Discussion
Nasendoscopy and biopsy have long been established as the gold standard for diagnosing NPC. However, NPC may be submucosal or contained within the pharyngeal recess and thus difficult to identify at endoscopy [
11], resulting in delayed diagnosis and treatment [
16]. Recent meta-analyses [
14,
15] have supported current practice guidelines in which MRI is employed for locoregional staging, with PET-CT playing a key role in staging of nodal and distant metastases [
6,
8,
12,
38]. However, the role of MRI in primary diagnosis of NPC is yet to be established.
Our meta-analysis supports the use of MRI as an accurate primary diagnostic tool, with a pooled sensitivity of 98.1%. MRI was also found to have excellent specificity and negative LR. MRI-detected abnormalities could direct biopsy planning whereas a normal MRI examination would establish a low post-test probability of disease. The applicability of these findings is strengthened by the MRI protocols involved, which included T2-weighted and pre- and post-gadolinium T1-weighted sequences on 1, 1.5 and 3 T magnets at slice thicknesses between 1 and 5 mm, reflecting real-world heterogeneity of protocols in clinical practice [
13].
The implication for current guidelines is that MRI could be the initial diagnostic test prior to endoscopic biopsy. Pre-operative MRI may increase the yield of endoscopic biopsy by alerting the surgeon to multiple lesions or lesions which may be endoscopically occult, and aid selection of patients who may benefit from intraoperative image-guided stereotactic approaches. The negative LR of MRI implies greatly reduced post-test probability of disease. MRI is a non-invasive technique which does not require anesthesia, yields excellent soft-tissue characterization of the lateral pharyngeal recess, submucosal soft tissues, and retropharyngeal nodes [
13], and already has an established role in locoregional staging [
6,
14,
15]. This raises the potential for use of MRI to exclude disease, reducing need for invasive biopsy.
The major pitfall for diagnosis highlighted in the included studies is false-positive results due to asymmetric nasopharyngeal hyperplasia [
39], reported in up to 14% of cases [
34]. Authors have attempted to distinguish this condition from NPC by assessing features such as lesion subsite, size, signal characteristics [
33], and anatomic features such as the deep mucosal white line [
19,
33]. Advanced imaging techniques [
40] have also been investigated as potential discriminators [
41‐
43]. Notably, several papers by King and colleagues [
19,
20], some of which are included in this meta-analysis [
16,
28,
30], have developed and applied multiparametric diagnostic criteria. Conversely, several included papers did not describe criteria for a positive test [
29,
31,
32,
34,
35], limiting potential for more subgroup analysis. Lesion size is known to limit endoscopic detection sensitivity [
9]. Unfortunately, there was insufficient data among the included papers to allow subgroup analysis of MRI accuracy based on lesion size or tumor stage. Further comparative studies are required for prospective validation of such grading systems.
The results of this systematic review and meta-analysis should be considered in context of several limitations. Only two studies [
28,
34] reported whether patients were enrolled randomly or consecutively, with one case–control study [
33] included. Combined with the inclusion of retrospective studies, and of multiple studies from a single institution, this introduces risk of selection and publication bias. Two studies [
16,
28] used clinical and imaging follow-up to document absence of NPC. Although such differential verification may overestimate test accuracy, it is difficult to ethically justify biopsy in patients with no visible lesion [
44]. One study [
34] was deemed at high risk of diagnostic review bias [
44] because endoscopy or histopathology interpretation was not blinded to MRI findings. When the initial endoscopy is negative, repeat endoscopy and biopsy targeted to imaging abnormalities is already the established standard of care [
6]. Furthermore, in clinical practice, histopathologic verification of suspected malignancy is still necessary and often interpreted with awareness of prior test results, such that the requirement for blinding is not so justifiable [
44]. Only one study reported time elapsed between MRI and biopsy [
28], whereas the others introduce risk of ascertainment or disease progression bias [
44]. Finally, the included cohort from the available literature is relatively small. Further prospective studies could assess the most robust sequences and imaging characteristics to enhance diagnostic accuracy.
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.