Incidence and predictive factors of retropharyngeal lymph node metastases in patients with oropharyngeal cancer undergoing multimodality treatment planning imaging

We investigated the incidence and predictive factors of retropharyngeal lymph node (RPLN) metastases in patients with oropharyngeal cancer (OPC) undergoing multimodality treatment planning imaging before radiotherapy.


| INTRODUCTION
Two main therapeutic options exist for nonmetastatic oropharyngeal carcinomas (OPC), namely radiotherapy (RT) with or without chemotherapy, and surgical excision, for which adjuvant RT and/or chemotherapy is frequently indicated. 1 Both approaches aim to treat the primary tumor and the lymph node regions at risk of metastases. However, in contrast to RT, surgery alone does not routinely treat nor sample the retropharyngeal space, yet retropharyngeal lymph node (RPLN) metastases have been shown to induce a worse prognosis in terms of disease control and survival among patients with OPC. [2][3][4][5][6] It is therefore critical to accurately assess the involvement of this space to guide treatment-related decisions. The retropharyngeal space is defined as the virtual area between the pharyngeal constrictor muscles and the prevertebral fascia, and typically extends cranially from the base of skull to the lower level of C3, caudally. 7 Unless massively involved, pathologic nodes in this area are consequently clinically occult and are not generally accessed surgically due to the imminent presence of the carotid artery. The assessment of this space for metastasis consequently relies on imaging, with reported rates between 9% and 25% for oropharyngeal cancers. 3,5,8,9 Computed tomography (CT) imaging alone is less sensitive for the detection of pathologic nodes, but its combination with magnetic resonance imaging (MRI) or fluorodeoxyglucose-positron emission tomography (FDG-PET) has shown increased sensitivity and overall accuracy. 10 Nonetheless, in the limited pathologic series investigating the incidence of RPLN metastases by surgical sampling or dissection for OPC, the rates were as high as 44%, illustrating the underperformance of current imaging modality protocols. 11,12 Surgical resection alone therefore risks forsaking residual subclinical disease in the retropharyngeal nodes with consequent high risk of cancer recurrence. However, none of the published radiologic studies systematically investigated patients with all three modalities.
As most patients with OPC treated with curativeintent RT in our institution routinely undergo a CT, an MRI, and an FDG-PET before their treatments, the aim of this study was to assess the incidence and predictive factors of RPLN metastasis in patients with OPC systematically undergoing multimodality imaging.

| Study population
All consecutive patients with previously untreated, nonmetastatic pathologically proven infiltrative squamous cell carcinoma (SCC) of the oropharynx planned for curative definitive RT with or without systemic therapy from May 2017 to June 2019 at our tertiary care center were included. Exclusion criteria were prior RT or therapeutic surgery to the head and neck, palliative intent, non-SCC histology, second concurrent head and neck malignancy, or evidence of distant metastatic disease at baseline. Demographic data, tobacco and alcohol consumption details, primary site, tumor extension and histopathology of cancer, tumor-nodemetastasis (TNM) stage according to the 8th edition of the American Joint Committee on Cancer (AJCC) staging system, and treatment planning details were retrospectively collected from electronic medical records to generate an anonymized database for this analysis. The study was conducted according to the Declaration of Helsinki and approved by the institutional ethical review board.

| Imaging evaluation
In our institution, all patients with oropharyngeal cancer planned for curative-intent RT undergo all three modalities (CT, MRI, and FDG-PET) as part of their treatment planning evaluations, unless medically contraindicated. FDG-PET images were acquired in the treatment position at the same time as the planning CT, while the MRI was co-registered thereafter. All treatment planning imaging studies were also interpreted by an expert head and neck radiologist and a nuclear medicine specialist. The planning and diagnostic imaging reports were reviewed for any documented radiologically positive RPLN. Disease extension to adjacent anatomic subsites, the number and laterality of positive RPLN as well as the distribution of other positive cervical nodes were also collected from the planning imaging reports, if applicable. The retropharyngeal nodal region was defined as the area between the pharyngeal constrictor muscles and the prevertebral fascia, extending cranially from the base of skull to the lower level of C3. 7 The distinction between medial and lateral RPLN was not routinely documented in the radiology reports as their respective anatomical limits are controversial in the literature. RPLN were regarded as positive if they had any of the following criteria: maximal axial diameter ≥10 mm, size asymmetry compared to the contralateral size, presence of central necrosis, evidence of extra-capsular spread, clustered RPLN, or hypermetabolic activity on FDG-PET with anatomic correlation to a lymph node on CT. Cervical lymph node distribution (level I-V) was reported according to the 2013 update of the consensus guidelines on neck node levels for head and neck tumors. 7 Low-, intermediate-, and high-risk prognostic subgroups were defined as per Ang et al.'s retrospective analysis of the RTOG 0129 study cohort. 13

| Statistical analysis
The rate of RPLN involvement as well as patient and tumor characteristics were reported using descriptive statistics. Univariate logistic regression was used to evaluate the association between RPLN involvement and patient and disease characteristics. Variables with a p-value lower than 0.10 in the univariate analysis were included Unknown HPV 20 7 in the multivariate analysis using the forward stepwise logistic regression model. A p-value lower than 0.05 was considered statistically significant. All statistical analyses were performed using IBSS SPSS Statistics, version 27 (IBM Corporation, Armonk, NY).

| Descriptive characteristics of the study population
Electronic medical records of 324 patients with OPC planned for curative-intent RT were reviewed and 300 patients met the eligibility criteria. Patients were excluded for the following reasons: concurrent head and neck malignancy other than OPC (n = 2), history of head and neck malignancy previously treated either by surgery or RT (n = 4), non-SCC histology (n = 5), treatment refusal by the patient or treatment with palliative intent (n = 12), and treatment received in another center (n = 1). Table 1 describes patient and disease characteristics for the 300 included patients. Median age was 62 years (range: 40-85 years). P16 tumor status was positive in 82% and unknown in 7% of patients. The majority had a tonsil (55%) or base of tongue (40%) primary, with only rare tumors originating from the soft palate (4%) or posterior pharyngeal wall (1%). All patients underwent a diagnostic CT as part of their initial disease workup, of which 8 patients also had a diagnostic FDG-PET and 1 patient also had a diagnostic MRI. During treatment planning, all patients underwent a planning CT, of which 297 (99%) were contrast enhanced. All but one patient had at least a dedicated planning MRI (n = 278) or a planning FDG-PET (n = 267) in addition to the treatment planning CT, and 246 (82%) had all three imaging modalities ( Table 2).

| Retropharyngeal lymph nodes
On planning imaging modalities, radiologically positive RPLN were reported in 66 patients (22%), of which 4 had no involved cervical lymph nodes, 9 had bilateral positive RPLN, and none had exclusive contralateral RPLN involvement without positive ipsilateral RPLN (Table 2). Of the four patients with a solitary ipsilateral RPLN, two had a primary soft palate tumor, and the two others had a tonsil primary, one of which invaded the soft palate. Eight of the nine patients with bilateral positive RPLN also had bilateral cervical lymph nodes. Most RPLN-positive patients had a single RPLN (n = 50), while 12 and 4 patients had two and three RPLNs, respectively. Figure 1 presents an example of a patient with two RPLN metastases reported on the planning CT, MRI, and FDG-PET, although better visualized on the MRI. Only 21 patients (7%) had reported RPLN metastases on diagnostic images acquired during their initial disease workup, of which 17 (6%) were detected on the diagnostic CT, while 4 were detected by a diagnostic FDG-PET (n = 3) or MRI (n = 1). RPLN metastases were therefore detected in 49 additional patients (16%) during treatment planning compared to diagnostic CT alone, of which 48 were reported on the planning CT or MRI and 21 on the planning FDG-PET (Table 2). Median time between diagnostic CT and treatment planning images was 52 days for patients with RPLN already detected on the diagnostic CT and 58 days for patients upstaged during treatment planning.
Results of the univariate and multivariate forward stepwise logistic regression are shown in Table 3

| DISCUSSION
To the best of our knowledge, this study is the first published experience of multimodality treatment planning imaging with CT, MRI, and FDG-PET for the detection of RPLN metastases in patients with OPC. With this approach, we detected an additional 16% RPLN-positive patients not detected on routine diagnostic CT, for a total rate of 22% among this population, highlighting the importance of MRI and FDG-PET in treatment planning. Several groups have described the incidence of RPLN metastases when combining two of the three stated modalities, with a rate of involvement ranging from 9% to 25%. 3,5,8,9 Of note, Iyizoba-Ebozue et al. conducted a similarly designed study which revealed a 10% rate of RPLN in patients with OPC, and posterior pharyngeal wall/soft palate primaries and contralateral cervical lymph node involvement were significantly associated on multivariate analysis. 14 However, only 10% of their patients underwent all three imaging modalities, compared to 82% in our study population. The sensitivity of multimodality imaging for the detection of RPLN was elegantly studied by Kim et al., who reviewed preoperative images of patients with head and neck cancer who underwent RPLN dissection during their curative-intent surgery. 10 The dissections revealed a 35% rate of RPLN among their population with a sensitivity of 90% for trimodality imaging compared to 83%, 74%, and 65% for FDG-PET, MRI, and CT alone, respectively. However, 15% of their cohort were recurrent cancers, 46% had nonoropharyngeal carcinomas, and only 60% of their patients benefited from the three imaging modalities. The study design was also fraught by potential selection bias since they specifically selected patients undergoing an RPLN dissection. In contrast, the present report was exclusively composed of patients with previously untreated OPC and included all consecutive patients treated with curative-intent RT during the study period, thereby limiting the potential selection bias associated with retrospective designs. Another strength of our study is that most imaging investigations were acquired in the neutral RT treatment position, enhancing the reproducibility and quality of the images for the evaluation of the involved nodal levels and anatomic extension to adjacent subsites.
In this cohort of patients, RPLN involvement was independently associated with tonsil, soft palate and posterior pharyngeal wall primaries, and with disease extension to the soft palate or vallecula. Lin et al. also found an association between tonsil primaries and RPLN metastases on univariate analysis, 5 while other groups did not. 2,14-16 A higher risk of RPLN metastases in patients with soft palate or posterior pharyngeal wall primaries  was also previously described. 14,15 From our research, it seems that we are the first to describe the extension of the primary tumor into adjacent anatomic subsites and its association with RPLN positivity. This association may be particularly important to assess for locally advanced tumors, for which the primary anatomic subsite alone may not be as declarative, considering the multiple secondary subsites involved. It is also important to adequately assess the retropharyngeal space in the era of treatment de-escalation and individualization for p16-positive OPC. Indeed, a phase II study of de-intensified RT for lateralized p16-positive OPC, sparing the contralateral RPLN and delivering a lower dose of 43.2 Gy in 24 fractions to the elective neck, showed excellent oncologic outcomes and rapid recovery of patient reported outcomes with this novel treatment strategy. 17 The randomized SHORT-OPC trial, investigating a lower elective dose of 40 Gy in 20 fractions with a stereotactic ablative RT boost of 14 Gy in 2 fractions to the macroscopic tumor, and the HN10 trial, evaluating elective volume-adjusted de-escalation RT, are two additional ongoing phase II trials omitting contralateral RPLN irradiation for patients with p16-positive OPC. 18,19 Furthermore, the 2019 international guidelines on lymph node target volumes selection by Biau et al. suggest elective coverage of the ipsilateral retropharyngeal space for patients with clinical N2 or N3 nodal disease, and bilateral coverage for posterior pharyngeal wall tumors. 20 A better comprehension of the incidence and predictive factors of RPLN metastasis could therefore guide treatment planning, notably by establishing a subset of low-risk patients for whom elective coverage of the retropharyngeal space could be partially or entirely omitted, thereby reducing long-term toxicities, especially chronic xerostomia. 21 On the other hand, positive RPLNs are usually treated to a higher dose of RT, further highlighting the importance of their detection. 1 Adequate pre-treatment assessment of the retropharyngeal space is equally crucial for patients eligible for surgery, as many now undergo transoral robotic surgery (TORS) or transoral laser microsurgery (TLM) as their sole treatment, which have shown comparable oncologic outcomes to RT for early-stage patients. 22,23 The identification of RPLN metastases in such patients would greatly affect the treatment plan as it would generally preclude these surgical approaches as a single therapy.
Despite our attempt at limiting selection bias by including consecutive patients treated in our department, and although the population consisted of patients of all disease stages, only RT-treated patients were included in the current study, consequently limiting the extrapolation of our conclusions to surgically treated patients. Another limitation of this study is the potential temporal bias associated with the time interval between the diagnostic and planning images, although relatively short in this study, which could theoretically allow for some disease progression. The isolated added value of the planning MRI for RPLN detection could also not be clearly established, as the planning CT and MRI were interpreted together and documented as a single report by the radiologists.
Finally, this study should be interpreted in light of its monocentric retrospective design and limited sample size. Larger prospective comparative trials are needed to better assess the role of multimodality imaging and the patient subgroups at higher risk of RPLN, since it remains logical that all imaging modalities will continue to underperform in the detection of microscopic RPLN involvement compared to histopathological series.

| CONCLUSIONS
Multimodality imaging with CT, MRI, and FDG-PET at the time of treatment planning reveals a 22% rate of RPLN metastasis in patients with OPC, compared to 6% for diagnostic CT alone. Patients with tonsil, soft palate or posterior pharyngeal wall primaries or disease extending to the soft palate or the vallecula appear at higher risk of RPLN involvement.