Clinical genomics: What’s real, what’s next?
Insights from the OptumLabs Research & Translation Forum
How can genomics improve health care?
Technology is making it easier and cheaper to analyze all or part of the human genome and obtain health insights. This is catalyzing a rapidly evolving field of medical genomics.
While genomics has helped diagnose unknown diseases in families, there are diverse views about the cost/benefit of using it to screen for prediction and prevention. There is ongoing discussion about what information is actionable now, how to apply and share it, and guide usage.
Below we share two expert perspectives on where genomic medicine and emerging technologies can add value in health care across diverse use cases, highlighting the challenges, benefits and future outlook.
The evolving landscape
Robert Green, MD, MPH, medical geneticist and professor, Harvard Medical School, provides an overview of the changing landscape of genomic medicine, and shares his views on where genomics is advancing medicine now, the impact of the consumer, and opportunities to expand its use going forward.
Opportunities and challenges
Listen to Robert Green, MD, MPH, and Sek Kathiresan, MD, CEO, Verve Therapeutics, discuss insights from their work and what aspects of medical genomics are ready for action, moderated by Paul Bleicher, MD, PhD, Strategic Advisor, OptumLabs.
- I am gonna say that we're really privileged today to have two of the really leading experts with us to talk about genomics. The people who are most likely if I open the New York Times or The Wall Street Journal on a given morning and there's a genomics article, one or the other of them or both of them are likely to be quoted. So this is really a great opportunity to learn for all of us and I consider myself a genomics dilatant and I'm an opinionated dilatant. so I'm gonna be asking questions and saying things and they may tell me I'm wrong about it, which is great because we'll all learn in that. But I do wanna frame this at the very beginning because we're in the early decades of the 21st century and I think we can really look at genomics in parallel with or we can use an analogy from physics at the early part of the 20th century. This just occurred to me this morning as I was thinking about how to start this. Because physics actually had a debate going on between deterministic physics, that is classic mechanics going up to relativity where Einstein believed that the universe could be fully explained if we understood the hidden variables that we were unaware of and probabilistic quantum physics, which Niels Bohr and others said it was really impossible to determine because it was a probabilistic model and things couldn't be nailed down precisely. Einstein in that debate famously said, "God doesn't play dice with the universe." Not because he was a deeply religious person, but because he was making the point that the universe has to be explainable in some ways. But we have to recognize that from the beginning of classical physics until today, we've used these concepts to develop models and those models are useful. And so in genomics, we're in the same place. A lot of people, you've heard the term that genomics is destiny, you've heard that genomics isn't destiny. There is the knowable, the idea that we could, that there are all these minute knowable pieces of genomics, and once we have everybody sequence to know their epigenetics and all that, we're gonna, we pretty much can predict everything, but clearly nutrition, lifestyle, medical care, the environment, those are a real large component of what goes on to make this up and then a component of randomness, which may or may not be explainable by hidden variables. And then there is the probabilistic idea, which is that there is this randomness and that is useful in understanding populations, but it means that we are challenged a little bit when we get down to individual patients. And that's a distinction that I think is really important to keep in mind. And I will just conclude that little introduction by pointing out that Gregor Mendel actually struggled with this when he invented genetics more or less in the mid-19th century when he was studying peas because Ronald Fisher, a famous statistician, Fisher's exact test, etc. in 1936 famously wrote a paper that showed that Mendel, actually, his results were far too good for what he should have been finding if you were probabilistically getting the data he got. And in fact, there was a component of error, a percentage error, which should've biased the number to around 225 and he knew it should have been 200 based on, and he basically pinned it at 200. So he fudged the data a little bit. The good thing was he was right about what he did and we're happy that he did it. Not to encourage anybody to fudge data. In any case, the topics - the way I formulate genomics are into a group of topics and we're gonna cover a bunch of them today. So I'm just gonna quickly run through them and then I'll sit down and we'll begin, I'll stop doing most of the talking. So preconception and prenatal planning, which I think is pretty clear. Newborn in adult screening mostly for high penetrance autosomal dominant diseases for which especially there is an intervention that's possible and also some autosomal recessive diseases as well. Tay Sachs disease, for example, which doesn't actually require genomics, but there are others like that. The diagnostic workup of diseases, this is becoming more and more important as we understand the genetics behind kidney disease or connective tissue diseases, etc. And also, obviously, that's genomic information, but somatic information as it relates to cancer has become a very significant, and possibly as of today, maybe a little bit over-hyped aspect of genomics or genetics. The diagnosis of syndromes, the diagnostic odysseys, patients who are, have been spending tens of thousand of dollars a month for months and years on end, often I think it's something like 60% or so can be actually identified by genomic testing. Pharmacogenomics, polygenic risk scores, which we'll gonna go into in great detail as well as the screening gene therapy and drug discovery. I know there are others that I've left out, but those I think are good categories. So I'm gonna start with, we're gonna do this just a little bit differently. I'm gonna focus on one and then the other of our guests and then we're gonna do a jump ball. And you guys don't have to both respond to each thing, but we wanna leave plenty of time for questions as well. So I wanna begin with Robert Green, and I encourage you to use your app to look at everybody's biography. I'm not gonna really go into it. But, Robert, I'm gonna ask you just to give us a brief sort of flavor for what you have been dedicating your genomic research and clinical work on.
- Thanks, Paul. It's a pleasure to be here. Many of you are already expert in genomics, so I'll back up just a moment for those who are not. I actually trained in neurology originally and got fascinated by Alzheimer's disease where I did a lot of clinical trials. And then I got into the epidemiology of Alzheimer's disease and from there, into the genetic epidemiology and from there, into genetics. Actually, went back, got an extra training in epidemiology, went back again, got extra training in genetics. And so, I have the zeal of the recently converted and I also have that public health feeling of the importance of making it clear to the world what are the benefits and complexities of particular problem. So I don't know about you, Sek, but as far as getting into the Wall Street Journal and so forth, I have had one of my bosses sit down with me and said, "Robert, I'd like a little more New England Journal, "a little less Wall Street Journal." Nonetheless, I persist. And , so my work has really been trying to bring that clinical trials orientation to these complex questions in genomics. And exactly as our previous panelists were talking about, that tension between observation and real clinical trials rigorous evidence, it's particularly tough in genomics where genomics has been built on rare diseases. So there are extremely small ends of almost every type of rare disease at least to try to work with. So we are probably best known for our MedSeq project and our BabySeq project. MedSeq is the first randomized trial of comprehensive whole genome sequencing in adults. We actually took a group of adults who wanted to get sequenced, randomized them into people who we interpreted fully and people we didn't interpret fully. The one finding was they were very disappointed on this side. And then a couple years later when the furor for doing that had died down, we took newborn babies, healthy newborn babies, and randomized them in the same way. And I would just say that the most startling finding is this: Rare mutations for dominant diseases are not actually rare. You're gonna find this hard to believe, but we've now replicated this in three different samples that 20% of this room is carrying a dominant mutation for a rare Mendelian condition. Think about that for a second. Now, your first instinct is gonna be, oh, yeah, but that's 'cause variable penetrance. Well, another paper that we did with Sek, in fact, suggested with very small numbers, but still suggested that if you follow that 20% over 20 years, a very high percentage of you will manifest some feature or fragment of that underlying monogenic disease. Now, put those two together and you can see that our vision of rare disease at the moment is extraordinarily parochial. We see it in terms of genetics clinicians who happened to see somebody in their clinic at one point in time or maybe a few years, but never followed them over decades. So rare disease is probably way less rare than we thought and it probably manifests over decades in ways that we haven't even imagined. I'll stop there for the moment. There's a lot of other nuances to our work but I think this is the most important sort of substantive discoveries that we've made.
- So I wanna dive in with a few questions about that. First of all, whole genome sequencing, that's great for a research study. The price has come down and down. I've heard $1,000 genome, I think we've passed, I think I've heard numbers like $300, $500 genomes, but that's not all the cost of that. There's a whole bunch of interpretation a lot of which is quite manual at this point because we all wind up with a bunch of mutations. So, can't we do this all with chips and arrays, which are really cheap and are sort of--
- Absolutely not, because - a couple of reasons - Number one, with chips and arrays, you're not going to get the super rare variance in any kind of accuracy. Number two, you're not going to get the private mutations that you have in your family and nobody else has. And number three, there is gonna be increasingly rapid knowledge about regulator variance and variance that create stochastic enhancements and so forth that are going to require different parts of the genome to explore. I think we're at a period now where one manufacturer has kept the price of genome sequencing a little bit higher than it needs to be, number one. Number two, you're right, interpretation becomes more expensive and more complex, the more genes you look at. And even people who are kind of knowledgeable about genetics forget this. If I do your whole genome and I say I'm gonna look at 59 genes, that's pretty easy and quick. If I do your genome we're gonna look at a thousand genes, that's a little tougher. If I'm gonna look at all 5,000 or so genes that have been reliably associated with human disease, that's gonna take a while and it's gonna have a manual component. So there is the sequencing cost, which is down to $300 to $500 to $600 for a CLIA sequence and there are the interpretation costs, which is entirely dependent on how many genes you analyze.
- So, and there is an argument to be made, and we can discuss it further but not necessarily, that in medicine, we do a lot of things where we get 80% and a long tail we miss and let go, especially as it relates to screening. but that's a policy debate more than a scientific debate because it's clear that there is that long tail. Screening has been a bit of a mess for us though in medicine for a long time. And more recently, whole body CTs with people getting lung nodules and lung biopsies and things like that. Thyroid cancer in Korea where you see a large increase in the incidence of thyroid cancer and no change in the death rate. And PSA screening is now a shared decision rather than something that is, along with CBCs and standard routine EKGs, ECGs. I'm dating myself. So tell me why this is different.
- Well, there may be people that object and that's cool, and I am a proud card-carrying public health professional as well as a genomics person, so I acknowledge the validity of concerns about unanticipated outcomes both in terms of cost and side effects. Absolutely a valid question. However, in the practice of medicine separated by screening programs in the practice of public health, we often get information that we take at a first cut and then we contextualize that based on family history, an enhanced physical examination, other sorts of diagnostic tests. So whereas I don't think we want, just to exaggerate, a public health program that says let's do whole genome sequencing and comprehensive interpretation of every newborn baby now, I don't think we're ready for that. What I would say is if any adult in this room wants to get sequenced and interpreted, we will find things in you at which point, we will circle back to you, we will recheck your heart, we will ask more about your family history, we'll do some additional diagnostic testing. We'll take that now new prior probability and we'll adjust it up and down with continuing iterative examination the way we do for any other risk factor. So I don't think this is a public health mandate like newborn screening or seatbelts, but I do think it's ready for contextualized medical practice like the rest of the practice in medicine.
- Great. So, Sek, let's turn to you. We've been talking about the screening for a whole bunch of rare diseases, which maybe turns into common diseases as we've heard, but you've actually been looking at using genetics - and we'll get to Verve later I think - but you've been looking at using genetics for a more common, one of the most common diseases is one you concentrated on and the risks that you can get from looking at a bunch of different genes. So, do you wanna tell us a bit about your work?
- Yeah, our story is I guess really focused. So I started about 15 years ago wondering why some people have a heart attack at a young age. And so we studied people who had a heart attack in the 30s, 40s. And when heart attack occurs at such a young age, we often think it's genetic, right? So we took 5,000 people, men and women who had suffered an early heart attack, and asked, could we look at their genome and figure out if there's a cause? And for the last 20, 30 years, the best understood model for how a gene could cause early disease is the monogenic model. So that's basically a single mutation that is sufficient to lead to early disease. So genes like BRCA1 or LDL receptor for heart attack where mutations in LDL receptor, for example, can elevate LDL. Pretty high level starting early in life and those patients are at markedly increased risk for heart attack. So when you look at the 5,000 people with early heart attack, we found that only about 5% of them could you actually find a mutation in these monogenic genes that have been previously known to, shown to cause heart attack. So that left a lot of people who had a heart attack at a young age and we're like, well, what's going on here? Certainly could be their lifestyle, smoking or could be other behavioral factors, but there's still a lot of genetics that was left to be explained. And what we have identified is that in a large fraction of the remainder of individuals, the predominant genetic model is not the monogenic model where a single mutation is sufficient to lead to early disease, but rather the polygenic model, the additive effect of many gene variants in aggregate basically can push a person to early disease. And so this model really is relevant for this disease and it turns out it's relevant for really most other common diseases as well. And what we're able to do is take all of the variation in the genome that you might carry, these are all common variations, and basically sum up that information into a single number, a susceptibility number for each person for each disease. And that number, if you plot it in the population follows a bell curve distribution like cholesterol or blood pressure or other quantitative traits. And so what you have now is you're able to reduce the risk of any given person for any given disease into a single number. That number has a bell curve. Some people are high, some people are low, most in the middle, and you can ask what happens if you happen to be on the high end of the bell curve? And what we found was that those early heart attack patients, if you look at the relative contribution of the monogenic and polygenic models, roughly two out of every 100 patients with early heart attack, it's because of a monogenic model, but at least 20 out of every 100 patients with early heart attack, it's because they have a very high polygenic score. So that's really been, in three minutes, the last 15 years.
- So, so we've been... It's fascinating. And as Robert has explained to me in private meetings before and I've heard in my reading, the polygenic risk scores are now approaching or even surpassing the what we consider a threshold for action in the autosomal dominant genes. They're like showing five to six X or even beyond the average risk, which is enough to make it make sense. But one question I always have, and I've seen literature on both sides of this, is we have Framingham risk studies, we have all sorts of risk studies, how much, in the statistical language of this, how much are the variance of the polygenic risk score is explained by the standard risk if you will? How much is left over once you've described standard risk?
- Yeah. So, of course, the unique value of genetics is it can be determined early in life. And then if there's actions that can be taken, then they can be worked on. If you think about the current way we approach risk prediction for heart attack, which is like the Framingham risk score or the pooled cohorts equation, which is the latest version and those are calculated typically in middle-aged adults. And we wondered whether the polygenic score, how correlated is it actually to the pooled cohorts equation or the Framingham risk score? It turns out the correlation coefficient between those two is like 0.02. So there's almost no correlation between this new genetic number, risk number, and what we use in clinical practice right now. And so this new risk number basically adds additional value in perspective models to what we already do. And the new risk number, the main difference is it can be calculated early in life. In fact, there's a very curious result where if you start your risk models for heart attack with the polygenic risk score, okay, you end up getting an AUC of about 0.72, okay. That's the same AUC if you start with the Framingham risk score. So in some sense, it's actually what you start with and you can imagine a world in the future where instead of starting with somebody's cholesterol in their 50s, you're starting with their understanding of their polygenic score earlier in life gives you a stratification in the population of people that are much higher risk than others and you can put them into bins.
- Jeffrey Drazen and David Hunter had a short editorial in The New England Journal of Medicine about polygenic risk score, and they pointed out a bunch of things, including the fact that behavior is a challenging thing to change and we find that with, especially around obesity, which is one of your polygenic risk scores that you worked on has recently been published on. I know that, Robert, you have some actually data on that, but also, that there's a lack of ethnic and racial diversity in the genomic database that's been used and their advice was let's wait a little longer. They said they were asked this 10 years ago and the implication was let's check-in again in 10 years. So, how would you answer them if you were--
- Yeah, I think it's gonna vary disease by disease. And I think for the disease that we've worked on, heart attack, which is still the leading cause of death in the world, I think it is ready actually because of two things. There are things you can do about it if you figure out you're at the tail of the distribution that you're at 99 percentile score where you don't wanna be. And the two things that we've shown that you can modify the risk that comes from the polygenic model, that DNA is not destiny here. And the two items that we've looked at, one is lifestyle. And what we showed is that if you actually have adherence to a healthy lifestyle, that you can cut the risk that comes from the polygenic model by at least a half, okay? But you could argue everybody should have a healthy lifestyle, but that is something that you can now tell a patient. You say you're a high polygenic risk, but we know that if you adhere to a healthy lifestyle, you can cut that risk. The second is actually cholesterol-lowering medication. So, statins. So we have looked at three randomized controlled trials that have all been designed to predict, or sorry, reduce the risk of a first heart attack with statin therapy, statin therapy versus placebo, we had genetic data and all those individuals, about 55,000 people, and then we simply look to see the benefit of statin therapy by polygenic risk strata and what we found was that those who had high polygenic risk actually had much more benefit from statin therapy. Both relative benefit and absolute benefit. And that was a little surprising actually. We expected the absolute benefit, the absolute risk reduction to be higher, but we do not expect the relative risk reduction to be higher in the high polygenic risk group. That's what we found and that's been replicated by a couple of other groups now. So here, you have a scenario where you're finding that this is the main factor in at least 20% of people who have early heart attack. And if you identify these individuals, they're currently flying under the radar. They're not being caught by the currently measured biomarkers. If you find that they're high, then there's something you can do about it. So that seems to me the right scenario where it is actionable now, you know? And I think with other disease, it's gonna vary, but we're actually, we've developed evidence for at least two other diseases, both breast cancer and colorectal cancer where we're getting to this level of kind of actionability. And breast cancer, that's particularly interesting. It's a very similar pattern to heart attack. Early heart attack. If you take 100 women with early breast cancer, in only about five will you find a BRCA1 or 2 mutation. Well, it turns out that a large fraction of the remainder is because they're on the very high end of the polygenic risk score. And so there, again, if you figure out that a woman has very high polygenic risk at a risk equivalent to carrying a BRCA1 mutation, then you might deploy the current recommendations for BRCA1 mutation carriers, which are kind of enhanced screening, for example.
- So if I could put this into just a little bit of statistical language, it seems to me what you're both saying is that rather than using Bayesian priors to determine who you run these scores or screen, you use these tests at birth. Ideally, these tests could be used at birth to set the Bayesian prior by which you then make other decisions.
- I mean, so for the, there's a sequencing which will get you, so currently somewhat expensive, but it would get you all of the variation in the genome, okay? The polygenic scores actually currently can be calculated with genotype of chips alone that costs about $20. And with that chip, you can actually calculate scores for every disease where it's possible. So let's say the top 10 common diseases. And you might have a report card where you say, you're 99 percentile for risk for heart attack, 95th for breast cancer and so forth. And then what you do with that information is where we all need to, you know...
- And I would just add to this, Sek's work in polygenomics has been extraordinary, and one of the things that I helped, at Mayo Clinic led a study where we randomized people into receiving a cardiovascular polygenic risk score or not, they and their doctors, then followed them through and there was a reduction in their LDL cholesterol. So using this in a medical context had a very clear proxy outcome effect in that randomized trial. And I don't know if there are others, but that's one of the few sort of randomized trials that's tested whether you can make a biological difference with a polygenic risk score.
- So, now is the jump ball section of this and we'll sort of do these quickly so that we can cover a bunch of them and leave time for people to, so the first is consumer testing. Anybody can go to 23andMe or ancestry.com or whatever. use the Illlumina strips, these are pretty, they have a lot of the stuff on--
- Everything you need actually for the score--
- For the polygenic.
- Not for the rare mutations.
- Right. But you can do this and get the data. The FDA have cracked down on 23andMe and they only let them show a few health risks, but you can download the data and put it into Promethease and/or a bunch of other things and you can get everything. I'll tell you personally, Robert and I shared a phone call where he could also tell you, it's a HIPAA thing that I couldn't handle it. It's stressful for a physician and as someone who has a background in molecular biology in the ancient past. Can the average consumer handle this? This gets to return of results. What do you tell people? What do you not tell people? And are we in a place where, where we need to have genetic counselors that people can just go to?
- I would say this is where the science is ready at least for heart attack I think. But I think that deployment into practice is gonna be challenging for some of the reasons you mentioned. A big issue is the standardization of these scores and the ability to report them in an informative way for patients and that is gonna require a fair amount of work. For at least the quantitative score part of it, I think patients and physicians will be able to understand because it's just like your LDL. If your physician says your LDL is 170, you're like, okay, well is that high? Is that low? Is that okay? And this number is gonna be very similar. You're gonna get a number and say you're in the 99th percentile. Only 1% of people are above you, 98 are below you. So I think it can be communicated, but we're not there yet.
- But things transfer out of being pathogenic and not being pathogenic over time.
- Well, that's a separate story.
- And I looked at mine, I just wanna, I have some like, I have like a 20 X increase of some eye disease, some corneal eye disease, which it would be manifest at this point. There is just a lot of I think anxiety and stress when you read through, even the ones that are pathogenic and that have a reasonable amount of evidence behind them, it comes down to penetrance and all that--
- And expressivity, yeah. Well, there are a couple of threads here. One is the hubris of a company to actually go outside the universe of medicine and offer health-related information to people and I think that raised hackles from the very beginning when these companies launched in 2007, but I think that they have done an extraordinary service by democratizing and breaking open the closed system which was holding back genomics. So I'm not a critic of direct-to-consumer or consumer-facing genetics, In fact, in full transparency, I've started, co-founded a company, Genome Medical, which sort of hopes to bridge that gap, provide genomics expertise through trained providers to people more easily because there's a six-month waiting list in Boston to get to see a geneticist right now. The other thread though is that genetics has a funny history. It started with Huntington's disease this fully deterministic, fully penetrant devastatingly traumatic neurologic disease and kinda terrorized everybody with that. And in fact, most of genetics is like most of medicine. It's probabilistic, you can talk about it with a patient, you can decompress their anxiety and you can put it in the context for them and try to help them manage it and manage those risks. So I think, actually, direct-to-consumer genetic testing is a phase. I think it's a phase that exploiting the gap between genetic discovery and true genetic integration into the practice of medicine.
- And one of the other risks of consumer testing will be manifest in a couple of weeks on Thanksgiving as people begin to discuss their results and the issues of who's parent was whom and who--
- That's true.
- And there's, I've been touched by some of this as well, from - It's messy.
- It is.
- So, the other thing, so many threads here. It's just a quickie. If the FDA is loos-- they allowed Promethease to do this and these other groups, but they've been real fussy about pharmacogenomics. They've like crackdown, it's come out of Promethease. It's very hard to get that information except for one or more, or its competitor. I've forgotten, YouScript or some, so what gives with that? Is that because it touches on drugs and the FDA is concerned or they worry that the people are gonna make the wrong decisions without information.
- Well, the FDA has some specific labeled drugs that they believe there should be pharmacogenomic testing for. And they are generally okay with the laboratories talking about those, but that's a small set of the drugs that we think pharmacogenomics can really help with. So it's kind of a bit of a turf battle I think in some sense and I think there's been a lot of pushback trying to get the FDA to loosen that up again. But they are saying that the interpretation of this is a practice of medicine. If a physician wants to do it or a trained pharmacist, it's okay, but they are objecting to the labs creating a message around the pharmacogenetics that could be interpreted as lower this dose, raise this dose because that's a practice of medicine and the lab shouldn't be participating in the practice of medicine.
- But I think what it reflects is that the genetics of disease onset is pretty robust, but the genetics of drug response is pretty dismal. Most of the literature is false positives and I think that's what the FDA is reacting to, is that it's not that robust. Or actually, only a handful of examples, or where you would change the dose of a medicine based on a gene variant. Most of the other claims are unsubstantiated.
- And the poster child for that was, of course, the Coumadin dosing, which in several studies showed really less than exemplary results in terms of using it that way.
- So, Sek, you've famously left your position at Harvard or whatever. I don't know what the--
- Rashly.
- Yeah. No, but, and to become the CEO and founder of a company that's using CRISPR for, I'm not sure and I don't wanna put you on the spot to actually be specific about that, you may be in stealth mode about that, but I'm actually gonna use that as a segue and feel free to talk about that in a minute. But when he said it's a segue to bring up something else, one of my retirement projects was to discu-- and I'm assuming it's somatic cell CRISPR therapy that you're talking about.
- That's right. Absolutely, adults.
- But one of my pet projects for retirement has been to write about for, in a popular context, eugenics. And eugenics, for those, and I find shockingly that a lot of really educated people are unaware of the horrible past that we had in the United States from the end of the 19th century through the middle of the 20th century of sterilization and of choosing Norwegian stock as the ideal stock and eliminating and doing various things to not mechanically change that, but to change that in a population basis which led to unfortunately some of, directly actually in his writings to some of Hitler's ideas that were put into practice. I'm reading a book by I think Edwin Land on this called the "War Against the Weak," where he talks about the fact that after the war, a lot of these people became the founders of what, they changed the name from eugenics to human genetics, They became the founders of that. Honestly, as you learn more about preconception stuff and prenatal things while weeding out disease is certainly something we want to do, not having a child with a terrible disease, we start to cross over with Down syndrome where there are a lot of people who object, or deafness, where there are a lot of people say this is just a normal variant and we shouldn't be selecting on that basis to the possibility of selecting gender, the possibility of selecting some of those 'desirable' characteristics. So, can you talk a little bit about that?
- Wow. No, I would say that, yes, there's a sordid histories, but there are new tools available, specifically CRISPR, that have potential for really dramatically affecting a human being's health, adult. And I think that's the balance that as a society, we have to strike. Right now, it's very clear that editing as a therapeutic tool in adults to relieve suffering is ready. And in fact, yesterday, there were two reports from Vertex and CRISPR Therapeutics, the first treatments for a woman in Tennessee with sickle cell disease. Her hematopoietic stem cells were taken out of the blood, edited. And then to increase the amount of fetal hemoglobin, put back in and the fetal hemoglobin came up to almost 50% and she now has normal hematocrit and no sickle crises for the last nine months. A similar story in Germany, beta thalassemia. I think like 60 transfusions in the previous few years, treatment with ex vivo gene editing, no transfusions for nine months. So, really remarkable. And so, I think that's the power of the new technology and now being applied to humans, but I think you're drawing out I think potential challenges in terms of misuse and ethical challenges.
- And just to add to that, I mean, I have a friend who is a former rock star turned investor, whose wife has Lynch syndrome. So she gets a series of colonoscopies every year and lives in fear of other cancers. They just became pregnant. And, of course, they used in IVF-PGD to select embryos without mutation. That was not covered by their insurance. They fought and fought and fought, but couldn't get it covered. But that child is gonna grow up without having to worry about that and that's being replicated in BRCA, it's being replicated in early onset Alzheimer's disease in a host of monogenic conditions. These are choices, not requirements. And these are, certainly, there are folks from the deaf community and the Down syndrome community who want to make sure that we can preserve the choice for people who do not wish to take extraordinary measures and I think we must respect that. But I think that the changes where it moves from some sort of choice to some sort of social or governmental expectation or requirement, that's the bridge too far.
- So last quick question and we'll open up to a few questions from the audience. Can we have another hour? Is that all right? No? So yesterday, I read an article about a woman in the UK who is suing three NHS sites because they were aware that her father had a diagnosis of Huntington's disease. She was pregnant and she wasn't told. And she had a baby, so I guess it's wrongful birth or whatever she would have had an abortion and possibly killed herself was her father's concern for not having them tell her. So very complex. There was an article in science in May, which I'd suggest people look up about the wild, wild west of responsibility for genomics. As things change, you now have this sequence which is not 100% of a genome, but it's at least 85% of a genome and you're holding onto it, whose responsibility is it? Is it the lab's responsibility, the doctor's responsibility to inform patients when new things are discovered? Maybe not in terms of a risk score as much as a disease, a serious disease--
- So, well, I'll just start real quickly, but one is that, first of all, I think from the monogenic to the polygenic is a somewhat artificial distinction. This, if you take one thing home that changes your day, your sort of concept of genetics, I would suggest that you don't see it anymore in terms of monogenic versus polygenic. See it in terms of markers, either one or many that puts you on a spectrum of risk from something that's fully penetrant to something that might be more penetrant over time to something that's rarely penetrant to something that's probabilistic, but still very robust. That's number one. And number two, I think that our western tradition of considering the individual and only the individual as the focus of our medical care is not going to be sufficient for the world of genomic medicine because the lowest hanging fruit that we are missing is when we do find a mutation in someone that we can do something about half their brothers, their sisters, their mothers, their fathers, and their children are carrying that same mutation. Cascade testing is the frontier that we are failing to explore. We could, right now, logistically find and alert those people and save their lives, and we're not doing that.
- I wanna just add one other comment actually kinda building on the last discussion about causal inference and whether randomized control trials versus observational epidemiology and the role of each in causal inference. It turns out genetics actually has a very useful application for causal inference and very powerful one and we've leveraged that to really understand whether a given relationship in the population, for example, HDL cholesterol and risk of heart attack, the correlation you see in the population, is that a causal relationship or is that just a mere correlation? You can get at that actually with randomized control trials, but also, prior to randomized control trials, you can get at that using human genetics. And the idea is very straightforward. If you have a mutation that raises HDL lifelong, for example, then those mutation carriers should be protected from heart attack if HDL was protective, causally protective for heart attack, which is what the observational epidemiology would suggest. Well, it turns out when we looked at that a few years ago, people who carried mutations that raised HDL cholesterol had the same risk of heart attack as those who didn't carry those mutations. And suggesting that HDL was just a marker of risk rather than a causal factor and suggesting that medicines designed to exclusively raise HDL cholesterol would not reduce risk of heart attack. And that's actually what happened when randomized controlled trials were done to really, medicines designed to raise HDL cholesterol, they raised HDL, but did not work to reduce risk of heart attack in several RCTs. So I wanna for all of you to kind of consider thinking about genetics kind of in that role as well as a tool for causal inference in terms of--
- So causality was on my list of things to ask about, so thank you for covering it. I was skipping over it. These guys who would certainly give up their lunch to answer questions from you all afterwards , but I do wanna-- Robert actually asked a question, so we're gonna do some active statistics here. I'd like people to raise their hands and keep them up. I'm gonna try and do this quickly. And only if you want to, only if you're comfortable, how many people have had whole genome sequencing in the room? Whole genome se-- so keep 'em up. Keep 'em up. How many people have had whole exome sequencing on top of that? Few more people. Keep your hands up. And how many people have had some sort of consumer testing where you've had a chip-based 23Me or...? Okay, not all that many so far. It's $79 to get to do, 23andMe, I have nothing to do with them, and then you can download your data and get all the clinical stuff. How many people would like to do that? I'm curious. Yeah? Yeah, a lot of people. How many people actively do not want to do that?
- Great. Thank you.