Surveillance tools for detection of recurrent nasopharyngeal carcinoma: An evidence-based review and recommendations

Endoscopy

Traditional diagnostic approaches in primary NPC involve nasopharyngeal endoscopy confirmed with a definitive biopsy of the primary tumor site.¹⁵ However, assessment of the nasopharyngeal space for recurrent disease may be confounded by post-radiation changes or obstructed by secretions and crusting. Improvements to the flexible nasopharyngoscopy have been developed such as narrow-band imaging (NBI), a novel optical technique that enhances the image of capillary vessels on the surface mucosa.¹⁶ In this review, three studies evaluating the effectiveness of enhanced endoscopy methods such as NBI and contact endoscopy were identified.^17-19

Flexible endoscopes allow for the visualization of the texture, color, and the extent of mucosal lesions; however unlike contact endoscopy they lack the ability to provide histopathological data during clinical examination.¹⁸ Contact endoscopy for the detection of primary NPC tumors has been verified previously and allows in vivo differentiation of normal and malignant cells of the nasopharynx by using in vivo microscopy that allow the evaluation of superficial cell layers of the epithelium.²⁰ Pak and colleagues were able to verify that this technique also extends to the surveillance of rNPC in patients treated with radiation for NPC.¹⁷ They reported a specificity and sensitivity of 100% in the detection of rNPC using contact endoscopy with a diagnostic accuracy of 92.1%. It should be noted, however, that only 5/64 patients enrolled in the study had recurrence and thus the sample size for verification of this technique is rather small. The authors recommended that contact endoscopy should be added as an intermediate procedure between nasal endoscopy and biopsy for patients with suspicious or normal nasal endoscopy results, to limit the need for more invasive biopsy and more accurate surveillance of rNPC.

In a case report presented by Lin and colleagues, they validated the effectiveness of NBI in rNPC surveillance for a 51-year old NPC patient treated with radiation.¹⁸ NBI was able to accurately detect recurrence at an earlier stage compared to nasal endoscopy when the patient did not present with clinical symptoms. Follow-up with MR imaging confirmed local rNPC and prompt treatment was initiated for the patient. The authors suggest that NBI may serve as the ideal surveillance tool in the early months post-radiation after NPC treatment as it uniquely detects lesion areas that are not visually apparent on nasal endoscopy alone. The authors themselves did not define what constituted the ‘early post-radiation period' in this study. However, several other papers share the consensus that this period ranges up to 4 to 6 months post-treatment.^21-24 Wang and colleagues verified the effectiveness of NBI in monitoring rNPC in a case study of 106 patients with NPC, although the rate of false positives was quite high (25/106) using NBI.¹⁹ Thus, Wang and colleagues suggested a modified NBI technique coined NBI closer view where areas suspicious on NBI were followed with further endoscopic closer view. Using this modified technique, the authors reported a sensitivity, specificity, and accuracy of 87.5%, 87.8%, and 87.7%, respectively, and conclude that NBI closer view had lower rates of false positives due to inflammation, necrosis, or non-neoplastic brownish spots mistaken for recurrence.

MR

In the background of post-radiation changes to the skull and submucosa after treatment, it has been argued that magnetic resonance imaging (MR) should be favored over conventional CT imaging because MRI has higher sensitivity for soft tissue structures. Indeed, in a prospective case series study performed by King and colleagues, the authors demonstrated that the specificity of MRI was equivalent to endoscopy with superior sensitivity in detecting primary NPC.²⁶ With respect to rNPC, five studies comparing the specificity, sensitivity, and accuracy of MRI to conventional imaging methods were identified.^27-31

In a case study performed by Gong and colleagues, 88 participants diagnosed with NPC were classified into either the control (n = 16) group if they had not yet received radiotherapy, or patient (n = 72) group if they had undergone radiotherapy.³⁰ The researchers then scanned each participant using both CT and MR imaging and characterized the signal intensity patterns of the MR images in patients with rNPC (n = 29). They concluded that MR images demonstrating signal intensity in T2, weighted, spin-echo pulse sequences were most accurate in distinguishing rNPC from fibrosis or primary NPC.³⁰ One drawback to the study is that it is fairly descriptive and qualitative in nature with no numerical data depicting accuracy rates or sensitivity and specificity values to compare and quantify the imaging modalities. In a 2 013 case study, Lin and colleagues took a similar approach and used MR imaging in 108 patients to evaluate post-radiation changes 1, 3, and 6 months after treatment.²⁸ The authors also concluded that residual tumor shows enhanced signal intensity on T2-weighted images and the appearance of tumor after radiotherapy correlated with rNPC up to five years following treatment.

Two case studies by Comoretto et al.²⁹ and Ng et al.³¹ both compared the effectiveness of MR vs FDG-PET imaging to identify rNPC. Comoretto and colleagues retrospectively examined 63 patients who were deemed high risk of residual disease for recurrence or restaging.²⁹ They reported the MR outperformed FDG-PET at detecting residual or recurrent NPC at the primary site (P = 0.16, accuracy = 92.1% for MR vs 85.7% for FDG-PET) but was inferior compared to FDG-PET at detecting rNPC in patients in regional lymph nodes. The authors attributed the limited use of the field of view and distant metastases to the inferiority of MRI at distant sites, suggesting that expansion of the MR scan to include supraclavicular lymph nodes may aid with detection of distant rNPC in MR imaging.²⁹ Using whole body MR imaging in 179 patients at high risk of residual or rNPC, Ng and colleagues found that MRI was similar to FDG-PET at detecting local rNPC (96.8% and 96.1% specificity for MRI and FDG-PET, respectively).³¹ In contrast, whole body MR was superior to FDG-PET in detecting distant rNPC (99.7% and 97.5% specificity, respectively). Since specificity was comparable between the two imaging modalities, the authors concluded that whole body MR and FDG-PET complement each other across different body sites. Whole body MRI is preferred over FDG-PET for brain and splenic metastases, intracranial invasion, and necrotic adenopathy.³¹ Ng and colleagues could not recommend the routine use of either modality for routine detection of rNPC due to lack of cost effectiveness evaluation.

In a study by Huang et al.²⁷ the authors noted the value of dynamic contrast enhanced T1W perfusion MR imaging in the detection of local head and neck primary tumors. However, patients with a history of iodine allergy would not be suitable for intravenous uptake of the contrast agent. Using 27 patients diagnosed with rNPC and 35 patients treated for NPC with tumor free post-chemoradiation fibrosis, the authors compared the diagnostic value of diffusion weighted MR imaging based on turbo spin-echo (TSE) and echo-planar (EP) imaging techniques. The latter is prone to distortion and susceptibility artifacts especially in areas of the head and neck, including the nasopharynx. Huang et al. found that TSE-DWI showed superior image quality and greater diagnostic accuracy for rNPC compared to conventional EP-DWI. TSE-DWI was able to efficiently differentiate between rNPC and post-chemoradiation changes (area under curve of 0.932).

FDG-PET

Radiation therapy is a mainstay component of NPC due to anatomical constraints and the high radiosensitivity of NPC.³⁴ However, long-term post-radiation changes are often observed such as crusting, mucositis, or varying degrees of trismus,³⁵ which compromise results using conventional techniques such as flexible endoscopy or CT imaging. Fluorodeoxyglucose positron emission tomography (FDG-PET) bypasses traditional imaging techniques by identifying head-and-neck malignancy through regions of increased glucose metabolism. FDG-PET has been shown to be sensitive and specific at detecting residual or recurrent tumors from post-radiation changes.^{36, 37} In this review, 5 studies evaluating the effectiveness of FDG-PET were identified.^{21, 23, 24, 29, 38}

The mainstay imaging modality for identifying mucosal recurrence has traditionally been MRI but arguments have been made for the effectiveness of FDG-PET, particularly in patients with indeterminant MRI findings. In case series by Tsai et al.²⁴ and Ng et al.,²³ patients who underwent radiotherapy and subsequently had non-conclusive MRI findings 4 months after completion of therapy were re-scanned using FDG-PET. Half of the patients (n = 14) enrolled in Tsai et al.'s study presented with recurrence, all which were accurately identified using FDG-PET, with only one false positive finding. Ng et al. enrolled 37 patients and while FDG-PET correctly identified 17/19 patients with residual or recurrent NPC, the rates of false-positives (8) and false-negatives (2) were higher, yielding an overall specificity of 89.5% and sensitivity of 55.6%.

Al-Amro et al.³⁸ compared the accuracy of FDG-PET to conventional CT imaging and found that FDG-PET was superior to CT with respect to specificity and sensitivity of rNPC detection at both primary and nodal sites. Similarly, Comoretto and colleagues²⁹ directly compared rNPC detection using both MRI and FDG-PET and concluded that FDG-PET had higher sensitivity and specificity at regional lymph nodes and distant metastases whereas MRI fared better at the primary site. The authors also found that combining imaging results from both MRI and FDG-PET was more accurate than each modality independently. In line with these results, Chan et al.²¹ reported similar findings such that FDG-PET was superior to MRI at detecting tumor recurrence in patients with initial T4 disease but not early stage (T1-T2) disease. Authors from the Tsai et al., Ng et al., and Chan et al. studies noted that the accumulation of FDG in inflammatory tissue may have led to the false positive readings in recently irradiated patients.^{21, 23, 24} Almost all (6/8) false positive cases from the Chan et al. study showed regression on follow-up imaging 3 months after the initial false positive finding, indicating the less pronounced influence of post-irradiation inflammation on imaging studies over time.²¹ For this reason, the authors highlighted that FDG-PET may not be a suitable surveillance tool during the early post-radiation period (~4–6 months).

Tc-99m MIBI and ²⁰¹Tl SPECT

Nuclear medicine techniques such as FDG-PET have shown to be highly sensitive and specific at distinguishing residual or recurrent NPC from post-radiation changes, but are relatively costly, have lower resolution, and have limited availability. The consideration of SPECT imaging using either Tc-99m MIBI or ²⁰¹Tl has been debated in the literature as alternatives or adjuncts to conventional endoscopy or imaging methods such as FDG-PET or MRI where conventional methods may fail due to post-radiation changes in the nasopharyngeal area. In this review, six studies were identified evaluating the specificity, sensitivity, and accuracy of Tc-99m MIBI or ²⁰¹Tl in surveillance of recurrent NPC.^{22, 40-44}

In a prospective study of 18 NPC patients, Kostakoglu et al.²² compared the effectiveness of post-therapy evaluation for rNPC using ²⁰¹Tl, Tc-99m MIBI, and MR imaging at 3 and 6 months post-treatment. Whereas MR demonstrated early false positives at the 3-month follow-up scan (15% specificity, 100% sensitivity, and 39% accuracy), MIBI imaging was superior at both time points with regards to specificity, sensitivity, and accuracy (85%, 100%, and 89%; 93%, 100%, and 94%, respectively at 3 and 6 months).²² In the 18 patients studied by Kostakoglu and colleagues,²⁰¹Tl yielded the lowest diagnostic accuracy (72% and 71% at 3 and 6 months post-treatment). In a later study with a larger sample of 46 NPC patients, Sobic-Saranovic and colleagues presented differing results demonstrating that the specificity, sensitivity, and accuracy of ²⁰¹Tl was actually similar to that of MIBI SPECT imaging (88%, 78%, and 82% versus 86%, 77%, and 81%).⁴⁴ It is unclear, however, if these results can be directly compared to the earlier study performed by Kostakoglu et al. as patients in the latter study were subdivided into two groups to receive different imaging modalities (either MIBI SPECT or ²⁰¹Tl) rather than both imaging modalities due to the limited availability of tracers and high radiation exposure.

The effectiveness of ²⁰¹Tl imaging in detection of rNPC was further evaluated by two case series by Shiau et al.⁴² and Tai et al.⁴³ Shiau and colleagues presented a case study of 30 NPC patients with indeterminant CT findings,⁴² whereas Tai et al. applied ²⁰¹Tl imaging to 26 NPC patients with indeterminant MRI findings.⁴³ In both studies, the use of ²⁰¹Tl imaging yielded effective detection of tumor recurrence when traditional imaging modalities failed. Since these two imaging studies were performed by the same group, it is unclear if there was overlap between the patients included in both studies whom had both indeterminant findings on CT and MR — and whether these patients represent a unique subset of NPC patients with anatomical difficulties yielding poor radiographic findings that may be especially beneficial to non-traditional imaging such as ²⁰¹Tl imaging techniques.

Evidence for the effectiveness of MIBI SPECT was also further elucidated by Kao and colleagues in a case series.⁴¹ The authors compared the efficacy of MIBI SPECT to traditional imaging techniques (CT and FDG-PET imaging) in 36 NPC patients and found that MIBI SPECT demonstrated comparable specificity to CT imaging (both 73%), though MIBI SPECT had slightly higher sensitivity and accuracy than CT alone (88% and 83% sensitivity and accuracy for FDG-PET vs 96% and 89%, respectively, for MIBI SPECT).⁴¹ In contrast, MIBI SPECT was not superior to FDG-PET in detecting rNPC (100% specificity, 96% sensitivity, and 97% accuracy). The authors concluded that while MIBI SPECT was more effective than CT at identifying rNPC, congruence between MIBI SPECT and CT imaging yielded the best predictive results. In patients with incongruent imaging results on MIBI SPECT and CT, FDG-PET was still the most effective surveillance modality.⁴¹

Epstein-Barr Virus DNA

Several risk factors have been identified in NPC, in particular the association between Epstein - Barr virus (EBV) infection and NPC pathogenesis is considered crucial in the etiology of NPC.⁴⁶ Using real-time quantitative PCR (rt-qPCR), Lo and colleagues demonstrated both quantitative and temporal correlations with circulating cell-free levels of EBV DNA and tumor recurrence in NPC.⁴⁷ Circulating EBV DNA has been evaluated as an adjunct to endoscopy and imaging in the pre-, mid-, and post-treatment surveillance of patients with NPC.^{48, 49} In this review, five studies were identified evaluating the effectiveness of EBV DNA detection in the surveillance of recurrent NPC.^50-54

Leung and colleagues quantified and compared plasma EBV DNA levels from 189 patients with NPC (140 with primary NPC, 24 with rNPC, 25 with metastatic NPC) to evaluate the sensitivity of using EBV DNA as a surveillance tool for local recurrence.⁵³ They demonstrated that detection of plasma EBV DNA was significantly lower in sensitivity in patients with local rNPC (15 out of 24 patients) compared with patients with primary NPC (135 out of 140 patients) and even in patients with distant, metastatic rNPC (24 out of 25 patients). Among the rNPC tumors positive for plasma EBV DNA, the authors quantified the median EBV DNA level and reported significantly lower levels of EBV DNA in rNPC tumors compared to primary NPC tumors. Thus, the authors suggest that EBV DNA is only sensitive for the diagnosis of primary NPC and metastatic NPC but should not be used as a sole surveillance technique in rNPC detection, especially in local sites. Leung et al.⁵³ hypothesized that the clearance of EBV DNA in rNPC tumors may be due to post-radiation changes that are found locally and not at distant sites, resulting in metastatic rNPC sites still demonstrating high levels of EBV DNA.

In a 2004 case series by Shao and colleagues,⁵² the authors used a similar stratification of 149 patients (120 patients with primary NPC, 8 with rNPC, 20 with distant metastatic NPC) to compare the effectiveness of EBV DNA versus EBV immunoglobulin A/viral capsid antigen and early antigen in diagnosing and monitoring patients with NPC. They found similar patterns of plasma EBV DNA levels to the study performed by Leung et al.⁵³ such that quantitatively, EBV DNA levels were lowest in local rNPC tumors compared to primary and metastatic NPC. When comparing the methods of plasma detection of EBV DNA versus immunoglobulin detection, Shao and colleagues report that plasma EBV DNA is a more sensitive and specific marker such that patients in remission demonstrate negative testing on EBV DNA (0 copies/mL) but not EBV immunoglobulin A/viral capsid antigen (levels remained high in all patients groups). The authors suggest that serologic techniques to diagnose and monitor NPC are only semi-quantitative and plasma EBV DNA surveillance is a more superior and sensitive technique in monitoring patients with NPC. These findings demonstrating the superiority of plasma EBV DNA to serology techniques in surveillance of rNPC were also verified by Stoker and colleagues,⁵¹ in a case study to evaluate the application of nasopharyngeal brush to detect EBV DNA. Stoker et al. compared pre- and post-treatment EBV DNA levels and serology obtained from traditional blood samples in 72 patients to those obtained from their newer technique, the nasopharyngeal brush, which is minimally invasive and easier to perform in the clinic compared to biopsy. They found that pre-treatment, EBV DNA levels acquired using nasopharyngeal brush outperformed plasma EBV DNA levels in detection of NPC, however, post-treatment, plasma EBV DNA levels were more accurate in diagnosis of rNPC compared to nasopharyngeal brush. The authors also validated previous studies in confirming that EBV serology did not demonstrate any prognostic value in detecting persistent or recurrent NPC. Stoker and colleagues suggested that monitoring EBV DNA levels using the nasopharyngeal brush is an effective technique in the early months after treatment for NPC and yields high prognostic value in distinguishing tumor from post-radiation damage. Using a smaller sample size of patients with rNPC (n = 14) and a control group (n = 15) of patients treated for NPC in clinical remission, Lam et al.⁵⁰ verified the nasopharyngeal brush technique as an additional surveillance strategy for rNPC. The authors found a statistically significant difference in Epstein Barr virus detection levels (EDL) between the recurrence group (mean = 2.38) and the control group (mean = 0.17). Further validation of this technique using a larger sample size is required in order to determine sensitivity, specificity, and accuracy.

Moreover, in a 2011 case study of 245 patients diagnosed with NPC, Wang and colleagues evaluated the effectiveness of the plasma EBV DNA assay prior to PET imaging for diagnosis of recurrence and treatment planning in rNPC.⁵⁴ Of the patients enrolled in the study, 41/245 developed rNPC. 36 out of 41 of these patients were accurately identified through the detection of abnormal EBV DNA levels, whereas 5/41 of these patients were falsely negative for EBV DNA but demonstrated clinical signs/symptoms suggestive of rNPC. Subsequent imaging with PET then revealed recurrence sites and guided treatment planning for these patients. The authors suggest that this surveillance approach using plasma EBV DNA assay prior to PET imaging yields high accuracy in detecting rNPC and should be used as a cost-effective measure in monitoring NPC patients as the EBV DNA assay is 77–88% less costly than PET imaging, without exposure to ionizing radiation, and did not identify any false positives. Nonetheless, clinical suspicion and expertise are still necessary as negative EBV DNA results (false negative) were obtained in 5 patients with rNPC. It should be noted that patients without recurrence identified by either EBV DNA levels or clinical signs/symptoms were not scanned with PET imaging in this study, and thus the true accuracy, sensitivity, and specificity of this tailored approach cannot be conclusively determined.

In addition to analysis for plasma EBV DNA, other EBV-derived genes involved in the immune response attributable to the pathogenesis of NPC have been shown to be promising tumor markers. These include the detection of EBV encoded latent membrane protein 1 (LMP1) and rightward transcripts known as BART microRNAs. LMP1 has been useful in the diagnosis of local rNPC, especially in cases of suspected osteoradionecrosis.⁵⁵ Changes in LMP1 status at the time of NPC reoccurrence have been observed to manifest earlier (~11.9 weeks) than morphological changes detected by nasopharyngoscopy.⁵⁶ The potential utility of EBV miRNA BARTS has also been studied. Tissue EBV miRNA BART7 was shown to be useful for identifying a subgroup of patients with histologically clear margins who are at an increased risk of subsequently developing tumor reoccurrence compared to those with negative histological margins.⁵⁷ Another study which profiled circulating plasma BART miRNAs, identified that BART8-3p and BART10-3p had the greatest efficacy in discerning rNPC patients with undetectable EBV DNA from non-cancer controls.⁵⁸