Systematic Review: Gene Expression Profiling Assays in Early-Stage Breast Cancer

Luigi Marchionni; Renee F. Wilson; Antonio C. Wolff; Spyridon Marinopoulos; Giovanni Parmigiani; Eric B. Bass; Steven N. Goodman

doi:10.7326/0003-4819-148-5-200803040-00208

Reviews |4 March 2008

Systematic Review: Gene Expression Profiling Assays in Early-Stage Breast Cancer Free

Luigi Marchionni, MD, PhD; Renee F. Wilson, MSc; Antonio C. Wolff, MD; Spyridon Marinopoulos, MD, MBA; Giovanni Parmigiani, PhD; Eric B. Bass, MD, MPH; Steven N. Goodman, MD, MHS, PhD

Article, Author, and Disclosure Information

Abstract

Background:

Three gene expression–based prognostic breast cancer tests have been licensed for use.

Purpose:

To summarize evidence on the validity and utility of 3 gene expression–based prognostic breast cancer tests: Oncotype DX (Genomic Health, Redwood City, California), MammaPrint (Agendia BV, Amsterdam, the Netherlands), and H/I (AvariaDX, Carlsbad, California).

Data Sources:

MEDLINE, EMBASE, and Cochrane databases (from 1990 through January 2007), Web sites of test manufacturers, and discussion with the manufacturers.

Study Selection:

Original data studies published in English that addressed prognostic accuracy and discrimination or treatment benefit prediction of any of the 3 tests in women with breast cancer.

Data Extraction:

Information was extracted about the clinical characteristics of the study population (particularly clinical and therapeutic homogeneity), tumor characteristics, and whether the marketed test or underlying signature was evaluated.

Data Synthesis:

The tests are based on distinct gene lists, using 2 different technologies. Overall, the body of evidence showed that this new generation of tests may improve prognostic and therapeutic prediction, but the tests are at different stages of development. Evidence shows that the tests offer clinically relevant, improved risk stratification over standard predictors. Oncotype DX has the strongest evidence, closely followed by MammaPrint and H/I (which is still maturing).

Limitations:

For all tests, the relationship of predicted to observed risk in different populations and their incremental contribution over conventional predictors, optimal implementation, and relevance to patients receiving current therapies need further study.

Conclusion:

Gene expression technologies show great promise to improve predictions of prognosis and treatment benefit for women with early-stage breast cancer. More information is needed on the extent of improvement in prediction, characteristics of women in whom the tests should be used, and how best to incorporate test results into decision making about breast cancer treatment.

Currently, 3 commercially available prognostic breast cancer tests based on gene expression (see Glossary) technology are available: Oncotype DX (Genomic Health, Redwood City, California), MammaPrint (Agendia BV, Amsterdam, the Netherlands), and H/I (AvariaDX, Carlsbad, California). Although measurement of gene expression is now a core research method, these commercial assays represent the first introduction of these technologies into clinical application.

Gene expression is the technical term to describe how active a particular gene is—that is, how many times it is expressed, or transcribed, to produce the protein it encodes (Figure 1). The transcription (see Glossary) of the gene's DNA into messenger RNA (mRNA) is the first step in this process; modern molecular biological tools measure this activity by counting the number of mRNA molecules in a given cell type or tissue. Because the mRNA molecule is translated within the ribosome to produce a complete protein, counting mRNA transcripts provides an estimate of the number of corresponding proteins. High-throughput technologies, such as DNA microarray (see Glossary) and real-time reverse transcriptase polymerase chain reaction (RT-PCR) (see Glossary), allow simultaneous counting of many gene transcriptions (up to tens of thousands). This creates a snapshot of a tissue's global gene activity, called the transcriptome.

Figure 1.

Technologies used for high-throughput gene expression analysis.

A. Breast cancer tumors are sampled at the treatment location and shipped to the central laboratory doing the assay, where pathologic review is done to assess cancer cell contents, followed by RNA preparation and integrity evaluation. Suitable samples are used to quantify RNA levels, thus assessing gene expression. When a gene is expressed, the transcription complex copies its DNA sequence into complementary RNA transcripts that are translated into proteins. High-throughput gene expression analysis aims to quantify messenger RNA (mRNA) populations in a given tissue. B. DNA microarray is the molecular biology technique enabling gene expression analysis in MammaPrint. RNA is labeled with fluorescent dye and hybridized against thousands of different nucleotide sequences corresponding to different genes and arrayed on a solid surface (that is, a modified microscope glass slide). On hybridization, fluorescence emitted by single locations on the microarray is used to estimate gene expression levels. In MammaPrint, a 2-color design is used, and RNA expression is estimated as a relative ratio between the sample and a reference RNA. For each patient, triplicate measurements are obtained from 2 microarrays inverting the labeling scheme. C. Real-time reverse transcriptase polymerase chain reaction (PCR) is the enabling technology to assess gene expression in Oncotype DX and H/I. This technique is based on reverse transcription (RT) (see Glossary) of a specific mRNA into the complementary DNA (cDNA) molecule, which is used as a template in PCR. The production of double-stranded DNA is accompanied by emission of light, which is recorded throughout the process and correlates to the amount of DNA that is produced. The higher the initial amount of RNA, the earlier light is emitted during RT-PCR, a measurable difference that allows gene expression to be quantitated. D. Gene expression levels are mathematically transformed into indexes predicting disease recurrence.

Gene expression measurements have been used to develop new biological concepts, refine disease classification, improve diagnostic and prognostic accuracy, and identify new molecular targets for drugs, especially in cancer research (1–9). Results are commonly reported in the form of a list of genes that are differentially expressed between normal and diseased patients or that correlate with different prognoses or phenotypes. These lists are called gene expression profiles or signatures (see Glossary).

“Breast cancer” is increasingly understood as an umbrella designation for various tumor subtypes that differ in their prognoses and responses to therapy. An important decision for many patients with early-stage breast cancer, especially patients who have tumors that express hormone receptors and will be given antiestrogen therapy, is whether they should also be treated with systemic chemotherapy. Although adjuvant chemotherapy is frequently recommended in this setting, many women will remain recurrence-free at 10 years without it, especially those with small, estrogen receptor–positive tumors without axillary nodal involvement. Patients and their physicians must weigh the possible benefit of chemotherapy in reducing recurrence against its toxicity and other attendant costs.

Practicing oncologists frequently base their decisions about therapy on prognostic clinical algorithms that include demographic data; tumor stage; and other tumor characteristics, such as grade and estrogen receptor expression. These conventional combination predictors include the National Institutes of Health (NIH) Consensus Development criteria (10, 11); the St. Gallen expert opinion criteria (12, 13); the National Comprehensive Cancer Network guideline (14–16); and a Web-based algorithm, Adjuvant! Online (17, 18). Gene expression profiling has been proposed to potentially augment or replace these prognostic tools.

Oncotype DX is based on a 21-gene profile developed by Paik and colleagues (19), MammaPrint is based on a 70-gene prognostic signature developed by van't Veer and colleagues (8), and H/I is based on a 2-gene signature (HOXB13–IL17BR) developed by Ma and colleagues (20). The gene sets on which these tests are based have minimal overlap. The 21-gene and the 70-gene expression signatures that form the basis of Oncotype DX and MammaPrints, respectively, share only 1 gene in common. Two technologies are used to determine gene expression: real-time RT-PCR (Oncotype DX and H/I) and DNA microarray (MammaPrint). All 3 tests use pathologic review of specimens to check tumor content and evaluate RNA preparation and quality. The 2 RT-PCR–based assays (Oncotype DX and H/I) are done in formalin-fixed, paraffin-embedded tumor tissues, whereas fresh unfixed tumor tissue is required for MammaPrint. We review evidence on the prognostic accuracy of these 3 tests and their ability to predict treatment benefit.

Methods

The Agency for Healthcare Research and Quality commissioned the review for the Centers of Disease Control and Prevention's Evaluation of Genomic Applications in Practice and Prevention program. Additional details about the methods and results are found in a comprehensive evidence report that is available through the Agency for Healthcare Research and Quality (www.ahrq.gov/clinic/epcindex.htm).

Data Sources

On 9 January 2007, we searched the MEDLINE and EMBASE databases by using Medical Subject Headings and other terms relevant to breast cancer, gene expression profiling, and Oncotype DX or MammaPrint. On 7 February 2007, we searched the Cochrane database, including Cochrane Reviews, CENTRAL, and CINAHL. We supplemented this search by updating searches in MEDLINE and by hand-searching added publications that appeared after the initial search (January 2007 to July 2007) and studies related to H/I. The test manufacturers were also asked to provide any published or unpublished data relating to our study questions. Searches were limited to publications in English.

Study Selection

Two investigators independently reviewed titles and abstracts to identify original data studies that involved the use of any of the 3 assays in women with breast cancer.

Data Extraction

We extracted and double-checked information on the clinical characteristics of the study population, tumor characteristics, and whether the marketed test or underlying signature was evaluated. To assess the quality of studies, we applied (where appropriate) the general principles of the REMARK (REporting recommendations for tumor MARKer prognostic studies) (21, 22) and Standards for Reporting of Diagnostic Accuracy (23, 24) guidelines.

We synthesized data on the ability of a test to accurately predict recurrence risk (clinical validity) and treatment benefit (clinical utility). We distinguished between gene expression signatures and the gene expression–based marketed tests. The gene signature is the collection of genes whose expression levels are measured in a given test. A gene signature can be measured by using various technologies (RT-PCR or complementary DNA [cDNA] array) and procedures (for example, different reagents, controls, sample acquisition, preparation, and transport procedures), which may not be identical to those used in the marketed test.

We report limited information on the technical performance characteristics of the tests, sometimes called analytic validity. The analytic validity of a test usually is assessed by determining how observed measurements differ from standard reference values. However, no reference standard exists for gene expression measurements outside of the technologies used for these tests. Because analytic validity affects predictive ability, our assessment of predictive ability incorporates the effect of less-than-perfect analytic validity. In the full evidence report (available at www.ahrq.gov/clinic/epcindex.htm), we summarize data on the reproducibility of a test when applied repeatedly to the same patient or when repeated over time, as well as variability as a function of tumor sampling and handling, specimen preparation, and biological variation within tumor samples. Appendix Tables 1, 2, and 3 show the evidence summary.

Appendix Table 1. Studies on the Oncotype DX Gene Expression Test

Appendix Table 2. Studies on the MammaPrint Gene Expression Test

Appendix Table 3. Studies on the H/I Gene Expression Test

Role of the Funding Source

The Agency for Healthcare Research and Quality and the Centers of Disease Control and Prevention's Evaluation of Genomic Applications in Practice and Prevention program helped formulate the initial study questions but did not participate in the literature search, determination of study eligibility criteria, data analysis, or interpretation.

Results

Figure 2 shows the number of studies considered at each phase of title, abstract, and article review. The final set of 26 studies was heterogeneous in focus and quality. Few reports addressed technical aspects of the tests. Ten reports focused on prognostic prediction. Only 1 study, involving Oncotype DX, examined the prediction of treatment benefit. Most of the published evidence available for Oncotype DX was conducted with the marketed assay. The evidence relevant to MammaPrint was a mix of studies of the underlying signature and of the marketed test. Only 1 study used the marketed version of H/I (41), and it was not clear whether the laboratory doing the assay was the same as the one with current rights to do the test. All other studies relevant to H/I examined the underlying 2-gene signature, using somewhat different measurement techniques and algorithms than those implemented in the marketed test.

Figure 2.

Systematic search strategy and results.

Analytic Validity and Variability

We found limited evidence about the laboratory procedures used for Oncotype DX (26, 27) and MammaPrint (34, 35), including information about their reproducibility. Such evidence was reported in methodological studies (26–35) and in clinical studies that focused on predictive validity (19, 30). In the case of MammaPrint, results from different laboratories depended on RNA labeling protocols (34), suggesting that MammaPrint results may not be identical if done in different laboratories (only 1 laboratory currently offers the test). The overall proportion of samples that were successfully tested with the various methods ranged from 67.7% to 98.9% for Oncotype DX and 67.7% to 80.9% for MammaPrint (36) (Appendix Tables 1 and 2). No reports investigated the reproducibility of H/I; Ma and colleagues (41) reported a success rate of 98%.

Predicting Disease Outcomes

Oncotype DX

Oncotype DX was developed on the basis of a prospectively chosen 250-candidate gene set, which was measured on 447 patients with breast cancer who were treated in 1 of 3 completed randomized trials with long-term follow-up. From these 250 genes, 21 genes (16 cancer-related and 5 references) were chosen to predict 10-year breast cancer recurrence. The expression levels of these genes are measured by using RT-PCR combined with a published quantitative algorithm to produce a number between 0 and 100, which is the recurrence score. In this review, “recurrence score” indicates the numeric value generated from Oncotype DX (19). The recurrence score is categorized into 3 risk strata: low (score <18), intermediate (score >18 but <30), or high (score ≥30).

Four studies assessed the clinical validity of Oncotype DX (19, 25, 30, 31). Paik and colleagues (19) studied 668 women in a randomized, controlled trial conducted by the National Surgical Adjuvant Breast and Bowel Project (NSABP) (Appendix Table 1). The parent study (the NSABP B-14 trial), which enrolled patients from 1982 to 1988, examined the effect of tamoxifen therapy versus placebo in women with lymph node–negative, estrogen receptor–positive, early-stage breast cancer (42–45). The recurrence score algorithm and risk categories for the group treated with tamoxifen were prespecified on the basis of the development studies. Stratification into the 3 risk categories yielded univariate actuarial 10-year recurrence risks of 7% (low), 14% (intermediate), and 31% (high) (P < 0.001) (Appendix Table 1). The recurrence score was the strongest predictor among all traditional risk factors, with an adjusted hazard ratio (HR) of 2.8 (CI, 1.7 to 4.6) for a 50-point change in the recurrence score.

Glas and colleagues (35) examined the clinical performance of the Oncotype DX assay to predict breast cancer death at 10 years in a community-based population of lymph node–negative, estrogen receptor–positive patients treated with tamoxifen. Two hundred twenty case patients (dead) and 570 matched control patients (alive) were selected, and 165 estrogen receptor–positive case patients and 55 case patients who received tamoxifen treatment (among estrogen receptor–positive patients) formed the final study sample. For patients treated with tamoxifen, results paralleled those of the NSABP B-14 trial (which examined recurrence): The probability of death at 10 years was 2.8%, 11%, and 16% in the low-, medium-, and high-risk groups, respectively. These rates were about 3 percentage points lower than those among patients not treated with tamoxifen. Prognostic value persisted after stratification by tumor grade and disease stage. The continuous recurrence score also showed a relationship with mortality risk in 52 estrogen receptor–negative patients after adjustment for tumor grade and disease stage (relative risk [RR], 1.4 per 10-unit increase in recurrence score [CI, 1.04 to 2.0]). Another study (36) showed no predictive value of the recurrence score in a small population of patients who received neither tamoxifen nor chemotherapy. However, worse tumor grade predicted better prognosis in that study, suggesting that the results were not reliable. A study on the signature alone, in which the main purpose was to contrast the different tests (33), is described in the “Comparison of Signatures” section.

No published study showed how the recurrence score reclassified patients into different risk strata after initial classification by conventional predictors. However, in 2004, Paik and colleagues (46) presented such information in poster form. They reported that the recurrence score had predictive power beyond that of the St. Gallen or National Comprehensive Cancer Network risk stratification guidelines, sufficient to change some patient decisions about chemotherapy. (The St. Gallen test did not include human epidermal growth factor receptor 2 at that time, which is included in the recurrence score). On the basis of the 2004 National Comprehensive Cancer Network guidelines, the study indicated that about half of the 92% of patients who were in the high-risk National Comprehensive Cancer Network category were reclassified as low-risk by the recurrence score, with a 10-year relapse risk of 7% (CI, 4% to 11%); this is similar to the risk seen in the low-risk recurrence score group without the National Comprehensive Cancer Network information. The same information for Adjuvant! Online was part of an oral presentation in 2005 (Table 1) (47). Compared with the Adjuvant! Online criteria, roughly 40% of women assessed to be at high risk (22% relapse) were reclassified into a group with an observed 8% risk if they had a low recurrence score. With Adjuvant! Online, 39% of women classified as high risk (31% recurrence) had a 9% recurrence risk after a low recurrence score. These findings showed that the greatest contribution of the test is probably the reclassification of patients from high to low risk (that is, reducing the number of patients who might unnecessarily undergo chemotherapy) and that combining this test with conventional predictors yields the most information.

Table 1. Patient Reclassification by Gene Expression Testing with Oncotype DX

MammaPrint

MammaPrint is based on the 70-gene signature derived from an initially unselected set of more than 25 000 candidate genes on a cDNA array. The test was developed in 2002 at the Netherlands Cancer Institute by using 78 lymph node–negative patients younger than age 55 years who did not carry a breast cancer gene mutation and who had tumors that were less than 5 cm in diameter (8). The end point for this training set was 5-year distant recurrence. Patients are classified by calculating the correlation coefficient between a patient's expression levels of the 70 genes and an average good-prognosis expression profile. If the correlation coefficient exceeds 0.4, the patient is classified as having a good prognosis; if less, they are classified as having a poor prognosis.

van de Vijver and coworkers (9) validated this signature in a series of 295 consecutive patients with stage I or II breast cancer and small tumors (<5 cm) who were younger than age 53 years. The population was mixed in terms of lymph node positivity, estrogen receptor status, and receipt of chemotherapy and tamoxifen. Sixty-one of the 295 patients in the validation study were also used to develop the signature. Patients with a good prognosis had dramatically better 5-year (95% vs. 61%) and 10-year (85% vs. 51%) recurrence-free survival and overall survival (95% vs. 55% at 10 years) than patients with a poor prognosis. Multivariable analysis showed that prognosis group, tumor size, and adjuvant chemotherapy were the strongest predictors of distant metastases, and patients with the poor-prognosis signature had the largest HR (4.6 [CI, 2.3 to 9.2]). Results of analyses excluding the 61 patients from the training cohort were similar. Fifteen percent of patients with the good-prognosis signature had recurrence by 10 years, demonstrating that when the 70-gene signature is used alone in this mixed population, the long-term risk in the good-prognosis group may not be low enough to justify forgoing chemotherapy.

Kaplan–Meier analyses showed the absolute risks associated with various predictors and combinations of predictors. Overall, the 70-gene signature placed 40% of the cohort (60 of 151) into the good-prognosis group, with a 10-year recurrence rate of about 15% (imputed from figures). The St. Gallen index placed only 15% (22 of 151) of the cohort into a low-risk group, with an estimated 10-year recurrence rate slightly greater than 20%, and the NIH criteria placed only 7% (11 of 151 patients) at low risk, with a long-term risk (based on small numbers) of slightly less than 20%. The 70-gene signature reclassified 33% (43 of 129) of St. Gallen high-risk patients and 38% (53 of 140) of NIH criteria high-risk patients into a lower-risk group, with a 10-year recurrence risk just a few percentage points greater than that of the 70-gene good-signature group without stratification.

Glas and colleagues (35) reanalyzed 145 patients from van de Vijver and coworkers' (9) cohort and all 78 patients from the training set by using the marketed MammaPrint assay instead of the signature (35). A different reference RNA and different quantification method were used. Although odds ratios (ORs) and HRs were similar to those found in the earlier studies, approximately 9% (7 of 78) of patients were placed into different risk categories, most of which had borderline correlations.

The MammaPrint test was validated in a multicenter European study of 302 patients not treated with chemotherapy or tamoxifen, and it provided prognostic information beyond that of Adjuvant! Online (36). Frozen tumor specimens from node-negative patients younger than age 60 years who did not receive systemic adjuvant chemotherapy were tested. Ninety patients were estrogen receptor–negative, and none of the estrogen receptor–positive patients received tamoxifen. The median follow-up was 13.6 years, and the overall rate of distant metastasis was 25%. The area under the receiver-operating characteristic curves indicated that both methods had similarly modest discriminatory power in absolute terms (0.68 for MammaPrint and 0.66 for Adjuvant! Online), but MammaPrint provided better reclassification of patients in risk groups (Table 2). Hazard ratio estimates between high- and low-risk categories for distant recurrence in van de Vijver and colleagues' study were substantially higher than those in this validation study (unadjusted HR >15 vs. 2.3, respectively; adjusted HR, 4.6 vs. 2.1). Compared with van de Vijver and colleagues' (9) study, this validation cohort was observed for a longer period (median, 13.6 vs. 6.7 years), included older women, and excluded patients who received adjuvant therapy. This study also found that HRs for all end points decreased steadily with an artificial increase in censoring time from 5 to 10 years.

Table 2. Kaplan–Meier Analysis of Survival Stratified by MammaPrint and Adjuvant! Online

The H/I Test

Ma and colleagues (20) identified the 2 genes that are the basis for H/I by screening 22 000 genes in 60 patients with estrogen receptor–positive, lymph node–positive or negative breast cancer treated with tamoxifen. High expression of HOXB13 predicted recurrence, and high expression of IL17BR predicted nonrecurrence; therefore, a higher ratio of the 2 genes strongly predicted recurrence in this training set (interquartile OR, 10.2; adjusted OR, 10.2).

Reid and colleagues (37) examined 58 tamoxifen-treated patients with estrogen receptor–positive breast cancer whose disease was more advanced than in Ma and colleagues' sample (48). No relationship between the expression of these genes and distant relapse was observed in these patients or in an additional 99 patients derived from a previously studied cohort (5) that the authors investigated after the initial negative result. In 2006, Goetz and colleagues (38) analyzed 206 estrogen receptor–positive patients treated in the tamoxifen-only arm of a phase III randomized trial. Expression values were normalized by using a different approach than that used by Ma and colleagues (20), and different cutoff points were calculated for the ratio that best predicted relapse-free survival, disease-free survival, and overall survival. The ratio had modest predictive strength in the entire cohort, with cross-validated HRs near 1.5 and P values around 0.05, and the predictive ability was restricted to node-negative patients.

In a large validation study, Ma and colleagues (41) examined a consecutive series of 852 patients with stage I or II breast cancer with a median follow-up of 6.8 years. The investigators used a slightly different method from the one they previously used (20) to combine and normalize the expression of the 2 genes into an index that is now the basis of the H/I assay. In a stratified analysis, the HOXB13–IL17BR ratio was predictive only in patients with node–negative, estrogen receptor–positive disease. The investigators optimized the threshold (maximizing the HR) differently in patients treated with tamoxifen and those who were not. The adjusted HR incorporating other risk factors was 3.9, regardless of tamoxifen treatment. Classification probabilities were not presented, and the incremental value of the HOXB13–IL17BR ratio compared with conventional combined predictors was not reported, although some components of those predictors were included in the multivariable analyses.

Jansen and associates (39) evaluated the ability of the HOXB13–IL17BR ratio to predict disease-free survival in 1252 patients with breast cancer who had undergone various surgical treatments. In this group, 73% of tumors were estrogen receptor–positive; 52% of patients were lymph node–positive, 14% were treated with tamoxifen, 17% received chemotherapy, and 55% received tamoxifen or chemotherapy after relapse. Jansen and associates (39) used different populations, protocols, normalization strategy, and ratio thresholds than Ma and colleagues (41). The overall relapse rate was high at 51% after a median of 6 years follow-up. The HOXB13–IL17BR ratio was examined in 468 patients with lymph node–negative, estrogen receptor–positive disease who did not receive adjuvant systemic chemotherapy. The ratio was associated with poor disease-free survival in a multivariable model (HR, 1.6; P = 0.02) and poor overall survival (HR not reported; P < 0.001). Prognostic value was also shown for untreated patients with estrogen receptor–positive, lymph node–positive tumors and for progression-free survival in patients with relapse who received first-line tamoxifen monotherapy (Appendix Table 3). The ratio was not compared with conventional combination risk indices, and classification probabilities for the models with and without the ratio were not provided.

Jerevall and coworkers (40) investigated whether the HOXB13–IL17BR ratio predicted a differential benefit between 264 patients with postmenopausal breast cancer who received tamoxifen for 2 and 5 years and 93 premenopausal patients who did not receive systemic therapy. Seventy-two percent of patients had lymph node–positive disease and 74% had estrogen receptor–positive disease (Appendix Table 3). The authors dichotomized the HOXB13–IL17BR ratio at the median, which differed from the approach used by Ma and colleagues (41). Jerevall and coworkers (40) concluded that IL17BR might be an independent prognostic factor in breast cancer and suggested that HOXB13 may be correlated with tamoxifen resistance. However, the HOXB13–IL17BR ratio had no prognostic value in postmenopausal patients with estrogen receptor–negative disease. Neither the patient profile nor the methods of calculation of the ratio were identical to those used in previous studies, and the results differed from previous reports because the HOXB13–IL17BR ratio predicted worse outcomes in patients with lymph node–positive disease.

Comparison of Signatures

Fan and colleagues (33) used the same data set to evaluate both the agreement between gene expression tests and other predictors and the individual performance of the tests. The Oncotype DX recurrence score and the HOXB13–IL17BR ratio were estimated from microarray gene expression data (that is, not RT-PCR) and thus were not obtained according to the protocols and methods used in the marketed assays. These data are therefore described as derived scores. Fan and colleagues (33) used the same 295 samples from patients with stage I or II breast cancer that had been used to develop the 70-gene signature (9). Therefore, Fan and colleagues' (33) comparison would be expected to favor the 70-gene profile over the derived recurrence score or the HOXB13–IL17BR ratio.

The 70-gene signature and the derived recurrence score predicted overall survival and disease-free survival, but the derived HOXB13–IL17BR ratio did not predict either (HR, about 1); however, measurement of the derived 2-gene ratio may have been flawed (49). The intermediate- and high-risk groups, as defined by the derived recurrence score, were combined and compared with the poor-prognosis group defined by the 70-gene signature. The agreement between MammaPrint and the derived recurrence score was 81% (239 of 295). All analyses were repeated for the 225 patients in the estrogen receptor–positive subset, with qualitatively similar results. Good correlation between predictions was found, which was of interest because classification was obtained by using different gene sets. The degree of prediction over and above standard combined clinical stratifiers was not clear, and a reclassification analysis of patients was not done.

Predicting Treatment Response

The ability of Oncotype DX to predict chemotherapy benefit was investigated in patients from the NSABP B-20 trial (50). In this study, Paik and colleagues (50) examined 10-year, distant recurrence-free survival in 651 patients with estrogen receptor–positive, lymph node–negative disease who were randomly assigned to receive tamoxifen alone or tamoxifen with chemotherapy. An overall benefit was seen from chemotherapy, but when the data were stratified by risk group, the benefit was restricted to patients with a high recurrence score (RR, 0.26 [CI, 0.13 to 0.53]), a finding that persisted in multivariate analyses. However, even though no benefit was seen in the low recurrence-score group, the point estimate had very wide CIs; a clinically relevant benefit could therefore not be excluded.

Two studies examined whether the recurrence score predicted pathologic response in patients receiving preoperative systemic therapy (28, 29). Neither study was done at the laboratory offering Oncotype DX (Appendix Table 1). One study found that the recurrence score predicted complete response (28), whereas the other study (29) found no such relationship. Finally, Chang and colleagues (32) assessed chemotherapy response prediction in 12 patients with complete clinical response among 72 women enrolled in phase II studies of docetaxel and found that a high recurrence score was associated with complete response (P = 0.008). When the recurrence score was used as a continuous variable, a 14-unit increase in the score (the difference between the low- and high-risk groups, as defined by the standard thresholds) were modestly predictive of a clinical complete response (OR, 1.7 [CI, 1.15 to 2.60]).

No study investigated the ability of MammaPrint to predict treatment response. One study reported that the HOXB13–IL17BR ratio could predict whether 5 years of tamoxifen therapy would provide survival benefit over 2 years of tamoxifen treatment in estrogen receptor–positive patients (40).

Discussion

This body of evidence on the 3 marketed gene expression tests for breast cancer prognosis shows that these tests have considerable potential for improving prognostic and therapeutic prediction. It also provides valuable lessons about the complexity of evaluating such tests.

Because the role of the genes included in these tests in determining prognosis is not completely understood, it is often unclear which clinical or tumor characteristics are being measured. Intrinsic tumor aggressiveness, ability to metastasize, and responsiveness to treatment (hormonal, radiation, or chemotherapy)—each of which might involve different genes—can determine prognosis. However, the characteristic being assessed in a particular study must often be inferred from the treatment, tumor, and clinical characteristics of the study population. Results from populations that are clinically and therapeutically heterogeneous may not be optimal in determining the prognosis or risk for a particular woman.

End points also varied. Survival was defined in the studies as disease-free, distant recurrence–free, or overall, measured at 5 or 10 (or more) years. Prediction strength varied considerably depending on what end point the test was optimized for. Finally, performance of the underlying gene signature is not necessarily identical to the marketed test, because many test procedures, including pretest sample preparation and transport, can differ. It is therefore critical to pay close attention to the test description, population, and end points for each study to understand which studies are mutually supportive, which are adding qualitatively different information, and to whom they apply.

Despite the clinical novelty of these tests, their development must follow the same principles and procedures as those for any multivariate clinical prediction rule. These principles are outlined in detail in the clinical literature (51, 52) and have been agreed on in reporting guidelines (22) and articulated with respect to expression-based predictors in various review articles (53–55).

Each of the 3 marketed tests is at a different point in the developmental pathway. Almost all studies of Oncotype DX have used the marketed test as opposed to the signature, whereas the evidence on MammaPrint comes from studies examining the signature or the assay (only the large multicenter validation by Buyse and colleagues [36] used the marketed assay). The study that compared the results of the marketed MammaPrint test versus its signature on the same samples showed that about 9% of the patients were placed into different risk groups when the marketed test was used. Almost all of the studies based on H/I calculated or implemented the test in subtly different ways, and only 1 seemed to use the marketed H/I test (48).

In terms of the clinical and therapeutic homogeneity of the underlying populations, Oncotype DX focused on a narrower, more clinically and therapeutically homogeneous group than the other tests, which is reflected in its claimed indications: patients with estrogen receptor–positive, lymph node–negative, stage I or II disease who receive tamoxifen. MammaPrint has been tested in heterogeneous populations including a mix of treated and untreated patients, patients with lymph node–negative and –positive disease, and those with estrogen receptor–positive and –negative disease. The claimed indications are therefore broader than those for Oncotype DX. Although the indications for MammaPrint match the populations in whom the test has been evaluated, whether to consider these populations prognostically homogenous is a critical question.

The gene signature underlying H/I has been investigated in large, heterogeneous populations, and differences were found in its prediction ability for specific subgroups. The signature has been variably formulated as a simple ratio or as an index, normalized to different sets of genes or standardized with calibration RNA, and stratified by using thresholds optimized within each study. Whereas plausible mechanisms support the test rationale, the marketed test requires further validation and comparison with conventional combination predictors. The eligibility criteria for H/I are similar to those for Oncotype DX.

The utility of these tests for clinical decision making is a critical question. No study has addressed whether MammaPrint or H/I can predict the clinical benefit of chemotherapy. Oncotype DX is the only gene expression test that can predict such a benefit. Although this claim is based on past data, the study design was strong and was based on a randomized, clinical trial of tamoxifen plus chemotherapy versus tamoxifen alone (50). This study suggested that chemotherapy had no benefit in women in the low-risk group, but wide CIs around the observed zero effect did not rule out a meaningful effect.

Even with few data on prediction of treatment benefit, the risk for long-term recurrence or death serves as an effective ceiling on the degree of chemotherapy benefit. If that risk is sufficiently low, some patients may forgo chemotherapy. Both the magnitude of the low-risk estimates and the proportion of patients who fall into those categories are therefore of considerable interest. Because various standard risk prediction tools are freely available, the question is how much the new tests add.

Only a few sources provide evidence on long-term absolute risks after conventional combination risk predictors are taken into account. For Oncotype DX, results of these analyses are published only in abstract form, although the findings are derived from the same NSABP B-14 cohort that provided main original validation evidence and are reported by the same authors (46, 47). This showed that Oncotype DX can reclassify patients in the highest-risk categories by conventional indices (18% for 2003 St. Gallen, 15% for 2004 National Comprehensive Cancer Network, and 22% for Adjuvant! Online) into clinically relevant lower risks (8%, 8%, and 9%, respectively), although the upper confidence limits on those new lower predictions all exceed 10%. For women at the lowest risk by conventional metrics, being placed in a low-risk stratum seems to lower the risk even further, information that patients might find useful; however, the number of patients in this group and on which this finding is based is small (approximately 30 to 60). For MammaPrint, the only reclassification data reported are those obtained in combination with Adjuvant! Online for a 10-year outcome (36). Adjuvant! Online had no predictive power for survival after the data were stratified by MammaPrint risk group, and it had only a very modest effect for 10-year distant recurrence. Similarly, findings were reported for the 70-gene signature in combination with the NIH and St. Gallen criteria (9). The risks in the good-signature MammaPrint groups were higher than in the Oncotype DX low-risk stratum, in part because the more heterogeneous validation population was at higher risk.

The exact values of the test results provide information that is lost when patients are assigned to risk categories, and the cutoffs for these categories may not correspond to optimal decision thresholds, particularly in combination with other predictors. How the results of such tests are conveyed to and understood by patients and physicians—for example, as absolute probabilities or as qualitative descriptors (“low risk”)—is critical as these tests become more widely used.

The ideal assessment of value of these tests would be to randomly assign patients to use them or not, as part their therapeutic decision making. The 2 ongoing prospective randomized trials, TAILORx (Trial Assigning IndividuaLized Options for Treatment) (56) and MINDACT (Microarray In Node-negative Disease may Avoid ChemoTherapy) (57), do not use such a design. The TAILORx compares disease-free survival among women with previously resected axillary node–negative breast cancer who had an Oncotype DX recurrence score between 11 and 25 and received adjuvant chemotherapy and hormonal therapy versus women who received hormonal therapy alone. All patients receive the test (56); those with a score of 10 or less do not receive chemotherapy, and all those with a score greater than 25 receive chemotherapy. These thresholds are lower than those conventionally used to designate high risk (score ≥30) and low risk (score <18).

The MINDACT study is a multicenter, prospective, phase III, randomized study directly comparing MammaPrint with Adjuvant! Online in selecting patients for adjuvant chemotherapy in node–negative breast cancer. Patients at low risk by both MammaPrint and Adjuvant! Online criteria do not receive chemotherapy; patients at high risk by both criteria receive chemotherapy; and patients with discordant criteria are randomly assigned to use either the MammaPrint or Adjuvant! Online results to determine treatment. This trial comes much closer to testing directly the clinical value of MammaPrint (versus Adjuvant! Online) than does TAILORx for Oncotype DX, although both studies will provide valuable evidence bearing on that question.

Our review identified important issues that may arise as these tests are applied in clinical practice, are modified, and as similar tests proliferate. These issues are described in Table 3. As these tests are modified and new ones are marketed, questions will arise about how the tests compare with one another and whether combining the tests has value. Answering these questions will require comparative effectiveness research. The U.S. government and industry often do not fund comparative effectiveness studies because such studies may not offer as much therapeutic promise as new discoveries and because industry is not eager to fund direct comparisons with competitive products. This same dynamic could take hold in the risk-prediction arena. Early in development, oversight of test development and research funding should encourage contrasts with existing expression-based predictors. Otherwise, new tests that all claim to offer similar guidance, or perhaps new guidance in previously neglected clinical subsets, will flood the market, and physicians and patients will have no way to evaluate the claims.

Table 3. Future Issues

In conclusion, the introduction of gene expression tests has ushered in a new era in which many conventional clinical markers may be seen merely as surrogates for more fundamental genetic and physiologic processes that can be measured with these tests. The multidimensional nature of these predictors demands that large numbers of clinically homogeneous patients be used in the validation process and that exceptional rigor and discipline be applied in evaluation. Every study provides an opportunity to modify a genetic signature, but we must find the right balance between speed of innovation and development of reliable tools. It will be important to preserve genetic and clinical information from tested patients to facilitate further evaluation and innovation in current populations. Although these tests show great promise to improve predictions of prognosis and treatment benefit for women with early-stage breast cancer, more must be learned about the extent of that improvement, in whom it is most improved, and how the tests are best incorporated into decision making about current breast cancer treatment.

Glossary

DNA microarray (also called gene chip or DNA chip): A collection of microscopic DNA spots (defined as features), commonly representing single genes or transcripts, arrayed on a solid surface by covalent attachment to chemically suitable matrices or directly synthesized on them. DNA microarrays use DNA as part of their detection system. Qualitative or quantitative measurements with DNA microarrays use the selective nature of DNA–DNA or DNA–RNA hybridization under high-stringency conditions and fluorophore-based detection. DNA arrays are commonly used for gene expression profiling (that is, monitoring expression levels of thousands of genes simultaneously) or for comparative genomic hybridization.

Gene expression: The translation of the information encoded in a gene into an RNA transcript. Expressed transcripts include messenger RNAs translated into proteins, as well as other types of RNA, such as transfer RNA, ribosomal RNA, micro RNA, and noncoding RNA, which are not translated into protein. Gene expression is a highly specific process by which cells switch genes on and off in a timely manner, according to their state. The study of messenger RNA expression in a cell is an indirect way to study the protein's counterpart.

Gene expression pattern: See gene expression profile.

Gene expression profile: Any set of genes that are known to be expressed in a specific sample. Gene expression profiles can be associated with various phenotypes.

Gene expression profiling: Any genomic technique that measures the subset of genes that is expressed in a specific sample. Currently, this definition refers to techniques that allow the assessment of more than 1 gene at a time, especially microarray and real-time reverse transcriptase polymerase chain reaction.

Gene expression signature: A specific gene expression profile, often a subset of expressed genes usually associated with a specific phenotype.

Polymerase chain reaction: A molecular biology technique for isolating and exponentially amplifying a DNA sequence of interest in vitro by means of enzymatic replication. This technique has been extensively modified to do a wide array of tasks, and it is now a common tool used in medical and biological research. Polymerase chain reaction is currently used to obtain the sequence of genes to diagnose hereditary diseases, identify genetic fingerprints (forensic medicine), detect infectious diseases, and create transgenic organisms. Coupled with reverse transcription, it is used to amplify RNA molecules.

Real-time reverse transcriptase polymerase chain reaction: A molecular biology technique that allows amplification and quantification in real time of defined RNA molecules from specific specimens. This technology has been used for several years in research and clinical settings. In brief, in the first step, DNA copies of the investigated RNA molecules are obtained by a process called reverse transcription, and DNA amplification is then obtained by using polymerase chain reaction. The quantification of the accumulating DNA product is accomplished by the use of specific fluorescent reagents. In this technique, the quantification of the target RNA molecule is based on the analysis of the accumulation curve of the complementary DNA, as measured by the fluorescence detected at each cycle of the reaction.

Reverse transcription: In biochemistry, the enzymatic reaction carried on by the RNA-dependent DNA polymerase. This enzyme, known as reverse transcriptase, is a DNA polymerase enzyme that copies single-stranded RNA into DNA. This process is the reverse of normal transcription, which involves the synthesis of RNA from DNA.

Transcription: The process by which DNA sequences are copied into complementary RNA molecules by the enzyme RNA polymerase. This reaction represents the transfer of genetic information from DNA into RNA. The RNA sequence that is transcribed from a DNA molecule is called a transcript.

References

Alizadeh
AA
,
Eisen
MB
,
Davis
RE
,
Ma
C
,
Lossos
IS
,
Rosenwald
A
,
et al
Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling.
Nature
2000
403
503
11
PubMed
CrossRef

PubMed
Alizadeh
AA
,
Ross
DT
,
Perou
CM
,
van de Rijn
M
.
Towards a novel classification of human malignancies based on gene expression patterns.
J Pathol
2001
195
41
52
PubMed
CrossRef

PubMed
Bittner
M
,
Meltzer
P
,
Chen
Y
,
Jiang
Y
,
Seftor
E
,
Hendrix
M
,
et al
Molecular classification of cutaneous malignant melanoma by gene expression profiling.
Nature
2000
406
536
40
PubMed
CrossRef

PubMed
Golub
TR
,
Slonim
DK
,
Tamayo
P
,
Huard
C
,
Gaasenbeek
M
,
Mesirov
JP
,
et al
Molecular classification of cancer: class discovery and class prediction by gene expression monitoring.
Science
1999
286
531
7
PubMed
CrossRef

PubMed
Sotiriou
C
,
Neo
SY
,
McShane
LM
,
Korn
EL
,
Long
PM
,
Jazaeri
A
,
et al
Breast cancer classification and prognosis based on gene expression profiles from a population-based study.
Proc Natl Acad Sci U S A
2003
100
10393
8
PubMed
CrossRef

PubMed
Sørlie
T
,
Perou
CM
,
Tibshirani
R
,
Aas
T
,
Geisler
S
,
Johnsen
H
,
et al
Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications.
Proc Natl Acad Sci U S A
2001
98
10869
74
PubMed
CrossRef

PubMed
Sorlie
T
,
Tibshirani
R
,
Parker
J
,
Hastie
T
,
Marron
JS
,
Nobel
A
,
et al
Repeated observation of breast tumor subtypes in independent gene expression data sets.
Proc Natl Acad Sci U S A
2003
100
8418
23
PubMed
CrossRef

PubMed
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
530
6
PubMed
CrossRef

PubMed
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
1999
2009
PubMed
CrossRef

PubMed
Thorlacius
S
,
Thorgilsson
B
,
Björnsson
J
,
Tryggvadottir
L
,
Börresen
AL
,
Ogmundsdottir
HM
,
et al
TP53 mutations and abnormal p53 protein staining in breast carcinomas related to prognosis.
Eur J Cancer
1995
31A
1856
61
PubMed
CrossRef

PubMed
National Institutes of Health
Adjuvant Therapy for Breast Cancer.
NIH Consensus Statement Online
2000
17
1
23
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
3817
27
PubMed
CrossRef

PubMed
Goldhirsch
A
,
Glick
JH
,
Gelber
RD
,
Coates
AS
,
Thürlimann
B
,
Senn
HJ
.
Panel members
Meeting highlights: international expert consensus on the primary therapy of early breast cancer 2005.
Ann Oncol
2005
16
1569
83
PubMed
CrossRef

PubMed
Carlson
RW
,
Anderson
BO
,
Bensinger
W
,
Cox
CE
,
Davidson
NE
,
Edge
SB
,
et al
National Comprehensive Cancer Network
NCCN Practice Guidelines for Breast Cancer.
Oncology (Williston Park)
2000
14
33
49
PubMed
PubMed
Carlson
RW
,
Anderson
BO
,
Burstein
HJ
,
Carter
WB
,
Edge
SB
,
Farrar
WB
,
et al
Invasive breast cancer.
J Natl Compr Canc Netw
2007
5
246
312
PubMed
PubMed
Carlson
RW
,
Hudis
CA
,
Pritchard
KI
.
National Comprehensive Cancer Network Breast Cancer Clinical Practice Guidelines in Oncology. American Society of Clinical Oncology Technology Assessment on the Use of Aromatase Inhibitors. St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer. Adjuvant endocrine therapy in hormone receptor positive postmenopausal breast cancer: evolution of NCCN, ASCO, and St Gallen recommendations.
J Natl Compr Canc Netw
2006
4
971
9
PubMed
PubMed
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
980
91
PubMed
CrossRef

PubMed
Adjuvant! Inc. Adjuvant Online. Decision making tools for health care professionals. Accessed at adjuvantonline.com/ on 12 December 2007.
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
2817
26
PubMed
CrossRef

PubMed
Ma
XJ
,
Wang
Z
,
Ryan
PD
,
Isakoff
SJ
,
Barmettler
A
,
Fuller
A
,
et al
A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen.
Cancer Cell
2004
5
607
16
PubMed
CrossRef

PubMed
McShane
LM
,
Altman
DG
,
Sauerbrei
W
,
Taube
SE
,
Gion
M
,
Clark
GM
,
et al
Statistics Subcommittee of the NCI-EORTC Working Group on Cancer Diagnostics
REporting recommendations for tumour MARKer prognostic studies (REMARK).
Br J Cancer
2005
93
387
91
PubMed
CrossRef

PubMed
McShane
LM
,
Altman
DG
,
Sauerbrei
W
,
Taube
SE
,
Gion
M
,
Clark
GM
,
et al
Statistics Subcommittee of the NCI-EORTC Working Group on Cancer Diagnostics
Reporting recommendations for tumor marker prognostic studies (REMARK).
J Natl Cancer Inst
2005
97
1180
4
PubMed
CrossRef

PubMed
Bossuyt
PM
,
Reitsma
JB
,
Bruns
DE
,
Gatsonis
CA
,
Glasziou
PP
,
Irwig
LM
,
et al
Standards for Reporting of Diagnostic Accuracy. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration.
Ann Intern Med
2003
138
W1
12
CrossRef

PubMed
Bossuyt
PM
,
Reitsma
JB
.
Standards for Reporting of Diagnostic Accuracy. The STARD initiative [Letter].
Lancet
2003
361
71
PubMed
CrossRef

PubMed
Cobleigh
MA
,
Tabesh
B
,
Bitterman
P
,
Baker
J
,
Cronin
M
,
Liu
ML
,
et al
Tumor gene expression and prognosis in breast cancer patients with 10 or more positive lymph nodes.
Clin Cancer Res
2005
11
8623
31
PubMed
CrossRef

PubMed
Cronin
M
,
Pho
M
,
Dutta
D
,
Stephans
JC
,
Shak
S
,
Kiefer
MC
,
et al
Measurement of gene expression in archival paraffin-embedded tissues: development and performance of a 92-gene reverse transcriptase-polymerase chain reaction assay.
Am J Pathol
2004
164
35
42
PubMed
CrossRef

PubMed
Cronin
M
,
Sangli
C
,
Liu
ML
,
Pho
M
,
Dutta
D
,
Nguyen
A
,
et al
Analytical validation of the Oncotype DX genomic diagnostic test for recurrence prognosis and therapeutic response prediction in node-negative, estrogen receptor-positive breast cancer.
Clin Chem
2007
53
1084
91
PubMed
CrossRef

PubMed
Gianni
L
,
Zambetti
M
,
Clark
K
,
Baker
J
,
Cronin
M
,
Wu
J
,
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
7265
77
PubMed
CrossRef

PubMed
Mina
L
,
Soule
SE
,
Badve
S
,
Baehner
FL
,
Baker
J
,
Cronin
M
,
et al
Predicting response to primary chemotherapy: gene expression profiling of paraffin-embedded core biopsy tissue.
Breast Cancer Res Treat
2007
103
197
208
PubMed
CrossRef

PubMed
Habel
LA
,
Shak
S
,
Jacobs
MK
,
Capra
A
,
Alexander
C
,
Pho
M
,
et al
A population-based study of tumor gene expression and risk of breast cancer death among lymph node-negative patients.
Breast Cancer Res
2006
8
R25
PubMed
CrossRef

PubMed
Esteva
FJ
,
Sahin
AA
,
Cristofanilli
M
,
Coombes
K
,
Lee
SJ
,
Baker
J
,
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
3315
9
PubMed
CrossRef

PubMed
Chang JC, Makris A, Gutierrez MC, Hilsenbeck SG, Hackett JR, Jeong J, et al. Gene expression patterns in formalin-fixed, paraffin-embedded core biopsies predict docetaxel chemosensitivity in breast cancer patients. Breast Cancer Res Treat. 2007 Apr 28 [Epub ahead of print]. [PMID: 17468949]
Fan
C
,
Oh
DS
,
Wessels
L
,
Weigelt
B
,
Nuyten
DS
,
Nobel
AB
,
et al
Concordance among gene-expression-based predictors for breast cancer.
N Engl J Med
2006
355
560
9
PubMed
CrossRef

PubMed
Ach
RA
,
Floore
A
,
Curry
B
,
Lazar
V
,
Glas
AM
,
Pover
R
,
et al
Robust interlaboratory reproducibility of a gene expression signature measurement consistent with the needs of a new generation of diagnostic tools.
BMC Genomics
2007
8
148
PubMed
CrossRef

PubMed
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
278
PubMed
CrossRef

PubMed
Buyse
M
,
Loi
S
,
van't Veer
L
,
Viale
G
,
Delorenzi
M
,
Glas
AM
,
et al
TRANSBIG Consortium
Validation and clinical utility of a 70-gene prognostic signature for women with node-negative breast cancer.
J Natl Cancer Inst
2006
98
1183
92
PubMed
CrossRef

PubMed
Reid
JF
,
Lusa
L
,
De Cecco
L
,
Coradini
D
,
Veneroni
S
,
Daidone
MG
,
et al
Limits of predictive models using microarray data for breast cancer clinical treatment outcome.
J Natl Cancer Inst
2005
97
927
30
PubMed
CrossRef

PubMed
Goetz
MP
,
Suman
VJ
,
Ingle
JN
,
Nibbe
AM
,
Visscher
DW
,
Reynolds
CA
,
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
2080
7
PubMed
CrossRef

PubMed
Jansen
MP
,
Sieuwerts
AM
,
Look
MP
,
Ritstier
K
,
Meijer-van Gelder
ME
,
van Staveren
IL
,
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
662
8
PubMed
CrossRef

PubMed
Jerevall
PL
,
Brommesson
S
,
Strand
C
,
Gruvberger-Saal
S
,
Malmström
P
,
Nordenskjöld
B
,
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
225
34
PubMed
CrossRef

PubMed
Ma
XJ
,
Hilsenbeck
SG
,
Wang
W
,
Ding
L
,
Sgroi
DC
,
Bender
RA
,
et al
The HOXB13:IL17BR expression index is a prognostic factor in early-stage breast cancer.
J Clin Oncol
2006
24
4611
9
PubMed
CrossRef

PubMed
Fisher
B
,
Dignam
J
,
Bryant
J
,
Wolmark
N
.
Five versus more than five years of tamoxifen for lymph node-negative breast cancer: updated findings from the National Surgical Adjuvant Breast and Bowel Project B-14 randomized trial.
J Natl Cancer Inst
2001
93
684
90
PubMed
CrossRef

PubMed
Fisher
B
,
Jeong
JH
,
Bryant
J
,
Anderson
S
,
Dignam
J
,
Fisher
ER
,
et al
National Surgical Adjuvant Breast and Bowel Project randomised clinical trials. Treatment of lymph-node-negative, oestrogen-receptor-positive breast cancer: long-term findings from National Surgical Adjuvant Breast and Bowel Project randomised clinical trials.
Lancet
2004
364
858
68
PubMed
CrossRef

PubMed
Fisher
B
,
Dignam
J
,
Wolmark
N
,
Wickerham
DL
,
Fisher
ER
,
Mamounas
E
,
et al
Tamoxifen in treatment of intraductal breast cancer: National Surgical Adjuvant Breast and Bowel Project B-24 randomised controlled trial.
Lancet
1999
353
1993
2000
PubMed
CrossRef

PubMed
Fisher
B
,
Dignam
J
,
Bryant
J
,
DeCillis
A
,
Wickerham
DL
,
Wolmark
N
,
et al
Five versus more than five years of tamoxifen therapy for breast cancer patients with negative lymph nodes and estrogen receptor-positive tumors.
J Natl Cancer Inst
1996
88
1529
42
PubMed
CrossRef

PubMed
Paik S, Shak S, Tang G. Risk classification of breast cancer patients by the recurrence score assay: comparison to guidelines based on patient age, tumor size, and tumor grade [Poster presentation]. Poster presented at the 28th Annual San Antonio Breast Cancer Symposium, December 2005, San Antonio, Texas.
Bryant J. Toward a more rational selection of tailored adjuvant therapy [Poster presentation]. Poster presented at the 9th International St. Gallen Oncology Conference on Primary Therapy of Early Breast Cancer, 26-29 January 2005, St. Gallen, Switzerland.
Ma
XJ
,
Salunga
R
,
Tuggle
JT
,
Gaudet
J
,
Enright
E
,
McQuary
P
,
et al
Gene expression profiles of human breast cancer progression.
Proc Natl Acad Sci U S A
2003
100
5974
9
PubMed
CrossRef

PubMed
Goetz MP, Ingle JN, Couch FJ. Gene-expression-based predictors for breast cancer [Letter]. N Engl J Med. 2007;356:752; author reply 752-3. [PMID: 17301312]
Paik
S
,
Tang
G
,
Shak
S
,
Kim
C
,
Baker
J
,
Kim
W
,
et al
Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptor-positive breast cancer.
J Clin Oncol
2006
24
3726
34
PubMed
CrossRef

PubMed
Harrell
FE
Jr
,
Lee
KL
,
Mark
DB
.
Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.
Stat Med
1996
15
361
87
PubMed
CrossRef

PubMed
Laupacis
A
,
Sekar
N
,
Stiell
IG
.
Clinical prediction rules. A review and suggested modifications of methodological standards.
JAMA
1997
277
488
94
PubMed
CrossRef

PubMed
Simon
R
.
Development and validation of therapeutically relevant multi-gene biomarker classifiers [Editorial].
J Natl Cancer Inst
2005
97
866
7
PubMed
CrossRef

PubMed
Simon
R
.
Roadmap for developing and validating therapeutically relevant genomic classifiers.
J Clin Oncol
2005
23
7332
41
PubMed
CrossRef

PubMed
Simon
R
,
Wang
SJ
.
Use of genomic signatures in therapeutics development in oncology and other diseases.
Pharmacogenomics J
2006
6
166
73
PubMed
CrossRef

PubMed
National Cancer Institute. The TAILORx Breast Cancer Trial. Accessed at www.cancer.gov/clinicaltrials/digestpage/TAILORx on 12 December 2007.
Breast International Group. TRANSBIG. MINDACT. Accessed at www.breastinternationalgroup.org/TRANSBIG/MINDACT on 12 December 2007.
Hedenfalk
I
,
Duggan
D
,
Chen
Y
,
Radmacher
M
,
Bittner
M
,
Simon
R
,
et al
Gene-expression profiles in hereditary breast cancer.
N Engl J Med
2001
344
539
48
PubMed
CrossRef

PubMed
Brazma
A
,
Hingamp
P
,
Quackenbush
J
,
Sherlock
G
,
Spellman
P
,
Stoeckert
C
,
et al
Minimum information about a microarray experiment (MIAME)-toward standards for microarray data.
Nat Genet
2001
29
365
71
PubMed
CrossRef

PubMed

Figure 1.

Technologies used for high-throughput gene expression analysis.

Figure 2.

Systematic search strategy and results.

Appendix Table 1. Studies on the Oncotype DX Gene Expression Test

Appendix Table 2. Studies on the MammaPrint Gene Expression Test

Appendix Table 3. Studies on the H/I Gene Expression Test

Table 1. Patient Reclassification by Gene Expression Testing with Oncotype DX

Table 2. Kaplan–Meier Analysis of Survival Stratified by MammaPrint and Adjuvant! Online

Table 3. Future Issues

Clinical Slide Sets

Terms of Use

The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.

This feature is available only to Registered Users

Subscribe/Learn More

PDF

Not Available

2008 American College of Physicians

Related Point of Care

Breast Cancer Screening and Prevention

Annals of Internal Medicine; 164 (11): ITC81-ITC96

Breast Cancer Screening and Prevention

Annals of Internal Medicine; 152 (7): ITC4-1

View MoreView Less

Reviews |4 March 2008

Systematic Review: Gene Expression Profiling Assays in Early-Stage Breast Cancer Free

Abstract

Figure 1.

Methods

Data Sources

Study Selection

Data Extraction

Appendix Table 1. Studies on the Oncotype DX Gene Expression Test

Appendix Table 2. Studies on the MammaPrint Gene Expression Test

Appendix Table 3. Studies on the H/I Gene Expression Test

Role of the Funding Source

Results

Figure 2.

Analytic Validity and Variability

Predicting Disease Outcomes

Oncotype DX

Table 1. Patient Reclassification by Gene Expression Testing with Oncotype DX

MammaPrint

Table 2. Kaplan–Meier Analysis of Survival Stratified by MammaPrint and Adjuvant! Online

The H/I Test

Comparison of Signatures

Predicting Treatment Response

Discussion

Table 3. Future Issues

Glossary

References

Figure 1.

Figure 2.

Appendix Table 1. Studies on the Oncotype DX Gene Expression Test

Appendix Table 2. Studies on the MammaPrint Gene Expression Test

Appendix Table 3. Studies on the H/I Gene Expression Test

Table 1. Patient Reclassification by Gene Expression Testing with Oncotype DX

Table 2. Kaplan–Meier Analysis of Survival Stratified by MammaPrint and Adjuvant! Online

Table 3. Future Issues

Clinical Slide Sets

This feature is available only to Registered Users

0 Comments

See Also

Related Articles

Journal Club

Related Point of Care

Related Topics

PubMed Articles

Content

Information For

Services

Awards and Cover

Other Resources

Follow Annals On

You need a subscription to this content to use this feature.