We performed this study to define the incidence of radiographic retropharyngeal lymph node (RPLN) involvement in oropharyngeal cancer (OPC) and its impact on clinical outcomes, neither of which has been well established to date.
We performed this study to define the incidence of radiographic retropharyngeal lymph node (RPLN) involvement in oropharyngeal cancer (OPC) and its impact on clinical outcomes, neither of which has been well established to date.
Our departmental database was queried for patients irradiated for OPC between 2001 and 2007. Analyzable patients were those with imaging data available for review to determine radiographic RPLN status. Demographic, clinical, and outcome data were retrieved and analyzed.
The cohort consisted of 981 patients. The median follow-up was 69 months. The base of the tongue (47%) and the tonsil (46%) were the most common primary sites. The majority of patients had stage T1 to T2 primary tumors (64%), and 94% had stage 3 to 4B disease. Intensity-modulated radiation therapy was used in 77% of patients, and systemic therapy was administered in 58% of patients. The incidence of radiographic RPLN involvement was 10% and was highest for the pharyngeal wall (23%) and lowest for the base of the tongue (6%). RPLN adenopathy correlated with several patient and tumor factors. RPLN involvement was associated with poorer 5-year outcomes on univariate analysis (P<.001 for all) for local control (79% vs 92%), nodal control (80% vs 93%), recurrence-free survival (51% vs 81%), distant metastases-free survival (66% vs 89%), and overall survival (52% vs 82%) and maintained significance on multivariate analysis for local control (P = .023), recurrence-free survival (P = .001), distant metastases-free survival (P = .003), and overall survival (P = .001).
In this cohort of nearly 1000 patients with radiographic RPLN adenopathy in OPC, RPLN involvement was observed in 10% of patients and indicates a negative influence on disease recurrence, distant relapse, and survival. In this cohort of nearly 1000 patients investigating radiographic RPLN adenopathy in OPC, RPLN involvement was observed in 10% of patients and portends a negative influence on disease recurrence, distant relapse, and survival. Cancer 2013;119:3162–3169. © 2013 American Cancer Society.
The retropharyngeal lymph node (RPLN) group, as initially described by Rouvière in 1936, lies within the retropharyngeal space, medial to the carotid artery, and anterior to the prevertebral musculature and fascia of the uppermost cervical vertebrae. Rouvière's anatomic dissections revealed that the most superior and lateral RPLNs (now referred to as “nodes of Rouvière”) received afferent lymph drainage from the pharynx and other sites, with efferent drainage to the upper jugular lymph node chain. The clinical importance of positive RPLNs in pharyngeal cancers was initially described by Ballantyne, and the presence of pathologic RPLNs has since been accepted as a harbinger of poor outcomes in surgically managed patients. More recently, modern series describing the use of intensity-modulated radiation therapy (IMRT) in the treatment of head and neck cancers highlighted patterns of failure with an emphasis on the importance of adequate radiation therapy target volume delineation pertaining to RPLN target coverage to avoid additional RPLN failures.[3, 4]
Despite the RPLN group being described as a nodal bed at risk for spread from oropharyngeal cancer (OPC), there are only a few reports in the medical literature that have documented the incidence of RPLN involvement in OPC,[5-7] and among these, the impact of RPLN involvement on oncologic outcomes has been ambiguous.[8-10] Because clinical RPLN involvement from OPC is a reportedly infrequent occurrence, and because RPLNs are neither readily amenable to direct clinical examination nor included in routine neck dissections, a large modern cohort of OPC patients with long-term follow-up initially staged with advanced imaging modalities inclusive of the skull base is necessary to accurately estimate both the incidence of clinical/radiographic RPLN involvement and subsequent impact on clinically relevant oncologic outcomes with sufficient power.
The goals of the present study were to 1) identify the benchmark incidence of radiographic RPLN involvement in OPC (and OPC subsites) within a large, modern radiotherapy database, 2) identify potential patient- and disease-specific correlates of radiographic RPLN adenopathy, 3) delineate potential endpoint differentials associated with radiographic RPLN involvement, and 4) generate testable hypotheses for future prospective studies.
The database maintained by the Department of Radiation Oncology at The University of Texas MD Anderson Cancer Center was searched to identify patients irradiated for OPC (squamous cell, poorly differentiated, undifferentiated, or not otherwise specified) between 2001 and 2007. Our institutional review board granted permission to conduct this retrospective study. The search (limited to patients with stage 1 to stage 4B cancer) identified 1026 medical records. Patients were excluded for the following reasons: concurrent malignancies (exclusive of a second malignancy of the oropharynx) at the time of diagnosis or history of any malignancy (excluding nonmelanomatous skin cancer) within 2 years of diagnosis (n = 16), a previously treated malignancy of the head and neck or previous radiation to the head or neck (n = 7), or treatment with chemotherapy prior to staging at MDACC (n = 7). Two patients had records that were not accessible. The medical records of the resultant 994 patients were reviewed to assess patients' demographic, clinical, pathologic, and radiologic data. Patient smoking status was classified as described previously. Because the primary aim of this study was to assess the incidence of radiographic retropharyngeal adenopathy, 13 patients without documented reports of their staging at presentation or stored radiographic images for review were therefore excluded. The final usable cohort for this study was thus 981 patients.
Patient disease was staged according to the American Joint Committee on Cancer (AJCC) 2002 staging system. Charts were reviewed to verify tumor site, size, and sites of invasion. Staging and disease variables of interest included T category, N category, and most caudal cervical nodal level. Patients who were staged Tx were typically those seen post-tonsillectomy and those for whom tumor size could not be determined after record review; these patients were staged T1 for the purpose of AJCC stage grouping. Patients who were staged Nx were those in whom a solitary node was excised for diagnosis and for whom size could not be determined. These patients were coded as N1 for the purpose of this analysis.
The initial pretreatment diagnostic radiology and clinical reports for all 981 patients were then reviewed for any documented radiographically positive RPLN. For those patients with available imaging, the pretreatment diagnostic staging studies were all reviewed to assess for radiographic RPLN involvement. Patients with any of the following RPLN imaging characteristics were considered to have radiographically involved RPLN: minimal axial diameter >5 mm12 or a maximal axial diameter ≥10 mm[13-15]; presence of central necrosis or hypodensity[12-15]; or presence of >1 lateral RPLN or hypermetabolic activity on fluorodeoxyglucose–positron emission tomography (maximum standardized uptake value >4.5). In addition, any identifiable median RPLN on imaging would be considered positive.
A chi-square test was used to compare proportions between subsets. A t test was used for comparison of means. The Kaplan-Meier method was used to calculate actuarial curves. Time of diagnosis was used as timepoint zero. Comparisons between survival curves were made using a log-rank test. Multivariate analysis was performed using the Cox proportional model, and included the following clinical variables: age, sex, smoking status, primary site, T category, N category, cancer stage, lowest involved neck level, and use of chemotherapy.
For this hypothesis-generating dataset, a non-Bonferroni-corrected α<0.05 was considered statistically significant for all measures.
Table 1 details the imaging used for initial staging of the 981 patients. Computed tomography (CT) of the head and neck was used to stage 942 (96%) patients. Among the remaining patients without a dedicated CT scan of the head and neck, 26 (3%) underwent magnetic resonance imaging of the head and neck, and 13 (2%) underwent positron emission tomography, often fused with a noncontrast CT. Overall, 927 (95%) patients had pretreatment scans available for re-review of radiographic RPLN status.
|Imaging Modality||No. (%)|
|CT of the head and heck||942 (96)|
|MRI of the head and neck||91 (9)|
|Ultrasound of the head and neck||130 (13)|
|Chest X-ray||932 (95)|
|CT of the chest||450 (46)|
Patient demographics, tumor sites, and staging are detailed in Table 2. The tonsil and the base of the tongue were the most common primary sites. Nearly two-thirds of patients had stage T1 or T2 primary tumors, but 94% had AJCC stage 3 to 4B disease.
|Characteristics||Entire Cohort (N = 981)||RPLN-Negative (n = 887)||RPLN-Positive (n = 94)||P|
|Mean (range)||56.3 (28–84)||56.2 (31–84)||57.7 (28–81)||.143|
|Men||843 (86)||764 (86)||79 (84)|
|Women||138 (14)||123 (14)||15 (16)|
|Current||227 (23)||191 (21)||36 (38)|
|Former||355 (36)||326 (37)||29 (31)|
|Never||399 (41)||370 (42)||29 (31)|
|Tonsil||447 (46)||396 (45)||51 (54)|
|Base of tongue||461 (47)||433 (49)||28 (30)|
|Soft palate||26 (3)||23 (3)||3 (3)|
|Pharyngeal wall||30 (3)||23 (3)||7 (7)|
|Oropharynxa||17 (2)||12 (1)||5 (5)|
|1||22 (2)||22 (3)|
|2||40 (4)||40 (5)|
|3||171 (17)||169 (19)||2 (2)|
|4A||633 (65)||570 (64)||63 (67)|
|4B||115 (12)||86 (10)||29 (31)|
|1||302 (31)||290 (33)||12 (13)|
|2||324 (33)||299 (34)||25 (26)|
|3||189 (19)||163 (18)||26 (28)|
|4||166 (17)||135 (15)||31 (33)|
|0||105 (11)||105 (12)|
|1–2a||273 (28)||266 (30)||7 (7)|
|2b-2c||516 (52)||445 (50)||71 (76)|
|3||87 (9)||71 (8)||16 (17)|
|Lowest neck level||<.001|
|0||111 (11)||105 (12)||8 (8)|
|1b and 2||510 (52)||472 (54)||37 (39)|
|3||253 (26)||229 (26)||24 (26)|
|4||96 (10)||71 (8)||25 (27)|
|XRT alone||410 (42)||394 (44)||16 (17)|
|ChT+XRT||345 (35)||309 (35)||36 (38)|
|Induction ChT→XRT||104 (11)||93 (11)||11 (12)|
|Induction ChT→ChT+XRT||122 (12)||91 (10)||31 (33)|
All patients in this cohort were irradiated. IMRT was used to treat 77% of these patients. The median radiation dose was 70 Gy (range, 4.4-74 Gy). Fourteen patients (2%) received <50 Gy. Seven of these 14 patients were treated with short palliative radiation schedules, and the remaining 7 patients discontinued therapy due to toxicity, choice, or death. Systemic therapy was used in 571 patients (58%). The mode of chemotherapy (induction and/or concurrent) is described in Table 2. All neoadjuvant regimens were platinum- and taxane-based. cis-Platinum and cetuximab were the most common agents used concurrently with radiotherapy.
Overall, 94 (10%) patients had radiographic evidence of RPLN involvement. All positive RPLNs were lateral RPLNs; there was no median RPLN identified in any patient. Of those patients with radiographically involved RPLNs, 7 had no involved lateral cervical nodes, 8 had bilateral retropharyngeal adenopathy, and 1 had contralateral RPLN involvement without ipsilateral RPLN involvement. Radiographic RPLN adenopathy was lowest in never smokers (7%) and highest in current smokers (16%). RPLN involvement was associated with anatomic primary site; seen in 11%, 6%, 12%, and 23% of patients with primary tumors of the tonsil, base of the tongue, soft palate, and pharyngeal wall, respectively. For those with radiographically positive RPLNs, all but 2 patients had stage 4 disease, and retropharyngeal adenopathy was associated with more advanced primary site (T category). RPLN positivity was also associated with the presence of involved nodes in the lower neck, as 27% of patients with radiographically positive RPLNs also had adenopathy in level 4 compared with 8% of patients who did not have positive RPLNs (Table 2).
The median follow-up time was 69 months. The actuarial 5-year overall survival rate was 82% for patients without radiographically involved RPLNs compared with 52% for those with RPLN adenopathy (P<.001) (Fig. 1). In multivariate analysis the presence of radiographically involved RPLNs was still associated with worse survival (P<.001). However, T category (P<.01), N category (P<.01), primary site (P<.01), smoking status (P<.01), use of chemotherapy (P = .03), and age (P<.01) were also statistically significant. RPLN adenopathy was also associated with higher rates of recurrence (local and nodal) and distant failure (Fig. 2) on univariate analysis. On multivariate analysis, RPLN involvement maintained significance for an association with local recurrence, disease recurrence, and distant failure (Table 3). The actuarial 5-year overall survival rate for never smokers without and with radiographic RPLN adenopathy was 90% and 70%, respectively, while for current smokers without and with radiographic RPLN adenopathy, it was 64% and 37%, respectively (P<.001) (Fig. 3).
|RPLN-Positive||RPLN-Negative||Univariate Analysis||Multivariate Analysis|
|Distant metastases-free survival||66%||89%||<.001||.003|
To establish a benchmark in a modern series of patient outcomes in radiographically positive RPLN cases, we report the incidence of clinical/radiographic RPLN involvement and its subsequent negative impact on disease control and survival in OPC. Cumulatively, the incidence of radiographically involved RPLNs in this cohort of nearly 1000 patients was 10%. Previously, the existing data regarding probability of RPLN metastases in OPC was limited by relatively small cohort size; for example, the largest evident report to date to our knowledge is approximately 440 OPC patients, making generalization difficult owing to sample size and power limitations. Our rates of clinical RPLN adenopathy by OPC subsite are notably similar to those reported by McLaughlin et al. Dirix et al reported a somewhat higher (16%) rate of RPLN involvement in their study of 208 OPC patients, potentially due to geographic/demographic phenomena or at least partially explained by the majority of patients in their cohort having advanced primary tumors (64% T3/4) compared with only 36% in our series, increasing the chances of RPLN positivity.
Similar to others who sought to identify predictors of clinical RPLN involvement in OPC,[5, 10] multiple clinical correlates of RPLN involvement were identified in our cohort. While we did identify a number of patient and disease characteristics that correlated with RPLN positivity (Table 2), of more importance to clinical practice is routine radiographic staging assessment of RPLNs in OPC to ensure detection and adequate treatment of clinically involved RPLNs, for which high-resolution, contrast-enhanced CT-based examination remains the hallmark of our current RPLN screening paradigm. While necessary and quite intuitive, radiographic evaluation of RPLNs can often be problematic compared with lateral cervical lymph node regions since no standard size criteria exist, data regarding radiographic and pathologic correlation are limited, and RPLNs may be continuous with the primary tumor or obscured by dental artifacts on CT, which may be potentially overcome by employing angled images through the oral cavity in such cases.
The presented data, in concert with previous reports across the literature, support the idea that higher RPLN positivity rates are observed for patients with pharyngeal wall primaries.[8, 10] The more common primary OPC subsites had lower rates of RPLN involvement, but clinical radiographic RPLN involvement was still seen in greater than 10% of patients with tonsillar carcinoma and 5% of patients with base of tongue cancer. Therefore, the risk of radiographically negative RPLN regions harboring subclinical microscopic metastatic disease for most OPC cases likely exceeds 5%, which is the threshold value classically used in head and neck cancer to determine whether lymph node regions are included in elective radiation treatment volumes. However, defining the risk of occult microscopic involvement of radiographically negative RPLNs in OPC is beyond the scope of the present study, in which all patients were irradiated and none had RPLN dissection or biopsy prior to radiation. Lack of histologic confirmation and sensitivity and specificity rates of radiographic RPLN status may be seen as a limitation of our study. However, we believe our results are applicable to modern-day routine clinical practice, because clinicians typically rely solely on diagnostic imaging findings when assessing RPLNs. Furthermore, because RPLNs are not commonly biopsied or routinely resected, large cohort imaging-based studies such as ours likely represent the best available evidence on this topic.
We believe that RPLNs must still be considered at clinically meaningful risk for occult microscopic involvement, and thus, even when radiographically uninvolved, still considered important when formulating therapeutic strategies. This point is of particular relevance because there has been a recent resurgence of surgical-based therapy for OPC, which typically does not involve resection of RPLNs and for selected patients often omits radiation therapy as part of a deintensification strategy. Therefore, in such cases, there is potential risk for unappreciated microscopic disease left untreated in RPLN regions, which could lead to subsequent disease recurrence.
The relatively high rate of distant metastatic spread in patients with radiographically involved RPLNs suggests a possible biologic differential in RPLN cohorts. Conceivably, cells that are imbued with the capacity to subvert anoikis, create invadopodia, and penetrate basement membranes have a higher capacity for both nodal dissemination as well as metastatic invasion. Consequently, phenotypic differences in the RPLN cohorts may be reflected in differences in distant relapse rates. The relative mortality decrement associated with RPLN involvement is not, however, uniformly noted across similar RPLN series. For instance, in contrast to the original report from Ballantyne, a more recent series of mostly OPC patients from Gross et al found no difference in outcomes for surgically treated patients with or without pathologic RPLN involvement. They hypothesized that any negative influence from RPLN adenopathy may have been overcome by multimodal therapy, as most received postoperative radiation. Results from our current study, which includes a larger cohort in which the majority of patients received systemic therapy as a component of their treatment, suggest that radiographic RPLN involvement indicates a poorer prognosis and increased risk for distant failure, despite the influence of systemic therapy. This finding suggests that radiographic RPLN involvement may affect outcomes similarly to other clinical markers of regionally advanced disease, with increased risk of recurrence, distant relapse, death, and, to a lesser magnitude, local failure.[20, 21] Based on the patterns of failure observed, patients with radiographic RPLN adenopathy may be considered candidates for future investigational strategies that seek to reduce risk of subsequent distant metastases or RPLN adenopathy might be used a stratification variable for distant failure reduction strategies (e.g. novel systemic therapy approaches).
Our study is limited by the lack of tumor-human papillomavirus (HPV) status, which (along with patient smoking history) has recently been established as an important predictor of outcomes in patients irradiated for OPC. We studied a modern cohort of patients with long-term follow-up irradiated between 2001 and 2007, an era when we were not routinely performing HPV analysis. Some have used patient smoking status as an indicator of tumor HPV status, and our data demonstrate a difference in survival between smoking cohorts dichotomized by RPLN status (Fig. 3). This finding suggests that RPLN involvement may indicate a negative influence on prognosis for both HPV-positive and HPV-negative cohorts. However, because radiographic RPLN status is highly correlated with a number of disease characteristics that influence patient outcomes, this finding must be considered exploratory and requires further investigation. We did note an increased incidence in radiographic RPLN adenopathy among current smokers compared with never smokers (Table 2); however, when controlled by T category, the differences were no longer significant (data not shown). Whether the incidence of radiographic RPLN adenopathy and negative impact on patient outcomes here reported is maintained when HPV status is incorporated will require future study.
Currently, patients at our institution with OPC and clinically involved RPLN are treated with a definitive radiation therapy–based approach, often in combination with systemic therapy, depending on primary tumor and other nodal characteristics. When clinically uninvolved, we have routinely and electively irradiated the bilateral lateral RPLN regions in all primary carcinomas of the pharynx to doses designed to eradicate subclinical disease, but for selected patients with well-lateralized tonsillar cancers, we have often restricted lymph node coverage to the ipsilateral cervical and RPLN basin. Our approach of routine elective treatment of RPLNs holds true for carcinomas of unknown primary site, when the pharyngeal axis is irradiated. This has been our preferred approach for the following reasons: 1) based on anatomic studies, the RPLNs are known to serve as primary drainage from the pharynx; 2) the incidence of radiographic RPLN involvement suggests that the risk of occult microscopic involvement is sufficient (ie, >5%) to justify standard practice of RPLN elective irradiation in most cases; 3) elective RPLN irradiation has been shown to be effective, and recurrence in electively treated regions is extremely rare; 4) recurrence at RPLN regions is more difficult to detect (without routine imaging) compared with lateral cervical lymph nodes, and salvage therapy may not be feasible; and 5) elective RPLN coverage can be safely achieved with modern radiation approaches (eg, IMRT) while still allowing for relative salivary gland sparing.
In our cohort, <1% of patients had bilateral radiographic RPLN involvement, and only 1 patient had contralateral radiographic RPLN involvement without ipsilateral RPLN involvement. Therefore, consideration of restricting elective RPLN irradiation to the ipsilateral RPLN basin in more than just selected lateralized tonsil cancer patients may be justifiable, but this approach has not been implemented into our practice to date. This conceivably would allow for a reduced dose to the contralateral parotid gland and a portion of the pharyngeal constrictor muscles. However, it is unclear whether this modification in RPLN target volume would translate into clinically meaningful reduction in toxicity balanced against the cost of a small but non-zero risk of subsequent contralateral RPLN failure, which is often nonsalvageable and fatal; therefore, this approach should only be studied in a judicious prospective fashion. Additionally, no patient in our study had an identifiable median RPLN, which further supports the dysphagia-reducing RPLN IMRT target volume delineation approach as described by Feng et al (targeting the lateral RPLNs but excluding the median RPLN regions), which allows for relative sparing of the more medial portion of the pharyngeal constrictor muscles.
In a cohort of nearly 1000 patients with OPC staged and treated in the modern era, radiographic RPLN involvement was observed in 10% of patients and conveyed a negative influence on disease recurrence, distant relapse, and patient survival. The rates of clinical radiographic RPLN involvement reported here, coupled with the existing literature regarding patterns of failure, support our current standard approach of routine elective coverage of uninvolved RPLN regions at risk for all irradiated OPC patients.
Supported in part by Cancer Center Support (Core) Grant CA016672 to The University of Texas MD Anderson Cancer Center.
The authors made no disclosures.