“We Have to Find Our Way”: Eric Topol of Scripps on Connected Health
(Page 2 of 3)
who I wasn’t sure was having a rhythm problem from the symptom, and I got the subject line, “I’m in atrial fibrillation, now what do I do?” The algorithm tells them their diagnosis. You have to have a circuit of the heart with two fingers. You can put two thumbs on your iPhone, or just carry it in your pocket.
X: I didn’t know there was something that simple to get actual information to the physician.
ET: Or it can just tell patients it’s normal, and they don’t have to bother the doctor. That’s what provides lots of reassurance. I give that example, because it’s the beginning of where this is all going. Automated interpretation.
X: Nike pulled out of FuelBand. People are down on that category of “quantified self” or wearable devices. Is that ominous for the type of applications you’re talking about?
ET: These Fitbits, FuelBands, Jawbones, they do help patients walk more. People have them and look at their phone 100 times a day, and they want to walk more. But the accuracy is a problem, and usually in a matter of weeks, people give up and say, “I can’t get my 10,000 steps.” It’s not the same as having a medical condition and the app is a way to get your arms around it. Almost every patient I have has high blood pressure. I used to ask them to write it down. I’d barely get a couple a week, if any. Now, they just have to press start with a wireless cuff on their wrist, keep that in their pocket. They get as many blood pressure [readings] as they want per day, and the app will tell them what percent is out of preset range that the patients and I would set.
The [genome] sequence is another part of the story. The biggest value of that today is pharmacogenomics. Let’s say you go to your doctor and he prescribes a medicine for you. Because you’ve had your sequence or your pharmacogenomic profile, it already knows that the medicine and you are incompatible because you’ll have a major side effect, or you should only take such and such dose, or it isn’t going to work. Everyone should have that, and it can be done inexpensively.
X: The arrythmia example is a simple layer of information that’s actionable. But with pharmacogenomics, isn’t it deeper? Is there a framework for getting from someone’s whole genome sequence to actionable data that you can apply with confidence?
ET: Not many people have had their whole genome sequenced. The tally a couple months ago in the world was 100,000 people. What I’m talking about is like out of a [particular] 23andMe genotype, they have 30 drugs they screen for. There are other companies, too, doing pharmacogenomic profiles inexpensively. If you get one of these direct-to-consumer panels, the whole kit and caboodle and it comes with upgrading, it’s cheap. Relatively cheap. $99 is not free, but for the information you’re getting it’s a pretty good value.
We need point-of-care genotyping, whether at a drug store or at your doctor’s office. Genotypes could be done for pennies. Quickly. In minutes. No one’s doing that yet.
X: Will it be common five years from now?
ET: I hope it doesn’t take that long, because a lot of people will get hurt with the wrong dose. It could be done today if there was a real will to do it. We’re talking about screening particular genotypes that are well established for drug interactions.
X: With all these data layers, will some become actionable sooner because they’re easy to mesh together?
ET: Every one of these layers—of this Google map of the person—you can make a case that if a person’s perfectly healthy, they’re not suited for all this high-definition drilling down. The problem has been that we want to jump to, “Oh, everyone will have any one or all of these ‘omics’ studied.” I think it’ll be different than that. I think it’ll be very specific and particularized. We’ll sort it out. For many people, obesity isn’t due to their gut microbiome. But for some people, it may well be. It’s going to be like a lab test when you work up a patient: You say I’ve ruled this out, I’ve seen this, I have these sensors for caloric intake, I have all these different layers of information. And when do I get this next layer? When do I go for a methylome of the whole genome? Or get sensors of the environment?
X: What do you mean, “sensors for caloric intake”?
ET: There are devices, hand-held spectrometers, you point at the food, and they tell you the exact caloric intake. These mass-spec devices, if you could tell exactly not just how many calories, but how much fat, protein, carbohydrate, it’s going to be pretty big. Supposedly the technology is getting pretty darn accurate and with multiple companies competing.
X: How do we start creating the protocols you’re talking about, where you don’t look at all the panels at once, you winnow things down?
ET: We have to find our way. We have powerful tools that are all coming alive in a short period of time. How to use them and how to compartmentalize those of academic interest and those of actionable potential, that’s really a big difference. If it’s not actionable, who’s going to pay for it? And we’re in an economic crisis. Doing this in a promiscuous way … Next Page »