JCO PO author Dr. Eric Klein shares insights into his JCO PO article, “Performance of a Cell-Free DNA-Based Multi-Cancer Detection Test in Individuals Presenting with Symptoms Suspicious for Cancers” Host Dr. Rafeh Naqash and Dr. Klein discuss how a multi-cancer detection test may facilitate workup and stratification of cancer risk in symptomatic individuals.

TRANSCRIPT

Dr. Rafeh Naqash: Hello and welcome to JCO Precision Oncology Conversations, where we bring you engaging conversations with authors of clinically relevant and highly significant JCO PO articles. I'm your host, Dr. Rafeh Naqash, Social Media Editor for JCO Precision Oncology and Assistant Professor at the OU Health Stephenson Cancer Center at the University of Oklahoma.

Today, we are excited to be joined by Dr. Eric Klein, Emirates Professor and Chair at the Glickman Urological and Kidney Institute at the Cleveland Clinic Lerner College of Medicine. Dr. Klein is also a distinguished scientist at Grail and author of the JCO Precision Oncology article titled "Performance of a Cell-free DNA-based Multi-cancer Detection Test in Individuals Presenting with Symptoms Suspicious for Cancer."

Our guest's disclosures will be linked in the transcript.

For the sake of our conversation today, we'll refer to each other using our first names. It's great to have you here today, Eric, and welcome to our podcast.

Dr. Eric Klein: Thanks, Rafeh. I'm happy to be here.

Dr. Rafeh Naqash: So today, we're going to try to delve into this very interesting paper. We've had a couple of very interesting podcasts on liquid biopsies, or plan to have a few more. And this is a different aspect of liquid biopsy assessment, and the context here is early cancer detection. Now, the story as it starts, is based on the methylation profile of cancer. Can you tell us, for the sake of our listeners, as we have a very broad audience ranging from trainees to community academic oncologists, what do you understand by methylation profile on a cancer?

Dr. Eric Klein: Sure. Happy to start with that. There are lots of cancer signals in the blood. Cancer cells secrete or otherwise supply the bloodstream with DNA that has methylation signals that are specific to cancer. That's a hallmark of cancer-specific mutations. You can look at chromosome fragments, you can look at proteins and mRNA and exosomes and that sort of thing. In Grail's development study, we focused on using methylation because that, as I mentioned, is a fundamental process. A fundamental property of cancer cells is altered methylation. And in our original development studies, that was the strongest signal, the one that allowed us to have the lowest limit of detection when cancer was present, and the one that allowed us to have the best predictive accuracy for the cancer signal origin. Some people think about that as predicting the tumor origin or the tumor type. And that's the basis of Grail's assay, a pan-cancer methylation profile.

Dr. Rafeh Naqash: Excellent. And now to understand some of the methodology that you used here, before we go into the details because there's a lot of sensitivity and specificity obviously associated with any cancer detection test, and you want a high sensitivity and specificity. And the idea here is that this would help in triaging patients appropriately using this non-invasive tool. Could you tell us the patient population that you were trying to enroll in this study? And I think there is, again, background to other studies that you have done using the Grail test. Could you put that into context of this specific study?

Dr. Eric Klein: Sure. The population in this particular publication was from substudy 3 of a much bigger study called the Circulating Cell-free Genome Atlas, or CCGA. That was a discovery, refinement, and validation study of this methylation-based signal. And in total, all three substudies together was about 15,000 people, and it was a case-control study. About 10,000 of the individuals enrolled had cancer and about 5000 were not known to have cancer and served as controls. In the first part of the study, substudy 1 of CCGA, we simply asked the question: In individuals with known cancer, could we detect a methylation-based signal? And the answer was ‘yes’. The second question was: In patients not known to have cancer, did we not see a signal? And by and large, the answer was ‘yes’. The second substudy was a refinement and validation of the original methylation-based test. And then this study, what we refer to colloquially as CCGA3, or substudy 3 of CCGA, was the final validation that underlies the methylation assay that is currently on the market.

So, in CCGA3, we determined what the performance characteristics of this test were in a case-control fashion, and what we found, importantly, was that the specificity was very high, at 99.5%, which means the false-positive rate is only half a percent. We found that the overall sensitivity for detecting cancer varied by stage, but when you included all stages 1 to 4, the overall sensitivity for detecting known cancers was about 51%. We found that the ability of this methylation-based test to predict the correct cancer signal origin was right around 90%. And finally, the final performance characteristic was really important, which is the positive predictive value. So in individuals who had a positive signal detected, the positive predictive value was 43%, which compares very favorably to existing screening tests, all of which are below 10%.

That was the background, and the development there was focused on eventually developing a test that will screen the general population, the asymptomatic population, at risk for developing cancer. This is a subset of CCGA3, or the substudy 3 of CCGA, where we looked at the performance characteristics of this test in individuals who had symptoms that could possibly be due to cancer and individuals who had underlying medical conditions that could result in a false positive, and individuals in particular over age 65, because the risk of cancer goes up over age 65.

Dr. Rafeh Naqash: Thank you for explaining that. So, again, going to some of the finer details in this study, you mentioned some very important numbers here, 99%, 63%, or something in that range for sensitivity and specificity. Could you explain a little more on that based on the cancer types? As you mentioned, stage 4, when I read the paper, has more true positives likely based on or related to how much cell-free DNA is released in the tumor. The tumor burden may be playing a role there. Could you explain that a little more for our listeners?

Dr. Eric Klein: A cancer that sheds cell-free DNA into the bloodstream is more likely to be aggressive, and that's been shown in multiple different studies using multiple different platforms. And the reason for that is that the ability to shed cell-free DNA into the bloodstream goes along with biologic processes that we know are related to tumor aggressiveness. So that's a higher mitotic rate, it's neovascularization or the angiogenic switch, it's the ability to be an invasive cancer. And so the fact that you can detect cell-free DNA in the bloodstream implies some degree of biologic aggressiveness, which is not to say that tumors that shed cell-free DNA into the bloodstream are not curable. They are, in fact, curable at the same rate as cancers in people who are not tested for cell-free DNA. We know that for sure. It's just a signal that is there for us to exploit for the detection of cancers in asymptomatic individuals. And the hope is when we screen the general population, the general asymptomatic population for cancer, as we do with mammography and colonoscopy and PSA and so forth, that we can detect cancers at earlier stages, when they are far easier to cure. So I mentioned in CCGA3 that the overall sensitivity across all stages for detecting the presence of known cancers was 51%. That varied from about 16% for stage 1 cancers to 40% for stage 2 cancers to over 80 and 90% for stage 3 and 4 cancers.

Dr. Rafeh Naqash: Right. And again, to provide more background to this, what we've come to understand gradually, as you mentioned, is that shedding is an important event in cancer trajectory. Do you think detection of cancers that are likely positive, driver mutation positive, have a lesser tendency to shed and maybe resulting in lesser tendency to earlier detection also, or is that not something that's true?

Dr. Eric Klein: No, I don't think it has anything to do with the presence of driver mutations. The methylation signal that we see is a reflection of the perturbation of methylation in normal cells. So normal cells turn genes on and off using methylation. That's well known. Cancer cells exploit that biologic process of methylation by - in a gross oversimplification, but in a way that makes it understandable - they use methylation to turn off all the genes that prevent cell growth and turn on all the genes that allow cells to proliferate and get all these other biologic properties that make them invasive and so forth. So it's really important to understand that the test that was used in this study and that was developed in CCGA3 measures a shared cancer signal across multiple different cancer types. In CCGA3, we were able to detect more than 50 different individual kinds of cancers. It's a shared cancer signal that is fundamental to the biology of cancers, not just a specific cancer, but cancers.

Dr. Rafeh Naqash: I see. I think what I was trying to say, basically was, when we do liquid biopsies in the regular standard of care clinic, and you're trying to assess VAFs or variant allele frequencies for a certain mutation, you tend to see some of these BRAFs or EGFRs that are very low VAF, and the data that I've seen is that you treat irrespective of the low VAF, if it's a driving mutation process. If your VAF is 0.1%, you still treat it with a targeted inhibitor. The context that I was trying to put into this is it all depends on shedding. So this liquid biopsy that we currently use, whether other platforms that are out there, if you're not shedding as much cell-free DNA or circulating tumor DNA, you're probably not going to catch that subclone or clone that is a driver. So, does that play a role in your test also? If you have, let's say, a lung cancer that is an EGFR stage 4, if the shedding is low, following a general conceptual context that these driver mutation-positive tumors do have less shedding in general than the non-driver mutation-positive, would you think that would somehow impact the detection using your test or your approach?

Dr. Eric Klein: So, generically speaking, any test that looks for a cancer signal in blood is going to have a lower limit of detection. So there are analytic variables that make it such that, if you have extremely low levels of cell-free DNA or your other target shed into the blood, it's not going to be detected by the test. That’s an analytical issue. Having said that, it's important to distinguish the fact that this test that we're developing isn't really a liquid biopsy. A liquid biopsy, really, if you think about it, is on patients who have known cancer, and you’re doing a biopsy of the blood to determine if you can see a signal in the blood. This test has been developed to screen asymptomatic individuals who are at elevated risk of cancer, who actually may not have cancer. So we don't really view it as a liquid biopsy. But conceptually, you are correct that every test is going to have an analytical lower limit of detection so that not every tumor that sheds minuscule amounts of cell-free DNA will be detected. But that's not really relevant to this particular paper, I would say. It’s not really relevant to the performance characteristics that we saw in this population.

Dr. Rafeh Naqash: Understood. Thank you for differentiating the usual liquid biopsy approach that we use currently in the clinic, and this approach, which is meant more for detection in asymptomatic individuals.

Going to some of the results, could you highlight some of the interesting findings that you had in this paper as far as performance is concerned?

Dr. Eric Klein: Sure. Let me put it in a clinical context because we were just discussing asymptomatic individuals. That's what the test is ultimately meant for - screening asymptomatic individuals. But a common problem in oncology is this: patients present to primary care physicians with vague or nonspecific symptoms. Someone with COPD, for example, who presents with a cough, the cough could be due to the COPD, but if they have an underlying lung cancer, the cough could also be due to the lung cancer. Or someone presents with GI symptoms, could be related to cancer, or it could be related to a whole host of other things. And so there is a challenge for primary care physicians to sort out who might have cancer and who does not, particularly if they present with vague symptoms. In fact, most cancer diagnoses in the United States and Great Britain are actually found by primary care providers.

In this paper, we looked retrospectively, after the fact, in CCGA3, the case-control study that we did, to see how this methylation-based test performed in individuals who had symptoms that could be associated with cancer, or could be due to cancer, or might not be, might be due to other things. What we found was that the performance characteristics were as good or better in this symptomatic population, where the physician is facing a diagnostic dilemma, as they were in the asymptomatic population. This is really important, specificity false negative rate across all the patients in the study was the same as it was in CCGA3. It was 99.5%. Again, the false positive rate was only 0.5%. We found, however, that overall sensitivity was better in the symptomatic population, and it was 64% instead of, or as compared to 43% in the asymptomatic population. That is not surprising because some patients who present with symptoms are more likely to have cancer.

We also looked at a subset of patients who had GI cancers because that’s a very, very common presenting symptom in primary care practice, and this test performs exceptionally well for detecting GI cancers. We found that the overall sensitivity was 84%. Finally, and importantly, in terms of the clinical utility of a blood-based test to detect cancer and direct a diagnostic workup, what we call the clinical signal origin accuracy - the likelihood or prediction that a positive signal was related to a particular tumor type - overall accuracy in this population was 90%. So if you had a cancer signal detected and you had a clinical signal of origin assigned to it, let’s say, the test came back with cancer signal detected, the CSO prediction was GI cancer, the overall accuracy in actually finding a GI cancer was 90%. Actually, it was a little higher for GI cancers, but overall, for all cancers, it was 90%.

Dr. Rafeh Naqash: You mentioned that GI cancers had a very high sensitivity, around 84% or so. Is that, again, related to the tumor shedding compared to some other tumor types?

Dr. Eric Klein: Yes, there is a broad range of shedding across tumor types. So if you look at our data from CCGA, cancers like thyroid, prostate, and kidney do not shed a lot of cell-free DNA into the bloodstream, whereas GI cancers, hematologic malignancies, ovarian and pancreatic cancers shed much more cell-free DNA, and therefore their sensitivity for detection of those cancers is better.

Dr. Rafeh Naqash: What would be the alternate approach? Your sensitivity here is 64%, which is pretty good, but it's not perfect. So the patients who potentially would be missed using this test, what would be the alternate approach capturing those patients also and hopefully avoiding a missed cancer diagnosis?

Dr. Eric Klein: Well, it would be whatever the standard workup is that a primary care physician orders for someone who has vague symptoms. So, he idea here was to develop this, what we call a diagnostic aid for cancer detection in the symptomatic population. The idea here is to make the workups more efficient and to lend a greater degree of certainty as to what the diagnostic pathway ought to be. So, if you have a patient with vague symptoms and you're not sure if they are due to cancer or not, you might order a pretty broad diagnostic evaluation that might not end up finding cancer. In fact, if you take all the patients in a primary care setting, only about 7% of those individuals have cancer. Whereas, if you have a blood test that has a sensitivity of 64% and a positive predictive value of 75%, and you did that blood test early in the diagnostic workup and it was positive, you can do a much more tailored and perhaps a more efficient evaluation in speeding the diagnostic resolution.

Dr. Rafeh Naqash: As you mentioned, perhaps avoid unnecessary testing, which adds to the overall cost burden in the healthcare field.

Dr. Eric Klein: Correct. This was tested in another study called SYMPLIFY, which was done in a similar population of patients as this study - symptomatic patients presenting with vague symptoms or GI symptoms or weight loss, fatigue, those sorts of things, to primary care practice in the UK. And that was a prospective study. And the performance characteristics were very similar to what we saw in this study, although the overall positive predictive value in that study was 75% if you look at all cancers. And that would be very useful to a primary care physician and a patient to know what the likelihood of their having cancer is at the time they present or within a few days of presenting.

Dr. Rafeh Naqash: Absolutely. And perhaps, to complement this approach with some of the other diagnostic approaches, maybe the possibility of detecting cancer earlier increases. So this is likely complementary and not necessarily the one-stop-shop.

Dr. Eric Klein: It's important to understand that even in the symptomatic population, this is a screening test. And so, like all screening tests, if you have a positive mammogram that shows a nodule, you need to have a diagnostic workup to prove whether or not you have cancer. This blood test does not make the diagnosis of cancer; it simply helps direct a diagnostic evaluation that’s necessary to confirm whether or not cancer is present or absent. That’s true for both the asymptomatic and symptomatic populations.

Dr. Rafeh Naqash: Could you tell us a little bit more about the CSO prediction in the general context of oncology and NGS, or the whole transcriptome sequencing that we do these days? We often see on a report that says,“What is the likely tumor of origin?” if you have an unclear primary. Can you explain that in the context of the approach that you guys use for CSO prediction? How does it differ from methylation versus mRNA prediction of tumor of origin or cell of origin?

Dr. Eric Klein: Methylation has a rich signal in it, and it can distinguish cancer cells from a non-cancer signal, and using a second algorithm, specific methylation patterns that are specific to given lineages can identify lung cancer versus colon cancer versus liver cancer.

Dr. Rafeh Naqash: Understood. Do you see this as becoming an approach that could be used, using, for example, urine or other sources that we can easily acquire versus blood?

Dr. Eric Klein: Possibly. There is a lot of work in the field looking at urine-based markers for cancers, particularly, obviously, urologic cancers. And so there are already some products on the market made by other companies using methylation and other specific mutation patterns, for example, in urine to detect bladder cancer and to determine bladder cancer aggressiveness. It is an area of active investigation.

Dr. Rafeh Naqash: This is definitely an exciting field, and the way the entire field of liquid biopsies in general is moving as it’s detecting cancers or identifying mutations, and then implementing appropriate approaches, whether it is more screening or more treatment and all the drugs, etc.

Are there any other interesting future approaches that you guys are planning as part of this paradigm shift that I envision will hopefully happen in the next few years?

Dr. Eric Klein: Yes, as a company, Grail is focused on using this methylation-based technology across the entire cancer spectrum. So that’s screening asymptomatic individuals, it’s helping to direct diagnostic workups in individuals who present with symptoms to primary care practice, and also in the post-diagnostic space and all the possible uses there. So the detection of minimal residual disease and the decision on whether or not additional treatment is necessary, predicting response to particular therapeutic agents, or even choosing the correct therapeutic agents. All of that is under development.

Dr. Rafeh Naqash: Definitely exciting. Now, the last portion of this podcast is specifically meant to highlight your career and know a little bit more about you. Could you tell us about your career trajectory and how you shifted focus towards a biomarker-driven approach?

Dr. Eric Klein: Sure. Biomarkers have been a part of my career for a long time. I am trained as a urologic oncologist and did my residency in urology at the Cleveland Clinic and a fellowship at Sloan Kettering. At the dawn of the molecular biology era, the lab I worked in bought one of the very first PerkinElmer RT PCR machines for $5,000. It took up a whole desktop. I got very interested in genomic science at that time. So I spent well over 30 years practicing urologic oncology at the Cleveland Clinic, primarily focusing on prostate cancer. In the course of my career, I had the opportunity to work on a number of blood-based, urine, and tissue-based biomarkers. I have always been interested in understanding how our ability to measure molecules in blood and urine can help improve patient outcomes either through a streamlined diagnostic process or understanding of the biology of the disease better, picking the appropriate therapy, and so forth.

In the course of that, I worked with someone at a company called Genomic Health in developing a biopsy-based RT PCR gene expression assay that helped select men for active surveillance. That individual subsequently joined Grail and he came knocking on my door in 2016 when Grail was just getting started to tell me about this exciting new technology. He said, “This isn't about urologic cancers in particular, but would you be interested in helping us accrue patients for this big clinical trial we're doing, CCGA, and determine if this technology would be useful in some way in helping patients.” And being the curious individual that I am, I said, “Sure.” And so I helped accrue lots of patients to CCGA. The results were shared, and I was quite excited by them and continued to work with the company on other studies, including PATHFINDER and some others, and eventually became a consultant for them.

When I reached what I thought was the end of my clinical career by choice, I decided to step away from clinical practice, I had the opportunity to join Grail as a scientist, and that's where it’s been. And what I would say, in the big picture, is this: as a surgeon, I was able to help a lot of patients on an individual basis. So I did about 10,000 major cancer operations in my career. So I helped those 10,000 people. As an academician, I was able to make certain observations and publish them in a way that taught people about different kinds of surgical techniques and how they may work better, and so I was able to expand my impact beyond the patients that I actually touched.

When I heard about and understood what Grail was trying to do, I thought, “Wow, if we could develop a screening test that detects lots of cancers that we don’t screen for - about 70% of all cancer deaths in the US are from cancers that we have no screening tests for - and if the screening population in the United States, individuals between ages 50 and 79, that’s how CMS defined screening populations, well over 100 million a year, if this works, think about the impact that that could have.” That is really why I got excited about it. It fit my scientific interest, and I could see the big picture.

Dr. Rafeh Naqash: Thank you for giving us some insights about your personal career. It is definitely a very interesting topic. I learned a lot, and hopefully, our listeners will find it equally interesting. Thank you again for being here today.

Dr. Eric Klein: My pleasure. Thank you for having me.

Dr. Rafeh Naqash: Thank you for listening to JCO Precision Oncology Conversations. Don't forget to rate and review this podcast, and be sure to subscribe so you never miss an episode. You can find all ASCO shows at asco.org/podcast.

The purpose of this podcast is to educate and inform. It is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions.

The guests on this podcast express their own opinions, experiences, and conclusions. Guest statements on the podcast do not reflect the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.

MultiCancer Detection Test Performance in Symptomatic Individuals

JCO Precision Oncology Conversations