Developing, Validating, or Optimizing the Staging
Classifications of OPC and NPC:
the viral induced pharyngeal cancers
Brian O’Sullivan Professor, Department of Radiation Oncology
The Princess Margaret Cancer Centre / University of Toronto
Radiation Oncology Chair United States NCI Head and Neck Steering Committee
Member, TNM Committee, UICC Chair, Prognostic Factors Committee, UICC
3:30-4:10 PM, Sunday, Sept 9, 2018
Hong Kong
Disclosure
None
Overview
Roles of different Professionals and the Different
Purposes involving Cancer Stage Classification (TNM)
How to create/validate a new TNM, or how to improve
an existing TNM classification
Use exampler of the two viral pharyngeal cancers
(NPC and HPV-related)
Challenges for the future and the incorporation of
factors beyond anatomic disease extent
Roles and Needs in TNM Classification
Numerous stakeholders:
IARC, WHO “blue books”
US NCI, CDC etc
Each group needs to be
aware of the needs and
purposes of the others
EBV:
• Responsible for majority of NPC in southern China and east Asia
HPV:
• Responsible for most OPC in western countries
EBV+ NPC and HPV+ OPC share similarity in:
• Multiple overlapping oncogenesis pathways
• Highly radio- and chemo-sensitive
• More immunogenic
• Circulating blood DNA with opportunities
to refine management and enhance
precision medicine
Viral Related Pharyngeal Cancer
Mesri 2014
EBV HPV
OS by 7th edition TNM Stage Groups [PMH Data 2000-2010]
Huang,
O’Sullivan, et
al. JCO,
March, 2015
HPV– OPC
(n=237) Inseparable OS between
stage I-IV (P=0.56)
Acceptable OS
performance (p=0.004)
HPV+ OPC
(n=573)
HPV(+) Stage IV disease does not have the ominous
outcome of smoking-related OPC
80.4%
Why Do We Need a New TNM for HPV+ OPC
HPV+ OPC is a new disease
7th Edition TNM was unable to adequately depict prognosis
Why HPV+ OPC Needs Its Own TNM
Necessary
• Consultation: relevant to discussion with patient / family
• Clinical trial eligibility and design (e.g. stratification)
• Framework for future clinical practice guidelines
Applicable
• For cancer surveillance
High HPV prevalence jurisdictions need an HPV+ TNM
Low HPV prevalence regions could theoretically continue to use HPV─ TNM
Brizel JCO 2015 Gillison JCO 2016
Developing and Validating A New Staging for
HPV+ OPC (cTNM and pTNM)
7 Centers 1907 HPV+ Patients • Training (PMH): n=661 • Validation: n=1246
RT: 98%; Surgery: 2%
O’Sullivan, Huang, Yu, Xu et al. Lancet Oncol 2016
cTNM
5 Centers 704 Patients • TORS/TLM: 100% • pT1-2: 81%; pN0-2b: 87%
Haughey, et al Oral Oncol 62:11-19, 2016
pTNM
5
Derivation cTNM for HPV+ OPC
PMH 2000-2010
• HPV+ OPC: 573
• HPV(-) OPC: 237
All definitive RT
Huang, Xu, O’Sullivan, et al JCO. 2015
Discovery PMH Study
- Feasibility - Statistical models - Evaluation criteria
Derivation ICON-S Study
- Training-validation - Derivation of a cTNM
7 Centers 1907 HPV+ Patients
• Training (PMH): n=661
• Validation: n=1246
RT: 98%; Surgery: 2%
O’Sullivan, Huang, Xu et al Lancet Oncol 2016
External Validation
Independent Datasets
1.Porceddu et al. Oral Oncol 2017 2.Husain et al Cancer 2017 3.Malm, Fakhry et al Cancer 2017 4.Wurdemann et al Front Oncol 2017 5.Cramer et al Head Neck 2018
1.Brisbane: n=279 2.NCDB Data: n=5626 3.Johns Hopkins: n=435 4.Giessen (Germany): n=150 5.NCDB Data: n=15116
HPV+ OS by 7th T- & N-category [multi-institutional dataset] (n=1907)
T: T1, T2, T3, and T4 separate well; T4a and T4b behave similarly
N: inferior OS in N2c and N3; Similar OS between N0 / N1-N2a and N2b
T Events/Total 5-year OS P value N Events/Total 5-year OS P value
Total 395 / 1907 80% (78-82) <0.001 Total 395 / 1907 80% (78-82) <0.001
T1 60 / 504 89% (87-92) N0 32 / 173 80% (73-87)
T2 116 / 716 83% (80-87) N1-N2a 59 / 416 87% (83-90)
T3 106 / 412 76% (72-81) N2b 136 / 749 83% (80-86)
T4a 93 / 231 58% (51-65) N2c 112 / 436 74% (70-79)
T4b 20 / 44 57% (44-75) N3 56 / 133 59% (51-69)
O’Sullivan,
Huang et al
Lancet Oncol,
2016
N3
N2c
T4b
T4a
Unilateral Neck
LNs
Bi-/contra-lat Neck LNs
Re-termed
ICON-S N
AHR: adjusted HR (adjusting for age, smoking, & chemo)
AHR Grid
Training Cohort (n=661)
5-year OS: 85%
5-year OS: 78%
5-year OS: 53%
Validation Cohort (n=1246)
5-year OS: 88%
5-year OS: 81%
5-year OS: 65%
AHR Grid
External Validation of ICON-S Staging Study Sample
Size
Primary
Tx
No. at risk and 5-yr OS by ICON-S Stage
I II III IV (M1)
ICON-S
(O’Sullivan)
N=1907
Sx: 2%
RT: 98%
85%
(n=962)
78%
(n=564)
53%
(n=381)
NA
Australia
(Porceddu)
N=279
Sx: --
RT: 100%
94%
(n=132)
82%
(n=82)
69%
(n=65)
NA
US JHH
(Malm)
N=435
Sx: 38%
RT: 62%
92%
(n=281)
87%
(n=77)
74%
(n=72)
40%
(n=5)
US NCDB
(Husain)
N=5626
Sx: 42%
RT: 56%
90% (3-yr)
(n=3631)
82%
(n=1242)
72%
(n=753)
NA
US NCDB
(Cramer)
N=15116 Sx: 44%
RT: 54%
90% (3-yr)
(n=8895)
81%
(n=3012)
68%
(n=1847)
31%
(n=320)
Germany
(Wurdemann)
N=150
Sx: 69%
RT: 31%
94%
(n=79)
77%
(n=31)
64%
(n=31)
25%
(n=9)
US 5 centers
(Haughey)
N=704 Sx: 100%
RT: --
90% 79% 70% NA
Comments About HPV+ cTNM
The 8th edition HPV+ cTNM reflects prognosis under current
treatment paradigms • But many had received intensified treatment
Stage I disease fares very well • Unknown if all stage I is suitable for less intensified treatment
• Uncertainty also about the optimal deintensification strategies
Stage I can be considered candidates for de-intensification
trials • Pre-mature to change treatment without trial data
Derivation of pTNM for HPV+ OPC Especially driven by the introduction of robotic surgery
Discovery
- LN No. (≤4 vs ≥5) was prognostic
Derivation
A multicenter study (incorporating LN # pragmatically)
External Validation
Pathologic (post-surgical) pTNM for HPV+ OPC
Washington University • HPV+ OPC: n=220 • TLM +/- adjuvant RT
Sinha et al Oral Oncol 51: 154-520, 2015
5 Centers (USA & UK) 704 Patients • TORS/TLM: 100% • pT1-2: 81%; pN0-2b: 87%
Haughey et al Oral Oncol 62:11-19, 2016
Zhan, et al, Oral Oncol 73:152-159, 2017
1. Zhan: NCDB (2010-2014) 2. Cramer: NCDB (2010-2013)
• For refinement of post-surgical reporting and framework to facilitate post-surgical management
• Cases first need to be classified by cTNM
Derivation of HPV+ pTNM
OS (UVA) N (%) HR p
pT1 (7th)
pT2
pT3
pT4
279 (40)
290 (41)
92 (13)
43 (6)
1.00
1.8 (1.08-3.05)
2.7 (1.38-5.16)
5.7 (2.67-11.31)
0.039
0.004
<0.001
pN0 (7th)
pN1
pN2a
pN2b
pN2c
pN3
44 (6)
91 (13)
141 (20)
337 (48)
52 (7)
39 (6)
1.00
0.9 (0.23-3.47)
1.2 (0.35-4.40)
1.9 (0.59-6.08)
3.6 (1.01-12.65)
2.7 (0.64-11.31)
0.870
0.737
0.285
0.049
0.174
pENE(-)
pENE(+)
233 (35)
427 (65)
1.00
1.61(0.98-2.63)
0.060
≤4 pN+
≥5 pN+
589 (84)
115 (16)
1.00
2.93 (0.21-0.53)
<0.001
Definition of pN
pN0 No regional LNs
pN1 ≤4 pN+ LNs
pN2 ≥5 pN+ LNs
Haughey el al Oral
Oncology 2016
MVA
MVA: adjusting for
age, treating center,
and treatment
8th Edition: cTNM vs pTNM Methodological Comparison
cTNM (O’Sullivan, TLO 2016) pTNM (Haughey, OO, 2017)
Study Cohort All M0 disease Resectable disease only
Study Period 7 centers, treated in 1998-2011 5 centers, treated in 1985-2015
HPV/p16 testing rate 50-81% unknown
Sample Size N=1907 N=704
7th T and N Category cT and cN
• cT3: 412 (22%); cT4: 275 (14%)
• cN2c: 436 (23%); cN3: 7%
pT and pN
• pT3: 92 (13%); pT4: 43 (6%)
• pN2c: 52 (7%); N3: 39 (6%)
Methodology • Training-validation
• Using models and evaluation
criteria from the “Discovery Study”
• Pooled dataset
• Adoption of “LN number” parameter
directly from the “Discovery Study”
Time point Assigned at the initial diagnosis After resection of both primary and neck
Application For multiple purposes For risk stratification
7th Edition cTNM 8th Edition cTNM
7th Edition pTNM 8th Edition pTNM
(M1)
(M1)
(M1)
(M1)
Uniqueness of HPV+ OPC TNM
(8th edition) - the 1st time in HNC:
• T4 or N3 M0 is no longer
Stage IV
• Different cN and pN
classification, and stage
groupings
Treatment guidelines should not
change in the absence of trials
--- TNM is not a Guideline
Optimizing the Stage Classification for
Nasopharyngeal Cancer:
Refined T- and N- Classification
19
Rationale for Refining NPC Stage Classification
NPC management has evolved substantially
• More accurate imaging:
Allows better delineation of tumor extent and early detection of occult
metastases
• Advances in radiotherapy:
Increases conformity of tumor coverage and sparing of non-involved
structures
• Combination chemotherapy:
Improved tumor control and cure rates especially for advanced loco-regional
disease
Based on data from 1609 NPC patients from Fujian and HK
Recommended changes as follows:
– Changing MP/LP involvement from T4 to T2
– Adding Prevertebral muscle involvement as T2
– Replacing SCF with lower neck and merging this with a maximum nodal
diameter > 6 cm as N3
– Merging T4 and N3 as stage IVA (formerly IVA and IVB)
Cancer 2015
Changes of T and N category (7th vs 8th Edition)
T2
T4
T-Category:
Extent of soft tissue involvement
• Adjacent - T2: including
Medial & lateral pterygoid muscle
Pre-vertebral muscle
• Extensive - T4
Beyond the lateral surface of the
lateral pterygoid muscle, parotid
Eliminated inconsistent terms:
• “masticator space”
• “infratemporal fossa”
N-Category:
Replacing s/c fossa (blue) with
lower neck LN
Merge N3a and N3b as N3
s/c
T2
Postero-lateral
infiltration
Incl. medial and
lateral pterygoid
invasion
Cancer 2015
lower neck LN: below
caudal border of cricoid
cartilage (Pink)
T4
7th T 8th T
OS for Adjacent vs Extensive Soft Tissue Involvement
Adjacent soft tissue involvement
T2
Extensive soft tissue involvement
T4
Validation of 8th Edition NPC Staging
Guangxi, China
7th edition T 8th edition T LRC
7th edition Stage 8th edition Stage
OS
DC
OS
7th edition N 8th edition N
OPC T1 T2 T3 T4
N0 I I II III
N1 I I II III
N2 II II II III
N3 III III III III
NPC T1 T2 T3 T4
N0 I II III IVA
N1 II II III IVA
N2 III III III IVA
N3 IVA IVA IVA IVA
8th Edition: Similar cTNM Grid for NPC vs HPV+ OPC
IVB: T any _ N any _ M1 IV: T any _ N any _ M1
8th Edition Stage Classification for HPV+ OPC and EBV+ NPC:
• Similar N classification (except lower neck LN parameter)
• Similar stage grid, except “arbitrary” numerical stage designation (i.e. I and II)
Viral Related NPC and OPC: RFS by Stage RFS: NPC I-II vs OPC I
OS after DM RFS: NPC IVA vs OPC III
RFS: NPC III vs OPC II
Huang,
O’Sullivan,
et al ESTRO
2017
Practicalities about TNM Cancer Staging
Anatomic TNM Classification
Advantages:
– Relatively versatile globally (world wide), including resource-limited regions
– Applicable to individual (used in clinic) and in groups of patients
(administrative and research)
Limitations:
– Variable assessment methods to determine anatomic extent
– Does not fully predict response to chemotherapy and targeted therapy
– Limited response to emerging information in a timely fashion
– Inability to describe tumor biology and other non-anatomic factors
T4bN2c HPV(+) SCC Oropharynx
Case 1:anatomic Stage predominant Case 2: tumour Profile predominant
T2N2c HPV(+) SCC Oropharynx
2 HPV + OPC cases with identical treatments: 70 Gy in 35 f with concurrent cisplatin:
“different” distant metastasis outcomes
HPV+ metastasis resected -disease free 7 years later
(potentially cured with very advanced stage)
Rapid fatal DM at
2-years, with less
advanced stage
lung, brain, bone,
liver, spleen,
mediastinum,
pleura Cystic brain
metastasis
18
months
later
18 months
later
Prognostic Grouping and Incorporating
Biomarkers
T N M
Categories
T N M
Stage Groups *
Prognostic
Groups
Anatomic
T1-4, N0-N3 Anatomic
Groups I-IV
Combines “stage” with:
• Profile (tumor)
• Patient (host)
• Social determinants of
Health (management)
• TNM needs to incorporate non-anatomic factors in addition to “Staging”
• Combines anatomic (“Staging”) with non-anatomic factors
• Example: EBV DNA for NPC; smoking pack-year for HPV+ OPC
UICC definition:
“Staging” components
UICC definition:
The “Stage”
Profile candidates in Head and Neck in Past 10 yrs:
EGFR, VEGF, HPV, EBV, “smoking”, now PD1, PDL1 etc
Prognostic Groups including Smoking
Pack-years
HPV(+) OPC (n=661)
5-year OS: 75% (72-78)
Stage III (T4 or N3) (n=163) 5-year OS: 53% (46-62)
Stage I-II (T1-3N0-N2) (n=498) 5-year OS: 82% (79-86)
Smoking >20 PY (n=170) 5-year OS: 67% (60-75)
Smoking <=20 PY (n=328) 5-year OS: 90% (86-93)
Prognostic Model of HPV(+)
OPC by 8th Edition Stage
Group I (n=328) Stage I-II, ≤20 PY
5-year OS: 90% (86-93)
Group III (n=163)
Stage III (T4 or N3M0) 5-year OS: 53% (46-62)
Group II (n=170)
Stage I-II, >20 PY 5-year OS: 67% (60-75)
OS by 3 Prognostic Groups (unadjusted) OS by 3 Prognostic Groups (adjusted*)
Group IV: metastatic (M1) HPV(+) OPC)
Prognostic Grouping
Group I T1-3 N0-N2 M0 Smoking: ≤ 20 PY
Group II T1-3 N0-N2 M0 Smoking: > 20 PY
Group III T4 Any N M0 Any smoking PY
Any T N3 M0 Any smoking PY
Group IV Any T Any N M1 Any smoking PY
Huang, O’Sullivan et al JCO 2015
* Adjusted for chemo
Issues About Smoking PY on HPV(+) OPC
Influence of smoking pack-years on OS may not directly
reflect altered tumor biology
May reflect:
• Competing mortality risk from smoking (late toxicity, second
primary, COPD, social problems, etc)
• Treatment tolerance due to comorbidities
• Radiotherapy efficacy (hypoxia) in current smokers
• Modified biology with p53 mutations and “hybrid” tumours
Incorporation of Circulating DNA into “Staging”
Technical Requirement: • Subject to testing conditions
• Large variation between laboratories
• Requires stringent harmonization
Some patients have undetectable copies • What to do for those undetectable EBV NPC vs non-EBV NPC?
Uncertainty concerning optimal cutoff
Same principles apply to circulating EBV and HPV DNA
Plasma EBV DNA is prognostic in patients
with NPC
Challenges of incorporating plasma EBV DNA in future TNM Staging System
• Various copy # cutoffs (1500-4000) have been used
• Likely reflected inter-laboratory variability
– Importance of testing standardization and harmonization
• Not elevated in some patients
Wang et. al. Cancer 2013
Potential Prognostic Value of
Pre- vs Post-RT HPV DNA
Plasma HPV DNA
52 HPV+ tumor
35 Pre-tx Plasma
HPV DNA+
17 Pre-tx Plasma
HPV DNA-
5 Post-tx
HPV DNA+
30 Post-tx
HPV DNA-
Recurrence:
4/5
Recurrence:
2/30
1 Post-tx
HPV DNA+
16 Post-tx
HPV DNA-
Recurrence:
1/1
Recurrence:
1/16
Ahn et al JAMA OHNS 2014 140;846-54
Challenges:
• Small sample size, unproven prognostication
• Low sensitivity
• Not available for others
• Test harmonization
Total M0 NPC (n=518)
• 5-yr PFS: 81% (77-85)
• 5-yr OS 82% (77-86)
N0-N2 (n=388)
• 5-yr PFS: 87% (83-91)
• 5-yr OS 85% (81-90)
N3 (n=130)
• 5-yr PFS: 63% (54-74)
• 5-yr OS 70% (60-81)
T1-4N0-2 EBV<500 (n=211)
5-yr PFS: 94% (91-98)
5-yr OS 89% (84-94)
N-Category
T1-4N0-2 EBV≥500 (n=177)
5-yr PFS: 80% (73-87)
5-yr OS 83% (77-90)
T1-2N3 (n=54)
5-yr PFS: 64% (51-81)
5-yr OS 83% (71-96)
T3-4N3 (n=76)
5-yr PFS: 63% (51-78)
5-yr OS: 60% (47-77)
EBV DNA T-category
PFS OS
RPA 4 Group Prognostic Model of NPC by 8th Edition Stage Classification
Adjusted for age, sex,
smoking, ACE-27 scores,
LDH and treatment modality
The addition of pretreatment plasma EBV DNA into the 8th Edition NPC TNM stage classification
Lee V, Kwong D, Leung TW, Choi CW, O’Sullivan B, Lam KO, Lai V, Khong PL, Chan SK, Ng CY, Tong
CC, Ying PP, Chan WL, Wong LS, Leung D, Chan SY, So TH, Luk MY, Lee A. IJC, accepted 2018
Group I Group II Group III Group IVA
N0-N2
partitioned by
EBV for PFS
N3 disease
trumps
EBV (PFS
& OS)
Evolution of NPC and HPV Stage classifications
HPV+ OPC and EBV+ NPC have similar N classification and
grid in the 8th edition TNM
• Developing or refining T & N classification is evidence-based
reflecting current knowledge
Different challenges in incorporating non-anatomic factors for
prognostic grouping
• Smoking for HPV+ OPC: unknown underline mechanisms and may
be trumped at more advanced stage (T4 and N3)
• EBV DNA for NPC: testing harmonization and may be trumped at
the advanced anatomic stage (N3)
• Blood HPV testing is more nascent
Global Consultation No 1 on Cancer classification
Addressed the challenges to a common understanding of
cancer staging and the needs of the different constituents
Reaffirmed the purpose of staging classification,
recognized the scope of TNM, and provided guidance for
its appropriate use
Consultation among UICC, AJCC, NCI, CDC, FIGO, IACR (International Association of
Cancer Registries), IARC (International Agency for Research in Cancer), and the ICCR
(International Collaboration on Cancer Reporting)
O’Sullivan, et al. July 2017
Next phase (Consultation No. 2 - April 2018)
Strategies and systems to develop sound prognostic classifications:
• May include non-anatomic factors sensitive to the state of the science and
availability of new treatments
• Will not replace descriptions of anatomic disease extent
May not be in the form of a classification:
• Systems to capture and analyze multiple relevant variables comprehensively
Respect personalized approaches:
• Deploy the most robust technologies
artificial intelligence, machine learning, and neural networks
to predict outcomes in individual cancer patients by using a more
comprehensive set of information than is used currently
Continued Challenges
Develop an agreed framework to include: • Non-anatomic factors:
Biology / predictive markers, molecular and genomic profiling, and imaging
• New technologies and strategies Address “Big Data” approaches
“Project based” if simple classifications not suited
Maintain world-wide cancer staging purposes: • Cancer control
specifically surveillance at the population level
• Research activities clinical trials eligibility and stratification, translational research
• Clinical guidance clinical care and decision-making in addition to communicating prognosis
Ongoing initiative in 2018 from UICC, AJCC, NCI, CDC, NCI, CDC, AJCC, IARC, IACR,
WHO, ICCR, to:
Acknowledgements
UICC TNM Project Committee
• Mary Gospodarowiz, Jim Brierley, Malcolm Mason, Anne Lee
• Shao Hui Huang, UICC Volunteer and Research colleague
• Julie Torode and Staff at the UICC Office, Geneva
AJCC and its Disease Site Commitees
• Carolyn Compton, Mahul Amin, Jatin Shah
• AJCC Office (at American College of Surgeons) especially Laura Meyer and Donna Gress
Hong Kong and Chinese colleagues
IARC, CDC, US NCI, numerous others
Our patients throughout the world