The Uromigos Episode 125: Paper of the Month—MRI Screening for Prostate Cancer

Martin Eklund discusses his papers published in the New England Journal of Medicine and Lancet Oncology.

Episode Transcript

Tom:
Welcome everybody to our latest paper of the month. I’m joined by Martin Eklund from the Karolinska. In fact, Martin could have won this twice, two consecutive papers of the month, which we haven’t had before. Martin had a publication in the New England Journal, looking at MRI and prostate cancer from a screening perspective. And we’re going to talk mainly about that.

But there was also a more recent Lancet Oncology paper with a slightly different slant on this. And maybe in the last five minutes, Martin, we’re going to jump into that. Martin, if you’d like to introduce yourself first and maybe go over any major conflicts, that’d be magic.

Martin Eklund:
Yeah. So, my name is Martin Eklund, and I am an associate professor of epidemiology at the Karolinska Institute in Stockholm. And I do have conflicts of interest for the second publication that you mentioned here, the one in Lancet Oncology, which looks into the performance of using the Stockholm-3 risk prediction algorithm for prostate cancer diagnostics in an MRI context.

And the Stockholm-3 algorithm is owned by a company called A3P Biomedical and I own shares in that company.

Tom:
Martin, I’m going to start if I may by asking you to give me a bit of a background about this. I’m aware, for example, that historically we used to do PSAs and then do almost ultrasound-guided random biopsies of the prostate.

I get the impression from recent New England Journal publications, and also from some Lancet publications, trials like PROMISE where we’ve identified already that that’s a suboptimal approach and a preferential approach is to use MRI scan first and then targeted biopsies from there.

Would you agree with that? And are there other facts which I missed out?

Martin Eklund:
No, I absolutely agree with that. There are a few papers that have been looking into this before of using MRI for improving prostate cancer diagnostics. And what is new with our paper, or what it adds to this sort of puzzle that we’re building with all of these papers, is the population-based screening approach.

Because we know from a lot of previous trials, specifically from the ERSPC trial, that doing prostate cancer screening, organized prostate cancer screening like we tend to in Europe, for instance for mammography screening for breast cancer, it reduces prostate cancer mortality, but it comes to a very high price in terms of over-diagnosis.

And therefore, it’s very important to look into this specifically. If you do organize prostate cancer screening and if you’re using MRI there, how does that perform?

Tom:
So, do you want to just talk a bit about what you did? Because this is slightly different in that you kind of went to patients who weren’t pitching up to urology clinics or had problems. You were looking for patients almost off the street. And you were doing a PSA test, and then from there progressing.

Do you want to just talk a bit about the methodology of what you did?

Martin Eklund:
Yeah, absolutely. So, what we did is exactly what you said, that instead of intercepting patients at the clinic who had already been referred for a prostate biopsy, we sent out the invitations to the general population. So, to general men in the age of 50 to 74, and we invited them for a prostate cancer screening within this trial.

So, if a man had an elevated PSA test, and if we talk specifically about the New England Journal of Medicine publication then we define that as a PSA of 3 or higher, then we randomized those men. Everyone who had a PSA above 3 were randomized to either undergo a standard ultrasound guided prostate biopsy or MRI, and then targeted and systematic biopsy for men who were positive on the MRI.

So, if you had a PI-RADS score of 3 or higher.

Tom:
Now, Martin, you recruited 6000 patients. What was the primary endpoint of the trial? And did you meet it?

Martin Eklund:
Yeah, so what we looked at was that first of all, by using MRI we don’t of course want to reduce the detection rate of clinically significant cancers. And then you can argue about, of course, what is a clinically significant cancer? What we used in this trial was a Gleason score of 7 or higher.

So, a 3+4 or higher Gleason score, that was defined or predefined in this study to be what we called a clinically significant cancer. So, that was our primary endpoint to look at non-inferiority with respect to detection rate of clinically significant cancers.

Martin Eklund:
Yeah. And on top of that, of course, we want to show that we’ll also show some improvement on other endpoints.

So, key secondary end points: we looked at reduced rates of clinically insignificant cancers, defined as Gleason score 6 cancers, and also a reduction in benign biopsies.

Brian:
And Martin, the primary end point was non-inferiority, or trying to show that this approach was non-inferior to the standard approach. Is that right?

Martin Eklund:
Exactly. For detecting clinically significant cancers, yeah.

Martin Eklund:
And why …

Brian:
And a couple …

Tom:
Why were you interested in a non-inferiority endpoint? If it’s no better than what we’re doing already, why would we want to change?

Martin Eklund:
Yeah, that’s a good point. So, what we want to do here is that we know, again, from the ERSPC trial for instance that if we do population-based screening we can reduce the mortality rate of prostate cancer. And we want to retain that mortality benefit that we know from previous trials that it exists.

But we want to then reduce the unintended negative consequences of doing screening. So, we want to show non-inferiority with respect to clinically significant cancers and having that as sort of a proxy endpoint for mortality. And making it plausible that if we detect all of these clinically significant cancers, we still retain the mortality benefit that we know exists from prostate cancer screening.

But we’d get rid… Or no, we don’t get rid of it, but we’d reduce markedly all of the negative side effects of over-diagnosis and large fractions of benign biopsies.

Brian:
So, maybe talk about that then, in terms of the results of the trial, about the reduction in, I guess, detection of clinically insignificant cancers.

Martin Eklund:
Yeah, absolutely. So, what we saw was that we had an even stronger, I think, effect in terms of reducing these unintended consequences of screening than we even had hopes for, I think, in the beginning. So, we saw that about 50% fewer men needed to undergo biopsy procedures overall in the MRI arm, compared to the standard biopsy arm.

Which also translated into a 75% reduction of men who were unnecessarily biopsied who had only benign findings on the biopsies. And about a 60 percent or even a 62 percent reduction in diagnosis of Gleason 6 cancers. So, what we typically view as over-diagnosis.

Brian:
And I guess the question is: is that worth it given the cost of the MRI, right? Because …

Martin Eklund:
Yeah, exactly.

Brian:
… even if we’re diagnosing Gleason 6 cancers, if we’re doing active surveillance that’s not a huge deal. Or maybe it is, I don’t know. That’s sort of the trade-off in my mind.

Martin Eklund:
Yeah, exactly. I mean… And that depends on how different people, patients, view active surveillance. To some it’s a huge deal, some cope with it very well. But irrespectively of whether it’s a huge deal to a patient or whether you cope with it well, we actually turn perfectly healthy men into cancer patients.

And that also has a cost from a healthcare point of view, of course. We need to manage these patients. So, whether it’s cost efficient is something that we’re looking into at the moment. And we’re working together on the cost effectiveness of doing MRI and targeted biopsies in a screening context.

And we hope to publish those results during the fall here.

Tom:
So, Martin, just to summarize, to crystallize the results. Essentially, you’ve done a lot of MRIs on people kind of off the street. If they had a PSA above 3 …

It looks like you can identify serious cancers and potentially miss those or potentially not identify less significant cancers, which ultimately is a better sifting process than the other system in which we have lots of biopsies and we identify lots of clinically insignificant cancers. Is that a reasonable summary?

Martin Eklund:
I would say that that’s a perfect summary. Very well expressed.

Brian:
It’s the first time in a 100 plus podcasts. But Martin, just to push back a little bit. I’m looking at a table from the paper now. I mean, there were only 22 less Gleason 6 cancers, diagnosed in the MRI group.

So, it wasn’t a huge absolute number of men out of 1500. It was less, but the absolute number was fairly small. Is that right?

Martin Eklund:
Well, you have to look also at the fraction here. Because there’s a three to two randomization here.

Brian:
Okay.

Martin Eklund:
So, there are more people in the MRI group here. So, you can’t really look at the absolute numbers per se. You need to look at the percentages.

Brian:
Fair enough.

Tom:
And Martin, in terms of the duration of follow-up of this trial, prostate cancer is a long game. Are you confident that your follow-up period for the study is long enough? And are there other analyses within this trial that you haven’t done that you need to do that might influence the result?

Martin Eklund:
Yeah, I mean … So, to break those two questions up into …

Martin Eklund:
Hello? It seems that I was cut off there for a while.

Brian:
Yeah.

Tom:
Martin, I’m really sorry. You got cut off there. And what we can do is we can actually edit out a midsection of this, which will allow us to not start again. So, we’ll have a few minutes in the middle which we’ll edit out.

I think the question I just asked, and you said there were two parts to it, is I said, “Is the follow-up long enough? And were there additional elements of data we need to analyze from this dataset?”

Martin Eklund:
Right, exactly. So, to answer those questions one by one. So, starting with the first one, whether the follow-up is long enough. That’s really sort of the million-dollar question here, I think.

In the paper we’re making the claim, or we hope to be able to make the claim or at least make it plausible, that by detecting these clinically significant cancers we will also have the mortality benefit from going on and treating these patients.

Of course, just like they showed in the ERSPC trial, that if you do that you get the reduction in prostate cancer mortality. But that’s essentially an assumption that we’re making in the paper. And we discussed that, of course, as a limitation in the paper as well. But it is a bit of a leap of faith. And we hope that to be true. And we think that it’s true.

But we can also turn this whole thing around. I mean, what if we had mortality as the end point of the trial? Well, first of all, we’d have to make the trial 10 times larger, and we’d have to follow the patients for 10 to 15 years or so. And by the time that the trial would be ready to read out, it would probably be obsolete.

So, I think that we have to sort of dare to trust proxy end points to make changes in the healthcare system, because otherwise the innovation cycle in the healthcare system would be way too long. And I think the patients will lose out there. So, yeah, it’s a very good question. And I think that’s something that we will continue to discuss for sure, in other contexts as well.

And the second thing here: are there other analyses that we want to make from this trial? And, yes, certainly. There are a number of those analyses that we haven’t made yet. And the one we have already touched upon, that’s health economy.

We know that the MRI is expensive, of course. And it’s a bottleneck in the healthcare system to get access to MRI. And especially in the US I think MRIs are much more expensive than they are in Sweden, at least. So, we need to look into health economy here, of course.

But there are also a lot of other interesting analyses that we want to make here. For instance, by doing targeted biopsies, you’re sampling much more intensively from one or a few lesions in the prostate. And that sort of ultimately inflates the Gleason score.

So, in a sense, we’re changing the endpoint a little bit by doing the targeted biopsies. And we hope to address that point, whether we can actually compare the biopsies from the targeted approach with the systematic biopsies in the follow-up publication. Because I think that’s one of the key questions that need to be answered as well.

Brian:
And Martin, one thing I noticed as well is that in the MRI group, half of the men didn’t get a biopsy because they had a negative MRI and a low Stockholm score.

So, that may almost be the biggest benefit, right? If you have a normal MRI, you can, in essence, rule out high-grade disease. And you probably don’t need a biopsy to diagnose your low-grade disease. Is that fair?

Martin Eklund:
Yeah, exactly. That’s the way we view it, at least, in this study. And that’s how we hope that it could be used in the screening context and within the screening context, and also in sort of a guideline-based approach like is typical in the US, for instance.

I think that also the idea here is that those patients, they would probably come back for another round of screening in let’s say two years or something. So, if there is any sign of disease …

Let’s say that we missed something there and there was some progression to the next time, we will probably capture it in two years and there is no big harm there, in all likelihood at least. So, that’s the way we view it, that we can rule out a lot of men from having a biopsy, therefore also reducing the number of negative biopsies, reducing the number of clinically insignificant cancers. And if we happen to miss something, we catch it at the next screening round.

Tom:
Martin, from a global perspective, in the traditional method of having a biopsy via ultrasound scan going on, and either having radiotherapy or surgery as the gold standard, in that environment what does this new approach have on the proportion of patients that go on to get potentially curative therapy?

Does the number increase? Or does the number reduce? Or are they different populations? Or are they getting more surgery and less radiotherapy? More hormones? What’s the distribution associated with subsequent therapy using this approach?

Martin Eklund:
Yeah, that’s also a great question. I think that we will see fewer patients on active surveillance because we get rid of so many of the lower grade cancers. And whether it will have impact on how many patients undergo radiation versus surgery, that remains to be seen.

I think … I can only speculate, but I think that potentially the fraction of patients undergoing radiation might be higher because we do more targeted sampling of the lesion and perhaps get, arguably, a more representative Gleason score of that particular lesion. And potentially also then a shift towards higher Gleason scores, meaning potentially more radiation. I don’t know. This is just speculation, but …

Tom:
Martin, what are the shortcomings of your work?

Martin Eklund:
Yeah, I think that there are a few ones, of course. We mentioned perhaps the key one and that’s our proxy end point. We’re trusting the Gleason score as our end point. And we don’t know what impact it will have on mortality.

We cannot say that definitively, at least. Secondly, we of course, as well, need to study this population over multiple rounds of screening, so that we can see that this benefit we see in terms of reduction of biopsies, in terms of reduction of low-grade disease, that that is retained over multiple rounds of screening. And we believe that that will be the case, but nonetheless, it needs to be shown as well. So, I think those are the two main limitations. To me at least.

Brian:
Tom, you want to move onto the Lancet Oncology publication?

Tom:
Yeah, so, [inaudible 00:18:31] a little bit about that and a bit about the Stockholm score, and pull that piece together, and how that adds to the current literature base.

Martin Eklund:
Yeah, I mean, the thing here is that, as we’ve already touched upon, what we do in this first publication … By first, I mean the one in the New England Journal … It leads to a lot of MRIs. And we know that it’s a bottleneck, again, in the healthcare system. We’d have to fight for MRI capacity, and we’d need to train radiologists to interpret the MRIs, et cetera, et cetera.

So, if we can reduce the number of men that need to undergo an MRI, so we have a better… Again, if I can sort of borrow your expression of a better sifting process to get patients into the MRI, then that would hopefully lead to a more streamlined diagnostic approach.

And again, we need to do that in such a way that we retain the number of clinically consequential or clinically significant cancers. So, that’s what we do in the second publication, published in Lancet Oncology.

That we use the Stockholm-3 score as a selection mechanism for patients to undergo MRI. And we compare that with the PSA approach that we used in the New England Journal paper.

Tom:
It’s important not to confuse the Stockholm scoring system with… Sorry, the Stockholm score, with Stockholm syndrome, which I think is falling out of your catch. These things are very different and I [crosstalk 00:20:07].

Brian:
Thank you, Tom. Thanks for that.

Tom:
Yeah, sorry.

Brian:
Can you describe … I think maybe not everybody’s familiar with the Stockholm score. What is it exactly? How is it derived? And how’s it used in this study?

Martin Eklund:
Yeah. So, it is a risk prediction algorithm. So, it’s similar to other risk prediction algorithms, like for instance the 4K score. There are numerous, or several of those at least, risk prediction or risk calculators, like the ERSPC score and things like that.

So, what we do is that we draw a blood sample from patients. And within the blood sample we measure certain proteins. So, PSA is one of them, but also free PSA and a MSMB, and MIC1, et cetera. So, there are a number of those that we measure. And we combine it with also looking at genetic markers. So, singular nucleotide polymorphisms, and using those to construct a genetic score.

And on top of that, we look at the patient’s age and things like that. And all of this is then being put into the Stockholm-3 risk prediction algorithm, which predicts your risk of having a Gleason score 7 or higher cancer.

Tom:
And Martin, so it’s fair to summarize that you’re conscious that PSA and MRI would potentially result in a big healthcare burden. The Lancet Oncology paper attempts to reduce that by having a better scoring system to identify patients who need an MRI scan. Do you think that scoring system will stand the test of time, or do you think we can do better?

Martin Eklund:
I think that there are probably still improvements to do there. But …

Tom:
And where are those improvements to be made?

Martin Eklund:
First of all, I think that there are probably some higher penetrance genes that we can look at and incorporate into the scoring system, like BRACA for instance, in particular BRACA2, of course.

So, those are very important for a few individuals. And there could also be other genes that we don’t know, of course, and other protein markers. And we could also combine it potentially with urine markers, for instance. But I think that there is also a limit. I think that state of the art risk prediction algorithms these days, like Stockholm-3 and like also the 4K score, and other of these algorithms that are performing well …

I mean, considering the variability that we have in the Gleason scoring system, there’s sort of a limit of how good we can get the AUC values of these prediction algorithms to become, because there’s such noise in the end point that they are evaluated against.

So, there is still room for improvement, but I don’t think it’s going to be a tremendous amount of improvement by adding new variables to those risks’ prediction algorithms.

Brian:
So, Martin, do you think in an ideal future state, that screening would be something like a Stockholm score? A blood test, an age, and family history, perhaps, et cetera, et cetera. As a wide screen for the whole population.

And then in men above a certain risk score, then undergo an MRI, and then obviously if positive undergo biopsy. Do you think that general paradigm will hold, even if the specific tools change?

Martin Eklund:
I think so, yes. Because I think to start with cheap tests that are easily available, like a PSA test, makes a lot of sense to me. And if the PSA test is elevated you go on and do some more refined testing, like Stockholm-3 score for instance.

And if that is also elevated, you’re going to do even more expensive and more refined testing, like MRI. And then you keep on doing that until the evidence sort of piled up that, okay, this patient actually needs a biopsy, and this patient does not need a biopsy.

So, yes, I think that that sort of sequential thinking will stand the test of time. But of course, we will add weapons in this arsenal, I think, for instance, circulating tumor DNA and those kinds of things. Those kinds of approaches are super interesting to follow the development of as well.

Tom:
Well, I just …

Brian:
Do you think we’ll get rid of PSA ever in our lifetime?

Martin Eklund:
Good question. That’s a challenging one. No, I don’t think so. I think it will …

Tom:
I was about to ask that question as well, by the way, Martin. And so …

Brian:
You were not. You were not.

Tom:
It would have been my good question as well.

Martin Eklund:
Right, yeah. But it was a good question, irrespective of who it came from. But no, I think …

Tom:
There’s no competition here, by the way.

Martin Eklund:
Well, some healthy competition is probably good, so. But I think PSA is so ubiquitous. It’s available everywhere. It’s cheap. The assay systems are well-developed and available in most parts of the world. So, no, I think we’ll be stuck with PSA for a long period of time, for sure.

Tom:
Martin, this has been terrific. Have you got anything else that you’d like to add? Brian, anything from you before we call it?

Brian:
I guess just maybe, you mentioned the circulating tumor DNA. And Tom, you have expertise there. I mean, could you imagine that sort of complimenting or even replacing some of these other things? Right? Because it’s obviously detecting cancer directly as opposed to maybe indirectly with some of these risk scores.

Martin Eklund:
Yeah. For sure. I mean, I think that’s a super interesting technology and I think we’ve only seen the start there. And I think the obvious first steps are around more on the treatment side, to make more targeted treatment approaches.

And I think that we’ll see a tremendous amount of development therefore for prostate cancer in the years to come. But also in the diagnostic setting, it needs to come down in price, of course. But there’s certainly a lot of potential there. But I think it will take a long period of time and it might not even completely replace things like we have today, for instance MRIs.

Also, especially with the imaging, you want to see where the lesion is. And it’s also for surgery planning, for all of those kinds of purposes, so.

Tom:
I agree with that, Martin. And I think the MRI scan is always going to be important because you need to know what’s going on. I think that your Stockholm score and PSA, I think there’s a big improvement that’s going on in that environment.

I think the CT DNA story in prostate cancer is very real though, because it’s a technology that works and the price will come down as we use it more and more. And I think we will be incorporating circulating tumor DNA.

One of the questions from my perspective is PSA is actually pretty good. And the question is: will it be more sensitive and specific then? Because it might be more sensitive but less specific or vice versa. So, I think that it’s part of the armory.

But actually, I can see a lot of patients going down this route. I can see MRI becoming more widely used in the population. And the really important point is how we address that because it’s going to be a complicated issue for the future. Martin, anything you’d like to add before we call it?

Martin Eklund:
No, I think that we’ve covered a lot of ground here. And it’s been a really fun and interesting conversation. So, I think …

Tom:
Congratulations …

Martin Eklund:
Nothing that I can think of at the moment.

Tom:
… on double paper of the month. Double paper month.

Brian:
Yes, congratulations. This is great.

Tom:
Martin, keep in touch.

Martin Eklund:
Thank you so much. Yeah, absolutely.