Tip:
Highlight text to annotate it
X
Debra Leonard: Thanks. So, it’s a little unnerving to stand
in front of this group who are more expert in genomic medicine and clinical genomics
than I. I have been working on this mostly from a policy and technology assessment perspective
with the College of American Pathologists, and also, however, yesterday, at 9:50 a.m.,
my proposal for starting a pilot project in clinical genomics through the hospital laboratories
was approved, so I am about to jump into the deep end of the pool. So I may be talking
with some of you [laughs].
So, I’d like to talk about just a framework on recommended pathway for omics test evaluation
to move information from research into clinical utility assessment, very briefly, that came
out of an Institute of Medicine report. And then mostly focus on the evaluation that the
College of American Pathologists did of genomics and genomic technologies. And then a few comments
on genomics from a molecular pathologist’s perspective.
So, this report about “Evolution of Translational Omics: Lessons Learned and the Path Forward”
was just completed in March by an IOM committee. I was a member of that committee. This is
very small print; I apologize. But basically it describes how you move from the discovery
phase of an omics discovery, and this would include genomics, but it’s also proteomics,
metabolomics, epigenomics, everything. So, there are recommended processes in the discovery
phase of how you validate, how you work with your data, which are very large data sets,
and how you validate that before it moves into a test validation phase. And that test
validation phase, it’s recommended be done in a CLIA laboratory setting so that when
you move it into the evaluation for clinical utility and use stage, it’s already been
validated in a CLIA environment with standard operating procedures and another independent
level of assessment from the research laboratory that made the discovery.
And then various pathways for assessing clinical utility are described, prospective/retrospective
studies, prospective clinical trials, where the test is not direct patient care. And either
of these first two clinical trial pathways do not need the testing to be done in a CLIA
environment. So the third pathway, prospective clinical trial, where the test will direct
patient care or patient management decisions, that testing needs to be done in a CLIA laboratory
environment. And actually, that is in the CLIA federal regulations, but very few people
understand this. Whenever a test is done for patient management, which means diagnosis,
anything, treatment selection, it needs to be done in a CLIA environment. So I recommend
this report to those of you who are looking at this translational highway.
The rest of my comments until the last slide are based on the College of American Pathologist’s
evaluation of genomics. We called it “Genomic Analysis.” I tried to remove that out of
this slide set and call it “Genomic Medicine” or “Genomics,” but...
So, understanding the human genome, basically the human genome project, was completed 2003,
led us to a generic genome, and individual patient genomes will allow us to make decisions
about directing patient care, hopefully improving outcomes, but we need data to demonstrate
that. And this is what we’re calling “Genomic Medicine.” So it’s made possible by the
ability to analyze individual patient genomes or large amounts of patient genome data.
Genomic medicine is driving a global molecular diagnostics market with an estimated annual
growth rate of about 14 percent annually. And a lot of this -- most of this -- is in
infectious disease testing, which I noted is part of what you’re doing with periodontal
work. But smaller parts and more rapidly growing parts of this growth are in genetics and oncology.
Obviously, you know this better than I do, being NHGRI, that the cost for genome sequencing
data has dropped precipitously, and this is a logarithmic scale, to boot. And now the
cost for doing a genome is less than $3,000, and still going down. I will put in a pet
peeve: I do not like the terminology whole genome or whole exome sequencing, because
it basically says that we can do everything now. The technology is advancing such that
what we do when we sequence a genome or an exome improves over time, which I think if
we’re already selling it as complete, we’re selling ourselves short. And the advances
in the sequencing technology is partly what’s driving adoption of clinical genomic analysis
in molecular pathology laboratories. So the first genome was done on ABI sequencers, taking
hundreds of them. We moved then into the HiSeq era of -- which is mostly used for clinical
research, where you can do 50 gigabases, and it can be used for gene panels, exomes, or
genomes.
In November of last year, at least at the molecular pathology, Association for Molecular
Pathology meeting, there was the roll-out of the MiSeq and the Ion Torrent PGM, in which
you can do sequencing -- you can buy these instruments at a clinically relevant price,
so $75- to $125,000 -- and the turnaround times are 27 hours to eight hours rather than
the week to week and a half of just data generation. So this begins to be clinically relevant cost
and turnaround times, and even moving on, the Ion Proton is coming out in the third
quarter that will allow an eight-hour turnaround time for an exome, genome, transcriptome.
And so this is really what we feel is driving clinical genomics to be possible today as
this technology continues to advance, and I don’t even have PacBio and Oxford, and,
you know, all the other technologies that are coming on the market.
So when we look at current molecular biology testing, we’re looking at a single or few
mutations, a single gene or a pathogen, or a few genes or a few pathogens, and clinical
genomics, we feel, is really doing gene panels, exome, or genome to analyze exome. There’s
also genome and transcriptome being done, and these slides, even from when they were
developed six months to a year ago, people are now doing genomes, exomes, transcriptomes
clinically. So we felt, at the time, that the genome and transcriptome data would mostly
feed back in our understanding of the clinical use of variants into clinically-targeted testing
with genome panels or exomes.
So for me, as a molecular pathologist, next-gen sequencing is just the next new technology.
I lived through PCR, capillary electrophoresis, microarrays. Next-gen is just the new cool
tool that we have in the molecular pathology laboratory, and it’s being used by many
labs today. In the setting of the molecular pathology laboratory, next-gen is not going
to replace everything that we do. So some things that we do -- deafness genetic testing,
many of the cancer mutations -- can be done on next-gen sequencing technology into gene
panels, exome, genome, transcriptome. But other things that we do -- viral loads, bone
marrow engraftment analysis -- there’re many tests that will stay on their current
platforms because that’s the only way they can be done, and that’s the most cost-effective
way to do the testing.
So, you know, I won’t go through all of this, but, you know, there are these various
stages when you think about clinical testing of the pre-analytical phases generating the
sequence data, doing the data interpretation, which Heidi talked about, reporting and billing,
and then clinical consultations. So we feel that molecular pathologists, molecular geneticists,
industry, others with a strong molecular biology or genetics knowledge base will be able to
do this sequence data generation and sequence data interpretation. And I think it’s interesting
that -- to hear Heidi saying that the sequence data generation is being outsourced from Partners
because it’s very complicated. But we see all pathologists being involved in this pre-analytical
test-selection phase, helping with the proper consenting of patients or documenting of proper
consenting, the reporting, billing, clinical consultations around how you use this information.
So we look at sequence data generation and sequence data interpretation. The data generation
has many quality issues from a laboratory perspective, but it’s really the sequence
data interpretation that is the pain point. And Heidi, better than this, described that
clinical-grade database and bioinformatics tools are a high-priority need in the clinical
laboratory. So both the clinical databases, where we know that the data generation was
done in a high-quality manner and not in a research laboratory -- not to denigrate research
laboratories, but there are different standards for operations in research settings and clinical
settings -- and then the software tools for interpretation and clinical usefulness are
needed. And we think that pathologists should be at the table in the development of bioinformatics
tools and these databases. And pathologists should be learning how to use these tools,
since we are the diagnostic testers along with molecular geneticists that will be doing
this.
So, when we think about all pathologists, we also need clinical decision support tools.
So this is different. So when people talk about databases, there are many different
things that they’re talking about, and that’s why we describe the clinical database repositories,
the interpretive tools, and then when you get into using this information clinically,
the clinical decision support tools that are needed. And it’s these clinical decision
support tools that will help all physicians understand the clinical usefulness.
So the speed of clinical adoption, we felt, hinged on several factors. The decreasing
costs are, of doing the testing, are driving this into molecular pathology, molecular genetics
laboratories, as well as the increasing -- the decreasing turnaround time, the increasing
speed of this testing. The bioinformatics is the pain point. We need clinical quality
databases and software tools, and the clinical usefulness is also a pain point. What is clinically
useful? Genomic analysis is used now with very targeted testing. We’re hopeful that
research and discovery, such as what you guys are doing, and clinical use will increase
the clinical applications.
There’s a lot of payment uncertainty. Currently, there are no specific CPT codes. I know we’re
talking about these new molecular codes that will be test-specific, but those test-specific
codes aren’t going to deal with a cancer gene panel or, you know, they are not going
to work for this next-gen sequencing technology, especially a genome or an exome. And it was
already brought up, the reinterpretation of already generated sequencing data, how are
we going to pay for that? And from talking with colleagues of mine who are already doing
this -- one colleague is doing a 28-gene panel, and talked to a payer to get it paid for,
and the payer said, but only three of those genes is relevant to the cancer type that
this patient has, and so we’ll only pay three-twenty-eighths of that test. And yet
if you did those three genes individually, it would be more costly than doing the 28-gene
panel. So we have to get them thinking in terms of chemistry testing, you know, where
you do the panels because that’s the most effective way, cost-effective way, to do the
test, and you use the parts of it that are relevant to that patient. But that’s -- maybe
the payers here understand that, but many of the payers do not understand that.
And then there is a great deal of regulatory uncertainty. It is not clear what the FDA
is going to do with next-gen sequencing technology in clinical laboratories. Quality standards
development is being led by the CAP in collaboration with AMP and ACMG. We now have checklist questions.
So when you are a clinical laboratory, there are a series of questions that you have, or
standards, if you will, that you have to meet in the clinical laboratory. Well, the CAP
accredits laboratories under CLIA with deemed status under CLIA, and we were accrediting
laboratories that were doing next-gen sequencing technology and had no standards or checklist
questions. We have now generated a starting point, trying not to be too restrictive in
the standards that we’re asking for, mostly focusing on documentation of what is being
done, what databases are being used, how things are functioning. And those will be in the
2012 checklist, so we do have standards, and that same group is now beginning to work on
proficiency testing around next-gen sequencing of anywhere from a gene panel to genome, transcriptome.
So, when we think about this, there are no IT standards for reporting in the LIS, the
EHR, and the personal health record. Many things flow directly into the personal health
record; interoperability standards are needed, terminology standards that Heidi already mentioned.
The molecular CPT codes are under revision but won’t address the issues, and there
aren’t next-gen sequencing codes available today or even planned. Early adopters are
finding that they’re negotiating coverage and reimbursement with each payer for each
patient. This is the early adopters setting.
I apologize. I already did these and was not able to update my slides, so. Sorry.
And the regulatory environment, the FDA held a meeting to understand the early clinical
users’ needs and concerns, but there is no FDA position or guidance available on next-gen
sequencing technology. There are no CLIA standards for genomic analysis. We have introduced those
into the CAP’s checklist, as I already said.
So, pathologists feel that they have an opportunity to lead in the medical community in genomic
medicine. There is no single medical specialty -- unless perhaps genetics, but that leaves
out cancer -- that is well-informed about genomic medicine as a specialty group. Pathologists
have an opportunity to be leaders in genomic medicine, considering that it is another diagnostic
testing modality, and that’s what we do. While genomic technology is rapidly advancing,
the discovery process for clinical genomics applications will be an evolution rather than
a revolution, because we’re going to be discovering one gene, one disease process
at a time. So we don’t want to ring alarm bells with pathologists, but they need to
keep up with the changing genomics landscape.
So, I’m just going to say here thoughts on genomics, and I’m not going to show you
all my bullet points because I’ve rethunk what I wanted to say --
[laughter]
-- based on listening to a few things, but also on my own thoughts. This testing, if
it’s used for patient care, has to done in a CLIA laboratory setting. And it’s interesting,
because when I walked in the room someone said to me, we’re trying to figure out how
to get the labs CLI certified. And I said, oh, that’s interesting, because I’m thinking
about how to get the CLIA labs to do next-gen sequencing. And it’s two different processes.
So I think one of the aspects of moving this into clinical care is really understanding
the business aspects of next-gen sequencing technology and what it provides, and how you
justify the costs. So I spent four years, intensely the last six months, working with
various hospital administrators to get my plan for a pilot -- it’s going to be a colorectal
cancer genomics project -- approved, to be able to buy the sequencers because we weren’t
a genomics sequencing institution, so we have to buy the sequencers -- how you justify that.
So I think business plans that include the clinical significance, and starting with a
pilot project is an interesting approach for those people who don’t already have huge
access. And so business plans are needed.
Also, when we talk about standards, I see that there’s a series of talks below, you
know, further on in the meeting, that are titled “Standards.” I see what they’re
going to be talking about, I think, as evidence generation. But I think of standards more
as the laboratory standards, the SOPs, the best practices, how you know that the data
you’re generating is accurate, what alignment protocols to use, what variant collars to
use, all of those technical aspects, as well as, I think we have to have discussions around
what are laboratory aspects that are regulated under CLIA versus what is the practice of
medicine in the interpretation of the sequences. And I think that that’s a very blurry line.
I don’t think -- I think you have to regulate the quality of the technical aspects, but
I think you also have to develop guidelines for that interpretive aspect, but I see that
more as the practice of medicine.
This is just the next technology for molecular pathologists, and I can’t tell you, I was
practically jumping up and down yesterday to be able to move from a talking head to
a molecular pathologist who will be doing this. So I think I’ll end my comments there.
I’ll just thank the IOM committee that I worked on, and the CAP committee, and the
CAP for letting me use their slides.
Male Speaker: David, and then John.
Male Speaker: I had a comment for Heidi, but it’s also
related to your representing CAP and the opinion on this. Going back to data sharing, I’d
like to suggest it’s not only about altruism versus proprietary databases. One of the unexpected
surprises with our IFCA database is NCBI wrote some pretty simple automated quality data
checks for it so that when I submit my data to NCBI they immediately send me back a notice
saying that I’ve classified the same variant two different ways, and would I like to look
at those and reconcile the discrepancy in my own data. And then they do a second check
where they compare my classification of a variant to all the other labs that have submitted
that same variant. And the proportion of discrepancies within a lab and between labs was shocking
to me, since we had all spent a huge amount of time on conference calls developing the
guidelines of how to classify pathogenic uncertain benign variants. And we thought we all understood
that we were doing it the same way, but lab directors within the same lab had different
interpretations, and between labs was even more significant. So I see this as a quality
control measure, and I see it’s something that CAP, CLIA, the ACMG, can all deal with
as a quality control recommendation or requirement to share data so that you get that kind of
feedback. And I’m curious your response to that.
Debra Leonard: I think that’s great. I don’t know what
Heidi’s thoughts are, since it was for both of us, but I think that would be extremely
useful quality check.
Heidi Rehm: I didn’t have time to go into some of the
depth, but David and I have talked about this before, and I think it’s a terrific concept
that if we could get the leverage within the standards communities and the regulatory communities
to leverage that fact, it would be useful.
Male Speaker: Do you have like training sets, and sort of
examples that you make people go through to sort of understand how to do that classification?
Male Speaker: Not so much training sets, but series of algorithms
about what kinds of data allow you to interpret something as pathogenic and …
Male Speaker: We could provide people with a set of sequence
data and say how would you classify this, and that would be a way of making sure that
everybody’s doing it the same way. John Harley?
Debra Leonard: Can I just make one comment, because there
was a point on my previous slide that I didn’t use, which is we need to think about how we
train the next generation who’s going to be expected to be doing this. And I think
we talk about how we’re practicing it today, but there aren’t a whole lot of groups talking
about how we’re training in residency training programs medical school; I mean, even pre-med,
is it relevant to do organic chemistry anymore? Or should we be requiring --
[laughter]
-- other courses? Sorry, sorry.
Male Speaker: Or the bones of the foot. Right.
Male Speaker: I’m struck by the comment you just made,
and the idea that before there were molecular pathologists, people used to look at slides,
in the old days, and make diagnoses on the basis of the slides.
Debra Leonard: They still do.
Male Speaker: And all the pathologists I know have these
huge collections of slides that they’ve had from all the patients they ever saw, and
this whole business of trying to go back and redo the data and reinterpret it has a direct
parallel to what those pathologists do. My pathology chairman still has a huge collection
of slides and tissues, and he’s constantly going back and getting patients from five
and 10 years ago, and reinterpreting those slides. And so I think we have a -- this idea
of having a data -- a large database that’s kind of fluid, we don’t know what to do
with it, and that that’ll be reinquired over and over again in the same sort of way
is just an evolutionary difference of the kind that you’re talking about.
Debra Leonard: Well, and in fact, looking at those glass
slides is evolving as well with digital pathology, which is kind of internal to pathology practice
but will allow algorithmic analysis of images and comparison of one image to the next; but
yes, pathologists do this.
[talking simultaneously]
Male Speaker: Soon we’ll be doing genomics on those slides,
too.
Heidi Rehm: We already are doing genomics on those slides,
so...
Male Speaker: [unintelligible] here. I direct the division
of informatics in a pathology department, so I could hear you singing for the past 20
minutes. It was wonderful. What I want to say is that the same way that we use annotated
images to train our statistical models, we need to capture this variability in the reporting.
It’s important that there is no sort of forcing of the classification. We need this
variability to be preserved in the reporting. Otherwise ,we’ll miss a lot of information
we need in the statistical modeling.
Debra Leonard: And yes, pathology departments are including
-- we have informatics faculty within our department and we collaborate very closely
with the Institute for Computational Biology. So it’s an aspect of the training, of having
people trained in genomics and computational biology or bioinformatics, and that combination
of experts is going to be needed as we move forward with genomic medicine.
Male Speaker: Jeff, last comment, and then we’ll have
a break.
Debra Leonard: You got me into this.
[laughter]
Male Speaker: Don’t blame me. [laughs] The FDA uncertainty
is somewhat stressful. I don’t know -- is there anybody from the FDA here?
Male Speaker: Yes.
Male Speaker: Okay. Oh, great. Well, welcome.
[laughter]
All right. Well, I was -- I was going to --
[talking simultaneously]
I was going to ask Debra to speculate where the FDA might put its stake in the ground
in this area because it seems that it could be at the level of the data generation but
it could equally be at the level of interpretation and analysis, and, or all, or the entire spectrum.
We’re talking about next-gen sequencing [laughs].
Debra Leonard: So am I to speculate when there’s an FDA
person in the room? [laughs] So the FDA did hold a meeting and looked at both the data
generation phase as well as the bioinformatics phase to hear from people doing this about
what is needed, and I think what came across at that meeting, although I only heard secondary
reports, was that regulation of the instrumentation would be helpful to clinical laboratories,
in that when you have the wild, wild, west and everything’s changing every month or
three months, it’s very hard to lock things down. And research instruments aren’t required
to do pre-notifications. They can change their chemistries, they can do anything they want
to us in a clinical laboratory setting that messes us up. So I think that’s one of the
aspects, and I don’t know what the FDA heard around the bioinformatics aspect, so maybe
you can comment.
Female Speaker: Yes, so we did --
Male Speaker: So could you introduce yourself, and then
talk into the microphone.
Zivana Tezak: Zivana Tezak, FDA. So I’m from office of
in vitro diagnostics, and I was one of the organizers of that meeting last summer, and
we asked a lot of questions, and many answers that we got back, as you all know, is this
is all in development, it’s changing all the time, and we have to -- we don’t know,
that was the answers for a lot of stuff. So we are working with the other government agencies.
We are working with CAP, and CDC, and NIST, and trying to get input from everybody to
try to figure out how to go ahead, because obviously we don’t want to stop developing
technologies, but also we want to make sure that the results people get back are correct.
So that’s in a nutshell, but we can talk about it.
Male Speaker: I think that this has been a really great
start to the two days, and we’ll break for coffee. It’s, we’ll be back at 11:00.
So instead of a half hour break, we’re going to have a 15-minute break.
[end of transcript]