Day 1 – Session 1: Building a consortium of cohorts

Day 1 – Session 1: Building a consortium of cohorts


>>Rick Lifton: Thanks very
much, Francis, for that truly inspiring introduction to
the festivities for the next several days. It’s my pleasure to moderate the
morning’s session and, as you’ve heard, leading up to with this
workshop, four groups have gotten together and put serious
work into devising white papers on specific avenues and defining
some of the opportunities and challenges going forward
for this project. In addition to discussing the
specific logistics of how this cohort might be put together and
operate, we need to keep in mind Francis’s comments about the
goals of the project because, at the end of the day it’ll be
critical, in my view, that we have well-defined goals and make
sure that the structures that we develop are going to be well
suited to answering the questions and the goals
that we seek to achieve. So to start off, first Mike
Lauer from NHLBI will discuss some of the issues we’ll need to
be thinking through to build a consortium of cohorts, and
I’ll just remind those of you who are online that the white
papers are available on the — on the website. Thanks very much. Mike?>>Mike Lauer: Great, Rick. Thanks so much and good
morning, everybody. I want to give particular credit
to the members of our working group: Eric Boerwinkle, who’s my
co-chair — please raise your hand — and Rebecca Baker, who’s
actually in the control room where all the wires and
computers are; Greg Burke — could you raise your hand? — Rory Collins; Michael
Gaziano; and Teri Manolio. Okay, so as already mentioned by
Francis, we now live in a world where there are
extensive resources. There are many cohorts. Some of them are quite large. And these — some of these
cohorts are extremely rich with well-phenotyped data, lots
of subjects, valuable longitudinal follow-up. But the part that is missing
is that we don’t have them all well-coordinated. There are consortia that have
been formed, but to a large extent, many of these are
working more in parallel than working together. And however, we certainly want
to take advantage of this and take advantage of the
extraordinarily enthusiastic research communities and
participants who are a part of these cohorts. So this is what it
might look like. One could imagine that we
have several large cohorts — Cohorts, say A, B, and C —
and these cohorts are brought together, and the data from
them are pooled and harmonized. This reminds of — reminds me of
that cartoon from The New Yorker where there was those very
complex mathematical equations, and one of the two people says,
“And then a miracle happens.” So this will then lead to the
development of what we’ll call Enhanced Cohort 1. And this Enhanced Cohort 1
already has the capability of generating new research that
could not have been done before, so we’ll call that Research Now,
and those might be some early victories. But this is not where
we’re going to stop. The vision now involves bringing
in new participants, new sequencing data, and new data
sources, and this is — a large part of the workshop is going
to deal with these particular components. This then leads to the
development of the mega-cohort, the Enhanced Cohort 2. Now, notice here that there is
this very thick bidirectional arrow. The idea here is that the cohort
is not some passive group of people who happened to be
followed, but rather there is rich bidirectional
communication happening between the cohort and the research communities. Now, this, of course — one of
the key requests of our working group was to think about what
might be the challenges and opportunities that would
underlie this vision. Here’s a picture of Royal
Dawber; he was one of the early directors of the
Framingham Heart Study. I had the great privilege many
years ago of during my research training at the
Framingham Heart Study. And you’ll notice here that he
is looking at a chest x-ray, which is on a large piece of
film, and that the record of the research subject is
on pieces of paper. This was one sight in one town. It was an incredible effort. And, in fact, the Framingham
Heart Study continues to be enormously productive
to this day. One wonders whether Dr. Dawber
might have imagined that 50 years later we should be
talking about entirely different ways of gathering data and
communicating with one another. This is an iPhone. This is a Samsung. And this is to remind us that
these are different platforms. They work on different
operating systems. And different people — I like
this; some people like those. And so we have to have a way
in which these platforms can interdigitate and communicate
with one another, and we also have to have a way in which
these platforms actually want to talk to one another,
that there are incentives behind the communication here. So these pictures bring out the
three major challenges that we see — or I should say four. One is the feasibility of doing
this — this is going to be a very — potentially, a very
expensive and time-consuming effort; the fact that we’re
dealing with rapidly evolving technologies; the fragmentation
that currently exists within our systems, both our research
systems as well as our health care systems; and the need
to have incentives so that researchers and research
participants want this thing to grow and will drive it. So we spent quite a bit of time
talking about these, and I’ll briefly summarize
our deliberations. So the first issue deals with
expense, time, and feasibility, the concern that this is going
to be something which is extraordinarily expensive, will
take a long time to accomplish, and maybe — well, will be
logistically challenging. And the key point here is that,
as has already been mentioned, we’re not starting
this from scratch. We leverage existing resources. There are existing cohorts. There’s already extensive data. There are groups of people who
have already been recruited and have indicated an interest,
an enthusiasm for participating in research. And so, by virtue of the fact
that we’re not building this cohort from scratch, we now
can enable the introduction of new participants
and new methods. Okay, now, the second
opportunity and challenge is technology. I remember and some of you may
remember that seven or eight years ago, when the iPhone first
came on the scene, there were some very smart people who said
that this thing is impossible; it can’t possibly do what they
claim it’s going to do, and who’s going to get it anyway? And we know that technologies
now are rapidly evolving. We can’t imagine what the
technologies are going to be four, five, or six
years from now. So how do we deal with this? So again, we leverage what
already exists, but we recognize that we’re going to
have to grow with it. We can’t just settle on one kind
of technological platform and say that’s what
it’s going to be. So part of this may involve the
use of bidirectional Internet- and cloud-based services. We’re going to hear a lot during
this workshop about mobile devices; a lot of excitement
about home monitoring technologies, and I think
Francis eluded to that in his hypothetical example, and
web-based portals; and then the introduction and integration of
electronic health records, as well as large scale digital
registries, which a number of, for example, professional
societies and patient advocacy groups have already
put together. Okay, now, a third problem is
coordination and governance. And this is — this is really
huge, and this has to do with the culture, the organization,
the way we think about how we’re going to all work together and
make this actually happen. And this kind of coordination
and governance can either drive it and make it move forward very
quickly or runs the potential risk of paralyzing
the whole thing. So we talk in our white paper
about what coordinating centers might do, what steering
committees might do, what their rolls might be. Ultimately, of course, one of
the key goals is to enable the conduct of research, both
observational research as well as clinical trials, and research
that deals with interests in precision medicine can
occur very rapidly and where the next cause, the next study would have an extraordinarily low
marginal cost. We spent a lot of time
talking about incentives. A critical part to this is that
the researchers who currently have their “own” cohorts — and
I put the word “own” in quotes — as well as patients and
patient groups and healthy participants who are interested
in participating in research, they need to have an incentive
— the system has to be engineered in such a way that
the incentives are there to drive it. And so one of the things we need
to think about is, how do we offer new value? How do we — how do we
communicate and make real a new value to researchers
and research subjects that currently does not exist? And these were some of the ideas
that we came up with: the fact that now you can
research your cohort, you can research many other
cohorts, with newly available whole genome sequencing data,
data on patient-reported outcomes that have been obtained through digital platforms, as well as from other
novel sources. Now you would be able to work
with much larger sample sizes and do much more robust analyses
by virtue of the fact that data from multiple different sites
and people have been harmonized together, and one can then
analyze their findings within broader context. But this is something that we
spent a lot of time talking about and probably will
have to be an important part of the equation. So let me end by, extending on
Francis’s and Kathy’s and Eric’s comments, talking about the
idea that this presents an opportunity for really big,
audacious new kinds of research. This could potentially be an
enormous scientific commons. I want to credit Eric
Boerwinkle for that term. This could entirely change the
dynamic that exists between researchers and participants,
that they now are working much more as partners rather than
more of the traditional kind of one-way relationship. Let me just end with a story. One of my favorite papers in the
medical literature was published in 1961 in Annals of
Internal Medicine. It was entitled Factors of Risk
in the Development of Coronary Heart Disease. Drs. Dawber and Kanell were the authors of this paper. It was the first time that the
term “risk factors” was used in the biomedical literature. And this was one of the
early reports from the Framingham Heart Study. It demonstrated that people who
have elevated cholesterol and blood pressure levels were
at increased risk for developing coronary
heart disease. In the discussion section
of the paper, there was something very interesting. There was a comment that said
that although we have shown that high cholesterol and high blood
pressure predicts a higher risk of coronary heart disease and
although it would be tempting to reduce levels of cholesterol
and blood pressure to prevent coronary heart disease,
we’re not at a point yet where we can say that. That’s something that’s going
to require further study and, specifically, experimental
studying in clinical trials. But at that time, that was going
to have to be something else. That was going to have to
be someone else’s problem. That was going to have
to come from an entirely different research platform. And that was something that
was potentially many, many years in the distance. Imagine now a situation where we
can make potentially critically important discoveries about risk
markers for disease or pathways for disease and then we want to
know whether or not acting on those leads to better health. We could then use the same
platform and the same network of very bright researchers and
research participants and data sources to now conduct the
definitive clinical trial and do it at very low cost and
not do it 10 years from now or 15 years from now,
but do it real soon. And in this way, this kind of
enormous scientific commons and research cohort could
function as a place in which breaking —
in which exciting observational research happens and
evidence-based precision medicine rapidly becomes a reality. I’d like to invite my partner
here, Eric, to come up here. And I told Rick that the way
we’ve decided to divide the labor up here for our
questions or comments is that if it’s easy, I’ll handle it — [laughter] — and if it’s difficult,
Eric will be glad to. [applause]>>Rick Lifton: Great. Thank you. So now we’ll open
it for discussion. Thanks.>>Marc Williams: Marc Williams,
Geisinger Health System. One of the challenges or
tensions of developing cohorts is what some people would
refer to as the least common denominator problem, that as
you’re trying to aggregate datasets and that, that you
always have to go down to the least common denominator. And so as you think about
conceptualizing building the cohort, are you envisioning
starting from what are the specifications that would be
needed to accomplish the science that you want to accomplish and
then put that out and say, if you want to participate, then
these are the things that you have to demonstrate, or do you
think it’s going to have to be more pragmatic, which to say, if
we want to get to the numbers, then we’re going to have to
accept lower abilities in terms of being able to share, and then
hopefully, we could fix that? I guess I’m trying to relate
that to your Cohort 1 versus your Cohort 2 conceptualization.>>Mike Lauer: Yes [laughs]. [laughter] We’ve talked about these —
about these issues, and that’s something which, actually, I
think will be an important part of our discussion here at this
— at this workshop because there’s something
to both issues. If we’re going to be expending
the effort now, we’d like to expend the effort in places
where we’re likely to draw fruit quickly. And that means certain kinds of
requirements will be necessary. But on the other hand, almost
certainly we’re not going to be able to find five perfect
cohorts that we can immediately put together in a week, and so
this will require a number of difficult decisions.>>Eric Boerwinkle: The only
thing I’ll add is, you know, clearly, there will need to be
entry criteria, not just of the cohorts but the participants,
and again, going back to this is going to be participatory, both
the study participants and the participants in the
investigator community. And so I think there will be,
again, minimum entry criteria, and I also think it’s not
feasible to have a square box where every individual is
measured for every variable nor is that necessary to
achieve the ultimate goals. And so I think once we have a
number of meetings kicked off with this one and we start to
articulate clearly what the goals are, I think then the
criteria and needed measurements will emerge from that.>>Mike Lauer: Yeah?>>Sekar Kathiresan: Sek
Kathiresan — is this working?>>Mike Lauer: Yeah.>>Sekar Kathiresan: From Mass General and Broad Institute. I just want to maybe take a
step back and say, what is the problem we’re
trying solve here? Because we — you know, it seems
like, as Rick mentioned, the design of a research cohort
could be dependent on exactly the specific questions that
are going to be addressed. And in terms of precision
medicine, I can at least see three or four different
overall goals. One is what the president
mentioned, which is CF and identifying a cause of a
rare disease and designing a drug toward that specific mechanism. Another could be what
Francis mentioned, PCSK9, which is identifying causal
biology, and the therapy developed is actually relevant for everybody. A third could be like BRCA1
where there, the issue is risk prediction and you identify
a therapy that actually has nothing to do with mutation
itself in terms of mastectomy. And the last, also Francis
mentioned, which is recall, identifying individuals’
specific genotypes or exceptional phenotypes and
having them available for specific physiologic
experiments. It seems like, you know, getting
the balance of what we want to accomplish among those, at least
— I’m sure there are others — these four goals, getting that
right will dictate kind of what kind of cohort you end up
designing, because there’s going to be strengths and weaknesses
to any of the different –>>Mike Lauer: So Sek
asked a really — this is a critical point. I think you’ve answered your
question very nicely, but we’ll continue to extend
on it, which is — [laughter] — what is the —
what’s the problem that we’re trying to solve? And we have to think about what
are our overall goals would be, and actually, within our working
group, we talked about use cases, rare diseases,
identifying people with special genotypes and phenotypes, being
able to call people back in with certain characteristics
to participate in follow-up studies. And so the question is how do
we think about those particular kinds of goals in designing
this because that would play a critical role.>>Eric Boerwinkle: I
think — first, I agree. I also think we have to be
a little bit careful not to articulate — try to
exhaustively articulate the goals too early; that
could become limiting. And so I think these four areas
that you mentioned really are the tip of the
proverbial iceberg. You know, I’ll give you
just two other examples. You know, we shouldn’t be too
— this may be heresy in this group. We shouldn’t be too genome-
and genotype-focused. We have an enormous opportunity
in other omics areas. Also, for many of us who’ve
struggled for years to really try to not about gene
environment interaction but actually do something about it,
it’s going to require very large sample sets. I think that, you know, some
other issues that are going to come up is, you know, we’ve
dreamed about having a national transferrable medical record,
and this really could be a pilot study so when someone moves from
Minneapolis to Florida to retire — this year we probably
should say Boston — [laughter] — to Florida just
to get a break — [laughter] — you know, they can, you know,
in quotes, “take their medical record with them,”
because we really have a national medical record. And those are just
a few examples. We could go on and on. So I really think this is a
resource that as we begin to build it, it’s not only going to
be valuable for these questions we articulate earlier, the
number of new questions that are going to emerge is
going to be spectacular.>>Mike Lauer: I got a call a
few weeks ago from a patient who was asking me about what whether
or not he really should be taking Coumadin for his atrial
fibrillation, and he pointed out that he had actually looked at
the risk calculator that we have up on our website for the
likelihood of stroke, and he said, “Boy, you know, this takes
me 3 percent to 2 percent. I’m not sure I’m so
excited about this. Why do I really
need to take this?” And this brought up the idea —
I was quite impressed that he was thinking
at this level. And this brings up this whole
idea of number needed to treat. I think this is one of the big
problems that we have is that you have to study huge numbers
of people or treat huge numbers of people in order to — in
order to benefit, and this kind of platform maybe a way in which
we could, instead of going from 3 percent to 2 percent, go from
30 percent to 5 percent and be able to then target our
treatments much better. Cashell?>>Cashell Jaquish:
Cashell Jaquish, NHLBI. I think this is getting to a
general issue of early discovery versus building a resource,
and they’re not exclusive, but the balance between the two. Because the study design for
early discovery may look very different, and I think we would
be naive to think that we could build a resource for everything. So I think, you know, we need
to discuss that a bit, that balance.>>Mike Lauer: So Cashell
Jaquish was asking about the balance between early discovery
and later applied research and how do we think about that
balance in putting this thing together because it can’t
necessarily be someone that would handle
absolutely everything.>>Eric Boerwinkle: But I — but
I also think we are going to need early discovery, and so
it is true that the final design may not look like the early
design, but I think keeping in mind early discovery is going to
be important in the early years. We’re going to need to document
to our constituents, to the participants that this is worth
it, to the scientific community that this is worth it and to,
ultimately, those that provide resources. However, as we design this early
discovery opportunities, we can then grow it to basically
fill out what ultimately the larger U.S. research cohort
would look like. But I think ignoring early
discovery and only creating a long-term resource would
likely be a mistake.>>Mike Lauer: Tom?>>Thomas Insel:
Tom Insel from NIMH. Mike, its clear the influence
of Framingham is all the way through here. Question I have about the way
you’ve conceptualized this as multiple cohorts coming
together, both current and new: Would any of those potentially
involve children and families? Is there some reason to think
about actually creating a different kind of cohort than
what Framingham was all about?>>Mike Lauer: Yes. No, actually, Framingham — one
of the strengths of Framingham is that there are multiple
generations, and so that actually enhances
the ability to study complex problems in smaller numbers of people. But yes, absolutely, this is
something we have talked about, and one could imagine — we
don’t have to imagine; we already see it — that when
one does family studies, one could generate useful
information in smaller samples. Eric, you’ve had
experience in this.>>Eric Boerwinkle: I also think
it’s — well, first, I am going to speak more to this later, but
I do think it’s important not to think of the U.S. research cohort to be a
Framingham on steroids. There are — there are many
other opportunities — for example, you know, having
population-based cohorts like Framingham. Also, the vision is to have
clinic-based cohorts or clinic-based patient
populations that can enrich. And so it’s much more than
simply bringing together and stitching together somehow
population-based cohorts. I think if we — it would be a
mistake to view this only as a consortium of cohorts. And again, I’ll speak
to this more later.>>Mike Lauer: All
the way in the back.>>George Yancopoulos: Yeah, hi,
this is George Yancopoulos from Regeneron. And I wanted to follow up
some of the comments that were already stated. And it gets back to, I think,
Francis’s great introduction which, in many ways, said that
we were starting somewhat agnostically and trying not to
have too many assumptions here. And I have to say I’m a little
disappointed in the way it starts because it starts with a
huge assumption, and people are already seemingly questioning
it, which is probably, you know, the most critical decision
that has to be made, which is how to start structuring this thing. And the assumption is to start
with these existing cohorts, and I — you know, I heard a lot
of words about leveraging and so forth, but — I don’t
know — a lot of us have dealt with similar sort of problems in
the past, and, you know, often it’s a lot more efficient,
it’s a lot more powerful not to be saddled with the baggage of the old. We heard the term from our
Geisinger partners about, you know, lowest common
denominator –>>Mike Lauer: Least
common denominator, yeah.>>George Yancopoulos: — which
I think is a really important perspective. But there is a lot of reasons to
consider the alternative, so I would say one needs to —
you know, Sek talked about going backwards a little bit. I think that we — I would say
we’re already starting with huge assumptions, and we’re starting
with, you know, a plan that’s based on something that a lot of
people might think is entirely the wrong
way to start. So I think that the whole group
has to start from the beginning and challenge the very first
assumption and say why are we even talking about these
approaches right now? Why don’t talk about
bigger picture? Why don’t we take a step back
and say, hey, if you have, for the first time, a chance to
build it and start on day one, are you really going to go
back before you go forward?>>Mike Lauer: So this is an
interesting question about whether or not the assumption
that we have about building up from existing cohorts makes
sense when it might make more sense to start something
completely from scratch. I want to — Rory, could
you take that one [laughs]? [laughter]>>Eric Boerwinkle: See? It’s the easy ones. I get the hard ones. You get the really hard ones.>>Rory Collins: I
was actually awake. [laughter] I think that — my view is that
what one wants is to work out what are the — as Eric put it,
the conditions for entry into the project. And what you want to do is
find people who are willing to participate, because it may well
be that existing cohorts or other collections may be good
ways of identifying people who would be willing to participate. But whereas the focus of the
discussion and the focus often among epidemiologists is around
characterizing the participants in these studies, I think what
we often fail to do is think about how we characterize
the health outcomes. And obviously, precision
medicine relates not to so much — or not solely to the
participant but also the particular disease they have. And so in thinking about
especially the long-term aims of this, I think the most
important condition is that one would be able to characterize
the health outcomes. And that means building it
within those systems where you would be able to find about what
happens to people, specifically, around their diseases. And I think in terms of the
point that Kathy made about people participating in this,
we should be thinking not just about electronic medical
records, which is about what doctors think is important to
patients and the population, but also about the things that
participants and the public think are important that doesn’t
get picked up through electronic medical records: cognitive
decline, mood, pain, quality of life. And I think we’ll come to this
around direct contact with participants and using
novel approaches, web-based questionnaires, other things
like that to really get full characterization
around health outcomes. So in UK Biobank, you know, if
you’re thinking about cancer, you want to be able to get
tissue samples, so you need to build your resource in such a
way that it would be feasible to get tissue samples for a
large proportion of the people who are in the study who
get cancer. So I think that’s the most
critical condition for the long-term aims, you need
to build it in such a way that you can get to that
information of characterizing the disease outcomes.>>Eric Boerwinkle: George, one
other point is, you know, the next four talks or these
early talks in the white papers, we should emphasize these
are draft white papers. This is a beginning to
facilitate discussion so the group, all of us, are
participating in this in terms of open to comments
and criticisms. And I think the other point is
— on the slide, is the U.S. research cohort would not,
again, simply be us stitching together existing sample
resources, both population-based and clinical. You saw on the slide
additional individuals. My guess is, at the end of the
day, this is going to be a hybrid approach of existing
sample resources and novel sample resources in order
to achieve these goals.>>George Yancopoulos: I’m
not arguing that it’s –>>Mike Lauer: We’ll take —
we’re going to — we’ll take one more question. Andrea.>>Andrea Downing: Hi,
I’m Andrea Downing. I am one of the few patient
advocates in the room. I have an ePatient site
called Brave Bosom. I just wanted to tag on to what
Rory Collins said and say we have to let patients lead. When we’re deciding the goals
and defining the success, we have to let patients actually
ask the research questions and define what’s important
to us. If we start with that principle
and design around that, I think we can be successful.>>Mike Lauer: That’s great. That’s a great way to end this.>>Eric Boerwinkle: Yes, agreed.>>Mike Lauer:
Thank you so much. Great. Thanks. [applause]>>Stephanie Devaney: And Rick,
if you don’t mind, I’ll just interject
for a second. That’s a great segue into the
participants’ work group. I just want to note that we have
over 130 people on the webcast — or WebEx, so thank
you for joining us. Please do submit your questions;
we want to hear from you. We’ve got over 200 people
listening in via videocast, and we’ve got a nice
conversation going on Twitter, so please join using
#pminetwork. Thank you.>>Rick Lifton: Perfect. Well, that was a
fascinating discussion, and I’m sorry we didn’t have two
more hours to continue it, and I’m sure it could have gone on a long time. I’ll just make the observation
that one of the challenges in starting with legacy cohorts
will, of course, be capturing ongoing data in real time and
being able to pull that in. And that seems to be the sort of
the challenge that George was raising about the challenge
of using legacy versus starting anew. And I think it will be
worth revisiting the cost of the UK Biobank and what it
took to get that started, more or less, from scratch. So now we’ll turn to the next
white paper, and this was — we’ll hear from Pearl O’Rourke
from Partners, who will discuss considerations for partnering
with research participants, and she’ll be speaking on
behalf of her group, and Laura Rodriguez from NHGRI.

You May Also Like

About the Author: Oren Garnes

Leave a Reply

Your email address will not be published. Required fields are marked *