Prognostic importance of radiologic extranodal extension in HPV-positive oropharyngeal carcinoma and its potential role in refining TNM-8 cN-classification

Radiotherapy and Oncology 144 (2020) 13–22

Contents lists available at ScienceDirect

Radiotherapy and Oncology journal homepage: www.thegreenjournal.com

Original Article

Prognostic importance of radiologic extranodal extension in HPV-positive oropharyngeal carcinoma and its potential role in refining TNM-8 cN-classification Shao Hui Huang a,b,⇑, Brian O’Sullivan a,b, Jie Su c, Eric Bartlett d, John Kim a, John N. Waldron a,b, Jolie Ringash a, John R. de Almeida b, Scott Bratman a, Aaron Hansen e, Andrew Bayley a, John Cho a, Meredith Giuliani a, Andrew Hope a, Ali Hosni a, Anna Spreafico e, Lillian Siu e, Douglas Chepeha b, Lt Tong a, Wei Xu c, Eugene Yu d a Department of Radiation Oncology, the Princess Margaret Cancer Centre/University of Toronto; b Department of Otolaryngology – Head & Neck Surgery, the Princess Margaret Cancer Centre/University of Toronto; c Department of Biostatistics, the Princess Margaret Cancer Centre/University of Toronto; d Department of Neuroradiology and Head and Neck Imaging, Princess Margaret Cancer Centre/University of Toronto; and e Division of Medical Oncology, the Princess Margaret Cancer Centre/University of Toronto, Canada

a r t i c l e

i n f o

Article history: Received 5 September 2019 Received in revised form 16 October 2019 Accepted 17 October 2019

Keywords: HPV Oropharyngeal carcinoma Staging cN-classification Outcome Prognosis

a b s t r a c t Purpose: This study examines outcome heterogeneity and potential to refine the TNM-8 cN-classification using radiologic extranodal extension (rENE) in a contemporary HPV-positive (HPV+) oropharyngeal carcinoma (OPC) cohort. Methods: All HPV+ OPC treated with definitive IMRT from 2010-2015 were included. Pre-treatment CT/ MR of cN+ cases were reviewed by a head-neck radiologist for rENE. Overall survival (OS) and disease-free survival (DFS) were compared between rENE-positive (rENE+) vs rENE-negative (rENE ). Multivariable analysis (MVA) for OS confirmed the prognostic value of rENE. Refined cN-classifications for new TNM staging proposals were evaluated against TNM-8 using established criteria. Results: A total of 517 cN+ (rENE+: 97; rENE : 420) and 41 cN0 cases were identified. The rENE+ proportion increased with rising N-category (N1/N2/N3: 11%/19%/84%, p < 0.001). Median follow-up was 5.1 years. Compared to rENE , rENE+ patients had a lower 5-year OS (56% vs 85%) and DFS (46% vs 83%) overall, and in N1 (OS: 57% vs 89%; DFS: 51% vs 87%) and N2 subsets (OS: 45% and 76%; DFS: 33% vs 74%) (all p < 0.001). MVA confirmed the prognostic value of rENE for OS (HR = 3.86, p < 0.001) and DFS (HR = 3.89, p < 0.001). We proposed two new cN-classifications: Schema1 reclassified any N_rENE+ as New_N3; Schema2 reclassified N1_rENE+ as New_N2 and N2_rENE+ as New_N3. Stage incorporating either Schema1 (ranked 1st) or Schema2 (ranked 2nd) cN-categories outperformed TNM-8. Conclusion: This study confirms that rENE is prognostically important and facilitates understanding of known outcome heterogeneity within TNM-8 in HPV+ OPC patients. rENE is a promising parameter to refine the TNM-8 cN-classifications. Ó 2019 Elsevier B.V. All rights reserved. Radiotherapy and Oncology 144 (2020)13–22

The recognition of improved prognosis in HPV-related (HPV+) oropharyngeal carcinoma (OPC) compared to traditional HPVnegative disease, prompted a new TNM classification for HPV+ OPC [1,2] in the 8th edition TNM (TNM-8). Compared to the 7th edition (TNM-7), the major change in TNM-8 was the definition of the N-classification; cTNM-8 now classifies ipsilateral neck nodes not greater than 6 cm as cN1, and bilateral neck nodes no more than 6 cm as cN2 [3,4]. Although many studies have shown

⇑ Corresponding author at: Department of Radiation Oncology, The Princess Margaret Cancer Centre, University of Toronto/University Health Network, 610 University Ave., Toronto, ON M5G 2M9, Canada. E-mail address: [email protected] (S.H. Huang). https://doi.org/10.1016/j.radonc.2019.10.011 0167-8140/Ó 2019 Elsevier B.V. All rights reserved.

that the TNM-8 outperforms TNM-7 in depicting prognosis, reservations remain [5–8]. For example, more than 50% of HPV+ OPC patients are now classified with cN1 disease, but heterogeneity exists in clinical presentation and prognosis for this subset. Although the new stage classification depicts overall prognosis well, it has not altered treatment decision-making in routine clinical care [9]. Identifying a subset with poor prognosis based on a widely available and reliable pre-treatment clinical/radiological assessment method has become increasingly important for current and future efforts to deintensify treatment in this population. Several radiological features have demonstrated potential to identify subsets with poorer prognosis. Radiologic extranodal extension (rENE) appears to be one of the most promising prognostic nodal

14

rENE refining HPV+ OPC cN Classification

features [10] and is consistent with the inclusion of more extensive regional node categorization in traditional (non-viral-mediated) head and neck cancers (HNC) in the TNM-8. The presence of rENE reflects tumor invasion through the nodal capsule and beyond, and can manifest in several forms [Table 1] [11–23]. A coalescent nodal mass (so-called ‘matted nodes’), a special form of rENE, is shown to be prognostic for distant metastasis (DM) and survival [24–26]. Ordinarily, HPV+ OPC patients with such adverse features have been excluded from de-intensification trials, such as the NRGHN002 (NCT02254278) study.

Previously, we showed that rENE can be reliably assessed by Head-Neck (HN) radiologists, demonstrating high specificity for pathologic ENE (pENE) [20] with excellent inter-rater/intra-rater concordance in HPV-negative HNC [20] and HPV+ OPC [10]. In addition, cN+ stage I HPV+ OPC with rENE+ portends poor prognosis [10]. This observation provides a real possibility to use rENE to refine the full spectrum of the cN-classification. We first confirm the prognostic value of rENE in all cN+ HPV+ OPC patients and subsequently develop new cN-classification Schema(s) incorporating rENE for consideration in the future. Finally, we explore

Table 1 Selected Literature Regarding rENE Definition and rENE-pENE Correlation in Head and Neck Cancer. First author published year case No.

Imaging slice thickness

Mancuso [11] 1983 N = 25 HNSCC

CT 4 mm

Carvalho [12] 1991 N = 28 HNSCC

CT 4 mm

Souter [13] 2009 N = 149 HNSCC

CT 3–5 mm

King [14] 2004 N = 17 HNSCC

CT 3–5 mm

Url [15] 2013 N = 49 HNSCC

CT 2–3 mm

Chai [16] 2013 N = 100 HNSCC

CT 2.5 mm

Kahn [17] 2014 N = 111 OPC

Prabhu [19] 2014 N = 432 HNSCC

Contrastenhanced CT 2.5 mm

Maxwell [18] 2015 N = 65 HPV+ OPC

CT 2.5 mm

Geltzeiler [21] 2017 N = 100 HPV+ OPC

Contrastenhanced CT

Amullar [20] 2018N = 483 OSCC

Contrastenhanced CT2 mm Contrastenhanced CT3 mm

Noor [23]2019N = 80 HPV+ OPC

McMullen [22] 2019 N = 92 OPC (HPV+: 70)

Not described

Definition of rENE  Ill-defined staining margin without clear distinction between it and surrounding fat, or  Evidence of edema or thickening of surrounding fibroadipose tissue or muscles  A node with irregular speculated borders, or  Loss of the fat planes around the node and thickening of the fascia adjacent to the node, or  Apparent invasion of an adjacent structure  Enhancing nodal margins, or  Alterations in adjacent fat, or  Loss of margin definition  Presence of indistinct margins, irregular capsular enhancement, and infiltration into surrounding structures  Apparent fat and soft tissue infiltration, or  Infiltration of sternocleidomastoid muscle, internal jugular vein or carotid artery.  Capsular contour irregularity, poorly defined nodal margins, and infiltration of adjacent fat planes  Certainty of rENE using 5-point scale: 1-definitely ENE–; 2-likely ENE–;3-equivocal ENE; 4-likely ENE+; 5-definitely ENE+  A thick-walled, enhancing nodal margin  Loss of outer nodal margin definition  Infiltration of the adjacent fat planes around portions of the node rENE based on initial radiology report:  Irregular borders and/or perinodal fat stranding, or  Invasion of adjacent structures  pENE grade:  Grade 1, tumor reaching the nodal capsule  Grade 2: 1 mm of pENE  Grade 3: >1 mm pENE  Grade 4: soft tissue metastasis    

Nodal capsular contour irregularity, or Poorly defined nodal margins, or Loss of intervening fat planes, and/or Invasion of adjacent structures Certainty of rENE using 5-point scale: 1-definitely rENE ; 2likely rENE ;3-equivocal rENE; 4-likely rENE+; 5-definitely ENE+  Regular, partially or severely irregular nodal border  Unequivocal ill-defined nodal borders

 Assessing internal characteristics, capsule contour, presence of perinodal fat stranding and invasion into surrounding structures Certainty of rENE using 5-point scale: 1-definitely rENE ; 2likely rENE ;3-equivocal rENE; 4-likely rENE+; 5-definitely ENE+  Poorly defined peripheral margins, and/or  Presence of ‘‘matted” nodes

rENE-pENE concordance  CT accurately predicted 9/11 patients with pENE

 Sensitivity 62.5%  Specificity: 60%

Two observers:  Sensitivity: 66% & 80%  Specificity: 91% & 90% CT & MR:  Sensitivity: 65% & 78%  Specificity: 93% & 86% Two observers:  Sensitivity: 73% & 76%  Specificity: 91% & 91%  Inter-rater Kappa: 0.86 % of pENE+ within rENE+ (two observers)  Likely rENE+: 79% & 61%  Definitely rENE+: 100% & 85%  Sensitivity: 61%  Specificity: 57%

 Sensitivity: 44%  Specificity: 98% Higher sensitivity with higher grade of pENE: - Grade 1–2: 19% - Grade 3: 53% - Grade 4: 72%  Adjacent structure invasion was a better predictor than irregular borders/fat stranding for pENE Two observers:  Sensitivity: 55% & 47%  Specificity: 70% & 85% % of pENE+ by certainty of rENE+  Likely rENE+: 62% & 67%  Definitely rENE+: 100% & 100% Severely irregular boarders:  Sensitivity: 21%  Specificity: 100%  Sensitivity: 52%  Specificity: 96% Two observers  Sensitivity: 56.5% & 60.9%  Specificity: 73.3% & 66.7% Inter-rater kappa for rENE: 0.40 % of pENE+ by certainty of rENE+:  Likely rENE+: 60% & 67%  Definitely rENE+: 85% & 64%  Sensitivity: 79%  Specificity: 52%

Abbreviation: rENE: radiologic extranodal extension; pENE: pathologic extranodal extension; HNSCC: head and neck squamous cell carcinoma; OPC: oropharyngeal carcinoma; HPV+: human papillomavirus positive; HPV–: human papillomavirus negative; OSCC: oral cavity squamous cell carcinoma; TORS: transoral-robotic surgery.

S.H. Huang et al. / Radiotherapy and Oncology 144 (2020) 13–22

feasibility in recognizing various rENE patterns evident from the scientific literature and assess their potential differential prognostic impact. Methods Study population With institutional ethics board approval, we reviewed all newly diagnosed non-metastatic (M0) HPV+ OPC patients treated between Jan 2010 to Dec 2015 using consistent treatment (daily volumetric image-guided IMRT) and diagnostic/staging image scanning protocols (contrast-enhanced CT with 2 mm slice thickness and/or MR with 3 mm slice thickness10) in our institution. Clinical characteristics, treatment, and outcome information were obtained prospectively at point-of-care [27] using similar scheduling to contemporary clinical trials. Tumor HPV status was ascertained prospectively using p16 immunohistochemistry staining as a surrogate marker (strong and diffuse nuclear and cytoplasmic staining in > 70% tumor cells) and supplemented by PCR for confirmation of presence of high-risk HPV DNA for cases with equivocal p16 staining. All patients were reclassified to TNM-8 [3–4]. Exclusion criteria for the cN+ cohort were: removal of lymph node (LN) (excisional biopsy or neck dissection [ND]) prior to staging CT/MR scan, or no available imaging within 2 months prior to radiotherapy (RT). The reason to exclude cases with a >2 months scan-to-RT interval was to avoid potential clinically-relevant tumor interval growth (e.g. possible rENE developing after staging CT/MRI) [20]. To reflect this, if multiple scans were available, the scan closer to RT initiation was used for analysis. Assessment of radiologic features including rENE Pre-treatment staging CT or MR of all cN+ cases performed in our institution were re-reviewed by a Head & Neck (HN)

15

radiologist (EY), blinded to clinical information and outcomes. Presence/absence of rENE (rENE+/rENE ) and certainty of rENE+ were recorded [Fig. 1] using the following criteria determined a priori by two experienced HN radiologists (EY and EB) and, to be more inclusive, discussed with the principal author (SHH) after reviewing and summarizing the available literatures:  Pattern_1: tumor invasion through a single LN capsule but confined to surrounding fat (clearly discernible loss of sharp plane between the LN capsule and its surrounding fat)  Pattern_2a: tumor invasion through the nodal capsule of 2 separate LNs  Pattern_2b: tumor invasion through the nodal capsule in 2 adjacent LNs resulting in the formation of a coalescent LN mass  Pattern_3: tumor invasion beyond surrounding nodal fat planes to invade or encase muscles and neurovascular structures Determination of rENE was made after reviewing all axial, coronal and sagittal plane slices. Equivocal/ambiguous rENE was classified as rENE . Other radiologic features were also recorded using criteria described previously [10].

Treatment and outcome assessment Patients were managed by a multidisciplinary team according to institutional protocols (based on TNM-7). All were treated with definitive IMRT ± chemotherapy: TNM-7 Stage I-II, and some Stage III patients received altered fractionation IMRT-alone while most Stage III/IV patients underwent concurrent high-dose cisplatin chemo-radiotherapy (CRT) [1,10], unless unfit for chemotherapy where altered fractionation IMRT-alone, occasionally enhanced by epithelial growth factor receptor inhibition (EGFRI), was used. Treatment decisions were audited at a weekly tumor board. All patients were treated with image-guided intensity-modulated RT or CRT. RT target delineation and plans of >80% cases were

Fig. 1. Various Patterns of Radiologic Extranodal Extension. Pattern_1: tumor invasion through a single nodal capsule into adjacent fat only (A); Pattern_2a: tumor invasion through the nodal capsule of two separable lymph nodes (B); Pattern_2b: tumor invasion through the nodal capsule of two adjacent lymph nodes forming a coalescent nodal mass (C); Pattern_3: tumor invasion beyond surrounding nodal fat plane into adjacent structures, such as, muscles and vessels (D).

16

rENE refining HPV+ OPC cN Classification

peer-reviewed in a weekly quality assurance round. Daily volumetric cone-beam CT ensured accurate RT delivery. Patients were assessed weekly or more frequently during RT by a team comprising radiation oncologists, nurses, dietitians, and other health care professionals. Based on our routine clinical practice [28], patients were evaluated at 3 months by a multidisciplinary team for treatment response (clinical exam and CT/MR of HN +/ PET). Patient with radiologic residual LN alone without adverse features (e.g. necrosis or positive PET) underwent close imaging-surveillance at 2–3 monthly intervals. Post-RT ND was undertaken if there was clinical/radiologic evidence of tumor re-growth and considered as regional failure if pathologically positive [29]. Disease surveillance was conducted 3-monthly or more frequently for the first two years, 4-monthly for the 3rd year, 6-monthly from year 4 to 5, and annually thereafter. Local or regional failures were confirmed histologically. Distant failures were recorded based on unequivocal clinical/radiological ± histological evidences. Solitary lung metastasis were confirmed by p16-positive IHC staining of metastatic lesions evaluated by biopsy or excision. Statistical analysis Clinical characteristics were compared between rENE+ vs rENE cohorts within cN+ patients using either the chi-square test for categorical variables or the non-parametric Kruskal-Wallis tests for continuous variables. Overall survival (OS) and disease-free survival (DFS) (any failure/death) were calculated using Kaplan-Meier methods. Loco-regional control (LRC) and distant control (DC) were calculated using competing-risk methods (deaths without sitespecific events as competing risk). ‘‘Time-to-event” episodes were calculated from date-of-diagnosis. Outcomes were compared using the log-rank or Gray’s Test. Cox regression was applied for univariable (UVA) and multivariable analysis (MVA) on OS and DFS. MVA models included the following variables +/ rENE: age, smoking pack-years, T-category, N-category, and systemic treatment. The concordance-index (C-index) [30] was calculated for MVA models with vs without inclusion of the rENE parameter. All tests were two-tailed using a significance level of 0.05. New cN-classification schema(s) were proposed based on adjusted hazard ratio (aHR) for risk of death conferred by rENE status in MVA. Candidate stage groupings (including cN0) based on rENE-refined new cN-classification(s) were evaluated compared to the TNM-8 using established criteria (OS as the end point) [1– 2,31,32]: 1). ‘‘hazard consistency (homogeneity)” within each stage group, 2). ‘‘hazard discrimination (distinctiveness)” across the different stage groups, 3). ‘‘explained variance”: the percentage of survival variation attributable to ‘‘stage” in the MVA, 4). ‘‘likelihood difference”: the difference of ‘goodness of fit’ between the MVA models with and without the ‘‘stage” variable, and 5). ‘‘sample size balance” across stage groups. Finally, to assess inter-rater reliability and intra-rater reproducibility in assessing rENE and its pattern, a subset of cN+ cases were randomly selected based on power calculation. Presence/ absence of rENE and rENE pattern was evaluated by a second HN radiologist (EB) independently and re-evaluated by the same initial radiologist (EY) after a three-month interval blinded to initial assessment to avoid memory bias. Kappa coefficients (j) were calculated. Results During the 2010–2015 period, a total of 606 HPV+ OPC patients were treated in our institution. Forty-eight were excluded due to LN removal prior to CT/MR scan (n = 15) and >2 months CT/MRto-RT interval (n = 33). The remaining 558 cases, including 41

cN0 and 517 cN+, were eligible for this study (Supplement-1: CONSORT Diagram). For the 517 cN+ cohort, median interval from CT (n = 168) or MR (n = 347) scan to RT initiation was 2.3 weeks (0.1–12 weeks). The clinical characteristics of cN0 (n = 41), cN +_rENE+ (n = 97), and cN+_rENE (n = 420) are shown (Table 2). The proportion of rENE+ increased with rising N-category [N1: 36/323 (11%); N2: 30/157 (19%); and N3: 31/37 (84%), p < 0.001] and higher TNM-8 stage [stage I: 24/218 (11%); stage II: 26/178 (15%); and stage III: 47/121 (39%), p < 0.001]. The frequency of rENE pattern was as follows: Pattern_1 (absence of capsule in a single LN): 8 (8%) (N1/N2/N3: 3/2/3), Pattern_2a (absence of capsule in two separated LNs): 2 (2%) (both N1), Pattern_2b (a coalescent LN mass): 49 (51%) (N1/N2/N3: 19/16/14), and Pattern_3 (invading beyond adjacent fat plane into surrounding structures): 38 (39%) (N1/N2/N3: 12/12/14) (Supplement-2). Compared to rENE , the rENE+ cohort had more retropharyngeal and necrotic LNs. A similar proportion in both cohorts received systemic treatment (p = 0.230). Median follow-up was 5.1 years (range 0.6–9.5) for the cN+ patients. A total of 30 patients with locoregional failures (LRFs) and 74 with distant metastases (DMs) were identified. Compared to rENE , the rENE+ cohort had inferior 5-year LRC (89% vs 96%), DC (61% vs 92%), DFS (46% vs 83%), and OS (56% vs 85%) (all p < 0.001) [Table 2]. Inferior 5-year outcomes (except LRC) with rENE+ vs rENE were also expected within the TNM-8 N1 and N2 subsets, respectively. There was no significant interaction between TNM-8N and rENE: OS: p = 0.96; PFS: p = 0.92. Of note the difference in outcomes between N3_rENE+ vs N3_rENE was not significant: 5-year LRC: 87% vs 100% (p = 0.387); DC: 74% vs 100% (p = 0.192); DFS: 53% vs 100% (p = 0.082); OS: 69% vs 100% (p = 0.188) [Fig. 2]. Exploratory analyses showed that Stage I RT-alone rENE patients (n = 63), compared to rENE+ (n = 6) had exceptionally high 5-year OS (86% vs 44%, p = 0.016), DFS (87% vs 33%, p = 0.001), LRC (100% vs 83%, p = 0.002), and DC (95% vs 61%, p = 0.014) compared to rENE+ (n = 6). Stage I CRT patients also demonstrated significant differences in 5-year OS (95% vs 74%, p = 0.006) and DFS (92% vs 70%, p = 0.011) between rENE (n = 131) and rENE+ (n = 18), but LRC was similarly high for both rENE subsets (97% vs 94%, p = 0.592) [Supplement-3]. UVA showed that all patterns of rENE+ had inferior OS and DFS compared to rENE with similar HRs: Patterns 1/2a/2b/3 vs rENE : 3.71/4.45/3.35/3.71 for death (OS); 3.50/3.92/3.96/4.15 for recurrence/death (DFS). MVA showed that rENE was the strongest prognostic factor for both OS [HR 3.86 (95% CI 2.49–5.97)] and DFS [HR 3.89 (2.62–5.77)] [Table 3]. Based on aHR (Supplement-4), two new cN-classification schemas were proposed: Schema1 reclassified any rENE+ as New_N3; Schema2 reclassified N1_rENE+ as New_N2 and N2_rENE+ as New_N3. The 5-year outcomes by Schema1_N and Schema2_N and TNM-8 are shown (Fig. 3). The c-index of the MVA models for OS and DFS, including Schema1 (OS: 0.748; DFS: 0.729) and Schema2 (OS: 0.740; DFS: 0.721) cN-classifications, were both higher compared to that using TNM-8 cN (OS: 0.717; DFS: 0.690) [Table 3]. The stage grouping algorithms with the Schema1 cNclassifications ranked first in ‘‘explained variance”, ‘‘likelihood difference” and ‘‘sample size balance”, while Schema2 ranked first in ‘‘hazard consistency”, and both outperformed TNM-8 in OS prediction [Supplement-5]. Since the MVA including Schema1 N had the highest c-index and the stage groups with Schema1 N had the highest overall evaluation ranking, we recommend the Schema1 cN-classification (reclassifying any rENE+ into New_N3) ahead of Schema2 (step-wise N reclassification) for consideration in subsequent revisions of TNM. Finally, we assessed inter-rater and intra-rater agreement in rENE and its pattern. Of the 120 randomly selected cN+ cases, intra-rater and inter-rater kappa for rENE+ (any pattern) were 0.77 (0.65–0.89) and 0.84 (0.74–0.95), respectively. The major dis-

17

S.H. Huang et al. / Radiotherapy and Oncology 144 (2020) 13–22 Table 2 Clinical Characteristics of cN0, cN+_rENE+ and cN+_rENE- Patients. Covariate

cN0 (n = 41)

cN+ (n = 517)

cN+_rENE

Age [Median (range)] Gender Female Male ECOG PS ECOG 0–1 ECOG 2–3 Smoking PY [Median (range)] History of Smoking Current Former-smoker Non-smoker T (TNM-8) T1 T2 T3 T4 N (TNM-8) N0 N1 N2 N3 Schema1_N N0 N1 N2 N3 Schema2_N N0 N1 N2 N3 Stage (TNM-8) Stage I Stage II Stage III CT/MR-RT Interval in weeks [median (range)]

61 (33–87)

59 (35–85)

7 (17) 34 (83)

Total LN [Median (range)] rENE Pattern rENE Pattern_1 Pattern_2a Pattern_2b Pattern_3 RPLN Yes No Cystic Yes No Necrotic Yes No Treatment Modality RT Alone RT + Chemotherapy RT + EGFRI RT Alone Regimen 70 Gy/35f/6 w, 6f/w 70 Gy/35f/7w, 5f/w 64 Gy/40f/4w, BID, 10f/w 60 Gy/25f/5w, 5f/w FU in years [Median (range)] 5-year Outcomes OS DFS LRC DC Vital Status Alive Deceased

(n = 420)

cN+_rENE+ (n = 97)

p-value (rENE+ vs rENE )

59 (36–85)

59 (35–82)

0.880

75 (15) 442 (85)

68 (16) 352 (84)

7 (7) 90 (93)

0.025

39 (95) 2 (5) 15 (0–55)

490 (95) 27 (5) 10 (0–108)

401 (95) 19 (5) 9 (0–108)

89 (92) 8 (8) 16 (0–90)

0.140

8 (20) 20 (49) 13 (32)

131 (25) 212 (41) 174 (34)

94 (23) 180 (43) 146 (35)

37 (38) 32 (33) 28 (29)

0.008

3 (7) 20 (49) 9 (22) 9 (22)

106 (21) 183 (35) 138 (27) 90 (17)

88 (21) 156 (37) 108 (26) 68 (16)

18 27 30 22

0.180

41 0 0 0

0 323 (62) 157 (30) 37 (7)

0 287 (68) 127 (30) 6 (1)

0 36 (36) 30 (31) 31 (32)

<0.001

41 0 0 0

0 287 (56) 127 (25) 103 (20)

0 287 (68) 127 (30) 6 (1)

0 0 (0) 0 (0) 97 (100)

<0.001

41 0 0 0

0 287 (56) 163 (32) 67 (13)

0 287 (68) 127 (30) 6 (1)

0 0 (0) 36 (37) 61 (63)

<0.001

23 (56) 9 (22) 9 (22) NA

194 (46) 152 (36) 74 (18) 2.3 (0.1–8.6) 3 (1–19)

24 (24) 26 (27) 47 (49) 2.2 (0.3–8.5) 10 (1–48)

<0.001

NA

218 (42) 178 (34) 121 (23) 2.3 (0.1–8.6) 4 (1–48)

NA NA NA NA NA

420 (81) 8 (2) 2 (0) 49 (9) 38 (7)

420 (100) 0 (0) 0 (0) 0 (0) 0 (0)

0 (0) 8 (8) 2 (2) 49 (51) 38 (39)

NA NA

90 (17) 427 (83)

57 (14) 363 (86)

33 (33) 64 (67)

<0.001

NA NA

195 (38) 322 (62)

151 (36) 269 (64)

44 (45) 53 (55)

0.100

NA NA

345 (67) 172 (33)

255 (61) 165 (39)

90 (93) 7 (7)

<0.001

33 (80) 8 (20) 0

136 (26) 334 (65) 47 (9)

115 (28) 264 (63) 41 (10)

21 (22) 70 (72) 6 (6)

21 (64) 4 (12) 5 (15) 3 (9) 5.0 (1.8–7.3) 83% (72–97) 75% (62–91) 95% (81–99) 94% (75–98)

101 (74) 9 (7) 23 (17) 3 (2) 5.1 (0.6–9.5) 80% (76–84) 76% (72–80) 94% (92–96) 86% (82–89)

83 (72) 9 (8) 21 (18) 2 (2) 5.1 (0.6–9.3) 85% (82–89) 83% (80–87) 95% (93–97) 92% (88–94)

18 (86) 0 2 (10) 1 (5) 5.2 (1.4–9.5) 56% (46–68) 46% (37–57) 89% (80–93) 61% (50–70)

34 (83) 7 (17)

419 (81) 98 (19)

363 (86) 57 (14)

56 (58) 41 (42)

(18) (28) (31) (23)

0.130

0.710 <0.001

0.230

0.250

0.800 <0.001 <0.001 0.009 <0.001 <0.001 (continued on next page)

18

rENE refining HPV+ OPC cN Classification

Table 2 (continued) Covariate

cN0 (n = 41)

cN+ (n = 517)

cN+_rENE

Cause of Death Index Cancer Other Cancer Other Cause Unknown

2 1 3 1

71 (72) 7 (7) 16 (16) 4 (4)

39 (68) 5 (9) 10 (17) 3 (5)

(29) (14) (43) (14)

(n = 420)

cN+_rENE+ (n = 97)

p-value (rENE+ vs rENE )

32 (78) 2 (5) 6 (15) 1 (2)

0.720

Abbreviations: rENE: presence of radiologic extranodal extension; rENE : absence of radiologic extranodal extension; ECOG PS: The Eastern Cooperative Oncology Group Performance Status Score; Smoking PY: smoking pack-years; EGFRI: epithelial growth factor receptor inhibitor; RT: radiotherapy; LN: lymph node; RPLN: retropharyngeal lymph node; FU: follow-up; 70 Gy/35f/6w, 6f/w: 70 Gy in 35 fractions over 6 weeks, 6 fractions per week; BID: twice daily, 6 hours apart. rENE Pattern: Pattern_1: tumor invasion through a single nodal capsule but confined to surrounding fat; Pattern_2a: tumor invasion through the nodal capsule of two or more separable lymph nodes; Pattern_2b: tumor invasion through the nodal capsule of two or more adjacent lymph nodes forming a coalescent nodal mass; Pattern_3: tumor invasion beyond surrounding nodal fat plane into adjacent structures, such as, muscles and vessels.

cordance was seen for Pattern_2b rENE (coalescent nodes) with intra-/inter-rater kappa of 0.49 (0.29–0.69)/0.39 (0.15–0.63) while Pattern_3 rENE had very high intra/inter-rater kappa 0.89 (0.77– 1.00)/0.89 (0.77–1.00). Kappa for Pattern_1 was non-assessable due to its rarity. Intra-/inter-rater concordance for rENE was similar on CT (n = 36) and MR (n = 84): Intra-rater: 0.73 vs 0.79; Interrater: 0.85 vs 0.84. [Supplement-6]. Discussion Our study shows that rENE most frequently presents as pattern_2b/3 while patterns 1/2a are rare in HPV+ OPC. The proportion of rENE+ cases increases with higher N; rENE is a strong prognostic factor for DM. Two new rENE-incorporated cNclassification schemas are proposed: Schema1 reclassifies any N_rENE+ to New_N3 while Schema2 reclassifies N1_ENE+ to N2 and N2_rENE+ to N3 (step-wised reclassification). Stage groupings using both new cN-classifications with the same staging algorithms outperform existing TNM-8 with the Schema1 stage grouping exhibiting the best overall performance. Exploratory analysis shows that TNM-8 stage I rENE subgroup has an excellent 5year LRC (100%) and DC (95%) even if treated by RT-alone (although small number), suggesting that this may be the optimal subset to evaluate for de-intensification. Addition of cisplatinbased systemic agents improved LRC in rENE+ but did not fully negate the impact of rENE+ on DM, suggesting that novel agents merit exploration in rENE+ patient. Finally, rENE can be reliably assessed by expert HN radiologists using stringent criteria with satisfactory inter-/intra-rater concordance, supporting the inclusion of rENE parameter in staging and clinical care. For a new parameter to be considered in a stage classification, the first criterion to be upheld is prognostic importance. The prognostic value of rENE for DM has been shown in stage I HPV+ OPC previously [10]. Our current study extends this to all cN+ patients and affirms the prognostic value of rENE for cN1, cN2, and cN3. The results echo findings from recent surgical series [33–35] indicating that pENE performs adversely in HPV+ OPC. The present study shows that rENE+ is the strongest prognostic factor in MVA with a aHR for death of 3.9, higher than reported surgical series [33,36]. It is plausible that the more overt soft tissue involvement embodied by the rENE parameter has greater prognostic impact compared to the full pENE rubric where additionally detected subclinical microscopic disease would dilute the effect. Our results also suggest that the prognostic value of pENE may need to be further evaluated by stratifying the extent of pENE. Besides strong prognostication, reproducibility and reliability in measuring such a parameter as well ease of procurement is also important. High specificity and modest sensitivity of rENE for pENE in HNC has been demonstrated in several studies [Table 1]. The modest sensitivity is not surprising since less extensive pENE identified by the microscope will not be readily visible on imaging. However, the modest sensitivity for pENE does not prevent its

use in clinical staging since the prognostic value in this setting is already very strong and the high specificity favors its inclusion. The aforementioned comments prompt consideration of the inconsistent reliability of identifying rENE versus pENE, potentially related to image quality, scanning protocol (slice thickness), and certainty of rENE judgement by radiologists. Moreover, pretreatment radiological assessment is not expected to fully compete with pathological assessment of microscopic disease in the same way that clinical N-staging remains fundamentally important, even though not as sensitive as pathological N-Staging, yet highly specific with few false positive cases. We recommend stringent definitions for rENE where only ‘‘unequivocal” or ‘‘definite” radiological evidence of tumor invasion through the LN capsule should be considered rENE+. Using such criteria, we demonstrated that rENE can be reliably assessed with satisfactory inter-rater and intra-rater concordance. However, rENE determination with lesser degrees of certainty, could reduce the concordance with pENE, thereby diminishing its prognostic value. This may partially explain why rENE is considered nonprognostic by some authors [37,38]. We recommend recording only unequivocal rENE as rENE+ and any ambiguity classified as rENE , since overcall might dilute its prognostic impact due to inclusion of unrepresentative findings. This is in consistent with the UICC and AJCC staging rule: where there is doubt, the lower (least ominous) stage level is applied. Thus, training and education should be emphasized before rENE is included in N-classification. Besides subjective certainty of rENE by radiologists, pattern of rENE may also influence reliability and reproducibility of rENE ascertainment. We found that intra-rater concordance for coalescent nodes is relatively low in Pattern_2b, which may reflect difficulty in assessing the capsule status when 2 LNs are intimately adjacent. Therefore, it may be difficult to differentiate a coalescent (so-called ‘matted’) nodal mass (rENE+) from clustered LNs (LN abutting each other but where nodal capsules are intact without rENE). To improve objectivity of assessing rENE, radiomics and machine learning with artificial intelligence [39] may provide a promising solution. A major concern about TNM-8 collapsing N1-N2b into a single N1-category and classifying T1-2N1 as early stage (stage I) is under-treating some patients if using numerically identified stage designated guidelines based on TNM-7 without adjusting for the prognostic heterogeneity within stage I. Although many studies have shown excellent prognosis of stage I, these outcomes reflect current treatment paradigms, in which many such patients received intensified treatment with chemoradiotherapy. Treatment decision-making should not alter without robust clinical trial data [9]. A small subset of stage I fare less well, mainly due to DM [10,24,38]. Inclusion of the rENE parameter in N-classification can migrate these patients into a more advanced stage, thereby facilitating clinical trial design to ultimately modify treatment and improve clinical care. In fact, several deintensification trials (e.g. NRG HN002, NCT02254278; ORATOR2, NCT03210103) have already excluded such patients from enrollment.

S.H. Huang et al. / Radiotherapy and Oncology 144 (2020) 13–22

19

Fig. 2. Outcomes by rENE within TNM-8N1, N2, and N3 Subsets. Abbreviation: OS: overall survival; DFS: disease-free survival; LRC: locoregional control; DC: distant control; rENE+: presence of unequivocal radiologic extranodal extension; rENE : absence of unequivocal radiologic extranodal extension.

We propose that any rENE+ be classified at the highest N, namely N3, which is similar to the existing cN definition in HPVnegative HNC. Step-wise N reclassification (i.e. N1_rENE+ to N2 while N2_rENE+ to N3) is also possible since stage grouping based on these two schemas have similar excellent performance for estimation of mortality risk. However it could perpetuate the valid concern about under-treating some patients who would be considered N3 if Schema1 was used.

In conclusion, this study shows that unequivocal rENE+ portends a higher DM risk with worse OS and DFS in HPV+ OPC. Addition of cisplatin-based systemic agents improves LRC in rENE+ but unable to fully negate the impact of rENE+ on DM. The presence of rENE refines the TNM-8 cN-classification where N1_rENE+ could be reclassified as either New_N3 (Schema1) or New_N2 with N2_rENE+ as New_N3 (Schema2). Schema1 appears to be preferable and mirrors the current cN-classification algorithm for HPV-negative

20

rENE refining HPV+ OPC cN Classification

Table 3 Multivariable Analysis for OS and DFS with vs without Inclusion of rENE in the cN+ Patients. Overall Survival Variable TNM-8 N2 vs N1 N3 vs N1 T3-T4 vs T1-T2 Age Smoking PY Systemic Tx C-index for MVA without

Disease-free Survival HR (95% CI) 1.78 (1.15–2.75) 1.72 (0.83–3.59) 2.57 (1.62–4.08) 0.99 (0.97–1.02) 1.01 (1.00–1.02) 0.43 (0.27–0.67) rENE:

Variable HR (95% CI) TNM-8 N2 vs N1 1.65 (1.06–2.56) N3 vs N1 0.67 (0.31–1.48) rENE+ vs rENE 3.86 (2.49–5.97) T3-T4 vs T1-T2 2.55 (1.61–4.04) Age 0.99 (0.97–1.02) Smoking PY 1.01 (1.00–1.02) Systemic Tx 0.38 (0.24–0.60) C-index for MVA with rENE: Variable HR (95% CI) Schema1_N*: New_N2 vs New_N1 1.71 (1.0–2.93) New_N3 vs New_N1 3.99 (2.44–6.50) New_N3 vs New_N2 2.33 (1.43–3.78) T3-T4 vs T1-T2 2.70 (1.72–4.22) Age 1.00 (0.97–1.02) Smoking PY 1.01 (1.00–1.02) Systemic Tx 0.40 (0.25–0.62) C-index for MVA with rENE: Variable HR (95% CI) Schema2_N**: New_N2 vs New_N1 2.17 (1.33–3.53) New_N3 vs New_N1 4.04 (2.34–6.97) New_N3 vs New_N2 1.86 (1.14–3.03) T3-T4 vs T1-T2 2.48 (1.58–3.89) Age 1.00 (0.97–1.02) Smoking PY 1.01 (1.00–1.02) Systemic Tx 0.40 (0.25–0.62) C-index for MVA with rENE:

p-value

Global p

Variable

0.030

TNM-8 N2 vs N1 N3 vs N1 T3-T4 vs T1-T2 Age Smoking PY Systemic Tx C-index for MVA without

0.010 0.150 <0.001 0.560 0.009 <0.001 0.717 p-value

Global p 0.016

0.027 0.320 <0.001 <0.001 0.670 0.010 <0.001 0.749 p-value

Global p <0.001

0.049 <0.001 <0.001 <0.001 0.720 0.016 <0.001 0.748 p-value

Global p <0.001

0.002 <0.001 0.013 <0.001 0.630 0.015 <0.001 0.740

HR (95% CI)

p-value

Global p 0.005

1.71 (1.15–2.54) 2.26 (1.25–4.11) 2.53 (1.70–3.77) 1.00 (0.98–1.02) 1.01 (1.00–1.02) 0.54 (0.36–0.82) rENE:

Variable HR (95% CI) TNM-8 N2 vs N1 1.55 (1.03–2.31) N3 vs N1 0.88 (0.46–1.69) rENE+ vs rENE 3.89 (2.62–5.77) T3-T4 vs T1-T2 2.38 (1.59–3.57) Age 1.01 (0.98–1.02) Smoking PY 1.01 (1.00–1.01) Systemic Tx 0.51 (0.34–0.78) C-index for MVA with rENE:

0.008 0.007 <0.001 0.850 0.089 0.004 0.690 p-value

Global p 0.051

0.034 0.700 <0.001 <0.001 0.990 0.120 0.002 0.736

Variable HR (95% CI) Schema1_N*: New_N2 vs New_N1 1.62 (1.0–2.62) New_N3 vs New_N1 4.14 (2.70–6.36) New_N3 vs New_N2 2.56 (1.64–3.99) T3-T4 vs T1-T2 2.51 (1.69–3.72) Age 1.00 (0.98–1.02) Smoking PY 1.01 (1.00–1.01) Systemic Tx 0.53 (0.35–0.81) C-index for MVA with rENE:

p-value

Variable HR (95% CI) Schema2_N**: New_N2 vs New_N1 2.09 (1.36–3.23) New_N3 vs New_N1 4.27 (2.65–6.88) New_N3 vs New_N2 2.04 (1.32–3.15) T3-T4 vs T1-T2 2.29 (1.54–3.40) Age 1.00 (0.98–1.02) Smoking PY 1.01 (1.00–1.01) Systemic Tx 0.52 (0.34–0.79) C-index for MVA with rENE:

p-value

Global p <0.001

0.051 <0.001 <0.001 <0.001 >0.999 0.130 0.003 0.729 Global p <0.001

<0.001 <0.001 0.001 <0.001 0.870 0.130 0.002 0.721

Abbreviations: systemic tx: systemic treatment; PY; pack-years; rENE: radiologic extranodal extension. *Schema1 N: New_N1 = N1_rENE-–; New_N2 = N2_rENE ; New_N3 = N1-rENE+ and N2_rENE+ and any N3. **Schema2 N: New_N1 = N1_rENE-–; New_N2 = N2_rENE and N1_rENE+; New_N3 = N2-rENE+ and any N3.

HNC. The newly proposed N-classification schema(s) improved OS staging performance and could facilitate future clinical trial design and subsequent treatment decision-making, if validated in multicenter datasets. rENE can be reliably identified by expert HN radiologists if stringent criteria are used. Patients with rENE are unsuitable for de-intensification trials. Standardized radiology taxonomy and reporting templates are expected to facilitate staging accuracy. Additional training and communication should also facilitate identification and inclusion of more radiologic features in future staging. Whether the rENE identification demonstrated in a single academic institution can be reproduced elsewhere remains to be confirmed and the proposed new N-classification described herein needs to be validated. Funding None. Conflict of interest  Dr. Hope reported non-financial support from Elekta, Inc., outside the submitted work.

 Dr. Giuliani reports non-financial support from Elekta, Inc., grants from Eli Lilly’s Research Grant, non-financial support from AstraZeneca, outside the submitted work.  Dr. Hansen reports Advisory/Consulting/Research for Genentech/Roche, Merck, GSK, Bristol-Myers Squibb, Novartis, Boston Biomedical, Boehringer-Ingelheim, AstraZeneca, Medimmune, outside the submitted work.  No actual or potential conflicts of interest exist for any other authors on this topic.

Author’s contributions statement  Conceptualization: Shao Hui Huang, Brian O’Sullivan, Eugene Yu, Jie Su, Wei Xu, Eric Bartlett.  Data curation: all.  Formal analysis: Shao Hui Huang Brian O’Sullivan, Eugene Yu, Jie Su, Wei Xu.  Data Interpretation: all.  Manuscript drafting: Shao Hui Huang, Brian O’Sullivan, Jie Su, Wei Xu, Eugene Yu.  Manuscript editing: all

S.H. Huang et al. / Radiotherapy and Oncology 144 (2020) 13–22

21

Fig. 3. OS and DFS by TNM-8, Schema1 and Schema2 N classifications and Stage Groups.

Acknowledgement We acknowledge the Bartley-Smith/Wharton, the Gordon Tozer, the Wharton Head and Neck Translational, Dr. Mariano Elia, and Petersen-Turofsky Funds, and ‘‘The Joe & Cara Finley Center for Head & Neck Cancer Research”, and the ‘‘Discovery Fund” at the Princess Margaret Cancer Foundation for supporting the authors’ (SHH, JNW, JS, WX, LT, BOS) academic activities. We also acknowl-

edge the O. Harold Warwick Prize of the Canadian Cancer Society for supporting the author’s (BOS) academic activities.

Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.radonc.2019.10.011.

22

rENE refining HPV+ OPC cN Classification

References [1] Huang SH, Xu W, Waldron J, et al. Refining American joint committee on cancer/union for international cancer control TNM stage and prognostic groups for human papillomavirus-related oropharyngeal carcinomas. J Clin Oncol 2015;33:836–45. [2] O’Sullivan B, Huang SH, Su J, 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. [3] O’Sullivan B. Head and neck tumours. In: Brierley J, Gospodarowicz M, Ch W, editors. UICC TNM classification of malignant tumours. Chichester, West Sussex, UK: John Wiley & Sons, Ltd; 2017. p. 17–54. [4] O’Sullivan B, Lydiatt W, Haughey BH, et al. HPV-Mediated (p16+) oropharyngeal cancer. In: Amin M, Edge S, Greene F, editors. AJCC cancer staging manual. Springer Nature: Switzerland; 2017. p. 113–21. [5] Beltz A, Gosswein D, Zimmer S, et al. Staging of oropharyngeal squamous cell carcinoma of the head and neck: Prognostic features and power of the 8th edition of the UICC staging manual. Eur J Surg Oncol 2019. [6] Yamashita Y, Ikegami T, Hirakawa H, et al. Staging and prognosis of oropharyngeal carcinoma according to the 8th Edition of the American Joint Committee on Cancer Staging Manual in human papillomavirus infection. Eur Arch Otorhinolaryngol 2019;276:827–36. [7] Sharma SJ, Wagner S, Reder HSF, et al. The 8th edition AJCC, UICC TNM staging for p16-positive oropharyngeal carcinoma: is there space for improvement?. Eur Arch Otorhinolaryngol 2018;275:3087–91. [8] Fakhry C, Zevallos JP, Eisele DW. Imbalance between clinical and pathologic staging in the updated American joint commission on cancer staging system for human papillomavirus-positive oropharyngeal cancer. J Clin Oncol 2018;36:217–9. [9] Adelstein DJ, Ismaila N, Ku JA, et al. Role of treatment deintensification in the management of p16+ Oropharyngeal cancer: ASCO provisional clinical opinion. J Clin Oncol 2019;37:1578–89. [10] Billfalk-Kelly A, Yu E, Su J, et al. Radiologic extranodal extension portends worse outcome in cN+ TNM8 stage I human papillomavirus-mediated oropharyngeal cancer. Int J Radiat Oncol Biol Phys 2019. [11] Mancuso AA, Harnsberger HR, Muraki AS, et al. 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;148:715–23. [12] Carvalho P, Baldwin D, Carter R, et al. Accuracy of CT in detecting squamous carcinoma metastases in cervical lymph nodes. Clin Radiol 1991;44:79–81. [13] Souter MA, Allison RS, Clarkson JH, et al. Sensitivity and specificity of computed tomography for detection of extranodal spread from metastatic head and neck squamous cell carcinoma. J Laryngol Otol 2009;123:778–82. [14] King AD, Tse GM, Yuen EH, et al. Comparison of CT and MR imaging for the detection of extranodal neoplastic spread in metastatic neck nodes. Eur J Radiol 2004;52:264–70. [15] Url C, Schartinger VH, Riechelmann H, et al. Radiological detection of extracapsular spread in head and neck squamous cell carcinoma (HNSCC) cervical metastases. Eur J Radiol 2013;82:1783–7. [16] Chai RL, Rath TJ, Johnson JT, et al. Accuracy of computed tomography in the prediction of extracapsular spread of lymph node metastases in squamous cell carcinoma of the head and neck. JAMA Otolaryngol Head Neck Surg 2013;139:1187–94. [17] Kann BH, Buckstein M, Carpenter TJ, et al. Radiographic extracapsular extension and treatment outcomes in locally advanced oropharyngeal carcinoma. Head Neck 2014;36:1689–94. [18] Maxwell JH, Rath TJ, Byrd JK, et al. Accuracy of computed tomography to predict extracapsular spread in p16-positive squamous cell carcinoma. Laryngoscope 2015;125:1613–8. [19] Prabhu RS, Magliocca KR, Hanasoge S, et al. Accuracy of computed tomography for predicting pathologic nodal extracapsular extension in patients with head-

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27] [28] [29]

[30]

[31]

[32]

[33]

[34]

[35]

[36] [37]

[38]

[39]

and-neck cancer undergoing initial surgical resection. Int J Radiat Oncol Biol Phys 2014;88:122–9. Almulla A, Noel CW, Lu L, et al. Radiologic-pathologic correlation of extranodal extension in patients with squamous cell carcinoma of the oral cavity: implications for future editions of the TNM classification. Int J Radiat Oncol Biol Phys 2018;102:698–708. Geltzeiler M, Clayburgh D, Gleysteen J, et al. Predictors of extracapsular extension in HPV-associated oropharyngeal cancer treated surgically. Oral Oncol 2017;65:89–93. McMullen CP, Garneau J, Weimar E, et al. Occult nodal disease and occult extranodal extension in patients with oropharyngeal squamous cell carcinoma undergoing primary transoral robotic surgery with neck dissection. JAMA Otolaryngol Head Neck Surg 2019. Noor A, Mintz J, Patel S, et al. Predictive value of computed tomography in identifying extracapsular spread of cervical lymph node metastases in p16 positive oropharyngeal squamous cell carcinoma. J Med Imaging Radiat Oncol 2019. Spector ME, Chinn SB, Bellile E, et al. Matted nodes as a predictor of distant metastasis in advanced-stage III/IV oropharyngeal squamous cell carcinoma. Head Neck 2016;38:184–90. Spector ME, Gallagher KK, Light E, et al. Matted nodes: poor prognostic marker in oropharyngeal squamous cell carcinoma independent of HPV and EGFR status. Head Neck 2012;34:1727–33. Vainshtein JM, Spector ME, Ibrahim M, et al. Matted nodes: High distantmetastasis risk and a potential indication for intensification of systemic therapy in human papillomavirus-related oropharyngeal cancer. Head Neck 2016;38:E805–14. Wong K, Huang SH, O’Sullivan B, et al. Point-of-care outcome assessment in the cancer clinic: audit of data quality. Radiother Oncol 2010;95:339–43. ‘‘Head and Neck Site Group”: Head and Neck - Oropharynx, (ed Sept, 2015). Princess Margaret Cancer Centre Clinical Practice Guidelines:, 2015 Huang SH, O’Sullivan B, Xu W, et al. Temporal nodal regression and regional control after primary radiation therapy for N2–N3 head-and-neck cancer stratified by HPV status. Int J Radiat Oncol Biol Phys 2013;87:1078–85. Harrell Jr FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996;15:361–87. Groome PA, Schulze KM, Mackillop WJ, et al. A comparison of published head and neck stage groupings in carcinomas of the tonsillar region. Cancer 2001;92:1484–94. Xu W, Shen XW, Su J, et al. Refining evaluation methodology on TNM stage system: assessment on HPV-related oropharyngeal cancer. Austin Biom Biostat 2015;2. An Y, Park HS, Kelly JR, et al. The prognostic value of extranodal extension in human papillomavirus-associated oropharyngeal squamous cell carcinoma. Cancer 2017;123:2762–72. Hobelmann KC, Topf MC, Bar-Ad V, et al. AJCC-8ed nodal staging does not predict outcomes in surgically managed HPV-associated oropharyngeal cancer. Oral Oncol 2018;82:138–43. Zhan KY, Eskander A, Kang SY, et al. Appraisal of the AJCC 8th edition pathologic staging modifications for HPV-positive oropharyngeal cancer, a study of the National Cancer Data Base. Oral Oncol 2017;73:152–9. Haughey BH, Sinha P, Kallogjeri D, et al. Pathology-based staging for HPVpositive squamous carcinoma of the oropharynx. Oral Oncol 2016;62:11–9. Tian S, Ferris MJ, Switchenko JM, et al. Prognostic value of radiographically defined extranodal extension in human papillomavirus-associated locally advanced oropharyngeal carcinoma. Head Neck 2019. Bhattasali O, Thompson LDR, Schumacher AJ, et al. Radiographic nodal prognostic factors in stage I HPV-related oropharyngeal squamous cell carcinoma. Head Neck 2018. Kann BH, Aneja S, Loganadane GV, et al. Pretreatment Identification of Head and Neck Cancer Nodal Metastasis and Extranodal Extension Using Deep Learning Neural Networks. Sci Rep 2018;8:14036.