1. Fatima Cardoso, MD Jules Bordet Institute & TRANSBIG 2006 European Breast Cancer Meeting Stockholm, Sweden 20–21 May 2006 USING PROGNOSTIC & PREDICTIVE FACTORS IN BREAST CANCER

2. PROGNOSTIC FACTOR % Treat. A Treat. B + -

3. PREDICTIVE FACTOR Case 1 Case 2 % % + + - - Treat. B Treat. A Treat. B Treat. A

4. PROGNOSTIC FACTORS Who needs a treatment? PREDICTIVE FACTORS Which treatment is best? THERAPEUTIC CHOICES AVOID UNDER AND OVER TREATMENT INDIVIDUALIZE TREATMENT WHY DO WE NEED PROGNOSTIC AND PREDICTIVE FACTORS

5. ER-ER+ RFS Basal-like1 HER-2-like Luminal1 Luminal2Luminal3 Basal-like2 Adapted from Sotiriou et al, PNAS, 2003 BC GENE EXPRESSION PATTERNS and OUTCOME Molecular (re-)classification of BC

6. PROGRESS IN ADJUVANT CHEMOTHERAPY FOR BREAST CANCER L-PAM, MF CMF x 6 AC x 4 FAC  FEC x 6 A(E)  CMF AC x 4  Paclitaxel x 4 TAC x 6 FEC  docetaxel AC  paclitaxel dose-dense ± + ++ +++ ++ +++ ± + Average treatment effect Financial toxicity 1970’s1980’s 1990’s 2000’s Successive generations of adjuvant CT regimens Adapted with permission from G. Hortobagyi d)  20.000 $ c)  13.800 $ b)  7.400 $ a)  800 $ +++ ADJUVANT TRASTUZUMAB +++

7. New prognostic factors accepted: HER-2, vascular invasion Node+ 1-3: in average risk group, if HER-2– and no vascular invasion St Gallen 2005 Consensus: What’s new? Beyond St Gallen 2005 … uPA, PAI-1 Cyclin E Genomic signatures Oncotype DX* (predictive & Px) Oncotype DX* (predictive & Px) Topo-II-  *Genomic Health

8. uPA-PAI-1

9. CLINICAL RELEVANCE OF uPA & PAI-1 IN PRIMARY BREAST CANCER uPA and PAI-1: first novel tumor biological factors in breast cancer with clinical relevance validated at highest level of evidence (LOE I)  Standardized quality assured ELISA tests: Sweep et al, Br J Cancer 78: 1434-41, 1998  Prospective multi-center therapy trial („Chemo N0“): Jänicke et al, JNCI 93: 913-20, 2001  EORTC RBG meta analysis (n=8,377): Look et al, JNCI 94:116-28, 2002  Recommended for clinical risk assessment: AGO Therapy Guidelines „breast cancer“ (since 2002):www.ago-online.de N. Harbeck – used with permission

10. uPA AND PAI-1 FIRST NOVEL TUMOR BIOLOGICAL FACTORS IN BC WITH LEVEL 1 OF EVIDENCE WHY ARE THEY NOT WIDELY USED? 1. ELISA not commonly used in pathological practice a.Biochemistry lab required b.Further personnel training required c.€€££$$ required 2. Frozen tumor specimen required 3. Large quantity (100 µg) required Target population = small tumors – feasible ? 4. Population used in validation studies: Interaction with ER status not well defined (?)

11. HOW CAN THEY BECOME WIDELY USED? 1.Refining ELISA test – less tissue 2.Alternative techniques – other protein assays – gene expression 3.Further validation according to ER status ALL ONGOING uPA AND PAI-1 FIRST NOVEL TUMOR BIOLOGICAL FACTORS IN BC WITH LEVEL 1 OF EVIDENCE

12. GENOMIC SIGNATURES

13. BIG-TRANSBIG Secretariat– Used with permission IMPROVED RISK ASSESSMENT OF EARLY BREAST CANCER THROUGH GENE EXPRESSION PROFILING microarray Gene-expression profile Good signature Poor signature N Engl J Med, Vol 347 (25), Dec. 2002 ~4% die of breast cancer ~96% survive breast cancer ~50% die of breast cancer ~50% survive breast cancer

14. TRANSLATING MOLECULAR KNOWLEDGE INTO EARLY BREAST CANCER MANAGEMENT

15. BIG-TRANSBIG Secretariat– Used with permission Audited clinical data INDEPENDENT VALIDATION : DESIGN RNA Achieved n = 307 Target n = 400 Amsterdam Gene expression profiling Agilent platform Agilent platform 70-gene prognostic 70-gene prognostic custom designed custom designed chip chip High or low gene signature risk Clinical data « Local » pathological data Brussels Comparison of clinical vs gene signature assessment of prognostic risk Endpoints 1. TDM 2. OS 3. DMFS, DFS Tissue samples  UK (Guy’s, Oxford) : 1984 => 1996  France (IGR, CRH) : 1978 => 1998  Sweden (Karolinska) : 1980 => 1990 Node negative, untreated Node negative, untreated < 60 years old < 60 years old > 5 years follow-up > 5 years follow-up T1, T2 T1, T2 Tumor cell % > 50% Tumor cell % > 50% Centrally reviewed path data (Milan)

16. BIG-TRANSBIG Secretariat– Used with permission OVERALL SURVIVAL by GENE SIGNATURE RISK Amsterdam/Agendia Signature 10-year OS 89% (81%-94%) 10-year OS 70% (62%-76%) Average Survival HR 2.66 Average Survival HR  2.66 M. Buyse et al. JNCI 2006. In press

17. BIG-TRANSBIG Secretariat– Used with permission TRANSBIG INDEPENDENT VALIDATION The best signature? Amsterdam’s Signature 70 genes Rotterdam’s Signature 76 genes TEST ALL IN VALIDATION SERIES & DECIDE Only few genes in common … But similar biological pathways Brussels’ GGI signature

18. BIG-TRANSBIG Secretariat– Used with permission OVERALL SURVIVAL by GENE SIGNATURE RISK Rotterdam/Veridex Signature 5-year survival: low risk group: 0.98 (0.88-1.00) high risk group: 0.84 (0.77-0.89) 10 year survival: low risk group: 0.87 (0.73-0.94) high risk group: 0.72 (0.63-0.78) C. Desmedt et al. Presentated at: EBCC 2006

19. CONCLUSIONS VALIDATION PHASE The Amsterdam 70-gene signature has been independently validated The Amsterdam 70-gene signature has been independently validated The Rotterdam 76-gene & Genomic Grade signatures have been independently validated using the same TRANSBIG validation series The Rotterdam 76-gene & Genomic Grade signatures have been independently validated using the same TRANSBIG validation series The performances of the signatures are similar The performances of the signatures are similar There is a strong time dependency of all signatures (better predictors of EARLY RELAPSE), which was not seen for the clinical risk There is a strong time dependency of all signatures (better predictors of EARLY RELAPSE), which was not seen for the clinical risk The Amsterdam 70-gene test is robust (laboratory reproducibility) and available for patient diagnostic testing The Amsterdam 70-gene test is robust (laboratory reproducibility) and available for patient diagnostic testing GREEN LIGHT FOR MINDACT TRIAL! GREEN LIGHT FOR MINDACT TRIAL!

20. Evaluate Clinical-Pathological risk and 70-gene signature risk Clinical-pathological and 70-gene both HIGH risk Discordant cases Clin-Path HIGH 70-gene LOW Clin-Path LOW 70-gene HIGH Clinical-pathological and 70-gene both LOW risk Use Clin-Path risk to decide Chemo or not Use 70-gene risk to decide Chemo or not 55% 32%13% R1 Chemotherapy N=3300 N=780 Endocrine therapy EORTC-BIG MINDACT TRIAL DESIGN 6,000 Node negative women N=1920 Potential CT sparing in 10-15% pts

21. GENOMIC GRADE

22. Histologic Grade G1 G2 G3 Genomic Grade GG1 GG2 GG3 Sotiriou et al., ASCO 2005 Poor inter observer reproducibility G2: difficult treatment decision making, under- or over treatment likely Findings consistent across multiple data sets and microarray platforms More objective assessment Easier treatment decision-making High proportion of genes involved in cell proliferation ! C. Sotiriou – used with permission

23. HistologicalGrade 3 Histological Grade 3 HG3 GENOMIC GRADE IN EACH OF THE HISTOLOGIC GRADE SUBGROUPS Genomic Grade 1 Genomic Grade 3 Histological Grade 2 HG2 Histological Grade 1 HG1 C. Sotiriou – used with permission C. Sotiriou et al. JNCI 2006

24. Oncotype DX NSABP & Genomic Health

25. MULTI GENE RT-PCR ASSAY FOR PREDICTING RECURRENCE IN NODE NEGATIVE BC PATIENTS 250 candidate genes Tested using RT-PCR Three studies 21 GENE PREDICTOR Recurrence score lowintermediatehigh

26. PROLIFERATION Ki-67 STK15 Survivin Cyclin B1 MYBL2 ESTROGEN ER PGR Bcl2 SCUBE2 INVASION Stromolysin 3 Cathepsin L2 HER2 GRB7 HER2 GSTM1 REFERENCEBeta-actinGAPDHRPLPOGUSTFRC Best RT-PCR performance and most robust predictors CD68 BAG1 Paik et al, N Engl J Med 2004 THREE BREAST CANCER STUDIES USED TO SELECT CANDIDATE GENES FOR A RECURRENCE SCORE UNDER TAMOXIFEN TREATMENT Recurrence score for TAM-treated pts established and subsequently validated

27. 338 pts 149 pts 181 pts B14-RESULTS DRFS—Low, Intermediate, High RS Groups Paik et al, N Engl J Med 2004

28. PREDICTIVE MARKERS

29. HER-2 neu 95% Negative predictive value <5% chances of responding to TRASTUZUMAB (HER-2) or to HT (ER) 30-70% Positive predictive value Accepted Predictive Markers In Breast Cancer ER/PgR OxfordOverview2000 St Gallen ConsensusPanel2003 NIHConsensusPanel2000 ASCOGuidelines2001 30%-70% chances of responding to HT (ER) & 40%-50% of responding to TRASTUZUMAB (HER-2)

30. PREDICTIVE MARKERS FOR CHEMOTHERAPY

31. FISH IHC ADJUVANT SETTING CMF vs. ANTHRA-BASED TOPO II  RESULTS Di Leo A et al, Clin Cancer Res, 2002 All pts with HER-2 amplification Di Leo A et al, Ann Oncol 2001

32. 4 x AC 60/600 mg/m 2 4 x Docetaxel 100 mg/m 2 6 x Docetaxel and Carboplatin 75 mg/m 2 AUC 6 1 Year Trastuzumab N=3,222 1 Year Trastuzumab AC  T AC  TH TCH Her2+ (Central FISH) N+ or high risk N- 4 x AC 60/600 mg/m 2 4 x Docetaxel 100 mg/m 2 Slamon D., SABCS 2005 BCIRG 006 Stratified by Nodes and Hormonal Receptor Status

33. Disease Free Survival % Disease Free 0.5 0.6 0.7 0.8 0.9 1.0 012345 Year from randomization 77% 86% 80% 73% 84% 80%86% 93% 91% Patients Events 1073147AC->T 107477AC->TH 107598TCH HR (AC->TH vs AC->T) = 0.49 [0.37;0.65] P<0.0001 HR (TCH vs AC->T) = 0.61 [0.47;0.79] P=0.0002 Slamon D., SABCS 2005 AC->TH AC->T TCH

34. DFS CO-AMPLIFIED TOPO II BY ARM % Disease Free Months 0.5 0.6 0.8 1.0 061218243036424854 Patients EventsTreatment 22723 AC->T 26513AC->TH 25221TCH Logrank P= 0.24 TCH AC->TH AC->T Slamon D., SABCS 2005

35. DFS NON CO-AMPLIFIED TOPO II BY ARM % Disease Free Months 0.0 0.6 0.8 1.0 061218243036424854 Patients EventsTreatment 45892AC->T 47245AC->TH 44654TCH Logrank P= <0.001 TCH AC->TH AC->T Slamon D., SABCS 2005

36. HER-2 AND TOPOISOMERASE-II  PROMISING POTENTIAL PREDICTIVE MARKERS OF ANTHRACYCLINE EFFICACY HOW TO OBTAIN LEVEL 1 EVIDENCE LARGE PROSPECTIVE TRIALS META-ANALYSIS

37. HER-2 AND TOPOISOMERASE-II  AS POTENTIAL PREDICTIVE MARKERS OF ANTHRACYCLINE EFFICACY: A META-ANALYSIS DANISH TRIAL FEC vs CMF UK TRIAL E  CMF vs CMF NCIC-CTG TRIAL CEF vs CMF BELGIAN TRIAL EC vs CMF Tampere University Laboratory Central evaluation of HER-2/TOPO II by FISH Central evaluation of HER-2/TOPO II gene amplification by FISH Correlation with outcome of CMF or anthracycline-based therapy with  4,500 tumor samples

38. TOP TRIAL OR « TRIAL OF PRINCIPLE » Operable tumors, > 2 cm ER-negative EPIRUBICIN 100 mg/m² x 4 SURGERY Docetaxel x 4 Radiotherapy ± HT Hypothesis :  pCr in HER-2 / Topo2 co-amplified tumors  pCr in HER-2 - / basal-like 1 tumors Incisional biopsy Snap frozen sample HER2/Topo2 FISH analysis (Vysis probe) Genomic signature of response to anthracyclines Inflammatory or LABC ER-negative EPIRUBICIN 100 mg/m² x 6 dose dense / 2w + G-CSF Gene expression analysis

39. EORTC-BIG-p53 TRANSLATIONAL RESEARCH TRIAL: STUDY DESIGN Target accrual= 1300 (872 p53 -, 436 p53 + ) Hypothesis: ↑ DFS at 3 y by 5% in p53 - and by 20% in p53 + RANDRAND Non Taxane arm FEC 100 or Canadian FEC Taxane arm T-T-T-ET-ET-ET Sample 1: standard fixation Incisional biopsy Sample 2: snap frozen. Loc. adv.. Infl.. Large Operable Local ± TAM therapy Local ± TAM therapy P53 pathway P53 analysis

40. FRAGRANCE trial 4 - 6 months Letrozole 15 days Microarray Analysis Microarray Analysis Genomic signature of de novo AI resistance Microarray Analysis Postmenopausal patients (no age limits) Non-candidates for CT T  2 cm Stages I, II & III ER and/or PgR+

41. INTEGRATING TRANSLATIONAL RESEARCH IN CLINICAL RESEARCH & PRACTICE Multidisciplinarity Multidisciplinarity Collaboration (between specialties, between centers…) Collaboration (between specialties, between centers…) Bench-to-bedside-to-bench Bench-to-bedside-to-bench Biological material collection (unethical not to do it!) Biological material collection (unethical not to do it!) Patient selection & treatment tailored to the individual Patient selection & treatment tailored to the individual New technologies, new statistical methods… New technologies, new statistical methods… Costs ?? Costs ?? INDISPENSABLE and already ongoing

42. Total expected costs:  €35, 000,000 EU funding €7,000,000 OTHER: National Funding Pharmaceutical Industry Biotechnology companies (Agendia) Other grants NATIONAL FUNDING FOR NATIONAL PATIENTS (indispensible) MINDACT & TRANSBIG FUNDING - 1

43. ACKNOWLEDGEMENTS BIG-TRANSBIG TeamBordet Fellows Translational Research Team M. Piccart