Background
More than 20% of patients with nonmetastatic nasopharyngeal carcinoma (NPC) will experience disease recurrence despite aggressive primary treatment, contributing largely to treatment failure and death [
1,
2]. The posttreatment surveillance is essential to detect recurrence timely when the tumor burden is minimal, which would maximize the efficacy of salvage treatment and might improve survival [
3‐
6]. However, surveillance can be challenging in NPC, and there is no consensus on the optimal follow-up modalities. Consequently, practice varies widely across clinicians and institutions. Although routine surveillance imaging, e.g., routine magnetic resonance imaging (MRI) and computed tomography (CT), is efficacious in detecting recurrence, it has been questioned due to the prohibitive resource consumption and economic burden [
7]. Hence, it is imperative to identify a follow-up strategy that can detect recurrence in a timely and cost-effective manner.
Epstein-Barr virus (EBV) infections were predominant in endemic NPC [
8]. Circulating cell-free EBV (cfEBV) DNA, short DNA fragments released by NPC cells, has been established as an NPC biomarker in population screening [
9], risk assessment [
8,
10], treatment evaluation [
11,
12], and follow-up [
13‐
16]. Several studies have shown that the posttreatment cfEBV DNA tests could facilitate early detection of NPC recurrence, especially for distant recurrence [
13‐
18]. Specifically, detectable plasma cfEBV DNA during follow-up indicates tumor recurrence, while undetectable cfEBV DNA demonstrates continuous remission [
15,
16,
18]. A recent work involving 1984 nonmetastatic NPC patients demonstrated that the sensitivity and specificity for cfEBV DNA to detect recurrence were up to 82.3 and 80.0%, respectively, highlighting the efficacy of cfEBV DNA surveillance [
16]. Another study investigating cfEBV DNA surveillance followed by positron emission tomography/computed tomography (PET/CT) indicated that such strategy could correctly identify NPC recurrence while saved approximately four-fifths in expenses [
14]. Therefore, the updated European Society for Medical Oncology (ESMO) guideline for NPC has recognized cfEBV DNA as a promising biomarker for recurrence and recommended evaluating it at least every year [
17], and the National Comprehensive Cancer Network (NCCN) guidelines also suggested EBV DNA monitoring for NPC [
19]. However, little evidence is available regarding how to integrate cfEBV DNA into the current surveillance series [
14,
17].
Considering that cfEBV DNA could promptly suggest disease recurrence with high accuracy [
14‐
16], we proposed a liquid biopsy-based stepwise surveillance strategy, starting with routine cfEBV DNA tests, followed by further imaging studies if the tests are positive. We estimated its cost-effectiveness and compared it with other surveillance strategies in each stage of NPC. We speculated that selected imaging studies for patients with positive cfEBV DNA was a cost-effective follow-up modality, demonstrating the potential application of circulating cell-free DNA in the follow-up of cancer survivors. We aimed to establish evidence-based, stage-specific NPC surveillance strategies and investigate the potential application of circulating cell-free DNA in the follow-up of cancer survivors.
Discussion
To our knowledge, this is the first study to evaluate the liquid biopsy-based surveillance in NPC patients. We found that the cfEBV DNA-guided imaging strategies were remarkably more cost-effective than routine imaging strategies. Specifically, cfEBV DNA-guided MRI + CT + BS was cost-effective for stage II and III patients, while cfEBV DNA-guided PET/CT was cost-effective for stage IV patients. However, routine follow-up was sufficient for stage I NPC patients due to their low recurrence probabilities. Overall, our results highlight the feasibility of cfEBV DNA in the surveillance of NPC.
According to the current NCCN guideline, routine surveillance imaging was not recommended in asymptomatic patients [
19], and our analysis agreed with it. However, directly halting imaging would result in delayed diagnoses of recurrence, especially for distant recurrence because many patients are asymptomatic [
19,
69]. In fact, surveillance imaging was the most widely used follow-up modality in NPC, and 79% of clinicians would order routine imaging in clinical practice for reassurance, patient requests, etc. [
70]. Consequently, undifferentiated imaging surveillance has resulted in considerable economic burdens and resource consumption. Given these circumstances, it is urgent to establish a convenient and reliable method that can distinguish patients at high risk of developing recurrence so as to focus imaging surveillance on them.
cfEBV DNA is an ideal biomarker with low expense and high sensitivity in suggesting disease recurrence. The ESMO and NCCN guidelines both recommended cfEBV DNA tests in NPC follow-up [
17,
19]. Therefore, cfEBV DNA-guided imaging strategies are particularly promising. Using cfEBV DNA, patients at risk of recurrence can be identified in order to concentrate surveillance imaging on those who will benefit, thus sparing the unnecessary expense and radiation exposure for those who will not. Compared with routine imaging surveillance, targeting imaging to patients with positive cfEBV DNA gained similar survival benefits but only required one quarter of the imaging to detect one recurrence. This liquid biopsy-based surveillance strategy underscores the practice of precision medicine in the follow-up of cancer survivors, which has been investigated in several malignancies [
71‐
73]. For example, Roschewski et al. [
72] found that surveillance circulating tumor DNA could accurately suggest recurrence in diffuse large B cell lymphoma, leading to reduced disease burden at recurrence.
Until now, few studies have investigated the optimal follow-up modalities for each stage of NPC [
30]. Consequently, despite the substantial heterogeneity of recurrence risk in different disease stages of NPC [
1,
22], the ESMO and NCCN guideline failed to recommend stage-specific follow-up strategies [
17,
19]. Zhou et al. [
30] evaluated the cost-effectiveness of surveillance MRI and recommended annual MRI for patients with advanced tumor stage and no MRI for early stage. However, they did not consider distant recurrence, the predominant failure pattern in NPC [
1]. Besides, they did not investigate cfEBV DNA and PET/CT, which have been widely adopted in clinical practice. Our study addressed these unmet needs. We found that cfEBV DNA-guided imaging surveillance is particularly preferable for stage II–IV NPC patients, where selected MRI + CT + BS or PET/CT were performed depending on their disease stages. However, for stage I patients, cfEBV DNA-guided imaging is not recommended due to their fairly good survival, with 10-year overall survival and recurrence-free survival both greater than 95% [
74]. Hence, routine follow-up would be sufficient for stage I NPC patients.
Evidence has indicated that PET/CT was more effective than other imaging modalities in detecting recurrent NPC, especially for distant recurrence [
19,
31,
34,
37,
75]. However, lacking guidance regarding when to perform PET/CT during NPC surveillance, clinicians have to be critically cautious about prescribing costly PET/CT when following up patients with relatively poor economic conditions. In this study, we found that cfEBV DNA could be a precise indicator to perform PET/CT for patients at risk of recurrence, which was cost-effective in stage IV NPC patients. In addition, we observed that the cost of PET/CT predominantly determined the ICERs between cfEBV DNA-guided PET/CT and cfEBV DNA-guided MRI + CT + BS in the sensitivity analyses. With the discounting of PET/CT, cfEBV DNA-guide PET/CT will be increasingly cost-effective. Therefore, the tradeoffs between the two strategies should be continuously recalibrated to reflect the updated cost of PET/CT in clinical practice, especially among stage III NPC patients where cfEBV DNA-guided PET/CT was near the most cost-effective strategy.
With the growing recognition that cfEBV DNA possessed favorable performance in detecting tumor recurrence [
13‐
16,
76], we proposed cfEBV DNA-guided surveillance strategies as references for patients and clinicians in NPC follow-up. Nevertheless, our recommendations should be implemented cautiously. First, although cfEBV DNA demonstrates high sensitivity in detecting RR and DM, it is relatively insensitive to LR [
13,
14,
16,
76]. Therefore, cfEBV DNA results should be interpreted with the history and physical examinations and nasopharyngoscopies to comprehensively evaluate the local conditions [
76]. Second, cfEBV DNA levels might elevate even before recurrences can be detected on imaging [
16,
76]. Hence, patients with detectable cfEBV DNA but negative imaging might require subsequent cfEBV DNA tests (e.g., another cfEBV DNA test 1–3 months later) to distinguish much earlier-stage recurrence from false-positive results, especially for those with high recurrence risk [
16].
Several limitations need to be noted. First, the inherent shortcoming of the decision-analytic model was the model inputs determined from various sources. We therefore conducted extensive sensitivity analyses to clarify the uncertainty, and the cfEBV DNA-guided imaging strategy was confirmed to be robustly cost-effective in the sensitivity analysis. Second, the model did not account for secondary recurrences because our study population was patients achieving CR to the primary treatment for NPC. Patients who have experienced recurrence will manifest distinct risks of disease failure, and thus may require different surveillance strategies [
8,
77]. Third, since the study is from the Chinese societal perspective, the results might vary with different medical and societal costs. Accordingly, similar analyses in other regions are warranted. Last but not least, the study does not account for non-endemic NPC. NPC in endemic areas, including southern China, Southeast Asia, and North and East Africa, constitutes more than 80% of new cases worldwide [
78‐
80], which are mostly the non-keratinising subtype (> 95%) and are invariably associated with EBV infection [
81]. However, the keratinising squamous subtype is more common in regions where NPC is non-endemic, e.g., North America and northern Europe [
81,
82]. Evidence has indicated that patients with the keratinising squamous subtype tend to have worse survival, poorer local control, and less distant failures than those with the non-keratinising subtype; and their association with EBV infection is relatively low [
81‐
84]. Therefore, the applicability of cfEBV DNA-guided surveillance strategies should be carefully re-evaluated in non-endemic areas.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.