Tip:
Highlight text to annotate it
X
William Pao: Can you hear me okay? I want to thank the
organizers for inviting me to speak today, and Dr. Swain for sharing the podium. So,
a lot of discussion today about point of decision -- point-of-care decision support, so I'm
just going to give you an example of what we've done at Vanderbilt, just as a representative
example of what can be done.
So many of you may know that cancer has been, for more than 100 years, really described
from where it came from on the body, as well as what it looks like under the microscope,
but we know, really, that at the genetic level, many cancers can differ, especially in terms
of what is so-called the driver mutation, i.e., the mutation that not only induces tumor
genesis, but then is required for that tumor cell to survive. And so, classically, melanoma,
for example, has been described where it comes from on the skin, but now we know from tumor
profiling at Vanderbilt, for example, and elsewhere, that there's really multiple subsets
that are clinically relevant.
At Vanderbilt, we've initiated routine tumor testing. We've tested over 2,100 specimens,
and, again, a large fraction of these patient's tumors will have specific driver mutations
that are potentially actionable with targeted therapies today. And, as a result of the genotyping,
61 percent of those patients have gone on to a clinical trial, specifically in melanoma,
a big comparison to the 3 percent nationally that go on.
So the reason I think I'm talking to you today is because there's a huge knowledge gap, even
in the oncology world, as to what to do with all these mutations. We build gene panels
that only we're testing for about 40 to 60 mutations, and even then some fellows and
staff at our -- at Vanderbilt were having difficulty deciding what to do with those
specific mutations. A lot of people have talked about this, but, essentially, there are surveys
that reveal a disconnect in the understanding of and communication about genetic mutation
testing among health care professionals and cancer patients.
And so one of the things we did was first to get rid of the old electronic medical record
for reporting gene mutation testing. As many of you may know, most -- in many hospitals,
there is a scanned .pdf from the molecular pathology lab that goes into a chart. It's
very hard to find that report. It's also hard to see where the result is. And the challenge
for us was, really, how do you start to report multiple mutations in multiple genes at the
same time, and whose role is to curate all that knowledge? And there's no clinical trial
information or therapeutic implications for the results.
And so at Vanderbilt we actually control our own medical records, called StarPanel, so
we can adapt it. And in conjunction with Mia Levy, a very talented biomedical informaticist,
we've essentially put into the chart the results of genetic testing of our patients. Here,
in particular, is what we call a dashboard or a whiteboard, where we can see by patient
the exact mutations that were tested for in this specific melanoma panel. And then we
put a lot of thought into saying what does a clinician really want? Do they just want
to know positive or negative? So you can see a yellow result is positive and a gray result
is negative. If you really want to know the exact mutation, you hover over the yellow,
and then it gives you specifically all the different kinds of mutations that you looked
for. And you can see that a clinician at the point of care with five minutes to see his
or her patient could get very confused, especially at the genomic level, of the significance
of all these mutations.
Now, again, I'm also a practicing oncologist. I know that you have five minutes sometimes
to go see your patient. You want to just look it up and be able to discuss it in an intelligent
manner. So within the chart, if you forgot what V600 is, you can just click on that,
and it takes you to My Cancer Genome, which is a freely available website designed to
enable a genetically-informed approach to cancer medicine. Now, at this particular website,
you can see in very brief detail. We wrote it -- or the mantra is that clinicians should
be able to catch up on all the information within five minutes. It will tell you, for
example, the location of the mutation, the frequency of that mutation in melanoma, the
frequency of the specific V600E mutation among BRAF mutant melanomas, and then the implications
for targeted therapy.
In addition, you can then -- what we're most excited about is that it can help you find
clinical trials, and so if you click on the clinical trials tab, it will take you to BRAF
mutation-directed melanoma clinical trials, not only at Vanderbilt within our own system,
but also in Tennessee, in the United States, and internationally. What Dr. Levy did was
to download all of clinicaltrials.gov, and then using natural language processing and
machine learning, used gene names to search that and annotate it so that you can search
the trials by gene status. And then if you click on, specifically, for example, that
number, it will take you to the clinical trial that is being done at Vanderbilt, and you
can check the eligibility. You can click on the United States and it will tell you which
sites around the world -- in the United States, particularly, have trials directed towards
BRAF mutant melanoma, and then you'll take -- click on the specific site, and it will
take you to nciclinicaltrials.gov.
So, again, this is a website that's designed to enable a genetically-informed approach
to cancer medicine called My Cancer Genome. We currently cover 17 cancers, 25 genes, and
289 disease-gene-variant relationships. We also have clinical trial search, where you
can search 36,000 cancer trials, 135 cancer diagnoses, and 437 cancer genes as defined
by COSMIC, which is the Catalog of Somatic Mutations in Cancer run by the Sanger. We
want to be the COSMIC for clinicians, which is essentially not available, except through
this website, currently, in the world.
We also built something called DIRECT, which is the DNA-Mutation Inventory to Refine and
Enhance Canter Treatment. We started with EGF receptor mutations in lung cancer, and
basically looked at all the papers since 2005, and have more than 1,000 patients in the database
with 188 different kinds of EGF receptor mutations. Again, as a person who was involved in the
discovery of EGFR mutations, I used to get emails all over -- every week from people:
"I have my patient's tumor genotype, it has this mutation, what am I supposed to do?"
Well, now they can just go to DIRECT, and then they can find out the significance of
that particular mutation in EGF receptor. And we hope to expand that out. And, actually,
we think there should be a national or international effort with this kind of database so that
we can actually keep track of all these rare mutations.
We also built the Internet's only complete list of targeted therapeutics, which you can
easily access on the website, and there are other educational things on the website as
well.
Right now, this is a worldwide collaboration. It's basically based on philanthropy and grants.
We have 49 contributors from 18 institutions and eight countries, and we now, starting
in about -- we started in January of 2011, and we're now more than -- we now have more
than 2,000 site users per week, according to Google Analytics, from 134 countries around
the world and all 52 U.S. territories.
And then this slide is just to show you that it's a huge collaborative effort. We do have
a lot of people at Vanderbilt, but our contributors are from all over the world, and we ask experts
throughout the world to help contribute on their specific disease and gene variants and
mutations.
Thanks.
Teri Manolio: Great. All right. Thank you very much. So
we have questions for our oncology colleagues. I think Marc is going to start off.
Marc Williams: Thank you. One comment related to the aggregation
of data. Again, we got a great example in oncology of how this can work, and that's
the Children's Oncology Group that, you know, for a variety of reasons, we're faced with
the challenges of rare tumors, and made the decision from the get-go to aggregate data,
and the number of children that are on trials approaches 100 percent, and so it's -- in
some ways I don't think we'll be able to scroll back and begin something like that, but it's
certainly a worthy goal.
The question I have for you, Bill, is tell me a little bit more about the outcome measures
that you're using to determine the impact and effectiveness of the MCG program.
William Pao: Yeah, that's a great question. Right now all
we have is a user survey that people can ask us questions or give recommendations, et cetera.
We had multiple survey requests. Right now the content is written for physicians, by
physicians, and physician scientists, but we've had a lot of requests to translate it
into patient-centric information. We've also had, and I'll get to your detailed question,
but we've also had requests to translate it into French, German, and then other medical
institutes wanting to link to us and/or suck in the content so that they can report their
outcomes.
In terms of actually whether it's influencing patient care, we have not measured that specifically.
Right now we're trying to help clinicians and oncologists really use the genetic information.
We know patients are getting their tumors tested. The bulk of mutations that are tested
for include, like, V600E, or actionable ones, like an EGF receptor, where people do know
what to do, but there are some rare ones where we don't have -- there's no evidence as we've
been discussing in terms of the guidelines.
Marc Williams: In terms --
Sandra Swain: And I think the reason, you know, I wanted
him to present it, too, is because we would like an ASCO to include these kind of things,
because our cancer link, one of the main goals is looking at outcomes for all these patients
who are treated, you know, not just with genomic information, but information in general.
Marc Williams: Yeah, and then one specific question or point.
You mentioned that 65 percent of your Vanderbilt patients are on trial. How much of that is
attributable to this, or was that just a cultural thing that pre-dated this?
William Pao: Yes. So some of that involves the fact that,
for example, melanoma, we have lots of trials that are genome-directed, and so I think there's
a huge referral bias for patients to come and get specific care for their specific mutation.
We have trials for all the different subsets of melanoma, for example.
Teri Manolio: Great. I'm wondering when you work with these
cancer genomes, are you getting more information than the candidate genes, or are you really
just focusing on the driver mutations in cancer?
William Pao: Yeah, so, right now, at least at Vanderbilt,
we're just focusing on the driver mutations, but we know there's, as other people have
talked about, there's going to be large groups of people who are doing exomes or genomes,
et cetera. Right now the website is only designed to help you with a single mutation, and a
single gene, and a single gene disease variant, so I think the next frontier will be how do
you deal with multiple mutations? However, there are companies that many of you may know,
for example, like Foundation Medicine, which profiles more than 200 cancer genes, full
exomes, including multiple fusions, et cetera, and they give much more extensive reports,
some of which our clinicians still have trouble interpreting if they're not necessarily pathway-oriented
in terms of cancer biology.
Teri Manolio: Yeah, one thing that you'll probably be struggling
with as well as everyone is how do you deal with the mutations that you find that are
not really into oncology? So, you know, you're dealing with the cancer, but then there are
all those other findings, and so have you run across any of those yet or are you making
plans to deal with other subspecialties? You want to describe that?
William Pao: Sure. Well, in cancer -- at Vanderbilt we're
only testing for known variants, but, obviously, everyone's moving toward full exome or finding
variants of unknown significance. Right now, as Dr. Swain alluded to, at least in cancer
you can say there's a mutation for which there's an FDA-approved drug and has an indication
in that specific disease. Then you may get a mutation, like V600E in a colon cancer,
for which there's no FDA approval, but which there's FDA approval in a different cancer,
and then -- and you can move on down the chain. You know, that's a mutation of clinical significance,
but not necessarily in that disease. Then you move on to variants of unknown significance,
et cetera. I think a lot of us are wrestling with that fact. I was just reading the Science
paper coming over about how you can also re-identify patients [laughs] by all of that data, and
I think we all have to grapple with that as we move towards these large databases.
Teri Manolio: Great. Gene.
Eugene Passamani: For Dr. Swain, it seems to me relative to
the other subspecialties of medicine, you really have a challenging task, because you're
having to put this all together as tons of data rolls in. Are there lessons learned about
how you've done that that could inform what other people are going to try to do and get
ready for it before it hits?
Sandra Swain: Well, I think the cancer link is the lesson
learned, and we are just still -- we're continuing to learn lessons and the prototype is just
being rolled out. So we definitely will and we plan to publish a paper on it on the issues.
It's not been easy, as you can imagine, using all different electronic medical records,
trying to get it all together, getting, you know, from my institution, for example, which
is MedStar, the lawyers are not excited, would be an understatement, about giving information
to a company to anonymize. That's a huge problem right there, the HIPAA issues. So there are
a lot of things that we will have to look at and we've worked with, but the plus is
that at least 130,000 patient's data is there, so the precedent will have been shown, and
hopefully in seeing that and seeing that we can show outcomes and that we can show benefit,
quality, looking at value and things like that, others will see that and actually agree.
Marc Williams: Yeah, so I think that that's a really important
point and something that I think we should highlight as we think about how to address
some of these issues. And this has come up at other meetings, is the concept of the trusted
broker or the safe harbor, and what role could perhaps the policy shop of genome develop
relating to defining those types of safe harbors where we could do this very important job
of aggregating data.
Sandra Swain: I think that would be terrific, really important.
William Pao: Can I comment?
Sandra Swain: Yeah.
William Pao: Quick comment on that. In terms of our DIRECT
database, the DNA-mutation Inventory to Refine and Enhance Cancer Treatment, you would be
surprised about how many barriers there are in terms of getting the mutation status from,
for example, trials, from cooperative group trials, et cetera. Originally we wanted to
make that database easily searchable by anyone, but then because groups would give us data
that had already been published but had not necessarily been published with the fine mutation
data in detail, they didn't want that sharable. So we had to basically make it that you ask
us a question, what is your specific mutation of interest, and then we would give the result
back.
And, for example, companies are also very reluctant to share that information, et cetera.
It would be good to have an honest broker in the middle that could get all that information.
Sandra Swain: I think the other thing that's really interesting,
if you actually talk to patients, they're very much for this, so they're the least of
the obstacle, you know, in getting this done, and I think it's important in this kind of
setting, too, as we go forward in talking about guidelines that patient advocates really
be involved, because they are our advocates. They want this done. They want quicker therapies
and all of that.
Teri Manolio: Great idea. Jonas, please.
Jonas Almeida: [unintelligible] exactly that our initiatives
like PatientsLikeMe, some help pushing for this to happen?
Sandra Swain: Patients pushing for what we're doing specifically?
Jonas Almeida: Yeah, so, initiatives like PatientsLikeMe,
for instance.
Sandra Swain: Yeah, I don't know that specifically, but
I know the patients really, you know, in looking at different surveys and all, have been very
willing to do this, and they're not pushing us specifically, but I think that they will
definitely be great advocates for it because it helps them in the long run. You know, you
have metastatic cancer, you're not cured in most situations. So, they all know that and
are very happy to contribute, and I think we just need to educate people, patients even
more about it, and make sure that the, you know, data is not used inappropriately.
Teri Manolio: Great. Other comments? Super. Well, this has
been wonderful. Thank you so much. Our next topic --
Marc Williams: We should thank our -- thank our speakers.
Teri Manolio: We should thank our speakers.
[applause]
Marc reminds me that my manners aren't glued too closely to me [laughs], so he sits next
to me instead. So, in terms of lunch, we had asked them to provide it a little bit early
in case we ended early. I don't think they're quite ready yet, but it should be out there
momentarily. It will be just to the side here. We are having a working lunch, so really we
need you to get your lunch and bring it back in, and with the size of this group, it's
going to take the full amount of time to get it back in. So we will reconvene here at 1:10,
hearing from the Heart Association. Thank you.