IMPAKT Breast Cancer Conference 2013
Update on the METABRIC consortium 2000 Breast Cancer Study
Professor Carlos Caldas - University of Cambridge, UK
Good to have you with us at ecancer.tv. You’ve actually developed techniques for classifying breast cancer; tell me all about the work you did some time ago and what this has come to now you’ve recently published.
Yes. Basically what we have identified is that by doing combined profiling of DNA and RNA in a large tumour set, about 2,000 tumours with complete clinical information, that breast cancer is not one disease, it’s really ten diseases. Current classification separates breast cancer roughly into four diseases: ER positive, HER2 negative; HER2 positive disease that can be either ER positive or negative and the so-called triple negative breast cancers. Intrinsic classifier, luminal A, luminal B etc. is really very similar to the immunohistochemical based classification. What we have shown is that in particular in ER positive disease there are at least six or seven different subtypes that are completely separate diseases because they are driven by different molecular mechanisms and different mutations. We also identified the largest subset of breast cancer, about 15%, are tumours that have very little in the way of copy number aberrations in their genome and in these tumours there is a subset that is particularly enriched for lymphocytic infiltration and that might explain why they have better prognosis. So I think this new classification of breast cancer has profound clinical implications in terms of stratifying patients, prognostication and eventually deciding on targeted therapies.
The old classifications, of course, were fairly easy, well they were well-established, certainly, and most cancer doctors could get their heads around them. What about these new ones, what factors are you adding and can you make it clear to us?
I think, as I said or hinted at, it’s what we call a driver based classification. In other words, drivers of the different subtypes have been identified by us and what we’re now trying to do, and in the next couple of years will be developing, is a test that can be done using paraffin-embedded material to classify breast cancers into one of these ten subtypes. Importantly, in some subtypes the prognosis is so good that women might be spared unnecessary adjuvant toxic therapy and in other cases we have identified very aggressive subtypes of disease for which clearly we need novel therapies.
Are there any names that you could name of these subtypes that would help?
We just boringly named them clusters 1-10. Cluster 10 is the one that contains most of the basal-like breast cancers; cluster 5 is the one that contains most of the HER2 positive tumours and then the other clusters are the subtypes of ER positive disease and cluster 4 is the one that has very few copy number aberrations. So I’m afraid they have quite boring names of 1-10.
But you’re doing all the hard work for us; you’re doing all the thinking. You’ve got the ten subtypes of breast cancer so it should be possible for the workers at the coalface to actually plug into your system and come out with one of the ten groups.
That’s absolutely correct, that’s exactly what we’re striving to do now. I hope that in the next couple of years we will come up with a test which we would like to be on a price range that is attractive to routine histopathology laboratories of the order of €100-200 and that uses paraffin embedded material to do it so that it can be widely applicable.
When that happens, how much of an impact clinically do you predict that this will make?
I think that it will have a major impact because, as I said, the ten subtypes are not just different diseases and therefore they have different drivers and they might require different therapies, but they have a completely different pattern of metastasis or a significantly different pattern of metastasis. They are very distinct clinical outcomes and so I think this classification will eventually take hold.
So do you think you could dial in a classification, having tested it, and then prescribe a line of treatment?
Well, I think eventually that’s the ultimate aim.
What about the role of things like a neoadjuvant therapy to find out, is that going to be in the past?
That’s a very good question and we actually have two very large phase III randomised clinical trials in neoadjuvant therapy that we’ve just completed in the UK. My group is in the fortunate position of leading the translational research for those two trials and we’re doing exactly what you are alluding to. We’re classifying the tumours in those trials into the ten subtypes and looking for differential response.
So on the one hand you can have your test which will give you a pretty good idea just by feeding in a sample at the beginning but also as a backstop you can also do neoadjuvant therapy to make some experiments?
Exactly, what you can do is the first question to ask is are there indications that the different ten subtypes will have different responses to neoadjuvant therapy and the answer is absolutely yes, we have preliminary evidence of that. We have not published that yet so I’m going to just hint at the promise on the early results that we have and we hope to be reporting that within the next 12-18 months.
Are you getting any clues as to new targets, whether they’re molecules or pathways?
We are and some of them are very interesting because they are targetable genes or gene products because they encode for enzymes. Particularly I will give you two examples: one is in chromatin modifiers, so these are enzymes that modify chromatin and so they are amenable to inhibition with small molecules. The second class is proteins that are evolved in what’s called ubiquitylation which is protein degradation and it’s increasingly being recognised that these proteins are also involved in cancer and because they are enzymes, again, you could devise strategies to inhibit them.
Are any of these going into the pipeline at the moment?
There are some that are being tested and again I would just stay tuned because I think we’re going to see things coming out in the next couple of years.
Do you see that this is going to make a big difference to survival, then, and cancer…?
I am not one that likes to make broad claims about things. I think that this is a very important step in us understanding breast cancer as more complex than what we have portrayed it until now. I think the recognition that there are these different subtypes will be a first step towards stratifying women into different risk groups and I think that there will be benefits on both sides of the equation. On one side identifying groups of women that have such good prognosis that you might spare them therapies that they are receiving now without getting any benefit because they don’t need it. And on the other, identifying subsets that are failed by all current therapies and very early on tell those women the current therapies will fail you based on this test and you should go for experimental therapies from the beginning.
To what extent is this predictive or prognostic information?
I think it’s both and I think that there is a lot that is being made about being predictive and prognostic; I think there is a lot of semantic emphasis in this. My opinion is that a prognostic test can also be predictive because if you have a prognostic test that tells you that a group has a very good prognosis, any therapy with a similar hazard ratio of benefit will have an absolute benefit that is very small in that group whereas in a group with very bad prognosis with the same hazard ratio the absolute benefit will be much bigger. So the argument is, in part, semantic. I would also say that most of the so-called predictive tests, they are very good negative predictive value tests. In other words, if you are ER negative you will not benefit from hormone therapy, if you are HER2 negative you will not benefit from trastuzumab, but they are very bad at positive predictive value.
Now it’s all very exciting. Let me fast forward if I may, with your colleagues, Sarah Jane Dawson and Dana Tsui, you’ve been actually developing a new test for looking for fragments of cancer cells or DNA in the bloodstream. This is looking quite exciting, isn’t it?
Yes, we are very excited about this. This is work developed at our institute in Cambridge and it’s a team effort with three principle investigators involved: myself, Nitzan Rosenfeld and James Brenton. In the past twelve months we’ve published three important papers in Science Translational Medicine, The New England Journal of Medicine and Nature coming out actually as a hard copy today; it was published online three weeks ago at AACR. What this shows is that fragments, as you say, of DNA that originate in cancer cells that are dying or have been damaged by chemotherapy find their way into the bloodstream as what we call cell-free plasma DNA and this tumour-specific cell-free plasma DNA can then be used as a barcode test, as a liquid biopsy if you want, to do several things. It can be used to monitor response to treatments; it can be used as a liquid biopsy, even to identify mutations if it’s very difficult to do a biopsy of a metastasis, for example; for targeted therapy and finally in the paper coming out as a hard copy in Nature today, we have shown that by sequencing the whole exome in plasma DNA you can actually identify the full complement of mutations in a tumour and identify mutations, for example, that are associated with drug resistance. So this is very exciting because doing biopsies in the metastatic setting is ethically questionable in some patients, it’s expensive, it’s logistically very difficult to do whereas a blood test is very attractive and appears that this can be a true alternative to a hard biopsy, what we are calling now a liquid biopsy.
Could you explain to me, though, why it’s better to get the fragments, the DNA, rather than, for instance, whole cells and antigens? These tests have been available but they’re not as sensitive, are they? The fragments are almost 100% sensitive.
Well that’s what we showed in the New England paper is we directly compare circulating tumour DNA with circulating tumour cells. Actually the median value showed us that there are 133-fold more units of circulating tumour DNA than circulating tumour cells. So even with significant improvements in circulating tumour cell technology, we think that this will always be a cheaper and more generally available test. Now I am not going to say that circulating tumour cell research and potential clinical application will not continue because I think that there are things to be learned, particularly if we could isolate circulating tumour cells as live cells and then manipulate them in a laboratory. So I would say they are complimentary but you are right, our data shows that circulating tumour DNA is more sensitive, is of course specific because the barcode tells you the mutations that are present in the tumour and, moreover, in the New England paper we showed some very exciting potential which is that in about half of the patients that progress, circulating tumour DNA was going up on average five months before you could see progression by RECIST criteria on a CT scan.
You save a lot of unnecessary therapy.
Yes. And you’ve actually found now, as you’ve just told me, that you are identifying some of the genetic abnormalities and giving you some steer in the direction of future therapies.
That’s correct. So, for example, you can identify mutations in EGFR that could be targeted by gefitinib or erlotinib that you would not have identified because they were not present in the primary tumour and a distant metastasis in a location where a CT scan is impossible and you can find that mutation in plasma. That would have immediate impact for that patient because that patient would be a candidate for targeted therapy.
And in The New England Journal Marc Lippman thought that this could be a way of monitoring for recurrence, so not just using it in metastatic disease but in primary breast cancer to watch out for the return; what do you think of that?
Well I think there is great potential there. I think that I would wait for the data; I think that our groups in Cambridge, and I know several other groups at Johns Hopkins and in other parts of the world, are very actively pursuing this area of research. There is a lot of promise but let’s wait for the data.
Certainly Cambridge is clearly a powerhouse for all of this. There seems to be huge progress being made in breast cancer. How would you describe this to the average cancer doctor, though? Is there something around the corner? Is this a big progress or is it a small amount of progress that is still worthwhile?
What I would say to the regular cancer doc, of which I am one too, I am a medical oncologist and I still see patients every week, is that multidisciplinary care has made a huge difference in managing breast cancer patients, that applying state of the art modern therapies the survival of breast cancer at ten years is over 80% in our unit and I think that in the best units in Europe that’s what the survival is and that’s very encouraging. But on the other hand, we are failing about 20% of women and we need to identify better therapies. And of the 80% that are alive at ten years I am certain that we are grossly over-treating a large percentage of them and so identifying these women that could be spared toxic therapies is another research priority.
Do you think over-treatment is one of the biggest priorities rather than not targeted treatment?
I think that they are both priorities, I would say that they are equal priorities.
Thank you very much.
You’re very welcome.