The prostate cancer 'super test'

Now, as we’ve been learning, a big part of the problem with prostate cancer is that we don’t have a simple solid test for disease. PSA can be misleading, and, even when it does detect a cancer, it can’t tell us who is at high risk and needs intervention: instead it triggers a slew of costly follow-up investigations. So can we improve on this? Well, a team in Cambridge think they can. Rather than rely on one single marker, they’ve used machine learning to comb through hundreds of molecules and markers, integrating genetics and clinical outcomes, to spot patterns of changes in a carefully chosen 100 of those markers that when considered together can more accurately, they say, predict prostate cancer outcomes. The company is called EDX Medical, and they hope to have the technology in front of regulatory bodies within the next 12 months. Chris Smith spoke with founder and scientist Sir Chris Evans who explained how it works…

Chris E - It's what it says on the tin, it's a super test. The aim was to look at a whole cluster of diverse biomarkers clinically linked with prostate cancer or the stage of prostate cancer, and we finally narrowed it down to a hundred and odd markers — those being hereditary genetic mutations, epigenetics, transcriptomics, RNA, and proteomics. We now look at all of those biomarkers in a single sample of blood and urine from a single man at a single point in time.

Chris S - As we've heard in this programme already, though, the issue with prostate cancer is often determining who has prostate cancer that will cause them a problem versus who has prostate cancer that is largely innocuous and will never cause them a problem in their expected lifetime. So can your test tell those two apart?

Chris E - Yes, there are sufficient biomarkers in the categories to actually characterise, first of all, the result — cancer is present. The second is: this cancer is very aggressive, and it's aggressive for these reasons, because of these biomarkers which have lit up. Or: your cancer does not appear to be aggressive, we have not found the aggressive biomarkers we were looking for, or they are at very low levels. And then thirdly, the type of prostate cancer you've got — so that again may influence how we decide to treat and monitor this disease going forward.

Chris S - So this would move us into a regime of what some surgeons and urologists are dubbing a sort of active monitoring state — where you can pick up the disease, and because you can gauge how aggressive or serious it is, you feel more confident saying, “Well, in this instance we're not going to need to do anything at this stage, we're just going to watch.” So can you do that active management, where you just keep an eye on people, monitoring them with this test over a period of time, and then if the tumour evolves and shows changes or switches to a more aggressive type, that would trigger a different kind of treatment?

Chris E - Yes, this test would be a brilliant monitoring test because it will pick up all those different characteristics as the disease progresses — so there's no question. And it can save quite a lot of time and money from having repeated MRIs, and particularly biopsies. If our blood and urine tests can be ready annually or biannually, we can generate data which is clinically significant and very reliable. That information may be all that consultants and doctors need to see.

Chris S - How have you validated this? And by that I mean, how have you compared what your test says about a patient with what the other methods we currently have — or long-term follow-up — say is going on in that patient, to prove that your test is getting it right?

Chris E - We need to go through at least the next 12 months of validation trials because we've done sufficient work to know that the test works. Okay, so we take 10 patients, knowing four of them have cancer and six haven't — we would get that right almost 100%. But in terms of comparing the diversity of people to start with — in terms of ethnicity, their age, the other phenotypic problems they have or don't have — and then the different types of prostate cancer that could be present and the different stages it could be present, we need to validate that through the trials we're now carrying out and submit all our findings to the MHRA and the FDA and get approval to sell our test to all, but particularly — and initially — to the private markets.

Chris S - One of the most important mantras in cancer is early detection, because the earlier we're able to intervene, the earlier the disease is in its course, the less likely it is to have gone anywhere, the less likely there are to be consequences and side effects of the treatment. So how far forward do you think, with this test, you might be able to bring early detection?

Chris E - Early-stage diagnosis at the very, very earliest stage is the ultimate holy grail — so that you can pick this thing up when it is ideally a single primary tumour in its smallest size, before it’s started doing anything else sinister. We know that our biomarkers — and particularly multiple biomarkers — can do that. Where you've got multiple biomarkers in multiple categories of completely different biochemical types — albeit genetic as well as protein, epigenetic, transcriptomic — you're increasing your chances of picking up the things you need to detect at the very early stage.

Chris S - And let's look at some numbers then. How good is it? If we look at the accuracy of distinguishing cancer from non-cancer first — just tell us, what is that number? If I give you a sample from a person, what's your ability to say cancer or not?

Chris E - What we've seen in terms of specificity and sensitivity — our figures have all been between 94 and 96 percent, and we've even seen categories as high as 99 percent.

Chris S - So it's a good rule-out test. If you find it, you're saying to someone there's a high likelihood this is a genuine cancer — or not — so that’s really good. When you find a cancer, what does it look like in terms of its ability to predict whether this is a “watch and wait” or “escalate this through imaging and possibly other kinds of treatment” case?

Chris E - Based on all of the samples we've done so far, you're talking about 90 percent. Again, these figures — around 90 percent — mean you're getting nearly nine out of ten cases right, in terms of what you're trying to do here: whether cancer is present or not, whether it's early stage or not, or whether it is aggressive or not. Now, we need to do quite a lot more computing work on the very aggressive stages versus the less aggressive cancers to publish the accurate number.