Boulder Leader

Imagine a jigsaw puzzle with thousands of tiny pieces spread across a table. The puzzle’s completion promises insights into better personalized patient care, but the pieces are from different puzzle-makers – their sides not fully matching up at first glance.

That’s the challenge Laura Wiley, PhD, MS, faces in her personalized medicine research.

“Our goal is to actually generate new knowledge about how best to care for patients,” said Wiley, an assistant professor in the Department of Biomedical Informatics at the University of Colorado School of Medicine. “But healthcare in the United States is really fragmented. There's a ton of patients who don't have health insurance, that don't have equitable access to care, and so they're underrepresented, or we only see small snapshots of them at a point in time,” Wiley said.

“And so part of my job, and part of the methods development work that I do, is trying to figure out: How do we actually use all this fragmented care?” said Wiley, who also has a secondary appointment in the Department of Biostatistics & Informatics in the Colorado School of Public Health and is a principal investigator in the Colorado Center for Personalized Medicine.

In the following Q&A, Wiley explains what personalized medicine is, how she approaches using electronic health records to help supplement ongoing patient care and health sciences research, and what this research means for patients.

Q&A Header

How would you define personalized/precision medicine?

The quick answer is: It depends on who you are talking to. Scientists and clinicians often have different answers. Defining precision or personalized medicine can be as holistic or as narrow as you want. Some scientists would say personalized medicine is only when you're creating an individual therapeutic for a single patient, like gene-based therapy. I tend to take the broader perspective on precision medicine: giving the best evidence available for any individual patient while taking into account their personal desires, goals for their care and life circumstances.

Can you talk about what makes up an electronic health record?

An electronic health record (EHR) has a lot of different components. Historically, you could think about an EHR as just the electronic version of your old paper chart, but in fact, it's so much more than that. They include not only details about any individual patient's health status, but it also captures all this information about the actual operations of hospitals. In a lot of cases, it has things like billing or personnel data of all your care providers: physicians, nurses, and all the ancillary health professionals.

How does your research interface with EHR data and personalized medicine?

Our goal is to generate new knowledge about how to best care for patients. Fundamentally: How do we generate new biomedical knowledge? I work with geneticists and folks with the Colorado Center for Personalized Medicine Biobank. And our focus is: How do we take all data from medical records and use it to learn about the underlying biology of health?

As we think about the patient's view and what information is there about them and their care, there's a lot of different types of data and detail. What kinds of clinical conditions does a particular patient have? In an EHR, there’s what treatments a patient is undergoing, their medications, what procedures or surgeries they might have had.

What does that translate to in practice? One general example is pharmacogenomic markers. Those are specific genetic markers that tell you whether a drug will work for a particular patient, if a different dosage is needed, or if it might cause side effects.

What research are you working on currently?

One collaboration is with the Department of Neurosurgery, Dr. Christopher Roark, that's looking for patients with unruptured brain aneurysms. For a lot of patients, unless you get brain imaging for another reason, there's no real way to know that you have an unruptured brain aneurysm, as there are typically no symptoms.

If a brain aneurysm ruptures, it's really serious. Subarachnoid hemorrhage has really dramatic effects, but the vast majority of aneurysms won't rupture. More people die with an unruptured aneurysm than die from them. The problem of course is: Which one is which?

Unfortunately, the best evidence that we have available right now tells you just the five-year rupture risk. But we care about the next 30 years and people living full, healthy lives. Researchers have tried to run a clinical trial in this area, but to test and identify exactly which aneurysms we should treat, some patients would need to forgo treatment, even for aneurysms that look really high risk based on our current knowledge.  The reality is there's just not enough patients or surgeons who are willing to participate under those circumstances,  which is understandable.

Our work is set up to say, ‘Let's look at all the patients that we've ever seen with a ruptured aneurysm and unruptured aneurysm. Let's see if we can use high-quality data methods to detect any patterns.’ Now, this is obviously not as good as a randomized control trial. But if you can't ever get a trial, this approach can give you some personalized evidence. You can look at the data and say, ‘Let's understand what this patient looks like, let's understand what their aneurysm looks like, and let's use that to see other patients like them and see if more of them are in the ruptured category or more of them in the unruptured category.’ It's not perfect, but again, it's better than what we have currently.

So this kind of data analysis helps supplement ongoing research and fill in the blanks and gaps in certain areas?

Yes. It can make a difference in real world care, but it can be something you won't see under the carefully controlled conditions of a clinical trial, because clinical trials tend to underrepresent different patient groups. Trials have really strict monitoring and controls that are often not 100% reflective of the real world. Our methods are trying to improve that and say, for example, ‘Hey, these algorithms might underrepresent women. They might not detect women or folks who have existing health disparities.’ If our algorithms are biased, and the evidence we produce is biased, then that's a problem – so how do we fix it?

We are trying to think through how we can use EHRs as a way to give a provider second-best evidence – things that are useful and meaningful for patients’ lives in situations where you’re not going to be able to run a clinical trial. That’s been a key ethos of our work.

How will research into EHRs be implemented into your doctor’s office experience? Will it be additional background information for the provider to consider or will it be more patient facing?

I think the answer is both. The work that I do generates long-term biomedical knowledge, which you may never see directly as an individual patient, but long-term, it will definitely have some effect on what we know about health and disease and how best to treat patients. Some of it is used to actually generate that knowledge right now, but that's at a population level. It's like running a clinical trial where we're saying, ‘Based on the patients that we saw in this trial, here's what we believe the best treatment is.’

And our goal ultimately it to lead to a situation where a provider tells their patient: ‘We could put you on Drug A, but actually, you signed up for the Colorado Center for Personalized Medicine Biobank, and it turns out that you actually have a genetic marker that puts you at a higher risk for this adverse reaction with Drug A, so we will prescribe you Drug B.’

Original source can be found here.