High sensitivity and negative predictive value of sentinel lymph node biopsy in a retrospective early stage oral cavity cancer cohort in the Northern Netherlands

In cT1‐2N0, oral squamous cell carcinoma (OSCC) occult metastases are detected in 23%‐37% of cases. Sentinel lymph node biopsy (SLNB) was introduced in head and neck cancer as a minimally invasive alternative for an elective neck dissection in neck staging. Meta‐analyses of SLNB accuracy show heterogeneity in the existing studies for reference standards, imaging techniques and pathological examination. The aim of this study was to assess the sensitivity and negative predictive value (NPV) of the SLNB in detecting occult metastases in cT1‐2N0 OSCC in a well‐defined cohort.


| INTRODUCTION
Regional metastases occur in 23%-37% of the early stage (cT1-2N0) oral squamous cell carcinomas (OSCC). [1][2][3] Lymph node status is an important prognostic factor for outcome and treatment decision-making of head and neck cancer. [1][2][3][4][5][6][7][8] However, not all metastases are clinically detectable with the current diagnostic modalities. [9][10][11] Occult metastases are conventionally treated by removal of the lymph nodes by elective neck dissection (END) after research showed higher rates of overall and disease-specific survival compared to a watchful waiting strategy. 12 However, an END has disadvantages: it leads to overtreatment in 63%-77% of the cases and has a risk of postoperative comorbidity (eg shoulder pain, reduced limb movement). 13 Therefore, there is a need for a better neck staging modality.
The sentinel lymph node biopsy (SLNB) was introduced in oral cavity cancer as a less invasive lymph node staging technique after successful implementation in melanoma and breast cancer. 5 The limited number of lymph nodes (LN) with the SLNB enables a more meticulous pathological examination incorporating step serial sectioning (SSS) and additional immunohistochemistry (IHC). 14 Recently, Liu and Wang reported a meta-analysis of 3566 early stage OSCC patients from 66 studies with a pooled sensitivity of 87% and negative predictive value (NPV) of 94% for SLNB in detecting occult metastasis. 15 However, many of these studies consist of small cohorts and differ in reference treatment, SLNB localisation technique (eg use of gamma probe, blue dye or single photon emission CT (SPECT-CT)) and pathological work-up (with or without IHC or SSS). Furthermore, several studies provide incomplete clinico-pathological information. This heterogeneity and lack of complete data underline the need for more studies using complete and homogeneous cohorts. The aim of this study was to determine the sensitivity and NPV of the SLNB in detecting occult metastases in a large, well-defined cohort. For this purpose, we used a retrospective cT1-2N0 OSCC cohort of 91 patients all treated by primary surgical resection, neck staging with the SLNB procedure and routine follow-up as reference standard for the SLNB negative neck.

| Ethical consideration
Sentinel lymph node biopsy was part of standard treatment and data were retrospectively gathered from existing data sources; therefore, no approval from the hospital research ethics board was required according to the Dutch ethical regulations. 16 Detailed information about the patient selection and the SLNB procedure are added to this manuscript as Appendix S1.
Briefly, inclusion criteria were as follows: clinically T1-2 and N0 staged OSCC (7th TNM classification); primary treatment by surgical resection and neck staging by SLNB. Clinico-pathological data of the 91 (100%) patients were retrospectively collected from the digital patients files (Table 1). Cases with a positive SNLB underwent a modified radical neck dissection (MRND) during a second surgery. Routine follow-up of the neck was used as reference standard in the SLNB negative patients and consisted of physical examination that was followed by ultrasound fine needle aspiration cytology (USFNAC) in case of enlarged (>1 cm) or otherwise suspicious lymph nodes.

| Study procedure
The SLNB procedure was described in detail before and was mostly the same in both centres. 18 Lymphoscintigraphy and SPECT/CT scans were made 1 day before surgery. Intraoperatively, SLNs were harvested after gamma probe assisted localisation. 18 SLNs were histopathologically examined by SSS with an interval of 500 lm and additional pan-cytokeratin antibody (AE 1/3) immunohistochemistry staining. Additional lymph nodes (non-SLNs) were harvested if they blocked the SLN or formed a conglomerate with the SLN.

| Statistical analysis
IBM SPSS Statistics 23 (Statistical Package for the Social Sciences, Inc., Chicago, IL, USA) was used for analysis. Categorical data are presented as number (n) and their percentages (%). Associations between categorical data were tested with the Fisher's exact or Chi-squared test. Continuous data were tested using the Student's t test or the Mann-Whitney U test for normally or skewed distributed data, respectively. False negative SLNB patients were defined as patients with isolated regional recurrence in the SLNB negative neck side and were used to calculate the sensitivity and negative predictive value. Significant differences were defined as a P-value ≤.05.

Keypoints
• The sentinel lymph node biopsy detected occult metastases in our cN0 cohort with an 85% sensitivity and a 94% negative predictive value.
• No additional metastases were seen in the neck dissection lymph nodes of patients with micrometastases or isolated tumor cells in the sentinel lymph nodes.
• The sentinel lymph node biopsy reveals unexpected lymphatic drainage patterns and therefore prevents patients for undertreatment of the neck.   Table 1. In all patients, at least one SLN was intraoperatively detected. However, in 4 patients (4%), additional hotspots were noticed besides the harvested SLNs on the SPECT-CT without intraoperative detectable radioactive LNs. In 1 of these 4 patients, the harvested SLN was positive and the neck was treated by MRND in a second operation. The other 3 patients were isolated regional recurrence (IRR) free after 10, 11 and 47 months of routine follow-up. In 1 patient with a ventral floor of mouth tumour, only a contralateral SLN was identified. The other patients had ipsilateral (n = 57, 63%) or bilateral (n = 33, 36%) located SLNs.
Positive SLNs were found in 25 (27%) patients. In 1 patient with a 1 mm metastasis in the SLN routine follow-up was chosen instead of a MRND. This patient was still recurrence free after 23 months.
In none of the patients with micrometastases or ITCs in the SLN, additional metastases were found in the MRND specimen ( Figure 1, Table 2, P = .024). Also, none of the 57 non-SLNs harvested during the SLNB were positive. Finally, skip metastases were not seen: all patients with positive SLNs had at least one positive SLN in level I-III. Infiltrative tumour border configuration (P = .008) and pT2 tumour stage (P = .036) showed an association with lymph node status (Table 1).

| Follow-up and regional recurrence
Overall the median FU was 32 months (IQR 21-47, Range 2-104, Table 1). All patients with a follow-up <10 months died. In total, 8 (9%) patients of this cohort died. Three patients died of disease, two 10 months and one 21 months after the initial treatment. Group too small to analyse.
Local recurrence and second primary tumours, with or without regional recurrence, were seen in 9 (10%) cases. Isolated regional recurrence was detected in 5 (5%) patients. One of these patients had IRR after a positive SLN and subsequent neck dissection at that neck side. The other 4 patients were diagnosed with IRR after 4, 6, 9 and 19 months. Their tumour, treatment and recurrence characteristics are shown in Table 3 Due to the four IRRs, the SLNB detected occult metastases with 85% sensitivity and 94% NPV.

| Synopsis of key findings
In our retrospective cohort of 91 patients treated for cT1-2N0 OSCC, 4 patients developed isolated regional recurrence on the side of a negative SLNB. This resulted in 85% sensitivity and 94% negative predictive value.

| Comparison to previous studies
The sensitivity and NPV are in agreement with the results of other studies with routine follow-up as a reference: sensitivity range 80%-94% and NPV range 88%-97.5% (number of patients 59-415). 1,[6][7][8]19 A recent meta-analysis also showed comparable results: sensitivity 87%, NPV 94%. 15 The slightly higher NPV of this cohort compared to these meta-analyses can be explained by the relative short fol- with other studies. 6,8 Retrospectively, the reason for missing these regional metastases remains unclear; shine-through phenomenon and aberrant lymphatic drainage due to metastatic tumour in the SLNs might be involved. Another possible explanation might be micrometastases in lymph nodes, other than the SLN (skip metastases).
Other studies reported a lower sensitivity of the SLNB procedure in FOM tumours compared to other oral cavity subsites due to the shine-through phenomenon. 7,8,21,22 One patient in this study had a FOM with an IRR resulting in an 80% sensitivity and a 96% NPV for FOM tumours. Retrospectively, this SLNB was overlooked because of this shine-through phenomenon ( Figure 2). To overcome shinethrough and subsequent regional recurrences, Stoeckli et al 23  , with a lymph node within the tumour hotspot (B) and an isolated regional recurrence after 19 mo (C) T A B L E 3 Characteristics of the 4 patients with isolated regional recurrence

Variables
Patients with isolated regional recurrence  27,28 In this study, intervals of 500 lm were used, because our head and neck SCC protocol was adapted from our vulvar SCC SLNB protocol. This is a protocol we have much experience with and has shown to provide accurate staging of vulvar SCC in our centre. [29][30][31] Besides this, the accuracy we found is comparable to that of most head and neck SLNB studies. 15  with multiple comorbidities, a second operation with general anaesthesia is undesirable due to a higher complication and mortality chance. 33 Moreover, in all positive cases, scar tissue makes the neck dissection surgery more challenging in the SLN levels. To avoid repeat surgery, the possibility of intraoperatively staging of SLNs with frozen sections has been studied. 34 However, frozen sections have a substantial false negative rate; therefore, frozen sections of the SLNs are not applied in our centres. Also, a substantial amount of the SLN is lost for the FFPE sections and thereby increasing the risk of missing ITCs and micrometastases. 34 In an ideal situation, patients at high risk of lymph node metastases are preoperatively selected for MRND or watchful waiting. In the current study, an infiltrative tumour border configuration or a pT2 tumour was significantly associated with more regional metastases. Our research group reported earlier infiltration depth and lymphovascular invasion as independent predictors for nodal status in pT1-2N0 and N-status determination by routine HKD and watchful waiting. 35 These markers are not associated with positive lymph nodes in this study. The lack of significance could be explained by the difference in patient selection between the mentioned study by Melchers (cN0 and cN+) and this study (cN0). 35 Therefore, the SLNB procedure is still more accurate in detecting occult metastasis in cT1-2N0 OSCC than the current clinical and pathological markers. In addition, it would be interesting to study the prognostic value of OSCC lymph node status associated biological markers such as WISP1, RAB25 or EpCAM in cT1-2N0 OSCC SLNB staged patients. [36][37][38]

| Study limitations
Limitation of this study is that the SLNB procedure was not part of the standard workflow for cT1-2N0 OSCC patients in the first years after introduction. If we analyse the accuracy without the 6 patients from this period, the sensitivity and NPV are still 85% and 94% respectively.