Background:

The updated American Joint Committee on Cancer (AJCC) 8th Edition staging manual restructured nodal classification and staging by placing less prognostic emphasis on nodal metastases for human papillomavirus (HPV)-positive oropharyngeal squamous cell carcinoma (OPSCC). However, there was no change for HPV-negative OPSCC. The purpose of our study is to examine the impact of nodal metastases on survival in HPV-negative OPSCC.

Methods:

HPV-negative OPSCC was queried from the National Cancer Database (NCDB) and Surveillance, Epidemiology and End Results program (SEER) databases. Univariable and multivariable models were utilized to determine the impact of nodal status on overall survival. These patients were reclassified according to AJCC 8 HPV-positive criteria (TNM8+) and risk stratification was quantified with C-statistic.

Results:

There were 11,147 cases of HPV-negative OPSCC in the NCDB and 3,613 cases in SEER that were included in the nodal classification analysis. Unlike nonoropharyngeal malignancies, increased nodal stage is not clearly associated with survival for patients with OPSCC independent of HPV status. When the TNM8+ was applied to HPV-negative patients, there was improved concordance in the NCDB cohort, 0.561 (plus minus) 0.004 to 0.624 (plus minus) 0.004 (difference +0.063) and the SEER cohort, 0.561 (plus minus) 0.008 to 0.625 (plus minus) 0.008 (difference +0.065).

Conclusions:

We demonstrated a reduced impact of nodal metastasis on OPSCC survival, independent of HPV status and specific to OPSCC.

Impact:

We demonstrate, for the first time that when nodal staging is deemphasized as a part of overall staging, we see improved concordance and risk stratification for HPV-negative OPSCC. The exact mechanism of this differential impact remains unknown but offers a novel area of study.

Nodal metastasis has been traditionally associated with poor prognosis in head and neck cancer (1). Recently, several studies examining the impact of nodal metastases have found that survival for oropharyngeal squamous cell carcinomas (OPSCC) is less dependent on nodal status compared with other head and neck subsites (2, 3). However, the majority of these studies have focused on human papillomavirus (HPV)-positive disease and confirmed that despite high rates of nodal metastases, there is a reduced impact on survival (2, 3). On the basis of these findings, the updated American Joint Committee on Cancer (AJCC) and Union for International Cancer Control 8th Edition staging manuals restructured nodal classification and overall group staging by placing less prognostic emphasis on nodal metastases for HPV-positive OPSCC and downstaged the overall group stages relative to other head and neck mucosal cancers (TNM8+; refs. 4, 5). However, unlike HPV-positive disease, HPV-negative OPSCC staging (TNM8–) remains unchanged other than the incorporation of extranodal extension (ENE) as a high risk feature (5).

Although the incidence of HPV-negative OPSCC is downtrending, these cancers still represent up to 36% of all OPSCC and contrary to their HPV-positive counterparts, morbidity and mortality remains poor (6, 7). Furthermore, HPV-negative OPSCC also presents with high rates of regional metastasis but it is assumed that the impact of nodal metastasis mirrors that of other head and neck cancers (i.e., oral cavity, larynx; ref. 8). Data from prospective multi-institutional studies have demonstrated a reduced impact of nodal burden on survival in oropharyngeal cancers; however, this effect has not been discussed in the literature or further validated for HPV-negative disease (2, 8). These findings require additional study to further expand on our current understanding of the mechanisms for metastases, the disparate survival when compared to HPV-positive OPSCC, and a larger, multi-center evaluation of the impact of nodal status in HPV-negative OPSCC remains limited.

Hereupon we analyzed two national cancer registries, the National Cancer Database (NCDB and the Surveillance, Epidemiology and End Results program (SEER), to investigate the impact of nodal metastasis on survival and recurrence in patients with HPV-negative OPSCC. A reference group of patients with HPV-positive OPSCC will be used as comparison given the previously studied impact of nodal status on outcomes. On the basis of previous institutional studies, we hypothesize that nodal burden is not associated with survival for HPV-negative OPSCC and that overall risk stratification can be improved with an alternative staging system.

NCDB/SEER databases

We used two NCDBs with HPV status as part of the analyses. The first was the NCDB from January 1, 2010 to December 31, 2015 (accessed April 4, 2020). This is a national cancer registry managed by the American Cancer Society and American College of Surgeons Commission on Cancer. These data are hospital based and represent up to 70% of newly diagnosed cancers in the United States each year and are collected from over 1,500 facilities accredited by the Commission on Cancer. The second was the SEER database from January 1, 2010 to December 31, 2016 (accessed May 8, 2020), which is maintained by the NCI and is a population-based cancer program comprised of 18 distinct geographical regions and cancer statistics from the source population. SEER integrates these population-based registries and captures about 48% of the U.S. population. This study utilizes the SEER Head and Neck database with HPV variables released for survival analysis. These are deidentified databases and determined to be exempt from review by the University of Michigan Institutional Review Board.

HPV status

The NCDB and SEER capture HPV status by both genotypic (i.e., PCR, in situ hybridization, etc.) and IHC (i.e., p16 expression; ref. 9). In this study, HPV positivity was defined as the presence of a high-risk profile (HPV DNA 16, 18, high risk—type not specified) or staining for overexpression of p16. HPV-negative cancers are those that were negative for HPV DNA, were negative for p16 overexpression, or were only positive for low-risk subtypes (HPV 6, 11, low-risk—type not specified). We excluded HPV-positive patients, those with distant metastases, those without survival data, and those who were missing clinical T, N, and M variables or were T0.

Nodal status

Baseline characteristics of HPV-negative patients included age at diagnosis, sex, race, Charlson–Deyo comorbidity score, hospital type, insurance status, diagnosis year, clinical T, N, and treatment. The private insurance group included NCDB patients that had private insurance and SEER patients that had unspecified nongovernment insurance. Comorbidity status and institution type were not encoded in SEER. Multivariable Cox proportional hazards models of clinical and pathologic T and N combinations were created to determine the differential impact of T and N status on overall survival (OS). These models were adjusted for age, sex, race, comorbidity score, hospital type, insurance status, diagnosis year, and treatment status. Treatment was separated by the following: radiation alone, chemoradiation, surgery alone, surgery + radiotherapy, surgery + chemo, and surgery + chemoradiation. Only procedure codes referring to definitive oncologic surgery were utilized (i.e., wide local excision, pharyngectomy with or without mandibulectomy, etc.). Additional analyses controlling for demographic variables and T status but not treatment were conducted to examine the effects of nodal status on outcomes.

To determine the effect of clinical lymph node burden on OS across other head and neck subgroups, we queried the NCDB for squamous cell carcinomas of the oropharynx (HPV-negative, HPV-positive, and HPV-unknown) as well as oral cavity and larynx. A separate multivariable analysis for each head and neck subgroup was constructed controlling for demographics, clinical T status, and treatment. The co-distribution of tumor (T) to nodal (N) status was analyzed with a log-linear model to determine whether HPV status has an impact on the patterns of local and regional spread.

Clinical restaging

Patients with HPV-negative OPSCC were recoded according to the AJCC 8 HPV-positive clinical staging criteria (designated TNM8+) and risk stratification was compared with the AJCC 7 staging system (TNM7; refs. 4, 10). Per TNM8+ staging, HPV-negative OPSCC T4a and T4b lesions were recoded to T4, and N1-2b and N2c disease were recoded to N1 and N2, respectively. Of note, TNM staging in SEER combines clinical and pathologic variables (i.e., cT1 pN1 cM0) and was considered to be clinical staging for the purpose of this study.

Pathologic restaging

Analysis of pathologic data was performed to evaluate the impact of pathologic staging and ENE on the outcomes and restaging of HPV-negative OPSCC. NCDB patients from the clinical staging analysis who underwent primary site surgery and had pathologic T (pT) and N status (pN) were included in the analysis. Of these patients, only those with known ENE status were staged according to AJCC 8 HPV-negative criteria (pTNM8–). Patients with quantitative data regarding the number of involved lymph nodes were restaged by pTNM8+.

For application of the pTNM8– staging criteria, patients with ECE were upstaged to pN2a and pN3 if they were previously pN1 and pN2 by AJCC 7, respectively. Otherwise, pathologic T and N classification were kept largely the same. To apply the pTNM8+ staging system, we grouped pTNM7 pT4a and pT4b lesions into a combined pT4 category. Nodal staging was restructured according to pTNM8+ criteria. Patients with negative lymph nodes were pN0. Patients with between one and four positive lymph nodes were coded as pN1, and patients with five or more positive lymph nodes were coded as pN2.

Statistical analysis

Bivariate analysis was used to analyze categorical variables. Observed proportions in each (T) and (N) status were visualized with balloon plots faceted by HPV status. The Breslow–Day statistic was employed to test homogeneity across HPV status. log-linear models were then used to evaluate the co-distributions of (T) and (N) status, and test whether they are dependent on HPV to confirm the significant heterogeneity identified by the Breslow–Day test. To further explore how the distribution varied, χ2 statistics and associated residuals from 2 × n tables were used to explore the distributions of (T) and (N) stage separately in the HPV-positive and HPV-negative settings. The residuals represent the difference in observed expected counts in each cell, where expected values come from the assumption of no association. Survival by N and overall TNM stage groups for various staging criteria (TNM7, TNM8–, and TNM8+) was analyzed by Kaplan–Meier analysis and compared with log-rank tests. Factors included in the multivariable model: age, sex, race, comorbidity score, hospital type, insurance status, diagnosis year, and treatment were identified pre hoc and confirmed on univariable analysis. Proportional hazards assumptions for the multivariable Cox models were assessed by a test of proportionality for AJCC7 nodal status. There was no significant deviation from proportionality in the HPV-negative oropharynx (OP) group (P = 0.9). Multivariable Cox proportional hazards models controlling for these demographic factors and treatment were constructed to compare risk by clinical TNM stage groups for the different staging systems. To verify whether misclassification of HPV status in the registry database impacted the effect nodal status has on survival outcomes, a sensitivity analysis using the simex() package in R to apply the SIMEX algorithm on the multivariable Cox model under various misclassification scenarios, ranging from 10% to 90% of misclassification among the observed HPV-negative cohort (11). The C-statistic was used to quantify the predictive ability of the TNM7 and TNM8+ staging systems and referenced to a cohort of patients with HPV-positive OPSCC. For pathologic data, univariable and multivariable analyses were conducted to examine the effect of pN and pTNM on survival with the pTNM7, pTNM8–, and pTNM8+ staging systems. C-statistic values were used to quantify the predictive ability of the three staging systems and referenced to a cohort of patients with HPV-positive OPSCC. Concordance indices were calculated utilizing STATA 16.1 (StataCorp, LLC). All other analyses utilized SPSS 26.0 (SPSS, Inc.). Bivariate analysis using χ2 was used to analyze our categorical variables. Kaplan–Meier survival analysis and multivariable Cox proportional hazards regression analysis were used to identify factors associated with OS. All tests were two sided and the threshold for significance was set at | $\alpha $ | < 0.05.

Data availability statement

Publicly available datasets (SEER and NCDB) were analyzed in this study. These data can be found through the following links:

https://seer.cancer.gov/.

https://www.facs.org/quality-programs/cancer/news/ncdb-puf-090220.

There were 11,147 cases of HPV-negative OPSCC in the NCDB and 3,613 cases in SEER that were included in the clinical nodal classification analysis, representing 32.8% and 27.2% of all OPSCC, respectively, in both databases (Fig. 1). Clinical demographics in the two cohorts are described in Table 1. From a staging standpoint, there were proportionally more T2 tumors in the NCDB cohort (NCDB, 37.3% vs. SEER, 33.9%; P = 0.003) and fewer T4b (NCDB, 4.2% vs. SEER, 5.8%; P < 0.001). T classification was not different between the two databases otherwise. Other than N2a (NCDB, 5.8% vs. SEER, 7.4%; P < 0.001) there was no difference in N classification between the two databases. The majority of patients in both cohorts underwent nonsurgical treatment (NCDB: 74.7% and SEER: 71.2%).

Figure 1.

Inclusion and exclusion criteria for clinical nodal staging survival analyses in NCDB and SEER.

Figure 1.

Inclusion and exclusion criteria for clinical nodal staging survival analyses in NCDB and SEER.

Close modal
Table 1.

Clinical demographic and clinicopathalogic factors of patients with HPV-negative OPSCC from the NCDB and SEER databases.

NCDBSEERTotal
11,147 (75.5)3,613 (24.5)14,760 (100)P
Age     <0.001 
 <50 1,254 (11.2) 352 (9.7) 1,606 (10.9)  
 50–60 3,783 (33.9) 1,136 (31.4) 4,919 (33.3)  
 61–70 3,706 (33.2) 1,214 (33.6) 4,920 (33.3)  
 >70 2,404 (21.6) 911 (25.2) 3,315 (22.5)  
Sex     0.287 
 Male 8,368 (75.1) 2,744 (75.9) 11,112 (75.3)  
 Female 2,779 (24.9) 869 (24.1) 3,648 (24.7)  
Race     <0.001 
 White 9,483 (85.1) 2,991 (82.8) 12,474 (84.5)  
 Black 1,327 (11.9) 458 (12.7) 1,785 (12.1)  
 Other 337 (3) 164 (4.5) 501 (3.4)  
Comorbidities      
 8,754 (78.5)  8,754 (78.5)  
 1+ 2,393 (21.5)  2,393 (21.5)  
Hospital type      
 Community 778 (7)  778 (7)  
 Comprehensive community 3,648 (32.7)  3,648 (32.7)  
 Academic 5,208 (46.7)  5,208 (46.7)  
 Integrated network 1,367 (12.3)  1,367 (12.3)  
 Other/unknown 146 (1.3)  146 (1.3)  
Insurance     <0.001 
 None 633 (5.7) 147 (4.1) 780 (5.3)  
 Private/NOSa 4,578 (41.1) 2,787 (77.1) 7,365 (49.9)  
 Gov't or Medicare 5,744 (51.5) 617 (17.1) 6,361 (43.1)  
 Unknown 192 (1.7) 62 (1.7) 254 (1.7)  
Year     <0.001 
 2010 1,006 (9) 259 (7.2) 1,265 (8.6)  
 2011 1,650 (14.8) 435 (12) 2,085 (14.1)  
 2012 1,940 (17.4) 496 (13.7) 2,436 (16.5)  
 2013 2,177 (19.5) 570 (15.8) 2,747 (18.6)  
 2014 2,133 (19.1) 618 (17.1) 2,751 (18.6)  
 2015 2,241 (20.1) 599 (16.6) 2,840 (19.2)  
 2016  636 (17.6) 636 (4.3)  
T status     <0.001 
 T1 2,525 (22.7) 863 (23.9) 3,388 (23)  
 T2 4,161 (37.3) 1,226 (33.9) 5,387 (36.5)  
 T3 2,421 (21.7) 769 (21.3) 3,190 (21.6)  
 T4a 1,574 (14.1) 547 (15.1) 2,121 (14.4)  
 T4b 466 (4.2) 208 (5.8) 674 (4.6)  
N status     0.007 
 N0 3,392 (30.4) 1,047 (29) 4,439 (30.1)  
 N1 2,050 (18.4) 652 (18) 2,702 (18.3)  
 N2a 641 (5.8) 266 (7.4) 907 (6.1)  
 N2b 2,860 (25.7) 960 (26.6) 3,820 (25.9)  
 N2c 1,759 (15.8) 538 (14.9) 2,297 (15.6)  
 N3 445 (4) 150 (4.2) 595 (4)  
T status (TNM8+   0.002 
 T4 2,040 (18.3) 755 (20.9) 2,795 (18.9)  
      
N status (TNM8+   0.109 
 N0 3,392 (30.4) 1,047 (29) 4,439 (30.1)  
 N1 5,551 (49.8) 1,878 (52) 7,429 (50.3)  
 N2 1,759 (15.8) 538 (14.9) 2,297 (15.6)  
 N3 445 (4) 150 (4.2) 595 (4)  
      
Treatment     <0.001 
 Nonsurgical 8,332 (74.7) 2,573 (71.2) 10,905 (73.9)  
 Surgical 2,685 (24.1) 1,023 (28.3) 3,708 (25.1)  
 Unknown 130 (1.2) 17 (0.5) 147 (1)  
NCDBSEERTotal
11,147 (75.5)3,613 (24.5)14,760 (100)P
Age     <0.001 
 <50 1,254 (11.2) 352 (9.7) 1,606 (10.9)  
 50–60 3,783 (33.9) 1,136 (31.4) 4,919 (33.3)  
 61–70 3,706 (33.2) 1,214 (33.6) 4,920 (33.3)  
 >70 2,404 (21.6) 911 (25.2) 3,315 (22.5)  
Sex     0.287 
 Male 8,368 (75.1) 2,744 (75.9) 11,112 (75.3)  
 Female 2,779 (24.9) 869 (24.1) 3,648 (24.7)  
Race     <0.001 
 White 9,483 (85.1) 2,991 (82.8) 12,474 (84.5)  
 Black 1,327 (11.9) 458 (12.7) 1,785 (12.1)  
 Other 337 (3) 164 (4.5) 501 (3.4)  
Comorbidities      
 8,754 (78.5)  8,754 (78.5)  
 1+ 2,393 (21.5)  2,393 (21.5)  
Hospital type      
 Community 778 (7)  778 (7)  
 Comprehensive community 3,648 (32.7)  3,648 (32.7)  
 Academic 5,208 (46.7)  5,208 (46.7)  
 Integrated network 1,367 (12.3)  1,367 (12.3)  
 Other/unknown 146 (1.3)  146 (1.3)  
Insurance     <0.001 
 None 633 (5.7) 147 (4.1) 780 (5.3)  
 Private/NOSa 4,578 (41.1) 2,787 (77.1) 7,365 (49.9)  
 Gov't or Medicare 5,744 (51.5) 617 (17.1) 6,361 (43.1)  
 Unknown 192 (1.7) 62 (1.7) 254 (1.7)  
Year     <0.001 
 2010 1,006 (9) 259 (7.2) 1,265 (8.6)  
 2011 1,650 (14.8) 435 (12) 2,085 (14.1)  
 2012 1,940 (17.4) 496 (13.7) 2,436 (16.5)  
 2013 2,177 (19.5) 570 (15.8) 2,747 (18.6)  
 2014 2,133 (19.1) 618 (17.1) 2,751 (18.6)  
 2015 2,241 (20.1) 599 (16.6) 2,840 (19.2)  
 2016  636 (17.6) 636 (4.3)  
T status     <0.001 
 T1 2,525 (22.7) 863 (23.9) 3,388 (23)  
 T2 4,161 (37.3) 1,226 (33.9) 5,387 (36.5)  
 T3 2,421 (21.7) 769 (21.3) 3,190 (21.6)  
 T4a 1,574 (14.1) 547 (15.1) 2,121 (14.4)  
 T4b 466 (4.2) 208 (5.8) 674 (4.6)  
N status     0.007 
 N0 3,392 (30.4) 1,047 (29) 4,439 (30.1)  
 N1 2,050 (18.4) 652 (18) 2,702 (18.3)  
 N2a 641 (5.8) 266 (7.4) 907 (6.1)  
 N2b 2,860 (25.7) 960 (26.6) 3,820 (25.9)  
 N2c 1,759 (15.8) 538 (14.9) 2,297 (15.6)  
 N3 445 (4) 150 (4.2) 595 (4)  
T status (TNM8+   0.002 
 T4 2,040 (18.3) 755 (20.9) 2,795 (18.9)  
      
N status (TNM8+   0.109 
 N0 3,392 (30.4) 1,047 (29) 4,439 (30.1)  
 N1 5,551 (49.8) 1,878 (52) 7,429 (50.3)  
 N2 1,759 (15.8) 538 (14.9) 2,297 (15.6)  
 N3 445 (4) 150 (4.2) 595 (4)  
      
Treatment     <0.001 
 Nonsurgical 8,332 (74.7) 2,573 (71.2) 10,905 (73.9)  
 Surgical 2,685 (24.1) 1,023 (28.3) 3,708 (25.1)  
 Unknown 130 (1.2) 17 (0.5) 147 (1)  

aNCDB factor is private insurance and SEER factor is nongovernment insurance not otherwise specified.

Nodal classification

The effect of clinical nodal status for patients with HPV-negative OPSCC was examined in the NCDB cohort. Survival analysis based on nodal status found significant differences based on the N-status (P < 0.001). Ranked pairwise comparisons from highest to lowest 5-year OS demonstrated unexpected results. Survival was highest for N2a (65%) followed by N1 (55%, P = 0.002), cN2b (54%, P = 0.22), N0 (49%, P = 0.001), N2c (40%, P < 0.001), and N3 (38%, P = 0.02), respectively (Fig. 2A). Clinical nodal classification in the SEER cohort (Fig. 2B) and pathologic nodal classification in the NCDB cohort demonstrated similar trends (Supplementary Fig. S1). This unexpected finding was not noted with clinical tumor (T) status, where survival decreased monotonically with increasing T status (pairwise P < 0.001; Supplementary Fig. S2).

Figure 2.

Kaplan–Meier analysis of clinical N status for patients with HPV-negative OPSCC in the NCDB (A) and SEER (B) databases. Cox porportional hazards models risk stratified by clinical N status based on HPV status for OPSCC (C) and oral cavity and larynx squamous cell carcinoma (D).

Figure 2.

Kaplan–Meier analysis of clinical N status for patients with HPV-negative OPSCC in the NCDB (A) and SEER (B) databases. Cox porportional hazards models risk stratified by clinical N status based on HPV status for OPSCC (C) and oral cavity and larynx squamous cell carcinoma (D).

Close modal

To determine whether this discordant nodal survival impact is OPSCC specific, multivariable models incorporating demographics, T status, and treatment (no treatment, radiation alone, chemotherapy alone, chemoradiation, surgery alone, surgery + radiotherapy, surgery + chemo, and surgery + chemoradiation) were used to determine the effect of nodal status on survival among various HPV-positive and HPV-negative head and neck subgroups (Fig. 2C and D). For oral cavity and larynx squamous cell carcinoma, increasing N status was associated with incrementally worse OS (Fig. 2D). However, this pattern did not persist among patients with OPSCC regardless of HPV type. Patients with OPSCC with N1 and N2a disease did not have decreased OS relative to N0 patients in the HPV-negative, -positive, or -unknown cohorts (Fig. 2C). Additional analyses controlling for tumor stage and demographic variables but not treatment were conducted to examine the effects of nodal status on outcomes (Supplementary Fig. S3).

This trend was consistent for patients with both low and high T stage. In the NCDB, patients with T1 disease and low nodal burden (N1 or N2) did not have decreased survival relative to those with clinically negative lymph nodes, (T1N1: HR, 0.66; 0.53–0.83, T1N2a: HR, 0.67; 0.55–0.79). Similarly, among T4b patients, those with early nodal disease did not have decreased survival relative to those with node negative disease (T4bN1: HR, 0.81; 0.65–1.01, T4bN2a: HR, 0.96; 0.81–1.14). Similar results were seen in the SEER clinical and NCDB pathologic staging cohorts (Supplementary Table S1).

Tumor (T) and nodal (N) co-distribution

The TxN distributions stratified by HPV status are presented in balloon plots (Fig. 3A and B). A log-linear model demonstrated a significant association of T and N status that is dependent on HPV status with increased heterogeneity and asymmetry of nodal distribution amongst HPV-positive OPSCC relative to HPV-negative OPSCC (P < 0.001). The heterogeneity was confirmed by evaluating the distribution of nodal disease across dichotomized T, N, and HPV status via Breslow–Day statistic. Again, we find a significant difference in nodal distribution in the HPV-positive group relative to HPV-negative (P < 0.0001). To better define this variant nodal distribution between HPV-positive and -negative OPSCC, we analyzed residual distributions based on the assumption of an equal distribution of nodal disease (i.e., observed versus expected T and N distribution). Tumor residual magnitudes in both SEER and NCDB found lower than expected T1 and T2 disease in HPV-negative disease compared with HPV positive. In more advanced T-stage, HPV-positive disease had lower than expected T3 and T4 disease relative to HPV negative (Fig. 3C). This variant distribution based on HPV status was even more pronounced when assessing nodal status. Here we found a higher-than-expected incidence of N2a and N2b disease in HPV-positive disease whereas there were a higher-than-expected number of N0 patients in HPV-negative disease (Fig. 3D). Taken together, these data further validate the assumption that HPV-positive OPSCC presents with more advanced nodal (N2a and N2b) disease in the setting of smaller primary tumors. HPV-negative cancers, however, often have more advanced primary tumors yet lower than expected nodal metastasis which suggests that T stage is the primary driver of poor survival.

Figure 3.

Balloon plots depicting proportion of patients presenting with various tumor (T) and nodal (N) status in the SEER (A) and NCDB (B) cohorts. χ2 residuals of tumor (C) and nodal (D) residual distribution based on HPV status.

Figure 3.

Balloon plots depicting proportion of patients presenting with various tumor (T) and nodal (N) status in the SEER (A) and NCDB (B) cohorts. χ2 residuals of tumor (C) and nodal (D) residual distribution based on HPV status.

Close modal

Validation and clinical restaging

On the basis of our findings that nodal status is less impactful in our HPV-negative OPSCC cohort and has similar trends to that of HPV-positive OPSCC reported in the literature, we applied the recently established TNM8+ staging criteria to our HPV-negative OPSCC cohort. Nodal restaging demonstrated the best 5-year OS survival for N1 (56%) compared with N0, N2, and N3, (49%, P < 0.001), (40%, P < 0.001), and (38%, P = 0.021), respectively. The restaged SEER cohort showed similar results with the exception that there was no survival difference between N2 and N3 disease (38% vs. 46%, P = 0.36), respectively. Per TNM7, the largest group was stage IVA (49.7%). After restaging with to TNM8+, patients were predominantly distributed to stage I (52.2%), followed by stage II (26.5%), and stage III (21.2%). There was no difference in the proportion of patients redistributed to each stage between the NCDB and SEER cohorts (all P > 0.05).

Risk stratification of overall stage was determined for patients in the NCDB comparing the TNM7 (Fig. 4A) to the TNM8+ (Fig. 4B) staging criteria. When restaged with TNM8+ criteria, there was improved risk stratification in both the NCDB and SEER cohorts (Fig. 4). Risk stratification was also improved with N status when utilizing TNM8+ criteria (Supplementary Fig. S4). Concordance indices (C-statistic) were calculated according to TNM7 and TNM8+ clinical staging systems for patients with HPV-negative OPSCC; this was repeated for HPV-positive patients to provide a comparison group. For HPV-positive patients with OPSCC, the TNM8+ staging system improved concordance within the NCDB cohort from 0.562 | $ \pm $ | 0.004 to 0.635 | $ \pm $ | 0.005 (difference +0.073). A similar improvement in concordance was seen for patients with HPV-negative OPSCC, from 0.561 | $ \pm $ | 0.004 to 0.624 | $ \pm $ | 0.004 (difference +0.063). In the SEER cohort, the C-statistic for patients with HPV-positive OPSCC improved with the TNM8+ clinical staging system, from 0.562 | $\ \pm $ | 0.008 to 0.651 | $ \pm $ | 0.008 (difference +0.089). The improvement in concordance was similar in the patients with HPV-negative OPSCC, from 0.561 | $ \pm $ | 0.008 to 0.625 | $ \pm $ | 0.008 (difference +0.065).

Figure 4.

Kaplan–Meier analysis of HPV-negative patients with OPSCC in the NCDB stratified by TNM7 (A) and TNM8+ (B) overall clinical staging criteria as well as in the SEER database utilizing TNM7 (C) and TNM8+ (D) overall clinical staging criteria.

Figure 4.

Kaplan–Meier analysis of HPV-negative patients with OPSCC in the NCDB stratified by TNM7 (A) and TNM8+ (B) overall clinical staging criteria as well as in the SEER database utilizing TNM7 (C) and TNM8+ (D) overall clinical staging criteria.

Close modal

Finally, given we were analyzing registry databases, a sensitivity analysis was performed to evaluate the impact of potential HPV-status misclassification on nodal status and survival outcomes in the multivariable regression models. Evaluating HPV-status misclassification under various scenarios we failed to find that misclassification across various ranges (10%–90%) of misclassification influenced our findings that oropharynx AJCC7 cN1 and cN2a subjects do as well (in some cases better) than N0 patients. We failed to find a proportion of misclassification of HPV in the HPV-negative group that would result in a significantly higher hazard for the N1 or N2a patients in the multivariable Cox model. We also performed a sensitivity analysis of the NCDB cohort multivariable Cox model among oropharynx subjects with observed HPV test results (either positive or negative.) We used the simex() package in R to apply the SIMEX algorithm on the multivariable Cox model under various misclassification scenarios, ranging from 10% to 90% of misclassification among the observed HPV-negative cohort and failed to find a proportion of misclassification of HPV-negative that would result in a significantly higher hazard for N1 or N2a in the multivariable Cox model. From this, we conclude that misclassification of HPV status, even if it exists in these data, did not affect our conclusion that oropharynx AJCC7 cN1 and cN2a subjects have similar survival (in some cases better) as compared with N0 patients, regardless of HPV status (Supplementary Table S2).

Pathologic restaging

To determine whether the impact of nodal status in OPSCC was related to an anomaly of clinical staging, we sought to evaluate the impact of nodal status based on pathologic staging, including ENE. NCDB patients who underwent surgical resection were pathologically restaged according to pTNM8+ criteria. The largest group per pTNM7 was stage IVA (46.7%). After restaging to pTNM8+, the most common stage was stage I (68.3%) followed by stage II (27.8%) and stage III (3.9%).

Risk stratification by 5-year OS for pN classification and pTNM staging was improved with the updated pTNM8+ criteria relative to either the pTNM7 or pTNM8– criteria (Fig. 5). Multivariable analysis controlling for potential cofounders and analyzed by stage also demonstrated improved hazard discrimination with pTNM8+.

Figure 5.

Kaplan–Meier analysis of overall pathologic staging for HPV-negative patients from the NCDB utilizing pTNM7 (A) and pTNM8– (B) and pTNM8+ (C) staging criteria.

Figure 5.

Kaplan–Meier analysis of overall pathologic staging for HPV-negative patients from the NCDB utilizing pTNM7 (A) and pTNM8– (B) and pTNM8+ (C) staging criteria.

Close modal

Concordance indices were calculated for the pTNM7, pTNM8–, and pTNM8+ pathologic staging systems for HPV-negative patients as well as a comparison group of HPV-positive patients. The pTNM8– staging system demonstrated increased concordance relative to pTNM7 criteria for HPV-positive OPSCC, 0.548 | $ \pm $ | 0.01 to 0.583 | $ \pm $ | 0.01 (difference +0.035) and HPV-negative patients 0.566 | $ \pm $ | 0.01 to 0.580 | $ \pm $ | 0.01 (difference +0.014). Utilization of pTNM8+ staging criteria was associated with a greater improvement in concordance for HPV-positive 0.630 | $ \pm $ | 0.01 (difference +0.082) and HPV-negative disease 0.629 | $ \pm $ | 0.01 (difference +0.063).

To confirm that the improvement in concordance from pTNM7 to pTNM8+ is unique to patients with OPSCC, we performed concordance analyses for oral cavity, larynx, and hypopharynx subsites. For oral cavity cancers, there was a decrease in concordance after restaging from pTNM7, 0.677 | $ \pm $ | 0.004, to pTNM8+, 0.634 ± 0.005. There was no significant difference in concordance after restaging for larynx and hypopharynx cancers. To control for the impact of neck dissection [defined as >10 lymph node (LN) removed], a subset multivariable analysis including patients that underwent surgery and controlled for demographics and overall pathologic stage, neck dissection was not associated with a statistically significant difference in survival (HR, 0.94; 95% CI, 0.78–1.14).

The decreasing predictive capacity of nodal burden in oropharyngeal carcinoma has been studied at the multi-institutional and population level, and is typically attributed to HPV-positive cases (2, 3, 12). In the current study, we demonstrated a reduced impact of nodal metastasis on survival for OPSCC, and it is an effect that is independent of HPV status.

It is well established that HPV-positive OPSCC is commonly associated with nodal spread early in the disease course and unlike traditional head and neck cancers has a limited impact on survival (1, 6, 8). This discovery ultimately led to the updated AJCC 8 staging system for HPV-positive disease which broadened the classification of cN1 and pN1 disease to encompass more and larger lymph nodes than its AJCC 7 predecessor. Under the new staging system, early nodal metastases can still be considered stage I given a small enough oropharyngeal primary. Thus, unlike traditional head and neck cancer, a single nodal metastasis, depending on its size, no longer automatically upstages to stage III or IVA.

Here we were able to confirm that HPV-negative cancers do not develop nodal disease until more advanced primary disease, unlike HPV-positive OPSCC that have a more heterogeneous distribution of nodal disease relative to primary tumor size. However, despite this variance in tumor-nodal distribution between HPV-positive and -negative OPSCC, nodal disease had low impact on survival for both HPV-negative and -positive OPSCC. Nodal staging of 696 patients in the HPV-negative cohort from the prospective, multi-institutional ICON-S trial demonstrated a similar finding to their HPV-positive cohort in that low-level nodal disease was not associated with decreased survival (8). Their N0 group (54%, 47–63) had a comparable 5-year OS to their N1-2a (58%, 50–67) and N2b groups (51%, 44–59). In the current study, we identify similar survival trends in clinical and pathologic nodal status for HPV-negative patients. Relative to N0, early nodal disease, N1-2b, is not associated with worse OS in either the NCDB or SEER cohorts. This effect appears to be limited to the oropharynx subsite and is independent of treatment type and HPV status. As such, it appears that nodal burden has a reduced impact on survival in OPSCC.

On the basis of this finding, we then sought to evaluate the impact of deemphasizing nodal disease as part of the staging of all oropharyngeal cancer, not just HPV-positive oropharyngeal cancer. Here we implemented the validated TNM8+ staging system to restage HPV-negative OPSCC and found qualitatively improved risk stratification base on Kaplan–Meier analysis as well as increased concordance based on C-statistic. We demonstrated improved risk stratification by reducing the impact of nodal disease on survival. This improvement in concordance was not seen after restaging oral cavity, larynx, and hypopharynx cancers which further highlights the uniqueness of this effect to the oropharynx. These data were confirmed in both SEER and NCDB as well as with pathologic data.

Oropharyngeal cancer is currently thought of as two separate diseases based on viral associated differences in etiology, demographics and survival outcomes. However, the findings in this study and others suggest that nodal disease, while highly prevalent in OPSCC, does not confer the same poor outcome as other head and neck cancers (2, 8). This is particularly interesting with regards to our findings for similar survival patterns based on nodal classification that is unique to OPSCC regardless of HPV status (Fig. 2A and B). This suggests a potentially conserved mechanism for metastatic spread specific to the oropharynx. A prudent question to ask is why? To date, knowledge of the biologic mechanism for metastasis in OPSCC remains limited. The few studies investigating metastatic mechanisms are based on molecular drivers of metastasis as well as the local anatomic and tumor microenvironment factors that may facilitate early metastasis. At the anatomic level, Waldeyers ring is the primary site for most oropharyngeal cancers and tonsillar tissue is derived from the endoderm. The convergence of ectodermal and endodermal layers in the oropharynx has been examined using high fidelity microscopy to potentially explain early capacity for metastatic spread. Histologically, normal tonsillar tissue is composed of epithelial cells in close proximity to infiltrating immune cells which leads to dynamic functional and morphologic rearrangement of epithelial cell-to-cell contacts. Moreover, the normal basement membrane is naturally porous allowing robust infiltration of lymphovascular cells (13). This intersection of ectodermal epithelial cells with lack of a true basement membrane and close proximity to endodermal blood and lymphatics vessels likely allows for earlier tumor invasion without the need for more aggressive tumor features to overcome substantial anatomic barriers. In addition, both HPV-positive and HPV-negative OPSCC have similar metastatic nodal topographic distribution (Fig. 3A and B), again suggesting that lymphatic drainage patterns are not an HPV-dependent process, but survival outcomes are (14). One of the key variations is T classification and N classification. While HPV-positive OPSCC has an asymmetric distribution of small primary tumors with large nodal burden, this correlate is not seen in HPV-negative OPSCC. The mechanism as to why lymph node status is less of an important driver of survival in both HPV-negative and HPV-positive oropharyngeal cancer remains unclear and deserves further research.

One of the major limitations of the NCDB and SEER databases is their joint utilization of the collaborative stage site-specific factor 10 (SSF-10) to encode HPV status and associated data collection errors. The intent of this variable was to capture HPV subtypes by genotyping and thus cancer associated HPV is defined in these databases as HPV type 16, 18, or high risk, not otherwise specified (NOS). Collaborative stage coding guidelines were formed prior to widespread usage of p16 IHC as a clinical marker and makes no specific mention of p16 IHC. However, patients undergoing p16 IHC are often miscoded as HPV type 16 and or HPV-positive high risk, NOS when they should be ideally coded as HPV-positive, NOS. Chart review of 824 patients from a single institution who had SSF-10 encoded in SEER found that the positive and negative predictive values of SSF-10 for HPV status were 94% and 95%, respectively (9). When only high-risk HPV subtypes are considered by excluding patients solely tested by p16 IHC, positive and negative predictive values of SSF-10 drop to 51% and 39%, respectively. Our study cohorts likely represent a heterogenous group comprised mostly of p16 IHC negative but HPV DNA unknown patients. To address this potential confounder, a sensitivity analysis addressing potential misclassification of HPV was performed and we failed to find a proportion of misclassification of HPV in the HPV-negative group that would result in a significantly altered hazard for the N1 or N2a patients in the multivariable Cox model. Furthermore, the OS of the OPSCC in both the HPV-negative and -positive groups remain consistent within the literature. Another potential confounder to explain the decreased impact of nodal metastasis in OPSCC is that treatment modalities increase with increasing stage. In our results, we found that N1-2b did better than N0 in HPV-negative OPSCC. One potential explanation is that treatment intensification, such as the addition of chemotherapy, may impart benefit on survival. Interestingly, this pattern was not seen in other HPV-negative subsites, despite similar treatment indications and paradigms for intensification. Furthermore, in patients treated with organ preservation regimens, any stage III or IV disease typically receives both radiotherapy and chemotherapy, thus N1-N3 receive similar treatment and should control for treatment intensification. However, given this is a national database study, it is impossible to know the exact regimen each patient was treated with and serves as a confounding variable. In addition to inaccuracies regarding HPV status, our study is limited by the retrospective nature of large national database studies. The intent of this study was to examine the impact of presenting nodal status on survival at a population level to avoid selection bias. Importantly, smoking is not included, however, it is less likely to be an important confounder given that we have stratified by HPV positivity and are focused on the impact of nodal status on survival. Finally, extranodal extension is a known predictor of worse survival in the head and neck and patients with ENE which is reflected in the improved concordance index utilizing the TNM8– staging system but are not classified differently according to TNM8+ criteria.

In conclusion, nodal burden has been associated with poor outcomes. However, while nodal disease in HPV-positive OPSCC has been shown to have less of an impact on recurrence and survival, there have been limited studies on the impact of nodal disease on survival in HPV-negative OPSCC. In our analysis of two large cancer registries, early nodal metastases are not associated with OS in all patients with oropharyngeal cancer, independent of HPV status. Further research is needed to better understand the biologic etiology of this behavior.

No disclosures were reported.

C.M. Chang: Conceptualization, resources, data curation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.M. Chen: Conceptualization, formal analysis, validation, writing–review and editing. E.L. Bellile: Formal analysis, validation, visualization. L.S. Rozek: Resources, data curation, formal analysis, supervision, validation, writing–review and editing. T.E. Carey: Resources, data curation, formal analysis, supervision, project administration. M.E. Spector: Supervision, project administration, writing–review and editing. G.T. Wolf: Conceptualization, supervision, project administration, writing–review and editing. J.M.G. Taylor: Formal analysis, supervision, writing–original draft. S.B. Chinn: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, project administration, writing–review and editing.

The authors would like to acknowledge the efforts of the NCDB Program and the NCI SEER Program tumor registries in the creation of the NCDB and SEER databases. S.B. Chinn's effort was supported in part by a grant from the NIH/NCI (1K08CA226350-01A1).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Schuller
DE
,
McGuirt
WF
,
McCabe
BF
,
Young
D
.
The prognostic significance of metastatic cervical lymph nodes
.
Laryngoscope
1980
;
90
:
557
70
.
2.
Keane
FK
,
Chen
YH
,
Neville
BA
,
Tishler
RB
,
Schoenfeld
JD
,
Catalano
PJ
, et al
.
Changing prognostic significance of tumor stage and nodal stage in patients with squamous cell carcinoma of the oropharynx in the human papillomavirus era
.
Cancer
2015
;
121
:
2594
602
.
3.
Dahlstrom
KR
,
Calzada
G
,
Hanby
JD
,
Garden
AS
,
Glisson
BS
,
Li
G
, et al
.
An evolution in demographics, treatment, and outcomes of oropharyngeal cancer at a major cancer center: a staging system in need of repair
.
Cancer
2013
;
119
:
81
9
.
4.
Lydiatt
WM
,
Patel
SG
,
O'Sullivan
B
,
Brandenwein
MS
,
Ridge
JA
,
Migliacci
JC
, et al
.
Head and Neck cancers-major changes in the American Joint Committee on cancer eighth edition cancer staging manual
.
CA Cancer J Clin
2017
;
67
:
122
37
.
5.
Amin
MB
,
Greene
FL
,
Edge
SB
,
Compton
CC
,
Gershenwal
JE
,
Brookland
RK
, et al
.
The Eighth Edition AJCC Cancer Staging Manual: continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging
.
CA Cancer J Clin
2017
;
67
:
93
9
.
6.
Ang
KK
,
Harris
J
,
Wheeler
R
,
Weber
R
,
Rosenthal
DI
,
Nguen-Tan
PF
, et al
.
Human papillomavirus and survival of patients with oropharyngeal cancer
.
N Engl J Med
2010
;
363
:
24
35
.
7.
O'Rorke
MA
,
Ellison
MV
,
Murray
LJ
,
Moran
M
,
James
J
,
Anderson
LA
.
Human papillomavirus related head and neck cancer survival: a systematic review and meta-analysis
.
Oral Oncol
2012
;
48
:
1191
201
.
8.
O'Sullivan
B
,
Huang
SH
,
Su
J
,
Garden
AS
,
Sturgis
EM
,
Dahlstrom
K
, et al
.
Development and validation of a staging system for HPV-related oropharyngeal cancer by the International Collaboration on Oropharyngeal cancer Network for Staging (ICON-S): a multicentre cohort study
.
Lancet Oncol
2016
;
17
:
440
51
.
9.
Kahl
AR
,
Charlton
ME
,
Pagedar
NA
,
Sperry
SM
,
Matt
B
,
Platz
C
, et al
.
Accuracy of the HPV status site-specific factor 10 (SSF-10) variable for patients with oropharyngeal cancers in the Iowa Cancer Registry, 2010–2014
.
Head Neck
2018
;
40
:
2199
209
.
10.
Edge
SB
,
Byrd
DR
,
Compton
CC
,
Fritz
AG
,
Greene
FL
Trotti
A
.
AJCC cancer staging manual
. 7th ed.
New York, NY
:
Springer
;
2010
.
11.
Küchenhoff
H
,
Mwalili
SM
,
Lesaffre
E
.
A general method for dealing with misclassification in regression: the misclassification SIMEX
.
Biometrics
2006
;
62
:
85
96
.
12.
Hobelmann
KC
,
Topf
MC
,
Bar-Ad
V
,
Luginbuhl
AJ
,
Keane
WM
,
Curry
JM
, et al
.
AJCC-8ed nodal staging does not predict outcomes in surgically managed HPV-associated oropharyngeal cancer
.
Oral Oncol
2018
;
82
:
138
43
.
13.
Perry
ME
.
The specialised structure of crypt epithelium in the human palatine tonsil and its functional significance
.
J Anat
1994
;
185
:
111
27
.
14.
Amsbaugh
MJ
,
Yusuf
M
,
Cash
E
,
Silverman
C
,
Wilson
E
,
Bumpous
J
, et al
.
Distribution of cervical lymph node metastases from squamous cell carcinoma of the oropharynx in the era of risk stratification using human papillomavirus and smoking status
.
Int J Radiat Oncol Biol Phys
2016
;
96
:
349
53
.

Supplementary data