Breast cancer dilemma: Adjuvant chemo—can
  new assays help?

March 2007
Cover Story

William Check, PhD

Like the ancient Roman god Janus, whose two faces looked back to the old year and forward to the new, adjuvant chemotherapy has two aspects—toxicity and benefit. For women with estrogen receptor-positive node-negative tumors that are more than 1 cm in diameter, current guidelines advise consideration of adjuvant chemotherapy. Yet chemotherapy helps only a minor fraction of these patients. As one editorialist wrote, “[A]re we willing to treat all patients with an excellent prognosis to benefit just 1 percent, 2 percent, 3 percent, or maybe 5 percent of patients?” (Swain S. J Clin Oncol. 2006;24:3717–3718).

For many years researchers have been looking for a way to identify early-stage breast cancer patients who will see only the toxic face of chemotherapy, thereby reducing harm without reducing survival. Yet no biomarkers of residual risk are currently recommended for guiding treatment. Now this situation may be changing: At least a half-dozen multianalyte assays at various stages of development and with various degrees of evidence claim to stratify patients by residual risk after localized treatment and hormonal therapy. One claims to select patients who will benefit from chemotherapy, potentially halving the number of early-stage patients for whom chemotherapy is recommended.

However, many do not find the evidence for their claims completely convincing. Like Janus, these skeptics tend to see the issue in a temporal perspective. One major argument is that all the data come from retrospective analyses, rather than meeting the gold standard—prospective randomized trials. Critics are also not convinced the new assays have been proved to be superior to accepted practice. They question whether molecular assays do a better job than a combination of classic pathologic markers—size, grade, stage, presence or absence of lymph node metastases and lymphovascular invasion, and determination of estrogen and progesterone receptors (ER and PR) and HER2.

“All of these new assays are very promising,” acknowledges Neal S. Goldstein, MD, staff anatomic pathologist and director of the Immunohistochemistry-Molecular Pathology Laboratory at William Beaumont Hospital, Royal Oak, Mich. “They have been a great boon to our understanding of breast carcinogenesis.” However, he notes that Oncotype DX (Genomic Health Inc.), the assay with the most extensive evidence and the only one commercially available in the U.S., “has been ‘validated’ against retrospective databases populated by patients with carcinomas that are much larger than they are today and against distant metastasis risk after adjuvant chemotherapy that included agents which are infrequently used in contemporary chemotherapy regimes.” In addition, these older database specimens had been obtained by surgical excision, as opposed to the widespread contemporary use of core needle biopsy. More generally, he adds, “For all of these assays, the big question is whether the additional information they provide is cost-worthy compared to the relatively low-cost information produced by standard predictive and prognostic markers.” In Dr. Goldstein’s view, that question will remain open until prospective studies have been completed and analyzed, which could be years away.

Genomic Health went about evaluating its assay in the right way, says Stuart J. Schnitt, MD, director of the Division of Anatomic Pathology at Beth Israel Deaconess Medical Center and professor of pathology at Harvard Medical School, Boston. “An enormous number of prognostic and predictive factors have been proposed over the past 20 years, most of which dropped off the map because the studies were done in a way that did not permit them to be validated,” Dr. Schnitt notes. ”Genomic Health raised the bar for tests like this by teaming up with NSABP [National Surgical Adjuvant Breast and Bowel Project].”

However, Dr. Schnitt agrees that even Oncotype DX has not been proved to be superior to standard histopathologic parameters. “I have made the argument that [published comparisons] are like comparing the New York Yankees as a team to Manny Ramirez as an individual,” he says. “Most extant comparisons show that the new assay is better than any one clinicopathologic factor. But clinicians don’t look at tumors as one factor,” Dr. Schnitt says. “They look at them as a constellation.” The pertinent question, he says, is how Oncotype DX stacks up against a combination of traditional pathologic factors, as embodied in the Nottingham Prognostic Index, for instance.

David Hicks, MD, vice chair of the Department of Pathology and director of anatomic pathology at Roswell Park Cancer Institute, Buffalo, NY, agrees that Genomic Health was “wise and fortunate” to obtain for its studies access to the NSABP tissue banks, composed of node-negative, ER-positive patients all of whom had received the same treatment. Also wise was to have focused its studies on formalin-fixed paraffin-embedded clinical samples, making it possible to examine retrospective archival material. He finds it “most significant” that chemotherapy benefit was largely restricted to the group that had a high recurrence score on the Oncotype DX assay. But, Dr. Hicks asks, “At what point do we have confidence that this can be brought into the clinic and used to make treatment decisions about patients today using the chemotherapeutic regimens that are in place in the modern era?” He, too, says the best evidence would be a prospective trial in which patients are triaged based on a molecular test. “[Ongoing trials] will take 10 years,” Dr. Hicks says. “What do we do in the meantime?”

Some say the new tests should be used while waiting for definitive evidence. “Genomics is providing new information not being provided by existing tests,” asserts Charles M. Perou, PhD, associate professor in the Departments of Genetics and Pathology and a genetics researcher in the Lineberger Comprehensive Cancer Center at the University of North Carolina at Chapel Hill. Dr. Perou developed one genomic signature, the intrinsic subtype model. Asked why predictive assays that will be used to guide treatment should not be held to the highest standard, prospective data, Dr. Perou acknowledges that is “the hardest question to answer.

“I’m glad to see that genomic tests are being put into prospective trials,” he says. “However, the answer will take 10 years to get. Do you wait that long before you utilize the test? My personal bias is that, if retrospective data are well performed and look good, that is probably enough to begin utilizing a test and paying close attention to the results.”

Daniel Hayes, MD, clinical director of the breast oncology program at the University of Michigan Comprehensive Cancer Center, Ann Arbor, holds a similar view. “It is worthwhile to do Oncotype DX in regular practice,” he says, even before the results of prospective trials are available. (Dr. Hayes volunteers that he was a paid consultant to Genomic Health until three years ago.) “My sense as a clinician is that [this assay has] pretty compelling data with regard to showing treatment effect.”

“At this point there is a clear indication that the Oncotype DX assay and clinical information are two independent sources that contribute to prognosis,” says Soonmyung Paik, MD, director of the Division of Pathology at the NSABP Foundation Inc., who was lead investigator on two crucial trials of this assay. “So when you treat a patient, you should consider both.” Dr. Paik’s only cavil is that “Oncotype DX is too expensive, so we need to encourage other companies to enter the market and drive the price down.”

Not everyone has a formed opinion on the utility of multianalyte molecular assays. “I couldn’t be more on the fence,” says Michael R. Pins, MD, associate professor of pathology and urology at Northwestern University’s Feinberg School of Medicine, who directs the reference laboratory for TAILORx, the prospective trial evaluating Oncotype DX. “There are no unequivocal data comparing Oncotype DX to current clinicopathology practice to say whether patients need it or not. That’s the whole purpose behind this study,” he notes. “We might find that Oncotype DX doesn’t add as much as four or five immunohistochemistry [IHC] stains.”

In the past few years there has been “a deluge of prognostic gene expression profiles for breast cancer,” Dr. Perou says. In addition to the intrinsic subtype assay and Oncotype DX, other clinical contenders include a Dutch assay called MammaPrint (Agendia), which is the only FDA-cleared candidate; an IHC test called Mammostrat (Applied Genomics); a FISH assay (Exagen); a two-gene RT-PCR assay (Aviara); and several others. As Dr. Hayes puts it, “These new assays take into account multiple markers, like a pathologist judging grade takes into account many things. It’s just done with molecules instead of eyeballs.”

In a parallel comparison study, Dr. Perou found strong prognostic concordance among five such assays (Fan C, et al. N Engl J Med. 2006;355:560–569). “Our motivation for doing this study,” Dr. Perou says, “was that we and many other people wondered whether different gene expression profiles were giving different answers to a similar question. Of the ones we tested, we found that most give a very similar answer, which we interpret to mean they must be tracking a strong and common biology.”

It’s worthwhile going further into the rationale and clinical data for three of these assays: Oncotype DX, MammaPrint, and Mammostrat.

Oncotype DX, the 500-pound gorilla in the room, is a real-time RT-PCR assay that measures RNA expression profiles of 21 genes (16 cancer related and five reference genes) using paraffin-embedded tissues. The result is expressed as a recurrence score from 0 to 100. Testing is done in Genomic Health’s CLIA-regulated and CAP-accredited reference laboratory. In 2006, the laboratory performed about 14,400 tests, double the number in 2005. At the end of 2006, insurers covered the test for more than 100 million lives.

“In my opinion, Oncotype DX is the furthest along,” Dr. Hayes says. “It was developed with two goals.” The first was to make the assay feasible and logistically easy to do in formalin-fixed paraffin-embedded tissue. The second was to make the assay clinically relevant. “They sat with a group of clinicians and asked us, ‘What is the question you want answered?’” Dr. Hayes says. “We said that one of the biggest difficulties for us in breast cancer medicine is what to do with patients who are ER-positive and node-negative. We know they will get hormone therapy. Should they also get chemotherapy?”

According to Dr. Hayes, the answer to this question depends on the absolute degree of benefit—for example, the reduction in the percent of recurrence a patient will face over the ensuing 10 years. “If the absolute benefit is five percent or greater, most otherwise healthy patients and clinicians in the Western world will say yes to chemotherapy,” he explains. “If it is two percent or less, most will say it is not worth it—risk and benefit are too close together.” For intermediate benefits, he says, “most of us throw up our hands and say, I don’t know what to do.” Assuming that chemotherapy reduces the risk of recurrence by about 20 percent, it will decrease a 10 percent risk of recurrence by about two percent and a 20 percent risk of recurrence by about four percent. So a recurrence score that defines a recurrence risk of 10 percent following hormone therapy forms the upper boundary of the low-risk group, those who can consider not getting adjuvant chemotherapy. And a score that defines a recurrence risk of 20 percent forms the lower boundary of the high-risk group, those who should consider chemotherapy. In the NSABP studies, “low” risk was found to be defined by a score of 18 or less and “high” risk by a score of greater than or equal to 31.

Recurrence score was shown to correlate with prognosis, with about 50 percent of patients falling into the low-risk group and about 25 percent each in the high- and intermediate-risk segments (Paik S, et al. N Engl J Med. 2004;351:2817–2826). In a further analysis of the NSABP database, results suggested that Oncotype DX may be predictive of adjuvant chemotherapy benefit: Patients with a high recurrence score showed a 27.6 percent absolute decrease in 10-year distant recurrence incidence with chemotherapy, while those whose tumors had a low score had minimal, if any, benefit.

Oncotype DX addresses two major issues with traditional clinicopathology, according to Dr. Paik. First is reproducibility. In the 2004 article, three independent pathologists evaluated tumor grade. “Agreement among the three was only about 50 percent,” Dr. Paik says. “We could improve it to 80 percent with training, but that’s still not good enough. It is worse for tumor type. And,” he adds, “array tests are objective.”

Second is what he considers the fallacy of categorizing individual tumors into subsets. “We clearly showed that prognostic response to therapy is a continuous variable,” Dr. Paik says. “Oncotype DX gives an individual risk prediction rather than a category.” An Oncotype DX report provides a numerical score with confidence intervals and a graph showing the relation between score and risk of distant recurrence. “It would be unfortunate if physicians used the score as a category,” Dr. Paik says.

But Dr. Goldstein at William Beaumont maintains that grading is not as subjective as Dr. Paik’s data suggest. “Their kappa values for interobserver variability were substantially lower than what standard of practice in surgical pathology allows,” he says. “I’d get fired real fast if I had reproducibility scores like that.” In 2001, Dr. Goldstein evaluated variability among himself and three other general surgical pathologists. “For grade three, two, and one carcinomas, our kappa values were 0.82, 0.64, and 0.39, respectively,” he told CAP TODAY. “These levels of interobserver agreement are in general agreement with some published studies and are substantially higher than the levels reported by [Dr. Paik and his colleagues].”

Sunil Badve, MBBS, associate professor of pathology at Indiana University School of Medicine, Indianapolis, agrees that “variability in grade as published in that paper was overstated.” Dr. Badve and colleagues have submitted to the upcoming meeting of the American Society of Clinical Oncology data comparing recurrence score to tumor grade in 700 contemporary cases. While Dr. Badve could not disclose the findings, he did say, “Even in Dr. Paik’s paper, tumor grade correlated wonderfully with prognosis, in spite of disagreement among the pathologists. Grade still seems to be a very important parameter.”

Federico Monzon, MD, director of the clinical genomics facility at the University of Pittsburgh Medical Center, is “very excited to see these multiplex tests happen.

“They can help us in areas where single analyte testing cannot. One of the major hurdles we face in cancer is that we are very good at diagnosis but not so good at prognosis or predicting therapeutic response.” Like Dr. Schnitt, he considers the development process for Oncotype DX to be well done.

However, Dr. Monzon, too, says published trials have not compared Oncotype DX to the current standard of care in breast cancer—reporting with the Nottingham score. “They did compare it to single parameters like nuclear size or degree of differentiation,” he says, “but the Nottingham score takes into account nuclear grade, mitotic activity, and histological pattern together. So at this point I don’t think they have shown that the recurrence score outperforms clinically or morphologically based predictive algorithms.” Dr. Monzon also points out that, in the 2004 publication by Paik, et al, tumor grade showed very similar prognostic power to the recurrence score. In multivariate analysis, poorly differentiated tumors had a hazard ratio of 3.34, compared with 2.81 for the recurrence score. “By evaluating tumor grade with mitotic activity and histological pattern, I would want to know whether the Nottingham score is outperformed,” he says. “And that doesn’t include HER2 status.” Dr. Monzon adds, “I am not saying that the test doesn’t outperform current prognostic indicators. I just don’t know based on the published data.”

Many of the criticisms of the Oncotype DX assay were addressed by a retrospective population-based study conducted by Genomic Health and the Kaiser system of Northern California, says Steven Shak, MD, Genomic Health’s chief medical officer (Habel LA, et al. Breast Cancer Res. 2006;8:R25. Epub 2006 May 31). “This study is important because it was done on more recent tumor specimens, patients diagnosed in the 1990s,” says Dr. Shak. “Since it was population-based, it reflected the distribution of tumors seen now—31 percent of tumors were 1 cm or smaller. Our confidence that the NSABP results are generalizable to women diagnosed today is strongly enhanced by [this] study.” In the Kaiser analysis, an interaction was seen between recurrence score and tumor size. “These tests should not be a recipe for treatment,” Dr. Shak cautions. “They need to be used in conjunction with all other factors.”

A small but provocative study comparing Oncotype DX with IHC will be presented at the March USCAP meeting. “We noticed in published data that the Genomic Health assay appeared to look primarily at three major axes—the ability of the tumor to proliferate, whether tumors expressed HER2, and whether the ER axis was activated,” says Kenneth J. Bloom, MD, chief medical officer and medical director of Clarient Inc., Aliso Viejo, Calif. Weights for genes composing these three axes were 1, 0.5, and 0.3, respectively. “It struck us that we measure all those things by IHC,” Dr. Bloom continues. “What if we looked at KI67 [MIB1] as a measure of proliferation, HER2 expression, and PR and BCL2 for activation of the ER axis?”

Collaborating with pathologists at West Virginia University, Dr. Bloom blindly assessed tissue blocks from 28 patients who had previously had Oncotype DX performed. “All 14 tumors that had a low recurrence score had high expression of PR and low expression of KI67 and did not overexpress HER2,” Dr. Bloom relates. “So if you want to find people with a low recurrence score, it appears that you could do that with three immunostains. Those also correlate very tightly with tumor grade, and low-grade tumors almost always fall into the low-recurrence category.”

Dr. Bloom calls the four tumors with a high recurrence score “more interesting.” One had 3+ expression of HER2 while a second did not express HER2 but had a proliferation index of greater than 50 percent. “In the other two cases IHC data would have predicted a low recurrence score,” Dr. Bloom says. Both expressed PR and had a low proliferation index. His initial conclusion was that IHC found all tumors with a low recurrence score but may have missed two with high recurrence scores.

However, detailed analysis of the two discrepant tumors showed something more intriguing. A key factor in both tissue blocks was a cavity from the patient’s previous needle core biopsy. “So the block contained tumor, but also adjacent to the tumor itself we could see the needle track with inflammatory tissue including macrophages,” Dr. Bloom says. “In the biopsy track there was granulation tissue proliferating like crazy. So lots of cells were expressing KI67, but they were not tumor cells.” Dr. Bloom also notes that a minor component in the recurrence score is CD68 expression, a macrophage marker. “We saw numerous CD68-positive macrophages in the tumor block,” he says, “but no tumor cells expressing CD68.” Taken together, he believes these findings can explain the discrepancies between the recurrence score and IHC results. Both patients with discrepant results elected to receive chemotherapy, he told CAP TODAY, “even though without this test they would have received only hormonal therapy. There is no doubt in my mind that both of these patients were significantly overtreated.”

(Dr. Shak says the GH clinical lab pathology group routinely performs manual microdissection in cases that contain biopsy cavities to exclude the contribution of granulation tissue to the recurrence score. Dr. Bloom counters that these two cases are real-world examples in which the microdissection was apparently less than optimal.)

Dr. Bloom cautions that the number of cases he looked at was quite small. And he acknowledges that the dynamic range of IHC is more limited than that of molecular methods such as RT-PCR. However, he asks, “Do we need more dynamic range? That is not clear right now.” He also emphasizes that IHC gives the ability to localize proteins in the tissue and within cells, which may have as much, if not more, importance than dynamic range.

“GHI [Genomic Health] has a great test,” he concedes, “but I scratch my head and say, Don’t I have most of this information already? They have largely proved what we already knew.” Dr. Bloom has been encouraging pathologists to look at their own IHC data on cases that have been sent for Oncotype DX testing and to review the data with their tumor boards. “You will be surprised at the information you already had available to you,” he says.

MammaPrint is a microarray assay that measures expression of 70 genes in fresh frozen tissue; it is the only assay currently FDA cleared. In a recent study, the 70-gene signature was compared with clinicopathologic scoring by Adjuvant software in ER-positive and -negative patients. Hazard ratios were significantly better for the gene signature than for Adjuvant, the authors reported (Buyse M, et al. J Natl Cancer Inst. 2006;98:1183–1192). In a previous study, this group had shown that the 70-gene signature correctly identifies patients who needed adjuvant chemotherapy and those who did not, based on U.S. and European guidelines (van de Vijver MJ, et al. N Engl J Med. 2002;347:1999–2009).

MammaPrint is a prognostic test that is intended to provide “further refinement of risk assessment for recurrence of breast cancer for patients diagnosed with node-negative disease,” says Laura J. van ’t Veer, PhD, head of molecular pathology at the Netherlands Cancer Institute, Amsterdam. Results are valid irrespective of ER status, says Dr. van ’t Veer, who is chief operating officer of Agendia and is responsible for development of the original scientific finding and for shepherding the resulting test through the FDA. As a prognostic test, MammaPrint was cleared as a class two device on a 510(k) application. If they had introduced predictive value taking into account a treatment, the FDA would more likely see it as a class three device, Dr. van ’t Veer says. “FDA especially valued that patients had not received systemic adjuvant treatment, so we were looking at the true biology of the tumor,” she adds. Patients were diagnosed up until 1998; they were below age 61 and not eligible for hormonal treatment or cytotoxic chemotherapy according to European guidelines at that time. “Adjuvant chemotherapy for node-negative patients was introduced later in Europe than in the U.S.,” she says.

About 5,000 clinical MammaPrint tests have been done in Europe. Results are given either as high risk or low risk for recurrence. “We have argued that in clinical practice it is most relevant and most useful to have a bimodal classification,” Dr. van ’t Veer says. “I think the continuum for both [MammaPrint and Oncotype DX] at the present time is not exact enough.”

Dr. van ’t Veer rejects the criticism that FDA clearance only validates MammaPrint’s technical aspects. “MammaPrint has shown its clinical utility,” she asserts. “That is one of the things that FDA reviewed. They stated in their press release that it is a clinically useful tool for risk assessment.” The FDA’s release said Agendia’s studies “confirmed that the test was useful in predicting time to distant metastasis in women who are under age 61 and in the two earliest stages of the disease (Stage I and Stage II) and who have tumor size equal to or less than five centimeters and no evidence that the cancer has spread to nearby lymph nodes.”

Many U.S. pathologists express doubt that the need for fresh frozen tissue could be met here. “We have made it into a very easy system,” Dr. van ’t Veer replies. “It should not take much more effort to get a small piece of tumor in a way it can be preserved and then fresh frozen. We have not experienced any difficulty of using fresh tissue [in Europe].”

MammaPrint is not available now in the U.S., but Dr. van ’t Veer says “Agendia is working to expedite the launch of MammaPrint in the near future.”

Mammostrat is an IHC test that measures five markers. One of its developers, Douglas T. Ross, MD, PhD, chief scientific officer and co-founder of Applied Genomics, says, “Gene expression profiling is a fantastic tool for discovery, but for translating those discoveries to clinical medicine we don’t have to measure thousands or even a hundred genes.” Making an IHC assay allows the use of paraffin blocks, which are good for validation and which facilitate distribution to countries that do not have access to more sophisticated technology. And it is much less expensive than an array test.

In one validation study, Mammostrat distinguished ER-positive patients with poor outcomes with a hazard ratio of 1.9 to 2.2 (Ring BZ, et al. J Clin Oncol. 2006;24:3039–3047). What Dr. Ross calls the most important validation study was presented last December at the San Antonio Breast Cancer Symposium. It was performed on the tamoxifen arm of NSABP studies B14 and B20, the same data set used to develop and test Oncotype DX. Results on Mammostrat were significantly associated with 10-year recurrence-free interval independent of tumor size. In an exploratory multivariate analysis, age was a significant factor. “Mammostrat appears to be most useful in patients greater than 60 years old,” Dr. Ross suggests. It “very robustly” identified older patients with worse prognosis following tamoxifen.

Only a small fraction of older women who have ER-positive node-negative tumors are treated with chemotherapy. Studies have shown a fairly small benefit in this group. “The strength of our data lies in identifying poor-prognosis or high-risk elder patients,” Dr. Ross says. “Mammostrat can identify older patients whose outcome is worse than average and who have a greater chance of benefiting from chemotherapy.” Stratification by age was not a pre-specified endpoint of the trial, he adds, but says, “We think this analysis was important because of the differences in treatment approach in younger as opposed to elder patients.”

Applied Genomics has now partnered with the Molecular Profiling Institute in Phoenix to launch Mammostrat as a reference-lab–based IHC test in the second quarter of this year. Dr. Ross says reimbursement for the test will be “consistent with the current cost of five-antibody IHC special stains.” Applied Genomics has also filed a pre-IDE with the FDA to discuss a test kit. “We intend to bring a test kit to market,” he says.

Based on the San Antonio presentation, Dr. Hicks calls Applied Genomics’ assay “a very promising approach that could reach pretty broadly to many institutions and other countries.” Currently, these studies provide clinical utility as a prognostic assay, he says, and their approach has paralleled the process that Genomic Health has used. “Now they can say it separates patients into prognostic groups. They still have to look at patients treated with chemotherapy and identify subsets likely to benefit and those who will not benefit.”

Two prospective trials of multianalyte molecular assays are now underway: TAILORx (Trial Assigning IndividuaLized Options for tReatment), which evaluates Oncotype DX; and MINDACT (Microarray In Node negative Disease may Avoid ChemoTherapy), testing MammaPrint.

TAILORx will enroll 10,000 patients who have ER- or PR-positive, HER2-negative, node-negative breast cancer that is 1.0 to 5.0 cm or 0.5 to 1.0 cm with unfavorable features. Patients with a low-risk recurrence score will receive hormonal therapy of their doctor’s choice. Patients with a high-risk recurrence score will receive hormonal therapy plus chemotherapy of their doctor’s choice. Those with a mid-range recurrence score will be randomly assigned to hormonal therapy plus or minus chemotherapy.

TAILORx has more conservative cutpoints than the commercial assay, says Sheila Taube, PhD, associate director of the cancer diagnosis program at the National Cancer Institute, which is cosponsoring the trial along with the Eastern Cooperative Oncology Group, or ECOG. For the upper boundary of the low-risk category, the score has been lowered from 18 to 11, while the lower boundary of the high-risk category has been lowered from 31 to 25. Dr. Taube says the net result will be to reduce the number of women in the lowest-risk group and put more patients into the high- and intermediate-risk groups.

“When we were planning this trial, we got data on actual patterns of care,” Dr. Taube says. “It turns out that lots of women choose not to have chemotherapy without any adjunct test.” According to guidelines, about 90 percent of women with early disease are recommended to get chemotherapy, but in reality a significant percentage of them do not receive it. Perhaps results of TAILORx will allow these decisions to be based on evidence.

There is no official arm comparing Oncotype DX to clinicopathological parameters, Dr. Taube says, but comparison will be done as an ancillary study. Dr. Badve, who is a co-chair of the pathology subcommittee for TAILORx, says recurrence scores will first be correlated with Adjuvant online. Other data will be collected, as will tissue and blood, which will compose a tissue bank that can be used for further analysis as new techniques become available.

Northwestern University’s Dr. Pins is the director of the ECOG Pathology Coordinating Office and Reference Laboratory, or PCO-RL, which oversees preanalytic sample processing for the TAILORx trial. He calls it “a monster trial.” At the PCO-RL he will receive leftover material not used in doing the Oncotype DX assay and any unstained slides. They will perform four IHC stains—PR, ER, KI67, and HER2—even though some of these stains will have already been done by local laboratories. “We’ve been a little bit surprised by occasional patients called ER-positive who are not positive by strict criteria,” he says. Some samples have been called ER-positive on the basis of two percent or three percent of cells staining. It has become clear, he says, that the definition of an ER-positive tumor varies somewhat between local pathology laboratories. The PCO-RL will also generate tissue microarrays and extract DNA and mononuclear cells and circulating tumor cells from blood samples.

“We are not prospectively computing a Nottingham Index,” Dr. Pins says, “but the data will be there for anyone to use. We have access to digital imaging, so we could scan the slides and send them.”

The MINDACT trial, sponsored by the European Organization for Research and Treatment of Cancer and the Translational Research Breast International Group, will compare MammaPrint with Adjuvant online in 6,000 patients who have node-negative disease and T1, T2, or T3 tumors. Patients concordant for high risk by both assessment tools will receive hormonal therapy (if ER-positive) followed by adjuvant chemotherapy. Those concordant for low risk will receive no further systemic treatment or hormonal treatment only. “In discordant cases, which are anticipated to be around 35 percent of patients, further treatment will be randomly assigned according to either Adjuvant online or MammaPrint risk estimates,” Dr. van ’t Veer says. The rationale for this design and the power calculations have been published (Bogaerts J, et al. Nat Clin Pract Oncol. 2006;3:540–551). One goal of the study is to determine whether patients with a low-risk molecular prognosis and a high-risk clinical prognosis can avoid chemotherapy.

Whatever the results of these trials, changes are in the offing. “One of the things about the Oncotype DX paradigm is that pathologists don’t like tests done by one company in a proprietary manner,” Dr. Schnitt says. “I think that is a mindset that has to change. More tests will happen that way. It will not be feasible for individual institutions to do every test. The only alternative will be for a commercial laboratory to do them.”

Technical changes will also occur. “By the time these trials give their answers, there will be tests we haven’t even imagined,” Dr. Perou says.

Dr. Bloom is onboard with this vision. “Oncotype DX and MammaPrint are a huge step forward,” he says. “They are quantitative assays with good dynamic range that allow us to inquire about many things simultaneously. Unfortunately, they don’t give us the ability to localize. And for a pathologist the ability to localize is most important.”

Dr. Bloom foresees new technologies, such as quantum dots and flexaphores, that may make it possible to localize and quantitatively assess multiple proteins and genes simultaneously with an improved dynamic range. “They are in their infancy right now,” he says. But by the time final results from TAILORx and MINDACT are published, these new technologies may make microarrays obsolete in the clinical arena.