Gene Expression Profiling in Breast Cancer

Transcription

1 Molecular assays can assist in choosing therapy and predicting outcomes in almost 50% of women with breast cancer. Nick Patten. Martha (detail). Oil on panel, 22" 28". Gene Expression Profiling in Breast Cancer Kiran Turaga, MD, MPH, Geza Acs, MD, PhD, and Christine Laronga, MD, FACS Background: Breast cancer is a heterogeneous group of different tumor subtypes that vary in prognosis and response to therapy. This heterogeneity has spawned an era of molecular assays striving to classify and thus predict outcome, thereby guiding the future in targeted personalized treatment strategies. Methods: This article provides an overview of the development and application of molecular assays as applied to breast cancer. Differences in the technology used for these tests as well as scientific evidence supporting the validity of the gene expression profile are discussed. Examples of the clinical applicability of these assays are provided, but these represent only a fraction of the potential uses yet to be discovered. A comparison of the three most commonly used assays is included. Results: Molecular assays have provided new genetic approaches to unravel the complexities of clinical specimens relevant to breast cancer treatment planning and assessment of outcome. In particular, on a molecular level specific to the woman s tumor, these assays allow a prediction of outcome (prognosis) in terms of low and high risk for the future development of distant metastatic disease. Additionally, one assay, Oncotype DX (Genomic Health Inc, Redwood City, CA), allows for the prediction of benefit of the addition of chemotherapy to hormone therapy alone. Conclusions: While incorporation of molecular assays into the treatment planning strategy of breast cancer continues to be a work in progress, this approach is evolving quickly due to strong scientific evidence to become standard of practice in the near future. The possibilities of these assays in terms of clinical investigation are limitless, but currently their general applicability is limited to less than half of the population of women presenting with breast cancer. Introduction Breast cancer is slowly being recognized as an umbrella designation of different tumor subtypes that differ in their prognosis and their responses to therapy. 1 Despite the significant heterogeneity, substantive ad vances have From the Comprehensive Breast Program, Department of Women s Oncology, H. Lee Moffitt Cancer Center & Research Institute, Tampa, Florida. Submitted September 29, 2009; accepted March 10, Address correspondence to Christine Laronga, MD, FACS, Comprehensive Breast Program, Department of Women s Oncology, MCC BRPROG, Moffitt Cancer Center, Magnolia Drive, Tampa, FL Christine.Laronga@moffitt.org Dr Laronga is on the speakers bureau for Genomic Health, Inc. The other authors report no significant relationship with the companies/organizations whose products or services may be referenced in this article. been made over the past decade in the care of patients with breast cancer. For instance, targeted therapies such as hormone therapy for estrogen receptor (ER) expression and the use of trastuzumab for inhibition of HER2 signaling have contributed significantly to these advances. It has become clear that the traditionally used one size fits all approach has important limitations. These include (1) drugs may be administered at suboptimal doses to drug-resistant patients, (2) tumors highly responsive to a medication may receive additional unnecessary treatments, and (3) accumulation of treatments may not be cost effective if we could identify the most responsive subset of patients. 2 Utilizing molecular predictors has been cited as a cost-effective way of directing targeted therapies to vulnerable populations with significant cost savings of $920,000 in a cohort of July 2010, Vol. 17, No. 3 Cancer Control 177

2 260 node-negative, ER-positive breast cancer patients. 3 The heterogeneity of outcome and drug sensitivity continues to drive the discovery of a second generation of molecular predictors. These have been developed using DNA-based arrays or reverse transcriptase-polymerase chain reaction (RT-PCR) and are expected to improve the quality of care for patients with breast cancer. We discuss briefly the current gene expression profiling tests, the validity of these tests, and the future of the second generation of gene expression profiling tests. Background Gene expression is a measure of a gene s activity, which is determined by the number of times it is transcribed into mrna and finally by the protein it encodes. A snapshot of a tissue s global gene activity (or expression) is captured by DNA microarray technology or RT- PCR and is called a transcriptome. Lists of genes associated with prognoses, responses to various treatments or phenotypes, are called gene profiles or gene signatures. The four major test platforms used for detecting gene profiles are immunohistochemistry (IHC), fluorescent in situ hybridization (FISH), quantitative reverse transcriptase polymerase chain reaction (qrt-pcr), and cdna microarray (quantitative cdna detection). While the former two platforms are semiquantitative and well established for detection of ER and HER2 status at low costs, the latter two are quantitative methods that require complex statistical methods to avoid false discovery. These two methodologies provide highly standardized and reproducible outcomes of uncertain prognostic value at this point. In addition, IHC has the advantage of directly measuring protein expression, not just mrna copy numbers, and it provides a visualization of the difference of protein localization and modification, which gene profiling cannot. All platforms can use formalin-fixed, paraffin-embedded tissues except cdna microarray, which preferentially uses freshfrozen samples only. Fresh-frozen tissues are expensive to obtain and store, require advance planning with informed consent at the time of surgery, and are currently not linked to any long-term outcome data, thus limiting their widespread applicability to the population at large. Currently available breast cancer molecular predictors based on the aforementioned platforms are listed in Table Predicting Disease Outcomes Below we discuss the three most commonly used gene profiling tests: Oncotype DX (Genomic Health Inc, Redwood City, CA), MammaPrint (Agendia Inc, Huntington Beach, CA), and HOXB13/IL17BR (H/I) expression ratio (Theros H/I; biotheranostics, San Diego, CA). Oncotype DX Assay Due to limitations aforementioned with fresh-frozen tissues, investigators sought to develop a genomic molecular assay using readily available and inexpensive paraffin-embedded tissue tumor blocks. After mastering the technique of reliably and reproducibly extracting mea- Table 1. Selected Molecular Predictor Tests Based on Platform Platform Assay Properties IHC ProEx Br 4 5-antibody panel/gene assay, overexpression of 2 linked to disease relapse Mammostrat 5 5-antibody panel/gene assay used to classify ER+, LN tumors treated with tamoxifen into high-, medium-, and low-risk groups FISH exagenbc 6 3 genes for ER+ tumors and 3 different genes for ER tumors, high-risk patients linked to high recurrence rates RT-PCR Oncotype DX 7 21-gene assay originally validated in ER+, LN patients to predict recurrence score and benefit of adjuvant chemotherapy. Now also validated in LN+, ER+ breast cancer H/I 8 6-gene assay, originally described as a ratio of 2 genes; prediction of recurrence in both tamoxifen-treated and -untreated patients, although effect in treated patients is challenged Celera metastasis score 9 14-gene assay test predicts 3.5-fold difference in risk between patients at the top and bottom quintiles of disease risk Breast bioclassifier classifier genes, 5 control genes; identifies patients independent of ER status who might benefit from chemotherapy DNA Microarray MammaPrint gene assay used for patients aged < 55 yrs with LN cancer (Amsterdam signature) Rotterdam signature gene assay used for patients with LN cancer independent of hormone status Invasiveness gene signature gene assay used irrespective of node and hormone receptor status Nuvoselect gene assay for chemotherapy response to T-FAC; 200-gene assay for hormone therapy Cytochrome P450 CYP2D gene assay that predicts benefit of aromatase inhibitors to tamoxifen in patients with ER+ disease Gene expression grade index gene assay that correlates with histological grade and classifies grade II patients into low- and high-risk recurrence groups T-FAC = paclitaxel, 5-fluorouracil, doxorubicin, and cyclophosphamide, LN = lymph node, ER = estrogen receptor. 178 Cancer Control July 2010, Vol. 17, No. 3

3 surable RNA from numerous tumor blocks of varying durations of storage, the developers panned the world s literature looking for candidate genes that have been linked to breast cancer and identified 250 candidate genes. A total of 447 archival tumor blocks were then retrieved from three separate clinical trials that reported data on long-term outcomes: the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-20 study (233 patients), the single-site study at Rush Presbyterian-St Luke s Hospital (78 patients), and the single-site study at Providence St Joseph s Hospital (136 patients). These tumor blocks were assayed for all 250 genes, with the result that all 250 genes could be measured reliably and reproducibly. The next step was to link combinations of expression of these genes with the outcomes data in such a way as to be able to separate patients into three cohorts: low-risk, intermediate-risk, and high-risk for the development of distant metastatic disease at 10 years. The combination of 16 cancer-related genes and five reference genes using a continuous variable algorithm allowed for the most informative separation into risk categories. In order to validate that this was the correct combination of genes, tumor blocks were retrieved from two additional prospective clinical trials in breast cancer patients with ER-positive, lymph node-negative disease (NSABP B-14 and Kaiser Permanente), and similar results were found. Of the multiple genes and their pathways, the most influential are genes related to the cell proliferation group (such as Ki-67), the ER expression and related genes (ER group), and the HER2 expression and related genes (HER2 group). Thus, low levels of ER and high levels of proliferation genes and/or HER2 predict a higher risk of recurrence. A recurrence score is generated on a continuum from 0 to 100 and further subdivides patients into low score (< 18), intermediate score (> 18 and < 30), or high score ( 31). 7 The score is then translated into a percent risk for the development of distant metastatic disease at 10 years, with a score of 18 signifying 10% risk and a score of 31 equal to 20% risk. The overall analytical validity of the Oncotype DX assay varies from 67.7% to 98.9%. 1 Four initial studies have assessed the validity of the Oncotype DX. 7,17-19 Studies on the NSABP B-14 cohort from the tamoxifen-treated arm of ER-positive, lymph node-negative patients revealed three risk categories with univariate actuarial recurrence risks of 7% (low), 14% (intermediate), and 31% (high). These are consistent with the overall study s 10-year disease-free survival rate of 85%. A 50-point change in the recurrence Table 2. Patient Reclassification by Gene Expression Testing With Oncotype DX Comparator Risk Group Patients in Risk 10-Year Risk for Group, n (%) Distant Relapse, % Oncotype DX Risk Group Patients in Risk 10-Year Risk for Group, n (%) Distant Relapse, % St Gallen expert criteria Low 38 (72) 0 Low 53 (8) 5 Medium 12 (22) 18 High 3 (6) 43 Low 134 (60) 5 Medium 222 (33) 9 Medium 51 (23) 6 High 37 (17) 21 Low 166 (42) 8 High 393 (59) 18 Medium 86 (22) 18 High 141 (36) National Comprehensive Cancer Network guidelines Low 38 (72) 0 Low 53 (8) 5 Medium 12 (22) 18 High 3 (6) 43 Low 300 (49) 8 High 615 (92) 15 Medium 137 (22) 14 High 178 (29) 30 Adjuvant! Online criteria Low 354 (53) 8 Low 216 (61) 6 Medium to high 138 (39) 13 Low 122 (39) 9 Medium to high 314 (47) 22 Medium to high 192 (61) 31 ANNALS OF INTERNAL MEDICINE. ONLINE by Luigi Marchionni, Renee F. Wilson, Antonio C. Wolff, Spyridon Marinopoulos, Giovanni Parmigiani, Eric B. Bass, and Steven N. Goodman. Copyright 2010 by AMERICAN COLLEGE OF PHYSICIANS - JOURNALS. Reproduced with permission of AMERICAN COL- LEGE OF PHYSICIANS - JOURNALS in the format Journal via Copyright Clearance Center. July 2010, Vol. 17, No. 3 Cancer Control 179

4 score was the single most important prognostic marker in the cohort (hazard ratio = 2.8; 95% confidence interval [CI], ). A study by Glas et al 20 revealed similar 10-year probability of death of 2.8%, 11%, and 16% in the low-, intermediate-, and high-risk groups, respectively, in a cohort of 165 ER-positive tamoxifen-treated patients. The prognostic value persisted after stratification by tumor grade and disease stage. Conversely, a small study revealed no predictive value of the recurrence score in patients who did not receive either tamoxifen or chemotherapy. 19 A recent systematic review reported the comparison of the performance of the Oncotype DX in the presence of stratification of other commonly used risk factors (Table 2). 1 It appears that in the presence of conventional risk predictors for breast cancer, molecular prognostics using the Oncotype DX changed the recurrence rate for half of the 92% of high-risk candidates by the National Comprehensive Cancer Network (NCCN) guidelines (2004) to low risk with a 10-year relapse risk of 7% (95% CI, 4% 11%). Thus, the use of Oncotype DX appears to better predict outcome than conventional predictors. Another facet of the Oncotype DX assay is the prediction of benefit of adjuvant chemotherapy. Several studies, including the original study NSABP B-20, showed that patients with a low-risk recurrence score (50% of the study cohort) had no benefit in terms of distant metastatic disease or survival with the addition of chemotherapy over hormone therapy alone. 21 Criticisms of this assay s utility initially centered on the fact that the validation studies used tamoxifen (not the modern aromatase inhibitors) and antiquated chemotherapy such as cyclophosphamide, methotrexate, and 5-fluorouracil (CMF). Recent studies presented at national and international breast cancer symposia have addressed these concerns but have not yet been published. One study used the tumor blocks from the ATAC trial comparing an aromatase inhibitor to tamoxifen (Arimidex, tamoxifen, alone or in combination) for the treatment of ER-positive breast cancer (patients Table 3. Kaplan-Meier Analysis of Survival Stratified by MammaPrint and Adjuvant! Online MammaPrint Prognosis Group Adjuvant! Online Risk Group Low Risk High Risk Good Patients, n year overall survival 0.88 ( ) 0.89 ( ) (95% CI) Poor Patients, n year overall survival 0.69 ( ) 0.69 ( ) (95% CI) ANNALS OF INTERNAL MEDICINE. ONLINE by Luigi Marchionni, Renee F. Wilson, Antonio C. Wolff, Spyridon Marinopoulos, Giovanni Parmigiani, Eric B. Bass, and Steven N. Goodman. Copyright 2010 by AMERICAN COLLEGE OF PHYSICIANS - JOURNALS. Reproduced with permission of AMERICAN COLLEGE OF PHYSICIANS - JOURNALS in the format Journal via Copyright Clearance Center. were lymph node-positive or -negative), and the Oncotype DX assay performed equally well with the aromatase inhibitor. 22 The ATAC trial was also important in extending the applicability of the assay to lymph nodepositive patients. 23 Although lymph node-positive patients overall do not fare as well as lymph node-positive patients, the Oncotype DX assay can assign a valid recurrence score and demonstrate the benefit of adjuvant chemotherapy in the high-risk category. 22 The data supporting the use of Oncotype DX assay in the lymph node-positive patient are rapidly amassing, but use of this assay in this setting has not yet gained widespread acceptance. MammaPrint Assay Following efforts to identify prognostic markers, MammaPrint was developed to seek gene expression profiles related to breast cancer by using fresh-frozen tissue. MammaPrint s top-down approach used a comprehensive search of marker genes among all candidate genes irrespective of the biology. Thus, a 70-gene signature was derived from more than 25,000 candidate genes on a cdna array. The original data set used for the test included 78 patients less than 55 years of age with invasive tumors < 5 cm in diameter, ER-positive or -negative disease, and the outcome recorded as distant recurrence at 5 years. Patients are classified based on correlation with a good-prognosis gene expression profile; a coefficient of greater than 0.4 is classified as good prognosis. In a validation study of 295 patients with stage I and II breast cancer, those with a good prognosis had dramatically better 5-year (95% vs 61%) and 10-year (85% vs 51%) recurrence-free survival rates. 11 Although the validation study included 61 of the 78 patients originally used for development of the prognostic signature, sophisticated statistical methods were employed including cross-validated classification to predict outcome among patients. Results excluding the 61 patients were presented in the study. An independent validation multicenter study of 302 patients not treated with tamoxifen or chemotherapy revealed that MammaPrint provided reclassification of patients in high-risk groups (hazard ratio for distant recurrence was 4.6 vs 2.1 for the low-risk groups). This demonstrated the power of discrimination among patients conventionally classified as high-risk patients (Table 3). 24 H/I Assay In a top-down approach similar to that used with MammaPrint, two genes were identified among 22,000 genes in 60 ER-positive patients treated with tamoxifen in a study published by Ma et al. 8 Expression of homeobox HOXB13 was positively correlated with recurrence, while interleukin 17B receptor gene (IL17BR) was negatively correlated with recurrence. This test essentially examines the ratio of these two genes. Initial validation studies by different authors revealed a prediction effect of none to modest, but the stage of disease was more advanced in these studies than in the 180 Cancer Control July 2010, Vol. 17, No. 3

5 original sample. 25,26 Subsequently, a large validation study was undertaken including 852 patients with stage I and II breast cancer, using a different method than originally described by the authors. 27 In this analysis, the ratio of H/I was predictive in node-negative, ER-positive disease, with an adjusted hazard ratio of 3.9 irrespective of treatment. Subsequent validation studies 28,29 suggested that HOXB13 was associated with tamoxifen resistance while IL17BR was independently associated with prognosis. Both studies varied from the protocol in the original study in the populations to which these were applied. Comparison of Signatures Fan et al 30 used the data from 295 patients with stage I and II breast cancer who were originally used to develop the 70-gene signature of MammaPrint to determine the recurrence score by Oncotype DX and the H/I test. Both of these were estimated from gene expression data and not from RT-PCR, and the test was expected to favor the MammaPrint test since it was the data set that was used to develop the signature. MammaPrint and the derived recurrence score from the Oncotype DX predicted overall and disease-free survival, but the H/I test did not (hazard ratio = 1). This may have been in part because the measurement of the gene ratio may have been flawed. 31 On comparison of the intermediate- and high-risk groups from the Oncotype DX to the poor prognosis group of the MammaPrint, the agreement was 81% (239 of 295). 1 Correlation was also noted among the ER-positive subset. As noted earlier, Oncotype DX is the only test to have been studied in the prediction of response. Paik et al 21 found that the benefit of chemotherapy for 651 patients with ER-positive, lymph node-negative disease was highest in those with a high recurrence score, ie, relative risk of 0.26 (95% CI, ). Additionally, in the neoadjuvant setting, three studies have investigated the use of Oncotype DX in predicting response to chemotherapy Two of them found that recurrence score (high) correlated with clinical response, while one study showed no such association The different features of the Oncotype DX, MammaPrint, and H/I tests are summarized in Table 4. Continuing Research The use of gene signatures has considerable potential for improving prognostic and therapeutic prediction. While the exact role of these genes is not understood, they may indeed be markers of already detected clinically relevant markers. The relevance of molecular markers would be in the reclassification of patients into risk strata after being classified by conventional methods such as Adjuvant! Online ( or the NCCN risk criteria. In addition, marketed tests do not reproduce the exact results from the gene signature; comparison of the MammaPrint gene signature to the marketed test revealed a 9% misclassification. Hence, information from genetic signature tests needs to be treated with caution and the tests need to be applied appropriately. It is now clear that breast cancer may be considered a molecularly heterogeneous disease, and genetic signatures should probably focus on more homogeneous test groups such as those used by the Oncotype DX test. In addition, treatment-specific signatures need to be developed. To date, the Oncotype DX is the only test that has shown benefit from the addition of chemotherapy to hormone therapy in high-risk groups. Two ongoing trials are focusing on genetic profiling of tumors: the TAILORx (Trial Assigning Individualized Options for Treatment trial) and the MINDACT (Microarray in Node Negative Disease May Avoid Chemotherapy trial). The TAILORx trial stratifies patients into three groups, using the recurrence score on Oncotype DX. Patients in group 1, with a score < 10, receive hormonal treatment only. Patients in group 2, with a score of > 25, receive combination chemotherapy and hormonal Table 4. Comparison of Commonly Used Breast Cancer Signatures Used for Prediction Variable Oncotype DX MammaPrint H/I Test Development of signature From 250 candidate genes in From candidate set of 25,000 genes From candidate set of 22,000 genes 447 patients in 78 patients in 60 patients Nature of signature 21 genes 70 genes 2 genes; now 6 genes Nature of test subjects Stage I and II breast cancer Stage I and II breast cancer ER+ breast cancer patients, ER+ LN treatment with tamoxifen LN < 55 yrs of age Receiving tamoxifen therapy < 5 cm tumors Outcome Distant relapse-free survival Distant metastasis as first Disease-free survival, relapse-free at 10 yrs relapse event at 5 yrs survival at 5 yrs Test results Recurrence score: Correlation coefficient: Ratio of H/I presented after Low < 18 > 0.4: good prognosis normalization of ratio, Intermediate : poor prognosis presented as hazard ratio High 30 Cost of test $3,978 $4,200 $1,400 Sample characteristics Formalin-fixed, paraffin- Fresh-frozen tumor samples or tissues Formalin-fixed, paraffinembedded tissues collected into RNA preservative embedded tissues July 2010, Vol. 17, No. 3 Cancer Control 181

6 therapy. Patients in group 3, with a score ranging from 11 to 25, are randomly assigned to receive either combination chemotherapy and hormonal therapy or hormonal therapy alone. The MINDACT trial is a multicenter trial comparing MammaPrint to Adjuvant! Online in patient selection for chemotherapy in node-negative disease. The data from these trials will determine the power of these signatures in reclassifying patients after conventional risk classification. Conclusions and Future Directions Several concepts under investigation today may change how gene prediction is performed. Recent discoveries of embryonic stem cell identity in poorly differentiated tumors suggest that cancer prediction using embryonic stem cell signatures may be an important direction for further research. 35 At this point, significant information gleaned from molecular tests needs to be tempered with clinical judgment and applied in the appropriate setting. The extent of the benefit offered by molecular assays, the role of these tests in clinical decision-making, and their impact on cost effectiveness must be rigorously evaluated prior to widespread use. References 1. Marchionni L, Wilson RF, Wolff AC, et al. Systematic review: gene expression profiling assays in early-stage breast cancer. Ann Intern Med. 2008;148(5): Desmedt C, Ruíz-García E, André F. Gene expression predictors in breast cancer: current status, limitations and perspectives. Eur J Cancer. 2008;44(18): Liang H, Brufsky AM, Lembersky BB, et al. A retrospective analysis of the impact of oncotype DX low recurrence score results on treatment decisions in a single academic breast cancer center. Presented at the 30th Annual San Antonio Breast Cancer Symposium; San Antonio, TX; December 13-16, Abstract Whitehead CM, Nelson R, Hudson P, et al. Selection and optimization of a panel of early stage breast cancer prognostic molecular markers. Mod Pathol. 2004;17:50A. 5. Ring BZ, Seitz RS, Beck R, et al. Novel prognostic immunohistochemical biomarker panel for estrogen receptor-positive breast cancer. J Clin Oncol. 2006;24(19): Davis LM, Harris C, Tang L, et al. Amplification patterns of three genomic regions predict distant recurrence in breast carcinoma. J Mol Diagn. 2007;9(3): Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med. 2004;351(27): Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell. 2004;5(6): Garber K. Genomic medicine. Gene expression tests foretell breast cancerʼs future. Science. 2004;303(5665): Perreard L, Fan C, Quackenbush JF, et al. Classification and risk stratification of invasive breast carcinomas using a real-time quantitative RT- PCR assay. Breast Cancer Res. 2006;8(2):R van de Vijver MJ, He YD, vanʼt Veer LJ, et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med. 2002;347 (25): Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet. 2005;65(9460): Liu R, Wang X, Chen GY, et al. The prognostic role of a gene signature from tumorigenic breast-cancer cells. N Engl J Med. 2007;356(3): Ayers M, Symmans WF, Stec J, et al. Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol. 2004; 22(12): Goetz MP, Knox SK, Suman VJ, et al. The impact of cytochrome P450 2D6 metabolism in women receiving adjuvant tamoxifen. Breast Cancer Res Treat. 2007;101(1): Sotiriou C, Wirapati P, Loi S, et al. Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst. 2006;98(4): Cobleigh MA, Tabesh B, Bitterman P, et al. Tumor gene expression and prognosis in breast cancer patients with 10 or more positive lymph nodes. Clin Cancer Res. 2005;11(24 pt 1): Habel LA, Shak S, Jacobs MK, et al. A population-based study of tumor gene expression and risk of breast cancer death among lymph nodenegative patients. Breast Cancer Res. 2006;8(3):R Esteva FJ, Sahin AA, Cristofanilli M, et al. Prognostic role of a multigene reverse transcriptase-pcr assay in patients with node-negative breast cancer not receiving adjuvant systemic therapy. Clin Cancer Res. 2005; 11(9): Glas AM, Floore A, Delahaye LJ, et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics. 2006;7: Paik S, Tang G, Shak S, et al. Gene expression and benefit of chemo - therapy in women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol. 2006;24(23): Albain KS, Barlow WE, Shak S, et al. Prognostic and predictive value of the 21-gene recurrence score assay in postmenopausal women with nodepositive, oestrogen-receptor-positive breast cancer on chemotherapy: a retrospective analysis of a randomised trial. Lancet Oncol. 2010;11(1): Howell A, Cuzick J, Baum M, et al. Results of the ATAC (Arimidex, Tamoxifen, Alone or in Combination) trial after completion of 5 yearsʼ adjuvant treatment for breast cancer. Lancet. 2005;365(9453): Buyse M, Loi S, vanʼt Veer L, 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(17): Reid JF, Lusa L, De Cecco L, et al. Limits of predictive models using microarray data for breast cancer clinical treatment outcome. J Natl Cancer Inst. 2005;97(12): Goetz MP, Suman VJ, Ingle JN, et al. A two-gene expression ratio of homeobox 13 and interleukin-17b receptor for prediction of recurrence and survival in women receiving adjuvant tamoxifen. Clin Cancer Res. 2006; 12(7 Pt 1): Ma XJ, Hilsenbeck SG, Wang W, et al. The HOXB13:IL17BR expression index is a prognostic factor in early-stage breast cancer. J Clin Oncol. 2006;24(28): Jansen MP, Sieuwerts AM, Look MP, et al. HOXB13-to-IL17BR expression ratio is related with tumor aggressiveness and response to tamoxifen of recurrent breast cancer: a retrospective study. J Clin Oncol. 2007;25(6): Jerevall PL, Brommesson S, Strand C, et al. Exploring the two-gene ratio in breast cancer: independent roles for HOXB13 and IL17BR in prediction of clinical outcome. Breast Cancer Res Treat. 2008;107(2): Fan C, Oh DS, Wessels L, et al. Concordance among gene-expression-based predictors for breast cancer. N Engl J Med. 2006;355(6): Goetz MP, Ingle JN, Couch FJ. Gene-expression-based predictors for breast cancer. N Engl J Med. 2007;356(7):752; author reply Mina L, Soule SE, Badve S, et al. Predicting response to primary chemotherapy: gene expression profiling of paraffin-embedded core biopsy tissue. Breast Cancer Res Treat. 2007;103(2): Gianni L, Zambetti M, Clark K, et al. Gene expression profiles in paraffin-embedded core biopsy tissue predict response to chemotherapy in women with locally advanced breast cancer. J Clin Oncol. 2005;23(29): Chang JC, Makris A, Gutierrez MC, et al. Gene expression patterns in formalin-fixed, paraffin-embedded core biopsies predict docetaxel chemosensitivity in breast cancer patients. Breast Cancer Res Treat. 2008;108(2): Ben-Porath I, Thomson MW, Carey VJ, et al. An embryonic stem cell-like gene expression signature in poorly differentiated aggressive human tumors. Nat Genet. 2008;40(5): Cancer Control July 2010, Vol. 17, No. 3