Reviewing patient records revealed that 22 patients (73.3%) had CT scans, while 8 (26.7%) underwent MRI. Among 30 patients, 21 (70.0%) had suspicious nodes; 9 (30.0%) did not. Out of 21 patients with suspected lymph nodes, Level I involvement was found in 18 cases (85.7%), while Level II was seen in 10 cases (47.6%), and Level III appeared in just 2 cases (9.5%). In 8 patients (38.1%), suspicious nodes were present across multiple cervical levels. Characterization of the radiological features of the 38 suspicious nodes showed 28 (73.6%) nodes had diameters greater than 9 mm, 2 (5.3%) nodes had a round shape, 6 (15.8%) nodes had abnormal hilum architecture, and 2 (5.3%) nodes were matted. The lymph node size ranged from 0.0 - 62 mm with a mean size of 12.7 mm (±13.3 SD), a median of 11.5 mm and a mode of 0.0 mm. Some of the patients exhibited more than one evaluation criterion. The radiological nodal categorization had 9 (30.0%) patients with cN0 followed by 8 (26.7%) and 7 (23.3%) patients with cN2b and cN1 respectively in Table 3. The radiological features of the patients are summarized in while the distributions of suspicious nodes by levels are summarized in Figure 1. The histological nodal classification indicated that 14 (46.7%) patients were categorized as pN0, while 6 (20.0%) patients were classified as pN3b. Additionally, 3 (10.0%) patients each were staged as pN1, pN2b, and pN2c. Only 1 (3.3%) patient was staged pN3a. The histological features of the patients are summarized in Table 4.

Table 3. Summary of radiological nodal assessment.

Figure 1.Number of histologically identified and involved lymph nodes by levels.

Table 4. Summary of histological assessment.

Out of a total of 343 lymph nodes identified during neck dissection for histopathological examination, 59 were confirmed to be positive for tumor. A paired t-test was conducted to compare the number of lymph nodes dissected out and the number of metastatic lymph nodes confirmed on histology. There was a statistically significant difference in the number of lymph nodes dissected out (M = 11.43, SD = 6.83) and the number of metastatic lymph nodes (M = 1.97, SD = 2.81); t (29) = 7.349, p < 0.001 from the patients. The effect size was large, with a Cohen’s d of 1.81, indicating that more than 96% of the number of lymph nodes involved would be below the average number of lymph nodes dissected. The comparison of the number of lymph nodes identified and involved is summarized in Table 5.

Table 5.Comparison of the number of lymph nodes identified and involved.

Note: Paired t-test was applied. *: The mean difference is statistically significant at the level of 0.05.

Histopathological evaluation of the number of nodes dissected per level showed Level I {146 (42.6%)} had the most nodes, followed by Level II {106 (30.9%)}, then Level III {62 (18.1%)} and Level IV {29 (8.4%)}. Of the 30 patients, 16 (53.3%) were positive for tumor in the cervical lymph nodes, while 14 (46.7%) did not exhibit any lymph node involvement. Analysis of the 16 patients with metastatic nodes showed 14 (87.5%) had clinically suspicious nodes but 2 (12.5%) were clinically negative nodes prior to surgery. Of the 16 (53.3%) histologically positive cases, the most common Tumor (T) categorization was T4a with 11 (68.8%) cases followed by T3 with 2 (12.5%) cases while T1, T2, and T4b had 1 (6.3%) case each. The distribution of Tumor (T) categorization among the histologic positive cases is summarized in Figure 2.

Figure 2.Distribution of radiological features of suspicious nodes by levels.

The histological nodal categorization had 14 (46.7%) patients at pN0 followed by 6 (20.0%) patients staged pN3b. pN1, pN2b and pN2c had 3 (10.0%) patients each. Only 1 (3.3%) patient was staged pN3a. Sixteen patients (53.3%) were found to have histologically confirmed tumours within their cervical lymph nodes (pN+). These histopathological characteristics of the nodes were described as follows; 14 patients had nodes > 9 mm in diameter (radiographically), 9 had with abnormal hilum architecture with 6 showing extra-nodal extension. Within this group of pN+ patients, the predominant nodal classification was N3b, identified in six patients Table 4.

Out of 59 metastatic lymph nodes, 44 (74.6%) were located in Level I, followed by 11 (18.6%) in Level II, 3 (5.1%) Level III and 1 (1.7%) in Level IV 3 (5.1%). On characterization of the 59 involved lymph nodes, 30 (50.8) nodes had abnormal hilar architecture, 24 (40.7%) nodes were greater than 9mm in diameter, 5 (16.7%) nodes were matted. The distribution of the metastatic nodes per criteria and cervical level is summarized in Table 4.

Due to the small sample size of 30, Cohen’s Kappa (κ) test was used to determine the patterns of agreement based on matched (paired) cases for the study. There was fair agreement between the two groups of patients, κ = 0.384, p < 0.05. The pattern of agreement between patients with suspicious nodes and patients with involved lymph nodes is summarized in Table 6. Cohen’s κ was run to determine the pattern of agreement between levels of suspicious nodes and levels of involved lymph nodes. There was a moderate agreement between the two groups of levels of nodes, κ = 0.512, p < 0.05. The pattern of agreement between clinically suspicious and histologically confirmed involved lymph nodes is summarized in Table 7. Cohen’s κ was run to determine the pattern of agreement between radiological and pathological nodal categories. There was a substantial agreement between the two groups of nodal categories, κ = 0.629, p < 0.05. The pattern of agreement between radiological and pathological nodal categories is summarized in Table 8.

Table 6. Pattern of agreement between patients with clinically suspicious nodes and patients with histological confirmation of involved lymph nodes.

Note: Cohen’s Kappa (κ) test was applied. *: Cohen’s Kappa (κ) is significant at the level 0.05.

Table 7. Pattern of agreement between cervical levels with suspicious and involved lymph nodes.

Note: Cohen’s Kappa (κ) test was applied. *: Cohen’s Kappa (κ) is significant at the level 0.05.

Table 8. Pattern of agreement between radiological and pathological nodal categories.

Note: Cohen’s Kappa (κ) test was applied. *: Cohen’s Kappa (κ) is significant at the level 0.05.

The sensitivity, specificity, false positives and false negatives of the two diagnostic tests (MRI and CT scan) were compared with the histopathological results as the gold standard. Evaluation of the MRI results showed a sensitivity (true positive) rate of 100.0%, a specificity (true negative) rate of 60.0%, a false positive rate of 40.0% and a false negative rate of 0.0%. A McNemar’s exact test determined that the difference in the proportions of MRI positive results and histological results was not statistically significant, p = 0.500.

Evaluation of the CT Scan results showed a sensitivity (true positive) rate of 83.3%, a specificity (true negative) rate of 44.4%, a false positive rate of 55.6% and false negative rate of 16.7%. A McNemar’s exact test determined that the difference in the proportions of CT scan positive results and histological results was not statistically significant, p = 0.453. A comparison between the two imaging modalities showed that there was a difference of 16.7% sensitivity rate between MRI (100.0%) and CT scan (83.3%). A McNemar’s exact test determined that the difference in the proportion of positive radiological results and histological results was not statistically significant, p = 0.180. The sensitivity, specificity, false positives and false negatives of the radiological investigations are summarized in Table 9.

Table 9. Sensitivity, specificity, false positives and false negatives of the radiological investigations.

Note: A McNemar’s exact test was applied.

This study found that CT scan was the most common radiological modality requested for assessing cervical lymph node metastasis. This was similar to other studies by Horváth et al. and Thoenissen et al. [9][18]. The preference for CT scans may partly stem from their greater availability, lower cost, and quicker procedure time when compared to MRI.

Approximately 30% of the patients did not have radiological evidence of cervical lymph node metastasis (cN0) but still underwent neck dissection to rule out occult metastasis. Previous studies have shown the prevalence of this prophylactic neck dissection to range from 31% to 60% [11][18]-[21]. Elective neck dissection is supported by evidence of occult metastasis from previous studies [11][22][23].

Cervical Level I had the greatest number of suspicious lymph nodes. This was similar to a prospective study by Narayana et al., of 24 patients which found Level I (combined Ia and Ib) to be the most prevalent suspicious cervical level. It is well demonstrated that level I has the most sentinel lymph nodes for primary tumors located in the floor of the mouth [24][25]. Thus, meticulous clinical assessment of Level I is very important [25].

In this study, the most common radiological feature of suspicious lymph nodes identified was an enlarged node of more than 9 mm in diameter. Most studies advocate for assessment criteria based on a combination of nodal size, architecture and signs of extra nodal spread like matted nodes [26]-[28]. Relying on size criteria for diagnosis of clinical cervical lymph node metastasis reduces the accuracy of CT scan to 45% compared to 95% - 100% accuracy when based on central necrosis [29]. This aspect is important in this study given that calculation of sensitivities and specificities was one of the objectives.

The most frequent clinical nodal categories in this study were cN2b and cN1. This differed from a German retrospective study of 242 patients by Voss et al. in 2022 which found cN1 to be the most prevalent clinical nodal category [20]. The higher nodal categorization in this current study could be due to the higher number of patients with higher T categorization. It could also be due to factors associated with delays in diagnosis of oral cancer, especially in developing countries [30].

In the current study, level I had the highest number of positive lymph nodes confirmed on histology. Several previous studies found similar results [11][21][31][32]. Thoenissen et al., found a near equal prevalence between Levels I and II (9). On the other hand, Nithya et al., when looking specifically at carcinoma of the tongue, found level II to be most commonly involved [33]. Levels I and II are known sentinel lymph nodes of primaries from the oral cavity [24]. These levels have to be thoroughly dissected out during neck dissection.

Almost half the patients who underwent neck dissection in the current study did not have cervical lymph node metastasis. Previous studies support this finding [11][23][31]-[35]. In contrast, Qiao et al., and Mehta et al., in retrospective studies found a lower prevalence of 30% and 20% respectively [19][31]. As demonstrated by Kligerman et al. in a randomized controlled trial of 67 patients with stage 1 and 2 OSCC of the floor of the mouth and tongue, survival rate is better when elective neck dissection is done [36].

In this study, the most common histopathological feature of positive lymph nodes was abnormal hilum architecture. Pandeshwar et al., found most metastatic cervical lymph nodes to have central necrosis. Presence of tumor distorts the architecture of the lymph node by causing necrosis, deposition of keratin pearls, among others. Most of these architectural changes can be seen on radiological examination and inform their assessment and subsequent clinical staging [21].

A third of all the positive nodes in this study had extra nodal extension. The prevalence of extra nodal extension in other studies ranges from 24% to 45% [20][32][37]. Extra nodal extension lowers the prognosis in OSCC [35]. It is recommended that adjuvant chemotherapy be administered after neck dissection in patients with extra capsular spread [38].

The most prevalent pathological nodal category in this study was pN3b. This was similar to studies by Rabie et al., and Voss et al., [20][37]. This, however, contrasted to the study by Thoenissen et al., who found N1 and N2b to be most prevalent [9]. N3b was introduced as part of TNM staging in the AJCC 8^th edition of 2018 and may not be captured in research done prior to 2018 [14]. N3b denotes extra nodal extension and has poor prognosis [35]. AJCC recommends adjuvant chemotherapy for N3b [14].

In this study, there was a fair agreement between patients with clinically suspicious nodes and the patients with histologically confirmed nodal metastasis. This low pattern of agreement could be due to the overreliance on size criteria in identifying suspicious nodes on radiology. Assessing lymph node architecture on imaging, rather than size alone, may improve detection of malignant invasion. The level of agreement increased to moderate when the unit of comparison was the cervical nodal level. The agreement increased to substantial when the comparison was between clinical (cN) and pathological (pN) nodal categories. This suggests that ultimately, the clinical (radiological) nodal assessment in TNM staging, which considers a combination of size, numbers, laterality and extra nodal extension, is an effective tool in predicting lymph node metastasis.

In this study population, CT had a sensitivity (true positive) of 83.3%. This was within the range of 52% to 83% found in other similar studies. However, the 44.4% specificity of CT scan in this study was lower than the range of 68% - 98% from other studies [9][26]-[28][39]. A common factor in the studies by Suryavanshi et al., Sumi et al., and Saafan et al., was their use of three or more criteria in assessing cervical lymph node metastasis (Central necrosis with peripheral enhancement, conglomeration of three or more lymph nodes and short axial diameter size criteria) [26]-[28]. The lower ability to exclude metastasis (specificity) in this study could be due to the overreliance on the size criteria. As a result, overtreatment of the neck may occur, leading to a considerable number of patients undergoing unnecessary neck dissections and experiencing related complications and increased morbidity.

The sensitivity (true positive) of MRI in this study was higher at 100%. The range observed in other studies was between 66% - 81%. On the other hand, the specificity of MRI in this study was 60%. This was lower than other studies which ranged from 68% to 80% [9][39]. The wide variation in sensitivity and specificity of MRI in this study could be due to the smaller number of patients who had MRI as their radiological investigation before surgery.

In this study, the false positive rates were 55.6% and 40% for CT and MRI, respectively (cumulatively 50% false positive for radiological assessment). This implies approximately half of the patients without metastatic nodal disease were found to have been falsely categorized as positive on radiological assessment. Other studies have shown false positive rates of 2% - 32% from radiological assessment. The relatively higher false positive rate in this study correlates to the lower specificity of CT and MRI found. On the other hand, the false negative rates in this study were 16.7% for CT and none for MRI. This is similar to previous studies which found a false negative rate of 17% - 48% [9][26]-[28][39]. This study’s small sample size means that the comparison between CT and MRI performance was not statistically significant and does not confirm an actual lack of difference between the two modalities.

Our study had a few limitations. First, the sample size was relatively small. However, this was similar to other cross-sectional studies where data was collected before and after surgery [25][26]. Secondly, there was possibility of bias in the reporting of the radiological images. This was mitigated by having an independent radiologist re-assess the radiological images for inter observer variability. Thirdly, the wide variation in MRI sensitivity and specificity could be due to the small number who were had an MRI investigation. Fourthly, the surgeries were performed by different surgeons thus raising the possibility of different qualities of neck dissection. The principal investigator was present at all neck dissections to ensure they followed ASCO guidelines [38].