SABCS 2012

Predicting late distant metastases in ER /HER2- breast cancer patients

Dr Peter Dubsky – Medical University of Vienna, Austria



Late metastasis in breast cancer is something that we’ve been talking about at this meeting and many other meetings for probably five to ten years now. It’s really a phenomenon that is quite common in ER /HER2- breast cancer, so the largest cohort of our breast cancer patients. What it really means to patients is it’s sort of the Damocles sword of these women that within the first five years they have a very good prognosis but then when you look at the second quinquennium and after fifteen years you still see recurrences. They’re not all the time, it is clearly below 2% type of recurrence rate per year that we’re seeing, but they don’t seem to decrease. So this persistent risk is something that we need to address and we have addressed. Obviously you have these large phase III randomised trials, the best known one is MA17 which looked at the extended letrozole after Tamoxifen, but even a small country such as Austria have done their own extended trial which was ABCSG6A and looked at how would patients do with a further three years of anastrozole after either Tamoxifen or an aromatase inhibitor in the first five years.

All over, when we look at that trial data what we see is that there’s sort of a 40% decrease in risk, which is great. Maybe we are sometimes able to delineate some subgroups such as the node positive subgroups which seem to benefit a little bit more. But all over there is a feeling we should not be giving every ER woman ten years, or even more, of extended endocrine therapy. So this is really an unmet medical need and the way that we try to address it is can we possibly use gene expression data to delineate, to predict late metastasis, to see who is more at risk than other women and could possibly these women have more benefit from the treatment? Possibly the way that we’d really like to address it, which is much simpler, is could we find a subgroup that had such a great prognosis, even at ten years after diagnosis, that we could say, “Your risk of late recurrence is so low that we do not recommend extended therapy.”

This is where our work with EndoPredict came in. The Austrian Breast and Colorectal Cancer Study Group has been working very close to Sividon for probably about five to seven years now and in our phase III randomised trial we validated the EndoPredict gene expression test and that was a very successful collaboration which has now long been published. We used that same gene expression data, that same expression test, that signature, to see if it works so well within the first five years. Could we actually do the analysis again and really look at those late years? This is what has been presented at this particular meeting and it’s quite extraordinary because when we use this signature which is a signature, you could very much compare it to the genomic health or the Agendia or the BCI signature. So it’s just another multi-gene signature but what’s different is that it had really been trained in a very, very large cohort of women and they were node positive and node negative and it had not been trained simply on the first five years of recurrences but on the entire follow-up that they had. So maybe not that surprisingly, when we used a very good biomarker sample of ER women we could delineate a group of women, which is actually a very large group of women, it’s 64%, and 98% of these women are free of late distant metastasis. So this is early data, obviously you’d have to use the same test again, go into a second biomarker sample, validate it again to really put this to clinical use but what we’re aiming at is really being able to specify a group of women and say, “Your prognosis is so good that weighed against the side effects you’re going to have with an aromatase inhibitor for ten years, obviously, it’s probably not worth taking that pill.”

Do you combine this with clinical parameters?

The EndoPredict, by itself, is an algorithm of eight genes and three reference genes. When you use that by itself it is clearly independent information by itself but when you look at the multi-variant models invariably you see that tumour size and nodal status is also independent information. So in this early training phase, this was not done in the validation studies, but in the training phase they quickly put together the EndoPredict scores, so the gene expression data, with tumour size and nodal status, it’s obviously an algorithm, not just an addition, and it’s actually called the EP, as in EndoPredict score, and the EP Clin meaning now tumour size and nodal status are included. This is actually what the company is marketing in the German speaking countries but obviously the marketing is now just for prognosis of women with ER disease, it is not marketing for the late metastasis story which is brand new.

What is the best treatment in these cases?

This is obviously, as always in breast cancer when you want a positive predictive factor, is this signature a factor? If we have the high risk group are these the ones that would really preferentially benefit from, for example, a late endocrine treatment or even we could think of maybe targeting the mTOR pathway or something like that in these women. We have no data on that. We are working in our trial, which is the ABCSG6A, this was a trial that compared extended anastrozole versus placebo and in that trial we are looking to see if we can actually predict, if we have a positive predictive factor using the signature but that data is just not ready and we’re actually analysing this right now but I was not able to do it for this meeting. Let’s see if we can get into the samples within the next month and maybe see something at ASCO.