Outcomes for patients with NPC have improved over the years with the introduction of chemotherapy and IMRT. However, local failure in the form of residual or recurrent disease still occur in 10-30% of cases (9, 10). The assessment of treatment response and clinical surveillance after definitive therapy for NPC is important, in order to permit earlier recognition of local failure and initiation of salvage therapies. Typically, patients are assessed clinically with cranial nerve examination, neck palpation and endoscopic inspection, in combination with imaging such as CT, MRI or PET-CT. Any suspicious lesions are then biopsied to obtain histological confirmation.
Given that radiation can lead to anatomical distortion within the treatment field, identification of residual or recurrent disease is often challenging. Palpation for cervical lymph nodes may be limited by fibrosis of neck musculature. Post radiation endoscopic examination may only reveal subtle mucosal changes such as fullness of postnasal space (PNS), or a mass which may represent fibrosis, crust or slough rather than residual tumour (11), and submucosal or deep-seated recurrences may be missed. Sensitivity of endoscopic examination in detecting persistent disease after RT is only 40.4% (12). Similarly, endoscopic biopsies run the risk of sampling errors as residual tumour cells are often scattered in small clusters, resulting in a sensitivity at 6 weeks post RT of 59.3% (12).
Radiological assessment faces similar difficulties, and to date there is no consensus regarding the optimal imaging modality in the post treatment setting. The utility of MRI, whilst well established in the initial staging of biopsy-proven NPC, is less clear post treatment. Compared to CT, MRI is able to better differentiate post radiation changes from recurrent tumour and delineate extent of disease (13, 14). The identification of skull base erosion is improved with contrast-enhanced fat-suppressed sequences (5).
However, when compared to PET-CT, MRI may be limited in its ability to distinguish between post RT changes often seen in the irradiated nasopharynx and neck e.g. tumour necrosis, tissue fibrosis and inflammation, from true viable tumour. Conversely, changes in tissue metabolism may precede changes in tumour morphology or volume. Liu et al in a systemic review concluded that PET-CT, with its ability for combined functional and anatomic assessment, had superior pooled sensitivity, specificity and overall diagnostic accuracy when compared to both CT and MRI (15). PET-CT also has the added benefit of uncovering any systemic metastases within the same examination, which in turn can determine goals of further treatment.
Nevertheless, there are drawbacks to relying solely on PET-CT to uncover residual or recurrent NPC. Disease at the primary site is more accurately demonstrated on MRI rather than PET-CT (92.1% vs. 85.7%) according to Comoretto et al (16). The latter produced false negative findings especially where there is when intracranial extension of disease. False positive results have also been associated with PET-CT when it is done too early post RT due to the residual inflammatory effects causing apparent increased glucose metabolism. It has been suggested that the PET-CT should be done 6 months or later post RT for optimal accuracy (sensitivity and specificity are 92% and 100% at 6 months or later vs 33% and 64% within 5 months) (17). Additionally, limiting factors such as cost and resource availability will need to be considered before the prevailing use of PET-CT takes place. In view of these, it is likely that MRI and PET-CT should be complementary, in order to improve overall diagnostic accuracy for recurrent or residual disease (16). Our data indicates that, within the boundaries of our institutional practice, most loco-regional recurrences are detected by MRI rather than non-MRI radiological modalities or clinical examination. If salvage surgery or RT are planned for LRR, the superior ability to determine extent of tissue invasion with MRI makes it preferable to PET-CT in guiding resectability or extent of re-treatment required.
Another issue worth addressing is the optimal timing of MRI in view of the potential diagnostic uncertainties post RT. Guidelines suggest varying timing of post treatment imaging ranging from 3 to 12 months (18-20). The time course of NPC regression after definitive treatment has been studied histologically and radiologically. In 1999 Kwong et al followed 803 NPC patients treated with 3D-CRT with or without neoadjuvant chemotherapy, through 2-weekly endoscopic biopsies. 93.2% achieved histologic remission by 12 weeks post RT (8). This formed the basis upon which many subsequent studies relied on when scheduling post treatment imaging. Li et al in 2017 challenged this paradigm with data from 556 NPC patients treated definitively with IMRT and followed up with serial MRIs (21). All MRI scans were reviewed independently by two radiologists with extensive experience in head and neck cancer imaging. In this group, 83.3% of patients achieved a clinical complete response (cCR) – defined as no evidence of residual tumour or nodal disease based on examination with MRI and flexible nasoendoscopy - at 3-4 months (early cCR), a figure which increased to 91.4% at 6-9 months (delayed cCR). Interestingly, prognosis of patients with a delayed cCR was no different to those with an early cCR, leading the authors to conclude that 6-9 months may be the best time point for assessment of maximal tumour response to IMRT.
The increasing use of IMRT as standard of care, as well as CCRT has been associated with delayed primary tumour regression through mechanisms which to date remain unclear (22). Whilst results from both Li’s and our study both suggest a lag time between histological and radiological tumour regression, a significantly lower proportion of patients in our series (32.8% vs 83.3%) achieved a radiological CR on the first ptMRI. This may be explained by the methodology of our study, which looks at the real-world radiology reports, rather than have one or more radiologists retrospectively re-looking radiology images and coming to a binary outcome regarding the absence or presence of residual disease. We believe our method more truly reflects real-life clinical practice, where limits of certainty in radiological interpretation is accepted, and collective decision-making is undertaken in a multidisciplinary setting for those cases showing indeterminate radiological changes. Even with this uncertainty, it may still be worthwhile having a first ptMRI done at an earlier date to provide a crucial baseline which can inform future scans. In addition, there may be a window period between 3-4 months and 6-9 months where further investigations may lead to earlier diagnosis of residual or recurrent disease, permitting prompt initiation of salvage therapies. Indeed, our study offers evidence that the optimal timing of the first ptMRI should be 4, rather than 3 months to reduce the rates of uncertainty in radiology reports.
We acknowledge that this study has some limitations. Firstly, the retrospective design meant that patients were not followed up based on a standardised protocol. ptMRIs were done at the clinicians’ discretion, resulting in a wide date range for the first (32-346 days) and subsequent ptMRIs. Secondly, the determination of ptMRI status was based primarily on review of actual radiology reports. Although these were all reported by a number of radiologists with a special interest in head and neck imaging, their level of experience may be differing.