This study documented the frequency and range in size of normal/reactive benign nodes in nodal groups of head and neck that have received less scrutiny in the literature, namely the parotid, submandibular, occipital, Level IIb nodes, facial and retroauricular groups. Measurable nodes (i.e., those ≥ 2 mm) were most frequent in the submandibular (90.3%), followed by Level IIb (79.2%), parotid (74.4%), occipital (27.7%) and retroauricular (2.8%) groups, and no nodes were identified in the facial group. Nodes in the parotid, occipital, Level IIb, and retroauricular groups were small ranging in mean SAD from 3.8 to 4.4 mm, while nodes in the submandibular group were slightly larger with a mean SAD of 5.5 mm. The range of nodal size in each of these groups only surpassed the threshold of ≥ 10 mm for a malignant node in 0.8% of submandibular nodes and none of the nodes in the other groups.
Nodal sizes for retropharyngeal, jugulodigastric and other Level IIa nodes were in keeping with sizes reported in the literature, suggesting our group is fairly representative of the expected size of normal/reactive nodes [8, 9, 13, 25, 26]. Of note, the largest node in the neck was the jugulodigastric node followed by the other Level IIa nodes. The larger size of jugulodigastric, Level IIa and submandibular nodes may reflect stimulation by dental disease or upper respiratory tract infections. The size of the retropharyngeal nodes was similar to nodes in the parotid, Level IIb, occipital groups and retroauricular groups. The smaller nodal size in the upper internal jugular chain in Level IIb compared to Level IIa, is of special interest in these patients being screened for NPC, because Level IIb nodes are commonly the first echelon of nodal spread in NPC [27]. The results in this study suggest that applying the current threshold of ≥ 10 mm to parotid, occipital, Level IIb nodes, facial and retroauricular nodes would result in a very high specificity for a malignant node but would compromise sensitivity. Recently Elsholtz et al. [28], proposed using a threshold of < 5 mm to denote normal nodes in the facial, parotid, retroauricular, occipital groups for a Node-RADS system that categorises the likelihood of a metastatic node from 1 to 5. Our findings support lowering the threshold from 10 mm to one that is similar to that already applied to retropharyngeal nodes, i.e., 5 mm or 6 mm [9, 13, 20]. Our current results suggest that 6 mm, rather than 5 mm, would reduce the number of false positive nodes and improve specificity in the parotid, occipital, Level IIb nodes, as well as the retropharyngeal group. However, thresholds may need to be adjusted 1- 2 mm higher or lower for submandibular and retroauricular groups respectively and the presence of any node in the facial group. Moreover, size thresholds are a trade-off between sensitivity and specificity and chosen thresholds may need to be adapted to the clinical scenario; for example to improve sensitivity when searching for small metastases in surgical candidates with a clinically N0 neck, thresholds as low as ≥ 4 mm [17, 29, 30] may be used for ultrasound guided fine needle aspiratory cytology of submandibular and upper internal jugular nodes.
Unilateral nodes or nodes larger on one side than the other is often used as a sign to heighten suspicion of a metastatic node in groups along the expected pathway of nodal spread. The results of this study support this practice for unilateral nodes, especially in the jugulodigastric region where only 6.9% were unilateral, compared to 19–28% in the other nodal groups. Interestingly, when nodes were bilateral, they were larger than when unilateral which may represent a general stimulation of nodes, in our group this may possibly be due to EBV infection causing reactive nodes. However, once nodes were bilateral nodes there was no significant difference in size between the right and left sides of the neck, again suggesting that asymmetry in size should be regarded with suspicion.
This study found age negatively correlated with the SAD of nodes in retropharyngeal group, and upper jugular group (jugulodigastric nodes, Level IIa and Level IIb nodes) as shown in previous studies [25, 31]. However, we found no correlation between age and size of nodes in the parotid, submandibular and occipital groups.
Although this study has focused on size criterion, it is worth remembering that the imaging diagnosis of a metastatic node, especially for those small nodes that do not surpass the size threshold, also takes account of other morphological features such as shape, necrosis, heterogeneity, extranodal spread, hilum, vascular pattern and functional features such as F-fluorodeoxyglucose activity and restricted diffusion.
There are some limitations in this study. First, evaluation of nodal groups was limited to those groups consistently covered in all scans of the upper neck, which unfortunately did not include submental nodes. Second, this study only evaluated measurable nodes (a SAD of ≥ 2 mm) which may result in under reporting of the frequency of the nodes in each nodal group. However, this assured the certainty in identifying a node rather than other structures (i.e., small vessels). Third, all patients in this study were from a single institution but this group should represent the expected range in size of benign nodes (normal and reactive). Although this group of patients with persistently elevated plasma EBV-DNA may have a potential bias towards reactive, and hence larger size, benign nodes, this adds to strength to the findings that a reduction in nodal size threshold for detecting malignant nodes should not compromise specificity. Fourth, the influence of outliers on the statistical significance between age and SAD is unknown. Fifth, the diagnostic performance of the SAD thresholds could not be fully assessed because we only evaluated the SAD of benign and not metastatic nodes. Radiology studies that include both malignant and benign nodes with pathological correlation are needed to explore the diagnostic performance of size thresholds for these under reported nodal groups, and this may require multicenter studies as data from a single center may be insufficient for analysis.