Abstracts #338 and 339
Jordan Berlin, MD
Ingram Professor of Cancer Research
Our goal is to improve the outcomes for those we can help while minimizing toxic
exposure for those for whom our treatments provide no
benefit
What do we know about stage II colon cancer?• It is a heterogeneous stage with an overall good
prognosis• Fluoropyrimidine adjuvant therapy provides benefit
to ~3-4% of unselected patients– FOLFOX does not appear to improve outcomes for low
risk patients– At time of last publication, FOLFOX did not statistically
improve survival of high risk stage II patients• The HR was 0.72, NS and more people were alive with relapse in
the 5FU arm at time of publication
• ASCO guidelines state that discussion of adjuvant therapy in stage II should be done
Clinicopathologic markers for high risk• Consider treatment if presence of 1 or more of the following
based on ASCO guidelines4
– T4 stage– Poorly differentiated histology– Bowel obstruction or perforation– <12 Lymph nodes resected– Lympho-vascular invasion– Perineural invasion– Close margins– Elevated preoperative CEA
• Also, we have Mismatch Repair (MMR) or MSI– This is prognostic for improved when MMR is deficient– This also appears predictive of 5FU effect in stage II
4. Benson AB et al. J Clin Oncol 2004.
Strategies to improve risk/benefit in stage II• Further refine the population at risk
– Theoretically these patients have the best chance for benefit
– Eg T4 and or at least 2 high risk features– Other prognostic markers
• Define the populations who will benefit from chemotherapy– Predictive markers
Oncotype DX: Primary Analysis: Recurrence Score Predicts Recurrence Risk in Stage II & III Colon Cancer Patients in NSABP C-07 (n=892)
Solid: 5FU Dashed: 5FU+Ox
Stage III C
Stage III A/B
Stage II
• With similar relative benefit of oxaliplatin added to adjuvant 5FU across the range of RS, absolute benefit of oxaliplatin increases with increasing RS, most apparently in stage II and stage IIIA/B patients
p<0.001
Solid: 5FUDashed: 5FU+Ox
O’Connell, ASCO 2012
Oncotype DX: 5-year Recurrence Risk in 5FU treated armCox Regression Analysis (n=892)
Stage II Stage IIIA/B Stage IIIC
RS Group* % pts Average Risk95% CI % pts Average Risk
95% CI % pts Average Risk95% CI
Low 39% 9% (6-13%) 41% 21% (16-26%) 33% 40% (32-48%)
Intermediate 36% 13% (8-17%) 34% 29% (24-34%) 37% 51% (43-59%)
High 25% 18% (12-25%) 25% 38% (30-46%) 30% 64% (55-74%)
* Pre-specified RS Groups: Low (RS<30), Intermediate (30≤RS<41), High (RS≥41).
• Recurrence risk is significantly higher in High vs. Low RS group: HR = 2.11, p<0.001
O’Connell, ASCO 2012
Contribution of RS Beyond Clinical and Pathologic CovariatesPre-specified Multivariate Analysis (n=892)
Variable Value HR HR 95% CI P value
Stage <0.001
(by nodal status) Stage III A/B vs. II 0.97 (0.55,1.71)
Stage III C vs. II 2.07 (1.16,3.68)
Treatment FU+Oxali vs. FU 0.82 (0.64,1.06) 0.12
MMR MMR-D vs. MMR-P 0.27 (0.12,0.62) <0.001
T-stage T4 st II & T3-T4 st III vs. T3 st II & T1-T2 st III
3.04 (1.84,5.02) <0.001
Nodes examined <12 vs. ≥12 1.51 (1.17,1.95) 0.002
Tumor grade High vs. Low 1.36 (1.02,1.82) 0.041
RS per 25 units 1.57 (1.19,2.08) 0.001
• RS is significantly associated with risk of recurrence after controlling for effects of T and N stage, MMR status, number of nodes examined, grade and treatment
Contribution of RS Beyond Clinical and Pathologic CovariatesPre-specified Multivariate Analysis (n=892)
Variable Value HR HR 95% CI P value
Stage <0.001
(by nodal status) Stage III A/B vs. II 0.97 (0.55,1.71)
Stage III C vs. II 2.07 (1.16,3.68)
Treatment FU+Oxali vs. FU 0.82 (0.64,1.06) 0.12
MMR MMR-D vs. MMR-P 0.27 (0.12,0.62) <0.001
T-stage T4 st II & T3-T4 st III vs. T3 st II & T1-T2 st III
3.04 (1.84,5.02) <0.001
Nodes examined <12 vs. ≥12 1.51 (1.17,1.95) 0.002
Tumor grade High vs. Low 1.36 (1.02,1.82) 0.041
RS per 25 units 1.57 (1.19,2.08) 0.001
• RS is significantly associated with risk of recurrence after controlling for effects of T and N stage, MMR status, number of nodes examined, grade and treatment
Kaplan-Meier plot of time to cancer-related death in the independent validation set.
Kennedy R D et al. JCO 2011;29:4620-4626©2011 by American Society of Clinical Oncology
Gene signature analysis
This prognosis was independent of known clinicopathologic factors
Data from QUASAR study using 13 cancer-related genes :Kaplan-Meier estimates of 3-year recurrence in surgery-alone patients by risk group.
Gray R G et al. JCO 2011;29:4611-4619
©2011 by American Society of Clinical Oncology
QUASAR prognosis stratification
Gray R G et al. JCO 2011;29:4611-4619
©2011 by American Society of Clinical Oncology
Risk groups did not predict for chemotherapy benefit
Use of Adjuvant Chemotherapy & Outcomes in Stage II Colon Cancer with vs. without Poor
Prognostic Features
Aalok Kumar, Hagen Kennecke, Howard Lim, Daniel Renouf, Ryan Woods, Caroline Speers, and Winson Cheung
Department of Medical OncologyBritish Columbia Cancer Agency
What was the study?• An exploratory analysis of prospectively
collected data in the British Columbia Cancer Agency Gastro-Intestinal Cancers Outcomes Unit (GICOU)– 10 year period of analysis– 1,697 patients divided into high vs low risk based
on ASCO guidelines• Note: these guidelines are well-considered and
literature based. Although 9 years old, they are still relevant
• 73% high risk and 27% low risk
Some baseline lessons
– More patients in the high risk category (29%) received adjuvant therapy than in the low risk category (13%)
– Patients ≥ 70 years of age were less likely to receive adjuvant chemotherapy
– Perforation and T stage appeared to play the largest roles in selecting patients for adjuvant chemotherapy
• CEA also played a role
Some issues to note• The performance status
– This may have been measured early, but PS = 3 were treated (surprising) and comprised a huge population
• Perforation/obstruction– Seemed like a high proportion of patients
presented with these findings
• Adjuvant therapy– We don’t actually know what the choice of
adjuvant therapy was for the patients
Putting this into perspective• Low risk group
– Nothing about adjuvant chemotherapy looked good– Univariate
• 3 year relapse survival, 5 year disease specific survival and 5 year overall survival were similar with or without adjuvant chemotherapy
– Multivariate• Just made chemo look worse in this setting.• 3 year RFS, 5 year DSS were worse with adjuvant chemo
• Note: The selection of low risk patient for chemo may have included some key factors not assessed that put them at higher risk for recurrence or death
High Risk Group• Very confusing• Univariate outcomes
– 3 year RFS and 5 year DSS were identical with or without adjuvant chemotherapy
– 5 year OS was improved with adjuvant chemo
Does adjuvant chemotherapy heal what is ailing you?
• The univariate analysis seems to suggest that while chemotherapy does not impact death from colorectal cancers, it does affect overall survival. – This is counterintuitive suggesting that
chemotherapy reduced deaths from other causes
– The multivariate outcomes analysis sheds more light
High Risk Outcomes
• Multivariate outcomes analysis showed that the high risk group does appear to benefit from adjuvant chemotehrapy (HR = 0.67 for OS)– However benefits vary
• T4 seems to have more benefit than T3• Multiple high risk features predicts for more benefit
from adjuvant chemotherapy• Single high risk feature T3 patients derive uncertain
benefit
What is the bottom line?
• This seems to confirm that stage II patients overall have a good prognosis, but– There are subsets with better and worse
prognoses– Clinicopathologic parameters can separate these
groups to an extent– But we don’t truly know who benefits from chemo
• That group is small– And to benefit these patients we need to expose a large
number of these patients to the risks
Is there a best way to use the data?• We could use prognosis to segregate
– Low risk individuals based on clinicopathologic criteria• Don’t appear to benefit from chemotherapy and may even be
harmed based on this data
– Very high risk (T4, multiple high risk features)• Appear to benefit from chemotherapy and maybe have the best
chance to derive benefit
– Intermediate risk (T3, one high risk feature)• These are still the most complex, and maybe they would derive
the most benefit from one of these recurrence risk panels
Summary• However, we still risk exposing many patients who
have no chance of benefit to adjuvant chemotherapy– While there are many gene signatures out there, these
need to be validated prospectively
• Ideally we need predictive markers—these markers would ideally completely separate those who benefit from adjvant chemotherapy from those who don’t– Currently, our only predictive marker appears to be MMR– ECOG 5202 may shed more light on prognostic groups
Summary II
• The abstract presented did not assess mismatch repair (MMR) phenotype (ie microsatellite instability/stability)– While mismatch repair biology and its impact on
treatment and prognosis is still emerging– Mismatch repair deficiency correlates with a
better prognosis in stage II colon cancer• The pathology often reads as poorly differentiated
– In stage II colon cancer MMR deficiency predicts for lack of benefit from fluoropyrimidines
A Molecular Profile of Colorectal Cancer to Guide Prognosis and Therapy after
Resection of Primary or Metastatic Disease
Joshua M. Uronis, Ph.D.Hsu Laboratory
Duke Institute for Genome Sciences and Policy
Duke Cancer InstituteDuke University
What did they do?
• They took microarray data from 850 primary CRC samples
• Looked at patterns of pathway activation/deregulation for 19 oncogenic pathways– 6 subrgroups came out of this
• Then evaluated 133 metastatic CRC samples– 6 subgroups again identified
What they learned
• Their 6 subgroups had different recurrence free survival– The same group did the worst in both resected
primary and resected liver metastases– These subgroups could not be differentiated by
any specific oncogene mutations– But these subgroups had potential to have
differential sensitivity to targeted agents.
Next step: prove the differential sensitivity to targeted agents--predictive
• They have now developed 50 explant models from patients into mice– These appear to maintain their characteristic
features over multiple generations
• They can place these explants into the 6 subgroups– Example given was from subgroups expected to
be resistant and expected to be sensitive to mTOR inhibitors based on their pathway activation analysis
Rad001
Control
Predicted Sensitive
Predicted Resistant
Control
Rad001
Predicting Drug Responses
What have we learned?
• This is early and needs further testing• However, they can potentially use the mouse
model to identify targets to “hit” and agents that are effective for each subgroup– This may allow us to find more active agents for
each subgroup– The authors hope to use this to find adjuvant
therapies
Are there advantages over other model systems?• This model would not require growing every
patient’s tumor in vivo,– It identifies putative drugs for subgroups– Analysis of a patient’s tumor specimen could
lump them into one of the 6 subgroups – And if clinical trials prove what their preclinical
models show,• Each subgroup will be treated differently, but with
potentially more active agents
Conclusions• Both abstracts add to our knowledge in colorectal
cancer– Kumar, et al showed that our clinicopathologic prognosis
models in stage II need more modification• However, there are groups that can potentially be designated to
treatment or no treatment based on these factors alone
– Uronis, et al have developed a model that may help us to subdivide CRC patients and treat them (personalized) based on their subgroup
• However, this needs further preclinical testing and clinical trial validation
• This provides an alternative to genomic analysis which may identify mutations, but we don’t know if this means pathway dependence/activation/dysregulation