Urine Markers for Bladder Cancer

I’m going to talk about urine markers, and you might think in the first half of my talk, these are just some disclosures. You might, none of these disclosures are relevant to this talk. You might think I’m a bit of a marker nihilist in the first half of the talk, but I’ll show you in the second half that I’m not really. But any talk about markers is really, it orients itself around cystoscopy and our use of cystoscopy, which is our gold standard for detection of bladder cancer in the first place and also for surveillance. But of course, cystoscopy comes with limitations. It does not have perfect sensitivity so we’re clearly missing tumors with cystoscopy, perhaps fewer with the new techniques like blue-light cysto. It’s also invasive, and it’s expensive. So, these are limitations that we try to overcome with the use of urine biomarkers. We want to enhance the performance of cysto, to enhance the sensitivity and make sure we’re not missing something, and we want to replace the use of at least some cystoscopy so we can decrease the burden on patients and also on payers.

Clinical utility of a biomarker can fit into different clinical scenarios. So, on the one hand the initial diagnosis and detection of bladder cancer in a patient who has either hematuria or symptoms of bladder tumor we can have markers that help in the surveillance of patients who have had a prior bladder tumor. Ideally, we like to reduce the frequency of cystos and enhance compliance. We’d also like to have markers reveal to us the tumors that we’re not seeing on cystoscopy, and also of course upper tract tumors. And some of the newer applications of markers include predicting the risk of progression and even their use for screening, and we’ll touch, I’ll touch on these in some slides.

The old working horse of the marker world is urine cytology, and what do we like about cytology? Well, first of all, it’s highly specific. If it shows cancer there usually is cancer there. It’s reasonably sensitive for high-grade disease. We like it I think most because it can reveal to us lesions that we don’t see, especially carcinoma in situ. So, we use it as an adjunct to cystoscopy. And of course it’s noninvasive, but it comes with significant limitations. It has a very poor sensitivity for low-grade disease. It does not help us at all for low-grade disease. It’s dependent on the expertise of a cytopathologist, and also the cost of a cytopathologist, and of course we are frequently dealing with indeterminant reports that we don’t really know what to do with.

The whole panel, actually no, I’ll come to that in a second. What we want in our ideal marker is something that ideally would have, we’d have the result immediately, it would be cheap, it would reduce the use of other markers, it would improve our diagnostic accuracy, and of course it would be, it would have a robust analytic performance. So easier, better, faster, and cheaper. And we have a panel of markers. These will be familiar to all of you. But I would say they don’t necessarily meet the criteria of the ideal marker that I just laid out. So, you’re familiar with the bladder tumor antigen, the BTA tests, the NMP22, ImmunoCyt, which is a Canadian marker, and UroVysion, or FISH. And I’ll just explain a little bit what these are. So, the NMP22 is a point of care assay. It looks like a pregnancy test you can do in the office, and some of you may be very experienced with this. I must say in Canada in general the use of markers is very little, is almost minimal because they’re not paid for by anybody. So, we generally do not use them widely, but you may in your practice use them fairly widely.

The BTA test is looking at a bladder tumor, or it’s looking at a complement in serum, and of course it doesn’t work well if there’s hematuria because the complement from the blood will cause false positives. The NMP22 is a nuclear matrix protein, and if there are a lot of cells in the urine, for example, if there’s inflammation, infection, and maybe there’s a stent in place, it also causes false positives.

The ImmunoCyt is a little bit more complex. It involves three specific monoclonal antibodies that are binding to three specific proteins on the cell surface of cells in the urine. So, it has a higher failure rate when the analysis is done.

UroVysion is familiar to you. This is identifying very specific chromosomal alterations. I think it has gained some use across the country for clarifying indeterminant cytology. So, it can help to sway you whether it’s truly positive or truly negative.

In general, why are we not using these tests more? Well, the main issue is that the studies generally start as case control studies, and they generally have too many positives, so there are too many patients with bladder tumor in the study. So, if you have a hematuria study and 50% of your patients have cancer, well that’s not what we’re actually seeing in practice. And so, it makes the tests look too favorable in the first report, and then often there’s insufficient evaluation and invalidation that allows us to adopt the test into clinical practice.

If we’re looking at the setting of surveillance, of recurrence, it’s the same issue that, well first of all, there are more tumors here so you’re more likely to have a more acceptable positive predictive value. The other issue here is that a positive test might actually mean that there is going to be a tumor in the future, you just haven’t seen it on cystoscopy. So, your test may be better than you think it is. Again, the main hurdle is just a lack of prospective validation studies.

This is from the guidelines, this table. This goes to what Sam was just talking about, and lays out very nicely the performance of these tests. And there are a lot of numbers up here. But in general, what you see is that the sensitivity of all of the tests really is not adequate to allow us to adopt them widely in practice. The tests that come closest are at the bottom there, the ImmunoCyt and the CXbladder where you have sensitivities around or even above 80% and specificities that are comparable.

But what do the patients want? How good does the test have to be? And this is a study from Memorial Sloan Kettering in 2007 where they took 200 patients and they asked them basically how sensitive the test would have to be to skip a cystoscopy. And 75% of patients wanted the test to be at least 95% sensitive, and the other, and 21% said 90% was good enough. Of course, none of these tests are that good, so all of these patients would prefer to have their cystoscopy than take the risk of skipping the cystoscopy for the test.

Some of the best data we have from the quality of the studies and the quality of the data is for NMP22. Bart Grossman led two multicenter trials, one for diagnosis and one for surveillance. And you see the diagnosis study here was prospective, 23 centers in the U.S., 1,300 patients, only 79 had bladder cancer. So, this shows you how we have to be realistic about what the patient population is. The sensitivity was only 55%, specificity was pretty good, and just it was also compared to cytology. But overall, I think we’d agree that this level of sensitivity is not adequate. The negative predictive value was approximately 96%, but since the vast majority of patients don’t have tumors it’s relatively easy to have a high negative predictive value. There were four patients who had tumor detected by marker only, and not by cystoscopy. In the surveillance setting it’s similar, again prospective multicenter 600 patients, 100 had tumors. So, there are more tumors, and you can see sensitivity is only around 50%. So, excellent studies, excellent quality of data, but the tests just underperform.

Coming back to these clinical utilities let’s just look at tests for predicting progression and what we have there. One of the interesting new concepts is one that the group at MD Anderson has presented where they’ve looked at FISH, so the UroVysion test, in patients getting BCG for high-risk disease. And you’ll see in the two rows highlighted in blue that the patients who have a positive FISH six weeks, so after induction of BCG, whether they started with positive or not, have a high risk of recurrence and progression. And so, Ashish Kamat has proposed this idea that we use FISH after induction BCG as a marker of failure, and prioritize the patients for alternative treatments, including for clinical trials.

I’ll just skip over there for the sake of time. Also, from the same group, from Ashish Kamat at MD Anderson they’ve looked again at patients receiving BCG for high-risk disease, also intermediate-risk disease, and they’ve looked at the urine at baseline before starting treatment, and then before and after BCG treatments, the sixth induction and their third maintenance dose. They looked at a panel of 12 cytokines and compared post-treatment to baseline, and they’ve developed a nomogram that’s based on nine of these markers that appear to be relevant, significantly different between post-treatment and baseline, and they can predict the likelihood of response based on cytokine production in the urine. So, this is a novel marker that requires validation in other studies to see if this can really predict response to BCG.

Now, what about screening for bladder cancer? This is one of the Holy Grails of bladder cancer. I think up to now we have no evidence to support screening, and indeed this large UroScreen study from Germany confirms that we do not have a lot of evidence right now.

This was a prospective trial started in 2003 that looked at 1,700 men who were perceived to be very high risk due to their exposure to aromatic amines at a factory, or a few factories, two large factories in Germany. They could also be smokers, etcetera, that was all captured, and they had multiple tests done. They had urinalysis, urine cytology, and then NMP22, UroVysion, and survivin. So, there’s 1,700 subjects, a large trial, 7,000 urine samples. There were 400 positive tests, but there were only 18 bladder cancers diagnosed. So, all of the effort, all of the positive tests, all of the workups that this would require only for the sake of 18 bladder cancers, most of which of course were low-grade noninvasive. So, it would be considered not necessarily worthwhile.

Now, what do these smart characters here, including Seth Lerner and Bart Grossman, say? This is a statement on the use of biomarkers, not only from the urologic community but also from the lab medicine community. And this is a little bit older, 2006/2010, but they’re basically saying that currently no markers can be recommended for use. And indeed, as Sam alluded to, the AUA more or less agrees with this in 2016. So, in the surveillance of non-muscle invasive disease regardless of grade and stage no marker should replace cystoscopy. So, that’s one key message I think we can all easily agree on. In patients with low-risk disease markers should not be routinely used. It does not add anything. And then the potential use is in patients, so the UroVysion test that I highlighted as measuring a response to BCG, or UroVysion, or ImmunoCyt for indeterminant cytology to help guide whether the patient needs further evaluation.

This is where you may ask why am I talking about markers if they’re no good, and as I said I come across as a bit of a nihilist, but there are a lot of interesting markers, or there are some interesting markers in the future, and I think that things will change in the not too distant future. So, one of the promising new markers is the AssureMDX from MDX Health, and it’s being studied in a couple of different contexts. The group in Rotterdam I think has developed this, or it’s spun off from their group. They’ve looked at hypomethylation of a panel of genes, but they’ve gotten it down to three most significant genes indicated here, not genes I would necessarily expect anyone to have heard of expect for maybe TWIST. But they’ve also combined it to mutation analysis of three genes that we know are very important, especially in non-muscle invasive disease.

In a discovery sense, so this was testing it for the first time in 154 patients with hematuria you see that 74 had bladder cancer, so again, we have this problem of really two high disease prevalence. The area under the curve, so the accuracy was 93%, the sensitivity was great at 97%, the specificity was also pretty good at 83%, the negative predictive value is worth highlighting 99.9%. This is assuming a prevalence of 5% bladder cancer in the hematuria population, even though the study was done almost 50%, so a bit of a disparity there.

But this group has just recently validated this in a three-center study, prospective study in Scandinavia. I think they’re all in Scandinavia. So, 200 hematuria patients, again about half of them had bladder cancer, 96% even better area under the curve in the validation study. Sensitivity is still very respectable at 93%, remember our patients for the most part want something above 90%. Specificity was good at 86%, and the negative predictive value 99.6. So, this looks very promising I think as something that may be clinically useful. They calculated potential for 80% reduction in cystoscopy, and there’s a large multicenter trial ongoing, or not ongoing but about to be launched in the U.S. and Canada. We’re taking part in this trial as well in 700 patients. So, I will be very interested to see what becomes of that. The test is not yet available in the U.S. but will be soon.

Now, there’s another multicenter perspective study that was published online just at the end of last year. This unfortunately, it’s from the same group in Rotterdam, but it uses a different mix of the same alterations, and it basically shows that it’s not that great, and I’m actually going to skip over this just so it doesn’t cause confusion.

The other marker of interest right now that’s developing is the CXbladder, which comes from a company in New Zealand, and it is looking at quantitative PCR for five mRNA, messenger RNAs in the urine. They started with a large panel, and again I’ve whittled it down to the five most important ones that you can see here. And they have done something interesting. They’re developing this as sort of three different markers. There’s CXbladder Triage, which is supposed to give you a black and white answer to scope or not to scope in a patient who presents with hematuria based on some clinical risk factors. So, they have an algorithm that integrates not only the score from these five mRNAs, but also the clinical factors. Then they have the CXbladder Detect, which is for potentially higher risk patients, and it just gives you a score the likelihood of having a bladder tumor. It doesn’t not incorporate the risk factors. And then there’s the CXbladder Monitor, which is for patients being monitored for prior history of non-muscle invasive disease, and the algorithm includes clinical information on their prior history.

And so, the Triage is interesting too that they, so they’ve done a study that was published in 2015. They looked at almost 600 patients, and they made a cutoff at 40%. So, 40% of the patients were going to have a negative test. That’s how they’ve defined it. And you can see on the left-hand panel the red triangles are a high-grade disease. So, they missed no high-grade disease with this cutoff. They did miss a few–low-grade disease are the green triangles. So, tumor detection for high grade is 100.0%, the negative predictive value is 98.5, and the overall test sensitivity is 95.0%. So, this is something new that clearly needs validation. This is just a first report.

For their CXbladder Detect they had 485 patients at multiple centers, 11 centers in Australia, patients with gross hematuria, and again this is just based on the score of the mRNA. There are no clinical factors incorporated into this one, and you see that it performs reasonably well with the sensitivity of 82% and the specificity of 85%, and it compares well with other markers. So again, this is something that needs further validation.

And then finally, the CXbladder Monitor, this was presented by Yair Lotan at the AUA in the plenary session last year, and was published just recently in the Journal of Urology. This is a prospective study in 11 centers in the U.S., 800 with prior muscle, non-muscle invasive disease. They had a training and a validation set they compared also to other markers. And as I mentioned they used an algorithm. Only 11% of patients had tumor, which is a reasonable number, what you’d expect. Sensitivity was 93%, so again this is a good performance. It outperformed all other tests, and it was higher for high grade versus low grade, as almost all of these tests are. Negative predictive value was 97%, which is a little bit lower perhaps than the AssureMDX. And the question posed by the authors is can this replace some of the cystoscopies, and that will remain to be seen.

Finally, I wanted to touch on, just on the last two slides, on some of the markers that are under development but in earlier stage development. And this is a paper that came out of Birmingham where the group was actually isolating DNA from urine and then sequencing it. And they’ve sequenced just a panel of six frequently mutated genes, and you can see which genes these are. They include things like FGFR3 and p53, things that we’re all familiar with. And they looked at 20 non-cancer and 120 cancer patients, as well as 89 who were being monitored after TURBT. And the sensitivity of 70% by itself is perhaps not so impressive, but you can imagine that it would be very easy to increase the number of genes that we’re looking at here, and you could imagine making this test better. And I know that there are various groups doing this, including Theresa Koppie from Oregon Health Sciences University. They had a panel of 99 genes and predicted based on TCGA data to capture 99.9% of patients. And so that’s something that we’ll look forward to seeing develop over the next few years.

And then finally, the group in Aarhus, Denmark has done a lot of really interesting work in genomics of bladder cancer, and they’ve come up with this model, and they’ve published it in European Urology just last year where they actually take bladder tumor tissue as well as peripheral blood, they sequence for tumor DNA, germline DNA, and they pick out specific somatic mutations in the tumor. And then they can develop a digital drop PCR test to look for these changes, and then they can monitor the patient with a simple blood test, or urine test, either way, to look for recurrence. So, it’s very elegant, although it is dependent on sequencing the patient’s original tumor. But I think we should all pay attention to what’s going on now, and hopefully we’ll have some better markers in the next few years. Thank you.