Gene signature for risk stratification and treatment of breast cancer: Incorporating tumor biology in clinical decision-making Drukker, Caroline

Transcriptie

1 UvA-DARE (Digital Academic Repository) Gene signature for risk stratification and treatment of breast cancer: Incorporating tumor biology in clinical decision-making Drukker, Caroline Link to publication Citation for published version (APA): Drukker, C. A. (2014). Gene signature for risk stratification and treatment of breast cancer: Incorporating tumor biology in clinical decision-making General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam ( Download date: 01 Jan 2018

2 Gene signature for risk stratification and treatment of breast cancer Incorporating tumor biology in clinical decision-making Caroline Drukker

3 Gene signature for risk stratification and treatment of breast cancer Incorporating tumor biology in clinical decision-making Caroline Drukker

4 The work described in this thesis was performed at the Netherlands Cancer Institute, Amsterdam, the Netherlands. In cooperation with the European Organization for Research and Treatment of Cancer (EORTC), Brussels, Belgium and the Dutch breast cancer screening facilities (BOB), Utrecht, the Netherlands. The research was funded by grants from BBMRI-NL and the EORTC. Unrestricted financial support for publication of this thesis was provided by: Netherlands Cancer Institute, Academic Medical Center, Roche, Sanofi-Aventis, Agendia NV, Chipsoft, and Amgen. Layout: Gildeprint - Enschede Printed by: Gildeprint - Enschede ISBN: Online: Caroline Drukker, Amsterdam, the Netherlands

5 Gene signature for risk stratification and treatment of breast cancer Incorporating tumor biology in clinical decision-making ACADEMISCH PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. D.C. van den Boom ten overstaan van een door het college voor promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op vrijdag 28 maart 2014, te uur door Caroline Anne-Marie Drukker geboren te Amsterdam

6 Promotiecommissie Promotores: Prof. dr. E.J.T. Rutgers Prof. dr. L.J. van t Veer Co-promotores: Prof. dr. S.C. Linn Dr. M.K. Schmidt Overige leden: Prof. dr. S. Rodenhuis Prof. dr. ir. F.E. van Leeuwen Prof. dr. G.J. den Heeten Prof. dr. ir. J.J.M. van der Hoeven Prof. dr. R. Versteeg Prof. dr. J.H.G. Klinkenbijl Faculteit der Geneeskunde

7 Table of contents Chapter 1 Introduction and outline 7 Chapter 2 Voorspelling van de prognose van patiënten met vroeg stadium 19 borstkanker: de bijdrage van een genexpressie-profiel. Chapter 3 A prospective evaluation of a breast cancer prognosis signature 33 in the observational RASTER study. Chapter 4 Optimized outcome prediction in breast cancer by combining 53 the 70-gene signature with clinical risk prediction algorithms. Chapter 5 Risk estimations and treatment decisions in 69 early stage breast cancer: agreement among oncologists and the impact of the 70-gene signature. Chapter 6 Long-term impact of the 70-gene signature 87 on breast cancer outcome. Chapter 7 Mammographic screening detects low risk 99 tumor biology breast cancers. Chapter 8 Gene-expression profiling to predict 119 the risk of locoregional recurrence in breast cancer. Chapter 9 General discussion and future prospects 139 Chapter 10 Summary 155 Samenvatting 161 PhD portfolio 167 Acknowledgements (Dankwoord) 177 Curriculum Vitae 185

8 Chapter 1

9 Introduction and outline

10 8 Chapter 1

11 Breast cancer Breast cancer is the most frequently diagnosed cancer among women worldwide. 1 Over the past two decades, the incidence rate in the Netherlands increased from 56 per women diagnosed with invasive breast cancer in 1991 to 83 per in This increasing incidence may (partly) be explained by the introduction of population-based screening programs in 1990, which resulted in an increase in the detection of early stage breast cancer after full coverage was achieved in Another important observation is a decrease in breast cancer mortality-rates, from 45 per women in 1991 to 38 per women in , which may be explained by early detection due to the implementation of screening programs as well as the improvement and more extensive use of adjuvant systemic treatment. 3,4 1 Adjuvant systemic therapy After primary treatment consisting of surgery with or without radiotherapy to achieve locoregional control most breast cancer patients are nowadays systemically treated in the adjuvant or neoadjuvant setting. Adjuvant systemic therapy (AST), including endocrine therapy, chemotherapy and/or trastuzumab, is used to control micrometastatic disease and improve long-term outcome. 5 Data from the Early Breast Cancer Trialists Collaborative Group (EBCTCG) confirmed the survival benefit of AST by showing a significant better disease-free and overall survival for patients treated with chemotherapy and/or endocrine therapy in different subgroups. 4 Guidelines in breast cancer treatment The selection of those patients at a high risk of recurrence who are most likely to benefit from AST has traditionally been based on clinicopathological factors such as age, tumor size, grade, estrogen-receptor status (ER), progesterone-receptor status (PR), Human Epidermal growth factor Receptor-2 (HER2) and the status of the axillairy lymph nodes. 6 There are multiple breast cancer guidelines and clinical tools that use these clinicopathological factors to estimate the risk of recurrence and provide a related recommendation for AST. 7-9 The components used in these guidelines are all very similar, but show slight differences in their definitions of high and low risk. Most guidelines only identify a small group of patients who are at a low risk of recurrence and for whom AST is of limited value. Consequently, a majority of patients are classified as high risk and therefore become eligible for AST. This will result in a substantial number of patients being treated with AST while they are unlikely to derive significant benefit from it. The differences in the definitions of low risk used by established guidelines create a non-overlapping group of patients at a low or high risk, indicating a suboptimal predictive accuracy for the individual patient. 7,10,11 For example, the online decision-making tool Adjuvant! Online uses age, tumor size, grade, ER and nodal status to estimate the 10-year survival probability of a given patient and the possible survival-benefit that can be derived from adjuvant endocrine therapy and chemotherapy, while the Nottingham Prognostic Index (NPI) only provides a high or low risk estimation based on a Introduction and outline 9

12 score which is calculated using only tumor size, grade and nodal status. Detailed information on prognostic factors used by established clinicopathological guidelines is summarized in Table 1. Even when using extensively validated clinicopathological factors, predicting the risk of recurrence for the individual patient remains challenging. Already for a long time, pathologists, clinicians and researchers are aware that breast cancer is a heterogeneous disease. Morphology, receptor expression and molecular subtypes all contribute to the clinical course of breast cancer in the individual patient. Variations in clinical behaviour and outcome have been described for several decades. 12 Guidelines and clinical tools have improved over the past years and are now including clinicopathological factors such as HER2 status and Ki67. 13,14 Nevertheless, HER2 and Ki67 only account for a small part of this heterogeneity and still most guidelines do not adjust for the heterogeneity entirely. Therefore, clinicopathological guidelines have only limited ability to predict individual patient outcomes. 15 Insight in the biology of breast cancer: introducing gene signatures Over the past decades, researchers identified many single genes involved in the proliferation and metastatic capacity of breast cancer. Breast cancer progression is a result however of multiple genetic aberrations, and thus one gene will never be responsible for the entire cancer process. 16,17 Therefore, researchers were looking for methods to evaluate the relationships among and within different cellular pathways. The introduction of micro-array analyses provided a way to evaluate multiple genes in multiple pathways at once in a more robust manner. Micro-array technology is used to develop gene signatures that are related to the metastatic potential of an individual breast cancer. These signatures can refine risk estimations based on standard clinicopathological guidelines. 18 One of these signatures is the 70-gene signature (MammaPrint ), which was developed by van t Veer and colleagues at the Netherlands Cancer Institute (NKI) in Amsterdam, the Netherlands. The 70-gene signature measures the level of expression of a set of genes by semi-quantitatively determining the level of messenger RNA (mrna) transcripts. 19 The intensity of the nucleic acids that hybridize to the individual gene probes are commonly shown in a two-color array. 19 Green reflects low expression and red reflects high expression of that gene in the tumor (Figure 1). After hybridization the slides are scanned with a dual laser scanner (Agilent Technologies) and the data is processed using a specific algorithm providing an index-score which originally ranges from 0 to The 70-gene signature was developed using frozen tumor samples from 78 patients who were diagnosed at the NKI with lymph node-negative breast cancer and who were up to 55 years of age at the time of diagnosis. 44 of these 78 patients remained free of distant metastases for at least 5 years. These patients were defined as good prognosis or low risk. 10 Chapter 1

13 Table 1. Clinicopathological factors used by breast cancer guidelines to estimate the risk of recurrence Guideline Age Size Grade Hist. type ER/PR HER2 Ki67 Nodal status Other factors Low risk is defined as Not specified. ER - - Yes Co-morbidities, CT regimen AOL Yes Yes Yes Ductal, in case of other hist. type, information is available online Luminal A; ER/PR +, HER2 -, low Ki67 Biological subtype - Yes - ER/PR Yes Yes Yes, more than 3+ nodes is high risk Pre- or post menopausal St. Gallen expert panel 2011 NPI - Yes Yes Yes - [0.2 x Size] + Number of nodes + Grade; low risk = score < 3.4 <35 or 35 Yes Yes - - Yes - Yes - 10-years survival probability 85%. N0, <35, grade I tumor 1 cm OR 35 yrs, grade I tumor 2 cm. NABON 2012 Not specified. Suggested: <3% survival benefit in 10-years no chemotherapy; 3-5% chemotherapy discussed as possible option Yes Yes Yes - ER Yes Yes Yes Method of detection, CT regimen PREDICT plus AOL = Adjuvant! Online; NPI=Nottingham Prognostic Index; NABON=Nationaal Borstkanker Overleg Nederland; ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal Growth factor Receptor 2; CT=chemotherapy. 1 Introduction and outline 11

14 Figure 1. Micro-array technology used for the 70-gene signature The remaining 34 patients developed distant metastases within 5 years after diagnosis and were defined as poor prognosis or high risk. 12 Tumors with an index-score >0.4 are classified as 70- gene signature low risk and tumors with an index-score <0.4 as 70-gene signature high risk. The signature was validated by van de Vijver et al. in a consecutive series of 151 lymph node-negative and 144 lymph node-positive breast cancer patients, diagnosed at the NKI, aged up to 53 years at the time of diagnosis. 21 Buyse et al. performed an independent validation in 302 lymph nodenegative patients from 5 European hospitals, aged up to 60 years at the time of diagnosis. 22 The prognostic value of the 70-gene signature has also been retrospectively confirmed in several patient subgroups, such as postmenopausal patients, patients with positive axillary lymph nodes and in case of HER2-positive disease Aside from the 70-gene signature, a few other gene signatures have found their way to the clinic. The characteristics of these tests, including PAM 50, Oncotype Dx, EndoPredict, Breast Cancer Index and MapQuant Dx, are described in Table 2. The analyses presented in this thesis focus on the 70-gene signature. Using the 70-gene signature in the daily clinical practice To prospectively evaluate the feasibility of implementation of the 70-gene signature in the community-based setting, the MicroarRAy PrognoSTics in Breast CancER (RASTER) study was conducted. 29 Between 2004 and eligible patients were included in 16 hospitals in the Netherlands. Implementation of the 70-gene signature appeared feasible, even though the test could only be performed on fresh frozen tumor samples at the time. Recently, also formalin fixed paraffin embedded (FFPE) tumor samples can be used to perform the 70-gene signature Chapter 1

15 Table 2. Characteristics of gene signatures currently available MammaPrint Pam50 Oncotype Dx EndoPredict Breast Cancer Index MapQuant Dx 97-gene (micro-array) or 8-gene (qrt-pcr) signature 11-gene signature 2-gene ratio HOXB13 and IL17R and molecular grade index (MGI) Assay 70-gene signature 55-gene signature 21-gene recurrence score Technique Micro-array qrt-pcr qrt-pcr qrt-pcr qrt-pcr Micro-array/ qrt-pcr Ipsogen (Marseille, FR) Biotheranostics (San Diego, USA) Sividon Diagnostics (Keulen, DU) Genomic Health (Redwood City, USA) ARUP Laboratories (Salt Lake City, USA) Provided by Agendia (Amsterdam, NL) Tissue sample Frozen or FFPE Frozen or FFPE Frozen or FFPE FFPE FFPE Frozen or FFPE 64 pt, ER+ 964 pt, ER+, HER2-588 pt, ER+, N0, treated with tamoxifen for 2-gene ratio. 410 pt ER+, N0 for MGI 447 pt, ER+ from NSABP B-20 study (tamoxifen-treated arm) 189 pt, ER+/-, HER2 +/-, T1-2N0-1 Training set 78 pt, T1-2N0, <55 yrs ER +/- 597 pt, ER+ 265 pt, ER+, N0, treated with tamoxifen 1702 pt, ER+, HER2-, treated with tamoxifen (2 series) 668 pt, ER+ from NSABP B-14 study (treated with tamoxifen) 761 pt for prognosis 133 pt for prediction ER+/-, HER2 +/-, T1-2N0-1 Validation set 295 pt, T1-2N0-1, <53 yrs, ER +/- Low grade GGI or high grade GGI Continuous variable divided in 3 groups; High, Intermediate, Low risk Continuous variable divided in 2 groups; High and Low risk (EP score) Output High and Low risk Continuous variable Continuous variable divided in 3 groups; High, Intermediate, Low risk Molecular grading of ER+, grade 2 tumors Prognosis prediction of ER+, N0 pt treated with tamoxifen Prognosis prediction of ER+, HER2- pt treated with tamoxifen EP clin score by combining EP score with clinicopathological factors Prognosis prediction of ER+, N0 pt treated with tamoxifen Prognosis prediction of N0 pt, ER+, treated with endocrine therapy Prognosis prediction of T1-2N0 pt, ER+/-,<61 yrs Initially developed for xx xx xx mrna levels ER, PR and HER2 mrna levels ER, PR and HER2 (TargetPrint), intrinsic subtypes (BluePrint) Additional information ER+ pt, treated with aromatase-inhibitors Possibly also prognostic for late recurrences ER+ tumor in postmenopausal women xx ER+ and patients with up to 3 pos. axillary lymph nodes Postmenopausal ER+ pt, treated with aromatase-inhibitors Patients with up to 3 pos. axillary lymph nodes, patients who are yrs, and for HER2+ disease Prognostic value in other subgroups FFPE=formalin fixed parafin embedded; pt=patient; ER=estrogen receptor; HER2=Human Epidermal Growth Factor receptor 2 1 Introduction and outline 13

16 Shortly after confirmation of its feasibility in the RASTER study, the 70-gene signature was subjected to an international, multicenter, randomized-controlled trial called Microarray in Node-negative and 1-3 lymph node positive Disease may Avoid Chemotherapy (MINDACT). Patients enrolled in the MINDACT trial had their risk of recurrence assessed by the known online decision making tool Adjuvant! Online and the 70-gene signature. In case of a concordant low risk estimation patients would only receive endocrine therapy, while in case of a concordant high risk estimation patients would receive adjuvant chemotherapy with or without endocrine therapy. If the clinicopathological risk estimation was discordant with the 70-gene signature risk estimation patients were randomized between treatment according to the risk estimation by the 70-gene signature or treatment according to the clinicopathological risk estimation. On July 1 st 2011, the required 6673 patients were successfully enrolled in the trial. The MINDACT trial will evaluate whether adjuvant chemotherapy can safely be omitted in patients with a tumor that is low risk according to the 70-gene signature, while clinical guidelines (in this case Adjuvant! Online) assessed this tumor as high risk. Meanwhile, better prognostication is desired in routine clinical practice and for this reason the 70-gene signature is increasingly applied when there is uncertainty regarding the indication of AST. Several studies from the Netherlands Cancer Institute have shown the impact of the introduction of the 70-gene signature on the quality of life of patients and the cost-effectiveness of genomic testing was confirmed multiple times On the other hand, the effect on clinical decision-making had not systematically been studied. Rationale and outline of this thesis The aim of this thesis is to evaluate outcome prediction and clinical relevance of the 70-gene signature for locoregional and distant recurrence, its influence on risk assessment and AST recommendations, and its additional value to established clinical guidelines used in breast cancer treatment. In addition, we used the 70-gene signature to gain better insight in the biological background of tumors detected in a population-based screening program. The first part of this thesis focuses on the current applicability of the 70-gene signature in daily clinical practice and the impact of the 70-gene signature on clinical decision-making. Chapter 2 of this thesis provides a current overview of the prognostic value of the 70-gene signature in different subgroups of patients as described in recently published, retrospective studies. Chapter 3 provides the first prospective evidence of the prognostic value of the 70-gene signature. The 5-year follow-up data of the RASTER study shows the outcome of patients for whom the 70- gene signature was used to decide whether or not an individual patient should receive adjuvant systemic treatment. In this chapter the 70-gene signature is compared to Adjuvant! Online. Because Adjuvant! Online is the most commonly used, but not the only guideline in breast cancer, we also compared the additional value of the 70-gene signature to other established guidelines in chapter 4. As described earlier, clinicopathological guidelines vary in their risk estimations. At 14 Chapter 1

17 this point in time, there is no data on the agreement among oncologists using clinicopathological factors for risk estimations and the impact of the 70-gene signature on clinical decision-making. Therefore, agreement among oncologists before and after providing the 70-gene signature result was evaluated in chapter 5. Also, we aimed to evaluate long-term outcome of patients for whom a 70-gene signature result was available. Therefore we updated the original consecutive series as published by van de Vijver et al. in 2002 (chapter 6). 1 The second part of this thesis focuses on new areas where the 70-gene signature may improve the biological understanding of breast cancer. Method of detection has proven to be an independent prognostic factor in breast cancer. Patients with a screen-detected cancer have more favorable outcome, independent of known clinicopathological factors such as age, size and ER-status. To investigate whether this observation is supported by a more favorable tumor biology in screen-detected cancers, we described the proportions of high, low and ultralow risk according to the 70-gene signature among screen-detected and interval cancers in the Dutch MINDACT cohort in chapter 7. Since a transition from film-screen mammography (FSM) to full field digital mammography (FFDM) took place at the same time as the MINDACT trial was conducted in the Netherlands we were also able to evaluate the impact of this transition on the biological background of the tumors detected in the nation-wide screening program. The 70-gene signature was developed to predict the risk of distant recurrence in breast cancer. Because of the correlation between distant and locoregional recurrence, we hypothesized that the 70-gene signature would also be able to predict the risk of locoregional recurrence after both breast conserving surgery and mastectomy. The results of analyzing this hypothesis in a pooled dataset of all patients included in one of the 70-gene signature validation studies, who were diagnosed and treated at the Netherlands Cancer Institute, is described in chapter 8.* This thesis ends with a general discussion and future prospects in chapter 9 and a summary of all results is presented in chapter 10. *All studies described in this thesis are performed in accordance with the FEDERA codes of conduct. 34 Introduction and outline 15

18 References 1 Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN Int J Cancer 2010; 127: Nederlandse Kankerregistratie beheerd door IKNL 3 Esserman LJ, Shieh Y, Rutgers EJ, Knauer M, Retel VP, Mook S et al. Impact of mammographic screening on the detection of good and poor prognosis breast cancers. Breast Cancer Res Treat 2011; 130: Early Breast Cancer Trialists Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 365: Sotiriou C, Piccart MJ. Taking gene-expression profiling to the clinic: when will molecular signatures become relevant to patient care? Nat Rev Cancer 2007; 7: Cleator S, Ashworth A. Molecular profiling of breast cancer: clinical implications. Br J Cancer 2004; 90: D Eredita G, Giardina C, Martellotta M, Natale T, Ferrarese F. Prognostic factors in breast cancer: the predictive value of the Nottingham Prognostic Index in patients with a long-term follow-up that were treated in a single institution. Eur J Cancer 2001; 37: Goldhirsch A, Glick JH, Gelber RD, Coates AS, Senn HJ. Meeting highlights: International Consensus Panel on the Treatment of Primary Breast Cancer. Seventh International Conference on Adjuvant Therapy of Primary Breast Cancer. J Clin Oncol 2001; 19: Ravdin PM, Siminoff LA, Davis GJ, Mercer MB, Hewlett J, Gerson N et al. Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 2001; 19: Goldhirsch A, Wood WC, Gelber RD, Coates AS, Thurlimann B, Senn HJ. Meeting highlights: updated international expert consensus on the primary therapy of early breast cancer. J Clin Oncol 2003; 21: Olivotto IA, Bajdik CD, Ravdin PM, Speers CH, Coldman AJ, Norris BD et al. Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23: van t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002; 415: Integraal Kankercentrum Nederland. NABON richtlijn mammacarcinoom Wishart GC, Bajdik CD, Dicks E, Provenzano E, Schmidt MK, Sherman M et al. PREDICT Plus: development and validation of a prognostic model for early breast cancer that includes HER2. Br J Cancer 2012; 107: Cardoso F. Microarray technology and its effect on breast cancer (re)classification and prediction of outcome. Breast Cancer Res 2003; 5: Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; 100: Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011; 144: Ross JS, Hatzis C, Symmans WF, Pusztai L, Hortobagyi GN. Commercialized multigene predictors of clinical outcome for breast cancer. Oncologist 2008; 13: Harris JR, Lippman ME, Osborne CK, Morrow M. Diseases of the Breast. Fourth ed. Lippincott Williams & Wilkins, Glas AM, Floore A, Delahaye LJ, Witteveen AT, Pover RC, Bakx N et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006; 7: Chapter 1

19 21 van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas AM et al. Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 2006; 98: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009; 117: Knauer M, Cardoso F, Wesseling J, Bedard PL, Linn SC, Rutgers EJ et al. Identification of a low-risk subgroup of HER-2-positive breast cancer by the 70-gene prognosis signature. Br J Cancer 2010; 103: Mook S, Schmidt MK, Viale G, Pruneri G, Eekhout I, Floore A et al. The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1-3 positive lymph nodes in an independent validation study. Breast Cancer Res Treat 2009; 116: Mook S, Knauer M, Bueno-de-Mesquita JM, Retel VP, Wesseling J, Linn SC et al. Metastatic potential of T1 breast cancer can be predicted by the 70-gene MammaPrint signature. Ann Surg Oncol 2010; 17: Mook S, Schmidt MK, Weigelt B, Kreike B, Eekhout I, van de Vijver MJ et al. The 70-gene prognosis signature predicts early metastasis in breast cancer patients between 55 and 70 years of age. Ann Oncol 2010; 21: Wittner BS, Sgroi DC, Ryan PD, Bruinsma TJ, Glas AM, Male A et al. Analysis of the MammaPrint breast cancer assay in a predominantly postmenopausal cohort. Clin Cancer Res 2008; 14: Bueno-de-Mesquita JM, van Harten WH, Retel VP, van t Veer LJ, van Dam FS, Karsenberg K et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007; 8: Sapino A, Roepman P, Linn SC, Snel M.H.J., Delahaye LJ, van den Akker J. et al. MammaPrint molecular diagnostics on Formalin Fixed Paraffin Embedded tissue. J Mol Diagn. In press. 31 Retel VP, Joore MA, Knauer M, Linn SC, Hauptmann M, Harten WH. Cost-effectiveness of the 70- gene signature versus St. Gallen guidelines and Adjuvant Online for early breast cancer. Eur J Cancer 2010; 46: Retel VP, Joore MA, Drukker CA, Bueno-de-Mesquita JM, Knauer M, van Tinteren H et al. Prospective cost-effectiveness analysis of genomic profiling in breast cancer. Eur J Cancer Retel VP, Groothuis-Oudshoorn CG, Aaronson NK, Brewer NT, Rutgers EJ, van Harten WH. Association between genomic recurrence risk and well-being among breast cancer patients. BMC Cancer 2013; 13: Dutch federation of Biomedical Scientific Societies. Codes of Conduct. 1 Introduction and outline 17

20 Chapter 2

21 Voorspelling van de prognose van patiënten met vroeg stadium borstkanker: de bijdrage van een genexpressie-profiel Accepted by Nederlands Tijdschrift voor Geneeskunde Caroline A. Drukker Marjanka K. Schmidt Thijs van Dalen Jacobus J.M. van der Hoeven Sabine C. Linn Emiel J.Th. Rutgers

22 Abstract Gene expression classifiers, such as the 70-gene signature, reflecting the biology of breast tumors, start finding their way into daily clinical practice. Retrospective validation studies in breast cancer have established the prognostic value of the 70-gene signature (MammaPrint ). The prospective observational RASTER study shows excellent 5-year distant-recurrencefree intervals in all subgroups. Patients with 70-gene signature low risk who had not received adjuvant chemotherapy despite poor clinicopathological factors, had an excellent 5-years distant-recurrence-free interval of 98.4%. Especially for patients aged 45 years or older, with an estrogen receptor (ER)-positive, HER2-negative tumor, diameter 1-2 cm, grade 2 there is prospective evidence that the 70-gene signature has additional value for adjuvant chemotherapy decisions. Samenvatting Genexpressie-profielen, zoals het 70-genen profiel, die informatie geven over het biologisch gedrag van borstkanker, hebben hun intrede gedaan in de dagelijkse klinische praktijk. Meerdere retrospectieve validatiestudies hebben de prognostische waarde van het 70-genen profiel (MammaPrint ) aangetoond. De prospectieve, observationele RASTER studie laat een uitstekende 5-jaars ziektevrije overleving zien voor klinisch hoog risico, maar 70-genen profiel laag risico patiënten, die geen adjuvante chemotherapie hadden gehad: een 5-jaars metastase-vrij interval van 98.4%. Met name voor patiënten vanaf 45 jaar met een oestrogeen receptor (ER)-positieve, HER2-negatieve tumor, 1-2 cm, graad 2 is er nu ook prospectief bewijs dat het 70-genen profiel kan bijdragen in de besluitvorming om al dan niet adjuvant chemotherapie te adviseren. 20 Chapter 2

23 Inleiding De daling in de mortaliteit van het mammacarcinoom in de afgelopen decennia wordt toegeschreven aan vroegere detectie door invoering van landelijke screeningsprogramma s, toegenomen bewustzijn bij patiënten waardoor sneller medische hulp wordt gezocht, maar vooral door verbetering en frequenter gebruik van adjuvant systemische behandelmogelijkheden. 1,2 De selectie van patiënten die in aanmerking komen voor adjuvante chemotherapie (ACT) wordt gebaseerd op klinisch-pathologische factoren, zoals leeftijd, tumorgrootte, lymfklierstatus, histologische graad, oestrogeen-receptor (ER) en Human Epidermal growth factor Receptor 2 (HER2) status. 2 Aan de hand van deze factoren wordt een inschatting gemaakt van het risico op het ontwikkelen van recidief ziekte, op basis waarvan richtlijnen de indicatie voor ACT bepalen. De huidige NABON richtlijn (2012) geeft aan dat ACT alleen gerechtvaardigd is als er een absoluut overlevingsvoordeel kan worden behaald van meer dan 5% in de eerste 10 jaar. 3 Voor de individuele patiënt blijft het echter moeilijk om een accurate risico-inschatting te maken, omdat patiënten met dezelfde klinisch-pathologische factoren een verschillend ziektebeloop kunnen hebben. 4 Daardoor krijgen veel vrouwen ACT, terwijl ze er waarschijnlijk geen baat bij hebben. Een meer op het individu gerichte risico-inschatting en behandeladvies kan zowel overbehandeling, met de bijbehorende (soms) ernstige toxiciteit, als onderbehandeling voorkomen. 4 2 Genexpressie-profielen Een mogelijkheid om de nauwkeurigheid van de risico-inschatting en het daaraan gekoppelde behandeladvies te verbeteren is het gebruik van genexpressie-profielen. 4 Bekende genexpressieprofielen zijn: het 70-genen profiel, het 76-genen profiel, PAM 50, MapQuant Dx, EndoPredict, de Breast Cancer Index en het 21-genen profiel van Oncotype Dx. Informatie over groepen patiënten waarvoor deze testen voor wat betreft hun prognostische waarde zijn gevalideerd is te vinden in tabel 1. Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 21

24 Tabel 1. Overzicht van bekende genexpressie-profielen en hun kenmerken MammaPrint Rotterdam signature Pam50 Oncotype Dx EndoPredict Breast Cancer Index MapQuant Dx Assay 70-genen profiel 76-genen profiel 55-genen profiel 21-genen recurrence score 11-genen profiel 2-genen ratio HOXB13 en IL17R en moleculaire graad index (MGI) 97-genen (micro-array) of 8-genen (qrt-pcr) profiel Analyse Micro-array Micro-array qrt-pcr qrt-pcr qrt-pcr qrt-pcr Micro-array/ qrt-pcr Op de markt gebracht door Agendia (Amsterdam, NL) Niet verkrijgbaar ARUP Laboratories (Salt Lake City, USA) Genomic Health (Redwood City, USA) Sividon Diagnostics (Keulen, DE) Biotheranostics (San Diego, USA) Ipsogen (Marseille, FR) Weefsel preservatie Vries of paraffine Vries Vries of paraffine Vries of paraffine Paraffine Paraffine Vries of paraffine 64 pt, ER+ Cohort waarop test is ontwikkeld 78 pt, T1-2N0, <55 jr ER +/- Validatie-cohort 295 pt, T1-2N0-1, <53 jr, ER +/- Weergave van uitslag 115 pt, ER+, N0 189 pt, ER+/-, HER2 +/-, T1-2N pt, ER+/-, N0 761 pt voor prognose 133 pt voor predictie ER+/-, HER2 +/-, T1-2N pt, ER+ uit NSABP B-20 studie (uit arm die alleen tamoxifen kreeg) 668 pt, ER+ uit NSABP B-14 studie (behandeld met tamoxifen) Hoog/laag risico Hoog/laag risico Continue variabele Continue variabele verdeeld in 3 groepen; hoog, intermediair en laag risico 964 pt, ER+, HER2-588 pt, ER+, N0, behandeld met tamoxifen voor 2-genen ratio. 410 pt ER+, N0 voor MGI 1702 pt, ER+, HER2-, behandeld met tamoxifen (2 series) Continue variabele verdeeld in 2 groepen; hoog en laag risico (EP score) 265 pt, ER+, N0, behandeld met tamoxifen Continue variabele verdeeld in 3 groepen; hoog, intermediair en laag risico 597 pt, ER+ Genomic Grade Index (GGI) laag of GGI hoog-gradig Initiële klinische doelgroep Additionele informatie Prognostische waarde in andere subgroepen Studies prognostische waarde Studies predictieve waarde Prognose van T1-2N0 pt, ER+/-,<61 jr mrna levels ER, PR en HER2 (TargetPrint), intrinsieke subtypes (BluePrint) Tot 3 pos. lymfklieren, jr, HER2+ van de Vijver et al. (NEJM 2002) Buyse et al. (JNCI 2006) Bueno-de-Mesquita et al. (BCRT 2008) MINDACT studie Knauer et al. (BCRT 2010) MINDACT studie Prognose van N0 pt Prognose van N0 pt, ER+, behandeld met hormonale therapie Prognose van ER+, N0 pt behandeld met tamoxifen n.v.t. n.v.t. mrna levels ER, PR en HER2 ER+, N0 pt behandeld met tamoxifen Foekens et al. (JCO 2006) Desmedt et al. (Clin Cancer Res 2007) n.v.t. ER+ en 1-3 pos. lymfklieren. Postmenop. ER+ pt, behandeld met aromatase-remmers Parker et al. (JCO 2009) Nielsen et al. (Clin Cancer Res 2010) n.v.t. Martin et al. (BCRT 2013) Paik et al. (NEJM 2004) Paik et al. (JCO 2006) Paik et al. (JCO 2006) Prognose van ER+, HER2- pt behandeld met tamoxifen EP clin score wordt verkregen door combinatie EP score met klin. path. factoren ER+ tumoren in postmenopauzale vrouwen Filipits et al. (Clin Cancer Res 2011) Dubsky et al. (Ann of Oncol 2013) Prognose van ER+, N0 pt behandeld met tamoxifen n.v.t. n.v.t. Mogelijk ook voorspellend voor late metastasen Zhang et al (Clin Cancer Res 2013) n.v.t. n.v.t. n.v.t. Moleculaire gradering van ER+, graad 2 pt ER+ pt behandeld met aromatase-remmers Sotiriou et al. (JNCI 2006) Reyal et al. (PlosOne 2012) qrt-pcr= quantitative reverse transcrip!ase polymerase chain reaction pt=patiënten 22 Chapter 2

25 Het 70-genen profiel (MammaPrint ) bepaalt de activiteit van de geselecteerde genen door de hoeveelheid messenger RNA (mrna) te meten. Met behulp van een algoritme wordt de 70-genen profiel indexscore berekend, gelegen tussen -1 en 1. 5 Hoog risico is gedefinieerd als een indexscore <0.4, laag risico als een indexscore >0.4. Figuur 1 laat zien hoe een 70-genen profiel testuitslag wordt weergegeven. De prognostische waarde van het profiel is uitgebreid gevalideerd in meerdere retrospectieve studies. 5-7 De test kon aanvankelijk alleen op ingevroren tumorweefsel worden uitgevoerd, maar nu ook op formaline gefixeerd paraffine ingebed materiaal (FFPE). 8 De Amerikaanse Food and Drug Administration (FDA) heeft het toepassingsgebied van het 70-genen profiel als prognostische marker goedgekeurd voor: vrouwen met een stadium 1 of 2 mammacarcinoom, <5 cm en een negatieve axillaire lymfklierstatus. 9 Prospectieve evaluatie van het al dan niet voorschrijven van chemotherapie op basis van het 70-genen profiel (de predictieve waarde) wordt momenteel onderzocht in de multicentrische, gerandomiseerde MINDACT studie, waarvan de inclusie per juli 2011 gesloten is. 10 Er zijn 6694 patiënten geincludeerd in de MINDACT, waarvan 2092 in 21 Nederlandse ziekenhuizen. In de studie is er, in geval van discordantie tussen de klinische risico-inschatting op basis van Adjuvant! Online en het 70-genen profiel, gerandomiseerd tussen behandeling conform Adjuvant! Online of conform het 70-genen profiel. 10 Het 70-genen profiel wordt in Nederland veel toegepast en door het merendeel van de Nederlandse zorgverzekeraars vergoed. In andere landen wordt het 21-genen profiel ook veel gebruikt. 2 Het 21-genen profiel van Oncotype Dx kent, naast een hoog en laag risico, ook een intermediaire risico-uitslag. De 10-jaars metastase-vrije-overleving van de laag, intermediair en hoog risico groep waren in het cohort waarop de test ontwikkeld is respectievelijk 93.2% (CI: ), 85.7% (CI: ) en 69.5% (CI: ). 11 De test is gevalideerd in 668 lymfklier-negatieve, ERpositieve patiënten die in de NSABP-B14 studie met tamoxifen waren behandeld. 12 Het 21-genen profiel is niet getest bij ER-negatieve tumoren, noch bij onbehandelde patiënten. In Nederland is de afgelopen jaren vooral ervaring met het 70-genen profiel opgedaan. Hieronder geven wij een overzicht van een aantal patiëntengroepen voor wie het 70-genen profiel kan bijdragen aan de besluitvorming tot wel of geen aanvullende chemotherapie zoals gerapporteerd in de recente literatuur. Gepubliceerde studies zijn gecategoriseerd aan de hand van mate van bewijs zoals voorgesteld door Simon et al (tabel 2). 13 Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 23

26 Amended R eport decoding breast cancer. CUSTOMER SPECIMEN PATIENT Doctor: Requisition #: Patient: Account: Collection Date: 19-Aug-2013 Test Request Date: 20-Aug-2013 DOB: Address: Date Received: 20-Aug-2013 Patient #: Report Date: 14-Sep-2013 Gender: Female City, St., Zip: Specimen Type: FFPE, Needle Core SSN: Country: Customer Ref.: G Clinicopathologic Findings LOW RISK The breast cancer tissue sample submitted was analyzed by MammaPrint, an IVDMIA 70 - Gene Pro e of Breast Cancer for Metastatic Risk that has been validated to correlate with high or low outcome risk for distant metastases in patients with invasive breast cancer. ¹ In the reference group as published, Low Risk means that a lymph node negative breast cancer patient has a 10% chance (95% CI 4-15) that their cancer will recur within 10 years without any additional adjuvant treatment, either hormonal therapy or chemotherapy. ² T he analytical measurement performed on the sample fell within a pre-defined area around the classification cut-off (i.e. borderline sample). B orderline samples have a less than 90% classification accuracy (i.e. > 10% chance of false classification). Tumor Cell Percentage: 50 % RNA Integrity : N/A The reported tumor cell percentage and pathology comments serve as a quality control for Agendia s genomic assays and should not be viewed as a diagnosis of the presence or absence of malignancy Assay Description The U.S. Food and Drug Administration (FDA) has provided IVDMIA clearance of MammaPrint with fresh tissue for Stage I and II, lymph node negative, invasive breast cancer, for patients of all ages who have a tumor of 5 cm or less, independent of estrogen receptor status (ER+/-), based upon the development and validation of the assay as reported in Nature, New England Journal of Medicine, Journal of the National Cancer Institute and BMC Genomics. 2-5The test is performed using a microarray-based gene expression pro e that was independently validated on 10 year outcome data on an untreated patient cohort. 2 An unbiased, supervised analysis of the entire human genome, ~25,000 genes, followed by a leave-one-out cross-validation procedure, revealed the 70 critical genes that distinguish patients at High Risk vs. Low Risk of metastasis. 3 Based on the analytical performance of MammaPrint, the accuracy of classifying a sample as High Risk or Low Risk is 98.9% with reproducibility of the measurement being 98.5%. 1 MammaPrint has been validated in over 774 breast cancer patients and shown to provide information independent of clinicopathological risk assessment. 2,4,5 MammaP rint B reast C ancer G ene P rofile3 TRANSBIG Validation Results Agendia NV Science Park XH Amsterdam The Netherlands phone: + 31 ( 0) fax: + 31 ( 0) customerservice@agendia.com FFP /Agendia (Netherlands) R-ROW-013-V1 Figuur 1. Resultaat formulier van het 70-genen profiel Figuur 1. Resultaat formulier van het 70-genen profiel 24 Chapter 2

27 Tabel 2. Categorieën voor mate van bewijs wetenschappelijk onderzoek 13 Categorie A B C D Prospectieve, gecontroleerde studie, gericht op het onderzoeken van de betreffende tumor marker, waarbij de patiënt prospectief geïncludeerd, behandeld en gevolgd wordt. Ook weefsel wordt ten tijde van inclusie verzameld en geanalyseerd. De studie heeft genoeg power om de hypothese over de betreffende marker te toetsen. Het is onwaarschijnlijk dat het toeval de resultaten beïnvloedt, waardoor validatie gewenst, maar niet vereist is. Prospectieve studie, niet specifiek gericht maar wel te gebruiken voor het onderzoeken van de betreffende tumor marker, waarbij de patiënt prospectief geïncludeerd wordt en behandeld en gevolgd volgens de standaard. Ook weefsel wordt ten tijde van inclusie verzameld en geanalyseerd. De studie heeft genoeg power om de therapeutische vraag te beantwoorden, maar niet om de hypothese over de betreffende marker te toetsen. Er is vooraf een plan t.a.v. de statistische analyses om deze vragen te beantwoorden. Het is meer waarschijnlijk dat het toeval de resultaten beïnvloedt, waardoor een of meer validatie-studies nodig zijn. Prospectieve observationele registratie. Patiënten zijn prospectief geïncludeerd in de registratie, maar behandeling en follow-up gaan volgens de standaard. Ook weefsel wordt ten tijde van inclusie verzameld, maar achteraf geanalyseerd. De studie heeft niet genoeg power om de hypothese over de betreffende marker te toetsen. Er is vooraf een plan t.a.v. de statistische analyses om deze vragen te beantwoorden. Het is vrij waarschijnlijk dat het toeval de resultaten beïnvloedt, waardoor opvolgende validatie-studies nodig zijn. Geen prospectief aspect in de studie. Geen prospectieve registratie. Weefsels worden verzameld en retrospectief geanalyseerd. De studie heeft prospectief niet genoeg power. Er is vooraf geen plan t.a.v. de statistische analyses om deze vragen te beantwoorden. Het is vrij waarschijnlijk dat het toeval de resultaten beïnvloedt, waardoor opvolgende validatiestudies nodig zijn. 2 Klinisch nut in subgroepen op basis van klinisch-pathologische factoren Leeftijd Steeds meer postmenopauzale vrouwen komen in aanmerking voor ACT ondanks dat deze groep vaker gunstige biologische tumorkarakteristieken heeft. 14,15 In een studie onder 148 systemisch onbehandelde patiënten tussen de 55 en 71 jaar met een T1-2N0M0 tumor, behandeld in het Nederlands Kanker Instituut (NKI) tussen 1984 en 1996, van wie 18% adjuvant hormonale therapie had gehad en niemand ACT, bleek er op basis van het 70-genen profiel een significant verschil in 5-jaars borstkanker-specifieke overleving tussen patiënten met een laag (99%, SE 1%) en een hoog risico (80%, SE 5%; p=0.036)(categorie C). 14 Op basis van het genexpressie-profiel zou bij de eerste groep veilig kunnen worden afgezien van ACT. Ook in een cohort van 100 patiënten met een gemiddelde leeftijd van 62 jaar en T1-2N0M0 tumoren, behandeld in het Massachusetts General Hospital tussen 1985 en 1997, waarvan 15% van de laag risico groep (n=27) en 23% van de hoog risico groep (n=73) ACT kreeg, had het 70-genen profiel een zeer goede negatief voorspellende waarde. 15 Na een mediane follow-up van 11.3 jaar ontwikkelde één van de patiënten met een 70-genen profiel laag risico tumor afstandsmetastasen (categorie D). Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 25

28 Tumorgrootte Ondanks dat in vele richtlijnen is opgenomen dat kleinere tumoren geassocieerd zijn met een goede prognose, 16 zien we in de praktijk dat kleine tumoren ook kunnen metastaseren. 17 Recent werd naar het 70-genen profiel gekeken van 964 patiënten met een T1 tumor (<2 cm). 18 In de 70-genen profiel laag risico groep (n= 525) was 10% behandeld met ACT en in de hoog risico groep (n=439) 37%. De 10-jaars borstkanker-specifieke-overleving van patiënten met een laag risico was 91% (SE 2%) en 72% (SE 3%) bij patiënten met een hoog risico (HR voor wat betreft borstkankerspecifieke sterfte na 10 jaar: 4.22 (95%CI: ); p<0.001). Op basis van het genexpressie-profiel (hoog risico) zou bijna de helft van de patiënten wel in aanmerking komen voor ACT bij de overwegende kleine tumoren (categorie D). ER-status Het 70-genen profiel is zowel voor ER-positieve als -negatieve tumoren gevalideerd. 5 ER-positieve tumoren zijn geassocieerd met een gunstiger 5-jaars prognose en hebben vaker een laag risico 70-genen profiel. 5 ER-negatieve tumoren zijn geassocieerd met een minder gunstige 5-jaars prognose, doordat deze vaker een hogere proliferatie vertonen en er geen effect te zien is van adjuvant hormonale therapie (AHT). In verzamelde validatiestudies is gevonden dat 3-6% van de ER-negatieve tumoren toch een laag risico heeft volgens het 70-genen profiel. Er zal verder onderzoek moeten plaatsvinden om te kijken of bij deze subgroep van ER-negatieve patiënten ACT veilig achterwege gelaten kan worden. HER2-status Een andere ongunstige prognostische factor is overexpressie van HER2. Vrijwel alle patiënten met een HER2-positief mammacarcinoom worden behandeld met ACT in combinatie met trastuzumab. 2 Trastuzumab is een duur middel, dat een lange behandelduur vraagt van een jaar en in combinatie met ACT gepaard gaat met een klein, maar niet verwaarloosbaar risico op ernstige cardiale toxiciteit in de eerste jaren na behandeling. Uit de HERA studie, waarin 1 of 2 jaar trastuzumab vergeleken werd met observatie bij 1698 HER2-positieve patiënten, is gebleken dat 72.2% (HR 0.76 (95% CI ); p<0.0001) van de patiënten 4 jaar ziektevrij bleven na ACT zonder trastuzumab. 19,20 In een recente studie identificeert het 70-genen profiel een laag risico groep met een relatief goede overleving binnen de HER2-positieve tumoren die niet behandeld zijn met ACT en/of trastuzumab (n=89). 21 Deze patiënten hadden een 10-jaars metastase-vrije-overleving van 84% in geval van een laag risico 70-genen profiel, hetgeen op zich nog steeds zal leiden tot behandeling met ACT en trastuzumab. Echter, voor patiënten die naast een laag risico 70-genen profiel een zogeheten hoog endocriene responsieve tumor (ER >50%) hadden en die noch trastuzumab, noch ACT hadden gekregen, was de 10-jaars metastase-vrijeoverleving 100% (categorie D) Chapter 2

29 Lymfklierstatus De aanwezigheid van axillaire lymfkliermetastasen is veelal een indicatie voor ACT, maar er zijn ook patiënten die zonder ACT een goede overlevingskans hebben. 22 Een onafhankelijke validatiestudie werd uitgevoerd in een groep van 241 patiënten met een operabel mammacarcinoom en 1-3 positieve lymfklieren, behandeld in het NKI of het European Institute of Oncology in Milaan. 23 In de 70-genen profiel laag risico groep (n=99) werd 41% behandeld met ACT en in de hoog risico (n=142) groep 67%. De 10-jaars metastase-vrije-overleving was 91% (SE 4%) in de laag risico groep en 76% (SE 4%) in de hoog risico groep (p=0.001). De 10-jaars borstkanker-specifiekeoverleving in dit cohort was 96% (SE 2%) in de laag risico groep en 76% (SE 4%) in de hoog risico groep (p<0.001). Met een multivariate HR van 7.17 (95%CI ; p=0.005) is het 70-genen profiel significant beter in het voorspellen van de overleving dan de bekende klinischpathologische factoren bij deze patiënten. 22 Het lastige van deze studie is dat een substantieel deel van de patiënten adjuvant systemisch behandeld is, waardoor het onduidelijk blijft wat de prognose van deze lymfklier-positieve patiëntengroep zou zijn geweest zonder ACT. Toch lijkt het binnen deze groep mogelijk om op basis van tumorload en genexpressie-profiel genuanceerd te denken over de indicatie voor ACT (categorie C). 2 Veel gebruikte klinisch-pathologische richtlijnen, zoals de NABON richtlijn, Adjuvant! Online, de St. Gallen richtlijn en de Nottingham Prognostic Index (NPI) zijn discordant met het 70-genen profiel in 7-40% van de patiënten. 23 Het profiel geeft aanvullende prognostische informatie bij een concordante laag risico-inschatting volgens de NABON richtlijn van 2004, St. Gallen en NPI of in geval van discordantie tussen de verschillende richtlijnen. 24 Indien een patiënt een hoog risico op afstandsmetastasen heeft volgens de drie richtlijnen, geeft het 70-genen profiel geen aanvullende informatie. Het grijze gebied In Nederland wordt bij gemiddeld vrouwen per jaar een invasief mammacarcinoom gediagnosticeerd. Jaarlijks zijn ongeveer 1400 van deze patiënten tussen jaar, met een T1 tumor (<2 cm), graad 2, en een negatieve lymfklierstatus of alleen micrometastasen in de oksel, zonder afstandsmetastasen. In 2008 ontving 15% van deze patiënten ACT. In 2010, na aanpassing van de 2008 richtlijn, is dit percentage bijna verdubbeld naar 27% (tabel 3). De huidige tendens is om bij twijfel een patiënt met ACT te behandelen. De vraag is of al deze patiënten ook daadwerkelijk profijt hebben hiervan. Immers, een groot deel van deze patiënten krijgt ook zonder ACT nooit metastasen. Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 27

30 Tabel 3. Adjuvant systemische therapie in Nederland voor patiënten jaar met T1N0i-/+M0 tumor graad 2 AST Geen % % % Hormonale Therapie (HT) % % % Chemotherapie+/- HT % % % Totaal Naast het uitvoerige retrospectieve bewijs dat het 70-genen profiel toegevoegde prognostische waarde heeft, zijn er nu ook de resultaten van de prospectieve RASTER studie (categorie B). 25 Deze observationele studie, waarbij het 70-genen profiel gebruikt werd in de besluitvorming om al dan niet adjuvant systemisch te behandelen, includeerde tussen 2004 en patiënten, < 61 jaar, met een T1-3N0M0 mammacarcinoom (70% T1, 29% T2, 1% T3). De 5-jaar follow-up resultaten laten zien dat patiënten met een laag risico profiel, waarvan 15% behandeld is met ACT, een uitstekende 5-jaars metastase-vrije-overleving hebben van 97% ten opzichte van 92% in de hoog risico groep, waarin 85% behandeld is met ACT. 25 Binnen de gehele patiëntengroep classificeerde het 70-genen profiel 22% minder patiënten als hoog risico dan de NABON richtlijn (2012). Binnen de groep T1 tumoren (n=301) was een reductie van 18% te zien (ongepubliceerde data). Dit percentage discordantie tussen de klinische risico-inschatting en het 70-genen profiel wordt ook in de MINDACT gezien. 10 De discordante groep bestond voornamelijk uit patiënten met een ER-positieve, HER2-negatieve tumoren van 1-2 cm, graad 2. Dit is juist de groep waarbinnen volgens gegevens van de Nederlandse Kanker Registratie de afgelopen jaren een forse stijging in behandeling met AST wordt gezien (tabel 3) ten gevolge van de recente aanscherping van de landelijke NABON-richtlijnen. Voor hen kan het 70-genen profiel dan ook het meest bijdragen in de besluitvorming om al dan niet adjuvant systemisch te behandelen. Ook werd een metastasevrije-overleving van 100% gezien in de groep die ACT noch AHT had gehad bij een laag risico volgens het 70-genen profiel, terwijl de huidige NABON-richtlijn deze patiënten als hoog risico had geclassificeerd. 25 Deze resultaten suggereren dat ACT veilig achterwege gelaten kan worden bij een patiënte in dit grijze gebied in geval van een laag risico 70-genen profiel. Voor het veilig achterwege laten van AHT zijn langere follow-up data nodig, omdat twee-derde van de recidieven in de ER-positieve groep tussen 5 en 20 jaar follow-up optreedt. 2 Ondanks dat er in de eerder genoemde RASTER studie niet gerandomiseerd was, geeft de studie wel belangrijke inzichten in het gebruik van het 70-genen profiel in de dagelijkse praktijk. 28 Chapter 2

31 Conclusie Naast retrospectief is er nu ook het eerste prospectieve bewijs uit de RASTER studie dat het 70-genen profiel additionele informatie geeft voor patiënten waarbij er, ondanks minder gunstige klinisch-pathologische factoren, toch twijfel blijft bestaan over de geschatte winst van ACT. In afwachting van de resultaten van de MINDACT studie, welke definitief uitsluitsel zullen geven of ACT veilig achterwege gelaten kan worden in geval van een laag risico 70-genen profiel, lijkt op basis van dit overzicht de toegevoegde waarde van het 70-genen profiel het grootst voor patiënten voor wie het stellen van de indicatie voor adjuvant chemotherapie op basis van patiënt- en tumorkarakteristieken niet eenduidig is. Het gaat hier om patiënten tussen 45 en 55 jaar oud met een tumor van 1-2 cm, graad 2, ER-positief en HER2-negatief. Voor andere patiëntengroepen kan in individuele gevallen het 70-genen profiel als extra hulpmiddel ingezet worden bij de besluitvorming omtrent het nut van adjuvant systemische therapie. 2 Dankwoord Met dank aan de Nederlandse Kankerregistratie beheerd door IKNL voor het verstrekken van de data aangaande het gebruik van adjuvant systemische therapie in Nederland. Met dank aan Stella Mook en Laura van t Veer voor hun bijdrage aan de vormgeving van dit manuscript. Leerpunten Genexpressie-profielen geven belangrijke aanvullende informatie over het risico op afstandsmetastasen bij borstkanker patiënten. Het 70-genen profiel kan voor zowel pre- als postmenopauzale patiënten, met of zonder aangedane axillaire lymfklieren en voor zowel HER2-positieve als -negatieve tumoren ingezet worden als hulpmiddel bij twijfel over de prognose inschatting en de daarmee gepaard gaande absolute overlevingswinst die met adjuvant systemische therapie verkregen kan worden. Vooral voor patiënten met ER-positieve, HER2-negatieve tumoren van 1-2 cm, graad 2 kan het 70-genen profiel bijdragen in de besluitvorming om al dan niet adjuvant chemotherapie te adviseren, naast adjuvant hormonale therapie. Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 29

32 Referenties 1 Esserman LJ, Shieh Y, Rutgers EJ, Knauer M, Retel VP, Mook S et al. Impact of mammographic screening on the detection of good and poor prognosis breast cancers. Breast Cancer Res Treat 2011; 130: Early Breast Cancer Trialists Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 365: Nationaal Borstkanker Overleg Nederland NABON, Kwaliteitsinstituut voor de Gezondheidszorg CBO, Vereniging van Integrale Kankercentra. Adjuvant Systemische Therapie. Richtlijn Mammacarcinoom Mook S, van t Veer LJ, Rutgers EJ, Piccart-Gebhart MJ, Cardoso F. Individualization of therapy using Mammaprint: from development to the MINDACT Trial. Cancer Genomics Proteomics 2007; 4: van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas AM et al. Validation and clinical utility of a 70 gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 2006; 98: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009; 117: Sapino A, Roepman P, Linn SC, Snel MH, Delahaye LJ, van den Akker J et al. MammaPrint molecular diagnostics on Formalin Fixed Paraffin Embedded tissue. J of Mol Diagn 2013; in press. 9 FDA Label - USFDA Clearance; Rutgers E, Piccart-Gebhart MJ, Bogaerts J, Delaloge S, van t Veer LV, Rubio IT et al. The EORTC 10041/ BIG MINDACT trial is feasible: results of the pilot phase. Eur J Cancer 2011; 47: Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 2004; 351: Tang G, Shak S, Paik S, Anderson SJ, Costantino JP, Geyer CE, Jr. et al. Comparison of the prognostic and predictive utilities of the 21-gene Recurrence Score assay and Adjuvant! for women with nodenegative, ER-positive breast cancer: results from NSABP B-14 and NSABP B-20. Breast Cancer Res Treat 2011; 127: Simon RM, Paik S, Hayes DF. Use of archived specimens in evaluation of prognostic and predictive biomarkers. J Natl Cancer Inst 2009; 101: Mook S, Schmidt MK, Weigelt B, Kreike B, Eekhout I, van de Vijver MJ et al. The 70-gene prognosis signature predicts early metastasis in breast cancer patients between 55 and 70 years of age. Ann Oncol 2010; 21: Wittner BS, Sgroi DC, Ryan PD, Bruinsma TJ, Glas AM, Male A et al. Analysis of the MammaPrint breast cancer assay in a predominantly postmenopausal cohort. Clin Cancer Res 2008; 14: Fitzgibbons PL, Page DL, Weaver D, Thor AD, Allred DC, Clark GM et al. Prognostic factors in breast cancer. College of American Pathologists Consensus Statement Arch Pathol Lab Med 2000; 124: Foulkes WD, Reis-Filho JS, Narod SA. Tumor size and survival in breast cancer--a reappraisal. Nat Rev Clin Oncol 2010; 7: Mook S, Knauer M, Bueno-de-Mesquita JM, Retel VP, Wesseling J, Linn SC et al. Metastatic potential of T1 breast cancer can be predicted by the 70-gene MammaPrint signature. Ann Surg Oncol 2010; 17: Chapter 2

33 19 Gianni L, Dafni U, Gelber RD, Azambuja E, Muehlbauer S, Goldhirsch A et al. Treatment with trastuzumab for 1 year after adjuvant chemotherapy in patients with HER2-positive early breast cancer: a 4-year follow-up of a randomised controlled trial. Lancet Oncol 2011; 12: Untch M, Gelber RD, Jackisch C, Procter M, Baselga J, Bell R et al. Estimating the magnitude of trastuzumab effects within patient subgroups in the HERA trial. Ann Oncol 2008; 19: Knauer M, Cardoso F, Wesseling J, Bedard PL, Linn SC, Rutgers EJ et al. Identification of a low-risk subgroup of HER-2-positive breast cancer by the 70-gene prognosis signature. Br J Cancer 2010; 103: Mook S, Schmidt MK, Viale G, Pruneri G, Eekhout I, Floore A et al. The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1-3 positive lymph nodes in an independent validation study. Breast Cancer Res Treat 2009; 116: Bueno-de-Mesquita JM, van Harten WH, Retel VP, van t Veer LJ, van Dam FS, Karsenberg K et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007; 8: Bueno-de-Mesquita JM, Sonke GS, van de Vijver MJ, Linn SC. Additional value and potential use of the 70-gene prognosis signature in node-negative breast cancer in daily clinical practice. Ann Oncol 2011; 22: Drukker CA, Bueno-de-Mesquita JM, Retel VP, van Harten WH, van Tinteren H, Wesseling J et al. A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 2013;133: Voorspelling prognose borstkanker: bijdrage genexpressie-profiel 31

34 Chapter 3

35 A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study International Journal of Cancer 2013;133: Caroline A. Drukker Jolien M. Bueno de Mesquita Valesca P. Retèl Wim H. van Harten Harm van Tinteren Jelle Wesseling Rudi M.H. Roumen Michael Knauer Laura J. van t Veer Gabe S. Sonke Emiel J.Th. Rutgers Marc J. van de Vijver Sabine C. Linn

36 Abstract The 70-gene signature (MammaPrint ) has been developed on retrospective series of breast cancer patients to predict the risk of breast cancer distant metastases. The microarray-prognosticsin-breast-cancer (RASTER) study was the first study designed to prospectively evaluate the performance of the 70-gene signature, which result was available for 427 patients (ct1-3n0m0). Adjuvant systemic treatment decisions were based on the Dutch CBO 2004 guidelines, the 70- gene signature, and doctors and patients preferences. Five-year distant-recurrence-free-interval (DRFI) probabilities were compared between subgroups based on the 70-gene signature and Adjuvant! Online (AOL) (10-year survival probability <90% was defined as high risk). Median follow-up was 61.6 months. Fifteen percent (33/219) of 70-gene signature low risk patients received adjuvant chemotherapy (ACT) versus 81% (169/208) of 70-gene signature high risk patients. The 5-year DRFI probabilities for 70-gene signature low risk (n=219) and high risk (n=208) patients were 97.0% and 91.7%. The 5-year DRFI probabilities for AOL low risk (n=132) and high risk (n=295) patients were 96.7% and 93.4%. For 70-gene signature low risk AOL high risk patients (n=124), of whom 76% (n=94) had not received ACT, 5-year DRFI was 98.4%. In the AOL high risk group, 32% (94/295) less patients would be eligible to receive ACT if the 70-gene signature was used. In this prospective community-based observational study, the 5-year DRFI probabilities confirmed the additional prognostic value of the 70-gene signature to clinic-pathological risk estimations such as AOL. Omission of adjuvant chemotherapy as judged appropriate by doctors and patients and instigated by a low risk 70-gene signature result, appeared not to compromise outcome. 34 Chapter 3

37 Introduction Over the last two decades breast cancer mortality has declined in Western countries. This decline has been ascribed to early detection due to the implementation of population-based mammographic screening programs and the introduction of adjuvant systemic therapy (AST). 1 Fifty percent of all breast cancer patients are cured with loco-regional therapy alone, while the other 50% recur in the absence of AST. The combination of adjuvant chemotherapy and adjuvant endocrine therapy halves the breast cancer mortality rate throughout the first 15 years after diagnosis. 2 Selection of those patients at high risk of relapse for AST is based on clinico-pathologic factors, such as age, tumor size, nodal status, histological grade, and hormone receptor status. Current guidelines and clinical tools, such as Adjuvant! Online (AOL), use these factors to estimate the risk of recurrence and the benefit of AST. However, when using these standard clinico-pathologic factors, individual risk assessment remains challenging. Many women are treated with chemotherapy, without deriving significant benefit. 3 To improve accuracy, several gene expression prognosis classifiers have been developed and validated on historic data to refine risk estimation based on current guidelines. 4 One of these is the 70-gene signature (MammaPrint ), for which its accuracy to select the right patient for AST is being compared to the accuracy of AOL in a randomized trial called MINDACT, that completed accrual and primary results are awaited. 5 Between 2004 and 2006 the 70-gene signature has been subjected to the first prospective study using a gene-expression prognosis classifier as a risk estimation tool, in addition to clinicopathological factors. The microarray prognostics in breast cancer (RASTER) study was conducted in 16 community hospitals in the Netherlands. 6 The primary aim of this multicenter observational study were to assess the feasibility of implementing the 70-gene signature in a community-based setting and to study the clinical impact of the 70-gene signature test result on AST decision making. 6 A secondary aim of the RASTER study was to assess the outcome of patients for whom a gene expression classifier was used to determine the need for adjuvant systemic treatment. Implementation of the 70-gene signature in daily clinical practice appeared feasible. A considerable discrepancy in risk estimations among different clinico-pathologic guidelines and the 70-gene signature was observed. 6 The addition of the 70-gene signature test result to standard clinicopathological factors led to a change in AST advice in 19% of patients. 6 Here, we report the 5-year follow-up data of the RASTER study. 3 Patients and methods The RASTER study design, patient eligibility criteria and study logistics have been described elsewhere. 6 In short, 812 female patients were enrolled after having given written informed consent. 427 patients were postoperatively eligible and for them a 70-gene signature 5-year follow-up of the RASTER study 35

38 (MammaPrint, Agendia NV) was obtained. All 427 patients were aged years old and had a histologically confirmed unilateral, unifocal, primary operable, invasive adenocarcinoma of the breast (ct1-3n0m0). Exclusion criteria were a history of a malignancy (with exception of basal-cell carcinoma or cervical dysplasia) and neoadjuvant systemic treatment. After enrollment of 242 patients, the maximum allowed age was changed to 54 years, because the 70-gene signature had been developed in patients under 55 years of age. At that time, the validation of the prognostic value in patients aged over 55 years was not yet available. 7 After enrollment, patients received surgery as their primary treatment. All patients underwent either breast conserving treatment or ablation of the breast. Within one hour after surgery, the tumor samples (stored without any preserving solution) were procured at the Pathology Department of the participating hospitals. To ensure (adapt to) routine clinical practice, the initial histopathology data were used for clinical risk assessment by the treating physician and in the statistical analysis, without central review of paraffin-embedded tumor samples. Details on tumor grading, assessment of hormone receptor status and HER2 status, RNA extraction and microarray analysis are described elsewhere. 8, 9 The RASTER study is registered on the International Standard Randomised Controlled Trial Register, number ISRCTN A summary of the study protocol is outlined online ( Established clinical risk classification indexes AST decisions in this study were based on the Dutch Institute of Healthcare Improvement (CBO) guidelines of 2004, 10 the 70-gene signature result, and doctors and patients preferences. The CBO guidelines used between 2004 and 2006 were more restrictive in selecting patients for AST as compared to other guidelines and were primarily based on the assumption that adjuvant chemotherapy is only justified if an absolute survival benefit of more than 5% at ten years can be expected. According to the 2004 CBO guidelines, low clinical risk was defined as age over 35 years, tumor of grade 1 and 30 mm or smaller, grade 2 and 20 mm or smaller, or grade 3 and 10 mm or smaller. Additionally, age less than 36 years with a grade 1 tumor of 10 mm or smaller was also defined as low risk. All other patients were defined as high risk. Notably, in the CBO guidelines, adjuvant endocrine treatment was advised only in clinically high risk patients with hormone-receptor-positive tumors in combination with chemotherapy. 10 To study how the addition of the 70-gene signature to a risk prediction tool used today influences clinical practice we used AOL software, version 8.0 to calculate 10-year survival probabilities based on the patient s age, tumor size, tumor grade, estrogen receptor status, and nodal status. 11,12 Patients were assigned to a high clinical risk if their calculated 10-year survival probability was less than 90%. 6 In addition, sensitivity analyses were performed for different AOL cutoffs ranging from 85% to 95%, including the cutoff used for the MINDACT trial Chapter 3

39 Statistical analysis For this analysis, we estimated 5-year distant-recurrence-free interval (DRFI), comprising distant recurrence and death from breast cancer. Overall survival (OS) and distant-disease-free-survival (DDFS) were also calculated. 13 Survival curves were constructed using the Kaplan-Meier method and compared using the log-rank test. In case of ordinal variables (age, pt-stage of TNM, histological grade and nodal status) with more than two groups, we tested for trends (using the Cochran-Armitage test). A significant finding was defined as a p-value below Analyses were performed using SAS version 9.2 and R version Results 3 Follow-up data of all 427 patients who were enrolled in the RASTER study were updated until September 15 th The first patient was enrolled January 22, 2004, the last patient December 18 th, Median follow-up was 61.6 months. Patient characteristics, AST and outcome stratified by 70-gene signature Supplementary Table 1 summarizes the patient characteristics defined by the result of the 70-gene signature as reported by Bueno-de-Mesquita et al gene signature high risk patients more often had large, poorly differentiated, estrogen receptor (ER)-negative, progesterone receptor (PR)-negative, and HER2-positive tumors than 70-gene signature low risk patients. Nineteen percent (9/47) of invasive lobular breast cancer patients had a high risk 70-gene signature, while 53% (183/345) of invasive ductal breast cancer patients had a high risk 70-gene signature result. Twelve percent (16/136) of grade 3 tumors were 70-gene signature low risk, while 83% (72/87) of grade 1 tumors were 70-gene signature low risk. After a median follow-up time of 61.6 months, 24 DRFI events and 11 deaths occurred. Nine patients died due to breast cancer. One patient died due to lung cancer and one patient due to cardiac disease (Supplementary Table 2). The 5-year DRFI probabilities for 70-gene signature low risk (n=219) and high risk (n=208) patients were 97.0% and 91.7% (p=0.03), respectively (Supplementary Figure 1). Importantly, this difference in outcome was observed despite the fact that in the 70-gene signature low risk group 15% (33/219) of the patients received adjuvant chemotherapy, versus 81% (169/208) in the high risk group. The administered chemotherapy regimens for low and high risk patients are described in Supplementary Table 1. Patient characteristics, AST and outcome stratified by 70-gene signature and AOL Table 1 shows the patient characteristics stratified by 70-gene signature and AOL risk prediction. Discordant risk estimations between 70-gene signature and AOL occurred in 38% of the cases (161/427). Most discordant cases were 70-gene signature low risk and AOL high risk (124/427=29%), while 37 cases (37/427=9%) had a high risk 70-gene signature result and a low 5-year follow-up of the RASTER study 37

40 risk AOL estimation. Figure 1 summarizes the AST received in the different categories defined by 70-gene signature result and AOL. Ninety-three percent (88/95) of patients who were 70- gene signature low risk and AOL low risk did not receive any AST (chemotherapy nor endocrine therapy). Fifty-six percent (70/124) of patients who were 70-gene signature low risk and AOL high risk did not receive any AST. In Supplementary Figure 1 Kaplan-Meier plots for DRFI, DDFS and OS are given for the whole group of patients, according to 70-gene signature, and according to AOL risk estimation. The 5-year DRFI probabilities for AOL low risk (n=132) and high risk (n=295) patients were 96.7% and 93.4%, respectively (p=0.24) (Supplementary Figure 1). Table 2 shows DRFI and DDFS probabilities according to the combined risk categories. Table 1. Clinicopathological characteristics of patient groups defined by 70-gene signature (70-GS) and AOL risk estimations Total 70-GS low- 70-GS high- 70-GS low- 70-GS high- (n = 427) AOL low (n = 95) AOL low (n = 37) AOL high (n = 124) AOL high (n = 171) Age <35 26 (6%) 5 (5%) 0 (0%) 2 (2%) 19 (11%) (10%) 12 (13%) 7 (19%) 2 (2%) 20 (12%) (20%) 19 (20%) 14 (38%) 18 (14%) 33 (19%) (33%) 28 (30%) 8 (22%) 58 (47%) 47 (28%) (23%) 27 (28%) 8 (22%) 29 (23%) 36 (21%) >55 35 (8%) 4 (4%) 0 (0%) 15 (12%) 16 (9%) pt (TNM) pt1 (<20mm) 301 (70%) 95 (100%) 37 (100%) 82 (66%) 87 (51%) pt2 (>20-50mm) 125 (29%) 0 (0%) 0 (0%) 42 (33%) 83 (48%) pt3 (>50mm) 1 (1%) 0 (0%) 0 (0%) 0 (0%) 1 (1%) Histological grade Good 87 (20%) 60 (63%) 12 (32%) 12 (10%) 3 (2%) Intermediate 204 (48%) 34 (36%) 19 (51%) 97 (78%) 54 (32%) Poor 136 (32%) 1 (1%) 6 (16%) 15 (12%) 114 (67%) Histological type Ductal 345 (81%) 73 (77%) 30 (81%) 89 (72%) 153 (89%) Lobular 47 (11%) 14 (15%) 2 (5%) 24 (19%) 7 (4%) Other 31 (7%) 7 (7%) 5 (13%) 9 (7%) 10 (6%) Unknown 4 (1%) 1 (1%) 0 (0%) 2 (2%) 1 (1%) ER status Negative 85 (20%) 0 (0%) 4 (11%) 3 (2%) 78 (46%) Positive 342 (80%) 95 (100%) 33 (89%) 121 (98%) 93 (54%) PgR status Negative 133 (31%) 9 (9%) 8 (21%) 24 (19%) 92 (54%) Positive 293 (69%) 86 (91%) 29 (78%) 100 (81%) 78 (46%) Unknown 1 (<1%) 0 (0%) 0 (0%) 0 (0%) 1 (1%) HER2 status Negative 358 (84%) 86 (91%) 29 (78%) 111 (90%) 132 (77%) Positive 48 (11%) 5 (5%) 5 (14%) 4 (3%) 34 (20%) Unknown 21 (5%) 4 (4%) 3 (8%) 9 (7%) 5 (3%) ER=estrogen receptor; PgR=progesterone receptor; HER2=Human Epidermal growth factor Receptor 2 38 Chapter 3

41 3 Figure 1. Distribution of patients (n=427) over the four risk categories defined by 70-gene signature and AOL risk estimations and proportion and type of AST received per category mlow = 70-gene signature low; mhigh = 70-gene signature high; clow = AOL low; chigh = AOL high; CT=adjuvant chemotherapy; Endo=adjuvant endocrine therapy; AST=adjuvant systemic therapy. Table 2. Kaplan-Meier risk estimations for DRFI and DDFS according to 70-gene signature and AOL stratification 70-gene signature AOL AST 5-year DRFI (%) (95% CI) 5-years DDFS (%) (95% CI) Low Low 7/95 (7%) 95.3 ( ) 94.3 ( ) High Low 32/37 (86%) 100 ( ) 94.6 ( ) Low High 54/124 (44%) 98.4 ( ) 97.6 ( ) High High 166/171 (97%) 89.8 ( ) 88.7 ( ) The difference in overall survival outcome between 70-gene signature low risk and AOL low risk is partly due to the two cases who died of non-breast cancer causes (Supplementary Table 2) who were categorized as 70-gene signature low risk and AOL high risk. Sensitivity analyses were performed for different AOL cutoffs ranging from 85% to 95%, showing a shift in the proportion of low risk patients without a substantial effect on DRFI, DDFS or OS survival probabilities (Supplementary Table 3). 5-year follow-up of the RASTER study 39

42 Outcome at five years according to AST in patients with a low risk 70-gene signature result Five-year DRFI was 98.4% in patients with 70-gene signature low risk AOL high risk (n=124), of which 76% (n=94) had not received adjuvant chemotherapy. The group that had not received adjuvant chemotherapy had a 5-year DRFI of 98.9%. The group that did not receive any systemic therapy (chemotherapy nor endocrine therapy) (n=70) had a 5-year DRFI of 100% (Figure 2a and 2b). No significant difference (p=0.29) was seen between systemically untreated patients with a concordant low risk assessment and patients with a 70-gene signature low risk result even with a high risk assessment by AOL. Table 3 shows the patient characteristics of patients who had a low risk 70-gene signature result and who had received adjuvant endocrine therapy only or no AST at all, split by AOL risk assessment. Table 3. Clinicopathological characteristics of 70-gene signature low risk patients who received no AST or ET only 70-GS low- AOL low 70-GS low-aol high No AST (n=88) No AST or ET only (n=92) No AST (n=70) No AST or ET only (n=94) Age <35 2 (2%) 3 (3%) 0 (0%) 0 (0%) (12%) 11 (12%) 0 (0%) 1 (1%) (22%) 19 (21%) 8 (11%) 8 (9%) (30%) 28 (30%) 32 (46%) 44 (47%) (30%) 27 (29%) 18 (26%) 26 (28%) >55 4 (5%) 4 (4%) 12 (17%) 15 (16%) pt (TNM) pt1 (<20mm) 88 (100%) 92 (100%) 62 (89%) 75 (80%) pt2 (>20-50mm) 0 (0%) 0 (0%) 8 (11%) 19 (20%) pt3 (>50mm) 0 (0%) 0 (0%) 0 (0%) 0 (0%) Histological grade Histological type Good 57 (65%) 60 (65%) 8 (11%) 9 (10%) Intermediate 30 (34%) 31 (34%) 60 (86%) 77 (82%) Poor 1 (1%) 1 (1%) 2 (3%) 8 (9%) Ductal 68 (77%) 72 (78%) 47 (67%) 68 (72%) Lobulair 13 (14%) 13 (14%) 16 (23%) 19 (20%) Other 7 (8%) 7 (8%) 6 (9%) 6 (6%) Unknown 0 (0%) 0 (0%) 1 (1%) 1 (1%) ER status Negative 0 (0%) 0 (0%) 2 (3%) 2 (2%) Positive 88 (100%) 92 (100%) 68 (97%) 92 (98%) PgR status Negative 9 (10%) 9 (10%) 15 (21%) 21 (22%) Positive 79 (90%) 83 (90%) 55 (79%) 73 (78%) Unknown 0 (0%) 0 (0%) 0 (0%) 1 (1%) HER2 status Negative 79 (90%) 83 (90%) 63 (90%) 85 (90%) Positive 5 (6%) 5 (5%) 2 (3%) 2 (2%) Unknown 4 (4%) 4 (4%) 5 (7%) 7 (7%) 40 Chapter 3

43 3 A) 70-gene signature - AOL in pts B) 70-gene signature - AOL in pts without any form of AST without AST or with ET only Figure 2. Five-year outcome of chemotherapy-naïve patients with a low risk 70-gene signature result Discussion The RASTER study provides the first prospective data on the outcome of patients with breast cancer for whom a gene expression prognosis classifier was used to determine the need for adjuvant systemic treatment. This community-based observational study confirms the potential of the 70-gene signature towards better selection of breast cancer patients who can forego adjuvant chemotherapy without compromising outcome. Use of the 70-gene signature reduced the proportion of high risk patients as classified by AOL by 20% (87/427). In the AOL high risk group, 32% (94/295) less patients would have received ACT if they had been treated according to the 70-gene signature risk estimation. Overall, the 5-year outcome of the whole cohort was favorable, taking into consideration that 39% (168/427) had not received any form of AST. Most importantly, the 5-year DRFI probabilities were excellent for patients who were clinically high risk but had a low risk 70-gene signature, even in the absence of any AST. In addition, there was no difference in DRFI between 70-gene signature low risk patients who were either clinical high or low risk. Patients with a high risk AOL result, but a low risk 70-gene signature result who did not receive any AST (Table 2) more often had ER-positive tumors with less often poor but more often intermediate histological grade than the total group of study patients. This group of patients had a 100% DRFI at five years. One limitation of the comparison between the gene signature and AOL is that the actual treatment decisions in this study were based on the restrictive Dutch guidelines of 2004 and doctor s and patients preferences. While this reflected clinical practice at the time of the study, equality of 5-year follow-up of the RASTER study 41

44 prognosis between groups that did or did not receive chemotherapy can not be guaranteed. Subtle selection mechanisms may therefore have influenced our results. The reduction in the number of patients eligible for AST when using the 70-gene signature can also be explained by the definition of low risk that was used for AOL. The cutoff we used here ( 90% overall survival probability at ten years is defined as low risk), which is also used in the Dutch national guidelines of 2012, classifies a relatively large proportion of patients as high risk. A lower AOL cutoff ( 85%) results in more low risk patients and thus fewer patients who require AST. Despite this lower cutoff, the outcome of patients in the AOL low risk group remained excellent. To our knowledge a cutoff below 90% is thusfar rarely used in clinical practice. Another possible limitation is that AOL risk estimations are based on 10-year outcomes, whereas we report on 5-year outcomes. The prognostic capacity of the 70-gene signature is best at a follow-up time of five years and has less discriminatory power in years From recent Oxford Overview data it is known that the carry-over effect of adjuvant chemotherapy gradually fades after five years. 2,15,16 Therefore, the data in this study can be considered relatively mature for the effect of adjuvant chemotherapy on outcome. The carry-over effect of five year adjuvant endocrine therapy remains present at ten years of follow-up. 2,17 Thus, the data presented here is immature regarding the effect of adjuvant endocrine therapy on long term outcome and needs to be reevaluated at 10-years of follow-up. Consequently, only the effect on outcome of the decision to omit adjuvant chemotherapy based on a low risk 70-gene signature can be derived from the current study. Theoretically, the best survival for the entire group of breast cancer patients will be obtained by offering AST to all patients, as long as our prognostic tests are not 100% accurate. 18 The mortality rate as a consequence of adjuvant chemotherapy toxicity is in the range of 1%. 19 For adjuvant endocrine therapy, this is in the order of 0.3%. Hence, the real question is how many unnecessary deaths we are generally accepting by erroneously foregoing AST based on a false low risk estimation to spare the large majority of breast cancer patients the unnecessary toxicity of adjuvant chemotherapy and consequent deterioration in quality of life based on a true low risk estimation. 20 In this study, 7 patients who developed distant metastases were low risk according to the 70-gene signature. Four of these patients were also low risk according to AOL. The other three patients were high risk according to AOL. One of these patients received both chemotherapy and endocrine therapy, one received endocrine therapy only, and one received no treatment at all. However, this AST untreated case developed a distant metastasis after 5 years (at 82 months of follow-up). Since 94 patients who had a 70-gene signature low risk - AOL high risk result did not receive chemotherapy and had a 98.9% (95%CI: %; Figure 2B) 5-year DRFI, one might infer that it costs about one avoidable distant recurrence (1.1%; 95%CI: 0-3.1%) to spare up to 94 patients unnecessary chemotherapy side-effects. When discussing the acceptable numbers-needed-to-treat and numbers-needed-to-harm, any prognostic factor that can improve the equation should be taken into account. The current data confirms that the 70- gene signature is such a prognostic factor. 42 Chapter 3

45 In conclusion, in a prospective community-based observational study, the 5-year follow-up data confirmed the additional prognostic value of the 70-gene signature to clinico-pathologic factors used in AOL risk estimations. Omission of chemotherapy as judged appropriate by doctors and patients and supported by a low risk 70-gene signature result appeared not to compromise outcome. Contributors SL, MvdV, WvH and LvtV were responsible for the study design and development of the protocol. WvH ensured financing. This study was financially supported by the Dutch Health Care Insurance Board. The funding source had no role in the study design, data collection, data analysis, data interpretation, in writing the report, or in the decision to submit for publication. JBdM coordinated the study. ER and RR participated in the patient accrual. JBdM, VR, MK and CD took part in data collection. MvdV, JBdM and JW took part in collection and processing of tumor samples. HvT and GS performed the data analysis. CD, GS, ER and SL took part in data interpretation and manuscript writing. All authors were involved in reviewing the report. 3 Conflict of interest The RASTER study was financially supported the Dutch Health Care Insurance Board (CVZ). LvtV and MvdV are named inventors on the patent for the 70-gene signature used in this study. LvtV reports being shareholder in and employed by Agendia NV, the commercial company that markets the 70-gene signature as MammaPrint. WvH is a non-remunerated, non-stake holding member of the supervisory board of Agendia NV. MK received unrestricted educational grants from Agendia NV. and the Austrian Society of Surgery for his research. LvtV was supported by the Dutch Genomics Initiative Cancer Genomics Centre. Acknowledgements We are indebted to the women who participated in this study; to the doctors, nurses, and data managers from the participating hospitals in the Netherlands that enrolled patients in the RASTER-study and contributed to the collection of follow-up data. Principal and co-investigators of the RASTER study The following clinicians entered patients and/or participated in the study (between the brackets, the number of accrued patients is mentioned): J. Meijer, J. Klinkenbijl, J. Douma, Alysis Care Group, Arnhem (31); J. Wijsman, D. van der Meer, P. de Wit, O. Loosveld, Amphia Hospital, Breda (4); S. Veltkamp, A. Baan, G. Timmers, K. van der Hoeven, Amstelland Hospital, Amstelveen (66); J. van der Bijl, A.M. Lenssen, I. Snijders, M. Nap, J. Wals, M. Pannebakker, Atrium Medical Center, 5-year follow-up of the RASTER study 43

46 Heerlen (13); L. Strobbe, F. van den Wildenberg, R. Berry, B. Dekker, E. Thunnissen, A. Uyterlinde, C. Mandigers, Canisius-Wilhelmina Hospital, Nijmegen (21); J.W. Arends, H. de Vries, A. Hemelsvan der Lans, A. Imholz, Deventer Hospital, Deventer (40); I. Burgmans, C.I. Perre, T. van Dalen, J. van Gorp, D. ten Bokkel Huinink, P. Thunissen, Diakonessenhuis, Utrecht (4); J. Roussel, C. Bernhart, E. Weltevreden, S. Radema, Gelre Hospitals, Apeldoorn (21); R. Roumen, P. Reemst, A. Brands, K. Vercoelen, M. van Beek, W. Dercksen, G. Vreugdenhil, Maxima Medical Centre, Eindhoven/Veldhoven (114); T. van der Sluis, A. Stam, Lotus Sterk, Medisch Spectrum Twente, Enschede (6); M.J. Baas-Vrancken Peeters, H. Oldenburg, I. Eekhout, H. Hauer, J. Schornagel, Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital, Amsterdam (172); H. van der Mijle, D. de Vries, I. Kruithof, S. Hovenga, Nij Smellinghe Hospital; Drachten (18); B. de Valk, M. de Boer, P.J. Borgstein, A. Walter, Onze Lieve Vrouwe Gasthuis, Amsterdam (16); C. van Krimpen, P.W. de Graaf, C. van de Pol, N. van Holsteijn, A. van Leeuwen, M.M.E.M. Bos, E. Maartense, Reinier de Graaf Group, Delft (124); A. Zeillemaker, G. van Leeuwen, J. Calame, W. Molendijk, G. Jonkers, F. Cluitmans, Rijnland Hospital, Leiderdorp (59); and F. Bellot, G. Heuff, A. Tanka, P. Hoekstra, K. van de Stadt, J. Schrama, Spaarne Hospital, Hoofddorp (103). 44 Chapter 3

47 References 1 Esserman LJ, Shieh Y, Rutgers EJ, Knauer M, Retel VP, Mook S, Glas AM, Moore DH, Linn S, van Leeuwen FE, van t Veer. Impact of mammographic screening on the detection of good and poor prognosis breast cancers. Breast Cancer Res Treat 2011; 130: Early Breast Cancer Trialists Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 365: Bedard PL, Cardoso F. Can some patients avoid adjuvant chemotherapy for early-stage breast cancer? Nat Rev Clin Oncol 2011; 8: Ross JS, Hatzis C, Symmans WF, Pusztai L, Hortobagyi GN. Commercialized multigene predictors of clinical outcome for breast cancer. Oncologist 2008; 13: Bogaerts J, Cardoso F, Buyse M, Braga S, Loi S, Harrison JA, Bines J, Mook S, Decker N, Ravdin PM, Therasse P, Rutgers EJ et al. Gene signature evaluation as a prognostic tool: challenges in the design of the MINDACT trial. Nat Clin Pract Oncol 2006; 3: Bueno-de-Mesquita JM, van Harten WH, Retel VP, van t Veer LJ, van Dam FS, Karsenberg K, Douma KF, van Tinteren H, Peterse JL, Wesseling J, Wu TS, Atsma D et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007; 8: Mook S, Schmidt MK, Weigelt B, Kreike B, Eekhout I, van de Vijver MJ, Glas AM, Floore A, Rutgers EJ, van t Veer LJ. The 70-gene prognosis signature predicts early metastasis in breast cancer patients between 55 and 70 years of age. Ann Oncol 2010; 21: Glas AM, Floore A, Delahaye LJ, Witteveer AT, Pover RC, Bakx N, Lahti-Domenici JS, Bruinsma TJ, Warmoes MO, Bernards R, Wessels LF, van t Veer LJ. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006; 7: van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW, Schreiber GJ, Peterse JL, Roberts C, Marton MJ, Parrish M, Atsma D et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: Kwaliteitsinstituut voor de Gezondheidszorg CBO VvlK. Adjuvante Systemische Therapie voor het Operabel Mammacarcinoom. Richtlijn Behandeling van het Mammacarcinoom 2004; Olivotto IA, Bajdik CD, Ravdin PM, Speers CH, Coldman AJ, Norris BD, Davis GJ, Chia SK, Gelmon KA. Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005; 23: Ravdin PM, Siminoff LA, Davis GJ, Mercer MB, Hewlett J, Gerson N, Parker HL. Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 2001; 19: Hudis CA, Barlow WE, Costantino JP, Gray RJ, Pritchart KI, Chapman JA, Sparano JA, Hunsberger S, Enos RA, Gelber RD, Zujewski JA. Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: the STEEP system. J Clin Oncol 2007; 25: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C, Meyers C, de Graaf PW, Bos MM, Hart AA, Rutgers EJ, Peterse JL et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009; 117: Clarke M, Coates AS, Darby SC, Davies C, Gelber RD, Godwin J, Goldhirsch A, Gray R, Peto R, Pritchard KI, Wood WC. Adjuvant chemotherapy in oestrogen-receptor-poor breast cancer: patientlevel meta-analysis of randomised trials. Lancet 2008; 371: Peto R, Davies C, Godwin J, Pan HC, Clarke M, Cutter D, Darby S, McGale P, Taylor C, Wang YC, Bergh J, Di Leo A et al. Comparisons between different polychemotherapy regimens for early breast cancer: meta-analyses of long-term outcome among 100,000 women in 123 randomised trials. Lancet 2012; 379: year follow-up of the RASTER study 45

48 17 Davies C, Godwin J, Gray R, Clarke M, Cutter D, Darby S, McGale P, Pan HC, Taylor C, Wang YC, Dowsett M, Ingle J et al. Relevance of breast cancer hormone receptors and other factors to the efficacy of adjuvant tamoxifen: patient-level meta-analysis of randomised trials. Lancet 2011; 378: Retel VP, Joore MA, Knauer M, Linn SC, Hauptmann M, Harten WH. Cost-effectiveness of the 70- gene signature versus St. Gallen guidelines and Adjuvant Online for early breast cancer. Eur J Cancer 2010; 46: Colozza M, de Azambuja E, Cardoso F, Bernard C, Piccart MJ. Breast cancer: achievements in adjuvant systemic therapies in the pre-genomic era. Oncologist 2006; 11: Park BW, Lee S, Lee AR, Lee KH, Hwang SY. Quality of Life Differences between Younger and Older Breast Cancer Patients. J Breast Cancer 2011; 14: Chapter 3

49 DDFS 70-gene signature DDFS AOL 3 DRFI 70-gene signature DRFI AOL OS 70-gene signature OS AOL Supplementary Figure 1. Kaplan-Meier plots for DDFS, DRFI and OS by 70-gene signature or AOL 5-year follow-up of the RASTER study 47

50 Supplementary Table 1. Clinicopathological characteristics by 70-gene signature result 70-gene signature 70-gene signature Low Risk (219) High Risk (208) Total Age <35 7 (3%) 19 (9%) 26 (6%) (6%) 27 (13%) 41 (10%) (17%) 47 (23%) 84 (20%) (39%) 55 (26%) 141 (33%) (26%) 44 (21%) 100 (23%) >55 19 (9%) 16 (8%) 35 (8%) pt (TNM) pt1 (<20mm) 177 (81%) 124 (60%) 301 (70%) pt2 (>20-50mm) 42 (19%) 83 (40%) 125 (29%) pt3 (>50mm) 0 (0%) 1 (0.5%) 1 (0.2%) Histological grade Good 72 (33%) 15 (7%) 87 (20%) Intermediate 131 (60%) 73 (35%) 204 (48%) Poor 16 (7%) 120 (58%) 136 (32%) Histological type Ductal 162 (74%) 183 (88%) 345 (81%) Lobular 38 (17%) 9 (4%) 47 (11%) Other 16 (7%) 15 (7%) 31 (7%) Unknown 3 (1%) 1 (0.5%) 4 (1%) ER status Negative 3 (1%) 82 (39%) 85 (20%) Positive 216 (99%) 126 (61%) 342 (80%) PgR status Negative 33 (15%) 100 (48%) 133 (31%) Positive 186 (85%) 107 (51.5%) 293 (68.6%) Unknown 0 (0%) 1 (0.5%) 1 (0.2%) HER2 status Negative 197 (90%) 161 (77%) 358 (84%) Positive 9 (4%) 39 (19%) 48 (11%) Unknown 13 (6%) 8 (4%) 21 (5%) CBO 2004 Low Risk 167 (76%) 76 (37%) 243 (57%) High Risk 52 (24%) 132 (63%) 184 (43%) Adjuvant! Online Low Risk 95 (43%) 37 (18%) 132 (31%) High Risk 124 (57%) 171 (82%) 295 (69%) Chemotherapy None 186 (85%) 39 (19%) 225 (53%) FEC/FAC* 25 (11%) 108 (52%) 133 (31%) AC** 7 (3,5%) 26 (12%) 33 (8%) TAC*** 0 (0%) 20 (10%) 20 (5%) AC-Paclitaxel 1 (0,5%) 15 (7%) 16 (4%) *Chemotherapy regimen consisting of fluorouracil, cyclophosphamide and either adriamycine or epirubicine ** Adriamycine and cyclophosphamide *** Docetaxel, adriamycine and cyclophosphamide 48 Chapter 3

51 Supplementary Table 2. Characteristics of patients with one or more events Type of Event IHC subtype 70-gene signature low risk 70-gene signature high risk Locoregional event ER+HER2-4 3 ER+HER ER-HER2-0 1 ER-HER ER+HER2 unknown 0 1 Median age at diagnosis 42 (27-46) 49 (33-59) (range) Distant metastasis event ER+HER2-6* 5 ER+HER2+ 1* 2 # ER-HER2-0 8 ER-HER Median age at diagnosis 46 (39-57) 50 (34-59) (range) Breast cancer-specific death ER+HER2-2 2 ER+HER ER-HER2-0 4 ER-HER Median age at diagnosis ND (46-51) 51 (45-59) (range) Death due to other causes 2 0 Contralateral breast cancer Δ 10 4 Second primary tumor IHC = immunohistochemistry One patient first developed an ipsilateral axillary recurrence, 10 months later followed by a contralateral breast cancer. *Of these 7 70-gene signature low risk cases, 4 cases were also low risk according to AOL. Of the three AOL high risk cases, only one did not receive any form of AST. This case developed a recurrence at 82 months of follow-up. # One case was AOL low risk and 70-gene signature high risk. No AST was administered. She had a recurrence at 5 months of follow-up. One patient died due to a cardiac cause. This patient had only received adjuvant radiotherapy of the breast. She had no signs of breast cancer recurrence. She had had an ER+HER2- breast cancer at the age of 57. The second patient died of rightsided primary lung cancer, proven by histology and ER-negative immunohistochemistry, two years after the primary diagnosis of invasive lobular breast cancer, ER+HER2-. She had only received adjuvant radiotherapy on the right breast after breast conserving therapy. She had stopped smoking one year before the diagnosis of breast cancer. Δ Out of patients with low risk 70-gene signature, only one had received AST. Of 4 high risk 70-gene signature patients, 3 had received AST. One AML (no adjuvant chemotherapy) and one lung cancer in the 70-gene signature low risk group. Four lung cancers, 2 colorectal cancers and 1 carcinoid in the 70-gene signature high risk group. 5-year follow-up of the RASTER study 49

52 Supplementary Table 3. Kaplan-Meier risk estimations for DRFI using different AOL cutoffs AOL cutoff 10-years OS AOL Number of patients 5-years DRFI (%) (95% CI) 85% Low 253 (59.3%) 97 ( ) High 174 (40.7%) 90 ( ) 90% Low 132 (30.9%) 97 ( ) High 295 (69.1%) 93 ( ) 88% (ER+) / 92% (ER-) Low 194 (45.4%) 98 ( ) High 233 (54.6%) 92 ( ) 95% Low 19 (4.4%) 100 ( ) High 408 (95.6%) 94 ( ) 50 Chapter 3

53 5-year follow-up of the RASTER study 51 3

54 Chapter 4

55 Optimized outcome prediction in breast cancer by combining the 70-gene signature with clinical risk prediction algorithms Submitted Caroline A. Drukker Matthijs V. Nijenhuis Jolien M. Bueno de Mesquita Valesca P. Retèl Wim H. van Harten Harm van Tinteren Marjanka K. Schmidt Laura J. van t Veer Gabe S. Sonke Emiel J.Th. Rutgers Marc J. van de Vijver Sabine C. Linn

56 Abstract Background Clinical guidelines for breast cancer treatment differ in their selection of patients at a high risk of recurrence who are eligible to receive adjuvant systemic treatment (AST). The 70-gene signature is a molecular tool to better guide AST decisions. The aim of this study was to evaluate whether adding the 70-gene signature to clinical risk prediction algorithms can optimize outcome prediction and consequently treatment decisions in early stage, node-negative breast cancer patients. Methods A 70-gene signature was available for 427 patients participating in the RASTER study (ct1-3n0m0). Median follow-up was 61.6 months. Based on 5-year distant-recurrence-free-interval (DRFI) probabilities survival Areas Under the Curve (AUC) were calculated and compared for risk estimations based on the six clinical risk prediction algorithms: Adjuvant! Online (AOL), Nottingham Prognostic Index (NPI), St. Gallen (2003), the Dutch National guidelines (CBO 2004 and NABON 2012) and PREDICT plus. Also, survival AUC were calculated after adding the 70- gene signature to these clinical risk estimations. Results Systemically untreated patients with a high clinical risk estimation but a low risk 70-gene signature had an excellent 5-year DRFI varying between 97.1% and 100%, depending on the clinical risk prediction algorithms used in the comparison. The best risk estimation was obtained in this cohort by adding the 70-gene signature to CBO 2012 (AUC: 0.644) and PREDICT (AUC: 0.662). Clinical risk estimations by all clinical risk prediction algorithms improved by adding the 70-gene signature. Conclusion Patients with a low risk 70-gene signature have an excellent survival, independent of their clinical risk estimation. Adding the 70-gene signature to clinical risk prediction algorithms improves risk estimations and therefore might improve the identification of early stage node-negative breast cancer patients for whom AST has limited value. In this cohort, the PREDICT plus tool in combination with the 70-gene signature provided the best risk prediction. 54 Chapter 4

57 Introduction For the past decade the selection of early stage breast cancer patients who are at a high risk of recurrence and eligible to receive adjuvant systemic treatment (AST) is based on clinicopathological factors, such as age, tumor size, nodal status, histological grade, and hormone receptor status. Several clinical risk prediction algorithms used in online tools and guidelines, such as Adjuvant! Online (AOL), the Nottingham Prognostic Index (NPI), the St. Gallen expert panel recommendations of 2003 and the Dutch National guidelines of 2004 and 2012 use these factors in specific algorithms for risk estimations and AST recommendations. 1-6 A relatively new online tool for outcome prediction in breast cancer patients is PREDICT plus. 7 This tool not only uses the clinicopathological factors mentioned above, but also incorporates Human Epidermal growth factor Receptor 2 (HER2) status and method of detection. Both of these factors have proven to be independent prognostic factors in overall and breast cancer specific survival. 7,8 Even with the aid of these clinical risk prediction algorithms, individual risk assessment remains challenging. Each of these clinical risk prediction algorithms may define a slightly different group of patients at a low or high risk, which are partly non-overlapping. This indicates that it is unclear which tool or guideline has the highest prognostic accuracy for the individual patient. 1,5,6,9 Moreover, online tools such as AOL provide a survival probability without stratification into high versus low risk. The choice for a specific cutoff point in risk clearly influences the concordance with other tools. 10 Gene-expression classifiers have been developed and validated on historic data to refine clinical risk estimations and related AST recommendations. 11,12 One of these classifiers is the 70-gene signature (MammaPrint, Agendia NV, Amsterdam, the Netherlands). 13,14 Between 2004 and 2006 the 70-gene signature has been assessed in the first prospective study using a gene-expression classifier as a risk estimation tool in addition to clinicopathological factors to determine the need for AST. A considerable discrepancy in risk estimations among different clinical guidelines and the 70-gene signature was observed. 9,15 Recently, the 5-year follow-up data of the RASTER study were reported showing an excellent distant-recurrence-free interval (DRFI) of 97% for patients with a low risk 70-gene signature. Patients with a high risk 70-gene signature showed a DRFI of 92%. 16 When compared to AOL, 70-gene signature low AOL high risk patients who did not receive any AST showed a DRFI of 100%. This indicates that omission of chemotherapy in these patients may not compromise outcome. Up to the evaluated 5 year median survival the number of events is small and the follow-up time relatively short. However, AOL is not the only risk estimation tool used in clinical practice today. Additionally, the 70-gene signature is more likely to be added to clinical risk prediction algorithms instead of replacing them. Therefore, we evaluated whether adding the 70-gene signature to clinical risk prediction algorithms can improve individual outcome prediction in early stage, node-negative breast cancer patients gene signature combined with clinical risk estimations 55

58 Patients and methods The RASTER study design, patient eligibility criteria and study logistics have been described elsewhere ( trials.com/isrctn ). 15 In short, 812 female patients were enrolled in 16 hospitals in the Netherlands. 427 patients were postoperatively eligible and for them a 70-gene signature (MammaPrint, Agendia NV) was obtained. All patients were between years old and had a histologically confirmed unilateral, unifocal, primary operable, invasive adenocarcinoma of the breast (ct1-3n0m0). All patients were primarily surgically treated with either breast conserving surgery or mastectomy. To ensure routine clinical practice, the initial histopathology data were used for clinical risk assessment by the treating physician and in the statistical analysis, without central review of the paraffin-embedded tumor samples. Details on tumor grading, assessment of hormone receptor status and HER2 status, RNA extraction and microarray analysis have been described elsewhere. 15 Decisions on whether or not to treat with AST (comprising chemotherapy and/or endocrine therapy) in the RASTER study were based on the Dutch national guidelines of 2004, the 70-gene signature, and doctors and patients preferences. 15 More detailed insight on the follow-up data of this cohort is described elsewhere. 16 Clinical risk prediction algorithms Hereafter, risk assessment by use of clinicopathological factors is referred to as clinical risk. Guidelines used in this study to assess clinical risk were Adjuvant! Online (AOL), Nottingham Prognostic Index (NPI), the St. Gallen expert panel recommendations (2003, current at the time the RASTER study was conducted), the Dutch National guidelines (2004, current at the time the RASTER study was conducted, and 2012) and PREDICT plus. Adjuvant! Online software, version 8.0, calculates the 10-year survival probabilities based on the age of the patient, tumor size, tumor grade, estrogen receptor (ER) status, and nodal status. 5,10 Patients were considered high risk if their calculated 10-year survival probability was less than 90%. 15 This cutoff was also used in the RASTER study and similar to the cutoff used in the MINDACT trial. The NPI computes a score with the algorithm: 0.2*size (cm) + grade + nodal status. A moderate or high risk was defined as a score greater than ,17 The St. Gallen expert panel of 2003 recommended to define low clinical risk as ER-positive or progesterone receptor(pr)-positive disease (or both) and all of the following criteria: tumor size of 2 cm or smaller, grade 1, and age 35 years or over. All other tumors were deemed to be associated with a moderate or high risk of distant metastasis and death. 2 The 2004 Dutch National guidelines define high clinical risk for node-negative breast cancer as age 35 years or younger (except for tumors grade 1 of 10 mm or smaller), a tumor of grade 3 and 10 mm or larger, or grade 2 and 20 mm or larger, and every tumor larger than 30 mm. Adjuvant endocrine treatment was advised only in clinically high risk patients with hormone receptor-positive tumors in combination with chemotherapy. 10 AST was justified for patients with a 10-year survival probability of less than 80%. The less restrictive Dutch guidelines of Chapter 4

59 define high clinical risk for node-negative breast cancer as age under 35 years except for tumors grade 1 of 10 mm or smaller, or age 35 years or older with a tumor of grade 2 or higher and 10-20mm in size, and every tumor larger than 20 mm. According to this 2012 guideline AST was justified for patients with a 10-year survival probability of less than 85%. The online PREDICT plus tool estimates the 5 and 10-year survival probabilities based on the age of the patient, method of detection, tumor size, tumor grade, number of positive nodes, ER and HER2 status. 7 We defined a 5-year survival probability of <95%, which is in line with the cutoffs used for Adjuvant! Online. All clinicopathological factors used by the guidelines mentioned above were summarized elsewhere. 18 In our analyses, a moderate or high clinical risk was considered an indication for adjuvant systemic treatment. Statistical analysis We estimated 5-year distant-recurrence-free interval (DRFI), comprising distant recurrence and death from breast cancer. 19 Survival curves were constructed using the Kaplan-Meier method and compared using the log-rank test. Survival ROC and AUC (c-index) analyses were performed to evaluate the additional value of the 70-gene signature to the clinical guidelines described in this manuscript. An ANOVA test was used to compare the model before and after adding the 70-gene signature. A significant finding was defined as a p-value below Analyses were performed using SAS version 9.2 and R version Results Patient and tumor characteristics, AST and outcome stratified by 70-gene signature Patient and tumor characteristics were described elsewhere. 15 After a median follow-up time of 61.6 months, 24 DRFI events occurred. Eleven patients died of whom nine due to breast cancer. The 5-year DRFI probabilities for 70-gene signature low risk (n=219) and high risk (n=208) patients were 97.0% (95%CI: ) and 91.7% (95%CI: ) (p=0.03), respectively (Supplementary Figure 1). 16 Additional value of 70-gene signature to clinical risk assessment Adding the 70-gene signature to clinical risk prediction algorithms improved outcome prediction. For most guidelines this was a borderline significant improvement of the c-index (Table 1). The c-index was highest for PREDICT plus (0.627), followed by NPI (0.591) and the Dutch National guidelines of 2004 (0.586). Adding the 70-gene signature improved the model to for NPI (p=0.05) and to for the Dutch national guidelines of 2004 (p=0.04). The best risk predictions were achieved when using PREDICT plus (0.662) or the Dutch guidelines of 2012 (0.644) in combination with the 70-gene signature. The c-index for AOL was lowest, before (0.532) and after adding the 70-gene signature (0.619). 70-gene signature combined with clinical risk estimations 57

60 Table 1. Survival AUC and proportions of low risk for clinicopathological guidelines and in combination with the 70-gene signature Low risk c-index guideline Low risk 70-gene c-index guideline + p-value guideline (95% CI) signature 70-gene signature AOL 132 (30.9%) ( (51.3%) ( ) 0.03 NPI 248 (58.1%) ( ) 219 (51.3%) ( ) 0.05 St. Gallen 73 (17.1%) ( ) 219 (51.3%) ( ) 0.05 CBO (56.9%) ( ) 219 (51.3%) ( ) 0.04 CBO (29.0%) ( ) 219 (51.3%) ( ) 0.05 PREDICT plus 228 (53.4%) ( ) 219 (51.3%) ( ) 0.27 Blue = proportion of low risk increased with the 70-gene signature Orange = proportion of low risk decreased with the 70-gene signature Discordance between clinical risk assessment and the 70-gene signature Discordant risk estimations occurred in 37% of the cases (161/427) for AOL, 27% for NPI (117/427), 39% for St. Gallen (168/427), 30% for the Dutch National guidelines of 2004 (128/427), 39% for the guidelines of 2012 (167/427) and 25% for PREDICT plus (107/427)(Table 2; Figure 1). Most discordant cases were 70-gene signature low risk and clinically high risk; 29% for AOL (124/427), 10% for NPI (44/427), 37% for St. Gallen (157/427), 12% for the Dutch National guidelines of 2004 (52/427), 31% for the guidelines of 2012 (131/427) and 11% for PREDICT plus at 5 years (49/427). Table 2 summarizes the AST given in the different categories stratified by 70-gene signature and clinical risk. When the 70-gene signature was used, 20% less patients would be eligible to receive ACT compared to AOL, 34% less compared to St. Gallen, 6% less compared to the Dutch guidelines of 2004 and 22% less compared to the guidelines of The 70-gene signature identifies 7% more patients eligible to receive ACT compared to NPI and 2% more compared to PREDICT plus. 58 Chapter 4

61 Figure 1. Risk estimations per case stratified by clinical risk prediction algorithms and the 70-gene signature. Cases were ordered according to their 70-gene signature gene signature combined with clinical risk estimations 59

62 Table 2. Distribution of patients (n=427) over the four risk categories defined by 70-gene signature and clinical risk and proportion and type of AST received per category 70-gene signature AOL No AST CT ET ET+CT Low Low 88/95 (93%) 0/95 (0%) 4/95 (4%) 3/95 (3%) High Low 5/37 (14%) 3/37 (8%) 11/37 (30%) 18/37 (49%) Low High 70/124 (56%) 1/124 (1%) 24/124 (19%) 29/124 (23%) High High 5/171 (3%) 73/171 (43%) 18/171 (11%) 75/171 (44%) 70-gene signature NPI Low Low 153/175 (87%) 0/175 (0%) 14/175 (8%) 8/175 (5%) High Low 7/73 (10%) 7/73 (10%) 23/73 (32%) 36/73 (49%) Low High 5/44 (11%) 1/44 (2%) 14/44 (32%) 24/44 (55%) High High 3/135 (2%) 69/135 (51%) 6/135 (4%) 57/135 (42%) 70-gene signature St. Gallen Low Low 59/62 (95%) 0/62 (0%) 3/62 (5%) 0/62 (0%) High Low 2/11 (18%) 0/11 (0%) 5/11 (45%) 4/11 (36%) Low High 99/157 (63%) 1/157 (1%) 25/157 (16%) 32/157 (20%) High High 8/196 (4%) 76/196 (39%) 23/196 (12%) 89/196 (45%) 70-gene signature CBO 2004 Low Low 152/167 (91%) 0/167 (0%) 13/167 (8%) 2/167 (1%) High Low 8/76 (11%) 10/76 (13%) 25/76 (33%) 33/76 (43%) Low High 6/52 (12%) 1/52 (2%) 15/52 (29%) 30/52 (58%) High High 2/132 (2%) 66/132 (50%) 4/132 (3%) 60/132 (45%) 70-gene signature CBO 2012 Low Low 83/88 (94%) 0/88 (0%) 5/88 (6%) 0/88 (0%) High Low 4/36 (11%) 6/36 (17%) 14/36 (39%) 12/36 (33%) Low High 75/131 (57%) 1/131 (1%) 23/131 (18%) 32/131 (24%) High High 6/172 (3%) 70/172 (41%) 15/172 (9%) 81/172 (47%) 70-gene signature PREDICT plus Low Low 141/170 (83%) 0/170 (0%) 16/170 (9%) 13/170 (8%) High Low 3/58 (5%) 1/58 (2%) 22/58 (38%) 32/58 (55%) Low High 17/49 (35%) 1/49 (2%) 12/49 (25%) 19/49 (39%) High High 7/150 (5%) 75/150 (50%) 7/150 (5%) 61/150 (41%) AST=adjuvant systemic therapy; CT=adjuvant chemotherapy; ET=adjuvant endocrine therapy; CBO=Dutch National guidelines The 5-year DRFI probabilities for AOL low risk (n=132) and high risk (n=295) patients were 96.7% (95%CI: ) and 93.4% (95%CI: ), respectively (p=0.24). For NPI low risk (n= 248) and high risk (n=179) patients the 5-year DRFI probabilities were 96.7% (95%CI: ) and 91.3% (95%CI: ) (p=0.03). The St. Gallen low risk (n=73) and high risk (n=353) patients showed 5-year DRFI probabilities of 98.5% (95%CI: ) and 93.5% (95%CI: )(p=0.08). For the Dutch National guidelines of 2004 low risk (n=243) and high risk (n=184) patients the 5-year DRFI probabilities were 96.6% (95%CI: ) and 91.5% (95%CI: ), respectively (p=0.11), while for the Dutch National guidelines of 2012 low risk (n=124) and high risk (n=303) patients the 5-year DRFI probabilities were 99.2% (95%CI: ) and 92.4% (95%CI: )(p=0.02). The 5-year prediction of PREDICT plus low risk (n=228) and high risk (n=199) patients showed DRFI probabilities of 96.8% (95%CI: ) and 91.7% (95%CI: ), respectively (p=0.004)(figure 2). Table 3 summarizes DRFI probabilities according to the combined risk categories. 60 Chapter 4

63 70-gene signature AOL 70-gene signature NPI 4 70-gene signature St. Gallen 70-gene signature CBO gene signature CBO gene signature PREDICT plus Figure 2. Five-year outcome of systemic therapy-naïve patients with a low risk 70-gene signature AOL=Adjuvant! Online, NPI=Nottingham Prognostic Index; CBO=Dutch National guidelines 70-gene signature combined with clinical risk estimations 61

64 Table 3. Kaplan-Meier risk estimations for DRFI and DDFS according to 70-gene signature and clinical risk stratification 70-gene signature AOL ACT 5-year DRFI (%) (95% CI) Low Low 3/95 (3%) 95.3 ( ) High Low 21/37 (57%) 100 ( ) Low High 30/124 (24%) 98.4 ( ) High High 148/171 (87%) 89.8 ( ) 70-gene signature NPI Low Low 8/175 (5%) 97.4 ( ) High Low 43/73 (59%) 95.3 ( ) Low High 25/44 (57%) 95.5 ( ) High High 126/135 (93%) 89.9 ( ) 70-gene signature St. Gallen Low Low 0/62 (0%) 98.3 ( ) High Low 4/11 (36%) 100 ( ) Low High 33/157 (21%) 96.5 ( ) High High 165/196 (84%) 91.2 ( ) 70-gene signature CBO 2004 Low Low 2/167(1%) 97.3 ( ) High Low 43/76 (57%) 95.5 ( ) Low High 31/52 (60%) 96.2 ( ) High High 126/132 (95%) 89.7 ( ) 70-gene signature CBO 2012 Low Low 0/88 (0%) 98.8 ( ) High Low 18/36 (50%) 100 ( ) Low High 33/131 (25%) 95.8 ( ) High High 151/172 (88%) 89.8 ( ) 70-gene signature PREDICT plus Low Low 13/170 (8%) 98.0 ( ) High Low 33/58 (57%) 93.9 ( ) Low High 20/49 (41%) 93.9 ( ) High High 136/150 (91%) 91.0 ( ) ACT = Adjuvant Chemotherapy; DRFI= Distant Recurrence Free Interval; DDFS= Distant Disease Free Survival; CBO=Dutch National guidelines Subgroup analyses of therapy-naïve patients Of the patients who had a low risk 70-gene signature 85% did not receive adjuvant chemotherapy. Only 27% of the 70-gene signature low risk patients received adjuvant endocrine therapy. Among the low risk systemically untreated patients, no significant difference was seen for most clinical risk algorithms (p=0.29 for AOL, p=0.66 for NPI, p=0.37 for St. Gallen, p=0.65 for the 2004 and p=0.14 for the 2012 Dutch National guidelines) between patients with a concordant low risk assessment and patients with a 70-gene signature low risk result but a high risk assessment by one or more of the clinical indexes (Figure 1). Only the PREDICT plus tool shows that patients with a concordant low risk assessment (n=141) at 5 years have a significantly better DRFI survival probability compared to patients with a low risk 70-gene signature and a high risk according to PREDICT plus (n=17)(p=0.002). 62 Chapter 4

65 Discussion The RASTER study was the first study to prospectively evaluate the outcome of patients for whom the 70-gene signature was used for risk estimations and AST recommendations. The recently published 5-year follow-up data of this study provide the opportunity to evaluate the additional value of a gene-expression classifier to risk estimations based on clinicopathological factors incorporated in clinical tools and guidelines. Of all clinical risk prediction algorithms used in this study, the online PREDICT plus tool provided the best risk estimation. Addition of the 70-gene signature to either the PREDICT plus tool or the Dutch National guidelines of 2012 resulted in the best risk estimations in this cohort. Interestingly, AOL showed the lowest c-index before and after adding the 70-gene signature. This might be explained by the fact that this guideline does not incorporate HER2 status, while the Dutch guidelines of 2012 and PREDICT plus do take this clinicopathological factor into account. In addition, as AOL does not provide a classification into high versus low risk, the choice for a specific cutoff point may influence these results. Previous analyses already showed that method of detection is an independent prognostic factor in breast cancer specific and overall survival. The fact that the PREDICT plus tool takes the method of detection into account may explain why this risk prediction algorithm performs so well in this cohort. When solely using the 70-gene signature, the number of patients at high risk of recurrence who are eligible for adjuvant chemotherapy would be reduced by 20% compared to AOL. As a similar comparison was made in the MINDACT trial (AOL in MINDACT does include HER2), one can hypothesize that a similar reduction in chemotherapy will be seen in this large, randomized controlled phase 3 trial. Analyses of the first 800 patients included in the MINDACT trial show a similar possible reduction in adjuvant chemotherapy of 18% (141/800). Overall, the 5-year outcome of this cohort of patients for whom the 70-gene signature result was prospectively used to guide AST decisions was favorable. One should take into consideration that a substantial proportion of patients, 39% (168/427) of this cohort, did not receive any form of AST. Most importantly, the 5-year DRFI probabilities were excellent for patients who were clinically at high risk but had a low risk 70-gene signature, even in the absence of any AST. 16 Therefore, omission of chemotherapy in patients with a low risk 70-gene signature appeared safe, even in case of a high risk estimation by one or more of the clinical guidelines. A larger number of patients in the untreated subgroups and longer follow-up is needed to draw firm conclusions. The only tool that was able to select patients at a slightly higher risk of recurrence among the 70-gene signature low risk patients was the PREDICT plus tool. However, in this subgroup the number of patients (n=17) was too low to draw any firm conclusions. A larger cohort is necessary to evaluate the additional prognostic value of the 70-gene signature to PREDICT plus tool. An advantage, but also a limitation of this study is that the actual treatment decisions were based on the Dutch guidelines of 2004, the 70-gene signature result and preferences of doctors and patients. The study design provides an optimal reflection of daily clinical practice, but subtle selection mechanisms may 4 70-gene signature combined with clinical risk estimations 63

66 be present and may have influenced our results. Another possible limitation is that all clinical tools and guidelines included in our analyses use slightly different definitions of high and low risk. These differences create an additional group of patients for whom the guidelines provide discordant risk estimations. Also, some guidelines base their risk assessment on 5-year survival probabilities, while others on 10-year survival probabilities. In our analyses we were unable to adjust for these differences, which makes a head-to-head comparison more difficult to interpret. Still, the guidelines as used in this study reflect the way they are used in current daily clinical practice. The c-indexes reported here leave room for improvement and this again underlines the need for more accurate personalized breast cancer care. Also, it should be kept in mind that the results of this study are based on a case mix of relatively young (<61 years) breast cancer patients. Finally, central pathology revision might have changed the results, since an earlier report showed that for 8% of the patients AOL risk estimations would change based on revised pathology. 20 In conclusion, our results indicate that adding the 70-gene signature clinical guidelines with the 70-gene signature improves risk estimations and therefore may help to identify early stage node-negative breast cancer patients for whom limited adjuvant systemic therapy might be appropriate and for whom overtreatment can be avoided. In this cohort, PREDICT plus appeared to be a promising tool to identify patients for whom limited adjuvant systemic therapy in case of early stage node-negative disease might be appropriate. Contributors SL, MvdV, WvH and LvtV were responsible for the RASTER study design and development of the protocol. JBdM coordinated the RASTER study. JBdM, VR, CD and MN took part in data collection. CD and HvT performed the data analysis. CD, HvT, ER, MKS and SL took part in data interpretation and manuscript writing. All authors were involved in reviewing the manuscript. Conflict of Interest The RASTER study was financially supported the Dutch Health Care Insurance Board (CVZ). LvtV and MvdV are named inventors on the patent for the 70-gene signature used in this study. LvtV reports being shareholder in and part-time employed by Agendia NV, the commercial company that markets the 70-gene signature as MammaPrint. LvtV was supported by the Dutch Genomics Initiative Cancer Genomics Centre. Acknowledgements We are indebted to the women who participated in the RASTER study; to the doctors, nurses, and data managers from the participating hospitals in the Netherlands that enrolled patients in the RASTER-study and contributed to the collection of follow-up data. 64 Chapter 4

67 References 1 D Eredita G, Giardina C, Martellotta M, Natale T, Ferrarese F (2001) Prognostic factors in breast cancer: the predictive value of the Nottingham Prognostic Index in patients with a long-term followup that were treated in a single institution. Eur J Cancer 37: Goldhirsch A, Glick JH, Gelber RD, Coates AS, Senn HJ (2001) Meeting highlights: International Consensus Panel on the Treatment of Primary Breast Cancer. Seventh International Conference on Adjuvant Therapy of Primary Breast Cancer. J Clin Oncol 19: Integraal Kankercentrum Nederland: NABON richtlijn mammacarcinoom Kwaliteitsinstituut voor de Gezondheidszorg CBO VvlK: Adjuvante Systemische Therapie voor het Operabel Mammacarcinoom. Richtlijn Behandeling van het Mammacarcinoom. 2004; Olivotto IA, Bajdik CD, Ravdin PM, Speers CH, Coldman AJ, Norris BD, Davis GJ, Chia SK, Gelmon KA (2005) Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 23: Goldhirsch A, Wood WC, Gelber RD, Coates AS, Thurlimann B, Senn HJ (2003) Meeting highlights: updated international expert consensus on the primary therapy of early breast cancer. J Clin Oncol 21: Wishart GC, Bajdik CD, Dicks E, Provenzano E, Schmidt MK, Sherman M, Greenberg DC, Green AR, Gelmon KA, Kosma VM, Olson JE, Beckmann MW, Winqvist R, Cross SS, Severi G, Huntsman D, Pylkas K, Ellis I, Nielsen TO, Giles G, Blomqvist C, Fasching PA, Couch FJ, Rakha E, Foulkes WD, Blows FM, Begin LR, van t Veer LJ, Southey M, Nevanlinna H, Mannermaa A, Cox A, Cheang M, Baglietto L, Caldas C, Garcia-Closas M, Pharoah PD (2012) PREDICT Plus: development and validation of a prognostic model for early breast cancer that includes HER2. Br J Cancer 107: Mook S, van t Veer LJ, Rutgers EJ, Ravdin PM, van de Velde AO, van Leeuwen FE, Visser O, Schmidt MK (2011) Independent prognostic value of screen detection in invasive breast cancer. J Natl Cancer Inst 103: Bueno-de-Mesquita JM, Sonke GS, van de Vijver MJ, Linn SC (2011) Additional value and potential use of the 70-gene prognosis signature in node-negative breast cancer in daily clinical practice. Ann Oncol 22: Ravdin PM, Siminoff LA, Davis GJ, Mercer MB, Hewlett J, Gerson N, Parker HL (2001) Computer program to assist in making decisions about adjuvant therapy for women with early breast cancer. J Clin Oncol 19: Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas AM, d Assignies MS, Bergh J, Lidereau R, Ellis P, Harris A, Bogaerts J, Therasse P, Floore A, Amakrane M, Piette F, Rutgers E, Sotiriou C, Cardoso F, Piccart MJ (2006) Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 98: Ross JS, Hatzis C, Symmans WF, Pusztai L, Hortobagyi GN (2008) Commercialized multigene predictors of clinical outcome for breast cancer. Oncologist 13: van t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415: van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW, Schreiber GJ, Peterse JL, Roberts C, Marton MJ, Parrish M, Atsma D, Witteveen A, Glas A, Delahaye L, van der Velde T, Bartelink H, Rodenhuis S, Rutgers ET, Friend SH, Bernards R (2002) A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 347: Bueno-de-Mesquita JM, van Harten WH, Retel VP, van t Veer LJ, van Dam FS, Karsenberg K, Douma KF, van Tinteren H, Peterse JL, Wesseling J, Wu TS, Atsma D, Rutgers EJ, Brink G, Floore AN, Glas AM, Roumen RM, Bellot FE, van Krimpen C, Rodenhuis S, van de Vijver MJ, Linn SC (2007) Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 8: gene signature combined with clinical risk estimations 65

68 16 Drukker CA, Bueno-de-Mesquita JM, Retel VP, van Harten WH, van Tinteren H, Wesseling J, Roumen RM, Knauer M, t Veer LJ, Sonke GS, Rutgers EJ, van de Vijver MJ, Linn SC (2013) A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 133: Todd JH, Dowle C, Williams MR, Elston CW, Ellis IO, Hinton CP, Blamey RW, Haybittle JL (1987) Confirmation of a prognostic index in primary breast cancer. Br J Cancer 56: Drukker CA, van den Hout HC, Sonke GS, Brain E, Bonnefoi H, Cardoso F, Goldhirsch A, Harbeck N, Honkoop A.H., Koornstra RH, van Laarhoven H.W.M., Portielje J.E.A., Schneeweiss A, Smorenburg C.H., Stouthard J., Linn SC, Schmidt MK: Risk estimations and treatment decisions in early stage breast cancer; agreement among oncologists and the impact of the 70-gene signature. (accepted EJC jan 2014) 19 Hudis CA, Barlow WE, Costantino JP, Gray RJ, Pritchard KI, Chapman JA, Sparano JA, Hunsberger S, Enos RA, Gelber RD, Zujewski JA (2007) Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: the STEEP system. J Clin Oncol 25: Bueno-de-Mesquita JM, Nuyten DS, Wesseling J, van Tinteren H, Linn SC, van de Vijver MJ (2010) The impact of inter-observer variation in pathological assessment of node-negative breast cancer on clinical risk assessment and patient selection for adjuvant systemic treatment. Ann Oncol 21: Chapter 4

69 70-gene signature Adjuvant! Online 4 Nottingham Prognostic Index st. Gallen 2003 CBO 2004 CBO 2012 PREDICT 5 year Supplementary Figure 1. Five-year outcome of RASTER patients stratified by 70-gene signature and clinical risk prediction algorithms. 70-gene signature combined with clinical risk estimations 67

70 Chapter 5

71 Risk estimations and treatment decisions in early stage breast cancer: agreement among oncologists and the impact of the 70-gene signature Accepted by European Journal of Cancer Caroline A. Drukker Ella H.C. van den Hout Gabe S. Sonke Etienne Brain Hervé Bonnefoi Fatima Cardoso Aaron Goldhirsch Nadia Harbeck Aafke H. Honkoop Rutger H.T. Koornstra Hanneke W.M. Laarhoven Johanna E.A. Portielje Andreas Schneeweiss Carolien H. Smorenburg Jacqueline Stouthard Sabine C. Linn* Marjanka K. Schmidt* *authors contributed equallly

72 Abstract Background Clinical decision-making in patients with early stage breast cancer requires adequate risk estimation by medical oncologists. This survey evaluates the agreement among oncologists on risk estimations and adjuvant systemic treatment (AST) decisions and the impact of adding the 70-gene signature to known clinicopathological factors. Methods Twelve medical oncologists assessed 37 breast cancer cases (ct1-3n0m0) and estimated their risk of recurrence (high or low) and gave a recommendation for AST. Cases were presented in two written questionnaires sent four weeks apart. Only the second questionnaire included the 70-gene signature result. Results The level of agreement among oncologists in risk estimation (κ=0.57) and AST-recommendation (κ=0.57) was moderate in the first questionnaire. Adding the 70-gene signature result significantly increased the agreement in risk estimation to substantial (κ=0.61), while agreement in AST recommendations remained moderate (κ=0.56). Overall, the proportion of high risk was reduced with 7.4% (range: %; p<0.001) and the proportion of chemotherapy that was recommended was reduced with 12.2% (range: %; p<0.001). Conclusion Oncologists risk estimations and AST recommendations vary greatly. Even though the number of participating oncologists is low, our results underline the need for a better standardization tool in clinical decision-making, in which integration of the 70-gene signature may be helpful in certain subgroups to provide patients with individualized, but standardized treatment. 70 Chapter 5

73 Introduction Clinicopathological guidelines are used to guide adjuvant systemic treatment (AST) decisions in early stage breast cancer patients. These guidelines combine clinicopathological factors such as age, tumor size, grade, hormone-receptor status, and nodal status to estimate the risk of recurrence and provide an AST advice. Commonly used clinicopathological guidelines are Adjuvant! Online (AOL), the Sankt Gallen expert panel recommendations and the Nottingham Prognostic Index (NPI). 1,2 In the Netherlands, the Dutch Institute of Healthcare Improvement (CBO) guidelines are used most often. 3 Nevertheless, correctly estimating whether an individual patient has a high risk of recurrence and is likely to benefit from AST remains challenging. 4 Most of the guidelines consider only a small proportion of patients at a low risk of recurrence. This may result in a substantial number of patients being treated with AST while they are unlikely to derive significant benefit. 5 Each guideline mentioned above defines a partly non-overlapping group of patients at a low or high risk, which indicates that predictive accuracy for the individual patient is not high. 1,6-8 Also, online tools such as AOL that provide a survival probability instead of a low/high risk estimation can be used with different cutoffs. Therefore, a variation in risk estimations made by oncologists who are guided by different guidelines is expected. The extent of this variation remains unclear. To refine risk estimations and provide a more tailored AST recommendation for the individual patient, gene expression prognosis classifiers have been developed. 9 One of these gene expression classifiers is the 70-gene signature (MammaPrint, Agendia NV, Amsterdam, the Netherlands). 10 The first prospective study, in which the 70-gene signature was used in addition to clinical guidelines, was conducted in the Netherlands between 2004 and This microarray prognostics in breast cancer (RASTER) study showed discordance in risk estimation between the 70-gene signature and clinicopathological guidelines in one third of the patients. 11 In daily clinical practice, medical oncologists are using the 70-gene signature the same way as it was used in the RASTER study, i.e. in addition to clinicopathological guidelines. 1,11 However, the impact of the 70-gene signature on risk estimations and AST decisions in daily clinical practice is unknown. The aim of this survey was to determine the agreement among oncologists risk estimations and AST recommendations based on clinicopathological factors as are used in clinical guidelines, and to assess the impact of the 70-gene signature. 5 Methods Two written questionnaires were developed (CAD, SCL, HCvdH, MKS) and reviewed by an independent oncologist (GSS). Thirty-seven cases of breast cancer patients were presented to 29 medical oncologists specialised in breast cancer in Europe. The oncologists were chosen because Agreement among oncologists 71

74 of their area of expertise and the country they work in. We included oncologists from all over Europe to not only demonstrate the situation among oncologists in one country, but for an entire continent. The oncologists were asked to indicate their use of clinical guidelines and to give their risk estimation (high/low) and recommendation of AST (none, endocrine therapy, chemotherapy, trastuzumab or a combination) for each case. Several weeks later, the same cases were presented in a randomly changed order in a second questionnaire. In this second questionnaire, the 70-gene signature result was provided along with clinical characteristics. Cases To provide a reflection of true clinical practice, thirty-seven cases of breast cancer patients were selected from the database of the RASTER study, with a 70-gene signature result. All cases involved women < 61 years, with unilateral, histological proven, operable breast cancer (ct1-3n0m0). Of each patient tumor size, histopathological grade, histological type, mitotic index, hormone-receptor status and Human Epidermal growth factor Receptor 2 (HER2) status were described (Supplementary Table 1). The actually received treatments were not mentioned in the questionnaire. Clinical risk estimation based on Adjuvant! Online Hereafter, risk estimations using clinicopathological factors will be referred to as clinical risk. In this survey, the clinical risk estimation was first assessed using AOL version 8.0. Patients were assigned to a high clinical risk if their AOL 10-year survival probability was less than 90% based on minor problems regarding overall health status, which is the default item of the online program. 11 Of the 37 cases, 10 cases were concordant high, 12 concordant low and, 15 discordant with the 70-gene signature result. The cases are a random selection from stratification of concordant low risk, discordant and concordant high risk with the 70-gene signature result. Clinical risk estimations by other guidelines Additional risk estimations according to the St. Gallen expert panel recommendations of 2003, NPI and CBO 2004 (all versions were used at the time of the RASTER study) were assessed previously. 6,11-13 Differences among clinicopathological guidelines, tool and expert panel recommendations are summarized in Table 1. Risk estimations were concordant with the 70-gene signature and all clinical guidelines in 12 cases, six were concordant high risk and six concordant low risk. There was discordance between the 70-gene signature and at least one of the guidelines in 25 cases (68%). 72 Chapter 5

75 Table 1. Clinicopathological factors used by breast cancer guidelines and risk estimation tools to define patients at a low risk of recurrence Guideline/tool Age Size Grade Hist. type ER/PR HER2 Ki67 Nodal status Other factors Low risk is defined as Not specified ER No No Yes Co-morbidities, CT regimen AOL 8 Continuous Yes Yes Ductal, in case of other hist. type, information is available online St. Gallen <35 or 35 Yes Yes Not used ER & PR No No Node-negative None ER + and PR +, grade I, 2cm and age 35 yrs High ER and PR, grade I, low Ki67, node negative, absence of PVI, 20mm, low score on multigene assay. Not used Yes Yes Not used ER & PR No Yes Yes PVI, multigene assays St. Gallen 2009 Luminal A; ER + and PR +, HER2 -, low Ki67 Biological subtype No Yes Not used ER & PR Yes Yes Yes, more than 3+ nodes is high risk St. Gallen Pre- or post menopausal NPI 6 35 Not used Yes Yes Not used No No No Yes None [0.2 x Size] + Number of nodes + Grade; low risk = score lower than 3.4 CBO <35 or Yes Yes Not used No No No Yes None N0, 35 yrs, grade I tumor 1 cm OR >35 years, grade 1 30mm OR grade 2, 20mm OR grade 3 10mm NABON <35 or 35 Yes Yes Not used No Yes No Yes None 10-years survival probability 85% N0, <35, grade I tumor 1 cm OR 35 yrs, grade I tumor 2 cm. Not specified. Suggested: <3% survival benefit in 10-years no chemotherapy; 3-5% chemotherapy discussed as possible option PREDICT Yes Yes Yes Not used ER Yes Yes Yes Method of detection, CT regimen AOL=Adjuvant! Online; NPI=Nottingham Prognostic Index; CBO & NABON=Dutch guidelines; ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal growth factor Receptor 2; CT=chemotherapy; PVI=peritumoral vascular invasion 5 Agreement among oncologists 73

76 Statistical analysis All data were analyzed using SPSS 20.0 (SPSS Inc.). Agreement among the oncologists as well as between each oncologist and the 70-gene signature result (low risk versus high risk) was assessed using kappa statistics. A kappa of 0 means random, slight agreement, fair agreement, moderate agreement, substantial agreement, almost perfect agreement and a kappa of 1 is perfect agreement. The paired samples t-test was conducted to compare the kappa means between the oncologists risk estimations in the first and second questionnaire. Logistic regression models were used to assess the likelihood of the 70-gene signature result leading to changes in risk estimations and AST recommendations. Covariants included in this model were age, tumor size, grade, histological type, estrogen receptor (ER) and HER2 status. In case of an unanswered question in either the clinical risk estimation or the estimation based on the 70-gene signature, these risk estimations were both excluded from the analyses. A significant finding was defined as a two-sided p-value below Results Participants and case characteristics Nineteen oncologists completed the first questionnaire (66%). Twelve oncologists (41%) also completed the second questionnaire. Mean age of these oncologists was 49 years (36-66 years) and they were practicing their current profession on average for 18 years (2-35 years). Six of the oncologists came from the Netherlands and six from other European countries (Germany, France, Italy and Portugal). Patient and tumor characteristics of the 37 cases included in the analyses as well as their risk estimations according to the 70-gene signature, AOL and other clinical guidelines are summarized in Supplementary Table 1. On average, for each case two risk estimations and three AST recommendations were missing per oncologist, i.e. not answered in the two questionnaires. Risk estimations and AST recommendations On average, the oncologists classified 51% (range 24-65%) cases as clinically low risk and 47% (range 32-76%) as clinically high risk. After adding the 70-gene signature result, the oncologists classified 59% (range 22-78%) of the cases as low risk and 38% (range 22-78%) as high risk (Figure 1). On average, an oncologist changed the given clinical risk estimation in 14.2% of the cases. In 10.8 % of the cases high risk changed to low risk and in 3.4% of the cases low risk changed to high risk (Table 2). This leads to a net reduction of 7.4% (range %) in high risk classifications. In the 12 cases in which all guidelines and the 70-gene signature were concordant significantly less changes in risk estimations were made (3.5%) compared to the 25 cases in which one or more of the guidelines and the 70-gene signature were discordant (18%) (p<0.0001). 74 Chapter 5

77 5 Figure 1. Changes in risk estimations per oncologist per case and risk estimations by clinicopathological guidelines and the 70-gene signature Table 2. Changes in risk estimation and AST recommendation after providing 70-gene signature result A. Changes in risk estimation (%) 70GS High risk Low risk Total CR CR $ High risk 149 (35.8) 45 (10.8) 194 (46.6) Low risk 14 (3.4) 208 (50) 222 (53.4) Total 70GS 163 (39.2) 253 (60.8) 416 C (100) B. Changes in AST-recommendation (%) 70GS No AST Chemotherapy A Endocrine CR $ therapy B Total CR No AST 16 (3.9) 1 (0.2) 5 (1.2) 22 (5.3) Chemotherapy A 2 (0.5) 144 (34.8) 57 (13.8) 203 (49) Endocrine therapy B 10 (2.4) 8 (1.9) 171 (41.3) 189 (45.7) Total 70GS 28 (6.8) 153 (37) 233 (56.3) 414 C (100) CR = Clinical risk, estimations based on clinicopathological factors, 70GS = 70-gene signature, result included in the questionnaire. A Chemotherapy alone or combined with endocrine therapy and / or trastuzumab. B Endocrine therapy alone. C Missing values not included Agreement among oncologists 75

78 The oncologists recommended AST based on clinicopathological factors in 95% (range %) of the cases, chemotherapy (alone or combined) in 48% (range 30-70%) and endocrine therapy (alone) in 46% (range 0-70%) of the cases (Table 2, Figure 2). After adding the 70-gene signature result to the clinicopathological factors provided in the first questionnaire, they recommended AST in 93% (range %) of the cases, chemotherapy (alone or combined) in 37% (range 22-68%) and endocrine therapy (alone) in 57% (range 11-78%). In 24% of the cases the oncologist adjusted the AST recommendation (Table 2). Adding the 70-gene signature resulted in 14.3% of the cases in a change from chemotherapy to either endocrine therapy or no AST at all. Only one oncologist advised more chemotherapy after knowledge of the 70-gene signature result. In 2.1% of the cases the advice of no AST or endocrine therapy only was changed to chemotherapy. This resulted in a reduction in chemotherapy use of 12.2% (range: %) after adding the 70- gene signature to known clinicopathological factors in the second questionnaire. In the 12 cases in which all guidelines and the 70-gene signature were concordant significantly less changes in AST recommendations were made (4.2%) compared to the 25 cases in which one or more of the guidelines and the 70-gene signature were discordant (20.7%)(p<0.0001). Agreement among oncologists There was moderate level of agreement among oncologists in risk estimations based solely on clinicopathological factors (κ=0.57; range: ) (Table 3). The level of agreement in AST recommendation was also moderate (κ=0.57; range: ). After adding the 70-gene signature result to clinicopathological factors, agreement in risk estimation increases slightly, but significantly to substantial (κ=0.61; range: ; p=0.035), while the level of agreement regarding AST recommendations remained moderate (κ=0.56; range: ; p=0.59). The agreement among oncologists after adding the 70-gene signature remained moderate for risk estimations (κ=0.44; range: ; p=0.39) as well as AST recommendations (κ=0.56; range: ; p=0,76). Opinion of oncologists about the use of the 70-gene signature Seven oncologists (58%) indicated the 70-gene signature result had additional value and adding the 70-gene signature result led to a slightly, not significantly larger decrease in the use of AST in these oncologists. On average, in 19% of the cases the result of the 70-gene signature was decisive according to the oncologists who indicated the 70-gene signature had additional value. 76 Chapter 5

79 5 Figure 2. Changes in AST recommendations per oncologist per case and the actual given treatment Agreement among oncologists 77

80 Table 3. Levels of agreement among oncologists in risk estimations and AST recommendations before and after providing the 70-gene signature result to known clinicopathological factors Legend Kappa Kappa Agreement <0 Less than chance Slight Fair Moderate Substantial Almost perfect Level of agreement among oncologists in risk estimation based solely on clinicopathological factors Oncologists ,30 0,33 0,39 0,29 0,29 0,36 0,33 0,20 0,44 0,36 0,46 2 0,54 0,71 0,68 0,70 0,64 0,77 0,62 0,55 0,63 0,57 3 0,73 0,71 0,73 0,56 0,78 0,56 0,73 0,35 0,58 4 0,76 0,78 0,72 0,73 0,61 0,47 0,55 0,64 5 0,88 0,70 0,59 0,80 0,49 0,49 0,51 6 0,61 0,62 0,83 0,57 0,56 0,64 7 0,67 0,66 0,41 0,49 0,82 8 0,56 0,62 0,61 0,58 9 0,42 0,51 0, ,41 0, ,45 12 Level of agreement among oncologists in risk estimation after providing the 70-gene signature result Oncologists ,26 0,23 0,21 0,21 0,19 0,21 0,32 0,14 0,35 0,30 0,36 2 0,46 0,75 0,93 0,81 0,86 0,54 0,68 0,61 0,48 0,73 3 0,69 0,65 0,61 0,65 0,60 0,73 0,43 0,54 0,66 4 0,93 0,80 0,78 0,65 0,79 0,60 0,49 0,72 5 0,92 0,92 0,73 0,92 0,73 0,52 0,83 6 1,00 0,59 0,54 0,85 0,44 0,86 7 0,68 0,92 0,60 0,48 0,92 8 0,58 0,61 0,58 0,74 9 0,53 0,35 0, ,52 0, , Chapter 5

81 Level of agreement among oncologists in AST recommendation based solely on clinicopathological factors Oncologists ,53 0,37 0,30 0,28 0,30 0,26 0,37 0,24 0,42 0,44 0,34 2 0,70 0,67 0,55 0,66 0,51 0,70 0,47 0,67 0,42 0,58 3 0,69 0,69 0,71 0,62 0,84 0,65 0,66 0,50 0,47 4 0,65 0,77 0,67 0,69 0,70 0,56 0,25 0,51 5 0,79 0,79 0,69 0,83 0,48 0,52 0,56 6 0,72 0,62 0,76 0,54 0,40 0,66 7 0,70 0,84 0,66 0,65 0,66 8 0,65 0,66 0,59 0,47 9 0,43 0,35 0, ,52 0, ,41 12 Level of agreement among oncologists in AST recommendation after providing the 70-gene signature result Oncologists ,40 0,45 0,28 0,23 0,25 0,27 0,36 0,18 0,34 0,41 0,30 2 0,48 0,75 0,63 0,64 0,65 0,54 0,52 0,72 0,50 0,55 3 0,54 0,44 0,49 0,51 0,51 0,51 0,48 0,59 0,51 4 0,88 0,83 0,81 0,63 0,75 0,52 0,50 0,72 5 0,82 0,80 0,56 0,73 0,44 0,44 0,75 6 1,00 0,58 0,80 0,42 0,46 0,88 7 0,64 0,86 0,50 0,48 0,93 8 0,51 0,49 0,62 0,65 9 0,39 0,34 0, ,47 0, , Discussion Only a moderate level of agreement for both risk estimations and treatment decisions was observed between oncologists when using the clinicopathological factors that are used in current guidelines, such as age, tumor size, grade and hormone-receptor status. After providing the 70-gene signature result the level of agreement in risk estimations among oncologists increased slightly from moderate (κ=0.55) to substantial (κ=0.61; p=0.035), showing that classification of patients into high and low risk groups based on the 70-gene signature result may be useful to guide AST recommendations. The participating oncologists classified more patients as high risk compared to the 70-gene signature. This was followed by recommendations of AST in 92% of the cases. In 10.8% of the Agreement among oncologists 79

82 cases a high risk estimation was changed into a low risk estimation after adding the 70-gene signature result. Overall, a reduction in the proportion of high risk patients of 7.4% and reduction of 12.2% in the use of chemotherapy was seen in this case-selection; these proportions may of course differ in populations with a different distribution of tumor characteristics. Previously reported specificity rates of the 70-gene signature (0.56) are higher than AOL (0.53) and St. Gallen (0.10) at 5 years of follow-up in a pooled dataset of 70-gene signature validation series of untreated patients with ER-positive, node-negative breast cancer. 14 This suggests that the 70-gene signature is a useful tool to reduce the risk of falsely classifying a patient as high risk and that the 70-gene signature may help to reduce overtreatment. An important observation is the variation among oncologists in risk estimation and AST recommendation. A similar study, where the Oncotype DX recurrence score was used as a prognostic tool, showed comparable results, demonstrating that oncologists only have fair to moderate level of agreement when predicting the recurrence score. 15 Adding the recurrence score resulted in a decrease in chemotherapy recommendation of 10.8%, which is comparable to the 12.2% seen in our survey. In our survey, 58% of the oncologists found the 70-gene signature of additional value. There are some limitations to this survey. The results of 12 oncologists are reported; 19 out of 29 responded to the first questionnaire and only 12 out of 29 also responded to the second questionnaire leading to a response rate of 41%. Unfortunately, because the number of participating oncologists was fairly low we were unable to perform subgroup analyses to evaluate if oncologists are adherent to the guidelines they indicated to use. The agreement among oncologists might also be partly explained by the presence of a few cases at such a high risk that chemotherapy might be considered standard of care. Even though not all guidelines included in this survey for example identify HER2-positive patients as high risk, the majority of the oncologists consider them eligible for chemotherapy. When excluding the HER2-positive cases from the analysis, the results show a moderate agreement in risk estimation and AST recommendation based on solely clinicopathological factors as well as after adding the 70-gene signature result. The changes in risk estimation and AST recommendations in this survey could also be due to practice patterns of oncologists and lack of adherence to guidelines in general. 16 Only oncologists in Europe were invited to participate in this survey. A larger survey, including a larger number of oncologists not only from Europe, but also from other continents would provide more detailed information on differences in breast cancer treatment between countries and continents. In daily clinical practice, the oncologist is faced with the challenge of tailoring adjuvant systemic treatment for each patient, taking the clinicopathological features of the tumor, the 70-gene signature result, the patients co-morbidities and preferences into account. Proliferation markers, like Ki-67, menopausal status and co-morbidity were unknown in our case-selection and were not presented in the questionnaires. Providing this kind of extra information may have further improved the ability to discriminate between high and low risk cases and may have influenced AST recommendation. On the other hand, providing more proliferation markers and pathological 80 Chapter 5

83 characteristics may not directly result in more agreement. 17 In clinical practice, gene-expression profiles will likely be used in addition to clinicopathological guidelines, like the way the 70-gene signature was used in the RASTER study and presented in the cases in our survey. 18 The follow-up of the RASTER study showed that patients treated according to the 70-gene signature who did not receive AST, despite poor clinicopathological factors, had a distant recurrence free interval of 100%. 18 Based on these data, the reduction in chemotherapy resulting from knowledge of the 70-gene signature result as presented in this survey, may be justified. Especially, since in the RASTER study not only the 70-gene signature result was decisive, but also the doctors and patients preferences. The St. Gallen 2011 recommendations and ESMO practice guidelines include the 70-gene signature as an indicator for AST. 19,20 In conclusion, this survey shows the variability in guidelines and oncologists risk estimations and recommendations of AST in early stage breast cancer patients. Providing the 70-gene signature result has a modest impact on risk estimation and AST recommendation. It may lead to a reduction in the classification of high risk patients and a decrease in the use of chemotherapy. Most importantly, this survey underlines the need for a better standardization tool in clinical decision-making. Acknowledgements The authors thank all participants who filled out the first questionnaire, but were not able to complete their participation in this survey, for their time and efforts. We thank Jolien Bueno-de- Mesquita for collecting and providing the baseline characteristics of the RASTER patients, Emiel Rutgers for evaluating the impact of the 70-gene signature from a surgeon s point of view and Philip Schouten for his efforts preparing the illustrations for this manuscript. 5 Funding source This work was supported by BBMRI-NL (complementation project no. 45), TI Pharma (project no. T3 502) and the Dutch Cancer Society (grant number NKI ). The RASTER study was financially supported by the Dutch Health Care Insurance Board (CVZ). Conflict of interest statement The authors have declared no conflicts of interest. Agreement among oncologists 81

84 References 1 Bueno-de-Mesquita JM, Sonke GS, van de Vijver MJ, Linn SC. Additional value and potential use of the 70-gene prognosis signature in node-negative breast cancer in daily clinical practice. Ann Oncol 2011;22: van de Vijver MJ, He YD, van t Veer LJ et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347: Integraal Kankercentrum Nederland: NABON richtlijn mammacarcinoom Cardoso F. Microarray technology and its effect on breast cancer (re)classification and prediction of outcome. Breast Cancer Res 2003;5: Cardoso F, van t Veer L, Rutgers E, Loi S, Mook S, Piccart-Gebhart MJ. Clinical application of the 70- gene profile: the MINDACT trial. J Clin Oncol 2008;26: D Eredita G, Giardina C, Martellotta M, Natale T, Ferrarese F. Prognostic factors in breast cancer: the predictive value of the Nottingham Prognostic Index in patients with a long-term follow-up that were treated in a single institution. Eur J Cancer 2001;37: Goldhirsch A, Glick JH, Gelber RD, Coates AS, Senn HJ. Meeting highlights: International Consensus Panel on the Treatment of Primary Breast Cancer. Seventh International Conference on Adjuvant Therapy of Primary Breast Cancer. J Clin Oncol 2001;19: Olivotto IA, Bajdik CD, Ravdin PM et al. Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005;23: Azim HA, Jr., Michiels S, Zagouri F et al. Utility of prognostic genomic tests in breast cancer practice: The IMPAKT 2012 Working Group Consensus Statement. Ann Oncol 2013;24: van t Veer LJ, Dai H, van de Vijver MJ et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415: Bueno-de-Mesquita JM, van Harten WH, Retel VP et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007;8: Goldhirsch A, Wood WC, Gelber RD, Coates AS, Thurlimann B, Senn HJ. Meeting highlights: updated international expert consensus on the primary therapy of early breast cancer. J Clin Oncol 2003;21: Kwaliteitsinstituut voor de Gezondheidszorg CBO VvlK. Adjuvante Systemische Therapie voor het Operabel Mammacarcinoom. Richtlijn Behandeling van het Mammacarcinoom 2004; Retel VP, Joore MA, Knauer M, Linn SC, Hauptmann M, Harten WH. Cost-effectiveness of the 70- gene signature versus St. Gallen guidelines and Adjuvant Online for early breast cancer. Eur J Cancer 2010;46: Kamal AH, Loprinzi CL, Reynolds C et al. Breast medical oncologists use of standard prognostic factors to predict a 21-gene recurrence score. Oncologist 2011;16: Foster JA, Abdolrasulnia M, Doroodchi H, McClure J, Casebeer L. Practice patterns and guideline adherence of medical oncologists in managing patients with early breast cancer. J Natl Compr Canc Netw 2009;7: Mehta R, Jain RK, Badve S. Personalized medicine: the road ahead. Clin Breast Cancer 2011;11: Drukker CA, Bueno-de-Mesquita JM, Retel VP et al. A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 2013;133: Aebi S, Davidson T, Gruber G, Cardoso F. Primary breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 2011;22 Suppl 6:vi12-vi Chapter 5

85 20 Goldhirsch A, Wood WC, Coates AS, Gelber RD, Thurlimann B, Senn HJ. Strategies for subtypes-- dealing with the diversity of breast cancer: highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer Ann Oncol 2011;22: Agreement among oncologists 83

86 Supplementary Table 1. Characteristics of the 37 breast cancer cases presented in the questionnaires Case Age yrs Size mm Histology Grade a Mitotic index ER b PR b HER2 c Primary treatment 70GS AOL NPI St. Gallen Ductal 3 30 per 2mm2 80% - Neg BCT + Rtx High High High High High Ductal 2 low 100% - Neg BCT + Rtx Low High Low High Low Ductal 3 16 per 8 HPF Ablation High High High High High Ductal Neg BCT + Rtx Low Low Low Low Low Ductal Neg BCT + Rtx Low High Low High Low Ductal 1 1 per 10 HPF 100% 100% Neg BCT + Rtx Low Low Low Low Low Lobular 2 unknown + + Neg BCT + Rtx Low High Low High Low Ductal 1 4 per 10 HPF + + Neg BCT + Rtx Low Low Low Low Low Ductal 2 unknown BCT + Rtx High High Low High Low Ductal 2 unknown + + Neg BCT + Rtx Low High Low High Low Ductal 2 unknown 70% 100% Neg BCT + Rtx Low Low Low High Low Ductal % 50% 3+ BCT + Rtx High High High High High Ductal 2 3 per 2mm2 + + Neg BCT + Rtx Low Low Low High Low Ductal 2 unknown + + Neg BCT+ Rtx Low High Low High Low Ductal Neg BCT+ Rtx Low High Low High Low Lobular 1 2 per 2mm2 30% 90% Neg BCT+ Rtx Low High Low High Low Ductal % 30% Neg BCT+ Rtx High High Low High High Lobular % - Neg Ablation Low High Low High Low Ductal 2 16 per 10 HPF - - Neg BCT+ Rtx High High High High High lobular 2 11 per 10 HPF 10% 40% Neg BCT+ Rtx Low High Low High Low lobular 2 4 per 10 HPF + + Neg Ablation Low High Low High Low ductolobular 2 5 per 10 HPF + - Neg Ablation Low High Low High Low ductal 2 18 per 2mm2 90% 90% Neg BCT+ Rtx High High Low High High ductal % - Neg BCT+ Rtx Low High Low Low Low ductal 2 5 per 10 HPF 100% 90% Neg BCT+ Rtx Low Low Low High Low ductal 2 8 per 2mm BCT+ Rtx High High Low High Low lobular 2 5 per 2mm2 + + Neg BCT+ Rtx Low High Low High Low ductal 1 2 per 2mm2 + + Neg BCT+ Rtx Low Low Low Low Low ductal % 100% Neg Ablation Low Low Low High High CBO Chapter 5

87 ductal 2 2 per 10 HPF 50% 50% 3+ Ablation High High High High High mucinous ductal Neg BCT+ Rtx Low High Low High Low invasive % 100% Neg Ablation + Rtx Low High Low High Low ductal Neg Ablation Low Low Low High Low ductal 1 unknown + + Neg BCT+ Rtx Low Low Low Low Low ductal 1 3 per 2mm2 60% 50% Neg BCT+ Rtx Low Low Low Low Low ductal 3 9 per 2mm BCT+ Rtx High High High High High ductal % 100% Neg BCT+ Rtx Low Low Low High Low a Histological tumor grade according to Elston and Ellis. b According to Dutch guidelines, oestrogen and progesterone receptors (ER, PR) were deemed positive if at least 10% of tumor cells stained positive in immunohistochemical assay. c Samples were deemed HER2-positive if the score was 3+ in immunohistochemical assay. If the score was 2+ in immunohistochemical assay and a fluorescent in-situ hybridisation result (FISH) was available, the FISH result (positive or negative) was used. Abbreviations: Neg or - = negative, + = positive, BCT = breast conserving therapy, Rtx = radiotherapy, 70GS = 70-gene signature, AOL= Adjuvant! Online version 8.0, NPI= Nottingham Prognostic Index, CBO= the Dutch Institute of Healthcare Improvement guidelines 5 Agreement among oncologists 85

88 Chapter 6

89 Long-term impact of the 70-gene signature on breast cancer outcome Breast Cancer Research and Treatment 2014;143: Caroline A. Drukker Harm van Tinteren Marjanka K. Schmidt Emiel J.Th. Rutgers Marc J. van de Vijver Laura J. van t Veer

90 Abstract Background Several studies have validated the prognostic value of the 70-gene prognosis signature (MammaPrint ), but long-term outcome of these patients has not been previously reported. Methods The follow-up of the consecutively treated cohort of 295 patients (<53 years) with invasive breast cancer (T1-2N0-1M0; n=151 N0, n=144 N1) diagnosed between 1984 and 1995, in which the 70-gene signature was previously validated, was updated. Results The median follow-up for this series is now extended to 18.5 years. A significant difference is seen in long-term distant-metastasis-free-survival (DMFS) for the patients with a low and a high risk 70-gene signature (DMFS p<0.0001), as well as separately for node-negative (DMFS p<0.0001) and node-positive patients (DMFS p=0.0004). The 25-year Hazard Ratios (HR) for all patients for DMFS and OS were 3.1 (95%CI: ) and 2.9 (95%CI: ), respectively. The HRs for DMFS and OS were largest in the first five years after diagnosis: 9.6 (95%CI: ) and 11.3 (95%CI: ), respectively. The 25-year HR in the subgroup of node-negative patients for DMFS and OS were 4.57 (95%CI: ) and 4.73 (95%CI: ), respectively, and for node-positive patients for DMFS and OS were 2.24 (95%CI: ) and 1.83 (95%CI: ), respectively. Conclusion The 70-gene signature remains prognostic at longer follow-up in patients < 53 years of age with stage I and II breast cancer. The 70-gene signature strongest prognostic power is seen in the first 5 years after diagnosis. 88 Chapter 6

91 Introduction Gene-expression signatures, such as the 70-gene signature (MammaPrint ), were developed to assess the risk of distant recurrence in the first five years after diagnosis to predict outcome of breast cancer patients. 1 The 70-gene signature was extensively validated in several retrospective studies. 2-4 The test was mainly validated in systemically untreated patients with estrogenreceptor (ER) positive and negative invasive breast cancer, <55 or 60 years, with no axillary nodal involvement. Subsequently, multiple studies validated this signature for additional subgroups such as postmenopausal patients, patients with up to three positive lymph nodes and for Human Epidermal growth factor Receptor 2 (HER2) positive disease. 5-7 More recently, the first prospective data on the 70-gene signature was published showing an excellent overall survival for patients with a low risk for recurrence estimation by the 70-gene signature. Even when these low risk patients did not receive adjuvant chemotherapy, despite poor clinicopathological factors, they had a 5-year distant recurrence free interval of 100%. 8 In studies on the prognostic value of the 70-gene signature published so far, the median followup was between 5 to 13.6 years. To our knowledge, no data on long-term survival of patients for whom gene-expression data is available has been published. We therefore set out to update the follow-up of the previously published 70-gene signature consecutive 295 patient cohort, as published by van de Vijver et al. in 2002, to investigate the long-term outcome of breast cancer patients and to evaluate the effect of the 70-gene signature after longer follow-up. 4 Patients and Methods 6 Follow-up was updated until September 2013 for a cohort of 295 consecutive patients diagnosed with primary breast cancer. Study design, patient eligibility and study logistics of the study have been described before. 4 In short, all patients were female, younger than 53 years with histologically proven, operable, invasive breast cancer (T1-2N0-1M0). All were diagnosed at the Netherlands Cancer Institute between 1984 and of the 295 patients had node-negative disease, 144 patients had node-positive disease. All patients were primarily treated with breast-conserving surgery or mastectomy. Adjuvant treatment consisted of radiotherapy, chemotherapy and/or endocrine therapy as indicated by guidelines used at the time of treatment. 70-gene signature For all patients included in these analyses a 70-gene signature result was available. Frozen tumor samples from each patient were processed at the Netherlands Cancer Institute and Rosetta Inpharmatics for RNA isolation, amplification, and labelling as described elsewhere. 1,4,9 Tumors were classified as a 70-gene signature low or high risk at the time of the initial studies. Low risk Long-term impact of the 70-gene signature 89

92 was defined as an index-score greater than 0.4. High risk was defined as an index-score lower than ,9 Statistical Analysis For this analysis, we estimated overall survival (OS) and distant metastasis free survival (DMFS). DMFS was defined as time from diagnosis to distant metastasis as first event. Data on all other patients was censored on the last date of follow-up, in the event of a second primary tumor including contralateral breast cancer, in case of death from any cause other than breast cancer or if there was a locoregional recurrence of the disease. In case a locoregional recurrence was followed by distant metastasis within 6 months, the event of distant metastasis was included in the analysis. Survival curves were constructed using the Kaplan-Meier method and compared using the log-rank test. Competing risk analyses were performed to adjust for patients having a type of event (for example death due to another cause than breast cancer) that makes them unable to develop the event of interest. The Hazard Ratio s (HR) of the 70-gene signature were calculated for the full follow-up as well as per 5-year intervals. A significant finding was defined as a p-value below Analyses were performed using SAS version 9.2 and R version Results Patient and tumor characteristics Patient characteristics are described in Supplementary Table 1. 4 Of the 295 patients, 115 patients had a low risk 70-gene signature and 180 had a high risk 70-gene signature. Patients with a low risk 70-gene signature were more often of older age and had more often smaller estrogenreceptor (ER)-positive tumors with lower grade. No significant difference between 70-gene signature high and low risk patients was seen for number of positive nodes, vascular invasion and treatment (type of surgery, adjuvant chemotherapy nor adjuvant endocrine therapy). Thirty seven percent of the patients received adjuvant chemotherapy and 14% received adjuvant endocrine therapy. After a median follow up of 18.5 years, 121 patients developed distant metastasis as first event. One hundred and twenty seven patients have died, of whom 114 due to breast cancer. Long-term prognostic value of the 70-gene signature Figure 1 shows DMFS and OS for the entire cohort (1A), and separately for node-negative (1B), and node-positive patients (1C). The Kaplan-Meier curves showed a significant absolute difference in DMFS and OS at 25 years between the patients with a low risk 70-gene signature (60.4% and 57.3% respectively) and the patients with a high risk 70-gene signature (41.6% and 44.5% respectively; p< for both OS and DMFS). This significant difference was observed for node-negative (p< for both OS and DMFS) as well as node-positive patients (p=0.03 for OS and p= for DMFS). 90 Chapter 6

93 DMFS All Patients OS All patients 6 DMFS Node-negative patients OS Node-negative patients DMFS Node-positive patients OS Node-positive patients Figure 1. Overall Survival (OS) and Distant Metastasis Free Survival (DMFS) for all patients and stratified by nodal status. Long-term impact of the 70-gene signature 91

94 Conditional survival probabilities for all patients and both subgroups for 5, 10, 15, 20 and 25 years are summarized in Table 1A. The 25-year HR for all patients for DMFS and OS were 3.1 (95%CI: ) and 2.9 (95%CI: ), respectively. The HR for DMFS in the first 5 years after diagnosis was 9.6 (95%CI: ) and 11.3 (95%CI: ) for OS. After 5 years the effect of the 70-gene signature on DMFS diminished, while the effect on OS from years 5 to 10 after diagnosis was still very significant with a HR 6.1 (95%CI: ). After 15 years, the effect of the 70-gene signature on OS slowly diminished (Table 1B). The 25-year HR for nodenegative patients for DMFS and OS were 4.57 (95%CI: ) and 4.73 (95%CI: ), respectively. The 25-year HR for node-positive patients for DMFS and OS were 2.24 (95%CI: ) and 1.83 (95%CI: ), respectively. Distant metastases and competing events Figure 2 demonstrates how competing events that occurred in this cohort in addition to distant metastases are divided over the 70-gene signature low and high risk groups. The 70-gene signature low versus high risk in this cohort was only significant for prediction of distant metastases, as shown in Figure 1 and Table 1. For the DMFS analyses, locoregional recurrence was considered a competing event and therefore follow-up was censored if occurred first, except when the locoregional event took place within 6 months prior to the distant metastasis. Reanalysing the data without including the locoregional events that occur within 6 months before the patient is diagnosed with distant metastases, gives no substantial difference in survival probabilities (data not shown). 70-gene signature low risk 70-gene signature high risk Figure 2. Competing risk analyses 92 Chapter 6

95 Tabel 1A. Distant Metastasis Free Survival (DMFS) and Overall Survival (OS) probabilities for all patients and stratified by nodal status. Group No. of patients DMFS (95%CI) 5yr 10 yr 15 yr 20 yr 25 yr All patients 295 Low risk ( ) 82.0 ( ) 78.1 ( ) 75.9 ( ) 60.4 ( ) High risk ( ) 50.0 ( ) 47.1 ( ) 44.8 ( ) 41.6 ( ) Node-negative 151 Low risk ( ) 85.6 ( ) 85.6 ( ) 81.3 ( ) 73.2 ( ) High risk ( ) 45.6 ( ) 44.0 ( ) 39.6 ( ) 39.6 ( ) Node-positive 144 Low risk ( ) 78.6 ( ) 70.3 ( ) 70.3 ( ) No pt at risk High risk ( ) 54.3 ( ) 50.1 ( ) 50.1 ( ) 44.5 ( ) OS (95%CI) 5yr 10 yr 15 yr 20 yr 25 yr All patients 295 Low risk ( ) 92.8 ( ) 83.0 ( ) 69.4 ( ) 57.3 ( ) High risk ( ) 55.7 ( ) 47.7 ( ) 42.0 ( ) 39.7 ( ) Node-negative 151 Low risk ( ) 93.2 ( ) 89.1 ( ) 82.1 ( ) 69.5 ( ) High risk ( ) 52.7 ( ) 44.3 ( ) 37.8 ( ) 33.6 (23-49) Node-positive 144 Low risk ( ) 92.5 ( ) 76.6 ( ) 54.5 ( ) 42.2 ( ) High risk ( ) 58.7 ( ) 51.1 ( ) 47.1 ( ) 47.1 ( ) Table 1B. Hazard Ratio s for the 70-gene signature for OS and DMFS At risk Events HR 95% CI DMFS 0-25 years 0-5 years 5-10 years years years years OS years 0-5 years 5-10 years years years years Long-term impact of the 70-gene signature 93

96 Discussion This update of the consecutive 295 patient cohort shows that the 70-gene signature is able to accurately differentiate between patients at a low and a high risk of distant metastases up to 25 years after diagnosis. This gene signature was designed to predict the risk of distant metastases in the first 5 years after diagnosis. Previous analyses by Buyse et al. already confirmed that the 70-gene signature has prognostic value in the first five years after diagnosis. 3 Their analyses also suggested that this effect might be present up to ten years after diagnosis. In our analyses the 70-gene signature has the largest prognostic value for DMFS and OS in the first 5 years (HRs 9.6 (95%CI: ) and 11.3 (95% CI: ) respectively). The significant prognostic value per 5 year intervals for OS remained from 5 years after diagnosis onwards and becomes smaller after 15 years. Meta-analyses of patients with breast cancer have shown that adjuvant chemotherapy reduces the rate of recurrence almost exclusively in the first 5 years. 10 Consequently, one would expect that patients with relapse in the first 5 years after surgery will benefit most from adjuvant chemotherapy. Thus, for the question of who should receive adjuvant chemotherapy, it is most relevant to identify the patients with relapse in the first 5 years after surgery. That the 70-gene signature has the highest HR for recurrence in the first 5 years supports the notion that this test can help identify those patients that are most likely to benefit from adjuvant chemotherapy. The Food and Drug Administration (FDA) 510-(k) cleared intended use of the 70-gene signature for prognosis prediction in the node-negative, systemically untreated patient population (IVDMIA k101454). The node-negative subgroup of this consecutive series, of whom over 85% did not receive adjuvant systemic treatment, most closely represents this population; it is shown here that for node-negative patients long term outcome can also be predicted using the 70-gene prognosis signature. Patients included in this cohort were all diagnosed between 1984 and Due to improvement of adjuvant systemic therapy and the introduction of nation-wide screening programs, which resulted in an increase in early stage breast cancer and a decrease in breast cancer mortality rates, one could hypothesize that the survival probabilities of this cohort if diagnosed today would be even better than shown here. 10 Also of note, the patients included were all younger than 53 years old, who tend to have a poorer prognosis compared to patients diagnosed at older age. 11,12 In conclusion, an update of the 70-gene signature consecutive 295 patient cohort shows that the 70-gene signature has long-term prognostic value in patients < 53 years old with stage I and II breast cancer. 94 Chapter 6

97 Contributors CD updated the follow-up data of this 70-gene signature validation study. CD and HvT performed the statistical analysis. CD, HvT, ER, MKS, and LvtV took part in data interpretation and manuscript writing. All authors were involved in reviewing the report. Funding sources This work was supported by the EORTC Breast Cancer Group (type 3 grant 2011/2012), BBMRI-NL, a research infrastructure financed by the Dutch Government (NWO , complementation project 45), and the Dutch Genomics Initiative Cancer Genomics Centre. Conflict of interest RB, MvdV and LvtV are named inventors on the patent for the 70-gene signature used in this study. RB and LvtV report being shareholder in and employed by Agendia NV, the commercial company that markets the 70-gene signature as MammaPrint. Acknowledgements We acknowledge the efforts of Sjoerd Elias, Stella Mook and Michael Knauer to keep the database used for this study updated. We are indebted to all women who participated in this 70-gene signature validation study. 6 Long-term impact of the 70-gene signature 95

98 References 1 van t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002; 415: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009; 117: Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas AM et al. Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 2006; 98: van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: Knauer M, Cardoso F, Wesseling J, Bedard PL, Linn SC, Rutgers EJ et al. Identification of a low-risk subgroup of HER-2-positive breast cancer by the 70-gene prognosis signature. Br J Cancer 2010; 103: Mook S, Schmidt MK, Viale G, Pruneri G, Eekhout I, Floore A et al. The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1-3 positive lymph nodes in an independent validation study. Breast Cancer Res Treat 2009; 116: Mook S, Schmidt MK, Weigelt B, Kreike B, Eekhout I, van de Vijver MJ et al. The 70-gene prognosis signature predicts early metastasis in breast cancer patients between 55 and 70 years of age. Ann Oncol 2010; 21: Drukker CA, Bueno-de-Mesquita JM, Retel VP, van Harten WH, van Tinteren H, Wesseling J et al. A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 2013; 133: Glas AM, Floore A, Delahaye LJ, Witteveen AT, Pover RC, Bakx N et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006; 7: Early Breast Cancer Trialists Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 365: Adami HO, Malker B, Holmberg L, Persson I, Stone B. The relation between survival and age at diagnosis in breast cancer. N Engl J Med 1986; 315: Chung M, Chang HR, Bland KI, Wanebo HJ. Younger women with breast carcinoma have a poorer prognosis than older women. Cancer 1996; 77: Chapter 6

99 Supplementary Table 1. Patient and tumor characteristics stratified by 70-gene signature 70-gene signature low risk n=115 (39%) 70-gene signature high risk n=180 (61%) p-value Age <0.001 <40 yrs 11 (10) 52 (29) yrs 44 (38) 41 (23) yrs 43 (37) 55 (31) >=50yrs 17 (15) 32 (18) No. of positive nodes (52) 91 (51) (37) 63 (35) >=4 12 (10) 26 (14) Tumor size =<20 mm 71 (62) 84 (47) >20 mm 44 (38) 96 (53) Histological grade <0.001 I 56 (49) 19 (11) II 45 (39) 56 (31) III 14 (12) 105 (58) Lymphovascular invasion 0.38 Absent 77 (67) 108 (60) 1-3 vessels 12 (10) 18 (10) >3 vessels 26 (23) 54 (30) ER status <0.001 Negative 3 (3) 66 (37) Positive 112 (97) 114 (63) Surgery 0.63 Breast Conserving 64 (56) 97 (54) Mastectomy 51 (44) 83 (46) Chemotherapy 0.79 No 71 (62) 114 (63) Yes 44 (38) 66 (37) Endocrine therapy 0.63 No 98 (85) 157 (87) Yes 17 (15) 23 (13) 6 ER=estrogen receptor Long-term impact of the 70-gene signature 97

100 Chapter 7

101 Mammographic screening detects low risk tumor biology breast cancers Breast Cancer Research and Treatment 2014 Jan 28. Epub ahead of print Caroline A. Drukker Marjanka K. Schmidt Emiel J.Th. Rutgers Fatima Cardoso Karla Kerlikowske Laura J. Esserman Flora E. van Leeuwen Ruud M. Pijnappel Leen Slaets Jan Bogaerts Laura J. van t Veer

102 Abstract Background Overdiagnosis of breast cancer, i.e. the detection of slow growing tumors that would never have caused symptoms or death, became more prevalent with the implementation of populationbased screening. Only rough estimates have been made of the proportion of patients that are overdiagnosed and identification of those patients is difficult. Therefore, the aim of this study is to evaluate whether tumor biology can help identify patients with screen-detected tumors at such a low risk of recurrence that they are likely to be overdiagnosed. Furthermore, we wish to evaluate the impact of the transition from film-screen mammography (FSM) to the more sensitive full-field digital mammography (FFDM) on the biology of the tumors detected by each screening-modality. Methods All Dutch breast cancer patients enrolled in the MINDACT trial (EORTC-10041) accrued , who participated in the national screening program (biennial screening ages 50-75) were included (n=1165). We calculated the proportions of high, low and among those the ultralow risk tumors according to the 70-gene signature for patients with screen-detected (n= 775) and interval (n=390) cancers for FSM and FFDM. Results Screen-detected cancers had significantly more often a low risk tumor biology (68%) of which 54% even an ultralow risk compared to interval cancers (53% low, of which 45% ultralow risk (p=0.001) with an OR of 2.33 (p<0.0001; 95%CI: ). FFDM detected significantly more high risk tumors (35%) compared to FSM (27%)(p=0.011). Conclusion Aside from favorable clinicopathological factors, screen-detected cancers were also more likely to have a biologically low risk or even ultralow risk tumor. Especially for patients with screendetected cancers the use of tools, such as the 70-gene signature, to differentiate breast cancers by risk of recurrence may minimize overtreatment. The recent transition in screening-modalities led to an increase in the detection of biologically high risk cancers using FFDM. 100 Chapter 7

103 Introduction The increasing incidence in breast cancer after implementation of population-based mammographic screening programs has been suggested to be partly due to the detection of slow growing tumors that would never have caused symptoms or death, i.e. breast cancer overdiagnosis. 1 This lead time bias is related to the phenomenon of length time bias, as slow growing tumors have a longer window of opportunity to be detected in screening and therefore they are overrepresented in screen-detected cancers. 1 Whether this actually results in an increase in the detection of low risk tumors or even clinically indolent disease is still being investigated. 2,3 The concept of overdiagnosis due to screening was first reported in 1982 by Lundgren et al. 4 Estimates of the proportion of overdiagnosis were made by different study groups and are reported between 1% and 54%, depending on the denominators that are used. 5,6 In the Netherlands, there is an estimated 2.8% overdiagnosis. 6 Previous analyses, including our own, reported that screendetection is associated with a better prognosis for overall and breast-cancer-specific survival, independent of other favorable prognostic clinicopathological factors. 7 Screen-detected cancers are more often tumors of smaller size, lymph node-negative, low grade, and estrogen-receptor positive than interval cancers. 7 Identification of the patients with screen-detected cancers that are likely to be overdiagnosed based on clinicopathological factors remains difficult. Therefore, the hypothesis was generated that knowledge of the biological background of the tumor may be helpful in the identification of patients with screen-detected tumors at such a low risk of recurrence that they are likely to be overdiagnosed. Nowadays, gene-expression classifiers are used in addition to clinicopathological factors to identify patients with a favorable prognosis based on the biology of their tumor. 8 One of these gene-expression classifiers is the 70-gene signature (MammaPrint ), developed to improve the selection of those patients who may benefit from adjuvant systemic treatment. 8 The prognostic value of the 70-gene signature has been validated in several studies, both retrospectively and prospectively We previously reported on the tumor biology of screen-detected cancers and suggested that screen-detection might also be associated with a higher likelihood of a biologically low risk or even ultralow risk tumor assessed by the 70-gene signature. 2 7 Over the past decade a transition in diagnostic imaging has occurred. Most screening facilities switched from film-screen mammography (FSM) to full-field digital mammography (FFDM). In the Netherlands this transition started in 2008 and as of 2010, 94% of the women participating in the Dutch screening program have been screened using FFDM. 14 Several studies have evaluated the performance of FFDM compared to FSM and showed comparable or even better results for FFDM in the detection of clinically relevant tumors. 15,16 FFDM showed a higher sensitivity compared to FSM and detects more ductal carcinoma in situ (DCIS) and invasive cancers 17, especially in women under the age of 50 years and in pre- or perimenopausal women with radiographically Screening detects low risk tumor biology breast cancers 101

104 dense breasts. 16,17 Recent studies indicate that FFDM-detected cancers are more often estrogenreceptor-negative tumors. 17,18 A more sensitive screening-modality such as FFDM may also lead to an increase in the detection of biologically high risk tumors as assessed by the 70-gene signature. No differences in other clinicopathological factors, such as tumor size or grade, are described in literature. 15,16 The aim of this study is to determine the proportion of biologically high, low, and among those ultralow risk tumors among the screen-detected and interval tumors and to evaluate the impact of the transition from FSM to the more sensitive FFDM on the biology of the tumors detected by each screening-modality. Patients and Methods Patients and Clinicopathological characteristics All Dutch breast cancer patients enrolled in the MINDACT trial (EORTC-10041) 19,20, who were invited for the Dutch screening program, were included in this study. The MINDACT trial enrolled women aged years with histologically proven operable invasive breast cancer, no distant metastases, and for whom a frozen tumor sample was available between 2007 and ,20 Eligibility criteria included tumor stage T1, T2, or operable T3, and unilateral; DCIS or lobular carcinoma in situ (LCIS) provided invasive cancer is present; surgery options included breastconserving surgery or mastectomy combined with either a sentinel node procedure or full axillary clearance; WHO performance status of 0 or 1 and adequate bone marrow, liver and renal functions. Main exclusion criteria were: previous or concurrent cancer, previous chemotherapy, anticancer endocrine therapy or radiotherapy, and clinically significant impaired cardiac function. The protocol was amended in April 2008 to allow inclusion of 1 3 lymph node positive (N1) disease and genomic test in samples containing >30% of tumor cells. 19,20 Clinicopathological characteristics were obtained from the EORTC trial database. In case of discordance between a patients clinical risk estimation (based on Adjuvant! Online) and 70-gene signature result, the patient was randomized between treatment according to their clinical risk estimation or according to the 70-gene signature result. Screening Program The Dutch Screening Program started April 1, First, women aged years old and from 1998 women up to 75 years old were invited to participate in the screening program based on area code regions. Full coverage was achieved in ,21 Women were invited for biennial mammography. Screening mammograms were performed in independent and (mostly) mobile screening units (3-8 units per region). The images are read double-blind by trained radiologists. 102 Chapter 7

105 The current attendance rate is around 80%. 7,14 FFDM was rolled out as from 2008 and fully implemented in From each patient in this study data was collected on whether the most recent screening was by FSM or FFDM. Method of Detection Data on the method of detection was retrieved from the database of the Dutch screening organization. Data of all five regions is centrally collected in the ibob database. 14 The screening data for the eligible Dutch MINDACT patients was derived from the ibob database based on demographic information. Patients were eligible if they were 49 years or older at the time of diagnoses and were invited to participate in the Dutch screening program (n=1475). One hospital excluded their patients (n=4) from the linkage protocol and 62 patients could not be matched to the ibob database, due to incomplete demographic information. Of the 1409 patients that were matched to the ibob database, 1165 were identified as participants of the screening program. Two types of breast cancer were identified based on the method of detection. First, the screendetected cancers, defined as breast cancers that were mammographically detected in the first (prevalent cancers, n=115) or a subsequent screening round (incident cancers, n=660) (total n=775). Second, the interval cancers, defined as symptomatic cancers that were diagnosed within 30 months of a negative screening (n=390). Screening is biennial, giving a window of 24 months for an interval cancer to become symptomatic after a negative screening mammography. When a woman moves to another area-code, her next screening could be delayed up to 6 months. Therefore the interval of 30 months was chosen. 70-gene signature In this study we used the 70-gene signature to evaluate tumor biology. For all patients included in the MINDACT trial a 70-gene signature result was available. The 70-gene signature, MammaPrint (Agendia NV, Amsterdam, the Netherlands), is a gene-expression classifier used to estimate the risk of developing distant metastasis. The result of the 70-gene signature is presented as a binary result (good or poor prognosis), which is derived from an index score (-1 to 1). 9,10 An index-score greater than 0.4 is classified as good prognosis (low risk) and an index-score less than 0.4 is classified as poor prognosis (high risk). For this study we also applied the previously set threshold to identify patients with an ultralow risk of distant recurrence (index-score >0.6). 2 Within the low risk group of the original 78 patients used to develop this classifier, no distant metastases were observed at five years in patients who had an index-score greater than ,9 7 Screening detects low risk tumor biology breast cancers 103

106 Statistical Analysis Baseline characteristics for Screen-detected and interval cancers were compared and the proportions of 70-gene signature high, low, and among the latter the ultralowrisk were calculated. We performed separate analyses for FSM and FFDM. Prognostic factors, such as age, tumor size, histological type, estrogen-receptor (ER), progesterone-receptor (PR), and HER2/neu-oncoprotein (ERBB2) were evaluated in a logistic regression model. Hereafter, tumor biology related factors are referred to as prognostic factors. Only factors that resulted in <10% change in the coefficient of association of the 70-gene signature with the method of detection were included in the multivariate analyses. Calculations were done using SPSS (version 19.0). A two-sided p-value of less than 0.05 was considered statistically significant. Results Patient characteristics The clinicopathological characteristics of the 1165 included patients are described in Table 1, stratified by method of detection, and in Supplementary Table 1 also stratified by 70-gene signature result. Screen-detected cancers were more often of smaller size (<2 cm), ER- and PRpositive, HER2-negative, grade I, without nodal involvement compared to interval cancers. 104 Chapter 7

107 Table 1. Breast cancer patients eligible to participate in the Dutch screening program: Patient and tumor characteristics stratified by method of detection Screen-detected cancers (SD) Interval cancers (IC) p-value # SD vs IC n=775 n= gene signature High risk 244 (32%) 185 (47%) < Low risk 242 (31%) 111 (29%) Ultralow risk 289 (37%) 94 (24%) Age (years) (27%) 103 (26%) (25%) 104 (27%) (26%) 94 (24%) (22%) 88 (23%) Size T1 (<20 mm) 613 (79%) 247 (63%) < T2 (20-50 mm) 160 (21%) 139 (36%) T3 (>50 mm) 2 (0.3%) 4 (1%) Lymph node status Negative 680 (88%) 315 (81%) positive nodes 95 (12%) 75 (19%) Histological type Ductal 643 (83%) 316 (81%) Lobular 76 (10%) 49 (13%) Mixed 28 (4%) 9 (2%) Other 28 (4%) 16 (4%) Grade Grade I 244 (32%) 59 (15%) < Grade II 356 (46%) 170 (44%) Grade III 174 (23%) 160 (41%) Undefined 1 1 ER status Negative 77 (10%) 80 (21%) < Positive 698 (90%) 310 (80%) PR status Negative 188 (24%) 138 (35%) < Positive 573 (74%) 242 (62%) Unknown HER2 status Negative 680 (88%) 325 (83%) Positive 94 (12%) 64 (16%) Unknown # Chi-square test ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal growth factor Receptor 2. Screening detects low risk tumor biology breast cancers 105

108 70-gene signature for screen-detected and interval cancers Among the screen-detected cancers, 32% had a 70-gene signature high risk and 68% a low risk tumor, of which 54% had a ultralow risk tumor (37% of total) (Figure 1A and Table 1). Among the interval cancers, 47% had a high risk and 53% a low risk tumor, of which 46% could be defined as ultralow risk tumor (24% of total). A significant difference was seen between screen-detected and interval cancers (px 2 test = 0.001) in 70-gene signature high, low, and ultralow risk groups. Of the prevalent tumors, detected in the first screening round, 19% had a 70-gene signature high risk and 81% a low risk tumor. Among the low risk prevalent tumors about 63% even had an ultralow risk tumor (51% of total)(figure 1B). Of the incident tumors, detected in subsequent screening rounds, 34% had a 70-gene signature high risk and 66% a low risk tumor. Among the low risk incident tumors, 52% could be defined as ultralow risk (35% of total)(px 2 test prevalent vs incident < ) (Figure 1B). When excluding the prevalent cancers from these analyses the significant difference between screen-detected and interval cancers remained (Supplementary Table 2). In a univariate analyses, patients with screen-detected cancers were two-times more likely to have an ultralow risk tumor compared to patients with an interval cancer (OR high vs ultralow: 2.33 (95%CI: ; p<0.0001)(table 3A). When adjusting for intermediate factors such as ER-status and tumor size, this significant association remained (Table 3A). However, when adjusting for grade the 70-gene signature was no longer a significant factor; likely due to a substantial correlation between the 70-gene signature and grade (ρ=0.393). The analyses mentioned above lead to similar conclusions in ER-positive patients only (data not shown). 54% ultralow among low 46% ultralow among low Figure 1A. Proportions of 70-gene signature result among screen-detected and interval cancers 63% ultralow among low 52% ultralow among low Figure 1B. Screen-detected cancers detected in first vs subsequent screening rounds 106 Chapter 7

109 Table 2. Breast cancer patients with screen-detected cancers: Patient and tumor characteristics stratified by film-screen or digital mammography Film screen mammography n=315 Full field digital mammography n=459 p-value # 70-gene signature High risk 85 (27%) 159 (35%) 0.04 Low risk 98 (31%) 143 (31%) Ultralow risk 132 (42%) 157 (34%) Age (years) yrs 77 (24%) 130 (28%) yrs 83 (26%) 110 (24%) yrs 81 (26%) 119 (26%) yrs 74 (24%) 96 (21%) Tumor size T1 (< 20 mm) 254 (81%) 358 (78%) T2 (20-50 mm) 60 (19%) 100 (22%) T3 (>50 mm) 1 (0 3%) 1 (0 2%) Lymph node status Negative 290 (92%) 389 (85%) Positive 25 (8%) 70 (15%) Histological type Ductal 269 (85%) 373 (81%) Lobular 28 (9%) 48 (11%) Mixed 11 (4%) 17 (4%) Other 7 (2%) 20 (4%) Grade I 111 (35%) 133 (29%) II 143 (45%) 213 (46%) III 61 (19%) 112 (24%) Unknown 0 1 ER status Negative 28 (9%) 49 (11%) Positive 287 (91%) 410 (89%) PR status Negative 79 (25%) 109 (24%) Positive 233 (74%) 339 (74%) Unknown 3 11 HER2 status Negative 271 (86%) 409 (89%) Positive 43 (14%) 50 (11%) Unknown # Chi-square test ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal growth factor Receptor 2. Screening detects low risk tumor biology breast cancers 107

110 Table 3A. Unadjusted and adjusted* Odd s Ratio s of the tumor biology among screen-detected vs interval cancers # Unadj. OR Adj. OR* (95%CI) p-value # (95%CI) p-value # 70-gene signature ultralow vs low 1.41 ( ) ( ) ultralow vs high 2.33 ( ) < ( ) gene signature + ER status 70-gene signature ultralow vs low 1.40 ( ) gene signature ultralow vs high 1.95 ( ) < ER status positive vs negative 1.68 ( ) gene signature + Grade 70-gene signature ultralow vs low 1.19 ( ) gene signature ultralow vs high 1.37 ( ) Grade I vs II 1.84 ( ) Grade I vs III 3.15 ( ) < gene signature + Tumor size 70-gene signature ultralow vs low 1.38 ( ) gene signature ultralow vs high 2.15 ( ) < T1 vs T ( ) < T1 vs T3 5.4 ( ) # Logistic regression model * Adjusted for grade, estrogen receptor status and tumor size Film screen versus full field digital mammography Between 2007 and 2011 a transition was seen in screening-modality used for the last screening before diagnoses. Supplementary Figure 1 displays this transition in this cohort over time. Among the screen-detected cancers, 41% were detected using FSM (n=315) and 59% were detected using FFDM (n=459). FSM detected 27% high risk and 73% low risk tumors of whom 57% could be defined as ultralow risk (42% of total). This is significantly different compared to cancers detected using FFDM (p X 2 test = 0.011), which detected 35% high risk and 65% low risk tumors of whom 51% could be defined ultralow risk (34% of total)(figure 2 and Table 2). Aside from a difference in tumor biology in tumors detected by FSM versus FFDM, there is also a difference in nodal involvement. For tumors detected by FSM 8% had one or more positive lymph nodes, while for tumors detected by FFDM 15% had one or more positive lymph nodes (px 2 test = 0.002). For other patient and tumor characteristics, such as age, size, histological type, grade, ER, PR, and HER2 status, no significant differences were seen between the two screening-modalities (Table 2). The association of nodal status with FFDM was at least partly attributable to the amendment of the MINDACT study in 2008, which allowed patients with 1-3 positive nodes to be included in the trial. This leads to an increase of nodal positive patients over the years (data not shown), however, nodal status was not associated with the 70-gene signature result (Supplementary Table 1). 108 Chapter 7

111 57% ultralow among low 51% ultralow among low Figure 2. Screen-detected cancers using film-screen vs digital mammography Overall, the proportion of interval cancers among the screened women within the Dutch MINDACT cohort was 33% (390/1164). In the FSM-screened population (n=624) the proportion of interval cancers was 49.5% (309/624), while for the FFDM-screened population (n= 540) the interval rate was 15% (81/540). Among the FSM interval cancers, which became symptomatic within 30 months after a negative FSM (n=309), 46% had a high risk and 54% had a low risk tumor of whom 54% had an ultralow risk tumor (Figure 3A). Among the FFDM interval cancers, which became symptomatic within 30 months after a negative FFDM (n=81), 54% had a high risk and 46% had a low risk tumor of whom 46% an ultralow risk tumor (Figure 3B). Odd s ratios for FSM and FFDM are shown in Table 3B. There was no effect-modification of screening-modality in the association between the 70-gene signature and screen-detected vs interval cancers. These proportions in tumor biology remained the same for FSM and FFDM when only including those patients that were diagnosed after the amendment. Sensitivity analyses in the period when FFDM screening was implemented in at least half of the population and potentially two years had passed for women with a negative FFDM screen in order for interval cancers to become manifest, i.e and 2010, showed similar proportions of high risk tumors among FSM- and FFDMscreened patients (26.2% FSM and 33.0% FFDM). 7 Screening detects low risk tumor biology breast cancers 109

112 57% ultralow among low 54% ultralow among low 57% ultralow among low 46% ultralow among low Figure gene signature proportions among screen-detected and interval cancers after A. film-screen mammography or B. full field digital mammography Table 3B. Unadjusted and adjusted* Odd s Ratio s of the tumor biology among interval vs screen-detected cancers for film-screen and digital mammography # Unadj. OR Adj. OR* (95%CI) p-value # (95%CI) p-value # Last screen FSM 70-gene signature ultralow vs low 1.59 ( ) ( ) gene signature ultralow vs high 2.84 ( ) < ( ) Last screen FFDM 70-gene signature ultralow vs low 1.29 ( ) ( ) gene signature ultralow vs high 2.56 ( ) ( ) # Logistic regression model * Adjusted for grade, estrogen receptor status and tumor size Discussion The effectiveness of breast cancer screening is extensively debated, particularly regarding the estimated proportion of overdiagnosed cancers. 3,22 Identification of these overdiagnosed screendetected cancers is challenging. Screen-detected cancers have shown to have more favorable clinicopathological factors and better outcome compared to interval cancers. 7 Our results also show that the majority of the cancers detected in screening (68%) are biologically low risk and over half of the low risk tumors are even ultralow risk. This indicates that knowledge of the biological background may help to identify those screen-detected breast cancers at such a low risk of recurrence that concerns about overdiagnosis can be raised. Especially for this subgroup of patients overtreatment with chemotherapy should be avoided. To determine whether the group with screen-detected ultralow-risk tumors is indeed overdiagnosed, a randomized controlled trial would provide further insight. Mammographic screening on the other hand, has proven to be an effective way to detect breast cancer at an early stage. 23 Our results confirm that screening 110 Chapter 7

113 also detects cancers with poor prognosis tumor biology, which are at a high risk of recurrence. Almost one third of the patients with a tumor detected in the screening program had a high risk 70-gene signature result. The 70-gene signature is likely to be a useful tool to separate patients at a high risk from those at a low or even an ultralow risk of recurrence. Patients with a screendetected cancer are two-times more likely to have an ultralow risk tumor compared to interval cancers. Even when adjusting for other prognostic factors with a substantial association with method of detection (in our population tumor size, grade, and ER status), the 70-gene signature remained an important prognostic factor. Previous analyses showed that the proportion of low and ultralow risk tumors among screen-detected cancers is higher compared to symptomatic cancers diagnosed before the introduction of screening. 2 Our current results validate this finding in a larger cohort, showing 68% low risk among screen-detected cancers of whom 54% had an ultralow risk. In literature it is still debated whether the prevalent screen-detected cancers should be included when analyzing screen-detected cancers. 23 In this study, we aimed to look at screen-detected cancers from a different, more biologically oriented perspective to evaluate the type of tumors that are detected in screening programs. Since prevalent cancers are also screen-detected and a substantial proportion of overdiagnosis may be present in this subgroup, they were included in our analyses. Good prognosis for prevalent cancers has been suggested by others 1, and our observation on the biological level support that notion, albeit not significant. The number of prevalent cancers in this cohort is low and in univariate analyses the screening round was not a significant prognostic factor. The recent transition from FSM to FFDM resulted in a larger proportion of high risk tumors among the screen-detected cancers, which may indicate that the introduction of FFDM leads to the detection of more aggressive cancers with a worse prognosis. It may also indicate that breast cancer screening using FFDM is even more effective than when using solely FSM. Given the possibility that high risk tumors that used to be missed in screening are now detected with FFDM, the introduction of FFDM might be responsible for an increase in the proportion of high risk tumors among the screen-detected cancers and decrease in the number of interval cancers. The fact that the proportion of interval cancers among FFDM-screened patients was low (15%) may therefore be a result of more sensitive screening, but can also explained by the fact that the accrual of women to FFDM was in transition from 2008 till Hence, for many women insufficient time had passed after a negative FFDM for the development of interval cancers. Thus, the ratio between the number of women at risk for a screen-detected tumor versus an interval cancer, is lower for FSM compared to FFDM. Therefore, no conclusions regarding the relative amount of interval cancers for FFDM versus FSM can be drawn based on the data presented here. Since the Dutch screening program is still collecting data on the effect of the transition from FSM to FFDM, we were not able compare our result to those of the entire screened population in the Netherlands. Of note is that the MINDACT trial currently only has available data of the tumor 7 Screening detects low risk tumor biology breast cancers 111

114 samples provided by the local pathology departments. Tumor-characteristics, especially grade, may change after central review of the samples. A limitation is the possibility of selection bias in the MINDACT trial itself. The novelty of gene-signatures and the limited experience of doctors with this new prognostic tool may have resulted in the inclusion of patients with more favorable tumor characteristics in the beginning of the trial. In conclusion, screen-detection was found to be associated with a higher likelihood of a 70-gene signature biologically low risk tumor, which prospectively validates our previous analyses. 2 Half of all screen-detected low risk tumors even had an ultralow risk of distant metastases. Especially for this screen-detected patient group the use of tools to differentiate breast cancers by risk of recurrence may minimize overtreatment. Second, the transition from FSM to FFDM resulted in the detection of a larger proportion of high risk tumors, which may indicate that FFDM is a more effective screening-modality than FSM. Contributors CD, MKS, LvtV, ER, FC, and JB were responsible for the study design and development of the protocol. CD coordinated the study and collected the data on method of detection. FC, LS and JB provided the MINDACT baseline characteristic data. CD and MKS performed the data analysis. CD, MKS, LE, KK, ER, LS, JB, and LvtV took part in data interpretation and manuscript writing. All authors were involved in reviewing the report. This study was approved by the MINDACT steering committee, and is confirmed to be in line with the principles of the sponsor of the trial, EORTC POL008 ( Release of data from EORTC studies for use in External Research Projects ). All patients have given written informed consent before enrolment in the MINDACT trial (EORTC 10041/BIG 3-04).This informed consent allowed linkage to the Dutch screening facilities. Funding source This work was supported by the EORTC Breast Cancer Group (type 3 grant 2011/2012), the Dutch Cancer Society (NKI ), BBMRI-NL (NWO , complementation project 45) and the Dutch Genomics Initiative Cancer Genomics Centre. The funding sources had no role in the study design, data collection, data analysis, data interpretation, in writing the report, or in the decision to submit for publication. CD, MKS and LvtV had full access to all the data. CD, MKS, ER and LvtV had final responsibility for the decision to submit for publication. Conflict of interest We have read and understood the BMJ Group policy on declaration of interests and declare the following interests: LvtV is named inventor on the patent for the 70-gene signature used in this study. LvtV reports being shareholder in and employed by Agendia NV, the commercial company that markets the 70-gene signature as MammaPrint. 112 Chapter 7

115 Acknowledgements We acknowledge the contribution of the European Organization for Research and Treatment of Cancer (EORTC) and the TransBig Consortium. We thank the Dutch Screening Facilities, Frank Yntema in particular, for providing the screening data used in this study. We thank Annuska Glas from Agendia for providing value information on the 70-gene signature results. We are indebted to all the Dutch women who participated in the MINDACT trial. 7 Screening detects low risk tumor biology breast cancers 113

116 References 1 Nagtegaal ID, Allgood PC, Duffy SW, Kearins O, Sullivan EO, Tappenden N, et al. Prognosis and pathology of screen-detected carcinomas: how different are they? Cancer 2011;117(7): Esserman LJ, Shieh Y, Rutgers EJ, Knauer M, Retel VP, Mook S, et al. Impact of mammographic screening on the detection of good and poor prognosis breast cancers. Breast Cancer Res Treat 2011;130(3): Esserman, LJ, Thompson IM, and Reid B. Overdiagnosis and Overtreatment in Cancer: an Opportunity for Improvement. JAMA 2013;310(8): Lundgren B, Helleberg A. Single oblique-view mammography for periodic screening for breast cancer in women. J Natl Cancer Inst 1982;68(3): Puliti D, Duffy SW, Miccinesi G, de Koning H, Lynge E, Zappa M, et al. Overdiagnosis in mammographic screening for breast cancer in Europe: a literature review. J Med Screen 2012;19 Suppl 1: de Gelder R, Heijnsdijk EA, van Ravesteyn NT, Fracheboud J, Draisma G, De Koning HJ. Interpreting overdiagnosis estimates in population-based mammography screening. Epidemiol Rev 2011;33: Mook S, van t Veer LJ, Rutgers EJ, Ravdin PM, van de Velde AO, van Leeuwen FE, et al. Independent prognostic value of screen detection in invasive breast cancer. J Natl Cancer Inst 2011;103(7): Azim HA, Jr., Michiels S, Zagouri F, Delaloge S, Filipits M, Namer M, et al. Utility of prognostic genomic tests in breast cancer practice: The IMPAKT 2012 Working Group Consensus Statement. Ann Oncol 2013;24: van t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415(6871): van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347(25): Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas A, et al. Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 2006;98: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C, et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009;117: Drukker CA, Bueno-de-Mesquita JM, Retel VP, van Harten WH, van Tinteren H, Wesseling J, et al. A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 2013;133(4): National Evaluation Team for Breast Cancer Screening in the Netherlands. NETB report Bluekens AM, Holland R, Karssemeijer N, Broeders MJ, den Heeten GJ. Comparison of digital screening mammography and screen-film mammography in the early detection of clinically relevant cancers: a multicenter study. Radiology 2012;265(3): Vinnicombe S, Pinto Pereira SM, McCormack VA, Shiel S, Perry N, Dos Santos Silva IM. Full-field digital versus screen-film mammography: comparison within the UK breast screening program and systematic review of published data. Radiology 2009;251(2): Pisano ED, Gatsonis C, Hendrick E, Yaffe M, Baum JK, Acharyya S et al. Diagnostic performance of digital versus film mammography for breast-cancer screening. N Engl J Med 2005;353(17): Kerlikowske K, Hubbard RA, Miglioretti DL, Geller BM, Yankaskas BC, Lehman CD, et al. Comparative effectiveness of digital versus film-screen mammography in community practice in the United States: a cohort study. Ann Intern Med 2011;155(8): Chapter 7

117 19 Cardoso F, van t Veer L, Rutgers E, Loi S, Mook S, Piccart-Gebhart MJ. Clinical application of the 70- gene profile: the MINDACT trial. J Clin Oncol 2008;26(5): Rutgers E, Piccart-Gebhart MJ, Bogaerts J, Delaloge S, Veer LV, Rubio IT, et al. The EORTC 10041/BIG MINDACT trial is feasible: results of the pilot phase. Eur J Cancer 2011;47(18): De Koning HJ, Fracheboud J, Boer R, Verbeek AL, Collette HJ, Hendriks JH et al. Nation-wide breast cancer screening in The Netherlands: support for breast-cancer mortality reduction. National Evaluation Team for Breast Cancer Screening (NETB). Int J Cancer 1995;60(6): Esserman L, Shieh Y, Thompson I. Rethinking screening for breast cancer and prostate cancer. JAMA 2009;302(15): Marmot MG, Altman DG, Cameron DA, Dewar JA, Thompson SG, Wilcox M. The benefits and harms of breast cancer screening: an independent review. Br J Cancer 2013;108(11): Screening detects low risk tumor biology breast cancers 115

118 Supplementary Table 1. Patient characteristics by method of detection and 70-gene signature # Screen-detected (n=775) Interval (n=390) 70-gene signature High risk Low risk Ultralow risk High risk Low risk Ultralow risk n=244 n=242 n=289 n=185 n=111 n=94 Age p=0.136 p= (23%) 60 (25%) 93 (32%) 49 (26%) 27 (24%) 27 (29%) (26%) 67 (28%) 62 (21%) 53 (29%) 24 (22%) 27 (29%) (27%) 61 (25%) 73 (25%) 44 (24%) 27 (24%) 23 (24%) (24%) 51 (21%) 60 (21%) 38 (21%) 33 (30%) 17 (18%) >70 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) Size p=0.025 p=0.014 T1 <20 mm 176 (72%) 199 (82%) 238 (82%) 107 (58%) 68 (61%) 72 (77%) T mm 67 (27%) 43 (18%) 50 (17%) 77 (42%) 42 (38%) 20 (21%) T3 >50 mm 1 (0%) 0 (0%) 1 (0%) 1 (1%) 1 (1%) 2 (2%) Histological type p=0.006 p=0.001 Ductal 216 (89%) 209 (86%) 218 (75%) 164 (89%) 87 (78%) 65 (69%) Lobular 15 (6%) 19 (8%) 42 (15%) 10 (5%) 17 (15%) 22 (23%) Mixed 5 (2%) 9 (4%) 14 (5%) 2 (1%) 3 (3%) 4 (4%) Other 8 (3%) 5 (2%) 14 (5%) 9 (5%) 4 (4%) 3 (3%) Grade p< p< Grade I 15 (6%) 80 (33%) 149 (52%) 7 (4%) 23 (21%) 29 (31%) Grade II 97 (40%) 133 (55%) 126 (44%) 50 (27%) 59 (53%) 61 (65%) Grade III 132 (54%) 28 (12%) 14 (5%) 128 (69%) 28 (25%) 4 (4%) Undefined 0 (0%) 1 (0%) 0 (0%) 0 (0%) 1 (1%) 0 (0%) ER status p< p< Negative 69 (28%) 8 (3%) 0 (0%) 79 (43%) 0 (0%) 1 (1%) Positive 175 (72%) 234 (97%) 289 (100%) 106 (57%) 111 (100%) 93 (99%) PR status p< p< Negative 111 (45%) 46 (19%) 31 (11%) 115 (62%) 12 (11%) 11 (12%) Positive 128 (52%) 190 (79%) 255 (88%) 68 (37%) 94 (85%) 80 (85%) Unknown 5 (2%) 6 (2%) 3 (1%) 2 (1%) 5 (5%) 3 (3%) HER2 status p< p< Negative 191 (78%) 214 (88%) 275 (95%) 136 (74%) 100 (90%) 89 (95%) Positive 53 (22%) 28 (12%) 13 (4%) 49 (26%) 10 (9%) 5 (5%) Unknown 0 (0%) 0 (0%) 1 (0%) 0 (0%) 1 (1%) 0 (0%) LN status p=0.649 p=0.354 N0 218 (89%) 211 (87%) 251(87%) 155 (84%) 87 (78%) 73 (78%) N1 26 (11%) 31 (13%) 38 (13%) 30 (16%) 24 (22%) 21 (22%) # Chi-square test ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal growth factor Receptor Chapter 7

119 Supplementary Table gene signature results for screen-detected and interval cancers with and without prevalent cases 70-gene signature Screen-detected incl prevalent Interval incl prevalent p-value# Screen-detected excl prevalent Interval excl prevalent p-value# High risk 244 (31.5%) 185 (47.4%) < (33.6%) 154 (47.8%) < Low risk 242 (31.2%) 111 (28.5%) 207 (31.5%) 88 (27.3%) Ultralow risk 289 (37.3%) 94 (24.1%) 230 (35.0%) 80 (24.8%) Total # Chi-Square test 7 Supplementary figure 1. Distribution of film-screen vs digital mammography over time Screening detects low risk tumor biology breast cancers 117

120 Chapter 8

121 Gene-expression profiling to predict the risk of locoregional recurrence in breast cancer To be submitted Caroline A. Drukker* Sjoerd G. Elias* Matthijs V. Nijenhuis* Jelle Wesseling Harry Bartelink Paula Elkhuizen Barbara Fowble Pat Whitworth Rakesh Patel Laura J. van t Veer Peter D. Beitsch Emiel J.Th. Rutgers *authors contributed equally

122 Abstract Background The 70-gene signature (MammaPrint ) has been developed to predict the risk of distant metastases in breast cancer and select those patients who may benefit from adjuvant treatment. Given the strong association between locoregional and distant recurrence, we hypothesize that the 70-gene signature will also be able to predict the risk of locoregional recurrence (LRR). Methods 1053 breast cancer patients primarily treated with breast conserving treatment (BCT) or mastectomy at the Netherlands Cancer Institute between were included. Adjuvant treatment consisted of radiotherapy, chemotherapy and/or endocrine therapy as indicated by guidelines used at the time. All patients were included in various 70-gene signature studies. Results After a median follow-up of 8.96 years, patients with a high risk 70-gene signature (n=492) had a LRR risk of 12.6 % (95%CI: ) at 10 years, compared to 6.1% (95%CI: ) for low risk patients (n=561)(p<0.0001), respectively. Adding the 70-gene signature to a Coxproportional-hazard model including clinicopathological factors, such as age, tumor size, grade, hormone receptor status, lymphovascular invasion, axillary lymph node involvement, surgical treatment and systemic treatment (endocrine and chemotherapy) resulted in a multivariable HR of 1.73 (95%CI: ; p=0.042). An increase of the C index from (95% CI: ) in a prediction model with solely clinicopathological factors to (95% CI: ) after adding the 70-gene signature is seen. Conclusion The 70-gene signature is able to predict the risk of LRR. A significantly lower incidence of LRR in patients with a low risk 70-gene signature result compared to those with high risk 70-gene signature result, independent of known risk factors was seen. 120 Chapter 8

123 Introduction For the majority of breast cancer patients locoregional recurrence (LRR) is becoming a less common problem. Improvements in patient selection, surgical treatment, radiotherapy techniques and (neo)adjuvant systemic therapy have led to a substantial decrease in LRR incidence rates. 1,2 Still, on average 3% of the patients do experience a LRR within 5 years after diagnosis. After 10 years of diagnosis this is around 6%. The favorable LRR rates raise the issue of overtreatment in women with a low or limited probability to recur locally and on the other hand undertreatment in those patients at a higher risk of LRR. Ideally, one would aim to identify patients at a high risk of LRR to better guide optimal locoregional treatment with more extensive surgery, radiotherapy and adjuvant systemic treatment, while at the same time identification of patients at a low risk of LRR can help to avoid unnecessary adjuvant radiotherapy in these patients. Currently, traditional clinicopathological factors such as age, grade, tumor size, lymphovascular invasion (LVI), hormone receptor status and involvement of axillary lymph nodes are used to predict the risk of LRR. Based on these factors, LRR rates as described earlier can be achieved, but further discrimation to assess LRR risk and to evaluate possible overtreatment seems not possible. Aside from these clinicopathological factors, the biological background of breast cancer may be of help in further assessing the risk of recurrence. 3 Gene expression classifiers, such as the 70- gene signature (MammaPrint, Agendia NV, Amsterdam, the Netherlands), have proven to be a useful additional tool for assessing the risk of distant recurrence in breast cancer. 4-6 The 70-gene signature has been extensively validated on historic data and has recently been prospectively evaluated in the microarray-prognostics-in-breast-cancer (RASTER) study. 5-8 The results show a favorable distant-recurrence-free interval for patients with a low risk 70-gene signature result after 5 years, even in the absence of adjuvant systemic treatment and despite poor clinicopathological factors. 5 The test has proven to be able to predict the risk of distant metastases in the individual patient based on the biological background of the tumor. 4,6 Since locoregional recurrence is an independent predictor of subsequent distant metastases 9, we hypothesize that the 70-gene signature will be able to predict the risk of LRR as well. The aim of this study is to evaluate the performance of the 70-gene signature in the prediction of LRR and its additional value to clinicopathological factors that are currently used. 8 Patients and methods Patients All 1053 individual breast cancer patients included in eight 70-gene signature studies, who were not included in the 70-gene signature training set and diagnosed and treated at the Netherlands Cancer Institute (NKI), were eligible for the current study (Study flow chart, Figure 1). Only NKI 70-gene signature predicts locoregional recurrence 121

124 treated patients were included to facilitate standardized ascertainment of locoregional events during the extended follow-up period and to allow an update of radiotherapy information. Details of study design, rationale, and patient eligibility of seven of the included studies have been described elsewhere. 6,7,10-14 In short, all patients were women with histologically proven, operable, invasive breast cancer (T1-3N0-1M0), diagnosed between 1984 and All patients were primarily treated with mastectomy or breast conserving therapy (BCT), comprising lumpectomy followed by whole breast irradiation. All patients had tumor free resection margins. Radiotherapy comprised whole breast irradiation, mostly consisting of 50 Gy in 25 fractions, with or without a boost dose of 16 Gy after BCT and chestwall and/or internal mammary chain radiotherapy after mastectomy. Systemic adjuvant treatment consisted of chemotherapy and/or endocrine therapy as indicated by guidelines used at the time. One of the studies included in this pooled analyses was not yet published at the time these analyses were performed. In this study Bedard et al. included 252 women aged 65 years and older diagnosed with early stage breast cancer (T1-3N0-1M0) at the NKI between 1987 and None of the patients received adjuvant chemotherapy. All individual studies complied with Ethical Review Board standards. The 78 patients included in the cohort used to develop the 70-gene signature were not included in these analyses. Median follow-up of this cohort of 1053 individual patients was 8.96 years. Figure 1. Study flow chart 122 Chapter 8

125 Clinicopathological factors and the 70-gene signature Information on age, grade, estrogen receptor (ER), progesterone receptor (PR), LVI, tumor size and involvement of axillary lymph nodes were derived from the original study data that included pathological review by an expert (Hans Peterse). LVI was not documented in all studies. Therefore, tumor samples of 150 patients were revised by an experienced breast pathologist (JW) for this study. Frozen tumor samples from each patient were processed at Agendia s laboratory (Amsterdam, the Netherlands) for RNA isolation, amplification, and labeling as described elsewhere. 4,6,15 To assess the mrna expression level of the 70 genes, RNA was hybridized to a custom-designed array, commercially available as MammaPrint. Agendia NV is ISO17025-certified, CLIA accredited and FDA-cleared. Tumors were classified as a 70-gene signature low or high risk at the time of the initial studies. Low risk was defined as an index-score greater than 0.4. High risk was defined as an index-score lower than ,15 Follow-up Follow-up for locoregional recurrence and death was updated through November 2011 using data from the NKI Tumor Registry complemented with review of the original patient records. Locoregional recurrence was defined as reappearance of breast cancer in the ipsilateral breast or chest wall or ipsilateral regional lymph node involvement, six months or longer after diagnosis. Ipsilateral supraclavicular lymph node involvement was included as regional recurrence throughout the follow-up period, although TNM-editions 4 and 5 considered such recurrences M1 instead of N3. Statistical analysis Detailed description of the statistical analyses is described in appendix 1. In short, analyses for time to LRR were performed using competing risk analyses as not to overestimate the absolute LRR risk. 16 For this, follow-up time started at diagnosis and ended at the first manifestation of LRR (event) or death (competing event), or at the end of follow-up without LRR or death (censored). Occurrences of distant metastases, contralateral breast cancer, or second primary tumors were not considered censoring events nor competing risks. The univariable 5- and 10-year absolute risk of LRR for the 70-gene signature high and low risk groups was estimated using the cumulative incidence function, 17 and compared using Gray s test. 18 Multivariable analyses were performed using Fine and Gray competing risk regression. 19 A multivariable model was constructed comprising solely of routine clinicopathological factors and treatment. To this model the 70-gene signature was added to evaluate its additional and independent prognostic value. The combined prognostic performance of the multivariable models was evaluated with regard to discrimination (Harrell s C index adapted to competing risk analyses 16 ) and calibration. Model improvement upon addition of the 70-gene signature was tested by the plr test, by the improvement in C index. Analyses were performed using R version All statistical tests were two-sided with a 8 70-gene signature predicts locoregional recurrence 123

126 cutoff for statistical significance of 5%. Estimates are reported together with 95% confidence intervals. For the C index, 2000-fold bootstrapping was used for statistical testing and standard error estimation. Results Patient and tumor characteristics by 70-gene signature 70-gene signature low risk patients (n=561) were more often of older age at the time of diagnosis, had smaller tumors of a lower grade, being ER-positive, PR-positive and HER2-negative as compared to the 70- gene signature high risk patients (n=495)(table 1). More chemotherapy was administered to 70-gene signature high risk patients, while no significant difference was seen in the administration of endocrine therapy between high and low risk patients. No significant difference was seen between high and low risk patients regarding their initial type of surgical treatment and no significant difference in the administration of radiotherapy after both types of surgical treatment. Type of LRR Through 10 years of follow-up (median 8.96 years; IQR ), 87 LRR events occurred; in 29 patients who had a low risk 70-gene signature primary cancer and 58 who had a high risk 70-gene signature cancer. Thirty-nine of the 470 patients treated with BCT developed an LRR event compared to 47 out of 555 patients treated with mastectomy. Of one patient with a LRR event data on surgical treatment is missing. Most common site for regional recurrence was the supraclavicular area for both patients treated with BCT and with mastectomy. Association between 70-gene signature and risk of LRR Patients with a low risk 70-gene signature tumor had a LRR risk of 2.7% (95%CI: ) at 5 years and 6.1% (95%CI: ) at 10 years (Figure 2A). Patients with a 70-gene signature high risk tumor had a LRR risk of 9.1% (95%CI: ) at 5 years and 12.6% (95%CI: ) at 10 years (p<0.001). Univariable probabilities of risk of LRR stratified by surgery and radiotherapy are shown in Figure 2B, 2C and 2D. 124 Chapter 8

127 Table 1. Patient and tumor characteristics stratified by 70-gene signature high and low risk. Age 50 year > 50 year GS low risk n=561 (%) (3.5) (66.5) GS high risk n=492 (%) (47.2) (52.8) p-value <0.001 Tumor size 22 mm 24 mm Nodal status Node positive Node negative Missing Grade Missing ER status Positive Negative Missing PR status Positive Negative Missing HER2 status Positive Negative Missing Surgical treatment Mastectomy Local RT (48) (51) (41.5) (46.1) (8.6) (3.7) (96.6) (2.9) (0.5) (79.8) (17.3) (2.9) (3.6) (78.6) (17.8) (53.2) (51% of MST) (54) (45.5) (10.5) (31.3) (55.9) (2.2) (64.4) (35.2) (0.2) (45.3) (51.8) (2.8) (20.3) (65) (14.7) <0.001 <0.001 <0.001 <0.001 (52.2) (57% of MST) 0.13 Breast conserving surgery Local RT Missing Radiotherapy RT RT + boost No RT Missing (45) (98% of BST) (1.8) (43.9) (27.3) (26.9) (1.9) (44.1) (100% of BST) (3.7) (47.1) (26.6) (22.8) (3.9) Endocrine therapy 263 (46.9) 212 (43.1) Chemotherapy 105 (18.7) 188 (38.2) < ER=estrogen receptor; PR=progesterone receptor; HER2=Human Epidermal growth factor Receptor 2 70-gene signature predicts locoregional recurrence 125

128 Figure 2. Risk of locoregional recurrence for the entire cohort stratified for 70-gene signature low risk and high risk. Additional value of the 70-gene signature to clinicopathological factors Patients with a high risk 70-gene signature had an 2.40 times higher risk of LRR than patients with a low risk 70-gene signature (univariable hazard ratio (HR) 2.40; 95%CI: )(Table 2). Other significant prognostic factors in the univariable model were age (HR non-linear; p<0.001), grade (HR 2.91 (grade 3 vs 1); p<0.001), LVI (HR 1.83; p=0.008), ER status (HR 0.55; p=0.014) and endocrine therapy (HR 0.51; p=0.004). In a multivariable Cox proportional Hazard model including the prognostic factors mentioned earlier (model 1), the factors age (HR nonlinear; p<0.001) and LVI (HR 1.94; 95%CI: ; p=0.012) were prognostic factors for the prediction of LRR based solely on clinicopathological factors. After adding the 70-gene signature 126 Chapter 8

129 to model 1, the 70-gene signature showed to be an independent prognostic factor for LRR with a multivariable HR of 1.73 (95%CI: ; p=0.042)(table 2; model 2). Other significant prognostic factors in model 2 were age (HR non-linear; p<0.001), LVI (HR 1.87; 95%CI: ; p=0.018) and adjuvant chemotherapy (HR 0.51; 95%CI: ; p=0.042). Adding the 70-gene signature to clinicopathological factors resulted in an increase of the C index from (95%CI: ) in a model with solely clinicopathological factors (model 1) to (95%CI: ) after adding the 70-gene signature in model 2 (Table 2) gene signature predicts locoregional recurrence 127

130 Table 2. Risk of locoregional recurrence within 10 years of breast cancer diagnosis according to the 70-gene signature and routine clinicopathological factors Fine and Gray competing risk regression analysis. Parameter Univariable analysis Multivariable Model 1 Traditional predictors Multivariable Model 2 Traditional + 70-gene signature 10-yr risk HR (95% CI) P HR (95% CI) P HR (95% CI) P (N events) 1 70-gene signature Low risk 6.1% (29) High risk 12.6% (58) 2.40 ( ) < ( ) Age at surgery (years) 2 Continuous Non-linear <0.001 Non-linear <0.001 Non-linear Surgical procedure Breast conserving 9.3% (39) Mastectomy 9.0% (48) 1.07 ( ) ( ) ( ) 0.31 Tumor size (per 5 mm) Continuous 1.04 ( ) ( ) ( ) 0.56 Tumor grade Grade 1 5.2% (13) Grade 2 8.8% (33) 1.76 ( ) ( ) ( ) 0.32 Grade % (40) 2.91 ( ) < ( ) ( ) 0.25 Lymphovascular invasion Absent 7.3% (47) Present 12.8% (40) 1.83 ( ) ( ) ( ) Axillary status (per positive node) Continuous 1.04 ( ) ( ) ( ) 0.26 Estrogen-receptor status Negative 13.3% (24) Positive 8.2% (63) 0.55 ( ) ( ) ( ) 0.89 Adjuvant chemotherapy No 8.9% (61) Yes 9.8% (26) 1.09 ( ) ( ) ( ) Adjuvant endocrine therapy No 11.6% (61) Yes 6.1% (26) 0.51 ( ) ( ) ( ) 0.12 Local radiotherapy No 10.5% (26) Yes 8.6% (61) 0.76 ( ) ( ) ( ) 0.31 Radiotherapy boost No 9.4% (66) Yes 8.5% (21) 0.83 ( ) ( ) ( ) 0.60 C-index ( ) ( ) 1 Events may not add-up due to averaging over multiple imputation datasets; 2 Modelled with linear tail-restricted cubic spline function with 4 degrees of freedom (see Supplementary Figure 1) 128 Chapter 8

131 Discussion The 70-gene signature is an independent prognostic factor in the prediction of LRR in breast cancer after adequate primary treatment with BCT or mastectomy. A significantly higher risk of LRR after 10 years is seen in patients with a high risk 70-gene signature result (12.6%) compared to patients with a low risk 70-gene signature (6.1%; p<0.001). In a multivariable competing risk model including known prognostic factors and treatment, the 70-gene signature is an independent significant predictor of LRR (HR 1.73; 95%CI: ; p=0.042). Other prognostic factors include age, LVI and adjuvant chemotherapy. Adding the 70-gene signature to clinicopathological factors improved the C index of this model, indicating the additional value of the 70-gene signature as a prognostic factor for LRR. The results of our study show potential clinical value for the 70-gene signature in the prediction of LRR. For instance, it is arguable whether there is added value of whole breast irradiation for 70-gene signature low risk patients treated with breast conserving surgery. A recent study already showed that whole breast irradiation after BCT might be of limited value in specific subgroups of patients. 20 In our study the LRR risk for 70-gene signature low risk patients treated with BCT was as low as 5.8% after 10 year after receiving whole breast irradiation. Even though their risk of LRR is lowered by half due to radiotherapy 21, the annual estimated risk of recurrence in this group would still be around 1%, which is a generally accepted risk of LRR. 22 Especially for those patients with ER-positive disease, who will also receive adjuvant endocrine therapy, the risk of LRR will be low. 23,24 Therefore, the 70-gene signature may aid in the identification of patients for whom radiotherapy can be safely omitted after BCT. On the other hand, the question rises whether 70-gene signature high risk patients who were treated with mastectomy without receiving radiotherapy should have received radiotherapy to reduce their risk of LRR which is 13.8% at 10 years in this cohort. This suggests that the 70-gene signature may also be helpful to identify patients at a high risk of recurrence who are eligible for radiotherapy after mastectomy. Validation of this finding is planned to evaluate whether these suggested clinical implications are supported in a larger, independent cohort. This study was conducted in the largest known patient cohort with long-term follow-up for whom gene-expression data as well as data on LRR was collected. There are two other studies reporting on the use of a gene-expression classifier to predict the risk of LRR. Mamounas et al. describe 73 LRR events among 895 patients for whom a 21-gene recurrence score was available from the NSABP B-14 and B-20 trials 25. Their results show a 10-year LRR risk of 15.8% for patients with a high risk recurrence score after treatment with tamoxifen and 4.3% among patients with a low risk recurrence score. Another study by Solin et al. describes only 30 LRR events among 388 patients for whom a 21-gene recurrence score was available from the Eastern Cooperative Oncology Group E2197 study. 26 This study also reports 10-year LRR rates in the 21-gene recurrence score high risk group of 8.7% and 3.7% in the low risk group. In the NSABP study all patients included were treated with tamoxifen and a large proportion of these patients (n=227) were also 8 70-gene signature predicts locoregional recurrence 129

132 used as a training set to develop the 21-gene recurrence score, which increases the probability of overfitting. 27 Both studies included only patients who were treated with BCT. The novelty in our study is the additional subgroup analyses of patients treated with mastectomy with and without receiving radiotherapy, showing additional value of the 70-gene signature especially in BCT treated patients and patients who received radiotherapy after mastectomy. Missing data in our database was imputed, which makes the analyses more reliable. Still, imputing missing data creates a margin of uncertainty. The studies that were included in this pooled dataset were conducted between 1984 and During these 22 years the management of breast cancer changed. Not only adjuvant systemic treatment options improved, but also population-based screening programs were introduced, leading to an increasing incidence of tumors with favorable prognostic factors. 28 The definitions for high and low risk used by clinical guidelines to guide adjuvant systemic treatment decisions were adjusted accordingly. It is not possible to adjust for all these time-related factors. In conclusion, the 70-gene signature is able to predict the risk of LRR, independent of known clinicopathological factors including the involvement of axillary lymph nodes. With this study the first step is taken in the search for a gene-expression profile that is of added value to traditional clinicopathological factors and can help select those patients who will have benefit of limited locoregional treatment and those who will have benefit of a more extensive locoregional treatment. Contributors CD, ER, and LvtV were responsible for the study design. No financial support was needed to perform this study. SE, CD and LvtV collected follow-up data on all validation studies collected in the 70-gene signature database used for this study. CD and MN collected data on surgical treatment, radiotherapy and resection margins. SE performed data analyses. CD, MN, SE, ER, PB, LvtV, HB, PE, and BF took part in data interpretation. CD, MN, SE, and ER took part in manuscript writing. All authors were involved in reviewing the report. Conflict of interest LvtV is named inventor on the patent for the 70-gene signature used in this study. LvtV reports being shareholder in and employed by Agendia NV, the commercial company that markets the 70-gene signature as MammaPrint. LvtV was supported by the Dutch Genomics Initiative Cancer Genomics Centre. HB is a non-remunerated, non-stake holding member of the supervisory board of Agendia NV. 130 Chapter 8

133 Acknowledgements We acknowledge the enormous efforts of M. van de Vijver, M. Buyse, P. Bedard, J. Bueno-de- Mesquita, S. Mook, M. Kok, M. Saghatchian, and colleagues to perform the various 70-gene signature studies included in this study. We thank F. de Snoo for her input in this collaboration and N. Russell for her input on the interpretation of the preliminary data. We especially thank the data-managers at the Netherlands Cancer Institute for all their efforts in collection of the follow-up data gene signature predicts locoregional recurrence 131

134 References 1 Clemons M, Danson S, Hamilton T, Goss P. Locoregionally recurrent breast cancer: incidence, risk factors and survival. Cancer Treat Rev 2001; 27: van der Heiden-van der Loo, Ho VK, Damhuis RA, Siesling S, Menke MB, Peeters PH et al. [Percentage of local recurrence following treatment for breast cancer is not a suitable performance indicator]. Ned Tijdschr Geneeskd 2010; 154:A van t Veer LJ, Paik S, Hayes DF. Gene expression profiling of breast cancer: a new tumor marker. J Clin Oncol 2005; 23: van t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002; 415: Drukker CA, Bueno-de-Mesquita JM, Retel VP, van Harten WH, van Tinteren H, Wesseling J et al. A prospective evaluation of a breast cancer prognosis signature in the observational RASTER study. Int J Cancer 2013; 133: van de Vijver MJ, He YD, van t Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: Bueno-de-Mesquita JM, Linn SC, Keijzer R, Wesseling J, Nuyten DS, van Krimpen C et al. Validation of 70-gene prognosis signature in node-negative breast cancer. Breast Cancer Res Treat 2009; 117: Buyse M, Loi S, van t Veer L, Viale G, Delorenzi M, Glas AM et al. Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer. J Natl Cancer Inst 2006; 98: Wapnir IL, Anderson SJ, Mamounas EP, Geyer CE, Jr., Jeong JH, Tan-Chiu E et al. Prognosis after ipsilateral breast tumor recurrence and locoregional recurrences in five National Surgical Adjuvant Breast and Bowel Project node-positive adjuvant breast cancer trials. J Clin Oncol 2006; 24: Bueno-de-Mesquita JM, van Harten WH, Retel VP, van t Veer LJ, van Dam FS, Karsenberg K et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007; 8: Knauer M, Cardoso F, Wesseling J, Bedard PL, Linn SC, Rutgers EJ et al. Identification of a low-risk subgroup of HER-2-positive breast cancer by the 70-gene prognosis signature. Br J Cancer 2010; 103: Mook S, Schmidt MK, Viale G, Pruneri G, Eekhout I, Floore A et al. The 70-gene prognosis-signature predicts disease outcome in breast cancer patients with 1-3 positive lymph nodes in an independent validation study. Breast Cancer Res Treat 2009; 116: Mook S, Knauer M, Bueno-de-Mesquita JM, Retel VP, Wesseling J, Linn SC et al. Metastatic potential of T1 breast cancer can be predicted by the 70-gene MammaPrint signature. Ann Surg Oncol 2010; 17: Mook S, Schmidt MK, Weigelt B, Kreike B, Eekhout I, van de Vijver MJ et al. The 70-gene prognosis signature predicts early metastasis in breast cancer patients between 55 and 70 years of age. Ann Oncol 2010; 21: Glas AM, Floore A, Delahaye LJ, Witteveen AT, Pover RC, Bakx N et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics 2006; 7: Wolbers M, Koller MT, Witteman JC, Steyerberg EW. Prognostic models with competing risks: methods and application to coronary risk prediction. Epidemiology 2009; 20: Aalen O. Nonparametric estimation of partial transition probabilities in multiple decrement models. Annals of Statistics 1978; Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Annals of Statistics 1988; Chapter 8

135 19 Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. JASA 1999; 94: Gatzemeier W, Andreoli C, Costa A, Gentilini MA, Tinterri C, Zanini V et al. Multi-centre randomised prospective trial on breast conservative surgery (BCS) with or without whole breast irradiation (WBI) in postmenopausal women aged years and low in-breast-recurrence risk: analysis after 9 years median follow-up - RT Study Group. [Abstract no ECCO 2013] Ref Type: Generic 21 Darby S, McGale P, Correa C, Taylor C, Arriagada R, Clarke M et al. Effect of radiotherapy after breast-conserving surgery on 10-year recurrence and 15-year breast cancer death: meta-analysis of individual patient data for 10,801 women in 17 randomised trials. Lancet 2011; 378: Rutgers EJ. Quality control in the locoregional treatment of breast cancer. Eur J Cancer 2001; 37: Fyles AW, McCready DR, Manchul LA, Trudeau ME, Merante P, Pintilie M et al. Tamoxifen with or without breast irradiation in women 50 years of age or older with early breast cancer. N Engl J Med 2004; 351: Hughes KS, Schnaper LA, Bellon JR, Cirrincione CT, Berry DA, McCormick B et al. Lumpectomy plus tamoxifen with or without irradiation in women age 70 years or older with early breast cancer: longterm follow-up of CALGB J Clin Oncol 2013; 31: Mamounas EP, Tang G, Fisher B, Paik S, Shak S, Costantino JP et al. Association between the 21- gene recurrence score assay and risk of locoregional recurrence in node-negative, estrogen receptorpositive breast cancer: results from NSABP B-14 and NSABP B-20. J Clin Oncol 2010; 28: Solin LJ, Gray R, Goldstein LJ, Recht A, Baehner FL, Shak S et al. Prognostic value of biologic subtype and the 21-gene recurrence score relative to local recurrence after breast conservation treatment with radiation for early stage breast carcinoma: results from the Eastern Cooperative Oncology Group E2197 study. Breast Cancer Res Treat 2012; 134: Azim HA, Jr., Michiels S, Zagouri F, Delaloge S, Filipits M, Namer M et al. Utility of prognostic genomic tests in breast cancer practice: The IMPAKT 2012 Working Group Consensus Statement. Ann Oncol 2013; 24: Mook S, van t Veer LJ, Rutgers EJ, Ravdin PM, van de Velde AO, van Leeuwen FE et al. Independent prognostic value of screen detection in invasive breast cancer. J Natl Cancer Inst 2011; 103: van Buuren S., Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equa- tions in R. Journal of Statistical Software 2011; 45: Moons KG, Donders RA, Stijnen T, Harrell FE, Jr. Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol 2006; 59: Marshall A, Altman DG, Holder RL, Royston P. Combining estimates of interest in prognostic modelling studies after multiple imputation: current practice and guidelines. BMC Med Res Methodol 2009; 9: gene signature predicts locoregional recurrence 133

136 Supplementary Figure 1. Relation between age at diagnosis and locoregionaal recurrence risk (linear+linear tail-restricted cubic spline function with 4df) 134 Chapter 8