Transcript RCB I

Breast Cancer Molecular Profiles
Predict Tumor Response of
Neoadjuvant Doxorubicin and
Paclitaxel, the I-SPY TRIAL
(CALGB 150007/150012, ACRIN 6657)
L. J. Esserman ,C. Perou, M. Cheang,
L. J. van 't Veer, J. Gray, E. Petricoin, K. Conway,
L. Carey, A. DeMichele, D. Berry, N. Hylton
I-SPY INVESTIGATORS
CALGB INTERSPORE ACRIN NCICB
Investigation of
Serial studies to
Predict
Your
Therapeutic
Response with
Imaging and
And
moLecular analysis
I SPY
WITH MY
LITTLE
EYE . . . . .
. . A BIOMARKER
BEGINING WITH
X
....
I-SPY 1 Clinical Trial Backbone
Layered Imaging/Molecular Biomarker Studies Onto
Standard Clinical Care
Anthracycline
Taxane
Surgery
& RT
Serial MRI Scans
Serial Core Biopsies
Tam if ER+
Trial Endpoints
• Early
(ASCO POSTER 529)
– MRI response after 1 cycle of chemotherapy
• Longest Diameter, Volume
• Intermediate
• pCR
Pathologic Complete Response
• RCB
Residual Cancer Burden
• % change in MR volume
• Late
• 3 year Recurrence Free Survival
• 3 year Overall Survival
Residual Cancer Burden
RCB = 1.4 x [fcell x (d1 d2)] 0.17 + [dmet x (1 - (1 -  ) LN ) / ] 0.17
PRIMARY TUMOR BURDEN
PRIMARY TUMOR BURDEN
Area (cm x cm)
Area (cm x cm)
% CANCER
CELLULARITY
% CANCER
CELLULARITY
+
AXILLARY NODAL BURDEN
Number of positive LNs
Diameter of largest metastasis (mm)
Symmans et al. J Clin Oncol. 2007 Oct 1;25(28):4414-22.
Residual Cancer Burden
• Integrates several pathologic features
– Lymph node status
– Extent of Tumor Bed
– Tumor size
– Tumor cellularity
• Output is continuous or 4 discrete categories
– RCB 0
– RCB I
– RCB II
– RCB III
pCR, no invasive tumor
scattered residual disease
moderate tumor burden
significant tumor burder
Symmans et al JCO 2007
I-SPY 1 Biomarker Platforms
Tissue: Core or Surgical
H&E,IHC,FISH
UNC, Penn
Expression Arrays
p53 GeneChip
UNC, UCSF, NKI
UNC
CGH
5
GMU
)
Serum
4
relative copy number (Log2)
Protein Arrays (RPMA)
3
1q
2
Id1 proteins
autoantibodies
phospho proteins
20q
1
0
-1
-2
1
1p
3
5
7
9
Genome location
11
13
15
17
19
17p 19p
21
X
UCSF
Total Accrual: 237
Institution Name
Accrual
University of Pennsylvania Medical Center
36
Georgetown University Hospital
4
University of North Carolina
36
Memorial Sloan Kettering Cancer Center
22
University of Washington
5
University of Alabama at Birmingham
Medical Center
51
University of Chicago
2
University of Texas Southwestern
14
University of California San Francisco
66
• 1042 frozen cores from 201 patients
• 1301 paraffin cores from 223 patients
• 948 serum samples from 158 patients.
Results
I-SPY: Poor Prognosis Tumors
70 significant prognosis genes
NKI 70 Gene Profile
“Good”
Signature
“Poor”
Signature
9%
91%
Mean Tumor Size= 6.0
Present as clinical mass
55% < Age 50
van´t Veer et al., Nature ,2002
Relationship of pCR and RCB with
Early Relapse for all I-SPY Pts
No pCR (n=157)
Relapse-free Proportion
Relapse-free Proportion
pCR (n=58)
Years since surgery
RCB 0 (n=56)
RCB I (n=18)
RCB II (n=86)
RCB III (n=41)
Years since surgery
pCR and RCB in context of
molecular features
pCR: IHC vs Molecular Subtypes
IHC
Distribution
(n = 190)
pCR
(n = 190)
HR+HER2HR+HER2+
HR-HER2+
HR-HER2-
48%
12%
12%
28%
10%
32%
50%
33%
Gene Profile
Distribution
Intrinsic Subtypes ( n = 149)
Luminal A
Luminal B
Her2-enriched
Basal
Normal-like
HR = Hormone Receptor
29%
19%
15%
32%
5%
P-value
pCR
(n = 144)
P-value
2%
15%
52%
34%
43%
<0.0001
pCR Rates: RNA Classifiers
Gene Profile
Distribution
( n = 149)
ROR-S
Low
26%
Moderate
38%
High
37%
NKI 70
Good Signature
9%
Poor Signature
91%
Wound Healing
Quiescent
23%
Activated
77%
p53 Mutation Gene signature
Wildtype
Mutation
50%
50%
pCR (n = 144)
P-value
5%
22%
40%
8.8 x 10-4
0%
27%
0.038
6%
30%
0.0049
11%
38%
3.7 x 10-4
pCR Rates: DNA Classifiers
DNA Profile
Distribution
MIP Arrays
( n = 118)
No Amplification 17q
86%
pCR (n = 144)
P-value
15%
<0.0001
Amplification 17q
14%
59%
P53 Gene Chip
(n=181)
Any mutation
43%
38%
0.003
missense
Zn-binding
17%
75%
0.02
0.0003
null
12%
57%
pCR and RCB are VERY
significant predictors of
early relapse in the context
of a poor prognosis profile
Among Basal-like Tumors
RCB I (n = 2)
RCB 0 (n = 16)
RCB II (n = 17)
RCB III (n= 9)
Log-rank P = 5.5 x 10-7
Among NKI-70 High Risk
RCB I (n = 10)
RCB 0 (n= 35)
RCB II (n = 55)
RCB III (n = 22)
Log-rank P = 5.9 x 10-5
Among Activated-Wound Signature
RCB I (n = 5)
RCB 0 (n = 33)
RCB II (n = 45)
RCB III (n= 20)
Log-rank P = 4.4 x 10-4
Among p53 Mutation Profile
RCB I (n= 4)
RCB 0 (n= 27)
RCB II (n = 24)
RCB III (n = 12)
Log-rank P = 4.5 x 10-7
Published RNA Signatures
• Identify good and poor risk subsets
• pCR and RCB are highly predictive of
outcome in the poor risk subsets of all
signatures
• Patients in the high and low subsets differ
among signatures
• A composite molecular signature can be
created
Integrated score is a good
predictor of prognosis
Integrated Score: Good Prognosis
Distributed across RCB 0-III
All do well REGARDLESS of RCB
Integrated score poor prognosis
patients associate with RCB
Integrated Score,
Intermediate
prognosis
P=0.16
p = 0.158
P=1.89e-07
Integrated Score,
Poor prognosis
p = 1.89e-07
Activated Proteins Provide Clues
for Future Targeting
• Method:
– Reverse Phase Protein Array (RPMA)
– All samples laser capture microdissected
• Preliminary findings
– pts with pCR: increased phosphorylation of 4EBP1,
eNOS, cAbl, STAT5, EGFR, AKT (p<0.05)
• all within a linked EGFR-AKT-mTOR pathway activation
– pts ER+ with poor response: increased phosphorylation of
pIRS, pIGFR, p706S (p<0.05).
Observations from I-SPY
• LABC have high risk biology
– Minimum tumor size 3cm, mean size of 6cm
– 91% are molecularly high risk as defined by NKI 70 gene profile
– Not screen detected: 84% are interval cancers (Lin, Abstract 1503)
• Molecular features identify low and high risk subsets
– Low risk subsets: low pCR rates, but good outcomes (<5 yrs)
– High risk subsets: high pCR rates (28-59%) to std chemo
– High risk subsets: response to therapy (pCR, RCB) is highly
predictive of early outcome
• Residual Cancer Burden (RCB)
– More refined way to measure pathologic response
– Highly correlated with RFS and OS
• MRI Volume change is a non-invasive way to measure pCR
– Highly correlated with path CR and RCB: (Hylton, Abstract #529)
Next Steps
• The molecular data, with the exception of HER2, does
not yet tell us how to treat poor responders
– Recurrence after pCR limited to HER2+ patients preTrastuzumab (6 of 7)
– The I-SPY repository is a resource for such discovery
• We should target improvement in pCR/RCB to
improve outcomes
– I-SPY 2 is an adaptive neoadjuvant trial designed to rapidly
screen agents and biomarkers to improve pCR/RCB
• Exclude patients with good prognosis profile
BACK-UP
Quantitative and serial measurement
of tumor response by MRI
Pre
Treatment
Complete response
Post
Treatment
Partial response
Progressive disease
Patients Accrued
n=237
Patients
Withdrawn
n=16
Patients who didn’t
have surgery
n=6
Patients
Available for
Analysis
n=221
Patients with pathology
assessment after
Neoadjuvant Therapy
n=215
Patients without
RCB
n=14
Patients with pCR and
RCB
n=201
Tissue Distribution & Analyses Schema
UNC:
Dressler Lab
2 Paraffin
Cores
Tumor
UCSF
Check for
Tumor Presence
2 Frozen
Cores
Tumor
Present
Check for
Tumor
Presence
Core Remainder
GMU:
Liotta/Petricoin Lab
UNC: Livasy, Dressler Lab
PENN: DeMichele Lab
Her2 Protein
Over expression
Storage
Proteomics
Initial H&E
IHC
Initial H&E
DNA
FISH
Her2, TopoII
Amplification
CGH
UCSF: Gray Lab
Gene Chip
For P53
UNC: Carey/
Dorsey Lab
RNA
Gene
Expression
UNC: Perou Lab
UCSF: Haqq Lab
MDACC: Pusztai/
Symmans Lab
NKI: van’t Veer
Lab
Data uploaded in:
NCI caIntegrator NCI: caBIG, Madhavan
UCSC Cancer Genomics Browser UCSC: Haussler, Kent, Zhu, Wang
Data Integration: NCI caINTEGRATOR
Among ROR-S High Risk
RCB I (n = 5)
RCB 0 (n = 21)
RCB II (n = 18)
RCB III (n= 7)
Log-rank P = 4.3 x 10-9
Integrated score based prognosis
classes: ER+/ER- distributions
Prognosis
(Counts)
ER-
ER+
Indeterminate
Total
Good
2
28
1
31
Intermediate
22
38
4
64
Poor
37
10
2
49
Poor RCB 0
13
6
1
20
Poor RCB I
2
0
0
2
Poor RCB II
12
4
1
17
Poor RCB III
8
0
0
8
Questions
• Does early response help us to predict early
relapse?
– Complete Pathologic Response: pCR
– Residual Cancer Burden: RCB
• How do the molecular signatures impact on
the interpretation of pCR and RCB?
Integrating Molecular Profiles
60
50
Frequency
40
30
20
10
0
-4
-3
-2
-1
0
1
2
3
4
Integrated score
Poor
prognosis
Intermediate
prognosis
Good
prognosis
•Based on NKI-70, ROR-S, Wound Healing Signature,, p53
mutation profile: +1 , 0, -1 based upon score; Sum the scores