Interestingly, the way this study was done was in a large group of Amish subjects.
They're all related to each other, and
the relatedness allows one to estimate the heritability of the trait.
Relatively accurately, so the heritability of this trait is large,
73% of the variability in the trait is heritable.
That said, the signal that you see on the GWAS is right in the CYP2C19 locus.
Locust and in fact when the authors corrected foe the fact that there were
CYP2C19*2 they included that as a covariant that signal now goes away
completely, what that means is that the entire signal
that you're looking at is a sub CYP2C19*2 signal.
Interestingly, only 12% of thew variability in
the signal is accounted for by sub CYP2C19*2.
So if you're a pessimist, you say, well, there's a huge amount of variability in
this trait and a little bit of it is due to CYP2C19*2, and
a lot of it is due to something else that we don't understand.
If you're an optimist, and geneticists are optimists,
they say well, It's extremely unusual to have a single snip that accounts for
12% of the variability in any trait,
be it cancer susceptibility or acute myocardial infarction or whatever.
So the genesis would say this is an enormous signal, and
if you infactor in known clinical factors like BMI and lipids.
You can increase the explanation for the variability, but
you still have a large chunk of the variability in this trait unexplained, and
that's a pretty common scenario across the genome-wide association world.
So people have gone on to look relatively retrospectively at
the outcome of clopidogrel therapy as a function of genotype in very,
very large clinical trials.
These are meta analysis showing the results of the individual trials,
they're confidence intervals and an over all.
Estimate of the affect size among subjects carrying
one copy of a loss of function allele.
And that generally means CYP2C19*2 at least in the non Asian populations and
the point estimate is about with an odds ratio about 1.5 for a recurrent event.
That means that there's 50% increase risk that patient taking clopidogrel
will have an adverse event at the normal dose after receiving clopidogrel,
like our patient did.
Patients with two copies of varied alleles, there are fewer of them, so
the point estimates have wider confidence intervals.
But the odds ratio is 2.8, so that's pretty high.
So the question is, do you in fact test everybody beforehand?
The optimist would say well, I want to know whether my patient is a star 2 star
2 perhaps even a star 1 star 2 because I might choose a different drug or
I might choose to increase the dose of the drug.
The pessimist would say well that's a pretty small effect.
Keep on telling about odds ratios of 100 and 200 and those are actionable.
Here's an effect size that is 50% so why should I get excited about that?
There's so much variability you haven't explained maybe what I should be doing is
testing the way in which platelets aggregate during therapy.
It turns out their patients who get clopidogrel and you test their platelets,
and their platelets still aggregate just fine as though the drug weren't there and
there are patients who aggregate just fine.
So maybe instead of measuring genotype I should measure net drug affect downstream.
And there are such tests that's controversial exactly where they fit in
and how good they are predicting outcomes.
The other thought is, well let me do the genotyping
among patients who are star one, star two, I'll just increase the dose.
And there's a little bit of support for that approach.
Not perfect, but some support for that approach.
What's interesting is, if you try to increase the dose in the stars two,
star twos who have no CYP2C19 activity, or very little activity.
You can't increase the dose enough to overcome the genetic defect.
So one strategy is to say, well, the star two, star twos, certainly,
have a defect and maybe they should be considered for an alternate therapy.
Star one, star twos, maybe, maybe not,
and that's the way many people approach that now.
So clopidogrel is a Pro-drug and it is bioactivated, and
this bioactivation occurs through a single pathway.
So the problem with a single path way of course is that if
anything interferes with its activity.
Then the Pro-drug accumulates and the active metabolites don't get made and
the pharmacological effect is missing.
So a Poor metabolizer trait is an example of how you might lose that bioactivation.
Are there other ways? Well,
it turns out there are drugs that inhibit CYP2C19.
The most commonly used inhibitors are the proton pump inhibitors,
notably on Leprosol.
And these are data, retrospective data, from a large pharmaceutical benefits
provider that looked at the incidents of recurrent events after
an initial prescription of clopidogrel for an acute coronary syndrome.
And what you can see is that among patients who ended up on clopideogrel at
a PPI there's a higher incidence of events compared to patients on clopidogrel.
This is retrospective, it's not randomized, but
it still suggests that there's a consistent signal.
One question you might ask your self is well,
maybe that's just because PPIs make people worse.
PPIs have some affect on platelet aggregation, but
it turns out they had a group of people that were only on PPIs.
They didn't seem to get worse than people who received neither drug.
So the answer is, it looks like from this retrospective examination,
that there really is an interaction.
