Conversation with The Cancer Letter
Hudis: ASCO Vying for Big Role in Big Data
The American Society of Clinical Oncology is in the final stages of developing a request for proposals for its CancerLinQ system.
The RFP is expected to be issued early in 2014.
ASCO made the announcement last week at a White House Office of Science and Technology Policy event called “Data to Knowledge to Action: Building New Partnerships” at the Ronald Reagan Building in Washington, D.C.
“ASCO is not a business,” the professional society’s president, Clifford Hudis, said to The Cancer Letter. “If our participation in the market ultimately drives the development of a superior tool, but from somebody else, then I would say we have accomplished our goal.
“We are an honest broker,” said Hudis, chief of the Breast Cancer Medicine Service at Memorial Sloan-Kettering Cancer Center. “Our goal here is to get this job done and to profit, in terms of society and societal benefit, the maximum number of people. If we support somebody who can do better, that’s fine. We just need this to happen for the good of the patients we serve.”
The multi-phase development process will encompass a series of quality improvement tools for physicians, with the first components becoming available by early 2015.
The total cost of the project’s phases is uncertain.
ASCO is seeking financial support through its Conquer Cancer Foundation, which has raised $7.8 million in commitments from Amgen, the Chan Soon-Shiong Family Foundation, Genentech BioOncology, Helsinn Therapeutics Inc., Eli Lilly & Co., Novartis Oncology, and Susan G. Komen for the Cure, as well as numerous individual supporters.
ASCO’s move into the full build follows completion of the CancerLinQ prototype, which demonstrated the feasibility of the system.
The first components of CancerLinQ will center on providing next generation quality measurement that builds on ASCO’s Quality Oncology Practice Initiative.
Future components will encompass more powerful quality improvement tools, real-time clinical decision support, and analysis of thousands of patient experiences to create a continuous cycle of learning.
It remains to be seen how ASCO’s participation will shape the power structure in bioinformatics in oncology, or whether it would be able to raise the funds the project would require.
Another umbrella group, National Comprehensive Cancer Network, which develops practice guidelines and pathways in oncology, is vying to expand its role as big data reshapes the field.
For-profit players, whose expertise covers at least some of the territory CancerLinQ would span, include the drug supplier McKesson, which has an alliance with NCCN to create clinical pathways and to produce software that will allow physicians to assess treatment options consistent with evidence-based standards.
The pathways and supporting software will also allow providers to consult coverage policies mandated by payers (The Cancer Letter, Nov. 30, 2012).
The market for clinical pathways and decision support and review systems, one of the segments of CancerLinQ’s activities, is competitive.
Players include the P4 Pathways, owned by CardinalHealth; Via Oncology Pathways; Eviti Inc.; ICORE Healthcare; and others.
Content libraries and software from these firms are used by a number of regional and national payers. IBM is also developing a decision support system using its Watson technology.
Other companies—including software developers, Internet companies and drug manufacturers—stand poised to jump in, insiders said.
Hudis spoke with Paul Goldberg, editor and publisher of The Cancer Letter.
Paul Goldberg: I guess what I am asking for is a tutorial: An idiot’s guide to CancerLinQ.
Clifford Hudis: That’s easy enough.
One: The simple issue is that 97 percent of adults in America who get treated for cancer—especially solid tumors—get treated outside of the context of clinical trials, and we—society—learn nothing from them.
Two: Over 60 percent of patients who have cancer in the U.S. get their care using electronic medical records. So the obvious opportunity is to start to mine all of that typing and clicking for outcomes, associations, safety—all the things you’d want to learn from medical records if you could mine them the way Google mines clicks and Amazon mines shopping patterns.
CancerLinQ is an attempt to do that. At first, you might say, isn’t that pretty straightforward, you take the data and you just crunch it. The problem is we don’t have standards for how the data is collected and stored in the various EMRs, so CancerLinQ is an attempt to solve that lack of interoperability. And, indeed, we’ve done a pilot study, which was able to demonstrate that we could draw data in from disparate EMRs, and then we could actually convert the data, analyze it, interpret it, and provide output.
And the output takes s several forms: quality assurance, but also feedback to doctors in the form of clinical decision support, and, of course, aggregation of outcomes in terms of both safety and toxicity.
So, that’s the thumbnail.
PG: This is massive. How much would it ultimately cost?
CH: Many millions of dollars.
PG: Many, as in 10?
CH: Many multiples of 10, many, many multiples.
CH: It’s all estimates at the moment about what the ultimate cost will be, because, of course, it’s all question about how much of it will be built. But to start with, I think it’s fair to estimate somewhere in the range of $80 million just to get this going through the first five years.
PG: What does $80 million get you to?
CH: That would get us, two years down the line, a stand-up product that would be in use, and it would get us, within several iterations over five years to version 2.0.
PG: And the money would come from?
CH: The money comes, obviously, from all the sources of revenue that ASCO currently has, and from the money that ASCO has been able to build up into its foundation over the years.
PG: You don’t have that kind of money; do you?
CH: Sure we do. Over time.
CH: It’s a little bit of a double-edged sword, because we don’t really think that ASCO should pay for it all alone. We think this has a huge potential societal benefit. We think that partners in pharma who will be able to benefit from some of the efficiencies that this provides should see the value in supporting it.
We, frankly, think that philanthropists should see its value. Maybe the government, if they had the money.
Here is a concrete example of what I mean. Right now, how do you do post-marketing drug surveillance for safety? You rely on voluntary reporting by scattered physicians who are or are not part of the clinical trial system, and it takes you forever to see important safety signals.
The poster child was Vioxx in terms of coronary artery events. It took about 62 months from approval until the drug was withdrawn from the market because of this. We estimate that if actually everyone were being cared for in a system like CancerLinQ, we would have seen the signal in six to eight weeks.
PG: You have some data in CancerLinQ now. How much data do you have?
CH: It’s actually pretty amazing. We wrote a pilot project. We set it up and we planned on roughly 30,000 individual patient records. We got, at last count, over 170,000.
This is certainly one of the larger databases of breast cancer patients ever assembled.
And what’s interesting about it is in some ways it’s different from others, because, in fact, it’s not clinical trials data. It’s real world data—treatment decisions and outcomes from private practices.
PG: Who owns these data?
CH: Again, the data are in the electronic medical records, so the doctors own the data. The aggregated data that we process, of course, we would own.
PG: How is it managed, how is it protected?
CH: We have a tremendous commitment, of course, to the safety of the data.
We have a series of steps that firewall it and de-identify it as it gets removed from the clinic to ASCO. And we have a data governance committee with robust legal counsel so that all the protections that need to be there are, in fact, there.
Now, having said that, this is ground that has been covered before. This is fundamentally a quality project, not fundamentally a research project.
That’s an important distinction. Because it, of course, means that it mirrors on a large scale the kinds of data gathering that have been done in an isolated or focused ways for years.
As a simple example—maybe it’s too simple—but the CDC tracks all the influenza reports every season. They know where the flu is being identified. They track certain communicable diseases. They do all that under the appropriate guise of quality and public safety.
They are able to maintain confidentiality, and transfer data, and so forth. I think we are learning from that, but we are upscaling it tremendously.
PG: Where will you get the guidance on clinical decision support? Would you develop your own? Would you use NCCN?
CH: That’s a very good question, and I don’t think it’s 100 percent resolved right now. Of course, we, at ASCO have a history of focus on quality, starting with our education programs, extending to Quality Oncology Practice Initiative, which is the largest quality assurance program for cancer in the US.
More than a third of practices participate in QOPI.
We write guidelines and guidance at times, but they are very focal and very in-depth. This system really does require something a little more generalized and faster to develop, something more like NCCN, in many cases.
PG: Would you work with them? Or would you make your own?
CH: Possibly. We are certainly in discussions with them. I’ll say a couple of things about that, though. Even within NCCN, certainly the majority of decision nodes that are enshrined in NCCN are not supported by high-level evidence.
They are just how doctors practice, and that’s made into a standard. One thing about [CancerLinQ] is that it’s a rapid learning system. Over time, as it learns what docs do and what the outcomes are, it would have the ability to inform better guidelines.
You are not locked in with what you start. The point is to improve outcomes through this system.
In terms of clinical decision support, the obvious question that comes up is [IBM’s] Watson. We are working closely with Watson. They have actually purchased extensive data, in terms of publications and guidelines from ASCO, and our program is modular, so it’s entirely possible that the clinical decision support module could be Watson, or something like Watson, which is providing feedback to doctors.
The difference is embedding that program in the larger CancerLinQ means that the program would not stay static, but would, in fact, be over time improved in terms of their treatment recommendations.
PG: I guess I am a little bit worried here. Really, the right word is anxiety. I am experiencing some of it as a political reporter, because I am seeing all these structures. Look at yourself, for example, you are ASCO, you are NCCN, you are Memorial, which has a deal with Watson. How do you keep these structures from colliding? How do you keep this from becoming a bumper car concession?
CH: I like to think one of the great things about medicine, and cancer medicine, is we are here to help patients. We are here to deliver the highest quality care most efficiently to the broadest array of patients.
All these people are working in a collaborative fashion normally anyway. The same people who are involved in this effort were involved in these various institutions historically.
Is it possible that we will find ourselves in competition?
Maybe so. But ASCO is not a business. If our participation in the market ultimately drives the development of a superior tool, but from somebody else, then I would say we have accomplished our goal.
PG: It’s almost a double-edged sword. There are huge advantages and huge potential disadvantages.
CH: One of the things about ASCO—and this can be true for other organizations—is that we are not for profit and we are an honest broker.
Our goal here is to get this job done and to profit, in terms of society and societal benefit, the maximum number of people. If we support somebody who can do better, that’s fine. We just need this to happen for the good of the patients we serve.
PG: There are private companies in that space, and IBM is only one of them.
CH: I have to point out with regard to Watson, it really represents a component of what we are trying to do. It’s not the whole thing.
Just as a quick aside, I am aware that there are others who are either in this space or claim to be in this space who have components of this. I met with one yesterday.
But again, we are more than happy to bootstrap. It’s important to point out that ASCO is not a software company. We are not writers of software. We are quality care deliverers. That’s what we do.
PG: What will the [RFP] look like? How close are you?
CH: We have an aggressive timeline. The request for proposals will be out early in 2014.
PG: That’s really soon. Do you know what it will look like? Could you, in general terms, describe it?
CH: No, I can’t yet, because it’s still being formed. We are in week 10 of a 12-week process, designing subjectively and qualitatively the capacity of the full build. And from that we will end up with the RFP.
PG: I am looking forward to covering it. I guess one last question is, are you running into any reluctance from anyone to give you data? Because data are the main currency…
CH: We anticipate that that will be a concern for some entities. Certainly, some competitive entities or some institutions I am sure will identify limits in what they can give or share. Maybe they will not participate at all.
On the other hand, one of the key take-home lessons from the pilot project was that a program that aimed for five or six practices and 30,000 patients in six or eight months garnered many more practices and 170,000 patients.
It’s pretty clear to us that there is some desire to take part in this. I’d expect that it will be biased a little more toward private practice and a little less toward academia, at least in the beginning.
PG: I am really looking forward to covering this. Is there anything we’ve missed?
CH: I don’t think so. You’ve given me a lot of time to cover it. I think the key thing is that it’s a quality project. The Q in CancerLinQ stands for quality, and that’s not just some marketing accent.
It’s really important to get the idea that what we want to do is improve the quality of care for all patients around the world.
Today many people today outside medicine are surprised that this kind of big data project doesn’t already exist and is not running.
And I hope we will look back and see that this is one of those turning points where we started to use all this work we put into recording data to actually get something out of it.