In Pursuit of Personalized Therapy for Breast Cancer: Gene Expression Signatures Jane S. Chawla, M.D. November 6, 2009
In Pursuit of Personalized Therapy for Breast Cancer: Gene Expression
Signatures
Jane S. Chawla, M.D.November 6, 2009
Objectives
Case: Chemotherapy or NOT Who should receive adjuvant chemotherapy? The birth of gene-expression signatures: Intrinsic
breast cancer subtypes Gene Expression Signatures under development
– 70-gene Signature– 21-gene Signature– 2-gene Signature– 50-gene signature
Case revisited
Should the patient receive chemotherapy?
Clinical Case
30 y/o WF with newly diagnosed IDC R breast MMG / Ultrasound 1.2 x 1.2 cm mass at 10 o’clock + surrounding
microcalcifications + several suspicious axillary LNs s/p bilateral mastectomies at OSH Surgical Path - 4.6 cm lesion consisting of IDC mixed with DCIS;
intermediate grade; + lymphovascular invasion; 11 (-) LNs; ER+/PR+/HER-2-; T2N0M0
Oncotype DX - -> low risk for recurrence Based on oncotype results her oncologist recommended adjuvant
tamoxifen x 5 years Pt presents for a second opinion
How else can we differentiate these tumors?
Who Should Receive Adjuvant Chemotherapy?
65% of women with invasive breast cancer have LN(-) disease Adjuvant chemotherapy improves DFS and OS in pre- and
postmenopausal women <70 with LN+/- breast cancer NSABP B-14 & B-20 showed the benefit of tamoxifen & chemo
in ER+/LN- breast cancer– Likelihood of distant mets at 10 years is about 15%– 85% of pts are overtreated if chemo given to all patients
Those with poor prognostic features benefit most from chemotherapy
Main clinical prognostic factors: age, tumor size, axillary LN status, tumor histology, grade, and hormone-receptor status
Sotiriou, C. et al. NEJM, 2009.
Luminal ALuminal B
HER2+Basal-like
Intrinsic Breast CancerSubtypes described by
Perou et al.
Express ↑ amountsOf luminal cyto-Keratins & geneticMarkers of luminalEpithelial cells ofNormal tissue
Express ↑ levels of EGFR, c-kit, & growth factors like hepatocyte growth factor and IGF
Gene Expression Profiles or “Signatures”
Composed of a selection of genes felt to provide prognostic or predictive information about tumors
A number of gene expression signatures have been developed to help identify those patients at highest risk for recurrent disease
This may avoid chemotherapy administration to patients at low risk
The 70-gene Assay: MammaPrint
Mammaprint: Development of the 70-Gene Signature
DNA microarray analysis of 78 breast primary tumors (untreated)– Pts were <55 years of age with T1-2/N0 disease– Pts selected based on outcome: Distant metastases within 5 years
Statistical analysis, “supervised classification,” identified 231 genes correlated with disease outcome Top 70 genes selected
Genes that regulate cell cycle, invasion, metastasis, & angiogenesis Patients categorized as “good prognosis” or “poor prognosis.” Found to be a better predictor of distant metastases within 5 years
than all clinical variables in this study Odds ratio (distant metastases): poor to good prognosis groups =
15
Van ’t Veer, L. Nature, 2002.
Retrospective Validation of the 70-Gene Signature
295 women ages ≤ 52 with T1-2, LN-/+ breast cancer– 226 ER+ / 69 ER-– chemo 31%; hormonal 7%; both 7%
61 pts included in the analysis were used to develop 70-gene signature
Van de Vijver. NEJM, 2002.
Prognosis Signature Predictive of End Points (10 years)
995% ±2.6%
555% ±4.4%
85 % ±4.3%
51 % ±4.5%
Probability of RemainingMetastasis-free Overall Survival
Multivariate Analysis
Independent predictors of risk of distant metastasis as a 1st event
70-gene signature Tumor diameter Chemotherapy
Lymph node status Vascular invasion
Authors report that gene signature is predictive of distant mets in LN+ pts
Multivariate analysis argues that LN status isindependent of gene signature as a predictor
Van de Vijver. NEJM, 2002.
TRANSBIG Independent Retrospective Validation Study
Retrospective evaluation of 302 pts from several sites in Europe (age<60, T1-2, LN-) previously untreated
Aim: to examine whether the 70-gene signature had prognostic value independent of the best clinical risk classifications (St Gallens, Nottingham Prognostic Index, & adjuvant online)
Buyse, M. J of NCI, 2006.
Does 70-gene Signature have Independent Prognostic Value?
Gene signature adds independent prognostic information to that provided by various risk classifications
The signature remained a statistically significant prognostic factor for time to distant metastases & OS even after adjustment for various risk classifications (HR 2.15 & 2.15, respectively)
Buyse, M. J of NCI, 2006.
Forrest Plot of HR for Time to Distant Metastases
Although, average HR of 2.32 is significant…there was wide variation in the HRs among clinical sites.
Therefore model is very sensitive to variation in patient population
ClinicalSites
Buyse, M. J of NCI, 2006.
Conclusions of Validation Studies
Studies were retrospective Concern for overfitting with supervised analysis of
relatively small sample sizes and large numbers of genes
The first validation study contained 61 patients from study on which the classifier was built
Claims exceed evidence?
Prospective Validation of Mammaprint: The MINDACT Trial
Accrual started 2/07 and is expected to be finished within 3 years
6,000 LN- patients
High risk Chemo +Endocrine
Risk assessed viaClinicopathologicalFactors (adjuvant)+ Mammaprint
Low risk Endocrine
Discordant cases:random assignmentto follow genomicvs clinicopathologicresult
Cardoso, F. JCO, 2008.
The 21-gene Recurrence Score: Oncotype DX
Oncotype Dx: The 21-Gene Assay Designed to quantify the risk of distant recurrence in patients
with LN(-), ER(+) tumors receiving tamoxifen RT-PCR was used to quantify gene expression from fixed,
parafin-embedded tumor tissue 250 candidate genes selected based on published literature,
genomic databases, & experiments based on DNA arrays on fresh-frozen tissue
Analyzed data from 3 independent studies (447 patients) including tamoxifen-only arm of NSABP trial B-20 to test relation b/w 250 genes and recurrence of breast cancer
From these studies, 21 genes were selected that correlated with proliferation and endocrine response
Levels of Gene Expression Determine Recurrence Score
21-gene assay = 16 outcome-related genes + 5 reference genes
Higher expression levels of“favorable” genes = ↓ RS
Higher expression levels of“unfavorable” genes = ↑ RS
A risk score is calculated from 0 -100Cutoff points chosen based onResults of NSABP trial B-20
Sparano, J & Paik, S. JCO, 2008.
Prospective Validation Study of 21-Gene Assay
Analyzed tumor blocks of 668 patients on NSABP trial B-14 study that were randomized to tamoxifen
RS was significantly correlated with relapse free interval & OS (p<0.001)
Rate of distant recurrence at 10 years:– Low risk – 6.8%– Intermediate risk – 14.3%– High risk – 30.5% (similar risk
to LN+ patients)
No distant recurrence (low risk) = 93.2%
No distant recurrence (high risk) = 69.5%
p<0.001
Paik, S. JCO, 2004.
Other End Points
RS predicted distantrecurrence independent of age & tumor size
The likelihood of distant recurrence ↑ continuously as the RS ↑
Multivariate Cox AnalysisDistant Recurrence as a
Function of RS
Paik, S. JCO, 2004.
Does RS Have Prognostic or Predictive Value?
Oncotype was evaluated in both TAM (n=290) & placebo (n=355) arms of the NSABP B-14 trial to determine its prognostic value in an untreated population– Patients with low (p=0.02) and intermediate (p=0.04) RS derived
benefit from tamoxifen, whereas those with high RS (p=0.82) did not– Although sample size limited, seems to indicate that ↓ RS predicts
greater benefit to TAM Population-based external validation study of women with ER+/LN- breast
cancer in The Northern California Kaiser Permanente tumor registry, age <75, untreated & TAM– RS was associated with risk of breast cancer death in tamoxifen-
treated (p=0.003) and untreated patients (p=0.03)– Prognostic in TAM-treated and untreated pts
Paik, S. ASCO Abstract, 2005.
Habel, L. Breast Cancer Research, 2006.
Does the 21-Gene Assay Predict Response to Chemotherapy?
NSABP B20 trial: 651 ER+/LN- pts randomized to chemo + TAM (n=424) v. TAM alone (n=227) (ALL PATIENTS)– Chemo – CMF or MF– No distant recurrence - 92.2% (chemo + TAM) v. 87.7% (TAM)– No local/distant relapse - 90.1% (chemo + TAM) v. 83.5% (TAM)– Overall Survival – 89.5% (chemo + TAM) v. 86.4% (TAM)
The TAM alone arm was used in the training set in development of the 21-gene assay
Paik, S. JCO, 2006.
Kaplan-Meier Plots for Distant Recurrence
All patients Low Risk RS <18
Intermediate Risk RS 18-30High Risk RS ≥ 31
88%
60%
92.2%
87.7%
Paik, S. JCO, 2006.
Relative & Absolute Benefit of Chemotherapy as a Function of RS
27.6%
Low RS mean absolute decrease in distant recurrence rate of -1.1%
High RS mean absolute decrease inDistant recurrence rate of 27.6%
Intermediate RS no substantial benefit,But the uncertainty in the estimate cannotExclude a clinically important benefit
Paik, S. JCO, 2006.
Prospective Validation of Oncotype DX: The TAILORX Trial
Dowsett, M. & Dunbier, A. Clin Cancer Res, 2008.
Low RS:HormonalTherapy
High RS:Chemo +HormonalTherapy
Hormonal Therapy Chemo + Hormonal
11,248 ER+/LN- patients
The Two-gene Ratio: HOXB13:IL17BR
Two-Gene Expression Ratio: HOXB13:IL17BR
Developed to predict response to tamoxifen in ER+/PR+, early stage , LN+/- breast cancers
Performed microarray gene expression analysis of tumors from 60 women treated with adjuvant tamoxifen– Patients selected based on outcome: early relapse or not– RNA isolated from frozen tumor-tissue 9 genes were
identified with p<0.001 2 selected– HOXB13 was overexpressed in tamoxifen recurrences & IL17BR
was overexpressed in tamoxifen nonrecurrences– HOX family of genes implicated in tumor invasion and
metastases– HOXB13 is expression is often up-regulated in breast cancer cells
Ma, X. Cancer Cell, 2004.
HOXB13:IL17BR Predicts Response to Tamoxifen
HOXB13 & IL17BR outperformed other predictive markers (quant-itative ER and PR, ERBB2, & EGFR)
HOXB13:IL17BR ratio was a better composite predictor of tamoxifen response
HOXB13:IL17BR was an indepen-dent predictor of treatment outcome when controlled for tumor size & expression of PGR and ERBB2
HOXB13:IL17BR ratio was highly correlated with clinical outcome in an independent 20 pt cohort
Ma, X. Cancer Cell, 2004.
Lack of Consistency of 2-gene Signature
STUDY PATIENTS HAZARD RATIOS OUTCOME CUT POINTS
Goetz et al. 206 ER+, LN+/- adjuvant TAM
RFS – 1.98DFS – 2.03OS – 2.4(LN- pts)
•Prognostic in TAM-treated LN- pts only
TAM -1.849
Ma et al. 852 pts (66% untreated & 34% TAM-treated )
RFS – 3.9 (p=.007) in TAM and untreated pts combined
•Prognostic in ER+/LN- pts•Unable to compare TAM & untreated groups
TAM -0.06Untreated – 1.0
Jansen et al. 1,252 LN+/-, 32% adj tx(44% hormonal, 53% chemo, & 3% both)
DFS (LN-, no tx) – 1.74 (p=.006)DFS (TAM) – 2.97 (p<.001)
•Prognostic in untreated, LN-•Predicts response to TAM
Untreated – 1.0TAM – 0.06
Limitations of These Studies
Extreme sensitivity to training set since there is such variability in the cut point
Does not seem to be effective in LN+ patients Needs prospective validation
Intrinsic Subtypes: PAM50
Intrinsic Subtypes: PAM50 Training set =189 breast tumor + 29 normal samples qRT-PCR An expanded gene set of 1,906 genes from other microarray studies
analyzed by hierarchical clustering significant clusters correlated to known intrinisic subtypes narrowed to 50 significant genes (PAM50)
Intrinsic subtypes are useful in predicting RFS of untreated patients & those stratified by ER status
Cox models tested using intrinsic subtypes alone and together with clinical variables (ER status, tumor size, & node status) – ROR-S Subtypes alone– ROR-C Subtype + clinical variables– Models containing both subtype & clinical variables better than either
clinical variables (p<0.0001) or subtype alone (p<0.0001)
ROR-C Predicts RFS in Untreated LN- Patients
Only Luminal A group contained low risk patients
ROR-C Scores Stratified by SubtypeKaplan-Meier Plots of RFS of Risk
Groups
Hig
h R
isk
Med
ium
Ris
kL
ow
Ris
k
Risk Groups Predict RFS
PAM50 Predicts Response to Neoadjuvant T/FAC
ROR-S v Probability of pCR
Low-risk tumors (luminal A) chemo-insensitive tumors with ↓ probability of pCR
Plateau chemo- resistance among high risk tumors
Sensitivity 94% & NPV 97% for identifying nonresponders toNeoadjuvant chemo
ROCC of ROR-S v pCR
↑ Probability of pCR with ↑ ROR-S
PAM50 Prognostic in TAM-treated Patients
786 ER+ TAM treated pts qRT-PCR, PAM50, and ROR calculation done– Multivariate analysis – Intrinsic subtype and ROR
score were independently prognostic (Grade and HER2 were not)
– Low ROR score good outcomes (even if +LN)– Found to be superior to immunohistochemical
markers & adjuvant! Online as a prognosticator
Ellis, M. ASCO, 2009.
5 Gene Signatures Compared
Is There Concordance Between Gene Expression Profiles?
295 Stage I/II ER +/-
Analyzes done withAll pts & ER+ only
165 untreated20 TAM20 TAM + chemo90 chemo
Fan, C. NEJM, 2006.
Relapse-free Survival Overall Survival
IntrinsicSubtype
RecurrenceScore
70-GeneProfile
WoundResponse
Two-GeneRatio
All models except the two-gene ratio modelwere significant predictors of both RFS and OS.
Summary of Gene Signatures
70-gene Signature
21-geneSignature
2-GeneRatio
IntrinsicSubtypes
AnalysisApproach
Supervised Supervised Supervised Unsupervised
Tissue Type Fresh or FrozenFormalin-Fixed,
Parafin-embedded
Formalin-Fixed, Parafin-
embedded
Formalin-Fixed, Parafin-
embedded
Technique DNA microarrays
Q-RT-PCR Q-RT-PCR Q-RT-PCR
Prognostic Untreated pts age<60, T1-2, LN-
Untreated & TAM-treated ER+/LN-
TAM-treated, ER+/LN- untreated
TAM-treated
PredictiveNO
Benefit to TAM +/- CMF/MF
Response to TAM NO
Validation Retrospective Retrospective Retrospective Retrospective
ProspectiveTrials MINDACT TAILORX NONE NONE
Case Revisited
Pathology report showed a poor-risk feature: lymphovascular invasion– Lee et al showed that those with peritumoral
lymphatic and blood vessel invasion had a 4.7x higher risk of relapse
Therefore, her oncotype DX results disregarded started on chemotherapy.
Conclusions
No gene signature is currently able to replace all clinicopathological variables in assessing risk of recurrence
What patient populations should be evaluated with gene signatures is actively being investigated
NCCN guidelines suggest that oncotype DX is an option for risk evaluation in 0.6-1 cm tumors with unfavorable characteristics or in >1 cm LN-, ER+/HER2 tumors
Resources Dowsett, M. & Dunbier, A. Emerging Biomarkers and New Understanding of Traditional
Markers in Personalized Therapy for Breast Cancer. Clinical Cancer Research 2008; 14: 8019-26. Sotiriou, C et al. Gene-Expression Signatures in Breast Cancer. New England Journal of
Medicine 2009; 360: 790-800. Van’t Veer, L et al. Gene Expression Profiling Predicts Clinical Outcome of Breast Cancer.
Nature 2002; 415: 530-36. Van de Vijver, M. et al. A Gene-Expression Signature as a Predictor of Survival in Breast Cancer.
The New England Journal of Medicine 2002; 347: 1999-2009. Paik, S. et al. A Multigene Assay to Predict Recurrence of Tamoxifen-Treated, Node-Negative
Breast Cancer. The New England Journal of Medicine 2004; 351: 2817-2826. Sparano, J and Paik, S. Development of the 21-Gene Assay and Its Application in Clinical
Practice and Clinical Trials. Journal of Clinical Oncology 2008; 26: 721-728. Paik, S. et al. Gene Expression and Benefit of Chemotherapy in Women With Node-Negative,
Estrogen-Receptor-Positive Breast Cancer. Journal of Clinical Oncology 2006; 24: 3726-3734. Paik, S. et al. Expression of the 21 genes in the Recurrence Score assay and tamoxifen clinical
benefit in the NSABP study B-14 of node negative, estrogen receptor positive breast cancer. Journal of Clinical Oncology 2005; 23: suppl abstr 510.
Resources Ma, X. et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients
treated with tamoxifen. Cancer Cell, 2004; 5: 607-16. Ma X. et al. The HOXB13:IL17BR Expression Index Is a Prognostic Factor in Early-Stage Breast
Cancer. Journal of Clinical Oncology, 2006; 24: 4611-4619. Jansen, M. et al. HOXB13-to-IL17BR Expression Ratio Is Related With Tumor Aggressiveness
and Response to Tamoxifen of Recurrent Breast Cancer: A Retrospective Study. Journal of Clinical Oncology, 2007; 25: 662-668.
Cardosa, F. et al. Clinical Application of the 70-Gene Profile: The MINDACT Trial. Journal of Clinical Oncology, 2008; 26: 729-735.
Intrinsic Subtype & Response to Paclitaxel
CALGB 9344 adding T to AC in the adjuvant setting led to a 5% absolute improvement in DFS
Of the 42% of pts with complete data, no benefit in the ER+/HER2- pts
Microarray done in 2039 pts and pts were categorized into intrinsic subtypes: 790 luminal A, 340 luminal B, 221 HER2-enriched, 444 basal, and 93 ER-/HER2-/nonbasal
Intrinsic subtype was prognostically significant in multivariate analysis (p<0.001)
No significant interaction of Ki67 with paclitaxel among ER+/HER2- patients
Basal subtype predicted benefit from paclitaxel in multivariate analysis (HR 0.75, p=0.033)
Variation in Hazard Ratios over Time
Cardoso, F. JCO, 2008.
Substantial variation over time of adjusted hazard ratio ability of gene signature to identify those who will develop distant relapse
is greatest within 5 years of diagnosis.
Sensitivity & Specificity of 70-gene Signature
Sensitivity is as good as other risk classification models
Specificity is slightly better than the other risk classification models
Buyse, M. J of NCI, 2006.
Classification of Tumor Samples According to Model
Basal-like, HER2+ and ER-, & luminal B all had a poor outcome by 70-gene, recurrence score, and wound –response models
Normal-like and luminal A had variability in the outcomes