Life After the FDA: Mining Data and Reducing Disparities

Interviewer: Robert A. Harrington, MD; Interviewee: Robert M. Califf, MD


July 20, 2017

Robert M. Califf, MD

Robert A. Harrington, MD: Hi. This is Bob Harrington from Stanford University on | Medscape Cardiology. We had an earlier discussion with my friend and colleague Rob Califf about his time at FDA. We're going to have an expanded conversation as to what's he going to do now, thinking about his new role as chair of a not-for-profit foundation spun out of PCORnet (the National Patient-Centered Clinical Research Network) and how this may be a way by which important healthcare questions can get answered at a massive scale at a fraction of the costs of current clinical-research methods. Maybe even more interesting, we'll spend some time exploring what he's going to be doing at Verily here in Silicon Valley.

Let's start with the Duke role. What are you going to be doing at Duke and why did you agree to do this? Why did you find this important?

Dr Califf: Well, Bob, I hope you don't mind I'll bring our relationship into this. As you vividly remember, you and I wrote a proposal to the university when you were here at Duke to create essentially a cross-institutional effort in health-data science. It was too early for Duke at that point.

Dr Harrington: I still think it was a good idea.

Dr Califf: It was the right idea. As I remember, when you were recruited to be chairman for Stanford, that plan was part of the picture and you're busy implementing it. I'm really proud of what you've put together there, partly going back to our discussions many years ago. Fast forward 6 years and the institution at Duke really is ready. The whole campus is on fire about making sense out of health data. We have all this talent and money and computational power out in Silicon Valley and a few other high-tech places. In the past month, there have been four key articles[1,2,3,4] about the disparities that exist in the US in general for longevity and specifically about cardiovascular disease, since this is largely a cardiology audience.

Essentially, you have areas living better and longer, mostly economically advantaged, well-educated people, and then you have the rest of the country, which is experiencing a reduction in life expectancy for the first time since 1918. Duke is at a very special position in the state of North Carolina in the Southeast, where the health is not looking too good. The ability to use information and develop technologies and tools to help people turn this around I think is a critical mission.

The focus here is not just on "academic" analyses, it's on actionable analysis, so it leads to implementation of strategies that improve health and healthcare.

Dr Harrington: There is, as you say, a lot of academic interest in the growing computational ability, in part made real by better computing power, by cloud computing, by techniques and technologies like artificial intelligence, machine learning, et cetera. It's all gotten better, and it's been, in some ways, largely an academic exercise in medicine, but now the real push is to apply it. Which brings me to your second role is as the chair of the board of this new not-for-profit foundation that spun out of PCORI and PCORnet. You're going to lead an effort thinking about application and testing of ideas that can then be widely applied or implemented. Let's first talk about PCORnet. What is this, and why a not-for-profit board, and why do you want to spend time doing this?

Dr Califf: I had started working with PCORI to develop this network before going to FDA. Working with my good friend Rich Platt and ultimately an amazing team at the coordinating center, but also approximately 34 health systems and 20 to 30 patient-powered research networks joined together to do practical, important research that met the needs identified by patients. While I was at the FDA, PCORI put a lot of infrastructure funding into this network. It was thrilling that I came back to Duke and pretty shortly thereafter I got a call asking if I was willing to be the chair of the board to take this entity into a not-for-profit foundation.

Essentially, it's a transition. You might liken it to a birdie in the nest being fed by a mother and now it has to fly. It is a not-for-profit run in the interest of patients. There's no motive to create shareholder value and all that, but there is a motive to do a better job in answering critical questions more quickly and at a lower cost so that patients can get what they need.

Dr Harrington: The idea that you and I had many years ago was to test the appropriate dose of aspirin. PCORI has funded the ADAPTABLE trial[5] to answer this question. You and I have both been involved in this. It's been an amazing exercise to watch how powerful it can be to do research on an infrastructure of aggregated EHR data. It allows you to think about answering questions at a scale that previously was just not possible.

Dr Califf: We had a meeting yesterday at the health policy center in Washington. When you cross-relay EHR data now being curated by health systems, like Stanford is doing and we're doing here at Duke and across the country, you cross tabulate that with claims data from the payers and registry data for the professional societies. I think we're going to be able to do research at a fraction of the cost and actually answer all these questions that really bother people, that we don't know the right treatment for a lot of things now, and we can find out.

Dr Harrington: If we can get this system honed to a point where it truly is generating information as to what are the most appropriate choices in practice for both physicians and patients, it would be an extraordinary step forward. Both of us were heavily influenced by Gene Stead's vision from many years ago, you pull out those papers[6] from the '60s and '70s and Dr Stead was talking about this then, wasn't he?

Dr Califf: He was definitely a pioneer in the early use of computers in medicine. Many of us who are disciples of Dr Stead have been waiting for the day when computation got to the point where it was really feasible to do it. We have to get the culture to catch up with the computation. Between PCORnet and a consortium of academic centers like Duke and Stanford, I'm really hopeful that we can get over this history that we've had of celebrating the individual and hoarding data, as opposed to using data in the most efficient way to get the answer to help people live longer and become more functional

Data Sharing and Working With Verily

Dr Harrington: Earlier today, I was on the phone with our mutual friend and colleague, Adrian Hernandez, and we were talking about open data and data sharing with a broader group, and Adrian said the clinical trialists want to share. They're just trying to figure out how. We need to deal with America's healthcare system. It's not a how issue, they just don't want to share.

[On data sharing] The superseding value needs to be what's best for patients.

Dr Califf: The view that's prevalent by those hoarding and hiding your data is that your business process is of great value and ultimately there will be all this intellectual property. The superseding value needs to be what's best for patients. There's nobody clearer about this than Joe Biden. When he took his son with glioblastoma around to cancer centers, it became obvious to him that our best cancer centers are not sharing data very well. No one sees enough people with glioblastoma to really do the work that's needed to find effective treatments. He's a real zealot about this, and we just have to figure out how to do it and get people over the hump.

Dr Harrington: Certainly, there are some subcultures within medicine like the pediatric cancer world, where they have figured out that for infrequently occurring diseases, banding together is in the best interest of everybody, most notably the patients.

Let's in the last couple minutes Rob, focus on Verily. People out there I think know that Verily is part of the Alphabet family. Part of the Google family. I think people are less familiar with some of the things they're doing. Why don't you talk about Verily, what their charge is, or how they see themselves in the health technology world, and then what you're going to be doing with them?

I may be the only Medicare recipient to go through Google orientation.

Dr Califf: Well, if you look at Google as a whole, it's founded on the basis that if you aggregated, stored, and organized information that the world would get better. Let's apply that to health, healthcare, and biology. That basically is what Verily is attempting to do. Although it has spun out of Google formally as a member of the Alphabet family, it has access to all the prowess that exists across the Google system, of all these really smart people and engineers that know how to use computers in ways that are frankly baffling to many but pretty exciting.

I think it's useful to think about Verily in two parts. One is a series of joint ventures. Some of them have been publicly announced. Some of them are in process now with more traditional pharma or device companies to marry the Google ability to put together technologies and sensors and information with traditional medical products companies' ability to distribute and market and manufacture

The other part is dealing more broadly with health information, which is where most of my focus will be. I just went through a Google orientation. I may be the only Medicare recipient to go through Google orientation. I'm not far enough along yet to really talk in detail about what we're going to do, except to mention the Baseline Study that you and I have been involved in since before I went to FDA, which is now enrolling. It's a Duke-Stanford-Verily project, being led at Verily by Jess Mega. I think of it as the study to measure everything you can possibly measure in about 10,000 people.

If you just look at the data system for this effort, it far exceeds anything I've ever seen in the ability to store and compute 6 terabytes of data on each participant, from each follow-up visit, but in between streaming of dozens of parameters into the cloud continuously. This was unimaginable a few years ago, but now it's being done.

Dr Harrington: We've been having a lot of fun on campus thinking about Project Baseline and just starting to put people into the trial and the data collection as you've said. Six terabytes worth of data, first visit taking more than a day because of the amount of information that's being collected. What's been extraordinary to see is that, in addition to the technical prowess, it is really generating great buzz in the community. People want to participate in this and want to go through this process of really being citizen scientists and providing their data for greater good.

Dr Califf: I'm very excited by this You would agree that it's just the beginning. The same data platform by contract will be used in the precision medicine initiative. It's an effort with [the National Institutes of Health] NIH and many, many academic medical centers. Verily and Vanderbilt are in a joint contract, enrolling over a million Americans. I would hope for 100 million, as I've told Francis Collins. This is just the beginning.

Dr Harrington: That's a good way to end our conversation. It is just the beginning. Rob, it's been fun to talk with you. For our listeners, we've explored everything from your time at FDA, to your new role thinking about health-data science at Duke, to the time you're going to spend with PCORnet in trying to do nationwide clinical studies involving integrated electronic health record systems. Finally, the time you're spending in Silicon Valley, including at Verily.

Dr Califf: It's been great talking with you. I hope you can display one of the JAMA heat maps so people can see it (Figure 1). For cardiologists listening, if we don't do something about the negative direction that cardiovascular risks is heading in, particularly in the economically and educationally deprived populations in the United States, we're going to be in big trouble, so we've got to work on that. I hope the kind of technologies we talked about will be tools that we can use to expand our reach as cardiologists to a much broader population.

Figure 1.. Reprinted from JAMA. 2017;317:1976-1992

Dr Harrington: My guest today has been Rob Califf from Duke University, Stanford University, and Verily. Rob, thanks for joining us today on Medscape Cardiology on

Dr Califf: You bet. Take care.


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: