Last month, GlaxoSmithKline researchers published data from a genome-wide association study [Nelson et al.] looking at hypersensitivity associated with the company’s HIV drug abacavir.
In the study, published in The Pharmacogenomics Journal, researchers reported that the known HLA-B*5701 gene’s association to the abacavir-related adverse drug reactions could be identified with sequential analysis of as few as 15 cases and 200 population controls.
In the article, entitled “Genome-wide approaches to identify pharmacogenetic contributions to adverse reactions,” Nelson and colleagues discuss how their study “demonstrate[s] through simulation that genome-wide genotyping, coupled with the use of clinical matched or population controls, can yield sufficient statistical power to permit the identification of strong genetic predictors of ADR risk in a prospective manner with modest numbers of ADR cases.”
Allen Roses, one of the study’s authors and the former head of genetics research at GlaxoSmithKline, this week discussed with Pharmacogenomics Reporter how the research into abacavir, also known as Ziagen,may affect the design and conduct of future studies trying to determine the appropriate patient population for a drug before it hits the market..
According to Roses, who is now director of Duke University’s newly launched Deane Drug Discovery Institute, a hypersensitivity test developed by the Laboratory Corporation of America that previously had a low adoption in the marketplace has seen a nine-fold increase in sales following the presentation of this new data.
Below is an edited transcript of the interview.
How is this particular genome-wide association study with abacavir significant for the design and conduct of future pharmacogenetic studies trying to gauge adverse reactions during drug development?
This is an eight-year project. It was begun in 1998 [at GSK] when hypersensitivity was noticed in patients. For the FDA, as part of its accelerated marketing agreement that allows you to put something for deadly diseases on the market, we agreed to do – which was incredible for 1998 – a pharmacogenetic study to predict who would be susceptible to the hypersensitivity syndrome.
The question we asked in this particular paper came after the report that we had in The Lancet in 2002 about HLA-B*5701 being found in a candidate gene scan. You have to remember the timing here. It was in 2002 that it was reported, it was found in 2001, and at the time there was no such thing as whole-genome screens or any of that stuff. …
Over the years we used much larger populations of patients retrospectively and prospectively obtained from our ongoing clinical trials with this drug. The data became more and more obvious that there was a difference between looking backwards on people who thought that they might have had hypersensitivity, versus looking forwards and actually looking at the diagnosis as prospectively.
We had previously published some work on a drug called Tranilast, which most people didn’t pay attention to because it ended up not being marketed for its adverse events. But what we wanted to put in context through [this abacavir] study … is, if we were going forward in clinical trials, how few patients would it take using genome-wide screening to detect the genetic basis for a serious adverse event.
So what we did in this paper was, we had information of HLA-B*5701, and what we actually did is, we took 22 patients [who had the abacavir-related ADRs] and some 200 controls who … had the drug but didn’t have the adverse event. … We ran the study in groups of five – five, ten, 15, 20 and then 22 – asking the question: When did we see what we would call a recognizable peak … and with 15 patients we saw what we knew would be there. We saw patients with HSR and also who had HLA-B*5701.
If you were doing a clinical study, as many big pharma companies do many clinical studies, if an adverse event showed up, there was now evidence that as few as 15 patients in 200 controls could give you useable information about an adverse event, and to do a predictive test.
Subsequently, after this work was done, we did a prospective study, which was just published in the New England Journal of Medicine. And that determined that HLA-B*5701 has greater than 99 percent specificity and greater than 97 sensitivity in detection of hypersensitivity syndrome.
Well, the next thing that people will say about that is, ‘Yeah, but we got lucky.’ Just like we got lucky with Alzheimer’s disease, right?
If you look at the literature, in Table 1 of the [Nelson et al.] paper, it shows how many patients the expectation would be that you would need, based on information about other diseases and other reports of adverse events. But there are two or three other adverse events immediately that you could look at using this methodology to make it your confirmation study. So you get your data retrospectively, and then you do a prospective study. And you’re looking at how few patients have to suffer the adverse event before you can establish it as a diagnostic.
That meets the criteria of the FDA, because we did it in the context of a clinical trial, which was the PREDICT-1 study that was just published in the NEJM.
There has always been the suggestion that pharmacogenetics would reduce the cost of clinical trials. Does this study, conducted in such a small cohort, essentially prove that?
It’s not about saving money. It’s about saving drugs.
This is the first time in the history of pharmacogenetics where … a serious allergic response to a drug can be diagnosed before the patient takes the drug.
Every drug has some adverse events. And some of those are going to interfere with continuing clinical trials, or withdrawal of a drug after clinical trials. If you could save those drugs by making the diagnosis during development as adverse events occur, you increase the benefit-to-risk ratio. And the drug stays on the market. That’s saving money, but it’s not saving money because it’s going to cost less to make the drug. It’s saving money because you have a drug out on the market from which you’re going to make money.
It’s important because you don’t risk people’s lives from now on and the drug would stay on the market. … And by the way, LabCorp put out a test [for hypersensitivity syndrome] on the market, based on the data that existed two years ago. And they had a very slow uptake of the test. … But, last July this [new] data were presented at an AIDS conference. Between that time and the end of 2007, the sales per month of LabCorp’s test increased nine-fold. And that is before the NEJM article was published.
If you can compare this with the … diagnostics for Coumadin – the doctors think they don’t need the test, that they can handle dosing patients without the test – for whatever reason, the uptake has been minimal at best.
This [test for hypersensitivity to abacavir] is inexpensive, and the uptake has been dramatic … from the point of view of diagnostic testing. And it’s been below the radar. It wasn’t done through newspapers, it wasn’t done through press. It was done by sheer science.
Other than LabCorp, do you know of any companies that may be developing a test for hypersensitivity to abacavir?
No, I don’t know of any other companies. GSK wouldn’t. What’s going to happen is, the drug is being reviewed by the FDA with this test. And it may be put in the label of the drug that the test should be done. We don’t know that yet. But it’s being reviewed by the FDA for that purpose.
GSK would not consider developing a companion diagnostic for abacavir?
No. GSK is not a diagnostic company. And they don’t have the outreach that LabCorp or Quest, one of these bigger [reference laboratories], have.
There are more diagnostic tests out there that gauge disease predisposition or recurrence than there are tests that gauge ADRs. Why is that? Is there a particular difficulty with conducting this type of study?
Yeah, absolutely. You have to collect DNA from people who have adverse events. So the only way you can really do that in clinical trials is if you have a company that’s doing clinical trials.
Of the susceptibility tests people are doing, none of them, come anywhere close to the susceptibility of APOE4 to Alzheimer’s disease. And just go back in the literature and see what a rough time that had.
So a lot of this is diagnostic business strategy. I’m not sure what the medical benefits are. But we were in these two cases, and there are others on the way by the way, able to [link] APOE4 for [susceptibility to Alzheimer’s disease] and HLA-B*5701 to adverse events [with abacavir].
What would you say to those who claim that you got lucky in this particular case with abacavir, that perhaps genome-wide screening with such a small cohort would not yield similar success with other drugs?
B.S.! No, seriously. It’s already worked with tranilast. It never came to market, because the drug had no efficacy. But we were able to diagnose with less than 20 patients, using three SNPs, who would get the adverse event.
Now, the problem with the people who do genome-wide screening, for the most part, they’re not biologists. They’re not physicians and they don’t have the opportunity to do these kinds of studies. It’s not that [GSK] has other things that are hidden in the company, we do, but it’s not that. But if you really want to see it, look at reports on tranilast from years 2000-2002. It was done without genome-wide screening, but we did an experiment where we mimicked what we would have found … had we done genome-wide screening … We used 1,000 controls … and we detected the ADRs at 20 patients and it got astronomical p-values by the time we were at 30 patients. By the time we were at 120 patients, it was double Avogadro’s number.
My answer to those people is, read the literature.
One of the problems with doing genome-wide screening doing 500,000 or 1 million points is that you have to correct for all the numbers of tests you are doing. So the signal has to be incredibly strong for you to detect it that way.
Are there other pharmacogenetic studies you’re conducting looking at ADRs?
Right now, I am consulting for other companies to set this up. The companies understand the [benefit] of having a test that can identify an adverse event can have in the life of their drug or their potential drug.