ASCO President Peter Yu, MD, on Big Data, Big Themes for Upcoming Annual Meeting

Gabriel Miller


May 27, 2015

Editor's Note: In advance of the 2015 Annual Meeting of the American Society of Clinical Oncology (ASCO), ASCO President Peter Yu, MD, spoke with Medscape and provided a brief primer on big data, one of the major themes of this year's meeting program. He also outlined the future of CancerLinQ™, ASCO's signature big data project.

He also highlighted a number of important meeting themes for ASCO attendees and made the case about why oncologists are exactly the specialists to lead medicine in the development of big data.

Medscape: Why did you choose "Illumination and Innovation: Transforming Data into Learning" as this year's theme?

Peter P. Yu, MD

Dr Yu: We are in an information age and there is a lot of information coming at us, a lot of data coming at us. Any physician who has gone on to electronic health records (EHRs) has become a 24/7 person. Data are so readily available that as soon as you order a test, it's virtually done and the results are back. You have to process that and decide, "What am I going to do about this?" It used to be that you would have a few days between ordering a test and when it came back to think about what you were going to do about it. Now it's instantaneous. Doctors are being inundated by data, data that they in fact generate themselves. The real question is, what are you going to do with all of that data and how do you make sense of it?

By the same token, in terms of precision medicine and molecular medicine, the amount of scientific discovery has greatly accelerated. Again, we are getting bombarded by data. If you order a next-gen sequencing panel on a tumor for a patient because you are not sure what the next best treatment might be, you may get a report of 300, 400 different genetic hotspots reporting on maybe 10 or 12 different mutations. Again, it's great to have that data, but what does it really mean to you and what does it mean to your patient? So we need to become smarter and more nimble about how we look at data and derive knowledge from that understanding, and then that drives actual decision-making that improves patient outcomes. The next generation of oncologists is going to be dealing with more and more data and trying to understand how we go about creating knowledge out of that.

That's what we do at the ASCO annual meeting. People come with data that they've obtained from their research work over the past year and in Chicago they present their data. They present not only the data but their conclusions—the knowledge that they've gleaned from their data—to the cancer world to say, "Do you agree with me or not?" Oncologists will look at that and come away, and in some cases these will be practice-changing presentations that will translate into something they've learned.

But I believe that data is just the start of it; we need to go through a process of understanding, whereby we develop knowledge and models about what the data are telling us, and then go through a process of intellectual discussion and discourse at the annual meeting. Then people come away from this with certain conclusions, and that is really their learning. So, what we do at the meeting is take data and transform it into practical learning. Only now we're doing this in the context of ever-increasing amounts of data from both a molecular point of view and a clinical point of view.

Medscape: "Big data" has been a buzzword in oncology for at least the past few years. Yet among community oncologists there's still a bit of confusion about what the phrase means. For practicing oncologists, what does "big data" encompass?

Dr Yu: It is a buzzword, and like any buzzword after it's been around for a couple of years, it begins to lose its meaning and becomes just a catchall phrase. Right now I would say that there isn't a lot of meaning behind the term "big data" because it refers to the ability to create large datasets and to mine that dataset. At this point, we've not really created those large datasets. We have EHRs; that's a digital representation of data. That's the first step.

But EHRs are not connected with each other. There is poor interoperability, meaning the ability of one computer system to talk to another computer system. Even if you had that—which we don't have right now—what is the foundation for doing that? Who is the honest broker? What's the mechanism for aggregating all of that data?

It's a fairly meaningless term right now to say "big data" in oncology. That's why CancerLinQ is being developed. ASCO is creating a platform whereby doctors can start sharing data. We want to create a large dataset that we can then mine for big data purposes. To do that, ASCO has set up CancerLinQ, which allows EHRs to be transferred into ASCO's database. The data will then be de-identified and aggregated. Then we will have created big data, allowing us to mine that data for learning purposes.

Medscape: What are the essential ideas or concepts that community oncologists need to know to get up to speed on big data?

Dr Yu: They should go to the exhibit hall where ASCO is having a special booth to explain and present this to our members—to explain the concepts of sharing data, creating large databases through CancerLinQ, and developing the ability to mine that data.

Data mining can be used for several purposes. It can be used very practically for a practice to analyze how it's performing on benchmarks compared with national averages. We could present the aggregate performance of oncology doctors and then how you specifically do on a given measure. That benchmarking can be used for quality improvement, of course. It can be used to analyze practice efficiency. For example, what's the degree of variability from doctor to doctor within your practice in terms of things like ordering PET scans or MRI scans, which drive up your cost of care? How much variation is there and are there potential areas that could become more efficient?

From a practical, practice point of view, it drives quality analysis as well as cost efficiencies, which will be essential for a practice to survive. As we learn best practices from those who are able to achieve the best outcomes at the least expenditure, we can then provide clinical decision support to help guide doctors when they're making decisions about patient management as to what they might consider as alternative approaches. From a practice point of view that's the pragmatic aspect.

This can supplement the knowledge that we get from clinical trials. Clinical trials are our bread and butter. They are how we frame up a question and how, in a scientific manner, we compare different treatments or different arms. That's really the gold standard, level-one evidence—the randomized clinical trial.

But you can't run a clinical trial to answer every important question. We need to have a more practical way of learning from what is actually done and not just rely on clinical trials, because they don't answer all of our questions. Clinical trials are long and expensive, so we don't have the bandwidth to ask every question. The patient population in a clinical trial is also a very homogenous population because of the eligibility and ineligibility requirements. When you try to apply that to real-world patients, very often they don't fit the criteria of the clinical trial. You have a result that may not be applicable to general patient populations. In practice we see this when oncologists use dose reductions, or they change the schedule and really don't know what that may do. Big data may allow us to supplement the knowledge we have from clinical trials and more rapidly understand how to treat real-world patients.

Medscape: There are examples of big data tools being used at places like Memorial Sloan Kettering and other well-known academic centers. Are there clinical tools that are available right now to community oncologists that harness big data and that they can access immediately?

Dr Yu: I don't think so. There is a gap that has not been addressed at this point.

Medscape: In light of that, what would you say are the main obstacles that have to be overcome in order to harness the data that you were just speaking about and make it available more widely?

Dr Yu: Practices first have to decide that they want to share their data. There are privacy issues, so physicians are obligated by HIPAA and by our ethical code to preserve patient privacy. With sharing data, the first questions you have to ask are, "Do I trust the person I'm sharing with? What are they going to do with that data, and who is going to have access to that data?"

We at ASCO feel that we do have the trust of the doctors. We're their organization. They can trust us because they are us and we are them. I think that's the most difficult thing—saying, "Do I believe in sharing data and do I feel safe doing so, and who do I believe I can share it with?" That's number one.

Number two is technical: How do we do this? How do I wire up my computer to download the information into the database? How much work is it going to take? ASCO has said that this has to be done in a way that isn't going to be very onerous or put a lot of work on the backs of doctors to provide their data. We're taking the data in as is. We're trying not to require practices to fill out special forms or to reconfigure their data in any way or change their EHR. We're bringing the data in as a data transfer. The cleanup of the data, the structuring of the data, making sense of it, mapping it to a common data format— this will happen on the ASCO end. We're making this very easy, the transfer of data, on the part of the practice. We're not requiring them to do a lot of work on their end. It's all going to be done on the back end. The data come in to ASCO and then we map that data to accommodate a structure so that it can be mined.

Once we've created this large dataset, we will then begin to see what kind of useful information we can learn from this and return to the practices. Some of the things are relatively straightforward, like the Quality Oncology Practice Initiative (QOPI) measures. We've been working for 10 years on QOPI and we have almost 200 measures of quality performance. Those are being put into a machine-readable format so that we can mine these out of the data. These we can turn around immediately and report back. We also have our vanguard practices, which are 15 practices that vary from small community practices to large hospital systems to cancer centers. We're working with the doctors in those groups to decide what is useful for them. What is it they want to know? What is possible? What is extractable? We can then generate the type of support tools that the doctors and practices feel are most important to them.

Medscape: Could you give us an update on the status of CancerLinQ and what we might expect for the rest of 2015 and beyond this year?

Dr Yu: The CancerLinQ prototype was released last year. This year we announced in January that we had started work on version one. We also announced in January that we are partnering with SAP, the large enterprise-data company. It's a German company with offices worldwide, and they are the world's leading software company for managing big data—enterprise-level data—in many industries, especially the finance industry. We evaluated the different technical platforms that we might build CancerLinQ on and came to the conclusion that SAP had the most robust product to do that.

We then approached SAP, and in January we both agreed that this was a strategic initiative for both ASCO and SAP, that it wasn't going to be a client/vendor relationship. We are going to work in co-development. We started in January taking their platform, which is called HANA, and building that out for medical use. The timeline is that we will complete that platform in the third quarter of this year. Simultaneously we have legal agreements with 15 of these practices that I mentioned earlier, our vanguard practices, which will transfer their data into the platform when it's ready in the third quarter. By the fourth quarter of this year we will have launched CancerLinQ with, we anticipate, about half a million patient records in this platform. Then we will begin to generate reports and learning out of that. We are following an agile development process with SAP. We expect every few months a rapid turnover with updated versions of the product. The idea is to move the product development very quickly over the timeframe of the following year. To answer your question, CancerLinQ will be built and launched by the end of this year, in 2015.

Medscape: There have been a few critics of the impact that big data stands to have on oncology. They often liken the promise of big data to that attributed to EHRs, and in doing so they feel that the promise of big data has been a bit oversold. Why should physicians who feel jaded about the influence of EHR technology on their practices believe in the promise of big data?

Dr Yu: I heard an interesting comment that was attributed to Bill Gates last month. I'm not sure whether he really said it, but it's a good quote. The quote was that people overestimate what can be achieved in 1 year, and they vastly underestimate what can be achieved in 10 years. Human nature, and maybe the hype that goes along with new things, makes one believe that something will work right out of the box the first time. What we've learned from EHRs is that they're pretty complicated and they're not going to work in the first year or even the first few years; they need some time to develop.

One of the reasons why EHRs have failed is because they did not start out from the beginning with the idea of sharing data and interoperability. Every vendor of EHRs created their own system and didn't want to share with anybody else's system because they were protecting their proprietary interests. Products were not developed together and had different functionality, and we weren't really going very fast because everybody was pulling in their own direction. The way to get beyond that is to think about sharing information and developing standards. I don't think the industry by itself is going to do that. I think it needs somebody, whether it's ASCO or a federal agency like the Office of the National Coordinator for Health Information Technology (ONC), to set the standards and pull people together. The future requires this kind of step forward. Whether a practice wants to be a vanguard practice, as we call them, and be at the front and help drive that, or whether it prefers to hang back and wait for the dust to settle—each practice will decide on its own. I think that this is inevitable. It's necessary and it's very, very important that we move ahead with this.

If we throw hands up and say, "It didn't work, I'm giving up, and it will never work"—that's just not an oncologist. Oncologists don't think that way. We wouldn't be oncologists if we said, "We didn't cure cancer in the first 10 years, so I'm giving up; I'm walking away." Can you imagine any oncologist saying that? It's just not in us to say, "This is too hard, this is too difficult, and it can't be done. I'm just going to live with the status quo."

Medscape: It might be why the field is leading in this discipline.

Dr. Yu: I think it is. It's always amazing that engineers love oncologists. I think it's for many reasons. One is what we just said. Second, I think we are very driven toward protocol thinking, logical thinking, which is what engineers need to have. They need to have somebody who will sit down and say, "I'm going to do this next and then that, and then if this happens, I'll do this." Then they can map out how to build it to do that.

One of the problems with EHRs is that doctors haven't been engaged, and engineers have been making assumptions about what they think doctors want. Of course, they created something that is not the right thing. Unless doctors themselves articulate what they want, the engineers are going to flounder. If they flounder, we're going to have lousy EHRs.

Medscape: Naturally you had an enormous guiding hand in the 2015 ASCO annual meeting program. What are you personally most excited about for this year's meeting?

Dr Yu: Of course there's always the new hot science, the new drugs that are going to be announced that appear to have significant breakthrough achievements. A lot of these drugs are already FDA designated as breakthrough drugs, so the FDA has already been saying these are promising and the results look good. We're going to accelerate the approval process. We're going to see several drugs that have been earmarked by the FDA and we've been waiting for their results. And I think they're going to be results that will make the FDA say, "Okay, we made the right call here. We will continue to accelerate this." There are several new drugs that are going to be in that class.

We're still exploring the boundaries of immunotherapy and the checkpoint inhibitors, and trying to understand whether this is a strategy or a class of drugs that will be effective across many cancers, or whether it is just a few cancers where immunotherapy is particularly important. We're starting to see some signs that this is a more generalizable strategy across cancers, which is very exciting. That will continue.

We will continue to focus on the value of new therapies in an era where the cost of care is threatening to make this an access problem. What good is it to have really great drugs if they're so expensive that people either go broke trying to get them or can't afford them and just can't get them? We're realizing more and more that with this bounty of new drugs at the prices they're coming out at, we really need to look carefully and say, "It's a great drug but where is its niche? It's not that everybody needs to get this drug. Where does it really add value?" That's something we started talking about last year, and we will continue to discuss that and educate our members about how this needs to be addressed as well.

We're looking more and more at global oncology, at healthcare in low- and middle-income countries. Again, it's the extension of the idea that " It's great to have wonderful therapies, but if people can't afford it or there isn't the infrastructure to deliver it, then we really haven't done the job." That's where I developed this theme of transforming data into learning. Initially I thought about calling it transferring data into knowledge, but then I realized that knowledge really isn't the answer. You may know something, but if you can't provide it to a patient then you really haven't learned anything. So you need to go beyond just knowing what you should do. You have to understand how to actually deliver that knowledge to benefit patients and improve patient outcomes. Until you do that, you really haven't learned anything. So we're going to be focusing on those situations. We have a special seminar on Friday afternoon on global health to look at that aspect of care delivery.

Our keynote speaker this year, Michael Porter, MBA, PhD, of Harvard Business School, is a healthcare economist. That's a first for us.He said we have to hammer on this theme: What are the outcomes that matter to patients? How is it from a patient perspective? We may think it's a wonderful, great new drug, but unless the patient lives longer; feels better; and has a more meaningful, productive life, it really isn't that great a drug. We need to think about how we evaluate that and how we reduce the cost of care so that we can deliver the care to more people.

Dr Yu has disclosed no relevant financial relationships.


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.