Tip:
Highlight text to annotate it
X
Mary Relling: Okay, Brandy, do you want to -- are those
my slides? Next slide. I agree with a lot of what Dan has said, and am so impressed
with what both Rex and Dan presented about potential for discovery in eMERGE using the
EMR. I do want to start off making the point that I think that it possibly may take even
more resources to fully utilize the data in EMR to do the kind of discovery that we can
do in clinical trials. Next slide.
So, an example of this is the example of Topoisomerase II inhibitor, such as a Teniposide- or Anthracycline-induced
secondary leukemia. This is an event that was discovered over 20 years ago, and it can
have as high of a frequency as 20 percent of patients treated with these drugs and uniformly
fatal disease. So, finding out the risk factors for this disease was critical. Next slide.
And I want to give you an example of a protocol. So this is the clinical trial that randomized
patients to two different schedules of exactly the same drugs. So these patients get treated
with weekly chemotherapy for 120 weeks, and the only difference between the patients randomized
to Arm II verses Arm II was that Arm II had the same pairs of drugs rotated every week
and Arm III had the same pairs of drugs rotated every six weeks. So the cumulative doses over
a two-year period of all these drugs were absolutely identical. Next slide.
But the frequency of secondary leukemia was much higher in the top curve in those who
had the six week blocks of drugs verses those who got the rotations every week. And I use
this as an example to point out that if we're going to do things like survey EMRs for drug-related
risk factors for adverse events, including fatal ones like this, looking at cumulative
dose alone is not enough. Next slide.
And the same thing was true in another fatal adverse event: a secondary brain tumor in
patients treated with leukemia. This slide depicts the frequency of secondary brain tumors
in four different frontline clinical trials as childhood AOL. They all went on for two
and a half years, and they all had exactly the same dose of cranial radiation, which
is the primary risk factor that was known for development of secondary leukemia. But
you can see in the top curve, this Total XII trial had a ridiculously high frequency of
severe fatal secondary brain tumors compared to the other trials even though the dose of
radiation was identical. Next slide.
And we found actually a genetic feature in that study, a defect in TPMT, which you just
heard about, which predisposed patients to develop this secondary brain tumor much worse
than the patients who did not have an inherited defect in TPMT. Next slide.
However, these patients that have a defect in TPMT have always been around, and it's
about 10 percent of the population that have that defect. So what was different about the
Total XII study that made that genetic variant shine through and have a penetrant effect
on this adverse drug affect that was not true in the other studies? Next slide.
So by looking carefully at exactly what therapy was given, her protocol among these four protocols
during the time period of a radiation week identified that it was the degree of antimetabolite
intensity only during the two weeks of radiation that was different about Total XII compared
to the other studies that had a reasonably low level of secondary brain tumor. So, again,
an overall chart review that looked at just cumulative doses of drugs or cumulative doses
of radiation never would've found this schedule-dependent interaction that was a strong risk factor
for this fatal complication. Next slide.
And so here we have this interaction of host genetics only shining through with its adverse
event on secondary brain tumor that was contributed to by the use of Thiopurine drugs during the
cranial radiation. Because the unfortunate mixture of these drugs during the period of
cranial radiation, very difficult to identify this kind of detailed drug-therapy interaction
from chart review alone. Next slide.
So, I guess the point that I would make is that I think that the EMR data being accumulated
by eMERGE is absolutely fantastic, and then parallel, that the tools that will be needed
to query it to come up with detailed information about schedule and timing of doses verses
other interventions is probably going to take even more resources.
Now, as Dan just alluded to, the other point is, can you implement something if you don't
already know that it works? Next slide.
And we are, of course, looking at things like some of these pharmacogenetics gene-drug pairs,
which we've worked on with the CPIC supported by the Pharmacogenetic Research Network. Right
now we've got about 13 genes affecting about 60 drugs that are affected by CPIC guidelines.
They are ready to implement now. However, I definitely agree with Dan that for rare
variants of these genes we don't know necessarily to implement them, and some of the data have
been generated in ancestrally non-diverse groups which would make implementation in
all ancestral groups potentially problematic. However, I do think that there has been a
difference between doing implementation for that which one can, with relatively common
variants in well-studied ancestral groups, and collecting new data, trying to find new
associations between rare variants in those same genes affecting those same 60 drugs.
I do think that some of the trials ongoing, not necessarily in eMERGE but elsewhere, where
patients are being randomized to have clinical implementation of genomics, especially pharmacogenomics
verses not, are potentially problematic because when things are ready to implement they should
really be implemented, and it wouldn't really be ethical to withhold something that's ready
to be implemented. And some sites are getting around this by sort of capitalizing on our
unfortunate non-uniform healthcare system to take advantage of natural randomizations
where some sites can do genomic medicine and others can't. But that's really just an accident
of our non-uniform healthcare system. And studies that purport to look at the effects
of genomic medicine on clinical outcomes verses historical controls I think really risk having
all of the problems of poor study design, coming up with misleading answers, because
the non-genetic clinical co-variants are so critical in deciding whether something works
or not that aren't controlled by these studies are really a problem and can never be addressed
by a historical controlled. And that also affects, of course, system randomization.
Next slide.
So I guess I would just end there by saying that I think we have to -- I think capitalizing
in eMERGE on both discovery and on implementation as it sounds as if you're doing is a fantastic
approach. But let's not mix up the two. When something is ready to implement, it's used
as a clinical tool, and generating additional clinical outcome data may not be the best
use of resources. If something's not ready to implement, it's worthy of clinical research,
and eMERGE is better poised to do that than anyone. I'll stop there.
Male Speaker: Okay. With that, thank you and we'll --