By Paul Nicolaus
November 15, 2021 | Geisinger researchers have been awarded a $5 million grant to develop a diagnostic tool for genetic disorders. With support from the National Institutes of Health’s National Human Genome Research Institute, the group plans to build a High Impact Phenotype Identification System. The aim is to trim the time between symptom onset and the discovery of a genetic basis to improve patient care and outcomes.
Diagnosing rare disease is an ongoing challenge in medicine, said Mark Williams, professor at Geisinger’s Genomic Medicine Institute and principal investigator for the project. And while not all rare disease is genetic, a significant proportion does have a genetic basis. When individuals present with a range of signs and symptoms as well as lab or imaging abnormalities but there is an ongoing struggle to put all that information together, the lack of diagnosis can lead to plenty of frustration—for patients, family members, and clinicians.
This project is essentially looking to shorten the duration of these frustrating diagnostic odysseys by achieving an appropriate diagnosis and implementing effective intervention much sooner. The endeavor “envisions a time when genomic sequence will be readily available,” he told Diagnostics World, which may not be all that far off. “It is not inconceivable that within the next 5 to 10 years, we would be at a point where everyone would have a genetic sequence that would be potentially available for use for a variety of indications.”
Three Big Questions
Williams pointed to the example of a child who presents to the clinic with a new-onset seizure disorder. “We know that there’s a variety of reasons why people can have seizures,” he said. “We can see it as a result of infection, we can see it as a result of trauma, but we also know there’s a very high genetic burden for seizures.”
So he and colleagues took a look at individuals within the Geisinger system who had a seizure diagnosis due to genetic abnormality. They returned to the electronic health record to determine how long it took, from the time the individual first presented with a seizure to the moment a diagnosis was made. In the roughly 20 patients they evaluated, it took an average time of approximately 3.5 years, with a range of about 2 to 12 years, to determine that the patient’s condition had a genetic basis.
While discovering this lengthy duration of time may have been troubling, looking back at the interventions used to attempt to control the seizures was perhaps even more concerning. Roughly half of those individuals were on medication at one time or another that would have been contraindicated had they known the actual cause. In other words, the therapeutic trial approach to drugs led to ineffective and potentially dangerous therapies, at least on a substantial subset of individuals.
This finding raised a series of questions, according to Williams. First, how can electronic health record data be used to identify individuals that may have a genetic condition? Second, how can that information be passed to this sequence (that he and colleagues are proposing will exist in the not-so-distant future)?
If an individual had a seizure, for example, that could lead to an in silico analysis of all the genes that are associated with the development of seizures and consistent with the other clinical features. This information, in turn, could be used to question whether any variants in any of these genes look like they might explain why the seizure occurred. It involves “doing a diagnostic genetic test using pre-existing genetic information and information from the electronic health record to provide guidance as to the indication for that test,” he explained.
Once there is a tentative diagnosis in place, the third and final question is: How can that information be communicated back to the clinician to help guide care? Collectively, these three questions are the driving force behind the three main aims of this research project.
The effort is funded by the grant award for a duration of five years and is essentially a pre-implementation project, Williams said. The goal is to provide answers to questions in each of those three areas that would allow the group, at the end of the five years, to have a system that could potentially be moved into the implementation space.
They are planning to use an electronic health record environment that is separated from a true clinical environment. This “sandbox” allows the research team to work on developmental projects of this nature without any fear of interfering with the day-to-day operations of a true electronic health record. This environment does reflect an actual electronic health record, meaning it uses the same standards and approaches. In addition, clinicians can interact within the environment, which enables the researchers to analyze workflow and user experience.
Targeting High-Impact Conditions
The research team used a set of criteria to come up with 13 “high-impact” conditions that would be well-suited for this project while acknowledging that it is by no means an exhaustive list. “I think we could identify rare genetic conditions in just about any specialty in medicine that you could imagine,” Williams acknowledged.
One of their key criteria centered on finding areas where many of the diagnoses are being missed. “We relied on both our own experience and the literature,” he said. They also looked at whether there is a sufficient genetic burden. “We wanted to focus on those where the genetic burden is much higher because then we would be able to really test out many more potential patients in this type of a system.” The potential for actionability was yet another area of importance.
One of the conditions they’ve looked at is monogenic diabetes. It looks a lot like Type 2 diabetes, which is very common. It represents only a small fraction of Type 2 diabetes, however, and is treated much differently. “So the impact on the patient, even though it’s relatively rare within this group of Type 2 diabetics, is so profound that we thought it would be a good target,” he explained.
There were other practical factors to weigh as well. “We already have a group within Geisinger that’s been working on monogenic diabetes,” he said, “so we knew that we could build on the work that they had already been doing.” That is also true in the case of some kidney conditions considering Geisinger has a nephrology group focused on genetic kidney disease.
In addition to kidney disease and diabetes, the overarching categories of focus include epilepsy and heart disease, among others. “Those are the ones that we’re going to initially focus on,” Williams said, “but we also recognize that genetic eye disease is something that is relatively common, and there are diagnostic challenges associated with that.”
The group has flexibility that may lead to alterations of the conditions they will focus on throughout the project depending on the interest of clinician end-users and the new knowledge being developed along the way. The hope is that this initial work on a relatively small number of conditions will ultimately lead to the development of generalizable principles. That way, the same approach could be used as the number of disorders examined expands.
Technical Challenges and Workflow Issues
So what are some of the biggest hurdles the researchers will need to overcome along the way? According to Williams, one of the biggest challenges pertains to the technical issues involved with extracting information from the electronic health record to define high-impact phenotypes.
“There are some emerging approaches that we think have promise, and we’re going to be testing those,” he said. Some testing companies have utilized gene panels that run an in silico panel off of exome data. But other approaches might be used to select the appropriate genes and then use automated or semi-automated variant analysis to identify genes and variants of interest. “That’s an area where there are a lot of unsolved issues, so those ethical and standards issues are definitely one category barrier.”
The aspect that he believes will have the most significant impact on whether or not this type of approach will be implementable, however, relates to clinician workflow issues. “We know that there’s a lot of good ideas that fail because they don’t take into account how clinicians actually do their work,” he said.
As part of this project, the researchers plan to engage with clinician end-users to identify the problems that they are experiencing. “In other words,” he explained, “we’ve come up with a group of disorders that we think are interesting. But does that really match the ones that are diagnostic problems for our clinician colleagues? We want to identify high-impact problems that they deal with so that we can provide solutions to the problems that they think are important.”
Williams and colleagues want to determine the best way to provide helpful information to clinicians. This includes figuring out when and how to deliver that information so it can be believed and acted upon. “We’ve really invested a lot in this project in end-user design and user experience because we know if we get that right, the likelihood that this will be acceptable as we move into implementation is significantly higher,” he said.
“The focus in the way that the grant is going to be conducted on the engagement with clinicians increases the likelihood of implementation and generalizability,” added Adam Buchanan, associate professor and director of Geisinger’s Genomic Medicine Institute. “It takes a long time for effective medical interventions to actually make their way into routine practice, and so I’m hopeful that the thoughtful engagement will shortcut that implementation timeline,” he told Diagnostics World.
While it is not yet addressed in this pre-implementation project, another dilemma highlighted by Buchanan pertains to reimbursement. Access to exome data and electronic health record data is an important step, he said. Still, if this is going to be sustainable, it will need to be reimbursable over time. For now, there is not a well-established method for insurers to know how to set their cost structure for this type of real-time review of exome and clinical data.
“At the present time, what we’re proposing really doesn’t have any reimbursement for the activity,” added Williams. The thought is that getting people on appropriate treatment early on could be beneficial to the system as a whole. “But it’s a higher-level economic question that would certainly need to be addressed.”
Paul Nicolaus is a freelance writer specializing in science, nature, and health. Learn more at www.nicolauswriting.com.