Thank you. I’m a clinician scientist pathologist from Vancouver. On behalf of my colleagues and my collaborators in the Danish Breast Cancer Group I’m pleased to present these results. Here are my disclosures.
The intrinsic subtypes of breast cancer is something that we’ve recognised over the past decade plus by molecular tests. These are the main subgroups that are not necessarily readily identified just by looking under the microscope. A lot of the older trials didn’t think about the subtype analysis as it hadn’t been figured out at that time. There’s a large body of data that’s built up now suggesting, for example, the luminal A subtype has a much better prognosis than the other subtypes and that the different subtypes may respond differently to particular types of chemotherapy. But the level of evidence behind these predictive questions is often from in vitro studies, animal studies, cohort studies. What we’ve really needed is randomised clinical trials to look at these findings, particularly with a really stringent statistical analysis like an interaction test for chemotherapy versus no chemotherapy and subtype. But it’s been very hard to do these studies because basically the question was settled back in the 1980s that adjuvant chemotherapy helps women with breast cancer. Since that time it’s been unethical to randomise women in most settings to no chemotherapy, most of the studies have been different types of chemotherapy as choices. So on these older trials unfortunately material wasn’t saved or in the few cases where it was it had been used up for other studies and there’s not enough cases available to make these distinctions. We were fortunate enough, after a lot of looking, to identify a clinical trial that did have material remaining where we could go back and address whether subtype was predictive for chemotherapy. This is a Danish trial that actually started way back in1977 and finished accruing patients by the mid-80s. This was a trial of over a thousand premenopausal women and they were high risk women, certainly by today’s standards and standards for a group you might consider not giving chemotherapy to. Almost all of them were node positive, otherwise they were large tumours or invasive into the chest wall. Women were taken regardless of hormone receptor or HER2 status into this trial and they did not receive endocrine therapy. They did, however, get what would be considered more than adequate local therapy by today’s standards – mastectomy supported by axillary node dissection and chest wall radiation. Then they were randomised to four different arms, two arms receiving no chemotherapy and two arms receiving chemotherapy. The chemotherapy was either a classic CMF or what was tested at that time was cyclophosphamide as a single agent at a higher dose. Both of those two chemotherapy arms had similar results and they showed that chemotherapy works. They were one of several trials at that time that demonstrated a clear advantage of giving chemotherapy in the adjuvant setting, in this case the hazard ratio was about 0.6 favouring chemotherapy over control arms.
So this material let us address the question that we’d really been wondering about, I think, and address it very well and thoroughly, is do luminal A tumours, the low risk intrinsic subtype, derive any benefit from chemotherapy. We could look at this with an interaction test between luminal A status and chemotherapy for the trial’s primary endpoint, which was ten year invasive disease free survival. So for these purposes we were able to combine the two chemotherapy arms and the control arm with an ineffective agent, levamisole, as a no chemotherapy arm in a formal retrospective prospective study. So that means that we had to lock down everything in advance, what the definitions of luminal A would be, how we would determine that, what the statistical plan would be, primary endpoint, much like a fully prospective clinical trial. My team in Vancouver, we did the pathology, we did the subtype calls but we had no access to the clinical data, that was all handled in Denmark by the central statistical office according to the pre-specified plan. So a strong study design, I would say. The material we were constrained with was tissue microarrays that had been built from the material from this clinical trial. That meant we could address this by immunohistochemistry.
My group, as well as others, have published extensively on how to identify these subtypes by immunohistochemistry and with six biomarkers, as we published in 2013, there is a locked down definition of these subtypes. But for the purpose of this study the main issue was are you luminal A or not luminal A. So luminal A is defined as being ER positive HER2 negative, also low proliferation rate, as measured in this situation by Ki-67 index being below 13%, and then high progesterone receptor expression which was a marker that the whole ER pathway is active in and critical to the biology of this tumour.
Here’s the consort diagram. The short version here is there were 709 of the original 1100 patients available on the tissue microarrays and we got results for 633, a little over a quarter of whom were luminal A patients. This group of women in the study set very much mirrored the original trial and again this is a trial of younger women with large tumours, almost all node positive, high grade invasive ductal carcinomas. That’s a group where the chemotherapy works and our study set certainly reflected the results of the whole trial that chemotherapy works. Where it works, however, is in the non-luminal A patients and here is the results, more or less expected results, in non-luminal A patients. The dotted line here is the patients on chemotherapy and the solid line is off chemotherapy and you can see a very large difference, hazard ratio of 0.5. The key result of our study is this: the luminal A patients having completely overlapping curves where although their prognosis is better, because the whole curve is higher, there is no advantage of being given chemotherapy. The hazard ratio is 1.0 and the formal statistical test, the interaction test, is significant. So there is a difference and this is predictive.
The main limitations of the study are that we used an immunohistochemical method which has issues around its own analytical validity and I’m presenting a poster, P10101, today on Ki-67 testing. We were using only the tissue microarrays, Currently we’re trying to get the original blocks and isolating RNA for gene profile analysis is another way to identify the subtypes. This is an older trial and so its direct relevance to today’s population has to be discussed; for example these women did not get anti-HER2 therapy, anthracyclines or taxanes. They are a high risk group and they received no endocrine therapy and you might expect that this would be a particularly difficult set of patients to find a group lacking benefit from chemotherapy, nonetheless we did find that group.
So the main result is that women with this common luminal A subtype get no benefit from chemotherapy even if node positive, even in the absence of hormonal therapy. I’d just like to thank actually the women from Denmark who many decades ago were part of this study because here we are decades later and they’re still informing us today on how to make the best decisions for managing breast cancer. Thank you.