Newsletter #80–Feb./Mar. 2004
Impressions from the 26th San Antonio Breast Cancer Symposium
by Musa Mayer
If you asked the advocates who were there what they thought of the 26th San Antonio Breast Cancer Symposium, you’d have gotten a decidedly mixed response. Some felt excited by and hopeful about what they were hearing, while for others there was little of genuine interest, and a growing disappointment that the promises of genomic research have produced so little that is useful so far. For some, the pace of emerging research was breathtaking; for others, it was vanishingly incremental and increasingly distorted by industry influences.
Yet we all watched the same 44 featured presentations on nine huge screens arrayed throughout the cavernous ballroom of the Gonzalez Convention Center. Over the four days of the conference, 120 of us advocates maneuvered among 6,000 researchers, scientists, clinicians and industry people from 80 countries. We all balanced plates of food as we peered at the poster displays detailing over 500 studies that had not been selected for feature presentation. While we tried to make sense of it all, the sheer size and diversity of research presented guaranteed that the sense we made would reflect our own expectations and interests as much as it did the research itself.
In the following report on the meeting, the names and numbers in brackets refer to abstracts available at www.sabcs.org (click on “Abstracts Online 2003,” log in as a guest, and search by author).
Let’s get real: Genomic models for prognosis, prediction, and treatment
Since the sequencing of the human genome, great excitement—and even greater hype—have accompanied the genomic revolution in cancer diagnosis and treatment. What became clear to me in San Antonio this year was just how premature any results and conclusions are, at this point in time.
After the successful development and approval of Herceptin® over five years ago, it seemed as if we had entered a brave new world of precisely targeted, less toxic, more effective combined therapies. However, developing a cocktail of targeted new drugs to switch off the offending proteins that are involved in cancer has so far proven an overly simplistic model. A major stumbling block is that while we know that different tumors have different genetic signatures, we don’t yet know how to identify these tumors in any individual woman, or to determine what a particular profile may mean in terms of prognosis or treatment. With the exception of HER2 and ER/PgR testing, scientists haven’t managed to identify which cancers are most likely to respond to targeted therapies. Giving so-called targeted therapies to all patients in hopes of benefiting a small fraction raises serious questions about toxicity, resources, research priorities, and ethics.
Faced with initial failures of their new, targeted therapies, some companies are now struggling to identify subgroups of patients for which their drugs may prove effective. One example is Biomira, manufacturer of the vaccine Theratope®, which announced the disappointing results of its Phase III trials in metastatic breast cancer patients last summer. In San Antonio, the company presented a subset analysis that, while not conclusive in itself, offers a possible direction for new trials. Patients with ER+ tumors receiving hormonal treatments while undergoing Theratope showed a trend (not a statistically significant difference) toward better time to disease progression and overall survival [Miles 36].
Determining which patients are most likely to respond can transform a drug failure into a success. The monoclonal antibody Herceptin would never have shown a benefit in breast cancer, Genentech researchers say, had they not been able to develop a test that predicted Herceptin response. To date, however, Genentech has been unable to devise such a test for its new drug Avastin®, which targets VEGF, a gene responsible for angiogenesis (the growth of blood vessels that feed tumors).
Straightforward tests for gene expression are not always useful, it seems. Many if not most of the new agents don’t show enough activity by themselves, at least not in a general patient population. So, for rational combinations to be designed, much more must be learned about the complex signaling pathways within cancer cells, crosstalk between genes, and messages from outside the cancer cell that govern cell proliferation and cell death (apoptosis).
Although it is now “hot,” the genomic research paradigm is not the only avenue worthy of exploration. Which cancer cells researchers target may be important, as Max Wicha, of the University of Michigan, suggested in a plenary talk in San Antonio [Wicha P1]. His research elegantly demonstrates that tumor stem cells, the original cells from which all other cancer cells differentiate, may turn out to be of crucial importance in explaining drug resistance and the incurability of many cancers. Pointing to the notable success in curing metastatic testicular cancer, Wicha suggested that therapies designed to target only stem cells may be much more effective.
Mining the Past to Predict the Future: Is It Possible?
With the introduction of microarray and related technologies a few years ago, where the entire genetic fingerprint of a cancer (or the most relevant genes) could be analyzed on a single chip, the door seemed to swing wide to precise individualization of treatment. A single test would have the power to predict the risk of recurrence, and then to tell a woman with breast cancer exactly which treatments she needed—or whether she needed no further treatment after surgery, an even more powerful revelation for the majority of newly diagnosed women whose breast cancers are actually cured by surgery alone. The concept was mind-boggling. Imagine no more overtreatment or undertreatment. No more one-size-fits-all. And for many if not most of us, there would be the certainty of knowing we were cured. Such were the hopes. So far, the possibilities have proven to be just that—possibilities.
The study poised to be this year’s breakthrough news involved the use of the multi-gene RT-PCR test to predict recurrence, a new 21-gene microarray developed by Genomic Health. This test is derived from a combination of genes known to govern primarily cell proliferation and hormonal regulation in tumor cells. The revolutionary premise here is that this test is done on standard diagnostic pathology specimens, namely, archived, paraffin-embedded tumor tissue. If this is proven useful, researchers will be able for the first time to correlate archived tumor tissue with known patient outcomes from studies initiated and completed years ago, avoiding lengthy and costly “prospective” clinical trials.
The validation study for RT-PCR was done through the National Surgical Adjuvant Breast and Bowel Project, one of the cancer cooperative groups responsible for many important clinical trials. Researchers hoped to predict who was most and least likely to have a recurrence, based on tumor tissue alone. Using samples of the tissue of 668 node-negative breast cancer patients with ER+ tumors who were treated with tamoxifen during the 1980s, the study authors said that they were able to predict outcome better than any single prognostic factor, apart from tumor grade. The “low risk” group they identified had a risk of recurrence of 6.8 percent at 10 years, while the “high risk” group had a recurrence rate of 30.5 percent [Paik 16].
Genomic Health, who will be marketing this test as Oncotype DX early in 2004 for a minimum of $3,000, claims that it will help women with node-negative ER+ tumors taking tamoxifen (as most such patients do) select treatment more wisely. But even if the test is confirmed in other studies, how much will it really help the individual woman? How will the results be confounded if Arimidex® is widely accepted as standard of care, rather than tamoxifen? Is a 6.8 percent risk of recurrence low enough for a woman to feel comfortable refusing tamoxifen? Wouldn’t a woman in the highest risk group be considering chemotherapy anyway? Today, oncologists combine factors when considering prognosis, including tumor size, involved lymph nodes, tumor grade, etc. Several complex algorithms that feed recent data from clinical trials into a computer program are freely available to assist in this task, such as Adjuvant at www.adjuvantonline.com. So, it’s unclear what this test will add to the picture, even if its results are widely confirmed.
And speaking of confirmation, another study presented at the conference, using the same test in a similar population, some of whom who did not receive tamoxifen, was conducted at M.D. Anderson Cancer Center. This study failed to show any predictive value for recurrence [Esteva 17]. So, much more work is clearly needed. A fundamental question to be asked here, according to statistician Donald Berry of the M.D. Anderson Cancer Center, is whether these incredibly complex analyses of genetic expression are likely to ever provide reliable, reproducible results, given the multiplicity of potential data points, the difficulties with uniform data collection and pathology, and a variety of other statistical and methodological issues that make interpretation of tests like this fraught with problems.
Musa Mayer is a 14-year survivor and the author of four books including, most recently, After Breast Cancer: Answers to the Questions You’re Afraid to Ask (O’Reilly, 2003). She provides information and support for women with metastatic breast cancer daily at www.bclist.org and www.bcmets.org.
Additional coverage of the 26th San Antonio Breast Cancer Symposium by Musa Mayer…