Can Molecular Profiling Lower Cancer Costs?

John L. Marshall, MD; Michael J. Pishvaian, MD, PhD; Kenneth Russell, PhD


December 24, 2014

Editorial Collaboration

Medscape &

This feature requires the newest version of Flash. You can download it here.

Molecular Profiling: Are We Ready?

John L. Marshall, MD: I'm John Marshall, director of the Ruesch Center for the Cure of Gastrointestinal Cancers at the Georgetown Lombardi Comprehensive Cancer Center in Washington, DC. Welcome to the 2014 Ruesch Center Symposium, "Fighting a Smarter War Against Cancer," where we bring experts together to do just that. At this year's symposium we are focusing on immunotherapy, but I want to do a flashback to last year's symposium where we targeted the concept of molecular profiling. Molecular profiling has entered your world, and I want to challenge us. Are we ready for molecular profiling, and what is the impact of molecular profiling on cancer cost? Could it in fact lower our cancer care costs over time?

Joining me are Dr Mike Pishvaian, assistant professor at Georgetown in the Hematology/Oncology Department at Lombardi Comprehensive Cancer Center, and Dr Kenneth Russell, director of clinical affairs international for Caris Life Sciences in Basel, Switzerland.

I want to set the stage for those who may be ordering molecular profiling tests and receiving reports back, and who may not be sure what to do with those reports. Can you give us a quick update? Where are we with molecular profiling? Is it genes? Is it proteins? Is it more than that? Is one company better than another? Is one test better than another? Is it ready for prime time?

Michael J. Pishvaian, MD, PhD: We are in an era in which we are able to profile not just the genes but the tumor cell in general—either by genomics (looking at the gene construction of the tumor cell), proteomics (looking at the enzymes that are acting), or phosphoproteomics (looking at which enzymes are active and inactive) and with newer technologies, such as RNA sequencing and metabolomics—to try and understand what is driving the cancer cell. Many of these tests are ready for prime time and are even in the retail market. Our ability to order these tests has become easy.

That doesn't necessarily mean that we should order them all the time, because part of being able to order them is being able to act upon the results. Getting actionable test results that would benefit our patients is something that we need to take into consideration. That is where much of the challenge comes in molecular profiling. We don't necessarily know whether everything we are finding out about the tumor cells is actionable, meaning whether it will make a difference in the patient's disease process. And we don't necessarily know how things mingle. For example, how do chemotherapy predictive markers mingle with genetic predictive markers, or with phosphoproteomic predictive markers? A tremendous amount of research needs to be done, but at the same time, these things are available for patients who are suffering from advanced disease today.

Newest Tumor Is Best to Test

Dr Marshall: Do I need new tumor tissue, and how much do I need?

Dr Pishvaian: I'm definitely of the camp philosophically that believes that the newest tumor that you can get access to is the most representative of that patient's cancer today. In a patient who had a colon cancer removed 5 years ago and whose tumor came back 2 years ago after the patient was on 2 years of chemotherapy, the tumor tissue from 5 years ago is probably not reflective of the patient's disease today. You should look at what is going on in the tumor today. How much tissue you need depends on the company or the internal process that it is testing. Some companies do well with very little tissue, and some companies need a lot of tissue. Obviously, the philosophy would be to use as little tissue as possible to get as much information as possible about the patient's tumor that we could act upon. That's ideal.

Dr Marshall: And the issue of tumor heterogeneity—we have seen evidence that you can biopsy one side and get one result, and get a different result from the other side.[1] How do you deal with that?

Dr Pishvaian: That is the 800-pound gorilla in the room that many of us don't even know to deal with. Tumor heterogeneity is much more at the genetic level, so many genes differ from one core to another within the same tumor. But we don't necessarily know whether that changes the phenotype, or how the actual cancer behaves. We certainly see patients who have a certain kind of mutation. In an epidermal growth factor receptor (EGFR) mutation in a non-small cell lung cancer, usually most of the tumor shrinks down in concert. Although there is a tremendous amount of heterogeneity, how that actually influences patient outcomes in response to a specific targeted therapy, we still don't know.

Dr Marshall: Part of our drill-down today is on cost, but you have just suggested that we need a new biopsy, usually CT-guided or similar, so there is risk to the patient, but also cost. We are doing this in the trenches. How is that getting paid for? Has that been a problem for you, and for us as a team?

Dr Pishvaian: We have not had much trouble getting it paid for so far, whether that is right or wrong. I am biased and think it's right. An insurance company should be willing to pay for something that will change the patient's management. That is why we are doing these biopsies, to see what we need to do to change that patient's therapy to maximize their options. So far it has not been a tremendous problem.

There is a problem, or at least there is education that has to occur on the part of the people who are doing the procedures. A good friend of mine who is an interventional radiologist told me that for his entire training, he was trained to use as little tissue as possible to make the diagnosis. Now we are going back to him and saying we need as much tissue as you can safely get from patients without harming them. It is a different mindset.

What Extent of Mutation Informs Treatment?

Dr Marshall: These assays are evolving too, right? What is normal, what is not normal, what is positive, and what is negative? Tell me a little bit about what we know.

Dr Pishvaian: Most of the people who are doing the assays, whether at the academic centers or retail companies, compare the result to a broad population norm. Whether people are looking at a protein (such as thymidylate synthase) or a gene (such as KRAS), they have a sense of what the normal gene should look like across a population, and they are looking for abnormalities. Much of the responsibility of the physician, the companies, and the academic institutions across the world is to keep looking at the evidence behind that and seeing whether we need to shift the bar on what we call a positive vs a negative finding.

I had an experience recently with one of the companies who said that they had looked at their internal data, and realized that they needed to shift the bar way down on what they are calling a negative to translate that into something that is useful for patients.

Dr Marshall: On the gene mutation side of things, is it one copy? Is that enough when you have a mutation, and that represents the entire tumor, or do we need higher numbers of copies? Right now, we seem to be on the low end.

Dr Pishvaian: One of the blessings of next-generation sequencing is the ability to use small amounts of tissue to pick up DNA mutations. The downside is that it can be too sensitive. Many of these blood-based tests that are picking up mutations with a blood sample of circulating DNA are super-sensitive. That's great, because you don't have to bother biopsying the patient's tumor, but it could also be overly sensitive and not representative. We just don't know yet.

Dr Marshall: As clinicians, we get these reports back. What is your level of certainty in using this evidence to either pick a medicine or say, "No, we are not going to use that medicine"?

Dr Pishvaian: Much of the data that we have come from large trials, usually retrospective, with the KRAS story in colorectal cancer being one of the best examples in which we had a large dataset,[2,3,4,5] and we went back and looked at those data. It was a great story because there was a very stark difference in response to EGFR therapy and KRAS–wild-type vs KRAS-mutant patients.[6] Unfortunately, our ability to scan a much broader gene profile or molecular profile also means that we are picking out needles in haystacks and a lot of anecdotal examples. We just need to keep building those anecdotal examples until they become larger subgroups.

The best application of this approach will be the NCI MATCH study,[7] in which they will look at a group (not a tremendously large group, but 20-40 patients) with a certain kind of mutation and treat them with what we think is the right drug, and look for at least an initial signal. It is similar to the I-SPY 2 trial model.[8] Dr Donald Berry, who worked on I-SPY, is once again involved in the statistics of MATCH. The MATCH study will be very informative, and it is going to happen very quickly.

Making the Educated Guess More Educated

Dr Marshall: Now for the hard part. We are excited about this. This is cool technology, and it is starting to make its way into the mainstream. We even see the American Society of Clinical Oncology (ASCO) officially saying that we need to be doing this[9] and to figure out how to get access to the medicine. We like this scientifically and medically. How do we justify the added cost of the molecular testing? How are we looking at this as a population process? How do we convince our payers and supporters that this is not going to cost more, that this will in fact save us money in the end? You have been doing lots of work around the world, outside of the United States, on this. Share with us where we are on this because you may represent one of the leading forces in this discussion. Take it away.

Kenneth Russell, PhD: Whether it will save or cost money is the million-dollar question. It's not even the 800-pound gorilla in the room. We have to think not about money wasted—because molecular profiling is providing valuable information—and look at it as money invested.

We have looked at a number of physician-led studies from which we are generating information on how broad molecular profiling is used. These are done in patients who are past the standard of care, so they are either refractory to the standard-of-care treatment options, or they have a rare cancer for which the treatment options are limited.

These are difficult patients to treat because the treatment decisions are empiric, meaning that we pick a therapy from what is available or from another tumor type on the basis of the physician's experience, the individual patient's performance status, how the patient responded to treatment in the past, and what the patient has been treated with. All of these factors go into a physician's decision. We are saying that by using tumor profiling, we can make this educated guess more educated, and we can use this extra information to help in the treatment decision.

In taking a broad approach to this—where we look at protein changes using immunohistochemistry, gene expression, in-situ hybridization, the position of mutations with next-generation sequencing, and other tumor profiling tests—we can build a comprehensive picture. By doing this, you can get the widest view, which will allow you to take information from other tumor types, to learn from your successes in other tumor types, and the anecdotal data that are building up. All of these factors contribute to showing that there is utility for this kind of approach in treating on the basis of tumor process rather than tumor type.

When all of this is taken into account, the report that we generate is actionable in about 95% of cases, and it is actually making an impact on the physician's decision in about 85% of cases. They are using the information to change what they actually administer to the patient. Obviously, that is important information.

The other thing we see is that most physicians will use chemotherapy rather than an expensive targeted therapy. The most popular treatments given are such cytotoxic agents as 5-fluorouracil (5-FU) and irinotecan. The agents that were actually given after the tumor profiling were predominantly the same cytotoxic agents—5-FU and irinotecan—but in a reallocation of the available chemotherapy resources to patients who are more likely to benefit. You see some patients in whom it is confirmed, some in whom it is given when it wasn’t previously considered, and some in whom it is avoided.

The idea of avoiding chemotherapy in patients in whom it is less likely to work and show benefit is where you will see some cost savings. Obviously, you are treating another patient in the population, but the important thing is the reallocation. Even if molecular profiling proves to be cost-neutral or slightly adds cost, we need to look at whether this cost is being offset by improved treatment outcomes, such that the patients who were using the resources that we have are more likely to benefit and we are avoiding it in patients who won't benefit.

Removing Costly Uncertainty

Dr Marshall: That is exactly what I was thinking. In some places around the world, a patient would not necessarily receive further therapy, or might just get "doctor's best-guess therapy," right? In a way, we are adding cost because we are seeking new options for a patient who has run out of options. If we find treatment options, the biopsy, the molecular profile, and that additional treatment add cost. Are we under more pressure to define the value of the treatment in that setting, to show that that treatment is having a positive impact? What are governments and payers expecting of us on that level?

Dr Russell: We need to test this prospectively and measure it. It's difficult to create a treatment model around tumor profiling, to go to a whiteboard or a blackboard and work out what the treatment algorithm should look like on the basis of outcomes. You need to actually use it. The payers want to see decision impact. If you are changing what the physician would have done, there is value in that. If you are having impact on a system or an approach, you are helping because you are removing the burden of uncertainty, and uncertainty is looked on as being costly by payers.

Sometimes the treatment decision is supported by evidence from another tumor type. The drug has not been approved for this indication, because in rare tumor types it is impossible for a pharmaceutical company to conduct a study in that patient cohort. If we can show using the best profiling that we have that the tumor is potentially susceptible to a particular treatment, and we put this together with one or two case reports, that supports the decision. That is what the payers are looking for. They are looking for an added degree of certainty.

Proposal to Pharma: Free Meds for Data

Dr Marshall: A very exciting position paper by Rich Schilsky,[10] presented at ASCO, said that we need to do this, and we want the pharmaceutical companies to provide us with the medicines free of charge in exchange for information to the payers, to the profilers, and to the physicians about whether it improved outcomes. In a way, it is innovating new outcomes. Is this where you see us going as a way to justify the added cost, by proving that added benefit? Did you like that position paper from Rich?

Dr Russell: Yes. Showing the added benefit is key, and we can only do it if researchers, tumor-profiling companies, and the pharmaceutical companies can work together to bring this forward. We need to also engage with the payers and look for the information that they are looking for. Only with a broad multidisciplinary approach are we going to actually prove this. We could be here for the next 20 years doing something that helps patients today and trying to answer a specific question. We could invest in a 3-year study, but then in 3 years we could be looking at completely different technologies; a completely different battery of tests would be available, with companion diagnostics, and drugs as well.

At the moment, we find with next-generation sequencing in particular that we can find a lot of actionable changes and alterations in the gene makeup that aren't "druggable." We might look back in 5 years and say, "We were able to find that mutation, but we never had the drug for it." Do we need to prove the health economics of this? Is it something that we can just work on today and prove forever or is it something we need to continually look at? These are all questions that we need to work out.

Dr Marshall: Can you share your experience with the National Health Service (NHS)? We think of the National Institute for [Health and Care] Excellence (NICE) as being one of the most discerning and strict groups for accepting expensive new therapies. You were sharing with me that you have been working with NHS to try to understand how they would incorporate molecular profiling.

Dr Russell: We have worked with the NHS and applied for a regional innovation fund in the UK to look at a group of patients with rare gynecologic cancers to see how tumor profiling can help these patients with their treatments. Obviously, NHS is interested in the impact that this will have on cost. We will work with them to show the value of the information and how it is improving outcomes, but also to provide some concrete evidence of what the added costs are. If you find a mutation that you weren't expecting, and there is a valuable drug against it (for example, a BRAF mutation where you were not expecting it), that will add to cost of care for that patient. But across the whole population, is it having an impact on resources, and is it something that can fit into practice?

La Carte vs Pan-Profiling

Dr Marshall: We are used to à la carte molecular testing—KRAS a year ago, (now it's extended RAS). We are ticking off à la carte mutations, and for each one of those events, a slide needs to be cut and material tested. Are there cost savings in the pan profiles that are available? We are learning about targets as we go. What if I already have it in the bank, and I find out 2 years from now that it is a good target? Are there cost savings that we can factor into our analysis?

Dr Russell: Yes. The illustration that you gave of the pan-RAS testing is very good. KRAS mutation has traditionally been looked at for cetuximab resistance, but you can also find NRAS or HRAS mutations that could be an escape mechanism from blockade. You are not just looking at the KRAS mutation; it could be that an NRAS mutation is driving the disease, and that is why you won't see effectiveness of an EGFR inhibitor. We also know that if we are looking at only one mutation—for example, if we try to find BRAF mutations in a population that has a 3% mutation prevalence—it's not sustainable to just look only at that one test and have 97% of your results prove noninformative for those patients. You need to take a broad view to catch that 3%, and also have something to help the other 97% of patients.

By taking as broad a view as possible, which we can do today, we might look back in 2 years and have that information in the bank. We can go back and say, "There is a new drug, so why don't we put the patient into this clinical trial," or "Why don't we use that information?" As I started by saying, it's an investment in what we can have for the patients. It's always going to be changing, it's always going to be broadening; but by doing the best that we can today, we are securing that option.

Dr Marshall: Even though we are doing 600 genes in next-generation sequencing, and lots of proteins and now phosphoproteins, are we still looking under the streetlight? Are there other genes out there that we are not testing, that we are not measuring? Or do we have most of the cancer-relevant genes under our umbrella now?

Dr Pishvaian: There is a lot that we don't know. There are 400 or 600 genes that have been deemed cancer related. There are 30,000 genes out there, and certainly when whole genome sequencing anecdotal presentations have been done, they find more to act upon that wouldn't necessarily have been captured in cancer-related exome sequencing or even whole exome sequencing. Someday we may get to a point when we are trying to capture all of the genes as well as all of the intronic sequences that can affect the ability of the genes to be transcribed appropriately. We don't necessarily look at phosphoproteomics right now—we look at the expression levels of certain key enzymes—but to look at what is actionable or activated in that phosphoproteomic signaling pathway may also be relevant. There are also thousands of those in the kinome.

Build the Haystack to Find the Needles

Dr Marshall: We are excited about this, but I can't help but feel that there is a risk that we might be going down the wrong path. Give me a sense of your assessment of that risk. If we keep doing this over the next 5-10 years, we can see how it might help. What are the risks that we are not going in the right direction?

Dr Russell: We have to remember what else we can do for these patients. For many patients, we won't have any other information that we can use. We can keep doing the same things, like the polar bear in the zoo who goes insane doing the same thing over and over, or we can try to use the best information that we have. The information is always changing, and the information that we are generating about the approach is always changing.

The initial sense that we have is that it is helping patients. We have seen evaluations by physicians; they have looked and have shown better outcomes. We need to start looking at that in more detail and become more confident with it. The thing that molecular profiling can offer a physician today is confidence in how they are treating their patients. You might look back in 5 years and say, what was I doing? At the time, however, you were using the best information.

Dr Pishvaian: The biggest obstacles might occur if we don't study this properly. I don't necessarily mean a detailed prospective clinical trial, although that would certainly be the gold standard, but we are going to be profiling lots and lots of patients, and we want to capture all of the data. The burden is on every one of the physician oncologists in the country and worldwide to collect those data, because many of us are using these reports to use off-label therapy. And if we are not telling anybody how those patients did, if we are not formally putting the information into a database somewhere, we are losing a valuable piece of information.

Dr Marshall: All that work that we do all week long of entering data into our electronic medical record—are we building the haystack?

Dr Pishvaian: Yes. We need to build that haystack big so that we understand where the needles are. And sometimes it's really important because there are key actionable mutations that don't mean anything in certain disease types. There are anecdotes that we think we have the actionable gene mutation, but for some reason in this disease it doesn't work, even though it worked well in another disease.

Dr Marshall: This has been terrific. Thank you very much. I hope our audience has gotten something out of this thought-provoking discussion and is inspired to begin incorporating this kind of process and doing the data entry that is so important for us to understand outcomes.

Ken and Mike, thank you so much for taking the time out from the symposium to join us and talk about what is going on in cancer care today. This is John Marshall, reporting from the 2014 Ruesch Center Symposium in Washington, DC.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: