| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© Oxford University Press 2006.
NEWS |
Breast Cancer Gene Microarrays Pass Muster
Biology may trump anatomy when it comes to predicting a breast cancer patient's outcome, according to a new study of gene expression tests.
The research, published in the New England Journal of Medicine, tested five different gene expression profiles developed within the last six years—two are now commercially available—to answer a big question about these tests: Do independently developed gene screens agree or disagree in predicting outcome for an individual patient?
The research team, led by scientists from the University of North Carolina in Chapel Hill, found that four of the five tests predicted the survival outcomes in a 295-patient Dutch group. More surprisingly, these tests agreed with each other even though each tested a different set of genes (only one gene overlapped between the two commercially available tests). A second analysis found that each of the four gene expression tests performed better at predicting disease-free survival and overall survival than clinical variables.
Even though the tests are substantially different, each one tracked similar biological characteristics that are probably present across breast cancers. The researchers suggest that knowing the molecular biology of a breast tumor could potentially be more important than, and certainly as valuable as, traditional prognostic markers like tumor size, nodal involvement, and the handful of protein tests now in common use, including cell receptors for estrogen and progesterone hormones (ER and PR) and the HER2 membrane receptor.
"These genomic tests all need more validation, but I believe that in the long run they will provide important new information that cannot be provided by standard clinical tests, even when optimally run," said the lead investigator, Charles Perou, Ph.D., an assistant professor of genetics, pathology, and laboratory medicine at UNC.
"We are not trying to throw away node status, tumor size, and other criteria but trying to make outcome predictions more accurate, providing additional information that oncologists can use to make treatment decisions."
If the results are proven correct—some people have questions about the tests' accuracy—they are part of a trend toward a new multifaceted view of breast cancer. This new view contradicts the decades-old take on breast cancer that Harold Burstein, M.D., Ph.D., calls the "M&M theory."
"Even though there were surface differences in how breast cancer cells stained for these markers, we thought that on the inside, they were all the same," said Burstein, a breast cancer clinician and researcher at the Dana-Farber Cancer Institute in Boston. "Until recently, we didn't appreciate how much difference there is below the surface, and these gene expression microarray profiles show profound differences in hundreds of different genes between the subtypes."
Beyond ER, PR, and HER2
The shift in the last decade toward thinking about breast cancer as a heterogeneous disease composed of many different molecular subtypes has been dramatic, making breast cancers the most studied solid tumors in terms of their varied biology. Ironically, developing the targeted therapies trastuzumab (Herceptin) and tamoxifen instigated the molecular revolution because they targeted subclasses of breast cancer that were only later understood.
Currently, most oncologists treat breast cancer patients based on the tumor's ER, PR, and HER2 status; cancer biopsy samples are routinely tested for these markers. nearly all of the estimated 212,000 new invasive breast cancer cases diagnosed each year are ER+— they are fueled, in part, by female hormones. Within that class are two types of cancer. One, believed to come from luminal epithelial cells, often responds to hormonal treatments alone. A second group, by contrast, often needs a second punch from chemotherapy. Perou was first to dub these subtypes as "luminal A" and "luminal B," respectively, based on their distinct genetic patterns.
The remaining cancers are ER–, and this category is split fairly evenly between subtypes that are positive for the HER2 receptors and basal-like breast cancers that can be negative for all three receptors (i.e., ER, PR, and HER2, also called triple negative).
This rough analysis can help direct targeted therapies to some subtypes—such as tamoxifen or newer classes of antihormonal therapy to ER+ and trastuzumab to HER2+. But picking out the cancers that will not respond to those therapies is not specific enough. It does not single out, for example, the substantial percentage of women who cannot benefit from tamoxifen because they have two copies of a variant gene, or the 50% of HER2 tumors that do not respond to trastuzumab.
With more refinement, gene expression profiles can tell you ahead of time how to treat the cancer, said Lisa Carey, an oncologist also at UNC. This knowledge is especially important for those cancer subtypes that now have no individualized therapy, such as basal-like tumors.
|
"It's a question of making these tests even better," said Carey, who recently reported for the first time that the basal-like subtype is more prevalent among premenopausal African American women than other groups. "We have done the easy stuff, and now it gets harder."
Digging Into Details
The five gene expression tests studied by Perou attempt to go beyond simple receptor status by zeroing in on the genes expressed in the different molecular subtypes (see box). The tests all exploit the ability of messenger RNA (mRNA) from biopsied tissue to bind to specific DNA arrays, revealing the active genes. A computer can measure the amount of mRNA stuck on the microarray to produce a quantitative profile of gene expression.
Although developers of these microarrays believe many of these genes drive growth and metastasis, they do not yet know the function of most of them, said Perou, whose own test was one of the five studied.
For all 295 patients in Perou's study, the currently available clinical markers—ER status, tumor grade and diameter, and nodal status—predicted relapse-free survival and overall survival. The researchers then evaluated the prognostic value of each gene expression test individually, as well as each test combined with the clinical variables. They found that all the gene tests except for the two-gene assay offered stronger predictive power than the clinical data.
"When we did the statistical analysis, we got a score on how much information each clinical variable along with the gene test variable was contributing to predicting outcome, and we found that the four gene expression models got the highest score," Perou said. "That can be interpreted to mean that these gene tests were the strongest predictors of relapse free survival and of overall survival."
They then compared the results of the Oncotype DX, MammaPrint, wound-response, and two-gene models with each other. They found that all were highly correlated with each other, though the two-gene assay was the least correlated. A comparison between the two commercially available tests showed that they predicted the same outcome in 77% of patients with ER+ cancer and 81% percent of all tumors.
Perou recognizes the shortcomings in his study, including that standard tests for PR and HER2 status were not done on these 295 patients so the researchers could not say exactly how accurate each gene assay was compared to the current testing regimen. He added that his failure to validate the two-gene predictor might be due to a technical failure because that assay had been successfully tested in other studies totaling about 1,000 patients.
| Testing Breast Cancer Genes Researchers at the University of North Carolina in Chapel Hill say that genetic expression tests can predict a patient's outcome better than currently available tests. Here are the five genetic expression assays they examined:
|
Other researchers have pointed out that the tests may be marred because some of the 295 tumor samples studied were also originally used to find predictive genes in three of the assays (MammaPrint, wound-healing, intrinsic subtypes profiles). But Perou said the gene choices for two of the three didn't come from this sample set, and the research team conducted analyses to evaluate biases and believe that did not disrupt the findings. "Further research is needed, but the whole point is that these data are promising but not definitive and need further validation," he said.
Larry Norton, M.D., agrees, saying that study shows that these tests can "separate patients into prognostic categories even though there is little overlap in the actual genes being used." He likens this concept to separating athletes into winners or losers based on tests of physical ability. One test of strength and another of speed "work equally well in separating the winners from the losers, but there may be no overlap between ‘strong’ and ‘fast’," said Norton, deputy physician in chief for breast cancer programs at Memorial Sloan-Kettering Cancer Center. "All of these tests are evolving rapidly and are certainly influencing how we think about cancer and how we will conduct clinical trials."
Carey said it is reassuring that the tests come to the same conclusion. "The paper tells us that both the lab scientists working on figuring out what creates or drives different kinds of breast cancer, and the clinically oriented folks looking for an assay to help us decide on treatment, are going in the same direction."
Wait and See, or What?
The literature on the breast cancer molecular assays is blossoming, but none of the tests—including the two commercially available ones—have been endorsed by major oncology societies or the FDA. However, oncologists seem to be separating themselves into two groups, Burstein said. "There is the early-adapters group and the wait-and-see group that would like to see years of validation research first."
Some high-end cancer centers are starting to use the Oncotype DX test routinely to decide which ER+ women will receive chemotherapy. Some insurance companies now cover the $3,500 cost as part of a pathology workup.
At Memorial Sloan-Kettering, Mark Robson, M.D., uses the assay to help women with small ER+ and lymph node–negative tumors—those who may have a favorable prognosis with tamoxifen treatment alone—decide whether they should also receive chemotherapy.
"It appears that, in this specific circumstance, the test provides incremental information beyond that derived from traditional factors, and I use it in the clinic quite often for just this reason," said Robson, clinic director of the clinical genetics service.
Burstein agrees that "this kind of test can stiffen the spine of clinicians" whose pathology reports are suggesting minimal therapy. "There is a different psychological twist when you throw these scientific tools into your decision making."
Gene expression profiles can also offer reassurance to physicians who know from clinical studies that routine receptor tests can disagree with each other, depending on whether they are conducted in-house or sent to a central laboratory, researchers say.
But others argue these gene expression tests, which have been dubbed "home brews" and do not need federal approval, are still a work in progress and need rigorous validation studies. Currently, the tests don't offer much more information than what can be obtained through the standard care, Joyce O'Shaughnessy, M.D., argued in an editorial that accompanied Perou's study.
At this point, it isn't clear that looking at the expression of potentially hundreds of genes provides much needed information about whether an individual's cancer will spread or how well it will respond to therapy beyond what is available from a sophisticated reading of the tests now available, said O'Shaughnessy, of the Baylor Sammons Cancer Center in Dallas.
"Many clinicians are cautious, and that is how it should be," Perou said. "But I am optimistic that within a decade, some version of these tests will be used on every breast cancer patient."
Correspondence about this Article
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2007 99: 572-573.
This article has been cited by other articles:
|
|
 |
|
  S. Shak, G. Palmer, and J. Baker Re: Breast Cancer Gene Microarrays Pass Muster J Natl Cancer Inst, April 4, 2007; 99(7): 572 - 573. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||