In this episode we’ll be looking at personalized medicine – an approach being increasingly used across clinical fields to provide tailored treatment for patients, on the basis of their individual genetic and clinical characteristics.
In some areas of medicine, this patient-tailored prediction and treatment of disease is common, but elsewhere advances in scientific understanding are harder to translate to a personalized medicine approach. In this podcast we’ve spoken with specialists across different areas to get their thoughts on where their field is at right now, and their predictions for the future.
But first we’ll speak to Eric Topol from the Scripps Translational Science Institute, California who thinks that one of the most exciting opportunities for personalized medicine is the emergence and spread of smartphones, e-readers and tablets. By using these devices to encourage patients to collect data over time, or even turning them into medical devices themselves, Topol thinks that this so called field of mobile health has the potential to revolutionize medicine.
BMC Medicine Senior Editor Claire Barnard discussed with Topol the various ways that mobile healthcare apps can contribute to individualized medicine and disease prevention.
How can mobile phones be used to personalize healthcare?
ET: Well the smartphone will become the hub of the future medicine, because it has this pluripotent impact. For one, it can be the conduit of sensor information, whether that’s blood pressure, glucose, heart rhythm, brain waves, the list is almost endless. But in addition to biosensors, there’s also the ability to change a mobile phone into a scanner – to an otoscope, ophthalmoscope, a microscope, any kind of scope – and in fact there’s even little ultrasound devices now that can function as a mobile phone. So you’ve got the sensor side. Then the next is that the smartphone as a lab on a chip, which is basically capable of doing almost any common laboratory assay. That includes kidney function, liver function, thyroid function, blood thinning International Normalised Ratio, potassium, and the list just goes on and on! Those are just a few ways that this can lead to collecting data, capturing data for a particular individual, to shape that person’s care.
Are medical smartphone apps subject to safety regulations by organizations such as the FDA?
ET: The FDA recently came out on that and they basically felt that most smartphone apps do not require their oversight. It’s only the ones that have critical measurements like blood glucose, blood pressure, heart rhythm – these are important devices. They’re not just apps, they’re also add-ons to the phone; they involve hardware and there has to be some demonstration that accuracy has been fulfilled. That I think is appropriate, it’s actually vital that we have an independent agency, a regulatory body that can assure that the things being measured are being done so in a highly rigorous, accurate way.
For example, if you had a glucose that was an error and you wind up taking a lot of insulin, you could be in a coma or you could have seizures or you could die. So whenever there’s something where the measurement is critical, that’s where the FDA is going to certainly weigh in as to whether it’s ready for approval or not.
What are the potential risks associated with this kind of app in healthcare?
ET: Well, there’s a few different risks that are important to mention. We’ve touched on one, which is lack of accuracy. For example, it could be accurate and then it starts to lose its calibration. That’s an issue that has to always be grappled with. Second would be security of the data – that there’s escapes or re-identification of data that’s supposed to be kept de-identified. That has to be grappled with because anything that’s digitised can be hacked or breached into, so that’s a concern. The third level of risk is inducing severe anxiety. That is, when a person has their own data being continuously displayed on their little device, for some people that’s very difficult, that’s not easy for them. We have to figure out ways of who would not be suitable for this, how we can reduce that type of anxiety. But these people tend to be highly anxious and wired anyway, so this is just one more means that they can add anxiety at a high level.
That’s Eric Topol highlighting some of the ways that mobile phones have the potential to personalize healthcare across different areas of medicine. But where are we right now? Claire Barnard spoke to Dan Hayes of the University of Michigan School of Public Health to find out more about progress in cancer diagnosis and treatment.
Here’s Hayes explaining why he thinks that the field of oncology has huge potential for personalized care.
DH: So oncology is an area that I believe probably has the greatest potential for personalisation of medicine, but probably has had the least application until recently. Over the last 40 or 50 years, especially in medical oncology, we’ve sort of taken a “one size fits all” approach, and that is that our only efforts at personalisation have basically been trying to tailor specific chemotherapies based on the tissue of origin. For example, we treat breast cancer with slightly different drugs than we treat lung cancer. But frankly, not very different. The only other effort is to give the chemotherapy based on the height and weight of the individual patient – so we try to get the dose kind of right.
In both cases, we almost certainly over treat many, many patients, we almost certainly give patients treatments that aren’t going to work, and we almost certainly under- or over-dose patients. So there’s a lot of room to move in medical oncology and I think we’ve got plenty of room to do it.
The reason I believe medical oncology has the greatest potential is, because unlike many other diseases, we have fairly easy access to the somatic tissue that we’re trying to address. So for example, the cardiologist can’t do serial heart biopsies very well, or any heart biopsies very easily. The general internists who are treating hypertension – there’s going to be some effort to do that based on inherited genetic single nucleotide polymorphisms – so-called snips – but so far that field hasn’t panned out very well. Whereas in cancer, we have access to the cancer and we can look for various things the cancer is doing that is different form normal tissue and try to focus on that. The parlance for that over the last five years has been so-called ‘targeted therapy’.
In fact actually oncology has done targeted therapy for the last 120 years or so, starting with George Beatson, who was a Scottish physician, surgeon, and also a gentleman. He raised dairy goats and he began to come up with a hypothesis that there might be some connection between ovaries and the breast. He took the ovaries out of three women and reported that two of them had what we would now call a “response’; they had breast cancer. Obviously, what he was doing was removing the growth factor oestrogen from tissue that needed oestrogen to grow because it made oestrogen receptor. It took a hundred years – or not that much, it took 60 or 70 years – to learn the molecular biology behind that’s allowed us to focus, for example, endocrine therapy on people who are likely to benefit from it; their cancer is oestrogen receptor-positive. And we don’t give it to people who are not likely to benefit; oestrogen receptor-negative.
More recently, in breast cancer, we’ve been able to do this with the HER2 molecule and a number of therapies, starting with trastuzumab and now many others directed against HER2. Those successes, frankly, I think led into a number of other diseases now. Especially in leukaemia, I’m sure you’re fully aware of the chronic myelogenous leukaemia story with the so-called “Philadelphia chromosome”. Brian Drucker and Janet Rowley and many others have shown us exactly what that is – the translocation is well known, the microbiology behind it is fully understood, and now we have many drugs – starting with gefitinib – against it, which have been quite successful. So I think the door is just starting to open and in fact there’s going to be a floodgate open in the next few years.
That’s Dan Hayes on the history of personalised medicine in oncology. Here he is again to tell us how the omics revolution has contributed towards personalised oncology in the last few years.
DH: In fact, there have been very few assays that have come out of the genomics revolution or evolution in the last 12 years. Based on that, in my opinion is that the one that is the most highly proven and adopted is the so-called ’21 gene recurrence score’ for breast cancer. We have really good evidence now that if an oestrogen receptor-positive node-negative patient has a so-called “low recurrence score”, her odds of recurring over the next 10 years – assuming she’ll get endocrine therapy – are so low that even if chemotherapy works, it won’t help enough people to outweigh the potentially life-threatening side effects of chemotherapy. That assay is pretty widely used in the United States and we’ve actually seen a reduction in the use of [...] chemotherapy for that group of patients, by about 20% over the last five years. Which in my opinion is exactly what we’re trying to do…
Now, the next step is so-called ‘next generation sequencing’, and that’s a whole different ball of wax…
This is really in its infancy; it’s a year or two old. We’ve gone from the original sequence of the entire human genome took 10-12 years and, I think, 10 billion dollars, to the point now where it can be done in a couple of weeks for about a thousand dollars. So there are many studies being generated on both sides of the Atlantic in which patients’ cancers are being biopsied and then fully sequenced in one way or another and then compared to their germ line DNA to see whether or not that patient might be more specifically treated.
Now there huge caveats to this approach. One caveat is just how accurate are these tests. It goes back to my original discussion. Just like any test, there can be mistakes made in genome sequencing, there can be mutations found that may or may not actually have significance. People are trying to worry about that. The second is still whether or not there are context-specific responses to our drugs. In other words, a mutation in, let’s say, HER2 may be important for drug X in breast cancer, but that same mutation in lung cancer may not have the same biology because there are many other things that go on relative to lunch tissue versus breast tissue where the cancer began.
This is a huge issue. Many people are just assuming if they find a mutation in cancer – and we know the drug worked against that mutation or at that amplification or whatever in cancer type A – that, oh well, it should work in cancer type B and cancer type C. That may not be true and one of the real issues we’re struggling with is how do we design the clinical trial to test this.
So cancer treatment is already being tailored towards the individual patient based on genomics and receptor expression, but in other areas of medicine, we are just beginning to see steps towards patient-tailored risk prediction and therapy. In the neurology, for example, there have been some studies published on the genetics of stroke, but individualized risk prediction is not yet employed in the clinic.
Hugh Markus works at the Department of Clinical Neurosciences at the University of Cambridge – here’s Claire Barnard asking him about recent research into stroke genetics.
How have recent results from genome-wide association studies, discussed in your review article, contributed to understanding of stroke?
HM: It’s still very early days in the genetics of stroke, but the studies from GWAS that we’ve got so far are telling us some quite interesting things about stroke. First of all, they’re telling us that different types of stroke, or different stroke mechanisms, have quite different genetic background or architecture. For example, the genes involved in stroke due to narrowing of the arteries to the brain – called large artery stroke – seem to be quite different form the genes involves in stroke associated with cardioembolism – or blood clots coming from the heart. So it’s telling us something about the pathophysiology of stroke. This may have important implications in that we need to tailor our treatment rather differently towards different types of stroke.
They’re also beginning to tell us about new pathways which may be involved in causing stroke. Perhaps the most exciting one so far is a pathway called [HDAC9], which appears to be associated with large artery – or atherosclerotic – stroke. This may give us a new way of understanding what causes this type of stroke, which could potentially be targeted with different treatment approaches. But this is a little way off at the moment.
So the field of stroke is not as advanced as other areas of medicine – for example oncology – in terms of personalised risk prediction and therapy. Could you explain why this might be?
HM: There are a number of reasons for that. First of all, the genes involved in stroke risk all contribute just a small amount of increased risk. Unlike some of the cancer genes where if you have a genetic mutation you’re really at quite high risk of, say, developing breast cancer. So there are lots and lots of different genes that are all contributing, presumably, a small amount of risk and therefore if you just look at one of these variants, it’s not going to give you the same sort of information as you will do from perhaps looking at a cancer gene. That’s one of the reasons.
Another reason is that we’ve really been quite late in doing genetic studies in stroke and we’re only just beginning to scratch the surface. Many other diseases have been looking at genetics, with the GWAS approach for rather longer and have got much larger sample sizes.
I think those are two of the main reasons. It may well be that we can predict risk in the future but at the moment, because each of the genes which we’ve identified only contributes a very small amount of risk, we’re still unable to explain much of the genetic risk factors. Therefore, looking at just a few genes when there are many, many we don’t understand or haven’t yet discovered is only going to contribute a limited amount to overall risk.
That’s Hugh Markus talking about the potential to assess an individual’s risk of stroke, and their treatment should they suffer one, become more personalized in the future.
Diabetes is another area of medicine that is moving towards patient-tailored treatment, and the increasing number of drugs available provides a challenge in selecting the best one for each patient.
Here’s Claire Barnard talking with David Leslie of St Bartholomew’s Hospital, London.
How is diabetes management becoming more tailored towards the individual patient?
DL: What’s happened is that the number of drugs available to us has increased substantially and the approaches to therapy to therapy have increased substantially. We now have more options in terms of drug therapy. Also, the drug therapy before was rather non-specific; everyone got the same treatment. But now we’re beginning to identify sub-groups of diabetes in whom some individuals are more responsive to some of these agents or these agents are more suitable for some than for others…
Last night, I saw a bus driver. This bus driver does not want to go on to insulin. Why? Because if he went on to insulin, he would lose his job and he’s relatively young. So in the event, I gave him a GLP1 agonist, which is an injection, and with this agonist, his blood glucose came down. But actually, if that agonist had not been available I would have had to give him insulin. So the introduction of these agents, such as GLP1 agonist, which do not cause hypoglycaemia have in this case meant that this man can continue with his job, and that is a perfect example of personalised medicine.
We still, unfortunately, are constrained by our approach, which is the prevention of complications. We could actually try and prevent the disease, but we don’t. And we could try and treat complications and reverse them, but we don’t. So what we’re doing is we’re limiting ourselves, still, to the prevention of complications. But even in that, we have expanded our horizons so that whilst before we had a glucose-centric approach, now we focus not only on glucose but on smoking, on hypertension, on high blood cholesterol and on exercise.
Do you think that personalizing diagnosis and treatment is more effective than a guideline-based approach?
DL: Well a guideline approach is very convenient for someone who is not an expert because it gives, quite literally, a guideline as to which direction you might go and what drugs you might use.
Trouble is that a lot of the guidelines are taken as tablets or commandments. This is a very real issue because a lot of people who one might recommend one therapy as a specialist to a primary care physician, you’re rebuffed because the primary care physician says, “But that’s not what the guideline says.” But the guideline is a guideline, it’s not a commandment. This is a big, big problem, I think, in trying to understand why we should allow specialists to treat the patient as an individual but understand that with the overwhelming number of cases of diabetes, we have to have people who are not so expert dealing with cases in which guidelines are probably quite valuable to them because it gives them a broad-brush approach to the condition, without being able to modify or refine their management policy.
Another issue is that the guidelines are only updated every few years or something and, as an expert, you’ll be reading new studies and new trials that come out all the time in diabetes and these might not necessarily be incorporated into the guidelines yet.
DL: You’re absolutely right. There’s a substantial lag between where we are as specialists and where the guidelines are. That’s inevitable, that lag. So inevitably you’re going to get positions taken by specialists which the people who produce guidelines are not going to have had time to illustrate.
That’s David Leslie with some examples of how diabetes treatment is tailored to the individual patient, and the advantages of a patient-centered approach.
So between oncology’s current use of genomics to guide patient-specific treatment, to the early stages of identifying the genes associated with different stroke pathologies, it’s clear that personalised medicine holds much promise but is by no means mature across much of medicine.
We asked each of the experts to outline their vision for the future of personalized medicine in their respective fields.
First, David Leslie gives his perspectives on how individualized diabetes management might develop in the coming years.
DL: I think the future of this field is very substantial – and when I say this field, I mean the concept of personalised medicine in general as compared with our current guidelines. I think this extends far beyond diabetes into a wide range of medical conditions.
Currently, most of our drugs address insulin secretion; very few address insulin sensitivity. So there is obviously a broad area where we might be able to target people who have a particularly severe problem with insulin sensitivity, where their requirement for drugs is distinct from those whose predominant problem is its secretion. And we already have this. We already know that patients with adult onset autoimmune diabetes who do not require insulin initially seem to do badly with sulfonylureas. And yet sulfonylureas would be considered to be probably the second line of treatment. Certainly all the guideline say as much. But they just don’t seem to do well with sulfonylureas. By way of contrast, there’s another type of diabetes associated with a genetic mutation of maturity onset diabetes of the young, in which sulfonylurea is particularly prone to causing hypoglycaemia and you therefore need to give a slightly different agent. However, there’s a really rare condition where the potassium channel of the beta cell has a mutation which can be circumvented by sulfonylureas. Those people, who are given insulin initially, actually can come off insulin if they’re given high doses of sulfonylureas. So here are examples of how our understanding of the pathophysiology of the disease and the genetics of the disease can take individuals with diabetes and actually target appropriate treatment.
David Leslie there. Earlier Leslie earlier explained why the greater variety of drugs available to treat diabetes has enabled personalized treatment for these patients. This is also true in the field of oncology, where an increasing number of drugs are available to target different genetic mutations. Dan Hayes gives us his perspectives on the future of personalized medicine in cancer treatment.
DH: Well I have, I think, three visions. One is: we just published a paper – we being myself and several of my colleagues – in which we outline something we call the “vicious cycle of tumour biomarker generation”. The vicious cycle involves both an inconsistent regulatory environment that has confused people as to how they should develop a new test […] insufficient reimbursement for a biomarker test. So for example, the third party payers in this country and in other countries – whether it’s governmental insurance or whatever – are used to providing drugs that have been shown with high [...] be effective. So they pay for those, even though it’s expensive. But they’re not used to paying for a test. In the old days, that was probably appropriate but if we’re using these biomarker tests to direct therapy, then they’re every bit as important as the therapy itself.
Second, I think, is the next generation sequencing we just discussed a minute ago. This is the future of oncology and I don’t know where we’re going but it’s pretty exciting, as I said, and we’re really looking forward to that.
DH: So the third area of the future, I think, is not so much a scientific revolution as it is a sociologic revolution, and that is the issue of being able to look at so-called “big data”. With the advent of the use of electronic medical records in many medical practices around the United States – and in fact it will be mandated by the Affordable Care Act – we’re going to have the ability to potentially review millions of patients’ outcomes in the future and apply the kinds of lessons we’ve learned from so-called “outcomes research” or “health services research” or “comparative effectiveness research”. Those, in my opinion are all just evolution of the same thing. That is – looking back at large datasets doing retrospective analyses of how patients do when they’re treated in various ways or approached in various ways.
Although that’s been around for a long, long time, most of the datasets that people have used are either highly specialised – like in the United States, looking just at the Medicare datasets and looking at billing codes, or the so-called “SEER” datasets – and those are relatively limited and have built-in biases in terms of what patients are actually put into the dataset. But in the future, I think we’ll be able to monitor millions of patients who are just seen in regular practice, with really high quality outcomes data.
Dan Hayes has described his thoughts on the areas that will be key in progressing personalised medicine in oncology– namely tumor biomarkers, next generation sequencing and big data. What about stroke, where personalised approaches are in their infancy? Here’s Hugh Markus explaining how he thinks personalized medicine for stroke might develop in the future…
HM: In a number of ways, possibly. One way which may be a little bit nearer is in looking at pharmacogenomics. This is where people respond differently to treatments according to their genetic makeup. This is already the case with drugs such as clopidogrel – which is an anti-platelet agent, which is frequently used to reduce recurrent stroke, or with warfarin – which is an anticoagulant and is frequently used to prevent recurrent stroke in patients, for example, with atrial fibrillation. For both of these treatments, there are genetic variants which determine how your body breaks down the drug and therefore what sort of dose of the drug you’re going to need or, in the case of clopidogrel, whether you’re likely to respond to it well or not.
So there’s already exciting data to suggest that if you look at these genetic variants you can tailor dose in individual people. Whether this actually has a major impact on outcome and allows us to reduce recurrent stroke risk is currently being determined in clinical trials. But this sort of pharmacogenomic approach is almost with us.
The other question is whether we can predict risk in individual people, so predict overall stroke risk in individual people. For the reasons I explained earlier, because we only know a few of the genes that are contributing to overall genetic risk, this is a little way off. When we’ve done many more genetic studies in larger sample sizes and we can explain more of the overall genetic risk, it may well be that we can start predicting risk in individual people. The question then is whether we can predict usefully more risks than we can with conventional cardiovascular risk factors. For example, we know that if you have high blood pressure or if you smoke or you’ve got high cholesterol, you’ve got an increased risk of getting stroke. The question is can genetics predict more risk than that really, and how useful is it? We’ll only be able to do that when we’ve discovered more of the overall genetic components for stroke.
So a key question is exactly how important an individual’s genetics can be in predicting their risk of disease, but how important will an individual’s own engagement in their health – lifestyle decisions and whether they engage in mobile health – prove to be?
Here’s Eric Topol again on the promise of mobile technology for monitoring patient data and encourage healthier living.
ET: I think this is the most exciting time in the history of medicine and it’s because of this. I think that, just as medicine became economically… moved into crisis mode, fortunately at the same time this whole area is blossoming, exploding with innovation. It has a lot of promise for lowering costs for the first time, which has never really been the case with new technology in medicine. But because it’s consumer-based – it’s a consumer cell phone, it’s their data – it’s really particularly enthralling for me to see this shift away from the doctor-dominated world of medicine to a much better parity and symmetry of information: the flow of information directly to the patient and then the guidance from the physician. I’m really, I think, impressed with where this already has gone in such a short time and where it can go.
Does Topol think that this will ultimately reduce the number of times patients have to visit the doctor?
ET: Well that’s the goal. The fact that so much of this information can be acquired anywhere anytime, why would you have to be, for example, going to the doctor to get a blood pressure check or a heart rhythm check or any kind of check? So much of that could be done by oneself, at the convenience, even much more data than ever before. So the hope is a marked reduction in the resource consumption, emergency room use, and ultimately you don’t need to have too many people in the hospital because they can have their vital signs and critical things measured anywhere. So the only people in the hospital are those who are truly acutely ill, in the intensive care unit, or they’re in the hospital for a special procedure or operation. But mostly, most people will be at their home, at the convenience of their home, which is much safer from the standpoint of infection and medication errors, and much cheaper. No less, being able to sleep in your own bed with your own pillow and everything. So the end of the hospital as we know it today, the end of office visits as we know them today – none of these entirely ending, but having a radical re-booting.
You can find more on personalised medicine online in BMC Medicine, their recent forum article has more from all of the contributors to this podcast, and is part of a full article series on personalised medicine available at biomedcentral.com/bmcmed.
Our thanks to Daniel F Hayes, Hugh S Markus, David Leslie and Eric J Topol.