Whole Genome Approaches to Complex Kidney Disease
February 11-12, 2012 Conference Videos

Communicating Research Information to Subjects: Research Results and Incidental, Actionable Findings
Ben Berkman, NHGRI

Video Transcript

00:00:01,133 --> 00:00:08,299
SARA HULL: Our next speaker is Benjamin Berkman. He’s the Deputy Director of the Bioethics Core of NHGRI and he also has a joint

00:00:08,300 --> 00:00:20,100
appointment as a faculty member in the NIH Clinical Center’s Department of Bioethics. He is trained as both a lawyer and in public health. He’s

00:00:20,100 --> 00:00:28,700
been very active in the role of research ethics consultant, helping many of the investigators and IRBs within NHGRI and the broader NIH

00:00:28,700 --> 00:00:36,633
community reflect on the ethical design of their research involving next generation sequencing and is a faculty member in the Department of

00:00:36,633 --> 00:00:46,099
Bioethics. His research interests span a broad range of legal and ethical issues associated with genetics and genomics research, particularly

00:00:46,100 --> 00:01:00,033
around the topic that he’s going to discuss with us today: obligations regarding the management and disclosure of genomic research findings.

00:01:00,033 --> 00:01:07,499
BENJAMIN BERKMAN: Good morning. I feel lucky to go second so Julie had to deal with you while your coffee was kicking in but hopefully it’s

00:01:07,500 --> 00:01:21,166
kicked in now and everyone is wide awake and ready to go. I want to take a step back a little bit. This is going to be a little bit broader, a little more

00:01:21,166 --> 00:01:33,132
theoretical and it’s about the ethics of next generation genomic research. It’s my standard disclaimer as a federal employee: everything I’m

00:01:33,133 --> 00:01:45,333
about to say is my view; it’s not the view of anyone important yet. So, what I want to do…I want to start with a framing case because

00:01:45,333 --> 00:01:54,999
you’ve heard a lot about how powerful this technology is. You’ve probably been thinking a lot about this question of incidental findings, so I

00:01:55,000 --> 00:02:03,266
want to give you one case which is actually based on a real consult that I did a couple of years ago, then I’ll take a step back. I’ll make an

00:02:03,266 --> 00:02:12,399
argument about the broad ethical implications of next-generation sequencing, then I will drill down a little bit and try to frame the fairly expansive

00:02:12,400 --> 00:02:24,200
scholarship on whether or not there’s an obligation to return incidental findings in genomic research. And let me pause here to make it clear

00:02:24,200 --> 00:02:33,033
that I’m going to mainly be talking about this question of whether there’s an obligation and any investigator can, at their option, do anything that

00:02:33,033 --> 00:02:42,299
they want. And so, the question I want to focus on: is there an ethical obligation, and if so, what are the contours of that obligation? And then…I

00:02:42,300 --> 00:02:52,066
could probably give a two-hour talk on this, but I just want to frame a couple of other things. I think I’ll briefly talk about this “right not to know” which I

00:02:52,066 --> 00:03:00,866
understand has come up a couple of times yesterday and I heard it this morning, and then even more briefly I’m going to talk about a project

00:03:00,866 --> 00:03:09,132
I’m working on now that I think is even harder than whether or not there’s an obligation to return incidental findings, and that’s the question of how

00:03:09,133 --> 00:03:17,333
hard—if you think there’s an obligation to return incidental findings—how hard do you have to look for them, given the Niagara Falls amount of

00:03:17,333 --> 00:03:28,433
data that we’re about to generate? So, let me start with a definition. This is not my favorite definition. I would be happy to quibble with it but

00:03:28,433 --> 00:03:35,633
it’s the one that is most commonly cited in the literature. So, what is an incidental finding? An incidental result is: a finding concerning an

00:03:35,633 --> 00:03:44,666
individual research participant that has potential health or reproductive importance and is discovered in the course of conducting research

00:03:44,666 --> 00:03:55,566
but is beyond the aims of the study. So, I don’t want to talk about the case now, but let me describe it to you and I’d love to hear your

00:03:55,566 --> 00:04:09,199
thoughts, maybe during the question-and-answer. So, imagine a breast cancer researcher who is looking for novel variants—BRCA

00:04:09,200 --> 00:04:17,900
variants—in a previously unstudied or understudied minority population. It says “diverse group” but let’s imagine that it’s a minority

00:04:17,900 --> 00:04:27,300
population. So, they want to take women that are presenting with breast cancer at an academic medical center. They want to sequence data from

00:04:27,300 --> 00:04:36,633
their surplus tumor tissue that was removed for clinical purposes and they’re looking for novel disease-associated variants; so not variants that

00:04:36,633 --> 00:04:45,133
are already characterized in the literature that are already available through clinical tests. So, they’ve got maybe 1,000 participants they’re

00:04:45,133 --> 00:04:53,999
looking for and it’s predictable, given the incidence of known variants and established variants that, in some of these women—some

00:04:54,000 --> 00:05:01,533
knowable percentage of the women—they’re going to find established variants. But that’s not their research aim. Their research aim is very

00:05:01,533 --> 00:05:10,133
specific and it’s very narrow and it’s to look for novel breast cancer-associated variants, and the question is: do they have an obligation to these

00:05:10,133 --> 00:05:20,399
women, who are going to have known variants, to return that information? I think when you describe cases like this—and I have lots of other

00:05:20,400 --> 00:05:29,400
cases that we could talk about—but when you describe cases like this when you have women who are presenting with a disease and you find

00:05:29,400 --> 00:05:38,466
out something about that disease, most people’s instincts are that you should have to return these findings for a variety of reasons which we’ll get

00:05:38,466 --> 00:05:48,932
into; there’s an obligation. But when you think about the time and money that would be required—even in this study—to return those findings,

00:05:48,933 --> 00:05:56,233
you might begin to question whether the researcher can even do their research and then you’ve defeated the purpose of the research

00:05:56,233 --> 00:06:04,299
endeavor. It gets much more complicated and it gets even more complicated. I mean, here you’ve got a relationship with women presenting with

00:06:04,300 --> 00:06:12,033
breast cancer but if you start to imagine the other kinds of researcher-subject relationships where there’s less of a nexus or maybe it’s just a

00:06:12,033 --> 00:06:23,866
blood-draw and you don’t really have any sort of real, meaningful connection to the subjects, the arguments for an obligation to return incidental

00:06:23,866 --> 00:06:33,232
findings might get more attenuated. So, keep this case in your mind as I go through some of my arguments. So, I’m not going to go into the

00:06:33,233 --> 00:06:42,133
science. Hopefully, you all saw these slides or some version of these slides earlier yesterday. The main point is: next-generation sequencing is

00:06:42,133 --> 00:06:53,233
an extraordinarily powerful new tool. It’s changing the way that research is done and it is enthusiastically being adopted in NIH and across

00:06:53,233 --> 00:07:03,699
the research world. The most relevant, the most salient characteristic of this research, as you’ve heard already, is the amount of data that will be

00:07:03,700 --> 00:07:13,300
generated and we’re moving from a…Laura Rodriguez might have talked about this yesterday. We’re moving from…we’re switching our

00:07:13,300 --> 00:07:22,466
bottleneck. Our bottleneck used to be one of that of production. We used to have dozens of people sequencing in labs to produce data and you’d

00:07:22,466 --> 00:07:30,032
have one or two people analyzing it. The bottleneck has switched now and it’s extremely easy and efficient to produce the data; it is much,

00:07:30,033 --> 00:07:42,299
much more difficult to know what to do with it and to analyze it. This is just to show you that we’ve moved pretty quickly from this notion of

00:07:42,300 --> 00:07:52,400
targeted genetic research. I think almost all of the studies that I’ve reviewed or consulted on now do some version of whole exome sequencing,

00:07:52,400 --> 00:08:04,533
and pretty soon they’re all going to do whole genome sequencing. So, genetic research and research ethics—this question of the ethical,

00:08:04,533 --> 00:08:14,033
legal and social issues raised by genetic research—is not a new problem; we’ve been thinking about this for decades. But our argument

00:08:14,033 --> 00:08:22,399
is that whole exome and whole genome sequencing—that WES/WGS will be my abbreviation throughout—it doesn’t raise novel

00:08:22,400 --> 00:08:30,200
ethical concerns. I mean, these are things we’ve been talking about forever, but it significantly magnifies and makes more concrete many of the

00:08:30,200 --> 00:08:37,700
risks that have been relatively theoretical to this point. I mean, we could have imagined the problem of incidental findings a decade ago but it

00:08:37,700 --> 00:08:48,366
wasn’t an actual problem, it was just a theoretical problem and now it is catching people by surprise, I will say, because these things that

00:08:48,366 --> 00:08:56,832
were kind of in the back of people’s heads, now they’re faced with them right in front of their faces. So, the fact that it makes many of the

00:08:56,833 --> 00:09:06,133
ethical risks more concrete has important implications for the way that we conduct ethical review of proposed whole exome and whole

00:09:06,133 --> 00:09:19,833
genome sequencing research. So, I have to pause to say that I’ve been thinking about these issues for a few years now and I’ve been talking

00:09:19,833 --> 00:09:30,699
to IRBs, I’ve been talking to investigators, I’ve been talking to institutions, the funders, NIH, and no one knows…I mean, everyone comes to me

00:09:30,700 --> 00:09:43,533
with this look on their face; they just don’t know how to handle this. And so, let me start by just saying that we’ve made a number of assumptions

00:09:43,533 --> 00:09:51,366
about genetic research and I think whole exome/whole genome research makes us re-evaluate some of these assumptions. So, the first

00:09:51,366 --> 00:10:00,532
assumption is that research is going to produce very few clinically significant incidental findings. I think that’s just wrong now. It’s no longer a

00:10:00,533 --> 00:10:13,799
question of whether or not clinically relevant findings will be found, but how many? So, that’s one assumption that I think we need to revise. A

00:10:13,800 --> 00:10:23,166
second assumption is that there’s a clear distinction between so-called incidental findings and findings explicitly related to the study

00:10:23,166 --> 00:10:37,666
hypotheses or disease focus, and this is a little bit of a nuance to point out—a complicated point—but if you’ve got an experimental approach that

00:10:37,666 --> 00:10:46,632
involves whole exome or whole genome sequencing, can you really say that it’s incidental to your methodology that you’ve found these

00:10:46,633 --> 00:10:54,366
other things? And this raises a complicated point about what counts. How do we define incidental? That’s part of the reason I quibble with

00:10:54,366 --> 00:11:02,599
the original definition of incidental findings that I showed you, but there are number of different ways that you can define “incidental.” Is it

00:11:02,600 --> 00:11:14,033
incidental to the disease you’re studying? Is it incidental to the specific methodology? Is it incidental to the analysis? Is the methodology the

00:11:14,033 --> 00:11:25,233
sequencing or is the methodology the analysis of the sequence data? And the answer to whether something is incidental changes, depending on

00:11:25,233 --> 00:11:35,499
how you want to define what your specific methodology is. But if we’re doing more and more research using this massively parallel genomic

00:11:35,500 --> 00:11:46,466
sequencing, to say that we’re going to find incidental findings gets increasingly difficult. And then a third assumption—and I’ll talk about this a

00:11:46,466 --> 00:11:56,632
little bit later—is this idea that we “don’t look, don’t tell,” that researchers generally don’t have an obligation to look for incidental findings, to act as

00:11:56,633 --> 00:12:06,133
clinicians. I mean, you can imagine a world someday, maybe 10 years from now, maybe 15 years from now, maybe 5 years from now,

00:12:06,133 --> 00:12:15,333
where the act of interrogating a genome just requires pushing a button, that we have robust databases and analytic tools and the act of

00:12:15,333 --> 00:12:25,833
analyzing someone’s genome and looking for clinically actionable known variants is that easy. Well, it gets harder to make the case that we

00:12:25,833 --> 00:12:32,066
shouldn’t do that—that we don’t have an obligation to that—as the technology gets better and as it gets easier to do that, and like I said, I’ll

00:12:42,666 --> 00:12:44,599
talk a little bit more about that today later. So, I should…I’ve been doing some qualitative research about how IRBs are actually

00:12:44,600 --> 00:12:54,733
approaching these issues, and I should start with the premise that they don’t know what they’re doing and there are more than one ethically

00:12:54,733 --> 00:13:00,233
defensible approaches to this. When I say they don’t know what they’re doing, they know that they don’t know what they’re doing. They

00:13:00,233 --> 00:13:13,966
acknowledge that they’re struggling with this and are looking for guidance from NIH; guidance from collections of experts in their field. So, one IRB

00:13:13,966 --> 00:13:21,666
Chair said, “It’s much more case-by-case. What are the protocols? What are the people? What’s the relationship between the investigators and

00:13:21,666 --> 00:13:27,532
the people whom they’re studying?” The second one said, “We certainly don’t have a policy and I don’t know that we’ve really come to a firm

00:13:27,533 --> 00:13:37,566
conclusion. I mean, it gets discussed every time, and there’s disagreement every time.” I’m finding that investigators and IRBs would love to have a

00:13:37,566 --> 00:13:48,099
consistent policy. The thing that I get asked for most of all—and I think this is consistent with Julie’s question they get—is, “Can I have template

00:13:48,100 --> 00:13:56,633
consent form language?” and I don’t want to give it to them because I don’t actually think there’s one answer; I think there are multiple answers.

00:13:56,633 --> 00:14:05,233
We don’t have an institutional policy. I think we’ve gone through several different discussions in our IRB for each specific protocol, but I think we’re

00:14:05,233 --> 00:14:12,233
still at the stage where we hear from investigators what their approach is and then we decide at the meeting if it sounds reasonable. So,

00:14:12,233 --> 00:14:22,033
the bottom line is IRBs are doing this on a case-by-case basis and don’t really have a consistent approach. Even within IRBs there’s considerable

00:14:22,033 --> 00:14:29,599
disagreement, so we’ve got disagreement between IRBs, within institutions that have multiple IRBs, they can have different policies,

00:14:29,600 --> 00:14:39,133
and even within an individual IRB, people seem to disagree about this issue. So, let me get a little more theoretical and talk about some of the

00:14:39,133 --> 00:14:48,499
arguments that have been made. So, here’s an argument for why there’s an obligation to return research results. There are a number of these, I

00:14:48,500 --> 00:14:58,500
could have framed any of them, but one of my colleagues at NIH wrote a paper arguing that it’s based on respect for persons; that, “It would be

00:14:58,500 --> 00:15:05,366
disrespectful to treat research volunteers as conduits for generating scientific data without giving due consideration to their interest in

00:15:05,366 --> 00:15:14,732
receiving information about themselves derived from their participation in research.” So, genetic information is important and if you give it to people

00:15:14,733 --> 00:15:24,533
and they can incorporate it into their decision making, you can enhance their autonomy and there’s something of a reciprocity argument here,

00:15:24,533 --> 00:15:33,933
too, that returning research results recognizes a participant’s contribution to research. So, this is one argument. Like I said, there are a number of

00:15:33,933 --> 00:15:43,866
them. But in some of my qualitative research I’ve found that there’s more than just a number of them; there’s a large number of them. And so,

00:15:43,866 --> 00:15:56,399
these are some preliminary statistics, unprompted—I didn’t come up with this list—unprompted, my subjects have articulated a duty-to-warn or you

00:15:56,400 --> 00:16:05,533
might frame it like a duty-to-rescue argument, respect for persons, the right to know, professional responsibility, doctor-patient

00:16:05,533 --> 00:16:19,799
relationship, reciprocity, legal liability, institutional reputation, public trust, beneficence, researcher-subject relationship, or 5% said “none.” Some of

00:16:19,800 --> 00:16:29,000
these, obviously, kind of collapse into each other but you can see that there’s really a wide range of arguments that you can make as to why there

00:16:29,000 --> 00:16:41,400
might be an obligation to return incidental findings. But there are some pretty clear objections to those views, too. So, you can challenge the

00:16:41,400 --> 00:16:50,966
conceptual notion that not returning incidental findings violates beneficence or respect for persons, reciprocity, etc. I mean, there are other

00:16:50,966 --> 00:16:59,766
ways to do good to a subject, there are other ways to respect your subjects, and there are other ways to recognize the reciprocal

00:16:59,766 --> 00:17:08,399
relationship; incidental findings isn’t the only way that you can do that. People make an argument about the distinction between research and

00:17:08,400 --> 00:17:17,966
clinical care. The purpose of research isn’t to benefit an individual but to produce generalizable knowledge. Some people have talked about the

00:17:17,966 --> 00:17:26,432
therapeutic or diagnostic misconception that there are actually risks associated with conflating research and clinical care, and these are all good

00:17:26,433 --> 00:17:36,233
arguments. I think the main one, though, is the last one, the resource limitations argument. This is hard, this takes a lot of time, you need to hire a

00:17:36,233 --> 00:17:48,733
genetic counselor, doing the annotation, doing the analysis is difficult, returning information is complicated, and especially as your Ns start

00:17:48,733 --> 00:17:59,633
inching up there are very real concerns that imposing an obligation to return incidental findings could reduce the ability for scientists to do good

00:17:59,633 --> 00:18:11,499
research. So, I don’t want to dwell on this too long, but for those of you not familiar with this field, there have been a number of attempts to put

00:18:11,500 --> 00:18:26,233
together guidelines and frameworks about what to do with genetic incidental findings. I’m not going to show you my fancy PowerPoint skills, but this

00:18:26,233 --> 00:18:35,166
is just to illustrate that people—the content of this slide isn’t important—but a lot of really smart people have spent a long time trying to put

00:18:35,166 --> 00:18:49,966
together ways to think about this and we haven’t really come to any clear consensus yet; but we have, sort of. Most of the research thus far has

00:18:49,966 --> 00:19:03,066
been focused on: if there is an obligation, what kinds of information should be subject to that obligation? So in general, people agree about the

00:19:03,066 --> 00:19:14,032
necessity that information should be analytically valid, should be clinically relevant, should be actionable and should be desired. So in general,

00:19:14,033 --> 00:19:22,699
we agree on those criteria as a broad framework. But I would argue that disagreements lurk. So, we don’t have any clear view on what

00:19:22,700 --> 00:19:33,533
counts as clinically relevant. There have been a number of different proposals. The net benefit idea, we talk about clinical utility versus personal

00:19:33,533 --> 00:19:46,333
utility versus general utility, can we define it more quantitatively with a relative risk number, even something as simple as who needs to be the

00:19:46,333 --> 00:19:54,466
subject of the clinically relevant information? So, there’s a big debate in literature about whether or not reproductive information is the kind of

00:19:54,466 --> 00:20:04,166
information that should be subject to an obligation to return, or should you only be returning…should there only be an obligation to return incidental

00:20:04,166 --> 00:20:17,132
findings when that information is directly relevant to the subject in front of you? There’s conflict over whether or not the information needs to be

00:20:17,133 --> 00:20:27,066
desired. Can you override the right not to know? And again, if I have a few minutes I’ll talk about that at the end. And there’s even some

00:20:27,066 --> 00:20:35,332
disagreement about, not whether it needs to be analytically valid, but how you go about establishing analytic validity. So, you all know

00:20:35,333 --> 00:20:47,599
about CLIA. The question is whether research results can be given back to people with a letter that clearly states, “These are research results;

00:20:47,600 --> 00:20:54,866
do not use for clinical decision making. Go get them verified in an independent CLIA lab under the supervision of your doctor.” Can you do that

00:20:54,866 --> 00:21:05,832
or is that in violation of CLIA? I think the law in this case—at least the letter of the law—is pretty clear; that you can’t be returning information to

00:21:05,833 --> 00:21:17,899
people that they could use for clinical decision making. CLIA is a profoundly flawed law particularly when it comes to the regulation of

00:21:17,900 --> 00:21:27,666
genomic testing; it’s really not well-designed for this new world of next-generation sequencing, but it’s the law that we have and so at least NIH’s

00:21:27,666 --> 00:21:43,266
policy—our office of legal counsel—we will not give information back to people labeled as research results without CLIA certification. So,

00:21:43,266 --> 00:21:51,999
just to complicate things even more—so, you’ve got this basic argument that there might be an obligation to return incidental findings—there’s

00:21:52,000 --> 00:22:06,033
this added dimension that you could think about that maybe that obligation is dependent—contingent—on the relationship between the

00:22:06,033 --> 00:22:14,999
subject and their investigator. So, you can think about study characteristics, you can think about population characteristics and the closer your

00:22:15,000 --> 00:22:27,566
nexus between the researcher and the subject, the more argument you might have for an obligation. I’m going to fly through that to get to

00:22:27,566 --> 00:22:38,566
the punchline here, that there are really three emerging models that we’ve seen. We’ve got a very few cases where people have said, “We’re

00:22:38,566 --> 00:22:50,032
not going to return anything. We’re going to define clearly the nature and scope of our analysis. We’re going to explain our data sharing plans and

00:22:50,033 --> 00:22:56,966
we’re going to tell you that we’re not going to disclose anything, even though we might generate information.” There’s the more limited

00:22:56,966 --> 00:23:10,399
incidental finding disclosure plan that says, “We might disclose information under a limited set of circumstances, although this is very, very

00:23:10,400 --> 00:23:22,566
unlikely,” and this method is nice because it leaves open the possibility that if you stumble upon something that is really important and you

00:23:22,566 --> 00:23:34,399
feel like you need to return it, you won’t feel boxed in by a preclusion to return. And then you’ve got people who are planning to…who

00:23:34,400 --> 00:23:43,000
have robust plans for disclosing incidental findings and they clearly define the circumstances under which they expect to return

00:23:43,000 --> 00:23:52,600
the incidental findings. You’ll notice the thing that’s consistent through all of these is that there is some language in the consent form or in the

00:23:52,600 --> 00:24:02,900
consent process that explains to people what next-generation sequencing will produce and that it is possible or even likely to produce incidental

00:24:02,900 --> 00:24:12,133
findings. The last thing you want is for people to be surprised. The most important thing you can do is to manage expectations to explain that

00:24:12,133 --> 00:24:20,699
incidental findings will be produced and then to explain whether or not and under what circumstances you will be returning incidental

00:24:20,700 --> 00:24:33,233
findings. So, let me just…I’ve only got maybe five or six more minutes, so let me just fly through a few last slides. The question of the right not to

00:24:33,233 --> 00:24:43,533
know has been raised a number of times. So, at least in the literature, there seems to be this consensus—this apparent consensus—that we

00:24:43,533 --> 00:24:52,966
should only return incidental findings when they’re desired. The obligation isn’t an obligation to force a piece of paper into someone’s hands,

00:24:52,966 --> 00:25:03,766
it’s an obligation to offer individual findings to research subjects, and what goes along with that obligation to offer is an obligation to solicit

00:25:03,766 --> 00:25:17,232
preferences and figure out what the person does or doesn’t want to learn. We see lots of examples of this in the consent forms. Most often it’s in a

00:25:17,233 --> 00:25:26,499
binary form like this. Do you want to know something or don’t you want to know something? Check one; and I think that’s problematic for

00:25:26,500 --> 00:25:36,166
obvious reasons. But I’m working on a project that is questioning whether these traditional conceptions about the right not to know are

00:25:36,166 --> 00:25:45,632
appropriate in a genomic research context, and there are really two questions. One is: how do you solicit preferences? That turns out to be very

00:25:45,633 --> 00:25:55,399
complicated. And then the second one, given that we’ve already collected lots of informed consent from people and some of them have checked

00:25:55,400 --> 00:26:04,366
“no”: are there any limited circumstances where it might be ethically appropriate to override an individual’s expressed wish not to know genetic

00:26:04,366 --> 00:26:14,999
information about themselves? And I think this is going to become an increasingly important problem as all of the studies that had been doing

00:26:15,000 --> 00:26:23,333
targeted genetic testing are now moving on to whole exome and whole sequencing. You’ve got these old consent forms, you’ve got consent

00:26:23,333 --> 00:26:33,233
forms that really haven’t thought through this problem about the right not to know, and you’re all of a sudden going to be faced with people that

00:26:33,233 --> 00:26:41,733
checked “no”—a cluster of people that checked “no”—on whom you found information that you really think you should return. Generally, I think

00:26:41,733 --> 00:26:51,366
Julie said as well, generally we want to respect autonomous choices but you’ve got the “I can’t sleep at night” problem, that investigators are

00:26:51,366 --> 00:27:05,632
going to be in possession of information that they really feel like they should return. So, I’m going to skip through this and say that there are lots of

00:27:05,633 --> 00:27:16,666
good reasons why someone might not want to know, and these should all be pretty familiar to people, and there are lots of good reasons why

00:27:16,666 --> 00:27:32,532
investigators might want to tell their subjects. So, figuring out how we solicit information for informed consent for future studies turns out to

00:27:32,533 --> 00:27:40,499
be really hard. So, you’ve got a few models. You could not solicit any preferences. You could say, look, you’re going to have to defer to us. We’re

00:27:40,500 --> 00:27:48,266
going to…the investigator gets to decide what we’re going to return or some sort of deliberative body; some sort of expert panel gets to decide.

00:27:48,266 --> 00:27:56,032
You could have a one-time binary choice: yes, I want to know; no, I don’t want to know anything. You could leave it to a Subject Discretion Model

00:27:56,033 --> 00:28:06,766
where you say, “We’re not going to make you choose now; we’re going to call you when we have information and you can decide then,”

00:28:06,766 --> 00:28:14,266
although I think that gets to be very problematic because how do you have that conversation where you say, “We have a piece of information;

00:28:14,266 --> 00:28:21,099
would you like to know it?” Well, what is it? “Well, I can’t tell you until you tell me if you want to know it,” and I think that gets to be very, very

00:28:21,100 --> 00:28:31,133
tricky. A lot of people have argued for a tiered choice model where maybe you can come up with all of the categories of information that one

00:28:31,133 --> 00:28:41,933
might or might not want to know. This is just the list that I came up with quickly, but you can see that if you’re really going to get into a granular

00:28:41,933 --> 00:28:50,233
level of detail, that this tiered list is going to take up two-thirds of your already too long informed consent document, and can people really

00:28:50,233 --> 00:28:54,999
differentiate between these different categories? And as Julie described, you could have a mandatory disclosure category. You could say,

00:28:55,000 --> 00:29:10,000
“You’re welcome to opt to know or not know any of the non-ticking time bomb things, but there is some limited category where we’re going to tell

00:29:10,000 --> 00:29:19,300
you whether or not you want to know.” Okay, so I’ve got lots more to say about the right not to know but I want to leave you with one last

00:29:19,300 --> 00:29:29,066
question which I think is a really, really tricky one that maybe we can discuss in the question-and-answer and maybe we can discuss in the

00:29:29,066 --> 00:29:35,799
breakout sessions. How purposefully, how intentionally should researchers have to interrogate genomic data with the goal of looking

00:29:35,800 --> 00:29:44,100
for incidental findings? If you believe that there’s an obligation to return incidental findings, the next question you have to ask is: how hard do you

00:29:44,100 --> 00:29:53,533
need to look for them? And so, some investigators subscribe to this notion of what I’ll call a “stumble strategy.” Whatever I stumble

00:29:53,533 --> 00:30:00,533
upon, well, then I’ll think about whether or not I need to return that, but I’m not going to look. I’ve got my research question; I’m only interested in

00:30:00,533 --> 00:30:15,199
my research question. But is that acceptable, particularly as these tools—the analytic genomic data analysis tools—evolve? Is that strategy

00:30:15,200 --> 00:30:25,766
appropriate or do we want to have researchers acting as clinicians? Do we want them to take a more proactive view towards looking for maybe

00:30:25,766 --> 00:30:36,332
some small defined list of specific variants, but do we want them to have to look for that defined list if we can come up with a list? And this really ties

00:30:36,333 --> 00:30:51,333
into the ancillary care ethical debate. Ancillary care is care that is not required for carrying out your research but would be morally

00:30:51,333 --> 00:31:02,233
praise-worthy if you could deliver it to your subject. And so, how far are we willing to extend the admittedly murky ancillary care argument into

00:31:02,233 --> 00:31:17,199
the genomic data field? So with that, I will take your questions. Thank you for your attention.

00:31:17,200 --> 00:31:23,133
BARRY FREEDMAN: Barry Freedman from Wake Forest. Thank you. That was a lovely overview and I think the questions are different now than

00:31:23,133 --> 00:31:29,066
they were before because of the uncertainty of these rare variants and the large amount of data, but lots of things are the same. I remember doing

00:31:29,066 --> 00:31:36,599
linkage studies 15 years ago for diabetic kidney disease and finding Klinefelter syndrome, which was associated with an increased risk of

00:31:36,600 --> 00:31:44,766
gonadal carcinoma and going to the IRB and begging them to let us tell the participants and being told, “No, you can’t.” Yet, if they had a mass

00:31:44,766 --> 00:31:53,932
or kidney failure on their labs or a mass on a CAT scan, that was no problem. So to me, I mean, this is a nuance of a new age, but I think the IRBs are

00:31:53,933 --> 00:32:01,299
evolving; it’s a different day for the IRB. The last point I’d make is that all of these studies, the researchers are all tied in with the clinicians.

00:32:01,300 --> 00:32:09,800
Somebody had to bring those samples in. Even in a population-based study of 100,000 people, there are MDs and people that collected those

00:32:09,800 --> 00:32:16,366
that, even though the funding may not be there, I think there are clinicians that could be notified who would make their best effort to notify

00:32:16,366 --> 00:32:22,132
patients when possible. So, I think it’s not just the researchers’ responsibility, you can go back down.

00:32:22,133 --> 00:32:32,933
BENJAMIN BERKMAN: I should make a more policy-related point. I agree. I think that there’s…I was talking mainly about investigator obligation

00:32:32,933 --> 00:32:39,999
but I think we need to think more broadly than that. There’s also a question of what funders need to provide. If there’s going to be a

00:32:40,000 --> 00:32:49,133
consensus that there needs to be…that investigators need to look for and/or return some of these incidental findings, then we need to think

00:32:49,133 --> 00:32:57,566
carefully about how we fund research studies and we need to explicitly and intentionally create funding streams to support this, and I think that

00:32:57,566 --> 00:33:08,432
responsibility also extends to research institutions. It doesn’t make sense to me that every study needs to hire their own genetic

00:33:08,433 --> 00:33:19,333
counselor and their own data analyst. I think that institutions can provide shared resources, core resources, to help their investigators deal with

00:33:19,333 --> 00:33:24,299
some of these common problems. MALE: Dr. Freedman was right. That was a

00:33:24,300 --> 00:33:32,966
lovely talk. Thank you very much. I think the issue—and it was buried in the middle of your talk—about clinical and analytic validity is

00:33:32,966 --> 00:33:40,599
extraordinarily important and Matthias raised this question: what do you do with uromodulin, which we don’t know is a cause or effect? What

00:33:40,600 --> 00:33:50,800
do you do with something that has a very, very small risk ratio? What do you do with a, perhaps, different population? As a clinician, I think that the

00:33:50,800 --> 00:34:00,566
point that’s brought back has to be that there has to be, besides the legal and the counseling aspect, there has to be a clinical aspect and it

00:34:00,566 --> 00:34:08,366
goes back to John Sedor’s point about: how are we going to do this in a clinical setting which has evolved into a research setting as well? So, it’s

00:34:08,366 --> 00:34:16,599
extraordinarily complex. I mean, I hadn’t really grappled with some of the things that you’ve been talking before, or at least in a focused way,

00:34:16,600 --> 00:34:30,600
but how you do it and with particular patients and which, you know, blip on the scan you’re going to go after is hugely critical and I want you to

00:34:30,600 --> 00:34:40,933
comment on the complexity because you did raise all these issues about having to be flexible and it’s a large set of data. But do we have some kind

00:34:40,933 --> 00:34:48,066
of…you mentioned road map. Do we some road map about how we can go forward or is it a question between IRB, investigator, and funding

00:34:48,066 --> 00:34:53,032
agency, case-by-case? BENJAMIN BERKMAN: So, defining what we

00:34:53,033 --> 00:35:03,766
mean by clinically significant or clinically actionable or clinically useful has proven to be a very, very difficult concept. There’s no

00:35:03,766 --> 00:35:12,099
consensus, and in fact, there’s a lot of conceptual muddiness. People use a lot of those terms interchangeably and they all mean different

00:35:12,100 --> 00:35:23,266
things. So, here’s the way I think about it. This is a framework; I’m not going to give you an answer. You need to decide first—well, not in

00:35:23,266 --> 00:35:33,632
order—but you need to decide what level of severity of condition is where your threshold lies for severity of condition. You need to decide how

00:35:33,633 --> 00:35:46,266
much evidence: how conclusively has it been proven that there’s a link between the given variant and human disease? You need to go back

00:35:46,266 --> 00:35:53,899
and make sure that the variant that you found is actually present in the person that you think you found it in. And then, if you want to take it one

00:35:53,900 --> 00:36:04,600
step further, you need to…So, you’ve defined a set of conditions that are severe enough, you’ve defined a level of evidentiary support that will

00:36:04,600 --> 00:36:13,966
make you comfortable reporting back a finding. But then, you might want to even think about actionability in terms of: is there evidence that

00:36:13,966 --> 00:36:25,599
intervening or a proposed intervention will actually change the clinical course of this disease? And so, people have different instincts

00:36:25,600 --> 00:36:36,800
about all of those questions. I think I mentioned before, some people think we should definitely be returning recessive carrier status to help people

00:36:36,800 --> 00:36:45,866
make reproductive decisions. Other people think that’s crazy because that opens you up to a whole world of variants and we should just be

00:36:45,866 --> 00:36:58,966
focusing on variants that are related to extremely severe diseases that that person will suffer from for which there is an available intervention that

00:36:58,966 --> 00:37:06,232
will change the course of their disease, and you can see that those two world views are vastly different but both positions are pretty strongly

00:37:06,233 --> 00:37:17,233
asserted in the literature. So, I can’t give you an answer but that’s the way I think. I personally fall more towards a narrow view of what counts as

00:37:17,233 --> 00:37:27,299
clinically significant. I think we should be focusing on very severe diseases and diseases for which we know we can do something effective right

00:37:27,300 --> 00:37:35,266
now and for the purposes of deciding what’s an obligation that we want to place on researchers. Again, researchers are free to expand the

00:37:35,266 --> 00:37:48,632
responsibilities that they take upon themselves; we’re talking about setting a floor.

00:37:48,633 --> 00:37:54,799
JEFFREY KOPP: Jeffrey Kopp, NIDDK. Thanks for that very nice talk. So, you made the comment during your talk that HHS OGC takes the stand

00:37:54,800 --> 00:38:05,733
that you cannot return non-CLIA certified data. So, I guess then the situation if somebody says, “I have an absolute right to my data,” the patient

00:38:05,733 --> 00:38:16,066
who takes that position, do you tell them something? Do you suggest something like: go to 23andMe or one of the other companies; that is

00:38:16,066 --> 00:38:20,432
your alternative approach to learning this data? BENJAMIN BERKMAN: It’s a complicated

00:38:20,433 --> 00:38:33,366
question. So first of all, it’s complicated by the fact that we are a federal agency and the Privacy Act. People have a right to data collected

00:38:33,366 --> 00:38:44,099
by the government about them, so I don’t think there’s a way that we could refuse to give them back their data, but what form do you give it back

00:38:44,100 --> 00:38:51,366
in? So, you could imagine just giving them a hard disk with all their As and Ts and Gs and Cs; that won’t mean anything to them. So I think if you did

00:38:51,366 --> 00:39:01,732
that, you could make a good faith argument that you’re not returning information that would be used for clinical decision making. The question

00:39:01,733 --> 00:39:13,599
gets more complicated when you’re talking about giving back annotated data. If I was an investigator I would not want to give back any

00:39:13,600 --> 00:39:23,066
sort of specific annotated piece of information that someone might use to make clinical decision making without first having it CLIA validated; that

00:39:23,066 --> 00:39:36,299
would be my position. And if they demanded that they get their hard disk back, I would try to talk them out of it, but if they demanded, I would give

00:39:36,300 --> 00:39:45,966
it back to them but I would give it back to them in an unannotated form and then they can do whatever it is that they want to do with it.

00:39:45,966 --> 00:39:54,066
JEFFREY KOPP: Yeah, I mean, their responses to it…and I like what you’ve said. So, one possibility would be the person who then says, well, the

00:39:54,066 --> 00:39:59,432
annotation is part of the research data. BENJAMIN BERKMAN: Yeah. I’m not sure how to

00:39:59,433 --> 00:40:09,866
answer that. That’s the crux of it but you wouldn’t have…the right to your information doesn’t include a right to force an investigator to

00:40:09,866 --> 00:40:18,366
annotate it before they give it back. So to the extent that they’ve already annotated it, you might be able to make an argument that that information

00:40:18,366 --> 00:40:28,199
should be included in whatever document or file or hard disk you give back, but you can’t make them analyze it. And so in my mind, it’s a limited

00:40:28,200 --> 00:40:35,633
right because they just have a right to see their sequence data if that’s all that’s been produced.

00:40:35,633 --> 00:40:39,899
JEFFREY KOPP: Are you aware of any researcher who has provided a CD with the sequence?

00:40:39,900 --> 00:40:46,966
BENJAMIN BERKMAN: We’ve had consults on this and we’ve generally discouraged…we had some investigators who were proposing to

00:40:46,966 --> 00:40:57,466
explicitly offer that in the consent form and we said we’d prefer for you not to make it clear that…you’re right. People have a right to it but

00:40:57,466 --> 00:41:06,499
let’s not advertise that because we don’t know what people are going to do with that information and I think there’s more harm than good that

00:41:06,500 --> 00:41:09,433
would come from sharing those CDs.

00:41:09,433 --> 00:41:15,733
MATTHIAS KRETZLER: I think very much along the same lines. Dr. Solomon yesterday in his presentation where [---] case in [---] he had 11

00:41:15,733 --> 00:41:23,966
diseases which he diagnosed by his exome chip which obviously were not present in the individual and 2 diseases remained on his list

00:41:23,966 --> 00:41:34,666
where, you know, they then went through further analysis. I think we also have to be humble in the way of what we understand and what we don’t

00:41:34,666 --> 00:41:36,666

00:41:36,666 --> 00:41:43,099
BENJAMIN BERKMAN: I think that’s very true and that’s why I made the point in response to the earlier question, that you want to think through

00:41:43,100 --> 00:41:54,100
what level of evidence is sufficient before you feel comfortable returning something. We’ve had a recent spate of cases where there have been

00:41:54,100 --> 00:42:03,533
published articles reporting a link between a variant and a disease and they’ve later been retracted or amended. I mean, we’re still in the

00:42:03,533 --> 00:42:11,733
very, very early stages of understanding what this genomic means for human disease, and so we do need to be very humble about what we

00:42:11,733 --> 00:42:15,199
feel comfortable returning to people.

Date Last Updated: 9/18/2012

General Inquiries may be addressed to:
Office of Communications and Public Liaison
Building 31, Rm 9A06
31 Center Drive, MSC 2560
Bethesda, MD 20892-2560
Phone: 301.496.3583