Abstract
BACKGROUND:
The objective of the current study was to characterize the incidence, pattern, and impact on oncologic outcomes of retropharyngeal lymph node (RPLN) involvement in HPV-associated oropharyngeal cancer (OPC).
METHODS:
Data regarding patients with HPV-associated OPC who were treated at The University of Texas MD Anderson Cancer Center with intensity-modulated radiotherapy from 2004 through 2013 were analyzed retrospectively. RPLN status was determined by reviewing pretreatment imaging and/or reports. Outcomes analysis was restricted to patients with lymph node-positive (+) disease. Kaplan-Meier survival estimates were generated and survival curves were compared using the log-rank test. Bayesian information criterion assessed model performance changes with the addition of RPLN status to current American Joint Committee on Cancer staging. Competing risk analysis compared modes of disease recurrence.
RESULTS:
The incidence of radiographic RPLN involvement was 9% (73 of 796 patients) and was found to vary by primary tumor site. The 5-year rates of freedom from distant metastases (FDM) and overall survival were lower in patients with RPLN(+) status compared with those with RPLN-negative (−) status (84% vs 93% [P = .0327] and 74% vs 87% [P = .0078], respectively). RPLN(+) status was not found to be associated with outcomes on multivariate analysis. Bayesian information criterion analysis demonstrated that current American Joint Committee on Cancer staging was not improved with the inclusion of RPLN. Locoregional and distant disease recurrence probabilities for those patients with RPLN(+) status were 8% and 13%, respectively, compared with 10% and 6%, respectively, for those with RPLN(−) status. RPLN(+) status portended worse 5-year FDM in the low-risk subgroup (smoking history of <10 pack-years) and among patients who received concurrent chemotherapy but not induction chemotherapy.
CONCLUSIONS:
RPLN(+) status was associated with worse overall survival and FDM on univariate but not multivariate analysis. In subgroup analyses, RPLN(+) status was associated with poorer FDM in both patients with a smoking history of <10 pack-years and those who received concurrent chemotherapy, suggesting that RPLN(+) status could be considered an exclusion criteria in treatment deintensification efforts seeking to omit chemotherapy.
Keywords: head and neck, HPV, oropharyngeal carcinoma, radiotherapy, retropharyngeal lymph node
INTRODUCTION
Tumor HPV status has been shown to be a powerful, independent prognostic factor of patient outcome in those with oropharyngeal carcinoma (OPC),1 most likely reflecting inherent differences in the underlying tumor biology and responsiveness to treatment.2 In addition to tumor HPV status, patient smoking history and/or intensity and the extent of local and regional disease form the basis of current clinical risk group stratification systems used in contemporary clinical trials. In addition, the recently updated eighth edition of the American Joint Committee on Cancer (AJCC) staging system groups patients with HPV-associated OPC based on the laterality and size of lymph node disease, yet the anatomic distribution of lymph node disease is not considered.
We previously published a retrospective analysis of a cohort of 981 patients who underwent radiotherapy (RT) for OPC and found that pretreatment radiographic retropharyngeal lymph node (RPLN) involvement had a negative influence on disease recurrence, distant failure, and overall survival (OS). However, given that these patients had been treated prior to routine tumor HPV testing, the lack of tumor HPV status limited study conclusions. Since our initial publication, other single-institution series evaluating patient outcomes for those specifically with HPV-associated OPC and RPLN involvement have demonstrated mixed results.3–5
Given the paucity and limited power of existing data and conflicting results regarding the impact of RPLN in patients with HPV-associated OPC, the current study was performed to: 1) characterize the incidence, patterns, and lymph node volume characteristics of RPLN involvement in patients with HPV-associated OPC; 2) assess the impact of RPLN involvement on oncologic outcomes and patterns of disease recurrence in patients with lymph node-positive (N+) OPC; 3) test the potential usefulness of RPLN status as a covariate to improve the predictive model performance of the current AJCC staging system; and 4) generate hypotheses for future investigation.
MATERIALS AND METHODS
A cohort of patients with HPV-associated OPC who were treated with definitive RT with or without systemic therapy at The University of Texas MD Anderson Cancer Center from 2004 through 2013 were analyzed retrospectively. The current study was performed after institutional review board approval. Demographic, clinical, and patient outcomes data were collected from the electronic medical record. Tumors were considered to be HPV associated if they were positive for either p16 by immunohistochemistry or high-risk HPV DNA by in situ hybridization. Patients were excluded from the current study if they were found to have distant metastatic disease prior to initial treatment (1 patient) or if they did not complete definitive treatment (2 patients).
RPLN status was determined by a review of initial staging and/or pretreatment diagnostic imaging reports; for those patients with available imaging, the pretreatment diagnostic studies were reviewed to assess for radiographic RPLN involvement. As with our previous study, the criteria for radiographic RPLN involvement included: RPLN size >5 mm in minimal axial diameter6 or ≥10 mm in maximal axial diameter7,8; the presence of any medial/ median RPLN8; central necrosis6,7; the identification of ≥2 clustered RPLNs6; and any RPLN with a maximum standardized uptake value >4.5 on positron emission tomography (PET).9
It is our practice to electively irradiate bilateral-lateral RPLN regions in patients with otherwise clinically uninvolved RPLNs, restricting the RT field in select patients with well-lateralized tonsillar cancers to the ipsilateral cervical and ipsilateral lateral RPLN regions, as previously described in further detail.10,11
Radiographically evident RPLNs were localized and manually segmented on pretreatment contrast-enhanced computed tomography (CE-CT) by 2 radiation oncologists (H.E. and B.E.) and confirmed by a third (A.S.M.). Three-dimensional reconstruction of segmented volumes and the calculation of the RPLN volume in cubic centimeters (cc3) was performed using Velocity AI 3.0.1 software (Varian Medical Systems, Palo Alto, California). Details regarding RPLN laterality and the level of involvement of other cervical lymph nodes were recorded. Patients who did not undergo a baseline CE-CT (2 patients), had RPLN that was inseparable from the primary tumor (1 patient), or whose RPLN was heavily obscured by dental artifact (1 patient) were excluded from volume analyses.
Patients were staged according to the eighth edition of the AJCC staging criteria. Patients with Tx or Nx disease were recorded as T1 or N1, respectively, for the purpose of analysis. Smoking status was classified using the same criteria as previously described.10,12 A cutoff value of 10 pack-years for the analysis of subgroups was chosen to conform with the risk stratification schema employed in the NRG-HN002 trial.13
Statistical Analysis
Descriptive statistics were used to determine the incidence of RPLN involvement. A chi-square test was used to compare proportions between groups, and a Student t test was used to compare means.
Patients with lymph node-negative disease were excluded from analyses of oncologic outcomes and survival endpoints to mitigate any potential confounding effect of lymph node-negative disease on outcome differences by RPLN status. Univariate and multivariate analyses were performed using the Cox proportional hazards model with respect to RPLN status. Univariate analysis examined age, sex, smoking status, primary tumor subsite, RPLN involvement, and the presence of a low-lying (level IV or level Vb) lymph node. Variables with an α ≤ .10 in univariate analysis were included in the subsequent multivariate analysis. The therapeutic combination (ie, RT alone, induction chemotherapy followed by RT [IC+RT], concurrent chemoradiotherapy [CC], and IC followed by CC [IC+CC]) was included in multivariate analysis to account for potential inherent clinical differences in treatment outcome related to more intensified treatments. Significance in multivariate analysis was determined by an α < .05. Kaplan-Meier survival curves were generated to calculate actuarial survival rates and were compared using the log-rank test. For the purpose of survival analysis, time point 0 was defined as the date of pathologic diagnosis. All analyses were performed using the JMP Pro statistical software (version 12.1.0; SAS Institute Inc, Cary, North Carolina).
Bayesian information criteria (BIC) were used to measure the effect of RPLN status on the performance of AJCC stage as a prognostic model. BIC measures the predictive performance of a set of covariates, with the addition of a penalty factor for each additional covariate added to the model. Numerically lower BIC scores indicate superior model performance and parsimony. Parametric models (Weibull distribution, Frechet distribution, exponential, log-normal, and log-logistic) and the semiparametric Cox proportional hazards model were evaluated. The model with the lowest BIC was selected for inclusion of the covariate RPLN status. BICs of the models before and after the addition of the RPLN covariate were compared. If the addition of RPLN status lowered the BIC value, this would suggest that RPLN added to AJCC stage performance and parsimony. Competing causes of failure (ie, locoregional vs distant) were analyzed in patients with and without RPLN involvement using competing risk analysis.
RESULTS
A total of 796 patients formed the current study cohort. Detailed demographic, disease, and treatment characteristics of the cohort are shown in Table 1. The median total RT dose and number of fractions was 69.96 grays (Gy) and 33, respectively.
TABLE 1.
Patient Demographics and Characteristics
| Characteristics | RPLN Negative N = 723 No. (Column %) | RPLN Positive N = 73 No. (Column %) | Total Cohort | P |
|---|---|---|---|---|
| Age, y | .2147 | |||
| Mean (range) | 58.7 (31–86) | 57.3 (34–76) | 58.6 (31–86) | |
| Median | 57.8 | 57.4 | 57.7 | |
| Sex | .0901 | |||
| Male | 629 (87) | 58 (79) | 687 | |
| Female | 94 (13) | 15 (21) | 109 | |
| Smoking (never/former/current) | .0487 | |||
| Current | 120 (16) | 21 (29) | 141 | |
| Former | 280 (39) | 25 (34) | 305 | |
| Never | 323 (45) | 27 (37) | 350 | |
| Smoking pack-y | .1275 | |||
| >10 | 460 (64) | 40 (55) | 500 | |
| <10 | 238 (33) | 32 (44) | 270 | |
| Unknown | 25 (3) | 1 (1) | 26 | |
| Low-lying LN status | .205 | |||
| Present | 81 (11) | 12 (16) | ||
| Absent | 642 (89) | 61 (84) | ||
| Tumor subsite | .002 | |||
| Tonsil | 318 (44) | 47 (65) | 365 | |
| Base of tongue | 347 (48) | 20 (27) | 367 | |
| Soft palate | 5 (1) | 2 (3) | 7 | |
| Pharyngeal wall | 2 (1) | 1 (1) | 3 | |
| NOS | 51 (7) | 3 (4) | 54 | |
| T classification | <.0001 | |||
| 1 | 258 (36) | 7 (10) | 265 | |
| 2 | 272 (38) | 35 (48) | 307 | |
| 3 | 117 (16) | 14 (19) | 131 | |
| 4 | 76 (10) | 17 (23) | 93 | |
| N classification (AJCC 8th edition) | <.0001 | |||
| 0 | 57 (8) | 0 (0) | 57 | |
| 1 | 507 (70) | 41 (56) | 548 | |
| 2 | 141 (20) | 27 (37) | 168 | |
| 3 | 18 (2) | 5 (7) | 23 | |
| AJCC 8th edition stage of disease | .0018 | |||
| I | 450 (62) | 31 (42) | 481 | |
| II | 181 (25) | 23 (32) | 204 | |
| III | 92 (13) | 19 (26) | 111 | |
| Therapeutic combination | <.0001 | |||
| RT Alone | 145 (20) | 6 (8) | 151 | |
| CCRT | 288 (40) | 21 (29) | 309 | |
| IC followed by RT | 109 (15) | 9 (12) | 118 | |
| IC followed by CCRT | 181 (25) | 37 (51) | 218 |
Abbreviations: AJCC, American Joint Committee on Cancer; CCRT, concurrent chemoradiotherapy; IC, induction chemotherapy; LN, lymph node; NOS, not otherwise specified; RPLN, retropharyngeal lymph node; RT, radiotherapy.
Incidence of RPLN and Patterns of Presentation
Radiographic RPLN involvement was identified in 73 patients, for an overall incidence of RPLN involvement of 9%. RPLNs for all 73 patients were identified at the time of the initial diagnosis and these patients were treated to the therapeutic radiation dose (median, 70 Gy; interquartile range, 66–70 Gy), with no untreated macroscopic disease. Positive RPLNs had been identified by a contrast-enhanced CT scan of the neck alone in 48 patients, PET/CT alone in 5 patients, both contrast-enhanced CT and PET/CT in 19 patients, and both contrast-enhanced CT and magnetic resonance imaging alone in 1 patient. Ultrasound was not used for the purpose of identifying RPLN involvement given the anatomic location of the retropharyngeal region, and no patient underwent transpharyngeal ultrasound.
Among patients with RPLN(+) status, 58% had T1/2 primary tumors, 16% had low-lying cervical lymph nodes, and 74% had AJCC stage I/II disease. With regard to therapeutic approach, 578 of 723 patients with RPLN(−) status (80%) received systemic therapy and 289 patients (40%) received IC as a component. Limiting the analysis to patients with RPLN(−) status who had N+ disease (667 patients), approximately 83% received systemic therapy and 42% received IC as a component. By comparison, 67 of 73 patients with RPLN(+) status (92%) received systemic therapy (P = .0304) and 46 patients (63%) received IC (P = .0013).
The incidence of RPLN involvement grouped by patient, tumor, and treatment variables is shown in Table 1. RPLN involvement was observed more frequently among patients with primary tumors in the tonsil compared with the base of the tongue (13% vs 5%; P = .0003). Advanced primary disease (T3/4 vs T1/2) and AJCC eighth edition stage of disease (stage II/III vs stage I) also were found to be associated with RPLN involvement (P = .0059 and P = .0012, respectively). The distribution of percentages for T classification, N classification, AJCC stage of disease, RPLN status, and low-lying lymph node status all were found to be statistically significantly different (P < .001 for all) by treatment group (see Supporting Table 1). Advanced T classification (ie, T3/4) was highest among those who received IC+CC (53%), followed by IC+RT (27%), CC (25%), and RT alone (1%). Regionally advanced disease (ie, N2-N3 disease) was most common in those treated with IC+CC (54%), followed by IC+RT (32%), CC (12%), and RT alone (2%). RPLN(+) status was most common among patients treated with IC+CC (17%), followed by IC+RT alone (8%), CC (7%), and RT alone (5%).
All involved RPLNs were noted in the lateral RPLN group. Isolated RPLN involvement was identified in 7 patients (ipsilateral in 6 patients and bilateral in 1 patient) who otherwise had no involved cervical lymph nodes (ie, would have been classified as having cN0 disease aside from RPLN involvement). Of these, 6 patients had primary tumors of the tonsil and 1 had a primary tumor of the base of the tongue. All patients but one were alive at the time of last follow-up with no evidence of disease recurrence; the one exception was a patient who died with distant metastatic disease recurrence. Bilateral RPLN involvement was observed in only 4 patients. Three of these patients had primary tumors of the tonsil, one of which was clinically well lateralized, whereas the fourth patient had a large tumor involving both tonsils, the palate, and the pharyngeal wall.
In 7 patients, a contralateral RPLN was involved without the presence of a positive ipsilateral RPLN. Of these, 5 were primary tumors of the base of the tongue and 2 were primary tumors of the tonsil, none of which were well lateralized, and 2 of 7 patients had T4 primary tumors. Of the 7 patients, 2 subsequently died, one with distant metastatic disease recurrence.
Retropharyngeal Lymph Node Volume
Overall, the median involved RPLN volume was 1.2 cc3 (range, 0.1–10.3 cc3). Based on the primary tumor subsite, the median positive RPLN volume was numerically highest in the tonsil (1.9 cc3), followed by the pharyngeal wall (1.8 cc3), soft palate (1.3 cc3), oropharynx-not otherwise specified (1.0 cc3), and base of the tongue (0.85 cc3).
Oncologic Outcomes
The median follow-up for the N+ cohort was 57 months (interquartile range, 42–76 months). The 5-year rates of local control (LC), regional control (RC), freedom from distant metastases (FDM), recurrence-free survival (RFS), disease-specific survival, and OS by radiographic RPLN involvement status are shown in Table 2. Patients with RPLN involvement demonstrated lower 5-year rates of FDM (84% vs 93%; log-rank P = .0327) and OS (74% vs 87%; log-rank P = .0078) compared with patients without RPLN involvement (Figs. 1A and 1B). The cause of death in 49 of the 114 patients who had died by the time of last follow-up (43%) was non-cancer related (ie, no evidence of OPC recurrence at the time of death). No statistically significant difference in 5-year outcomes was observed for LC, RC, RFS, or disease-specific survival.
TABLE 2.
Five-Year Treatment Outcomes in Patients With Lymph Node–Positive OPC by RPLN Status
| Treatment Outcomes | RPLN Positive (%) | RPLN Negative (%) | Pa |
|---|---|---|---|
| Local control | 96 | 94 | .5671 |
| Regional control | 95 | 93 | .4736 |
| Freedom from distant metastases | 84 | 93 | .0327 |
| Recurrence-free survival | 81 | 85 | .3848 |
| Disease-specific survival | 86 | 92 | .3515 |
| Overall survival | 74 | 87 | .0078 |
Abbreviations: OPC, oropharyngeal cancer; RPLN, retropharyngeal lymph node.
Bold type indicates statistical significance.
Figure 1.
(A) Freedom from distant metastases in patients with lymph node-positive disease by retropharyngeal lymph node (RPLN) status. (B) Overall survival in patients with lymph node-positive disease by RPLN status. +, indicates positive; -, negative.
Subgroup Analyses
In patients with a smoking history of ≤10 pack-years, patients with RPLN(+) status were found to have worse rates of 5-year FDM (83% vs 94%; P = .0176) and OS (76% vs 88%; P = .0478) compared with patients with RPLN(−) status. There were no statistically significant differences in examined 5-year outcomes based on RPLN status noted among patients with a smoking history of >10 pack-years, although there was a strong trend for worse 5-year OS (71% vs 86%; P = .0594) in patients with RPLN(+) status (Fig. 2).
Figure 2.
Overall survival in patients with lymph node-positive disease by retropharyngeal lymph node (RPLN) status and smoking intensity. +, indicates positive;−, negative.
Outcomes by treatment combination are shown in Supporting Table 2. There were no statistically significant differences in 5-year outcomes noted by RPLN status among patients receiving RT alone (122 patients), IC+RT (118 patients), or IC+CC (217 patients). In the group of patients who received CC (283 patients), the 5-year FDM rate was 75% and 93%, respectively, in the patients with RPLN(+) and RPLN(−) status (P = .0198). Among patients who received any concurrent chemotherapy as a component of treatment (500 patients), the 5-year FDM rate was 92% compared with 82% (P = .0519) and the 5-year OS rate was 84% compared with 72% (P.0344) for patients with RPLN(−) compared with RPLN(+) status. Among patients who received any IC as a component of treatment (335 patients), the 5-year FDM rate was 92% compared with 86% (P = .2517) and the 5-year OS rate was 85% compared with 77% (P = .0766) among patients with RPLN(−) status compared with those with RPLN(+) status.
Outcomes for those patients with RPLN(+) status by receipt of IC are shown in Supporting Table 3. Among patients with RPLN involvement, there were no statistically significant differences noted with regard to treatment outcomes by receipt of IC. Specifically, the 5-year FDM rate for patients with RPLN+ status with and without receipt of IC was 81% and 86%, respectively (P = .7420).
When stratified by N classification, numerical differences in the 5-year FDM and OS rates between patients with RPLN(+) and RPLN(−) status still were observed (Fig. 3). Specifically, among patients with N1 disease, those with RPLN(+) status had numerically lower 5-year FDM (86% vs 94%; P = .1352) and OS (79% vs 88%; P = .5772) rates than those with RPLN(−) status. Among patients with N2-N3 disease, patients with RPLN(+) status had lower 5-year FDM (82% vs 90%; P = .2201) and OS (67% vs 84%; P = .0140) rates compared with patients with RPLN(−) status.
Figure 3.
Freedom from distant metastases by N category and retropharyngeal lymph node (RPLN) status. +, indicates positive;−, negative.
Univariate and Multivariate Analysis
The results of univariate and multivariate analysis by Cox proportional hazards analysis are provided in Supporting Table 4. RPLN was selected for inclusion in multivariate analysis of FDM and OS based on univariate P values < .10 in each case. Multivariate analysis of both FDM and OS included age, sex, low-lying lymph node status, RPLN status, therapeutic combination, and AJCC stage. The statistical significance of RPLN involvement was not maintained in the multivariate analysis of either treatment outcome (FDM: P = .1112 and OS: P = .1167).
Model Performance Using BIC
When examining models using AJCC stage without accounting for RPLN status, the best performing parametric model for each treatment outcome was the Frechet distribution, with the exception of OS, in which the log-normal distribution performed best. However, the best overall model (parametric or nonparametric) for all tested treatment endpoints was the Cox proportional hazards model. When RPLN status was added as a covariate to the Cox model, BIC increased for every measured treatment outcome, indicating that the addition of the RPLN covariate did not improve model performance and parsimony within this cohort (see Supporting Table 5).
Patterns of Failure
For patients with RPLN involvement, the probability of locoregional and distant failure at the 5-year follow-up was 8% and 13%, respectively (Fig. 4A). In contrast, for patients without RPLN involvement, the probability of locoregional and distant failure at the 5-year follow-up was 10% and 6%, respectively (Fig. 4B).
Figure 4.
Probability of failure (locoregional vs distant vs aggregated) in patients with lymph node-positive oropharyngeal cancer (A) with retropharyngeal lymph node (RPLN) involvement and (B) without RPLN involvement.+, indicates positive; −, negative.
All patients with RPLN(+) status had control of their retropharyngeal disease. RPLN recurrence was observed in 3 patients with RPLN(−) status (<1%); 1 recurrence was ipsilateral to the primary tumor, 1 was contralateral, and 1 was bilateral. All 3 patients initially had stage II disease; 2 had primary tumors of the tonsil and the third patient had a primary cancer in the base of the tongue. Each area of disease recurrence had been targeted to elective RT doses. All 3 patients subsequently died of their disease.
DISCUSSION
In the current study, we retrospectively analyzed the patterns of RPLN involvement and its impact on patient outcome in a large cohort of patients who underwent RT for HPV-associated OPC with long-term follow up. Overall, the current study findings bear similarities to our previous work regarding the subject of radiographic RPLN involvement in patients with OPC,10 such as the overall and tumor subsite-specific incidences of RPLN involvement, but important differences between these 2 studies should be noted. First, the current study was restricted to a contemporary cohort with a known positive HPV association, whereas the previous work examined patients treated from 2001 to 2007 and whose HPV-associated status was unknown at the time, and was likely to contain a relatively higher percentage of non-HPV-associated cases. For example, compared with our 2013 report, in the current study we noted an increase in classic HPV-associated subsites (eg, tonsil and base of the tongue) and a decrease in typically non-HPV-associated subsites (eg, soft palate and pharyngeal wall). In the initial study, RPLN involvement was found to be significantly associated with lower 5-year rates of LC, RC, FDM, RFS, and OS and maintained this significance in multivariate analysis for LC, FDM, RFS, and OS, and numerical outcome differences between those patients with RPLN(+) status compared with RPLN(−) status ranged from 13% to 30%. Overall, the disease control and survival rates in the current study are substantially higher than those in our 2013 study, both for those with and without RPLN involvement, which is a somewhat expected finding given that the current study was restricted to a group of patients with HPV-associated disease yet is notable in that numeric differences in outcomes in the current study between patients with RPLN(+) status and those with RPLN(−) status ranged from −2% to 13%, suggesting that statistically significant group differences in those with HPV-associated disease who are managed using contemporary standards would be difficult to detect.
Based on the results of our studies and others, RPLN involvement as a single variable appears informative of outcome in patients with OPC. However, within the context of other variables, the potential relationship between RPLN and outcomes may be difficult to demonstrate. Tang et al5 studied a cohort of 165 patients with OPC, 134 of whom had HPV-associated disease, and observed a 12% incidence of RPLN involvement, similar to the 9% incidence observed in the current study cohort and the 10% incidence observed in our 2013 study, although lower than the incidence of 16% to 21% reported by others.3,14–16
The discrepancy in the RPLN incidences reported in these study populations compared with that in the current study may be related to differences in the percentage of patients with locally advanced OPC in each study cohort. For example, the percentage of patients with T3-T4 disease in the referenced studies ranged from 46% to 64%, which is higher than the rate of 42% observed in the study by Tang et al5 and the rate of 28% (224 of 796 patients) observed in the current study. With regard to oncologic outcomes, Tang et al5 demonstrated that RPLN involvement demonstrated a nonsignificant trend toward worse OS (P = .08). Similarly, Baxter et al found increased odds of disease recurrence or death in patients with HPV-associated OPC with radiographic RPLN involvement but this did not remain statistically significant in multivariate analysis.4 In contrast, Samuels et al found lower 5-year rates of OS and increased distant failure in patients with RPLN involvement compared with those without, findings that maintained statistical significance on multivariate analysis.3
Tumor HPV status now is considered in the AJCC eighth edition staging criteria for OPC. The exploratory results in the current study using BIC to compare multivariate models of AJCC stage group with and without the addition of RPLN as a covariate demonstrated that the addition of RPLN status could not improve the predictive performance of the current staging system within this cohort. However, this finding could reflect the high overall curability of HPV-associated OPC and/or the robustness of the current staging system in terms of survival, not a diminution of the potential effect of RPLN on the risk or pattern of disease recurrence. O’Sullivan et al helped to define candidates for treatment deintensification among patients with HPV-related OPC according to their risk of distant disease recurrence.17 Based on the AJCC seventh edition staging system, they found that patients with T1-T3 N0-N2c disease were at low risk of distant disease, but noted that patients with bilateral lymph node disease had an increased risk of distant disease recurrence if treated with RT alone, thereby recommending that patients with N2c disease be excluded from deintensification strategies that exclude chemotherapy. Although not included in the new iteration of staging, we recommend a similar caution for patients with RPLN disease, particularly given that among patients with N1 disease in the current study there was an observed 8% numerical difference in 5-year FDM rates between patients with RPLN(+) status and those with RPLN(−) status.
Given our previous findings of an elevated risk of distant failure associated with RPLN involvement and the potential for IC to mitigate that risk, we were interested to explore further subgroup outcomes by RPLN status, treatment combination, and subsequent distant failure. In the current study, we found that patients who received CC had statistically significantly worse 5-year FDM among patients with RPLN(+) status compared with RPLN(−) status (75% vs 93%; P = .0198), whereas the 5-year FDM rate was only numerically different in either IC group (7% difference for patients receiving IC+RT and 5% difference for patients receiving IC+CC). To further enhance the power and to evaluate for statistical significance, patients were grouped by receipt of any IC (ie, IC+RT or IC+CC) or any CC (ie, CC or IC+CC). It is interesting to note that for the “any CC” cohort, the difference in the 5-year FDM rate trended toward statistical significance (92% vs 82%; P = .0519), whereas the 5-year OS rate was significantly different (84% vs 72%; P = .0344) for patients with RPLN(−) status compared with those with RPLN(+) status. Conversely, in the “any IC” cohort, the 5-year FDM rate was found to be nonsignificantly different (P = .2517), with a numerical difference of only 6% noted between the RPLN groups, whereas the 5-year OS rate trended toward statistical significance (85% vs 77%; P = .0766). Among patients in the “any IC” cohort, there was a statistically nonsignificant difference in the 5-year FDM rate of 6%; however, among patients in the “any CC” cohort, there was a 10% difference in the 5-year FDM rate that trended toward statistical significance (P = .0519). From these results, a few observations are noted. A difference in outcomes for patients treated with RT alone was not observed, most likely due to the favorable prognostic factors and outcomes characteristic of this treatment cohort. Second, given that locoregional control rates were similar by RPLN status, the apparent negative influence of RPLN(+) status on survival may be driven by distant metastatic disease recurrence. Third, given that patients who received CC had worse 5-year FDM in the RPLN(+) cohort compared with the RPLN(−) cohort, whereas the same difference was not observed in either of the IC treatment groups, it is possible that the negative influence of RPLN involvement on distant failure may be at least partially mitigated by IC. Although additional subgroup analysis demonstrated no statistically significant difference in 5-year FDM rates among patients with RPLN(+) status based on receipt of IC, the potential that IC mitigated the subsequent risk of distant failure nevertheless cannot be ruled out.
In an additional subgroup analysis of patients with a smoking history of ≤10 pack-years, a criterion used to define low-risk OPC clinical subgroups, 5-year FDM rates were worse in patients with RPLN(+) status compared with those with RPLN(−) status at a statistically significant level. Within the context of higher distant failure probabilities in patients with RPLN(+) status compared with RPLN(−) status, these findings further underscore the need for caution when considering patients with RPLN(+) status for deintensification strategies, even in patients otherwise considered at low risk by markers such as a shorter smoking history and/or lower smoking intensity.
RPLN involvement was not found to be associated with worse locoregional control. Furthermore, all patients with known RPLN disease had their RPLN disease controlled. Disease recurrence in the RPLN region in patients with RPLN(−) status was observed in only 3 patients, and all 3 patients subsequently died of disease. All of these patients received subclinical doses (50–57 Gy) of RT to this area. Given the incidence of radiographic involvement observed in the current study, the expected risk of microscopic involvement for those without radiographic involvement,18 difficulty in the early detection and salvage of skull base recurrences, and the overall low incidence of RPLN recurrence after elective RT, we continue to electively irradiate the lateral RPLN regions as part of our routine clinical practice, irrespective of HPV tumor status. However, in patients with HPV-associated OPC, it may be possible to consider sparing elective irradiation to the contralateral RPLN in selected low-risk populations. Selected subgroups of patients with OPC should be at low risk of RPLN involvement and particularly contralateral RPLN involvement, such as more lateralized tumors of the base of the tongue with little to no lymph node disease ipsilateral to the RPLN station being considered for sparing. Sparing of the contralateral RPLN region previously has been shown to be associated with improved quality of life with no significant difference in observed failure rates,19 and resulted in a lowered RT dose to the nearby pharyngeal constrictor muscles and contralateral parotid gland, which should reduce RT-induced toxicities. These strategies should be evaluated prospectively and patients committed to routine surveillance incorporating imaging with attention to RPLN regions.
Several important limitations of the current study need to be considered when interpreting the results. In addition to those limitations inherent to single-institution retrospective series, the current study was limited by the lack of a matched comparator HPV-negative group and notable treatment heterogeneity in terms of the sequence of systemic therapy. In an attempt to address these limitations, beyond multivariate analysis, numerous subgroup analyses were included to explore for potential differences in outcomes according to disease characteristics or treatment received according to RPLN status. It is interesting to note that although CC was the most common approach used, many patients in the current study received IC as a component of their care, which, although reflective of practice patterns at the time these patients were treated, is a practice that has since decreased at the study institution and elsewhere given the negative results from prospective studies reported in the interim. As the incidence of HPV-negative OPC continues to decline, further study of HPV-negative OPC with respect to RPLN status will require longer collection periods and would be aided by the use of multi-institutional and international shared databases. A limitation of the BIC analysis of the use of the AJCC eighth edition staging system as a prognostic tool was that model performance is specific to our single-institution database and should not be generalized to other data sets with potentially differing patient characteristics. Most important, the potential negative influence of RPLN on outcomes and patterns of failure reported in the current study should be examined further within the context of a prospective study with treatment homogeneity.
CONCLUSIONS
The overall incidence of radiographic RPLN involvement in this cohort of approximately 800 patients with HPV-associated OPC was 9%, although it was significantly higher (P = .0003) for patients with cancers of the tonsil compared with those with cancers of the base of tongue at 13% and 5%, respectively. RPLN as a single variable remained associated with increased distant failure and worse OS, but within the context of other prognostic variables (multivariate analysis) and the excellent outcomes noted in patients with HPV-associated OPC, the magnitude of this difference was smaller compared with our previous report and was not found to be statistically significant. Nevertheless, on subgroup analyses, RPLN status demonstrated a differential in outcomes even for those patients with a minimal smoking history and/or intensity and could be considered to be an exclusion criteria for investigational treatment deintensification strategies, specifically those seeking to omit systemic therapy.
Supplementary Material
FUNDING SUPPORT
Supported by the Andrew Sabin Family Foundation.
Footnotes
CONFLICT OF INTEREST DISCLOSURES
Steven J. Frank has acted as a paid consultant and member of the advisory board for Varian, is the founder and director of and has received grants and personal fees from C4 Imaging, has received grants from Eli Lilly and Elekta AB, has received grants and honoraria from Hitachi, and has received honoraria from Augminex for work performed outside the current study. Clifton D. Fuller is a Sabin Family Foundation Fellow and has received grants to his institution and conference travel support from Elekta AB for work performed as part of the current study. Dr. Fuller also has received book royalties from Demos Publishing and personal fees and nonfinancial support from the Greater Baltimore Medical Center as well as grant salary and funding from the National Institutes of Health (NIH) including the National Institute for Dental and Craniofacial Research Award (1R01DE025248/R56DE025248); a National Science Foundation (NSF) Division of Mathematical Sciences Joint NIH/NSF Initiative on Quantitative Approaches to Biomedical Big Data grant (NSF 1557679); the NIH Big Data to Knowledge (BD2K) program of the National Cancer Institute Early Stage Development of Technologies in Biomedical Computing, Informatics, and Big Data Science Award (1R01CA214825); National Cancer Institute Early Phase Clinical Trials in Imaging and Image-Guided Interventions Program (1R01CA218148); a National Institute of Biomedical Imaging and Bioengineering Research Education Program award (1R25EB025787); and grants K12 CA088084, P50CA097007, R01CA225190, and R01CA218148 from the NIH. Dr. Fuller has received travel and honoraria for an external advisory panel from the University of California at Los Angeles and travel for the T32 EAB Panel from the University of Texas Health Science Center at San Antonio for work performed outside of the current study. In addition, Dr. Fuller has a patent UTSC.P1315US.P1/62533321 pending.
Additional supporting information may be found in the online version of this article.
REFERENCES
- 1.Rosenthal DI, Harari PM, Giralt J, et al. Association of human papillomavirus and p16 status with outcomes in the IMCL-9815 phase III registration trial for patients with locoregionally advanced oropharyngeal squamous cell carcinoma of the head and neck treated with radiotherapy with or without cetuximab. J Clin Oncol. 2016;34:1300–1308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kim MJ, Ki MS, Kim K, et al. Different protein expression associated with chemotherapy response in oropharyngeal cancer according to HPV status. BMC Cancer. 2014;14:824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Samuels SE, Vainshtein J, Spector ME, et al. Impact of retropharyngeal adenopathy on distant control and survival in HPV-related oropharyngeal cancer treated with chemoradiotherapy. Radiother Oncol. 2015;116:75–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Baxter M, Chan JY, Mydlarz WK, et al. Retropharyngeal lymph node involvement in human papillomavirus-associated oropharyngeal squamous cell carcinoma. Laryngoscope. 2015;125:2503–2508. [DOI] [PubMed] [Google Scholar]
- 5.Tang C, Komakula S, Chan C, et al. Radiologic assessment of retropharyngeal node involvement in oropharyngeal carcinomas stratified by HPV status. Radiother Oncol. 2013;109:293–296. [DOI] [PubMed] [Google Scholar]
- 6.Zhang GY, Liu LZ, Wei WH, Deng YM, Li YZ, Liu XW. Radiologic criteria of retropharyngeal lymph node metastasis in nasopharyngeal carcinoma treated with radiation therapy. Radiology. 2010;255:605–612. [DOI] [PubMed] [Google Scholar]
- 7.Chua DT, Sham JS, Kwong DL, Au GK, Choy DT. Retropharyngeal lymphadenopathy in patients with nasopharyngeal carcinoma: a computed tomography-based study. Cancer. 1997;79:869–877. [PubMed] [Google Scholar]
- 8.Mancuso AA, Harnsberger HR, Muraki AS, Stevens MH. Computed tomography of cervical and retropharyngeal lymph nodes: normal anatomy, variants of normal, and applications in staging head and neck cancer. Part II: pathology. Radiology. 1983;148:715–723. [DOI] [PubMed] [Google Scholar]
- 9.Matsubara R, Kawano S, Chikui T, et al. Clinical significance of combined assessment of the maximum standardized uptake value of F-18 FDG PET with nodal size in the diagnosis of cervical lymph node metastasis of oral squamous cell carcinoma. Acad Radiol. 2012;19:708–717. [DOI] [PubMed] [Google Scholar]
- 10.Gunn GB, Debnam JM, Fuller CD, et al. The impact of radiographic retropharyngeal adenopathy in oropharyngeal cancer. Cancer. 2013;119:3162–3169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Garden AS, Dong L, Morrison WH, et al. Patterns of disease recurrence following treatment of oropharyngeal cancer with intensity modulated radiation therapy. Int J Radiat Oncol Biol Phys. 2013;85:941–947. [DOI] [PubMed] [Google Scholar]
- 12.Pytynia KB, Grant JR, Etzel CJ, Roberts DB, Wei Q, Sturgis EM. Matched-pair analysis of survival of never smokers and ever smokers with squamous cell carcinoma of the head and neck. J Clin Oncol. 2004;22:3981–3988. [DOI] [PubMed] [Google Scholar]
- 13.ClinicalTrials.gov. ClinicalTrials.gov identifier: NCT02254278. Reduced-dose intensity-modulated radiation therapy with or without cisplatin in treating patients with advanced oropharyngeal cancer. www.clinicaltrials.gov/ct2/show/NCT02254278. Accessed June 4, 2018.
- 14.Dirix P, Nuyts S, Bussels B, Hermans R, Van den Bogaert W. Prognostic influence of retropharyngeal lymph node metastasis in squamous cell carcinoma of the oropharynx. Int J Radiat Oncol Biol Phys. 2006;65:739–744. [DOI] [PubMed] [Google Scholar]
- 15.Spector ME, Chinn SB, Bellile E, et al. Exploration for an algorithm for deintensification to exclude the retropharyngeal site from advanced oropharyngeal squamous cell carcinoma treatment. JAMA Otolaryngol Head Neck Surg. 2016;142:313–318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tauzin M, Rabalais A, Hagan JL, Wood CG, Ferris RL, Walvekar RR. PET-CT staging of the neck in cancers of the oropharynx: patterns of regional and retropharyngeal nodal metastasis. World J Surg Oncol. 2010;8:70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.O’Sullivan B, Huang SH, Siu LL, et al. Deintensification candidate subgroups in human papillomavirus-related oropharyngeal cancer according to minimal risk of distant metastasis. J Clin Oncol. 2013;31:543–550. [DOI] [PubMed] [Google Scholar]
- 18.Chung EJ, Kim GW, Cho BK, Cho SJ, Yoon DY, Rho YS. Retropharyngeal lymph node metastasis in 54 patients with oropharyngeal squamous cell carcinoma who underwent surgery-based treatment. Ann Surg Oncol. 2015;22:3049–3054. [DOI] [PubMed] [Google Scholar]
- 19.Spencer CR, Gay HA, Haughey BH, et al. Eliminating radiotherapy to the contralateral retropharyngeal and high level II lymph nodes in head and neck squamous cell carcinoma is safe and improves quality of life. Cancer. 2014;120:3994–4002. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.