November 2007: Volume 33, Number 11
Gene Expression Profile Tests for Breast Cancer Recurrence
Do They Inform Treatment Decisions?
By Deborah Levenson
Adjuvant chemotherapy for breast cancer significantly reduces risk of recurrence for certain women, but others experience the drugs’ harmful side effects without any benefit. Current guidelines from the National Comprehensive Cancer Network recommend that oncologists consider adjuvant chemotherapy for women with tumors larger than 1 cm or smaller tumors with troublesome histologic features. While 80–85% of women with estrogen receptor (ER) positive, lymph-node negative cancers are adequately treated with hormonal therapy alone, the guidelines suggest that chemotherapy can reduce their relative risk of recurrence by about 25%. But research shows that chemotherapy confers an absolute reduction in risk for individual patients ranging from just 1–5%. Consequently, many studies have focused on better diagnostic tools to help individualize breast cancer chemotherapy. One promising approach is gene expression profiling, which includes tests that would help oncologists determine in conjunction with classic markers which women will actually need adjuvant chemotherapy. A few currently available and forthcoming gene expression profile assays are intended to reduce unnecessary treatments, yet oncologists are not always sure what to do with the data, especially if the results show the cancer carries a moderate risk of recurrence.
Using these tests poses another, more basic clinical question: How strong is the data? Most companies back up their claims about these tests—which cost from about $1,000 to more than $3,000—with data from retrospective analyses rather than prospective trials. Large ongoing trials for two of these tests may increase practitioners’ comfort with them, but final data for both is years away. “The scientist in me says there’s not data to say these assays have a prospective role in therapy decision making. But if my wife or daughter had breast cancer, I would want to them to get an assay,” commented pathologist Michael Pins, MD, former director of the Eastern Cooperative Oncology Group Coordinating Office and Reference Laboratory, a solid tumor repository for one of the ongoing retrospective Trial Assigning Individualized Options for Treatment (Rx), commonly known as TAILORx, and Coordinating Officer of the pathology department reference lab at Northwestern Hospital in Chicago, Ill. He is now a practicing pathologist at Lutheran General Hospital in Chicago and Visiting Professor of Clinical Pathology, University of Illinois at Chicago College of Medicine.
Current widespread insurance coverage for the Oncotype DX assay developed by the reference lab Genomic Health (Redwood City, Calif.) and FDA clearance for marketing approval for Agendia’s MammaPrint assay (Amsterdam, Netherlands) may make these two tests attractive options. As more oncologists and laboratorians decide what to make of these tests, they must decide if the tests will add useful information to the usual prognostic markers and if a patient’s disease, age, and ethnicity is similar to a particular test’s validation cohort, say experienced users.
Oncotype and MammaPrint: A Comparison
Since January 2004, more than 6,000 physicians have ordered the Oncotype test for over 33,000 patients, according to Genomic Health. However, Agendia will not formally launch MammaPrint in the U.S. until later this year. Not surprisingly, U.S. oncologists and pathologists tend to be more familiar with Oncotype. The test determines 10-year risk of recurrence using formalin-fixed, paraffin-embedded tumor tissue—which can be stored for years—to analyze the expression of 21 genes, five of which are reference genes. Samples sent to Genomic Health receive DNase 1 treatment followed by verification that samples are contaminant-free before they undergo reverse transcriptase-polymerase chain reaction (RT-PCR). The results of the test are used to assign a risk score that relies upon an algorithm which weighs certain genes more than others. Results are available within 10 to 14 days, according to the Genomic Health Web site.
Initial validation data for Oncotype, published in December 2004 (NEJM 2004; 351: 2817–2826), came from 668 patient samples from ER-positive, node-negative, tamoxifen-treated breast cancer patients of various ages who were enrolled in the National Surgical Adjvant Breast and Bowel Project clinical trial B-14. In that study, researchers classified 51% of patients as having low risk of recurrence, with a 6.8% rate of recurrence at 10 years. A smaller proportion—27% of women tested—was deemed high risk and had a 10-year recurrence rate of 30.5%. Researchers concluded that the recurrence score predicts both overall survival and relapse-free survival, commenting that “few assays have been rigorously validated for use as prognostic or predictive tests in oncology.” More validation data came from a study conducted at 14 Northern California Kaiser Permanente Hospitals. (Breast Cancer Res 2006; 8: R25).
Beyond the validation studies, published research shows that Oncotype is indeed predictive. In a study of tumors from 651 patients treated with either tamoxifen alone or with both tamoxifen and chemotherapy, investigators from the University of Pittsburgh, Alleghany General Hospital (Pittsburgh, Pa.), and Genomic Health found that patients who had high Oncotype risk scores had large benefit from chemotherapy, while those with low risk scores had minimal, if any benefit. (J Clin Oncol 2006; 24(23) 1–12).
At least one study has shown that Oncotype risk scores affect treatment decisions. In a study of 93 patients in one community and three academic oncology practices, risk scores from the assay changed oncologists’ recommended treatment plan in 31.5% of cases, according to data presented at the 2007 annual meeting of the American Society of Clinical Oncology by researchers from Loyola University in Marywood, Ill., University of Michigan at Ann Arbor, University of California-Davis, and Mount Sinai Medical Center in New York, N.Y.
In contrast, MammaPrint is prognostic rather than predictive, involves more genes, is applicable to a larger but younger group of women, and requires fresh or frozen tissue. The DNA microarray test uses a 70-gene expression signature consisting of genes regulating cell cycle, invasion, metastasis, and angiogenesis to assess risk of recurrence in ER-positive or negative, stage 1 and 2, lymph-node negative women under age 61. To Agendia’s credit, MammaPrint is the first device approved under FDA’s controversial new in vitro diagnostic multivariate index assay (IVDMIA) classification (CLN, March 2007). The FDA initially approved the test for use on frozen samples analyzed in Agendia’s Amsterdam labs; however, in June, FDA granted the company a second approval for use of fresh samples preserved in a special solution and sent to the company’s lab. In its decision summary, FDA notes that the assay is indicated for use only as a prognostic marker, along with other clinicopathological factors.
To assess the global gene expression pattern of a breast tumor sample, Agendia’s lab extracts messenger RNA from the sample and labels it with a fluorescent dye. The labeled mRNA, together with the labeled mRNA from a reference sample, is hybridized to the DNA microarray. The customized arrays also contain several hundred carefully selected normalization genes. Finally, negative control genes are present on each microarray; these are DNA sequences to which human mRNA does not hybridize and that are used to monitor various technical aspects of the microarray process. Results are available within 10 working days of sample receipt, according to Agendia’s Web site.
The MammaPrint validation study (Journal of the National Cancer Institute 2006; 98: 1183–1192) divided 307 patients into high- and low-risk groups based on their gene signature and clinical risk classifications—with 137 events after a median follow-up of 13.6 years by five European centers. Patients were assigned to the gene signature low-risk group if their 5-year distant metastasis-free survival probability as estimated by the gene signature was greater than 90%. Patients were assigned to the clinicopathologic low-risk group if their 10-year survival probability, as estimated by algorithm software from Adjuvant! Online, was greater than 88% for ER-positive patients and 92% for ER-negative patients. Researchers from 39 European institution members of the TRANSBIG Consortium found that the signature outperformed the clinicopathologic risk assessment in predicting all endpoints. The FDA summary for MammPrint notes that MammPrint’s clinical performance has been supported by three other prior studies that used TRANSBIG data.
The tests’ technologies and validation cohorts raise many issues, noted both Genomic Health and Agendia officials. Gary Palmer, MD, Genomic Health’s Senior Director of Medical Affairs pointed first to predictive data. “Most importantly and unlike any other tests, Oncotype DX has a predictive component as published in the Journal of Clinical Oncology in 2006.” Then there is the practical matter of the tumor sample itself. Oncotype’s use of paraffin-embedded tumor samples reflects most U.S. hospitals’ collection methods, whereas MammaPrint’s use of fresh samples would present difficulty at many institutions, Palmer noted “There’s no other test done on fresh tissue at most community hospitals,” he said, noting that most institutions are not set up to collect it.
But Rene Bernards, Agendia’s Chief Scientific Officer, regarded his company’s use of fresh tissue as a strength. “Each tumor starts fresh. Paraffin embedding used to preserve tissue is a technique that dates to the nineteenth century. It’s not the wave of the future.” He predicted that as genomic tests become more common, paraffin embedding will become “a thing of the past” because it degrades RNA. “FDA approval takes into consideration how you ensure RNA quality,” he said. “The bottom line is yes, paraffin is more convenient. The question you need to ask yourself as a scientist is whether to go for accuracy or convenience.”
Genomic Health has taken measures to ensure RNA quality in its assay, Palmer asserted. Oncotype “was optimized to measure RNA fragments found in formalin-fixed, paraffin-embedded tissue and the assay includes reference genes that essentially measure the amount of RNA degradation, since all specimens are not the same with respect to fixation,” he explained. Research by the company shows that the assay’s five reference genes both compensate for sample variation in extracted RNA attributable to the tissue fixation process and normalize expression of the 16 cancer-related genes prior to calculation of the recurrence score (Clinical Chemistry 2007; 53: 1084–1091).
Which patients may potentially benefit from the test is another key issue to consider. According to Palmer, Genomic Health’s test is designed for ER-positive patients. “Since these women are likely to respond to hormone therapy like tomoxifen or aromatase inhibitors, oncologists often have trouble deciding whether additional chemotherapy is necessary. This situation makes a test geared for ER-positive women valuable,” he explained. “MammaPrint’s validation studies didn’t separate out ER-positive and ER-negative patients. Physicians really question how to treat ER-negative patients.” A substantial number of breast cancer patients are also older than 61 and would not benefit from MammaPrint, he added.
According to Bernards, MammaPrint’s FDA approval confers a major advantage. “FDA looked at our data more deeply to make sure it agrees with the claims we make. So clinicians know the test is based on solid science. Physicians with no insight into the details of a (reference lab’s) test will have difficulty knowing if the claims are sufficiently founded. The dilemma for physicians is whether to wait for prospective validation (of MammaPrint) or to go with our test based on its FDA approval.”
More Gene Profiling Assays for Breast Cancer
Lab directors and oncologists considering gene profiling assays to help determine risk of recurrence or tailor treatment of breast cancer patients have an expanding list of options to consider. In addition to the Oncotype DX (Genomic Health, Redwood City, Calif.) and MammaPrint assay (Agendia, Amsterdam, Netherlands), other tests are available or forthcoming.
Quest Diagnostics (Lyndhurst, N.J.) has offered the Breast Cancer Gene Expression Ratio since December 2006. Results of the six-gene, PCR test are based on the ratio of the expression of the homeobox gene-B13 and the interleukin-17B receptor gene. The test assesses the likelihood of recurrence for ER positive, node-negative patients.
Mammostrat, developed by Applied Genomics (Huntsville, Ala.) and currently offered by the Molecular Profiling Institute (Phoenix, Ariz.), relies on five genes to assign a risk value in postmenopausal, node-negative, ER positive patients.
ARUP Laboratories (Salt Lake City, Utah) is submitting its first 510(k) application for the Breast Bioclassifier, a 50-gene assay designed to identify five biological breast cancer subtypes and assign a risk score for women with ER-positive and ER-negative tumors. Like Oncotype, it’s intended to identify which ER-positive tumors are unlikely to recur. Unlike other assays, it will identify subtypes of ER-negative tumors, such as Basal-like tumors, that have shown “exquisite” sensitivity to particular chemotherapeutic regimens.
Exagen Diagnostics (Albuquerque, N.M.) is awaiting 510(k) approval for the eXagenBC assay, a fluorescence in situ hybridization (FISH) test that assigns risk for stage I ER-positive patients.
The Users Weigh In
Pathologists and oncologists interviewed for this article generally viewed gene expression profiling tests as promising, but their varied opinions about Oncotype and MammaPrint show that profiling tests haven’t achieved widespread acceptance. Some of the physicians emphasized that they still consider the tests secondary to traditional markers used to determine risk of recurrence but prefer Oncotype for its use of paraffin embedding and its predictive ability, while others preferred MammaPrint for its larger number of genes and use in both ER-positive and negative cancers.
“Personally, I’m a skeptic about these sorts of tests in general,” said Pins. “I need to see good, solid prospective data.” Pins prefers Oncotype over MammaPrint because getting fresh tissue “is a laborious process with little standardization between sites.” Richard Bender, MD, Medical Director, Hematology and Oncology at Quest Diagnostics’ Nichols Institute in San Juan Capistrano, Calif.—which markets its own gene profiling test— agreed that using fresh tissue is a downside to MammaPrint. “It’s not based on the way we practice medicine here in the U.S,” he maintained. David Hyams, MD, National Director of Clinical Research for Aptium Oncology, Inc. (Los Angeles, Calif.), Medical Director of Surgical Oncology at Comprehensive Cancer Center at Desert Regional Medical Center in Palm Springs, Calif., and a consultant to Genomic Health, explained that with paraffin-embedded tissue, “You don’t have to make a decision about an assay before a patient comes to surgery.”
Oncotype provides more information about a tumor than does MammaPrint, in Hyams’ opinion. “Oncotype tells you more about the biology of tumors, while MammaPrint basically tells you only if a patient is at high- or low-risk. It doesn’t rank patients, whereas Oncotype does.” He added that although MammaPrint looks at more genes, it gives less information about each one because it uses microarrays rather than PCR. “It can’t tell you the degree to which a gene is over or under expressed. It tells you if a patient if at low or high risk of recurrence, period. It doesn’t say how high,” he explained. Hyams also prefers Oncotype because it provides predictive information, and he considers the assay more rigidly validated than MammaPrint.
Pathologist Neal Goldstein, MD, Director of Advanced Diagnostics and the Molecular Oncology Lab at William Beaumont Hospital (Royal Oak, Mich.) said he looks to the future and the availability of more data on gene expression profile assays, but in the meantime, he prefers MammaPrint because it’s applicable for both ER-positive and negative cancers. He’s concerned about the current lack of standards for determining if samples are ER positive. “What’s considered positive in one institution may not be in another. So if you’re using Oncotype, there’s a question regarding the stringency of what you’re selecting for and who is eligible for the assay.” Genomic Health’s Palmer agreed that determination of ER status varies from hospital to hospital, but he maintained that despite any irregularities, oncologists must make decisions about treatment based on ER status, whatever a hospital’s methods or quality assurance.
Another reason for Goldstein’s preference for the MammaPrint test is that it encompasses many more genes than the Oncotype test. “The clinical utility of assays using smaller numbers of genes is dubious in my opinion. Cancers are complex, and oncologists are trying to identify not only the genes involved, but also what signaling pathways are being disturbed by being turned off or on by the cancer. Therapy must match those alterations.” That said, Goldstein noted that William Beaumont hospital oncologists most often order the Oncotype test to convince reluctant patients to undergo chemotherapy. “It appears that nobody withholds chemotherapy based on a low recurrence score,” he commented.
The Ethnicity Question
Gene expression tests may be more useful in smaller community hospitals, but decisions must take into account the ethnicity of particular patients and validation cohorts, according to Philip Bernard, MD, Assistant Professor in the Department of Pathology at the University of Utah in Salt Lake City. He is now preparing a 510(k) submission for a new prognostic test he hopes ARUP Labs (Salt Lake City, Utah) will offer sometime in 2008. “There’s a place for multianalyte tests. They may be more useful for oncologists in community institutions that don’t have access to a panel of experts like their counterparts in academic medical centers. But it’s important to be aware of whether the patient is similar to the assay validation cohort. All of these tests are better validated in Caucasians than in other ethnic groups,” he advised.
Allen M. Gown, MD, Clinical Professor of Pathology, University of British Columbia (Vancouver, Canada), and Medical Director and Chief Pathologist, PhenoPath Labs (Seattle, Wash.), which is participating in a study of the Oncotype assay, has another view. “Is Oncotype giving us information we don’t already have?” he asked. “There are other ways to stratify patients, and many are part of the genes examined by Oncotype. But Oncotype is more accurate and standardized than immunohistrochemistry. If the test only gives numbers that are accurate and reproducible, instead of novel, that may be enough to make it worthwhile. If you get better, more reliable numbers, it’s worth the price, especially if insurance pays,” he asserted.
The Bottom Line: Useful Information
Whatever their preference for a certain assay or opinion about gene expression profiling tests in general, these clinicians and lab directors agreed that the bottom line on using the gene profiling tests is whether they yield useful information. “The issue is whether stratification of risk and related change of treatment affects outcomes. What’s important isn’t so much stratifying risk but showing that the information allows change of intervention that reduces recurrence,” summarized Nichols Institute’s Bender. “The problem is that there’s still a number of younger patients who are ER positive, node negative, and with slightly higher grade cancers who are concerned that hormone therapy is not enough. If there’s a high likelihood that cancer will recur, that information will help them make an informed decision about chemotherapy.”
A Long Wait for More Data
Data from large, ongoing studies may provide better insight about using the information provided by the Oncotype and MammaPrint tests and aid in choosing a test, but trials won’t conclude any time soon. Preliminary prospective data on the benefit of chemotherapy to women with intermediate Oncotype risk scores won’t be available until at least 2013, according to Joseph A. Sparano, MD, Professor of Medicine and Women’s Health at Albert Einstein College of Medicine, Director of the Breast Evaluation Center at the Montefiore-Einstein Cancer Center (Bronx, N.Y), and chair of the National Cancer Institute-sponsored TAILORx trial. Started in April 2006, the trial will enroll 10,000 women with ER-positive, HER2-negative, node-negative breast cancer tumors ranging from 1.0–5 cm or 0.5–1.0 cm with unfavorable histologic features.
Preliminary prospective data for the MammaPrint trial is a minimum of 5 years away, according to a recent newsletter published by the Treatment of Cancer and the Translational Research Breast International Group (TRANSBIG), which is conducting The Microarray in Node Negative Disease May Avoid Chemotherapy trial (MINDACT) with support from the European Organization for Research. The trial will compare MammaPrint with predictions of recurrence risk made by the Adjuvant! Online algorithm using such factors as patient age, tumor size, nodal involvement, and histologic grade. MINDACT aims to enroll 6,000 women, and has thus far enrolled 67, according to the newsletter.
Palmer and Sparano maintained that the two trials ask different questions. While MINDACT is a validation trial to determine if the assay works, TAILORx assumes OncoType DX works and is using it as a tool to assign treatment, as well as whether women with mid-range risk scores need chemotherapy, they said.
A collaboration between the research groups performing the two trials is in the works. In an attempt to perform a cross-validation between the two trials and their tools, TRANSBIG recently decided to collect 100 frozen samples from selected U.S. centers for inclusion in MINDACT. If all goes well, certain U.S. centers will give TRANSBIG 100 samples, both frozen and paraffin-embedded.
While lab directors and oncologists are divided on the current utility of these assays, Hyams believes they represent “the dawn of a new age” in breast cancer treatment. “These are the ‘Model Ts’ of gene profiling assays. Ultimately, we will get a variety of gene panels that will help us understand the particular disease and customize therapy.”