Welcome to Master Class for Oncologists Miami, FL December 18, 2009 Session 1: 1:00 PM - 1:45 PM Towards Personalized Medicine in Breast Cancer: Understanding Molecular Subtypes and the Role of Diagnostics Lisa A. Carey, MD Associate Professor Medical Oncology Presenter Disclosure Information The following relationships exist related to this presentation: Dr Lisa A. Carey is an uncompensated consultant or speaker for for Eli Lilly and Company, Wyeth, Glaxo SmithKline, Genentech, and Sanofi-Aventis She / UNC receives research grants from Genentech; GlaxoSmithKline; Novartis; and Bristol-Myers Squibb. Off Label/Investigational Discussion In accordance with Pri-Med Institute policy, faculty have been asked to disclose discussion of unlabeled or unapproved use(s) of drugs or devices during the course of their presentations. Making Understanding Molecular Profiles Less Painful Genomic profiling studies reveal: the heterogeneity of breast cancer the biologic pathways driving prognosis the biologic pathways defining response We will focus upon those that have relevance in the clinic today and maybe a few for tomorrow. How many known intrinsic subtypes of breast cancer are there? 1. 1 2. 2-3 3. 4-5 4. 10-15 5. As many as Dr. Burstein says there are Genomic Profiling Assays in Breast Cancer: The Good, The Bad, The Uninterpretable Good: Expand existing knowledge Identify relevant biology Not reliant on single markers Bad: Too much information Gene protein function Uninterpretable: What population / indication? Prognostic? Predictive? Validation requires prospective or exceptional datasets 6 1
Unsupervised Analysis Give Us Breast Cancer Intrinsic Subtypes Genomic Profiling Analyses Unsupervised are samples different? Good for identifying biologically distinct groups Not good for predicting anything Example: Intrinsic subtypes Intrinsic Gene Clusters: Hormone receptor related Supervised can gene sets predict clinical event? HER2 related Generally good for predicting what they were designed to predict. Tricky validating is hard, need homogeneous population and Rx. Genes may be part of pathways, not themselves relevant Example: 70-Gene/Mammaprint, Recurrence Score ER + subtypes Basal ER negative subtypes Proliferation 7 8 Courtesy Chuck Perou Hormone Receptor-Driven: Luminal Subtypes Breast Cancer Subtypes and Prognosis Luminal B Heterogeneity in ER+ prognosis Luminal A Majority of tumors high expression hormone receptor related gene cluster. HER2 + or Can be proliferative or not Relative uniformity of ER, at least until HER2 targeting Most heterogeneous group N= 311 Courtesy C. Perou 9 p < 0.0000001 10Sorlie 9 T et al, PNAS 2001 Basal-like (Also Known as Triple Negative ) Subtype HER2-Enriched Subtype About 15% of tumors HER2 Low HER2 cluster expression 15-20% of tumors Basal High basal cluster (CK 5, 17, High HER2 cluster expression EGFR, αb crystallin, c-kit etc) Basal Low ER (and related genes) cluster expression Luminal expression Low ER (and related genes) cluster expression ER+ is a different subtype Very proliferative Luminal ~50% are p53 mutant Very proliferative Evidence of genomic instability Abnormal Majority of tumors in BRCA1 DNA repair Proliferation Proliferation 11 12 2 mutation carriers Evidence of BRCA1 dysfunction even in sporadic tumors
Triple Negative Phenotype and Basal-Like Expression Profile Targeting Basal-Like Breast Cancer: BRCA1, PARP1 and DNA Repair KEEP IN MIND: When we talk about triple negative breast cancer, we are mostly but not entirely talking about the basallike molecular subtype BRCA1 DNA Repair Mechanism Homologous (error free) recombination Triple negative but not basal 10 30% Can also include claudin low, a subtype notable for high expression of stem cell markers IHC Triple negative and Basal like Array Basal but not triple negative 15 40% are ER+, PR+, or HER2+ DS DNA breaks Naturally occurring or induced by chemo PARP1 Nonhomologous recombination PARP inhibitors work in BRCA+ cancers. Appear to work in sporadic triple negative. Intrinsic Subtypes: How To Identify? How many known intrinsic subtypes of breast cancer are there? ACCURACY Expression profiling Complex IHC profiles PAM50: Centroid based bioclassifier Available in the clinic EASE ER, PR, HER2 IHC proxies 1. 1 2. 2-3 3. 4-5 4. 10-15 5. As many as Dr. Burstein says there are 50 (not 500) genes Can be used in fixed tissue Parker et al, JCO 2009 Intrinsic Subtypes and Therapy Genomics and Prognostication What we already do: Luminal subtypes receive endocrine therapy HER2-enriched subtype receives HER2-targeted Luminal B (which have high Recurrence Scores) often receive combined chemoendocrine therapy Basal-like (which lacks known targets) receives chemotherapy What we may do in the future: Chemotherapy targeted to subtype? New targeted agents relevant in certain subtypes? 55-year-old healthy woman has a newly diagnosed breast cancer and undergoes breast conservation surgery. Final pathology: grade 2 infiltrating ductal carcinoma, 2.5cm, ER+ (75%), PR+ (30%), HER2- (1+). 5 of 17 lymph nodes contained tumor. She comes to you for adjuvant therapy recommendations 3
How do we counsel her? A More Pessimistic View of Our Counsel Risk of relapse untreated: 80% With optimal endocrine treatment: 40% Over Treatment (60%) Optimal Treatment (15%) Under Treatment (25%) Combined chemoendocrine Rx: 25% Patients who would have been fine with no therapy or endocrine therapy only Chemotherapy made the difference Cancer recurred despite standard treatment Unsupervised vs Supervised Profiling Analyses Basal like HER 2+/ER Luminal B Normal Luminal A? Good prognosis signature Poor prognosis signature Genomics and Prognostication 21 Sorlie T et al, PNAS 2001 Van de Vijver, NEJM 2002 22 48-year-old otherwise healthy woman is diagnosed with a 2.0 cm grade 2, ER+ PR- HER2- node-negative right breast cancer. She comes to you for adjuvant therapy recommendations. Validated genomic tools for prognostication in the nodenegative setting include:? Top-Down Assay Development: The MammaPrint 70-gene Prognosticator Outcome by Gene expression signature 1. Mammaprint 70-gene profile 2. Intrinsic subtype 3. Oncotype Dx Recurrence Score 4. 23andme 5. 1 and 3 6. All 4 Tumors Genes Gene discovery from 2 cohorts relapsed early and not Apply to dataset with known outcome N Engl J Med, Vol 347 (25), Dec. 2002 24 Van t Veer, Nature 2002 76-gene (Rotterdam) signature similar approach but with merging of separate ER+ and ER- gene sets 4
70-Gene Prognosticator Validation 76-Gene Prognosticator Validation Studies 326 node-negative patients < 61 years old, NO adjuvant therapy. HR lower than in initial series 85 of 90 ER-negative tumors had poor signature (limited utility in ER-) Validation #1-180 node-negative pts, no adjuvant therapy Validation study included very few ER- tumors (16) Validation #2-198 node-negative pts < 61 years old, no adjuvant therapy 64 ER- tumors, despite profile design only 14 low risk gene signature Gene signature HR adjusted for clinical risk Gene profile adds to prognosis regardless of clinical risk = independent But modest (~ 2) HR ~5 Gene signature HR adjusted for clinical risk Cumulative events Neither profile can identify good-risk ERdisease Both predict early, not late ER+ relapse 25 High clinical risk = Probability of 10y OS less than Buyse M, JNCI 2006 26 It works.. to predict early relapse Foekens J et al, JCO 06; Desmedt C et al, Clin Cancer Res 07 27... clone MGC:13188 hypothetical protein FLJ23468 (FLJ23468) Consensus includes gb:aa772093 /neuralised (Drosophila)-like chromatin-specific transcription elongation factor, 140 kda subunit (FACTP140) Consensus includes gb:u07802 /DEF=Human Tis11d gene Rho GDP dissociation inhibitor (GDI) β (ARHGDIB) proteasome (prosome, macropain) 26S subunit, ATPase, 2 (PSMC2) hypothetical protein DKFZp434E2220 (DKFZp434E2220) Consensus includes gb:r39094 /KIAA1085 protein Similar to CD44 antigen (homing function and Indian blood group system) Consensus includes gb:al117652.1 /DEF=Homo sapiens mrna solute carrier family 35 (CMP-sialic acid transporter), member 1 (SLC35A1) cyclin E2 (CCNE2) Consensus includes gb:bf055474 / putative zinc finger protein NY-REN-34 Ag Why Don t the Gene Sets Overlap? A Selection of the 76 Genes. It s about pathways, not individual genes Allows us to be ignorant 250 candidate genes (literature) Dependence on HER2- and proliferation genes - why RS not useful in HER2+ 28 Bottom Up Development Using Tailored Gene Lists: The Recurrence Score RT PCR on fixed specimens 3 datasets + 1.04 + 0.47 0.34 PROLIFERATION Ki 67 STK15 Survivin Cyclin B1 MYBL2 INVASION Stromolysin 3 Cathepsin L2 HER2 GRB7 HER2 GSTM1 CD68 BAG1 Tailored to 16 genes ESTROGEN ER PGR Bcl2 SCUBE2 REFERENCE Beta-actin GAPDH RPLPO GUS TFRC Paik et al, SABCS 2003 Validation subsets from prospective NSABP datasets N0, ER+ Tamoxifen treated Paik, NEJM 2004 The Recurrence Score Differentiates Luminal A versus B Relapse in NSABP B14 by Recurrence Score (Node-Negative, HR+, Tamoxifen-Treated) 29 Courtesy C. Perou HER2 and GRB7 ER, SCUBE2, BCL2 Ki-67, STK6, Survivin, Cyclin B1 and MYBL2 NSABP B 14 DRFS 30 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Distant relapse-free survival at10 years: Risk NCCN St. Gallen s RS N DRFS N DRFS N DRFS Low 8 93% 8 95% 51 93% Intermed - - 33 91% 22 86% High 92 85% 59 81% 27 69% Low RS 7% distant mets Int RS 14% distant mets High RS 31% distant mets RS better separation of risk than clinical However you would not use one or other 2 4 6 8 10 12 14 16 Years Paik S et al, NEJM 04, SABCS 04 5
Is Part of the RS Just Improved Clinical Variables? Too Many Prognostic Profiles? Grading by local pathologist Grading by pathologist at central lab 100 Tumor Grade (Site) P<0.001 100 Tumor Grade (Pathologist B) P<0.001 Recurrence Recurrence Score Score 80 60 40 20 0 12% 22% 42% 16% 22% 22% 75% 56% 36% Well Moderate Poor N=77 N=339 N=163 Paik S, et al. J Clin Oncol. 2006;24:3726 3734. Recurrence Score 80 60 40 20 0 5% 12% 83% 12% 24% 64% 61% 19% 19% Well Moderate Poor N=119 N=340 N=190 Amsterdam 70 gene Signature Van de Vijver NEJM 2002 Rotterdam 76 gene signature Wang, Lancet 2005 Especially in 70 gene poor Wound Response Signature Chang, PNAS 2005 Take home: 1. Soon Paik is better than your pathologist 2. Even Soon s grading is often discordant with genomic grade 5% when he calls low, 19% when he calls high Oncotype Recurrence Score 32 Paik, NEJM 2004 Independent of wound response Stem Cell IGS signature Liu, NEJM 2007 Levels of Biomarker Evidence Node-Negative Prognostic Profiles Level I II III IV Type Prospective, high power, designed to test marker (does not need to be randomized) OR Meta-analysis of Level II/III Prospective clinical trial not designed to test marker. Good specimen ascertainment. Large but retrospective studies OR Variable proportion of specimens obtained Small retrospective studies, case-control studies Name Mammaprint (70-gene) 1 76-gene 2 Validation population 302 LNno Rx, > 10y 180 LN- Any Rx, >5y Outcome HR Time to death 2.6 Needs fresh/frozen tissue Distant met-free survival Needs frozen tissue 11.3 V Pilot studies Recurrence Score 3 668 ER+ LN- Rx tam, >10y Distant mets 2.8 Validated in population based sets Reliance on standard gene sets? 33 34 1 Buyse JNCI 06, 2 Wang Lancet 05, Desmedt CCR 07; 3 Paik NEJM 05 Prognostic Profiles Work in HR+ Regardless of Type of Endocrine Therapy Or Nodal Status. Distant Recurrence Increases with the Number of Positive Nodes Regardless of RS TransATAC 1231 of 9366 tested by RS 306 (25%) N+ Tamoxifen vs anastrozole No chemotherapy E2197 (Goldstein, JCO 08) 465 HR+ RS prognostic in N- and N+ All chemoendocrine Rx 70-gene profile (Mook, BCRT 09) 347 pts 1-3 LN+ Heterogeneous tumors and Rx Most ER- had poor profile 70 gene profile HR 7.2 SWOG 8814 (Albain, SABCS 07) 367 tested (of 927 postmenopausal HR+ LN+) CAF-tam vs tam RS prognostic (and predictive of CAF benefit) All (n=1231) Node-negative (n=872) Tamoxifen (n=432) Anastrozole (n= 440) Node-positive (n=306) Tamoxifen (n=152) Anastrozole (n=154) Symbol size proportional to # events 0.1 1.0 4.35 10.0 Adjusted HR (for 50-pt change)* (central grading) Dowsett et al., SABCS 2008, Abstract # 53 9-Years Risk of Distant Recurrence (%) 0 10 20 30 40 50 60 70 80 90 100 RS and clinical variables are significant and independent Good risk N+ (small proportion of n=306) 4+ LN 0 5 10 15 20 25 30 35 40 45 50 Recurrence Score 1-3 LN Node negative Dowsett et al., SABCS 2008, # 53 6
RS-Driven Decision-making and False Dichotomies Molecular Profiles Are Often Concordant Subtype N Recurrence Score 70-gene Wound healing 2-gene Does this number help you? B-14 High RS only: < 1cm ~ 80% 1-4 cm ~ 70% 4+ cm ~ 45% Adding them together didn t prognosticate any better than each alone they are not measuring independent processes Basal-like 53 100% 100% 94% 79% HER2+/ER- 35 100% 91% 100% 80% Luminal B 55 91% 84% 93% 45% Luminal A 123 29% 29% 63% 36% 38 Fan C et al. NEJM 2006 Profile Concordance in Multivariable Setting 48-year-old otherwise healthy woman is diagnosed with a 2.0 cm grade 2, ER+ PR- HER2- node-negative right breast cancer. She comes to you for adjuvant therapy recommendations.? 198 node negative untreated patients 70% concordance among profiles Hazard Ratio Distant Met Free Survival 70-gene 76-gene GGI Age 1.5 1.8 1.7 T size 1.3 1.3 1.2 ER+ 0.8 0.6 0.8 Grade 0.9 1.5 0.8 Genomic signature 7.1 3.4 6.4 Validated genomic tools for prognostication in the nodenegative setting include: 1. Mammaprint 70-gene profile 2. Intrinsic subtype 3. Oncotype Dx Recurrence Score 4. 23andme 5. 1 and 3 6. All 4 Haibe Kains, BMC Genomics 2008 50-year-old woman has a 2.3cm node-positive, ER/PR+, HER2 3+ IDC? Which is the best genomic test to determine which chemotherapy to use? Genomics and Adjuvant Therapy 1. Mammaprint 70-gene profile 2. Oncotype Dx Recurrence Score 3. 76-gene profile 4. MDACC T/FAC profile 5. Intrinsic subtypes 6. None of the Above 41 7
Chemotherapy Advances Benefit ER-Negative More Breast Cancer Relapse is Heterogeneous 43 Improvement in DFS 70 60 50 40 30 20 10 36% 14% 25% 12% 23% 10% 63% 0 CALGB Trial 8541 9344 9741 Overall Optimizing Adding taxane Optimizing anthracycline taxane 32% ER- ER+ Berry DA et al, JAMA 2006 44 Higher risk of early relapse Constant risk of relapse Anderson et al, Breast Cancer Res Treat 2006 Chemotherapy Affects Early Relapse Intrinsic Subtypes and Chemotherapy Sensitivity ER ER+ CALGB 8541: Optimizing anthracycline Pathologic CR to neoadjuvant anthracycline/taxane: T-FAC (N=82) AC-T (n=107) CALGB 9344: Adding taxane Method of assessment Gene expression microarray IHC proxy Luminal A/B 7% 7% Normal-like 0 NA 45 CALGB 9741: Optimizing taxane Berry DA et al, JAMA 06 HER2+/ER- 45% 36% Basal-like 45% 26% Rouzier, CCR 05; Carey, CCR 07 pcr is uncommon in luminal subtypes (regardless of therapy) pcr Relationship to Outcome Survival by pcr in Triple Negative and Non Triple Negative pcr Not pcr ~ 90% 5yr DDFS Lum A pcr do well, regardless of subtype Lum B Basal like pcr has good outcome regardless of subtype ER-negative subtypes worse prognosis due to high risk of residual disease. Non pcr identifies poor prognosis subset pcr is less relevant as a prognosticator in ER+ subtypes 47 * P<0.01 Residual disease only HER2+/ER 48 Non pcr do not do well, especially if triple negative Liedtke, C. et al. J Clin Oncol; 26:1275 1281 2008 8
Straver M, BCRT 2009 High clinical risk = few good profile (even in ER+) Good profile (ER+) = chemoresistant but good outcome How does this help you? 70-Gene And Neoadjuvant Response (Heterogeneous Rx) pcr in poor signature Dose response 50 PROLIFERATION Ki 67 STK15 Survivin Cyclin B1 MYBL2 INVASION Stromolysin 3 Cathepsin L2 Chemotherapy Benefit in ER positive: Role of the Recurrence Score Several of the genes reflect markers already associated with chemosensitivity -0.34 + 1.04 + 0.47 HER2 GRB7 HER2 GSTM1 CD68 BAG1 Paik et al, JCO 2006 ESTROGEN ER PGR Bcl2 SCUBE2 REFERENCE Beta actin GAPDH RPLPO GUS TFRC NSABP B 14 N0, ER+ Recurrence Score Validation Studies Placebo Tamoxifen Low RS <18 CMF/MF Chemotherapy Benefit in Node Negative: NSABP B 20 Absolute benefit of MF or CMF added to tamoxifen n = 353 First report Paik, NEJM 05 Risk of distant relapse despite tamoxifen Int RS 18 30 n = 134 51 NSABP B 20 N0, ER+ Tamoxifen Tamoxifen + MF or CMF Second report Paik, JCO 06 Benefit of chemo added to tam Paik et al. N Engl J Med. 2004;351:2817 2826. High RS 31 52 n = 164 0 10% 20% 30% 40% % Increase in DRFS at 10 Yrs (mean ± SE) Paik et al. J Clin Oncol. 2006. Recurrence Score and Pathologic Response to Anthracycline Plus Taxane CAF Benefit in Node-Positive HR+ Postmenopausal: SWOG 8814 Probability of pcr 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Pathologic complete response to neoadjuvant AT No pcr pcr Low RS (<18) P=0.97 at 10 yrs SWOG 8814: Postmenop. N+, ER+ Tamoxifen Tamoxifen + CAF As in node negative, the benefit of adding CAF to tam was mostly in the high RS group Intermed. RS High RS ( 31 ) 0% 0 10 20 30 40 50 60 70 80 90 100 Recurrence Score P=0.48 at 10 yrs P=0.033 at 10 yrs 53 Gianni L et al, JCO 2005 54 Albain. SABCS. 2007 (abstr 10). 9
Chemotherapy Decision-Making in Node- Positive Using the RS: Uncharted Territories? Recurrence Score Marks General Chemosensitivity What is the risk of recurrence in N+ postmenopausal women with low RS treated with tamoxifen only? Predicts benefit of CMF/MF in node-negative Predicts pcr to AT in locally advanced Predicts benefit of CAF in node-positive SWOG 8814 40% relapse at 10y ATAC 10% relapse at 10y It does NOT tell you which regimen to use. It does NOT identify those that need no adjuvant therapy. Selection matters. S8814 N+ study with chemotherapy randomization Chemotherapy benefit greater in high RS Cannot say no benefit in low RS 10y DFS low RS: 60% tam 64% tam + CAF 95% CI consistent with 40% 75% Reasonable to use RS in certain N+ groups (e.g. micrometastases) but do so with caution. In high clinical risk (e.g. node +) be careful. RS fails to identify a low risk group Existing analyses underpowered to identify chemotherapy benefit in in any stratum. 56 Adjuvant Endocrine Therapy Profiles Several gene sets developed to predict hormonal sensitivity, none yet adequately validated Response to tamoxifen in advanced disease (n=69): Anatomic Factors Optimizing Use of All Factors? Integration into clinical practice Biologic Factors Prospective Clinical Trials TailoRx MINDACT 57 Kok M et al, BCRT 08 58 TAILORx Study Design EORTC BIG MINDACT TRIAL Secondary Study Group 1 RS < 11 ~29% of Population Pre-REGISTER ONCOTYPE DX ASSAY REGISTER Specimen Banking Primary Study Group RS 11-25 ~44% of Population Secondary Study Group 2 RS > 25 ~27% of Population N= 6000 node negative women Evaluate Clinical-Pathological risk and 70-gene signature risk Clinical & 70-gene HIGH / HIGH 55% 32% 13% Discordant Clin-Path HIGH 70-gene LOW 70% N=1344 Clinical and 70-gene LOW / LOW Clin-Path LOW 70-gene HIGH 30% N=576 59 ARM A Hormonal Therapy Alone RANDOMIZE Stratification Factors: Tumor Size, Menopausal Status, Planned Chemo, Planned Radiation ARM B Hormonal Therapy Alone ARM C Chemotherapy Plus Hormonal Therapy ARM D Chemotherapy Plus Hormonal Therapy 60 R1 Use Clin-Path risk to decide Chemo or not N=1920 Use 70-gene risk to decide Chemo or not Potentially spares chemotherapy in 10 15% pts 10
Phases of Biomarker Development TET/FEC predictor (Bonnefoi Lancet Oncol 2007) Oncogene signatures (Bild, Nature 2006) 61 62 Adapted from Andre & Pusztai, Nat Clin Prac Oncol 2006 Genomics and Specific Chemotherapy Decisionmaking Chemotherapy- Specific Response Signatures Are going to be hard to develop. Most patients receive polychemotherapy Cell lines are not cancers Cancers are heterogeneous 4 cell line-based predictors of responsiveness (anthracycline, taxane, cyclophosphamide, fluorouracil) Failed to predict neoadjuvant response in human breast cancers. Sensitivity score 0 should match up with RCB = 0 Sensitivity score 4 with high RCB. Neoadjuvant trials with dedicated research biopsies will be the most likely to bear fruit..support those studies! RCB index = residual cancer burden, measure of pathologic response to therapy Liedtke et al, BCRT 2009 50-year-old woman has a 2.3cm node-positive, ER/PR+, HER2 3+ IDC? Which is the best genomic test to determine which chemotherapy to use? 1. Mammaprint 70-gene profile 2. Oncotype Dx Recurrence Score 3. 76-gene profile 4. MDACC T/FAC profile 5. Intrinsic subtypes 6. None of the Above Summary Genomics opens the window further on the heterogeneity of breast cancer. Intrinsic subtypes give us information about biologic differences across the spectrum of breast cancer. Several profiles are validated for prognosis, they often are measuring the same pathways You cannot replace anatomic considerations with biologic ones. Both are important. Thank you for attending Master Class for Oncologists Much more information about the clinical use of prognostic profiles will be answered by prospective clinical trials. Predictive profiles are still in their infancy wait a few years. 65 11
Many Signatures Will Not Work Across Phenotypes: Genomic Grade Index (GGI) Questions & Answers T/FAC chemotherapy +/- HT N = PPV NPV GGI High predicts pcr to T/FAC 229 40% 88%? ER Positive / HER2 Negative ER Negative / HER2 Negative Liedtke et al. Manuscript in press Cohort Used To Identify A Response Predictor For T/FAC Chemotherapy HER2-Negative Breast Cancer T/FAC Cases N=274 Predictive Profiles for T/FAC Response (pcr/rcb-i versus other) Cross-Validation N = PPV NPV Overall 227 58% 87% Non-Usable Cases N=14 T/FAC Cases Available N=260 pcr: 21.5% RCB 0/I: 32.3% RCB III: 24.2% ER+ Subset (IHC) 130 66% 90% ER- Subset (IHC) 97 78% 77% Hatzis et al ASCO 2008 pcr: 39.4% RCB 0/I: 45.5% RCB III: 15.2% HER2 Positive N=33 HER2 Normal N=227 pcr: 18.9% RCB 0/I: 30.4% RCB III: 25.6% pcr: 6.1% RCB 0/I: 16.2% RCB III: 27.7% ER Positive N=130 ER Negative N=97 pcr: 36.1% RCB 0/I: 49.5% RCB III: 22.7% Courtesy WF Symmans NEEDS TO BE VALIDATED! 12