At A Glance
Name: David Cox
Title: Chief Scientific Officer of Perlegen Sciences
Background: Consulting Professor of Genetics, Stanford University — 2002-Present; Professor of Genetics, Stanford University — 1993-2002; Associate Chair, Department of Genetics, Stanford University — 1996-2000; Co-director, Stanford Human Genome Center, Stanford University — 1993-2000; Co-director, UCSF Human Genome Mapping Center, University of California at San Francisco — 1991-1993; Professor, Psychiatry, Biochemistry and Pediatrics, UCSF — l990-1993
Education: MD from the University of Washington — 1975; PhD in Genetics from the University of Washington — l975; MMS in Medical Science from Brown University —1968-70; AB in Biology from Brown University — 1968
As part of the 54th annual American Society of Human Genetics meeting in Toronto last month, the invited session “Pharmacogenetics in the Clinic, the Laboratory, and Society” featured the lectures of six important figures in the field, including David Cox of Perlegen Sciences. In his talk, “Whole-genome scans, large-scale genotyping, and the study of variable drug response,” Cox presented his case for the future of pharmacogenomics in the healthcare systems of the world.
He is an elected member of the National Academy of Science’s Institute of Medicine, a member of the Health Sciences Policy Board, and a former member of the National Bioethics Advisory Commission and the advisory council of the National Human Genome Research Institute, among other professional service organizations. Cox is the author of dozens of scientific articles spanning the field of genetics.
What is your impression of the future of pharmacogenomics?
First of all, I think pharmacogenomics is used, and will be used in many, many different ways. So, what one of the things that has really focused that, in terms of the pharmaceutical industry, is ways of making it cheaper and faster to get drugs through the pipeline and to have them work. And the second aspect of that is, once they get through the pipeline, use pharmacogenomics to get more things approved by the FDA to change the risk/benefit.
But at the end of the day, there is this other vast amount of data that’s involved with human health, which is after all of those things have happened, there’s a drug that looks like it works, it gets approved by the FDA, people start taking it — but then this vast accumulation of information about how well it works — things you didn’t know when you got it approved, but that you learn later — and you can sort of titrate based on that knowledge. And right now, those data aren’t really collected, and they’re not really used. But if you had those data, then you would be in a position to take existing treatments and use them much more effectively.
Now, from the pharmaceutical perspective, the pharma industry isn’t really that interested in that, but from the point of view of human health of a particular country — I know a few countries that wouldn’t be interested that because most of the expense of medicine is a result of a small fraction of all the people who get sick. So, if you could identify all those people early on, in a prospective way, and do something about it to prevent them from getting sick, you’d save so much money that you could use that and improve the health of everyone.
Now, this sounds pretty Pollyanna, but then it’s not so crazy as long as you have a healthcare system in your country that can take advantage of it.
So, let me give some examples. In the United States, we don’t have single-payer healthcare, but we have a Medicare system. What if the doctors who treated patients with Medicare got economic incentives if their patients got better and got less money if their patients didn’t. So it’s not based on how many people they see, but it’s based on the outcomes of the overall health of the population that they’re looking at. Well, that’s dangerous because then doctors will only want healthy people. This is not so simple, but that would be the idea. But even more important is that it’s not really researched — it’s clinical medicine these people are getting, the treatments — but if you would go through and then use DNA, as well as other information, to identify, to be able to predict who’s going to do well and who’s not going to do well. Then you can use that in individual patients. What you have to do, though, is address the questions with a set of 300–400 people: people who the drug works on; who it doesn’t work on; who gets an adverse effect; and who doesn’t. Test that again in a separate set of people, and then that’s the information that you need that can be applied to individuals.
So, what we don’t have in this country, is — we have plenty of people that are being treated — all that data that’s available — but it’s not being collected or utilized in any way. In the [United Kingdom], people have realized this, so they said, ‘We’ll have a biobank. What we’ll do is we’ll start off and collect 500,000 people that are sort of random. And we’ll start off with them as adults and we’ll get their DNA and we’ll see what happens to them. We’ll see how many people get heart attacks, see how many people really do get ulcers with respect to a particular drug, or how many people get reactions.’ And you have all that information, you have the DNA genotypes, and you can use it to mix and match medicine better.
Sounds great, right?
What’s the problem?
The problem is that it’s pretty expensive to get the consents and the DNA from 500,000 people. But that’s cheap compared to the cost and accurately keeping track of who gets what. So, it’s a false promise in many ways, because the funding is never there for the part that really matters, which is the phenotype. The funding is there for the DNA. So you’re going to have all these genotypes and have no place to go, because you won’t have the right questions that you’re trying to deal with.
So, am I being too pessimistic? I don’t think so.
How do you change the situation?
One way to change it is to take advantage of the existing medical care system, where we already have to keep track of the data. And we use that. We don’t use a separate research surrogate of a cohort of 500,000 people that are the research subjects, where later on that stuff gets applied to clinical medicine.
There’s a good reason why research is separate from clinical medicine — because you don’t want to put individuals at risk for research. But we’re not talking about research here, we’re talking about people who are already getting the drugs, they’re already at risk. We’re just saying, ‘Use that information and apply the DNA to that,’ rather than take a separate set of people with the DNA and have to characterize their clinical information.
How about privacy? Well, you don’t need people’s names for this, you don’t need to track anything about people for this. What you do need, though, is to have some way — there is longitudinal data; I don’t mean to blow off the privacy issue on this, but I think it’s doable.
So you think it’s possible, then?
Oh, I’m really optimistic. But I think that in societies like the American Society of Human Genetics, the pharma industry isn’t going to drive this. The NIH could drive this — it hasn’t because it’s a research arm. But if NIH can work with Medicare and healthcare reimbursement in this country, and they go together, then that’s a new opportunity.
Many people see [lagging] reimbursement as one of the major stumbling blocks to wider implementation of pharmacogenomics.
Yeah, it is. I’m not an economist, but I will tell you one thing is that if you can go in and you can get that small fraction of people that use 90 percent of the money, and you can keep them from getting sick, I’ll be laughing all the way to the bank with the amount of money that will be there to help our healthcare system.
This isn’t a novel idea on my part. There are a number of leaders in the country that are more articulate on the topic than I am. Mark McClellan [former FDA commissioner who now oversees Medicare] for example.
What is Perlegen doing to bring pharmacogenomics into healthcare?
Perlegen is a fairly young company — we just got started in 2001. But during that time, what we’ve done is develop the technology. Not to sell information to people, not to actually sell technology to people. What we do since we have that technology and can execute it in a rapid and cheap way — it allows us to work with individuals by ourselves to address questions that one couldn’t address in the past.
What Perlegen does is, it uses this technology that allows us to look at DNA variation across the genome and we pick questions where being able to answer what the genetic contribution is, it will have an economic impact.
So, it doesn’t do you any good if you have to account for 100 percent of the differences between people to solve a particular problem, because genetics is only half of the story.
But let’s say you had a drug side effect, where a company spent $700 million on a drug and it had twice the side effect of the drugs already on the market in the class. It didn’t kill people, but it was a serious enough side effect that the drug didn’t get approved at all. But if you can go in and reduce that risk from a factor of two down to the risk of [the other drugs], that goes from a zero-dollar drug to being a multibillion-dollar drug. It’s like selling houses in expensive neighborhoods — you don’t have to sell many and you make a really good living.
More importantly for medicine, you go out and try to make predictions about how you can prevent diarrhea in people that are taking these cancer drugs, or prevent other side effects. And you put this stuff together and people have better lives. And if you improve people’s lives, you have a good business.
Some people think about genomics, ‘Well, it’s a commodity business, it’s all about genotyping.’ Genotyping is only the technology, but the business is being clever enough to figure out what are the important problems and which ones your technology can basically get you over the hurdle of, and if it’s important enough to people, it can make a big economic difference.
And Perlegen doesn’t sell its genotyping data, right?
Nope. In fact, because we really believe in using the information, we’ve recently made all of our SNPs publicly available. We genotyped 71 individuals: 24 Caucasians; 24 African-Americans; 23 people of Chinese ancestry. We put all this stuff in a public database. That’s stupid, right? No. Because what it does is, it shows that the name of the game isn’t hoarding the information or selling people the information, it’s using it.
And how does the company make money, for example?
An example of that — [there is] a high blood pressure drug I mentioned earlier — [Perlegen is] working with pharma companies in being able to predict who’s going to have a side effect and who’s not, so now that drug that wasn’t under consideration is now back at the FDA, with the pharma company taking another shot at it.
Can you say anything about collaborations that Perlegen is going to be doing?
I think it’s six of the top eight pharma companies we’re doing stuff with. Each company has a different problem.
In addition, we in-license drugs that people think aren’t very useful and by doing these types of approaches, we figure out the people that they work in and that’s like being a junk dealer. What you do is get what’s junk to someone else and you turn it into a useful drug that brings real value to your company.
Reviving drugs that companies have given up on, or have pulled off the market?
Well, it’s hard to revive them if they’ve been pulled off the market, but if a company has given up on them, then you can license them and do them. Now, we don’t do that in anybody’s face, we can do it in collaboration with companies, we can do it on a retainer.
And the final thing we do is, we do a lot of stuff in the public sector on the basis of grants or foundations to work with academics to be able to get the fundamental genetic components of diseases — like with the Michael J. Fox Foundation, we’ve got a big grant to do Parkinson’s disease with collaborators at the Mayo Institute.
When did that start?
It was announced about a month ago, and by the end of the year we should have some results and all that’ll be public.
We have a grant from the National Institutes of Aging working with academic collaborators on Alzheimer’s for a whole-genome scan. That was recently, within the past month or so.
We have a contract with NIDA on a study of nicotine addiction that should [start by the end of 2004].
And so these are things that the company does — by doing it, it doesn’t cost us money — we may make some money from this, but the main idea is not making money on this sector, it’s going in and demonstrating that the technology provides useful approaches.
Since Lawrence Lesko [director of the FDA's Office of Clinical Pharmacology and Biopharmaceuticals] was here, I was hoping you’d tell me what you want to see from the pharmacogenomics voluntary submissions guidelines.
I think Larry was really very straightforward. I think what I’d like to see is guidelines based on empiric data, and we need to get in more empiric data. But it’s not going to be like a statistical odds score, it’s a cost-benefit ratio. So depending on the question that you have, you’re going to have different hurdles to get over. And I think that the more different examples we have, the more people will be able to see ways that they’ll be able to win, both economically and in terms of utility, with pharmacogenomics.
Are you worried about submitted data affecting drug trials?
Not at all. In fact, I think the more data that gets in there, the smarter we’ll all be in terms of where this is going to work and where this not going to work.
Is there something you’d like to add?
I think that people should keep their eye on the ball, and not be swayed by people who are too optimistic or too pessimistic. But that until we see some practical examples where this really impacts people’s lives, no one should be celebrating. On the other hand, people shouldn’t be too pessimistic because I think we have opportunities that we haven’t had before.