Patient-Knows-Best Research Trials

Robert Harrington, MD; Adrian F Hernandez, MD, MHS


May 11, 2015

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

Challenges and Limitations of Randomized Clinical Trials

Robert A Harrington, MD: Hi. This is Bob Harrington on Medscape Cardiology and

We all know that cardiology is a very evidence-driven field and that cardiologists are a group that believes passionately in helping to generate that evidence that will both guide practice as well as policy. We also know, and there has been a lot talked about, including on Medscape Cardiology, some of the challenges and limitations of randomized clinical trials (RCTs): the insufficient size to answer questions, the lack of generalizability due to very strict inclusion/exclusion criteria, and a variety of other things. Nonetheless, the process of randomization and what one learns in the assessment of a therapeutic through randomization has remained the gold standard for assessing a therapy.

We also simultaneously have the ability to tap into enormous amounts of data through electronic health records. While it's attractive to use electronic health records in research, there are limitations there when one thinks about the assessment of the benefit of a therapy or the risks of a therapy.

To try to overcome some of these dual limitations of trials and electronically collected clinical-practice data, several groups, but most notably our colleagues in Sweden, have looked at innovative approaches to randomized clinical trials that involve utilizing routinely collected clinical data. They pulled off a trial called TASTE,[1] which we've talked about on this program and that I think was really groundbreaking in its approach to the methods of clinical trials.

Well, in the US Patient-Centered Outcomes Research Institute [PCORI], and the NIH have collaborated to form an entity called PCORnet. We're going to come back to PCORnet, what it is and what its purpose is, and how it may play a role in understanding how to do certain types of clinical trials that have been deemed pragmatic trials.

With me today is a long-time friend and colleague, Adrian Hernandez from Duke University. Adrian is an associate professor of medicine in the cardiovascular division at Duke University and he's the director of health services and outcomes research at the Duke Clinical Research Institute [DCRI].

Adrian has been instrumental in the workings of PCORnet and is one of the leaders of a new clinical trial being planned dealing with the dose of aspirin for patients with coronary disease. Adrian, thanks for joining me here on Medscape Cardiology.

Adrian F Hernandez, MD, MHS: Thanks for having me here, Bob.

Patient-Centered Outcomes Research Institute (PCORI)

Dr Harrington: Adrian, you heard my preamble about this issue of evidence and some of the challenges of RCTs, but we'll come back to that in the context of our conversation. Instead, let's jump right into this notion of PCORnet. Maybe, for the listener, you could first describe what is PCORI's mission, what is the NIH collaboratory, and how did they come together to form this entity PCORnet. Then we'll get into some of the details of PCORnet.

Dr Hernandez: PCORI came together to do a couple of different things in terms of research. One, it brings together interest and expertise from the scientific community, interest in improving population health, interest in addressing questions for patients with patients, and also interest in trying to do studies better, more effectively and more efficiently, so that patients and clinicians can get real-world answers to the questions that they face daily.

The other thing that PCORI is very interested in is trying to think about new ways to approach these common questions, untangled by any prior history or legacy about how things used to be done. It doesn't have to be done that way. It may have fit a purpose at that time, but moving forward, if you were to start with a blank slate, how would you do it in a way that was most important or has the biggest impact on population health?

NIH Collaboratory

Dr Harrington: The NIH collaboratory, what was the NIH's interest in this arena?

Dr Hernandez: The NIH is also interested in the same way. They care about public health and they also recognize that the current model of doing trials, where you have an enrollment of, for instance, one patient per site per month or less, makes it really hard to answer questions in a timely fashion. The other thing is that it doesn't really leverage the systems that we have now in development for healthcare delivery. In principle, how do you actually leverage healthcare system data with electronic health records so you can reach out to large populations, recruit those large populations in a quick and efficient manner, and then be able to do randomized studies within health systems or among health systems to answer questions quickly? And then also disseminate those answers back seamlessly into the healthcare system?

The NIH, through the common fund, formed the collaboratory, which is really a testing bed to do demonstration projects to transform how clinical research is done leveraging healthcare systems. Many of the studies there are addressing common questions, but the key thing is how those approaches are generalizable to doing large studies that affect large populations and answer important questions. And on a scale that is vastly different compared with what we normally do for a given clinical trial.

Being Patient Centered

Dr Harrington: It's really interesting, isn't it Adrian? You're bringing up a lot of the limitations of a clinical trials, including the generalizability issue—if someone had selected a topic that was pretty narrow in its scope and was only able to enroll, as you suggest, a very small fraction of the patients who might have a disease in that category or something related. The collaboratory was designed, in part, to overcome some of those limitations.

Now Adrian, PCORI also has the unique focus in that the first two letters of PCORI relate to patient-centeredness. They're really interested in bringing the patient's perspective. While that seems intuitive and obvious, it's not something we've always considered over the years in clinical research, is it?

Dr Hernandez: No, and it's actually been pretty fascinating to work with patients in all facets of understanding what questions they feel are most important, understanding the outcomes that are most important, and actually having them on our steering committees and developing and then reviewing results and actually interpreting them in terms of why this is meaningful to someone like them with a given condition.

I can tell you from personal experience that they've changed how we've approached certain questions and also how we selected outcomes to study. They've also have been very direct about how certain studies may not be of impact or they may not fit their routine. That's been really helpful.

Dr Harrington: It reminds me that we did a trial years ago—I'm going to guess it was back in the late 1990s early 2000s when we were studying the oral glycoprotein IIb/IIIa inhibitors. We did a study[2] where we were enrolling patients with prior stroke and prior heart attack. As part of the exercise of understanding the tolerability of the agent regarding bleeding, we asked patients during every visit, did you experience bleeding with the study drug, and if you did, would you keep taking this drug if it had a certain ability to prevent a second event? If you had had a heart attack, prevent a second heart attack, if you had had a stroke, a second stroke.

It was really interesting that the patients who had had a prior stroke, not surprisingly, were a lot older than the patients who'd had a prior heart attack and had a lot more comorbidities. They had a higher risk of bleeding, not surprising, but they had a much greater willingness to take the drug, to adhere to the drug despite the bleeding risks if they could prevent a second stroke, compared with the MI group. It really points out that while in a clinical study we may consider heart attacks and stroke as comparable events in terms of the outcome measures, the patients very appropriately may not. Is that the kind of thing you're getting at with the patient-centeredness?

Dr Hernandez: Yes. The clearest examples are when questions are prioritized, they actually come from patients—what's going to be the most meaningful question for what they do in terms of facing decisions they have. The most important thing, though, is patient preferences on which outcome matters the most, so you're describing every therapy or strategy of care in terms of both risks and benefits.

We as clinicians feel like we understand what that may mean, but really it's in the patients' views, or through their eyes, that we have to understand the true balance between benefit and risk. That balance may vary based on the patient's personal characteristics, whether they are older or younger, what other medical problems they may have. Having an understanding through the lens of a patient really helps define what's most meaningful and that's really the heart of patient-centeredness.

PCORnet: Pragmatic Trials

Dr Harrington: This is a perfect transition to get a little more specific about the kind of work that you're doing. PCORI and the NIH collaboratory have come together to form this entity called PCORnet, and we'll talk about that in a moment. PCORnet put out a solicitation for ideas for trials that they've called "pragmatic trials" to test in large patient populations using electronic health records.

Adrian, can you talk a bit about what PCORnet is? You and I are involved in a trial utilizing aspirin—do you want to talk about how that particular question was selected as part of this PCORnet solicitation process?

Dr Hernandez: In brief, PCORnet, the vision for PCORnet, is that it's the research engine that helps improve care in real-world healthcare systems. It's imagined that this would be a public utility or a research infrastructure that can be accessed from clinical investigators within PCORnet, investigators outside of PCORnet, other partners such as NIH, etc. That this is an infrastructure that hopefully allows research to be done with more efficiency in terms of addressing the questions, and getting away from this idea where you do a trial, you build a network, you break it down, you rebuild it again. As well as the parallel processing that we often have, where we have literally study coordinators who have a case report form that they're filling out, and they're cutting and pasting from the electronic medical record from another screen. It's a way to take that away.

Now for the aspirin study, here again, it comes from the patient. One of the key things for these early demonstration projects for PCORnet is what are questions that would have a public-health impact? What are the questions that are highly feasible in a pragmatic real-world trial setting? And what are the questions that could actually test the infrastructure for PCORnet and pave the way for future studies, because if you think about the simple questions that we face every day, there are hundreds of those? Aspirin is an example of that. If you can develop the systems and approaches from electronic data, to follow-up, to regulatory and ethics issues, then you can have a reusable infrastructure that can truly be a public utility for clinical research here.

The Aspirin Question

Dr Harrington: Let's talk about the aspirin question, which is something the cardiology community, including you and me, have talked about for many years. It seems astonishing that there is no definitive evidence that would demonstrate that one dose of aspirin, in this case a full-dose aspirin of 325 mg, would be better than a lower dose of aspirin, 81 mg or 162 mg, on reducing the risk of ischemic outcomes. That's coupled with the notion that we're pretty certain we know that lower doses of aspirin are safer from a bleeding perspective. Is that the kind of question that fits well in the PCORnet framework?

Dr Hernandez: Absolutely. If you just consider the context here where a patient is going to their drugstore or their grocery store and literally getting a bottle off the shelf or over the counter, that choice actually could translate potentially to a higher risk for bleeding or a higher risk for death and heart attacks. Those simple questions—when you start thinking about it on a national basis, then you start thinking of thousands of patients that are affected by that choice every day, or on an annual basis—are what we're aiming to address.

Dr Harrington: We certainly know from work done by a variety of investigators, including those using large professional society databases, that on this aspirin question there is equipoise in the field. We know that there is a distribution of aspirin doses that providers use, and that presumably, as you said, that patients select when they go to their grocery store or drugstore. You want to give some context to that, Adrian?

Dr Hernandez: If you just consider the national landscape in terms of a point of discharge from a hospital in patients who have cardiovascular disease or recent MI, there is a 25-fold variation in the choice at discharge for high-dose vs a low-dose of aspirin.[3] That really illustrates to me that hospitals and practices don't know exactly the right dose that we should have. Then, when you look nationally and sum it all up, basically two-thirds of patients are on full dose, or 325 mg, and a third to 40% are on 81 mg. That highlights the equipoise, and again, this is one of the best questions in terms of known potential benefits, known potential risks, but what is the right balance is, especially for a given patient in terms of their own personal profile?

The Aspirin Trial

Dr Harrington: Let's review the specifics of the trial and the question, and how it will be answered. This went through a selection process through PCORnet. The aspirin-dosing question was chosen as an appropriate one for a demonstration project within PCORnet. You, I others, and patients were involved in a lengthy process where the protocol was developed. We now have what I would call a very simple randomized trial—selecting a group of patients with coronary disease and some risk factors and randomizing them to receive low-dose aspirin or high-dose aspirin, and then following them over time for both ischemic and bleeding events.

Pretty simple question, but what's really going to be the interesting test is how to pull it all together and answer the question. Do you want to talk about the framework, Adrian, for how the question will actually be answered?

Dr Hernandez: This builds on what health systems have naturally, which is they have large electronic health records systems that form the basis of data marks that healthcare systems use for understanding their own healthcare delivery and how physicians and patients are actually interacting with the healthcare system.

We take advantage of that, those data marks, to identify a cohort of eligible patients within the health system and through that can approach them electronically to tell them the story about aspirin and the uncertainty in that question and invite them to participate in this trial.

We'll also take advantage of programs like Health eHeart, which is really trying to engage the cardiovascular patient community. Then, through that, consent patient electronically and then follow them up electronically every 4 months through the ADAPTABLE study portal via Health eHeart. We'll also have follow-up through the health systems and their data marks as patients have encounters in the healthcare system and through a call center that we have centrally. Another asset here in terms of follow-up is through payers, such as Medicare and the Medicare enclave. We'll have different components to ensure that we have complete follow-up and same outcomes.

Dr Harrington: We know that randomized trials have limitations, we know that electronic-health-record observational research has its limitations—this is trying to take the best of both worlds and bring them together in a way that allows us efficiently to answer the questions.

Adrian, our listeners will be amazed to hear that the intention is to randomize 20,000 patients to the two doses of aspirin, follow them for a period of time to understand this risk/benefit ratio, but to do it for a very modest amount of money considering the scope of the question.

Dr Hernandez: To give a sense of what that scope is going to be—we're talking about having six to eight of these research networks within PCORnet, that's anywhere from 2000 to 4000 patients. For a given healthcare system within these nodes, that's about 400 patients per healthcare system. That's randomizing 20,000 patients, US only, at about 40 to 50 so-called traditional sites, in a timeframe of 18 to 24 months. Then in terms of the cost, we're talking about a cost of one-fifth to one-tenth of what it would take for a normal trial that's done through, say, the NIH. If you start thinking about that kind of scale and if we're able to answer five more, 10 more questions, given the same type of investment, then you really see a big impact on population health.

The "Adapters"

Dr Harrington: That's the astonishing part isn't it? A paper[4] a number of years ago now came out of DCRI by Pierluigi Tricoci, Rob Califf, and others looking at the quality of evidence that we have that accompanies the guideline recommendations from ACC and AHA. As you've pointed out at the beginning, the vast majority of things we do every day in practice are not accompanied by a high level of evidence. This would allow us, potentially, to fill some of those evidence gaps in ways that we couldn't have done 5, 7, 10 years ago.

In closing, I just want to get your views on how it's been engaging the patients in this particular project and the role that they're going to play in the trial, which I understand has a name—it will be called ADAPTABLE?

Dr Hernandez: The trial name is ADAPTABLE, and that was again, selected by patients and stakeholders. One of the key things here is that because it's a new model of research, we understand that we will have to adapt to a different framework and adapt along the way. The patient group, they have named themselves "the adapters." They are really driving the engagement here from the patient community and also driving engagements of the clinical community, saying, hey, this question matters to me, and we need to do something about it. It's uniting everyone together for the good of public health.

Dr Harrington: The phrase "citizen scientist" has been talked about when talking about engaging patients, and the patients are truly involved in the science here. They've been involved in the protocol. They'll be involved in the safety monitoring. I love the word "adapters" to put a name on this really novel way of doing clinical trials. Congratulations.

Adrian, this has been a fantastic discussion on new ways of thinking about clinical trials and a clinical trial on aspirin dose, an old but relevant question that's going to be addressed in this PCORnet clinical trial.

I want to thank Adrian for joining me. Adrian is an associate professor of medicine and cardiology at Duke University and the director of health services and outcomes research at the Duke Clinical Research Institute. Adrian, thanks for being on Medscape Cardiology.

Dr Hernandez: Thanks for having me.


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: