Extracapsular spread and adjuvant therapy in human papillomavirus-related, p16-positive oropharyngeal carcinoma


  • Parul Sinha MBBS, MS,

    1. Department of Otolaryngology-Head and Neck Surgery, Washington University School of Medicine, St. Louis, Missouri
    Search for more papers by this author
  • James S. Lewis Jr. MD,

    1. Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri
    Search for more papers by this author
  • Jay F. Piccirillo MD,

    1. Department of Otolaryngology-Head and Neck Surgery, Washington University School of Medicine, St. Louis, Missouri
    Search for more papers by this author
  • Dorina Kallogjeri MD, MPH,

    1. Department of Otolaryngology-Head and Neck Surgery, Washington University School of Medicine, St. Louis, Missouri
    Search for more papers by this author
  • Bruce H. Haughey MBChB

    Corresponding author
    1. Department of Otolaryngology-Head and Neck Surgery, Washington University School of Medicine, St. Louis, Missouri
    • Department of Otolaryngology-Head and Neck Surgery, Washington University School of Medicine, 660 S. Euclid Avenue, Campus Box 8115, St. Louis, MO 63110

    Search for more papers by this author
    • Fax: (314) 362-7522

  • We gratefully acknowledge Sue Pagano for help with the pathology slides and the support of Kathryn Trinkaus of the Biostatistics Core, Siteman Comprehensive Cancer Center.



Extracapsular spread (ECS) is commonly used to justify adjuvant chemotherapy in patients with head and neck cancer. The role of ECS as a prognosticator and adjuvant therapy determinant in surgically resected, human papillomavirus-related oropharyngeal squamous cell carcinoma (OPSCC), however, has never been determined.


Of 210 oropharynx patients in a prospective transoral laser microsurgery database, 152 patients who had p16-positive primary OPSCC and pathologically positive necks were eligible for the study. ECS was measured from routine reporting (ECSreport) and by using a novel histologic grading system (ECSgraded). Proportional hazards models and matched analyses were used to compare the impact of ECS and adjuvant therapy on disease-free survival (DFS). Patients with and without graded ECS were matched for T-stage, surgical margins, and adjuvant therapy.


At a median follow-up of 43 months, the presence of ECS was not associated with poorer DFS in multivariate analyses (ECSreport: hazard ratio [HR], 3.42; 95% confidence interval [CI], 0.45-25.88; P = .23; ECSgraded: HR, 2.54; 95% CI, 0.88-7.34; P = .09). T-stage and high-grade ECS, ie soft tissue metastasis (STMgraded) were prognostic. Overall and in the presence of ECS or even STM, adjuvant CRT was not associated with better DFS over radiotherapy alone (HR, 0.25; 95% CI, 0.06-1.13; P = .07). In addition, matched analyses demonstrated no significant reduction in DFS for the presence of ECS versus the absence of ECS or reduced DFS for the administration of adjuvant radiotherapy alone versus CRT in ECS-positive patients.


Routinely reported ECS was not prognostic in this study. Adjuvant CRT versus radiotherapy alone produced no improvement in DFS for ECS-positive patients. The authors propose that de-escalated adjuvant therapy should be considered for patients with p16-positive OPSCC who undergo surgery and that routinely reported ECS should not be used to justify adjuvant chemotherapy. Cancer 2012;3519–3530. © 2011 American Cancer Society.


Extracapsular spread (ECS) in head and neck squamous cell carcinoma (HNSCC) upgrades the disease to “high-risk” status,1 which often is targeted with adjuvant chemoradiotherapy (CRT) to improve locoregional control and survival.2-4 A shift in HNSCC epidemiology toward human papillomavirus (HPV)-related oropharyngeal squamous cell carcinoma (OPSCC),5, 6 a distinct group of tumors characterized by favorable clinical outcomes,7-11 warrants a reappraisal of traditional prognosticators. These prognosticators currently intensify adjuvant treatment, adding toxicity. This reappraisal is especially pertinent with the growth of transoral laser microsurgery (TLM) as a primary treatment modality.12-14

Available reports on ECS in HNSCC demonstrate inconsistencies that limit their extrapolation to HPV-related OPSCC. ECS is described as a negative prognosticator in studies that combine multiple subsites, including oral cavity, larynx, and hypopharynx.15-19 The studies that do analyze oropharyngeal cancer fail to measure the impact of ECS stratified by HPV status.20-22 A significant gap, therefore, exists in the all-important identification of prognosticators for p16-positive oropharynx cancer.

In our recent studies on outcomes of patients with predominantly p16-positive, advanced-stage OPSCC who underwent primary TLM and neck dissection with or without adjuvant therapy, neither presence of lymph node metastasis, high N stage, nor ECS was associated with reduced survival.13, 14 This confirmed the need for a detailed evaluation of the impact of ECS in exclusively p16-positive OPSCC. In the current study, we capture ECS as recorded from the original pathology report and also use a novel grading system of ECS.23 We assess both prognostic and treatment implications, or lack thereof, of ECS in a large cohort of patients with p16-positive, surgically treated OPSCC. To further control for the effect of potential prognostic confounders, we used a matched analysis.


Patient Population

A prospective TLM database of HNSCC patients was searched for patients with oropharyngeal cancer who were treated consecutively from June 1996 through June 2010. Washington University School of Medicine's Human Research Protection Office approved the protocol, and written informed consent was obtained from participants. Eligibility criteria were the presence of untreated, biopsy-proven OPSCC; TLM resection for the primary; pathologically positive lymph node disease; and p16 positivity in the specimens (Fig. 1). Tumor (T) stage was clinical unless it was upstaged by operative or histopathologic findings. Demographic and tumor characteristics, disease stage, treatments, complications, and outcomes for each patient were recorded as they transpired. Patients with prior history of head and neck cancer or distant metastasis were excluded.

Figure 1.

This flow chart illustrates the composition of the study cohort (N = 152). OPSCC indicates oropharyngeal squamous cell carcinoma; TLM, transoral laser microsurgery; N0, negative lymph node status; IHC, immunohistochemistry; ECS, extracapsular spread.

Adjuvant Therapy

The decision for adjuvant therapy after primary TLM resection and neck dissection was made in a multidisciplinary tumor board based on extensive lymph node disease, presence of ECS, positive margins, patient preference, performance status, and metabolic criteria. During the course of patient accrual in the TLM database, a change occurred in adjuvant therapy policy, and patients who had ECS-positive and/or margin-positive disease usually received CRT after May 2004 (66% after, vs 10% before this date).

Extracapsular Spread and p16 Expression

We used 2 different measures to assess the impact of ECS: 1) ECSreport comprised routinely reported data available from several pathologists, and 2) ECSgraded was a stringent categorization that excluded simple capsular thickening without infiltrative borders.23 ECSreport was determined according to the presence or absence of ECS and/or soft tissue metastasis (STMreport) recorded from the pathology report. ECSgraded was determined from a novel lymph node ECS grading system23 that was applied by a single study pathologist who was blinded to treatments and outcomes. In this system, grades 0 and 1 correspond to the absence of ECS; and grades 2, 3, and 4 represent the presence of ECS, with grade 4/STMgraded considered the highest grade.

  1. Grade 0, tumor within a lymph node surrounded by lymphoid tissue;

  2. Grade 1, tumor filling the subcapsular sinus with no intervening lymphoid tissue and with a thickened capsule/pseudocapsule, smooth leading edge, and no extension beyond the capsule;

  3. Grade 2, tumor extending ≤1 mm beyond the capsule;

  4. Grade 3, tumor extending >1 mm beyond the capsule; and

  5. Grade 4, STMgraded, ie, irregular masses of tumor with no histologic evidence of residual lymph node tissue or architecture.

p16 Immunohistochemistry was performed on the surgical specimen using a monoclonal antibody (MTM Laboratories CINTEC, Westborough, Mass) on a Ventana Benchmark immunostainer (Ventana Inc., Tucson, Ariz) following standard protocols.23 Ninety-nine percent of p16-positive specimens had >75% of cells stained positive.

Study Endpoints

The primary endpoint was disease-free survival (DFS), which was calculated from the date of surgery to the date of either death or first recurrence. The secondary endpoints included overall survival (OS), disease-specific survival (DSS), and patterns of recurrence (local, regional, distant). OS was calculated from the date of surgery to the date of death from any cause. DSS was calculated from the date of surgery to the date of death from either oropharyngeal cancer or any direct treatment effects.

Statistical Analysis

Statistical analysis was conducted using the SAS statistical software package (version 9.2; SAS Institute Inc., Cary, NC). Heterogeneity between any 2 groups was investigated using the chi-square test or the Fisher exact test for categorical data and the independent t test or its nonparametric equivalent Mann-Whitney U test for continuous data. All statistical tests used were 2-sided. The survival probability with 95% confidence interval (CI) was estimated using the Kaplan-Meier method and was compared with the log-rank test. Cox proportional hazards regression analysis was used to generate a hazard ratio (HR) with 95% CI for each variable. Multivariate models were created to adjust for covariates that were identified as significant predictors of survival. The assumption of proportionality was tested and met for the Cox analysis. For all analyses, statistical significance was indicated at a P value < .05.

Patients were grouped into those with or without ECSgraded and were matched for T-stage (T1-T2 vs T3-T4), margins status (presence or absence), and type of adjuvant therapy (CRT or radiotherapy [RT]). We also grouped patients with ECSgraded into those who received RT alone versus those who received CRT to perform an additional matched analysis (T-stage and margin status). Similar matching was performed for STMgraded. Factors that were not matched were tested for heterogeneity between groups.


Patient and Tumor Characteristics

One hundred fifty-two patients were eligible for the study. The baseline characteristics were similar in patients with and without ECSreport (Table 1). The median follow-up was 43 months (minimum-maximum, 12-173 months). In the entire cohort, 137 of 152 patients (90%) survived, 9 patients (6%) died of disease, and 6 patients (4%) died from causes unrelated to oropharyngeal cancer.

Table 1. Baseline Characteristics of the Study Cohort Stratified According to Extracapsular Spread Measured by Routine Reporting (ECS report) and by Novel Histologic Grading (ECS graded)
  ECSReport: No. of Patients (%) ECSGraded: No. of Patients (%) 
CharacteristicTotal No. (%)AbsentPresentPAbsentPresentP
  • Abbreviations: AJCC, American Joint Committee on Cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; ECS, extracapsular spread; ECSGraded, extracapsular spread measured by novel histologic grading; ECSReport, extracapsular spread measured by routine reporting.

  • a

    Comorbidities were captured using the Adult Comorbidity Evaluation 27 (see Piccirillo et al24).

  • b

    Patients were classified as current smokers if the period of cessation was <6 months before presentation and as heavy smokers if the lifetime tobacco exposure was ≥20 pack-years (see Hyland et al25 and D'souza et al26).

  • c

    After the change in treatment policy for “high-risk” criteria, 32 patients with positive ECSreport did not receive chemotherapy because of age ≥75 years (N = 3), refusal (N = 25), or death before the administration of any adjuvant therapy (N = 4).

Total152 (100)28 (18)124 (82) 73 (48)79 (52) 
Age: Median [range], y56 [27-83]58 [40-79]55 [27-83].2155 [35-83]56 [27-86].39
Sex   .53  .75
 Men134 (88)26 (93)108 (87) 65 (89)69 (87) 
 Women18 (12)2 (7)16 (13) 8 (11)10 (13) 
Disease site   .89  .56
 Base of tongue71 (47)14 (50)57 (46) 36 (49)35 (44) 
 Tonsil and soft palate81 (53)14 (50)67 (54) 37 (51)44 (56) 
Comorbidity scorea   .38  .41
 0, None72 (47)15(54)57(46) 36 (49)36 (46) 
 1, Mild55 (36)7 (25)48 (39) 27 (37)28 (35) 
 2, Moderate19 (13)4 (14)15 (12) 6 (8)13 (17) 
 3, Severe6 (4)2 (7)4 (3) 4 (6)2 (2) 
ECOG PS   .07  .98
 0-1140 (92)25 (89)115 (93) 65 (89)75 (94) 
 2-48 (6)3 (11)5 (4) 4 (6)4 (6) 
 Unknown4 (2)04 (3) 4 (5)0 (0) 
Smoking statusb   .88  .47
 Never smoked64 (42)12 (43)52 (42) 34 (47)30 (38) 
 Former smoker55 (36)11 (39)44 (35) 25 (34)30 (38) 
 Current smoker32 (21)5 (18)27 (22) 13 (18)19 (24) 
 Unknown1 (1)0 (0)1 (1) 1 (1)0 (0) 
Smoking dose, pack-yearsb   .02  .37
 <20, Light25 (29)9 (56)16 (23) 13 (35)12 (15) 
 ≥20, Heavy58 (67)7 (44)51 (72) 24 (63)34 (69) 
 Unknown4 (4)0 (0)4 (5) 1 (2)3 (6) 
T-stage   .15  .09
 T152 (34)12 (43)40 (32) 29 (40)23 (29) 
 T251 (34)12 (43)39 (31) 26 (35)25 (32) 
 T328 (18)3 (11)25 (20) 13 (18)15 (19) 
 T421 (14)1 (3)20 (17) 5 (7)16 (20) 
N-stage   .04  .05
 N124 (16)9 (32)15 (12) 17 (23)7 (9) 
 N2118 (77)18 (64)100 (81) 52 (71)66 (83) 
 N310 (7)1 (4)9 (7) 4 (6)6 (8) 
AJCC stage   .007  .003
 III22 (14)9 (32)13 (10) 17 (23)5 (6) 
 IV130 (86)19 (68)111 (90) 56 (77)74 (94) 
Surgical margins   .99  .13
 Negative138 (91)26 (93)112 (90) 69 (94)69 (87) 
 Positive14 (9)2 (7)12 (10) 4 (6)10 (13) 
No. of lymph nodes   <.001  <.001
 1-285 (56)24 (86)61 (49) 56 (77)29 (37) 
 >266 (43)4 (14)62 (50) 17 (23)49 (62) 
 Unknown1 (1)0 (0)1 (1) 0 (0)1 (1) 
Level of lymph nodes   <.001  <.001
 Single51 (34)18 (64)33 (27) 37 (51)14 (17) 
 Multiple80 (52)5 (18)75 (60) 22 (30)58 (74) 
 Unknown21 (14)5 (18)16 (13) 14 (19)7 (9) 
Median size of metastatic lymph node [range], cm3.5 [1-10]3 [1-7]3.5 [1-10].423.25[0.6-7.6]3.5 [0.5-9.8].45
Median tumor depth [range], mm7 [0.5-25]9 [0.5-13.0]7 [0.8-25.0].937 [0.5-15.0]6 [1.2-25.0].99
Perineural invasion   .74  .03
 Present19 (12)2 (7)17 (14) 4 (5)15 (19) 
 Absent116 (76)19 (68)97 (78) 56 (77)60 (76) 
 Unknown17 (11)7 (25)10 (8) 13 (18)4 (5) 
Angioinvasion   .25  .03
 Present26 (17)2 (7)24 (19) 7 (10)19 (24) 
 Absent109 (72)20 (71)89 (72) 54 (74)55 (70) 
 Unknown17 (11)6 (21)11 (9) 12 (16)5 (6) 
Lymphatic invasion   .12  .05
 Present38 (25)3 (11)35 (28) 12 (16)26 (33) 
 Absent96 (63)18 (64)78 (63) 48 (66)48 (61) 
 Unknown18 (12)7 (25)11 (9) 13 (18)5 (6) 
Adjuvant therapyc   <.001  <.001
 None19 (13)8 (29)11 (9) 13 (18)6 (7) 
 Radiotherapy alone66 (43)18 (64)48 (39) 41 (56)25 (32) 
 Chemoradiotherapy67 (44)2 (7)65 (52) 2 (26)48 (61) 

Extracapsular Spread

Routine reporting (ECS report)

According to pathology reports, 124 of 152 patients (82%) had ECSreport, and 80 of 152 patients (53%) had STMreport.

Novel histologic grading (ECS graded)

The presence of ECS was detected in 79 of 152 patients (52%), and the numbers were reduced by 45 (37%) compared with ECSreport. Grades were assigned as follows: Nineteen patients (12%) had grade 0, 54 patients (36%) had grade 1, 12 patients (8%) had grade 2 ECS, 22 patients (14%) had grade 3 ECS, and 45 patients (30%) had grade 4 ECS/STMgraded.

Overall, the ECSgraded system recorded fewer ECS-positive patients (N = 79) compared with ECSreport (N = 124). This reduction is accounted for by exclusion from ECSgraded of those patients who had tumor filling the subcapsular sinus or a thickened capsule (grade 1).23 Such cases were frequently diagnosed as bona fide ECS in routine reporting (ECSreport).

Forty-six pairs and 42 pairs were generated randomly for the evaluation of ECSgraded and STMgraded, respectively. For adjuvant CRT versus RT analyses, matching yielded 24 pairs of ECSgraded-positive and 10 pairs of STMgraded-positive cases.

Adjuvant Therapy

Overall, 133 of 152 patients (87%) received adjuvant therapy, including 66 patients (43%) who received RT and 67 patients (44%) who received CRT. Because ECS was a criterion for adding chemotherapy, significantly more patients in this category received CRT than RT (Table 1). Between the RT and CRT groups, there were no significant differences in timing, dosage, or field size and technique of RT administration (Table 2). The planned RT dose was delivered in 131 of 133 patients (98%). Of 67 patients who received CRT, 62 (93%) received cisplatin, and 5 (7%) received cetuximab. In total, 32 of 67 patients (52%) completed 3 cycles of cisplatin, 27 (43%) completed 2 cycles, and 3 (5%) tolerated only 1 cycle.

Table 2. Radiotherapy Parameters by Type of Adjuvant Therapy (Chemoradiotherapy vs Radiotherapy Alone) in All Patients Who Received Adjuvant Therapy, N = 133
 RT Type: No. of Patients (%)Primary SiteIpsilateral Neck (N = 133)Contralateral NeckRT Initiation >42 daysTotal No.
Adjuvant TherapyConventionalIMRTNo. of Patients (%)Median Dose [Minimum- Maximum], GyMedian Dose [Minimum- Maximum], GyNo. of Patients (%)Median Dose [Minimum- Maximum], GyNo. of Patients (%) 
  • Abbreviations: CRT, chemoradiotherapy; Gy, grays; IMRT, intensity-modulated radiotherapy; RT, radiotherapy.

  • a

    RT was terminated before the completion of planned dose at 49 Gy for 1 patient because of severe acute toxicity.

  • b

    RT was terminated before the completion of planned dose at 36 Gy for 1 patient because of severe acute toxicity.

CRTa166 (99)65 (97)66 [54-70]66 [54-70]57 (85)56(54-66)29 (43)67
RTb957 (86)56 (85)66 [54-70]66 [54-70]54 (82)56 (50-66)31 (47)66
Total10123      133

Patterns of Recurrence

Ten patients (6.6%) developed recurrences, including 2 primary recurrences and 2 regional recurrences, for an overall locoregional control rate of 97%. There were 6 distant metastases (4%). The median time to recurrence did not differ significantly according to the presence or absence of ECSreport (33 months vs 40 months, respectively; P = .40) nor did the recurrence rates (difference, 0.03; 95% CI, −0.06 to 0.12; P = .689). There was no significant difference between positive versus negative ECSgraded groups for recurrences (difference, 0.07; 95% CI, −0.3 to 14.9; P = .10). Two of the 10 recurrences developed in patients who received no adjuvant therapy (N = 19). The pattern of recurrence by type of adjuvant treatment was assessed (Fig. 2). In patients who received RT alone (N = 66), there were 5 recurrences (7.5%), including 1 local recurrence, 2 regional recurrences, and 2 distant metastases. In the concurrent CRT group (N = 67), 3 patients (4.5%) developed recurrences, all of which were distant metastases. This difference of 3% (95% CI, −4.8% to 10.8%) in recurrence rates between the RT and CRT groups was not statistically significant (P = .492; Fisher exact test).

Figure 2.

This Kaplan-Meier estimate illustrates the probability of disease control in patients who had extracapsular spread according to adjuvant therapy type (P = .96). RT indicates radiotherapy alone; ChemoRT, chemoradiotherapy.

Distant metastasis occurred in 5% of patients in the ECSreport group (N = 6 of 124 patients), 7.5% of patients in the STMreport group (N = 6 of 80 patients), 6% of patients in the ECSgraded group (N = 6 of 79 patients), and 11% of patients in the STMgraded group (N = 5 of 45 patients). Of the 6 patients who had distant metastasis in the ECS-positive group, 3 had received CRT, 2 had received RT alone, and 1 had refused any adjuvant therapy.

Survival Outcomes

Extracapsular spread: Routine reporting (ECS report)

DFS in patients with ECSreport was not significantly reduced compared with patients without ECSreport (P = .21), with 3-year Kaplan-Meier estimates of 89% (95% CI, 84%-95%) and 94% (95%CI, 83%-100%), respectively (Fig. 3, top left). The 95% CI for this 5% difference ranged from −3.6% to 13.6%. The 3-year DFS outcome did not differ significantly for ECSreport-positive patients who received CRT (91.8%) compared to those who received RT (94.5%; P = .74). The 95% CI for this 2.7% difference ranged from −6.5% to 11.9% (Fig. 3, top right). Similarly, there were no significant differences in 3-year DFS for STMreport-positive patients versus STMreport-negative patients (90% vs 92%; P = .48) nor for patients with STMreport results who received with RT versus CRT (95% vs 92%; P = .86) (Fig. 3, bottom).

Figure 3.

(Top Left) Kaplan-Meier disease-free survival (DFS) estimates are shown for the study cohort stratified according to the presence (N = 124) or absence (N = 28) of extracapsular spread (ECS) measured from routine reporting (ECSreport) (P = .21). (Top Right) Kaplan-Meier DFS estimates are shown for patients who had positive ECSreport results stratified according to the receipt of adjuvant chemoradiotherapy (ChemoRT) (N = 65) versus radiotherapy alone (RT) (N = 48) (P = .74). (Bottom Left) Kaplan-Meier DFS estimates are shown for the study cohort stratified according to the presence (N = 80) or absence (N = 72) of soft tissue metastasis (STMreport) (P = .48). (Bottom Right) Kaplan-Meier DFS estimates are shown for patients who had positive STMreport results stratified according to the receipt of adjuvant chemoRT (N = 46) versus RT (N = 26; P = .86).

Extracapsular spread: Novel histologic grading (ECS graded)

DFS did not differ significantly according to the presence versus the absence of ECSgraded (92% vs 97%, respectively; P = .08) or for ECSgraded-positive patients who received CRT versus RT alone (87.8% vs 89.4%; P = .98). However, patients with STMgraded had significantly reduced DFS compared with patients without STMgraded (80% vs 93%; P = .02). However, for patients with STMgraded, DFS was no better with CRT than with RT alone (80% vs 83%; P = .32).

Multivariate Analysis

In multivariate analysis (Table 2), ECSreport, ECSgraded, and STMreport were not significant predictors of worse survival for any of the 3 survival outcomes. The presence of the highest degree of graded ECS (ie, STMgraded) was a significant determinant of reduced survival outcomes. When evaluated in the patients who received adjuvant therapy (N = 133), however, STMgraded was no longer a prognostic factor for DFS, DSS, or OS.

Early T-stage and the administration of any adjuvant therapy were associated significantly with improved DFS, DSS, and OS; whereas negative margins remained significant for better DSS and OS. Adjuvant CRT versus RT was not associated significantly with better outcomes for either DFS, DSS, or OS.

Smoking status/dose, lymph node status, the level or size of metastatic lymph nodes, perineural invasion, comorbidity, and performance status were not associated significantly with any of the study endpoints and node number was colinear with STM (Table 3, c).

Table 3. Univariate and Multivariate Cox Proportional Hazard Regression Analysis of Disease-Free, Disease-Specific, and Overall Survival
 Univariate AnalysisMultivariate Analysis
VariableHR (95% CI)PHR (95% CI)P
  • Abbreviations: CI, confidence interval; CRT, chemoradiotherapy; DFS, disease-free survival; DSS, disease-specific survival; ECOG PS, Eastern Cooperative Oncology Group performance status; ECS, extracapsular spread; ECSgraded, extracapsular spread measured by novel histologic grading; ECSreport, extracapsular spread measured by routine reporting; HR, hazard ratio; OS, overall survival, RT, radiotherapy alone; STM, soft tissue metastasis; STMgraded, soft tissue metastasis measured by novel histologic grading; STMreport, soft tissue metastasis measured by routine reporting.

  • a

    Variable with a statistically significant HR.

  • b

    In the model that was confined to patients who received adjuvant therapy (N = 133), STMgraded was not associated significantly with reduced DFS (HR, 1.73; 95% CI, 0.45-6.74; P .43), DSS (HR, 2.4; 95% CI, 0.30-19.03; P = .41), or OS (HR, 2.71; 95% CI, 0.63-11.63; P .18).

  • c

    The number of lymph nodes was excluded from the final multivariate model for DFS because of collinearity with STMgraded (P < .001).

  • d

    Only variables with an HR that reached statistical significance and those related to ECS and STM are reported for DFS and OS.

  • e

    This variable had an insignificant HR with very wide 95% CIs.

 Age: Continuous1.06 (1.01-1.12).030a1.03 (0.98-1.10).26
 Sex: Men vs women1.06 (0.24-4.65).94  
 Comorbidity score: 2-3 vs 0-11.13 (0.32-3.98).85  
 ECOG PS: 2-4 vs 0-12.33 (0.52-10.42).27  
 Smoker: Ever vs never1.99 (0.63-6.27).27  
 Smoking dose: Heavy vs light3.88 (0.49-30.70).20  
 T-stage: T3-T4 vs T1-T28.0 (2.68-23.84)<.001a7.13 (2.17-23.46).001a
 N-stage: N2-N3 vs N11.68 (0.38-7.34).49  
 Surgical margins: Positive vs. negative4.82 (1.52-15.3).008a3.20 (0.92-11.08).07
 ECSreport: Yes vs no3.42 (0.45-25.88).23  
 STMreport: Yes vs no1.43 (0.52-3.95).49  
 ECSgraded: Yes vs no2.54 (0.88-7.34).09  
 STMgraded: Yes vs no3.15 (1.17-8.44).02a4.39 (1.39-13.82).01a,b
 Adjuvant therapy    
  Any vs none7.64 (2.73-21.4)<.001a6.63 (1.85-23.8)0.004a
  CRT vs RT0.25 (0.06-1.13).0711.85(0.36-9.42)0.46
 No. of lymph nodes: >2 vs ≤24.96 (1.59-15.43).006a,c  
 Lymph node level: Multiple vs single4.54 (1.01-20.32).05  
 Size of metastatic lymph node: Continuous0.67 (0.44-1.02).06  
 Tumor depth: Continuous1.04 (0.89-1.22).57  
 Perineural invasion: Yes vs no1.8 (0.5-6.5).35  
 Angioinvasion: Yes vs no1.61 (0.51-5.05).42  
 Lymphatic invasion: Yes vs no1.2 (0.41-3.51).74  
 T-stage: T3-T4 vs T1-T219.3(2.36-158.3).006a10.2 (1.1-97.8).04a
 Surgical margins: Positive vs negative11.3 (2.83-45.5)<.001a9.14 (1.6-52.88).01a
 ECSreport: Yes vs noe.40  
 STMreport: Yes vs no1.4(0.34-6.04).62  
 ECSgraded: Yes vs noe.15  
 STMgraded: Yes vs no4.74 (1.13-19.9).03a24.0 (1.7-340.2).01a,b
 Angioinvasion: Yes vs no5.92 (1.32-26.5).02a1.33 (0.12-8.28).76
 Adjuvant therapy    
  Any vs none12.2 (2.96-49.9)<.001a17.43 (2.41-126).005a
  CRT vs RT1.16 (0.16-8.25).883  
 T-stage: T3-T4 vs T1-T213.7 (3.64-51.5)<.001a8.1 (1.9-33.8).004a
 Surgical margins: Positive vs negative5.73 (1.74-18.9).004a4.6 (1.2-18.2).03a
 ECSreport: Yes vs noe.26  
 STMreport: Yes vs no2.15 (0.68-6.88).20  
 ECSgraded: Yes vs no4.22 (1.17-15.2).03a3.98 (0.95-16.71).06
 STMgraded: Yes vs no4.17 (1.44-12.1).009a3.34 (1.12-9.9).03a,b
 Adjuvant therapy    
  Any vs none9.09 (3.09-26.7)<.001a9.44 (2.46-36.3).001a
  CRT vs RT0.9 (0.2-4.10).93  

Matched Analyses

Assessment of extracapsular spread/soft tissue metastasis

Even after matching for T-stage, margin status. and type of adjuvant therapy, ECSgraded was not a significant predictor of worse DFS (HR, 1.61; 95% CI, 0.51-5.06; P = .42) nor was STMgraded (HR, 2.1; 95% CI, 0.63-7.4; P = .22) (Fig. 4, top).

Figure 4.

(Top Left) Kaplan-Meier disease-free survival (DFS) estimates are shown for the study cohort matched for T-stage, surgical margins, and adjuvant therapy type (N = 46 pairs) stratified according to the presence or absence of extracapsular spread (ECS) based on a novel histologic grading system (ECSgraded) (P = .42). (Top Right) Kaplan-Meier DFS estimates are shown for the cohort matched for T-stage, surgical margins, and adjuvant therapy type (N = 42 pairs) stratified according to the presence or absence of soft tissue metastasis (STMgraded) (P = .22). (Bottom Left). Kaplan-Meier DFS estimates are shown for the matched ECSgraded-positive cohort stratified according to the receipt of adjuvant chemoradiotherapy (ChemoRT) versus radiotherapy alone (RT) (P = .68). (Bottom Right). Kaplan-Meier DFS estimates are shown for the matched STMgraded-positive cohort stratified according to the receipt of adjuvant chemoRT versus RT (P = .53).

Assessment of chemoradiotherapy versus radiotherapy alone in patients with ECS/soft tissue metastasis

CRT did not confer a significant risk reduction over RT alone for DFS in patients with ECSgraded (HR, 0.65; 95% CI, 0.06-7.29; P = .73) or STMgraded (HR, 2.78; 95% CI, 0.14-54.68; P = .50) (Fig. 4, bottom).


We observed no significant impact for ECS on survival outcomes in our institution's routine pathologic reporting. This lack of prognostic effect was not associated with chemotherapy-intensified adjuvant treatment. In an effort to apply a more discriminating pathological analysis, we used ECS grading, which, in general, also did not correlate with survival outcomes except for the highest grade (STMgraded), which was associated significantly with reduced survival outcomes. However, STM's negative impact was not observed in patients who received adjuvant therapy, either RT or CRT. This probably explains the mechanism of benefit to survival conferred by adjuvant RT, ie, the reduced negative prognostic impact of STMgraded, notably without additional benefit from chemotherapy.

Our study demonstrated that, although the absolute number of disease-related deaths (N = 9) differed according to the presence or absence of ECS, this variable did not correlate with reduced DFS or DSS. It is critical to note that the 9 patients who died of disease harbored other independently significant, negative prognosticators (7 patients had T3/T4 primary tumors, 4 patients had positive surgical margins, 5 patients had angioinvasion, 6 patients had STMgraded, and 4 patients did not receive any adjuvant therapy). Thus, we observed that the presence of ECS, among other important negative prognostic factors, was associated significantly with a greater number of disease-related deaths. However, after controlling for these other important prognostic factors through multivariate analysis, ECS, either by report or graded, was not a statistically significant predictor of survival.

Overall, therefore, we conclude that, in patients with p16-positive OPSCC, the prognostic effect of ECS is confined to the most severe grade, STM, ie, when tumor has obliterated the lymph node with no histologic evidence of residual lymph node tissue or architecture.23, 27, 28 Its negative impact was obviated by radiotherapy. Proposals to explain the lack of prognostic effect of routinely reported or low grades of ECS on survival include the possibility that these patients already had benefited from intensified adjuvant therapy. We addressed this possibility first by controlling for adjuvant treatment type (CRT vs RT alone) in our Cox regression multivariate models. Second, the paired analysis, matched for the presence and type of adjuvant therapy, failed to demonstrate a significant reduction in survival outcomes according to the presence of ECS versus its absence (See also next section for matched CRT vs RT study, Fig. 4). These analyses further reinforce our observation of weak or no prognostic relevance of ECS in patients with p16-positive OPSCC and caution against speculation that equivalent survivorship in the presence of ECS was caused by intensified adjuvant treatment. Thus, surgery followed by adjuvant RT to the neck attenuates the negative prognostication of ECS, an observation that is supported by previous reports.29, 30

Equivalence of Adjuvant Treatment Types in p16-Positive Oropharyngeal Carcinoma With Extracapsular Spread

Despite results from a meta-analysis of nonheterogeneous trials that demonstrated no significant benefit from adjuvant chemotherapy,31 the intensification of adjuvant treatment in the presence of ECS by CRT often is advocated. This is on the basis of Radiation Therapy Oncology Group (RTOG)3 and European Organization for Research and Treatment of Cancer (EORTC)4 trials in patients with head and neck cancer that included a pooled mix of primary sites. The majority of primary sites (58%3 and 70%,4 respectively) used in those investigations were non-oropharyngeal, and HPV status was unknown. ECS also was mixed with positive margins and other negative prognosticators, for which the improved outcomes were not stratified. Therefore, the precise impact of prognosticators and their treatment implications in the oropharyngeal cancer population, specifically those associated with HPV, is currently unpublished.

In our population of patients with p16-positive OPSCC, adjuvant CRT was not associated with increased survival. This was evident both for the entire study cohort and exclusively for the ECSreport and ECSgraded patients. CRT versus RT in patients with ECSgraded lacked any difference in survival after matching for T-stage and margin status. Furthermore, the matched groups with STMgraded, the highest degree of ECS, had the same survival with the 2 different adjuvant regimens. Therefore, our results diverge from what is reported on patients with non-HPV-related head and neck cancer in the available literature, which frequently advocates the use of postoperative chemotherapy versus RT in the presence of ECS. Our results are not surprising given the unique biology of p16-positive disease, and are intuitively reasonable given the high prevalence of ECS in a precisely defined cohort of patients with excellent survival.

Patterns of recurrence in our study indicated that distant metastasis (6 of 10 recurrences) was more common than locoregional recurrence (4 of 10 recurrences). A higher incidence of distant metastasis has been associated with ECS,32, 33 but the role of adjuvant chemotherapy for systemic control in such patients remains unclear. For example, the RTOG trial3 identified no significant reduction in the incidence of distant metastasis in “high-risk” patients who received adjuvant CRT (20%) versus RT alone (23%), and in the EORTC trial, the rates were 21% versus 25%, respectively.4

Compounding these issues, the trials document a 2-fold or greater increase in the incidence of severe acute toxicity from 34% in the RT group to 77% in the CRT group in the RTOG trial3 and 21% to 41%, respectively, in the EORTC trial.4 In addition, the administration of concurrent CRT has been associated with increased odds of long-term swallowing dysfunction in both definitive34, 35 and adjuvant14 treatment settings.

Our current study does not focus on evaluating adjuvant treatment-related toxicities but strongly supports the hypothesis that, in patients with p16-positive OPSCC who undergo surgery, the mere presence of ECS is not an accurate indicator of a poor prognosis. Therefore, the automatic application of ECS as justification for an intensified adjuvant protocol comprising CRT, particularly in the background of the toxicity described above, should be questioned. Just as the negative prognostication of ECS in the past led to addition of chemotherapy, conversely, subtraction of chemotherapy should be considered in patients with p16-positive OPSCC who do not have a poor prognosis in the presence of ECS, when managed surgically as indicated by our current results.

We demonstrate here and also have confirmed from previous studies13, 14 in predominantly p16-positive OPSCC that high T-stage is the strongest prognosticator for poor outcomes. Therefore, it is more logical to consider high T-stage as well as STMgraded and positive margin status as possible indications for intensified adjuvant therapy. A second treatment-related implication from our current study introduces the more stringent pathologist's interpretation of ECS, preventing an “upgrade” of risk status that is not justified by our data. This study is unique in characterizing both the prognostic impact and the extent of ECS in exclusively p16-positive oropharyngeal cancer and its effect on survival.

There are certain limitations in our study. First, this is a single-institution study. However, our recent multicenter study,14 with a greater sample size, had similar findings regarding ECS. Second, the current study is nonrandomized. However, adequate accrual for a randomized controlled trial to achieve sufficient statistical power to detect a significant survival outcome difference would be formidable. This is because the biologically driven good prognosis and low rates of failure (97% locoregional control, 6.5% recurrence rate in our study) in surgically treated patients with p16-positive OPSCC results in too few outcome events.14-16, 19 For the purpose of planning a future study, we estimate that, to be able to detect a statistically significant 5% difference (3-year DFS in ECS-negative patients vs ECS-positive patients, 94% vs 89%) in 3-year DFS between ECS-negative and ECS-positive patients with 80% power at an alpha of .05, in total, 2058 patients would be required. To detect a statistically significant 2.7% difference (3-year DFS with RT vs CRT, 94.5% vs 91.8%) between RT and CRT in ECS-positive patients with 80% power at an alpha of .05, we would require a total of 3252 patients. Unfortunately, there are multiple reasons why the accrual of such a large sample size from 1 or even several institutions is impractical, including patient concern for chemotherapy-related toxicity.

Third, a new, nonvalidated grading system to measure ECS may lack reproducibility. However, the method is uncomplicated, teachable, and clearly more discriminating for the detection of ECS, important features when adverse outcomes are rare and the factor under study has direct treatment implications. Furthermore, STMgraded emerged as a significant independent prognosticator. Finally, other histologic criteria may exist in these tumors that are prognostic but went unobserved. Although a small minority of patients recurred or died, it would be highly desirable to discover predictors of these events.

In summary, we conclude that ECS, as a generally reported pathologic classification, did not predict a poorer prognosis in surgically treated patients with p16-positive OPSCC. This is in marked contrast to previous studies, albeit with cohorts that included heterogeneous groups of patients with head and neck cancer and risk factors that were nonspecific to HPV-related etiology. The 1 exception is graded STM, although its effect was negated by adjuvant RT. These results were obtained independent of treatment variables; any potential benefit from chemotherapy was controlled for by using different levels of analysis. Adjuvant CRT versus RT alone conferred no additional benefits to any of the study endpoints in the presence of ECS or STM, even in closely matched groups. On the basis of our data, it is reasonable to propose that, in surgically managed patients with low T-stage, adjuvant RT alone is sufficient.


No specific funding was disclosed.


The authors made no disclosures.