV DNA but not eligible for randomization, the time origin of all endpoints was calculated from the date of consent for first pEBV DNA.

Correlations of first and second post-radiotherapy pEBV DNA and clinical outcome were analyzed by Spearman correlation, Mann–Whitney U test, or χ² test, as appropriate. All time-to-event data were analyzed by Kaplan–Meier method, log-rank statistics, and Cox proportional hazard model. Age, sex, UICC T, N and overall stage, primary treatment (radiotherapy/chemoradiation), adjuvant chemotherapy (yes/no), and pEBV DNA level were included in multivariate analyses. All statistical tests were two sided, and we considered P < 0.05 as significant. We used SAS version 9.4 and GraphPad Prism, RRID:SCR_002798, version 8.0 for statistical analyses.

From September 2006 to July 2015, we enrolled 789 patients with NPC for pEBV DNA screening at 6 weeks post-radiotherapy. Patients were categorized into three groups according to their first post-radiotherapy pEBV DNA level (Fig. 1). Group A included 573 patients (72.6%) whose first post-radiotherapy pEBV DNA was undetectable. Group B included 112 patients (14.2%) whose first post-radiotherapy pEBV DNA was detectable, but they were excluded from randomization for reasons listed in Fig. 1. Group C included 104 patients whose first post-radiotherapy pEBV DNA was detectable and they were eligible for randomization.

All patients with undetectable first post-radiotherapy pEBV DNA (group A) underwent clinical observation and routine follow-up. Patients in group B who had residual/metastatic disease detected at restaging received standard of care at discretion of their attending oncologist; otherwise, they underwent clinical observation and routine follow-up. All 104 patients in group C were randomly allocated to receive adjuvant chemotherapy (group C1, n = 52) or clinical observation (group C2, n = 52). At 6 months post-randomization, 98 patients (94%) completed second post-radiotherapy pEBV DNA test per study protocol (Fig. 1). The median duration of follow-up was 6.2, 6.7, and 6.6 years for patients in groups A, B, and C, respectively. The distribution of sites of progression and death was summarized in Supplementary Table S1.

The distribution of first and second post-radiotherapy pEBV DNA in randomized patients with NPC who completed second pEBV DNA was shown in the scatter plot (Supplementary Fig. S1A, n = 98). There was a weak correlation between the levels of first and second post-radiotherapy pEBV DNA (Spearman correlation coefficient 0.38, P = 0.0001). The levels of first and second pEBV DNA correlated significantly with distant metastasis (P = 0.003 and 0.01 for first and second post-radiotherapy pEBV DNA, respectively) but not with local or regional recurrence (Supplementary Fig. S1B and S1C).

Table 1 summarized the performance of first and second post-radiotherapy pEBV DNA (at cut-off value of >0 copy/mL) in predicting disease progression. The first post-radiotherapy pEBV DNA had a modest sensitivity of 0.48, specificity of 0.81, area under receiver-operator characteristics curve (AUC) of 0.65, false positive (FP) rate of 13.8%, and false negative (FN) rate of 14.4%. Compared with first post-radiotherapy pEBV DNA, the second post-radiotherapy pEBV DNA had improved sensitivity of 0.81, specificity of 0.75, AUC of 0.78, similar FP rate of 14.3%, and reduced FN rate of 8.1%.

In general, pEBV DNA has a higher negative predictive value (NPV, 0.80 and 0.84 for first and second post-radiotherapy pEBV DNA, respectively) than positive predictive value (PPV, 0.50 and 0.71 for first and second post-radiotherapy pEBV DNA, respectively). The performance of pEBV DNA was superior in predicting distant metastasis (AUC 0.67–0.81) than local (0.48–0.54) or regional progression (AUC 0.56–0.64). The improved AUC of second pEBV DNA was mainly due to improved prediction of distant metastasis, with no significant change in prediction of local or regional progression (Table 1).

The second PET-CT had a sensitivity of 30/41 (0.73), specificity of 20/56 (0.36), AUC of 0.54 for disease progression (Supplementary Table S2). The FP rate was high (37.1%) particularly for local progression (48.5%). Of 33 patients with a true positive second pEBV DNA and disease progression, 4 were FN on PET-CT. Of 14 patients with a FP second pEBV DNA without disease progression, 9 were also FP on PET-CT.

Of 42 patients with true negative second pEBV DNA without disease progression, 27 were FP on PET-CT. Of 8 patients with a FN second pEBV DNA and disease progression, only one had a true positive PET-CT, a patient with local relapse at 11.4 months. The other 7 were also FN on PET-CT: 2 diagnosed with relapse at 3 and 11.4 months and 5 after 1 year (12.6–55.0 months).

Patients can be stratified into three risk groups based on their first post-radiotherapy pEBV DNA level. Patients with undetectable pEBV DNA at 6 weeks post-radiotherapy (group A) had superior survival with 5-year rate of PFS 77.1%, OS 87.3%, LRFS 80.3%, and DMFS 83.1% (Table 2; Supplementary Fig. S2). Patients with detectable pEBV DNA at 6 weeks post-radiotherapy and eligible for randomization (group C) had intermediate survival with 5-year rate of PFS 52.1%, OS 66.0%, LRFS 56.8%, DMFS 61.4% (HR for PFS, OS, LRFS, DMFS ranged from 2.4–2.7, all P < 0.0001 when compared with group A). Patients with detectable pEBV DNA at 6 weeks post-radiotherapy but were ineligible for randomization (group B) had poor survival with 5-year PFS 38.9%, OS 54.8%, LRFS 46.3%, DMFS 47.4% (HR for PFS, OS, LRFS, and DMFS ranged from 3.8–4.0, all P < 0.0001 when compared with group A). The poor survival of patients in group B was mostly attributed to 54% (61/112) of group B patients already having residual disease (n = 36) or distant metastases (n = 25) detected at restaging (Fig. 1).

Among randomized patients who completed second post-radiotherapy pEBV DNA (n = 98), 50 patients (51.0%) had pEBV DNA clearance (undetectable second pEBV DNA), 13 patients (13.3%) had pEBV nonclearance but decrease from baseline (second pEBV DNA detectable but ≤ first pEBV DNA), and 35 patients (35.8%) had pEBV DNA nonclearance and increase from baseline (second pEBV DNA > first pEBV DNA). The median baseline pEBV DNA (1.45 log copies/mL, interquartile range 1.20–1.81) in the pEBV DNA clearance group was lower than the median value in the nonclearance group (1.80 log copies/mL, interquartile range 1.47–2.44; Fig. 2A). For patients with pEBV DNA clearance, 66% (33/50 patients) had first pEBV DNA < median compared with 34% (17/50 patients) had first pEBV DNA ≥ median (P = 0.0006 by χ² test). The likelihood of clearance of second pEBV DNA was related to the absolute level of the first pEBV DNA (OR 0.36; 95% CI, 0.18–0.73; P = 0.0045 by logistic regression).

The pattern of pEBV DNA clearance correlated with distinctive PFS (Fig. 2B). The 5-year of PFS was 85.5% for patients with pEBV clearance, 51.9% for patients with pEBV DNA nonclearance but decrease from baseline, and 14.3% for patients with pEBV DNA clearance and increase from baseline (log-rank P < 0.0001).

The clearance rate of pEBV DNA was 75% (42/56) in patients in clinical remission (without progression), which was significantly higher than 19% (8/42) in patients with progression (Fig. 2C and D, P < 0.0001 by χ² test). In contrast, there was no significant difference in pEBV DNA clearance rate between patients randomized to adjuvant chemotherapy (24/48, 50%) versus clinical observation (26/50, 52%), consistent with the lack of efficacy of adjuvant chemotherapy compared with clinical observation (Fig. 2E and F, P = 0.84), as reported previously (5).

Patients with pEBV DNA clearance had superior 5-year PFS of 85.5% when compared with 23.3% for patients with pEBV DNA nonclearance (HR, 9.6; 95% CI, 4.4–20.9; P < 0.0001). Similarly, the 5-year OS of 95.8% for patients with pEBV DNA clearance was superior when compared with 38.8% for patients with pEBV DNA nonclearance (HR, 15.7; 95% CI, 4.7–51.8; P < 0.0001). Superior survival outcome of LRFS and DMFS was also observed for patients who had pEBV DNA clearance compared with patients with pEBV DNA nonclearance (HR 6.7 for LRFS and 20.2 for DMFS, respectively, both P < 0.0001; Table 2; Fig. 3). We observed a similar magnitude of significant survival difference in all survival endpoints (PFS, OS, LRFS, and DMFS) in patients with pEBV DNA clearance compared with patients with pEBV DNA nonclearance in each of the two study arms regardless of randomization to either adjuvant chemotherapy or clinical observation (Table 2; Fig. 4). On the other hand, adjuvant chemotherapy did not have any impact on pEBV DNA clearance when compared with clinical observation (Figs. 2 and 4).

For patients who had detectable first post-radiotherapy pEBV DNA but subsequently experienced pEBV DNA clearance at 6 months post-randomization, their long-term survival [group C with pEBV DNA clearance, 5-year PFS of 85.5%, HR 0.61 (95% CI, 0.30–1.24) when compared with group A; 5-year OS of 95.8%, HR 0.31 (95% CI, 0.10–0.99) when compared with group A] was comparable with patients who had undetectable first post-radiotherapy pEBV DNA (group A, 5-year PFS of 77.1% and 5-year OS of 87.3%), regardless of receiving adjuvant chemotherapy or clinical observation (Table 2; Fig. 4).

Finally, we included all clinical parameters together with second post-radiotherapy pEBV DNA and second PET-CT in the analysis of prognostic factors for progression free of patients with NPC with detectable post-radiotherapy pEBV DNA and randomized to adjuvant chemotherapy or clinical observation (Supplementary Table S3). UICC T stage (HR, 2.15; P = 0.016), second post-radiotherapy pEBV DNA (HR, 9.44; P < 0.0001) were the significant prognostic factors in univariate analysis. Post-radiotherapy pEBV DNA clearance was the only significant prognostic factor for PFS in multivariate analysis.

To our knowledge, this is the first multi-center, prospective pEBV DNA screening study to identify patients with NPC with detectable post-radiotherapy pEBV DNA for randomization to adjuvant chemotherapy versus clinical observation. We previously reported that adjuvant chemotherapy with cisplatin-gemcitabine did not improve the survival outcome in these high-risk patients (5). Here, we reported the dynamic changes of post-radiotherapy pEBV DNA in the randomized patient cohorts.

Our findings showed that current pEBV DNA assay provided only modest sensitivity (48%) in the post-radiotherapy adjuvant setting, consistent with previous reports of 28%–46% in prospective NPC patient cohorts (8, 9). This is not surprising given that the concentration of pEBV DNA correlated strongly with tumor load (21, 22), which is expected to be low in patients with NPC after modern radiotherapy/chemoradiation. Serial monitoring of post-radiotherapy pEBV DNA improved sensitivity and reduced the rate of FN results. In predicting disease progression in NPC, pEBV DNA has higher NPV (0.80–0.84) than PPV (0.50–0.71) that undetectable post-radiotherapy pEBV DNA can help to select patients with a good prognosis. As the performance of post-radiotherapy pEBV DNA was superior for prediction of distant metastasis but suboptimal for prediction of local or regional recurrence, regular nasopharyngeal endoscopy remains essential in locoregional surveillance (9). In subjects with detectable pEBV DNA, we have previously demonstrated that MRI is complementary to nasopharyngeal endoscopy and can enable the earlier detection of endoscopically occult NPC (23).

In this study, a low SUV_max (>2.5) was chosen to indicate a positive result to maximize the sensitivity for identifying sites of relapse, resulting in a high FP rate for PET-CT (Supplementary Table S2). Despite using this low threshold, the sensitivity of the second PET-CT (0.73) was lower that of second pEBV DNA (0.81). Our findings are in keeping with a previous study in which patients with positive pEBV DNA underwent PET-CT (24). We also performed PET-CT in our patients who had a negative pEBV DNA. Only 16% (8/50) of these patients had disease relapse and except for one case, the PET-CT was negative. As most relapse arose after 1 year, it is possible that the disease was subclinical at the time of the investigations.

Recently, a risk adjusted surveillance strategy based on patient's tumor–node–metastasis (TNM) staging and pretreatment pEBV DNA level outperformed standard guideline strategies and maximized the early detection of NPC recurrence in a simulation study (25, 26). Our results suggested that plasma EBV DNA clearance during post-radiotherapy follow-up can further improve the risk stratification and support the development of personalized surveillance of patients with NPC according to dynamic change in individual risk profile (27).

The presence of post-radiotherapy detectable pEBV DNA likely represents the existence of minimal residual disease following definitive radiotherapy/chemoradiation, which might serve as surrogate for subclinical persistent or recurrent disease long before clinical progression (1). This is supported by the strong correlation of risk of progression with the absolute level of first post-radiotherapy pEBV DNA in our previous report (5). Here we further tested the hypothesis that post-radiotherapy pEBV DNA clearance as a surrogate biomarker of long-term survival in NPC.

To date, the clearance of pEBV DNA after adjuvant chemotherapy as compared with clinical observation has not been reported from a randomized controlled trial in NPC. Here we observed a strong association of post-radiotherapy pEBV DNA clearance and absence of clinical progression. This correlation holds true for all survival endpoints (including PFS, OS, LRFS, and DMFS). Our result confirmed the hypothesis that post-radiotherapy pEBV DNA clearance is associated with improved survival outcome. Lack of clearance of pEBV DNA in patients after adjuvant chemotherapy indicated absence of clinical benefit on long-term follow-up. Our findings suggested that liquid biopsy tracking by serial monitoring of pEBV DNA clearance may aid in real-time monitoring of treatment response and could potentially serve as an early surrogate endpoint for OS. With the increasing evidence on the clinical utility of circulating tumor DNA (ctDNA) analysis reported in several other cancer types (28–31), we recommend that future clinical trials of adjuvant therapy in NPC should incorporate post-radiotherapy pEBV DNA at enrolment (6). The use of pEBV DNA clearance as one of the coprimary endpoints could potentially improve trial efficiency by reducing the number of patients needed to treat and the duration of following up.

This study had several limitations. First, 6 and 7 patients in the randomized study arms did not complete the second pEBV DNA and PET-CT study, respectively (Fig. 1; Supplementary Table S1), for reasons of clinical progression (n = 4, one in adjuvant chemotherapy and three in observation arm), death from treatment complications (n = 2, one in adjuvant chemotherapy arm and one in observation arm) or intercurrent illness (n = 1, in adjuvant chemotherapy arm). To assess the potential statistical impact of participant drop out (6%) due to early clinical events occurring in the first 6 months post-randomization, we have conducted sensitivity analysis for three possible scenarios: (i) all 6 drop out patients had pEBV DNA clearance, (ii) all 6 drop out patients had pEBV DNA nonclearance, and (iii) all 3 patients with clinical progression had pEBV DNA nonclearance and 3 died of treatment complication or intercurrent illness had pEBV DNA clearance (Supplementary Table S4). The range of HR for PFS was between 5.2 and 10.4 for the best and worst scenario, respectively. These values were still covered in 95% CI of HR for PFS (4.4–20.9). Therefore, the small number of participants drop out in this study would not have significant impact on the conclusion.

Second, the second post-radiotherapy pEBV DNA was only performed in patients with detectable first post-radiotherapy pEBV DNA and eligible for randomization (group C), but not in patients excluded from randomization (group B). As group B had more progression events, the true performance of second pEBV DNA for predicting progression will again most likely to be underestimated. Third, we regarded group A patients with undetectable first post-radiotherapy plasma EBV DNA as low risk that adjuvant chemotherapy was not indicated, second plasma EBV DNA was not performed in group A in the study design. Although group A patients enjoyed a good prognosis, there remains a nonnegligible rate of relapse and death in this group (5-year rate of PFS 77.1%, OS 87.3%, Table 2; Supplementary Fig. S2). In a previous retrospective study of 273 patients with NPC who completed radiotherapy or chemoradiation, plasma EBV DNA was undetectable in 254 (89.4%) of patients at end of radiotherapy (± 1 week) but became detectable (change from negative to positive) in 10 (3.7%) of patients at 3 months after treatment (32). Three of these 10 patients developed disease failure within 3 months after treatment and another 5 patients developed disease failure at 3.2–27.5 months after detection of plasma EBV DNA at 3-months after treatment. Although not a common scenario, our study design cannot address this possibility of reemergence of plasma EBV DNA from negative to positive during follow-up. Fourth, because of the lack of efficacy of adjuvant cisplatin and gemcitabine to improve PFS in the original randomized controlled trial (5), we cannot ascertain the predictive value of pEBV DNA clearance and efficacy of adjuvant chemotherapy in this study. The establishment of pEBV DNA clearance as a formal surrogate endpoint of the efficacy for adjuvant therapy would require validation in additional large-scale prospective randomized trials.

We have previously demonstrated that patients with NPC with detectable post-radiotherapy pEBV DNA carried a poor prognosis and integrating post-radiotherapy pEBV DNA with TNM stage could refine the risk stratification (5, 6). By monitoring the dynamic change of post-radiotherapy pEBV DNA in this study, we identified a favorable prognostic subgroup among patients with initially detectable post-radiotherapy pEBV DNA. Our result provided reassurance to patients who experienced subsequent clearance of pEBV DNA; they enjoyed a superior long-term survival outcome comparable with patients with initially undetectable post-radiotherapy pEBV DNA.

E.P. Hui reports grants from Pfizer and Merck Sharp & Dohme, and personal fees from Merck Sharp & Dohme outside the submitted work. W.K.J. Lam reports other from Grail, Inc. outside the submitted work; in addition, W.K.J. Lam has multiple patent applications on molecular diagnostics using cell-free nucleic acids pending. K.C.A. Chan reports grants and personal fees from Grail and personal fees from Take2 during the conduct of the study, as well as grants and personal fees from Grail and Take2 outside the submitted work; in addition, K.C.A. Chan has a portfolio of patents on molecular diagnostics pending, issued, licensed, and with royalties paid from Grail, Take2, Illumina, Xcelom, and DRA. Y.M.D. Lo reports a Theme-Based Research Grant from Hong Kong Research Grants Council (grant no. T12-401/16-W), an endowed professorship from the Li Ka Shing Foundation, and a collaborative research agreement from Grail Inc during the conduct of the study, as well as other from Take2 Group of companies, DRA Limited, Illumina, Sequenom, Decheng Capital, and Grail Inc outside the submitted work; in addition, Y.M.D. Lo has numerous patents and patent applications in the cell-free DNA field pending, issued, licensed, and with royalties paid from Grail, Take2, Illumina, Sequenom, Xcelom and DRA Limited. A.T.C. Chan reports grants from Pfizer, Eli Lilly, and Novartis; grants and personal fees from Merck Serono and MSD; and personal fees from Cullinan outside the submitted work. No disclosures were reported by the other authors.

E.P. Hui: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. B.B.Y. Ma: Resources, data curation, investigation, writing–review and editing. W.K.J. Lam: Investigation, writing–review and editing. K.C.A. Chan: Investigation, writing–review and editing. F. Mo: Data curation, software, formal analysis, validation, visualization, methodology, writing–review and editing. Q.-y.H. Ai: Investigation, writing–review and editing. A.D. King: Investigation, writing–review and editing. C.H. Wong: Data curation, software, investigation, writing–review and editing. K.C.W. Wong: Resources, writing–review and editing. D.C.M. Lam: Resources, writing–review and editing. M. Tong: Resources, writing–review and editing. D.M.C. Poon: Resources, writing–review and editing. L. Li: Resources, writing–review and editing. T.K.H. Lau: Resources, writing–review and editing. K.H. Wong: Resources, writing–review and editing. Y.M.D. Lo: Funding acquisition, investigation, writing–review and editing. A.T.C. Chan: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

The authors would like to thank all study patients for their participation, and all coinvestigators for their contribution in the Hong Kong NPC Study Group 0502 trial.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.