CHICAGO – At last month's American Medical Informatics Association (AMIA) annual symposium in Washington, D.C., a panel representing molecular laboratories, health systems, vendors of electronic medical records, and creators of clinical decision support content seemed to agree that pharmacogenomics is one of several necessary components of precision medicine, but the promise cannot be realized until PGx becomes an unobtrusive element of clinician workflow.
Present at that session was Nephi Walton, associate medical director of precision genomics at Intermountain Healthcare in Salt Lake City. During the Q&A portion, Walton stood up to say that no frontline clinician is going to read pharmacogenomic reports unless there is "active" clinical decision support suggesting specific, actionable steps practitioners can take.
He later told GenomeWeb that clinical decision support is of limited effectiveness without discrete data from EMRs, genetic tests, and other lab results.
Days after the AMIA panel, a paper appeared in the Journal of Personalized Medicine describing early efforts to build a bioinformatics infrastructure at Intermountain that seeks to facilitate delivery of this kind of knowledge to frontline clinicians.
The informatics infrastructure works in concert with HerediGene, a global population genomics initiative led by Decode Genetics and Intermountain Healthcare that launched in June 2019 to collect samples from 500,000 participants and discover new connections between genetics and human disease.
Walton said recently that the program has enrolled more than 150,000 participants to date. The Journal of Personalized Medicine paper covers the first 26,302 participants, all of whom had whole-genome sequencing through HerediGene.
Even with this cohort representing barely 5 percent of the planned size of the HerediGene database, the Intermountain researchers including Walton wrote that they "have demonstrated tremendous potential to impact patient care."
The paper talks more about processes than technology. Walton is in charge of the return of results for HerediGene, but the infrastructure is about a lot more than just returning results. "Once you get to a certain stage, it's really hard to scale this because you really need to be coordinated with cardiology and with oncology," he said.
"You need to be able to track and see if the patient is getting the appropriate care and you're actually impacting outcomes," Walton explained. That includes offering effective clinical pathways and the pros and cons of each alternative to multidisciplinary clinical teams, as well as steps patients can take in their daily lives.
The initial implementation of the infrastructure targeted the US Centers for Disease Control and Prevention tier one conditions — breast cancer, ovarian cancer, Lynch syndrome, and familial hypercholesterolemia — as well as hereditary hemochromatosis.
"We developed our framework with the implementation of [hereditary hemochromatosis] return of results and then scaled our processes to both cardiology and oncology domains simultaneously," the researchers wrote.
These results are being augmented by other clinical genomics programs at Intermountain. "With the increased uptake of genome and exome sequencing, in addition to increased diagnostic yield, there are a significant number of secondary findings that are being returned to patients," the authors wrote.
They said that this population genomics framework within Intermountain overcomes a key hurdle to large-scale WGS by providing resources and help clinicians to manage findings from genetic testing because it "reduces the cognitive load, effort, and stress for a testing provider," according to the paper.
"Some of the pushback on adoption of genome sequencing is all of this extra data and what to do with it," Walton told GenomeWeb. "Once you have a population health genomics infrastructure in place, that takes the burden off the ordering provider and allows them to order whole-genome sequencing because … there's a system to deal with that [data] and they don't have to worry about it."
Walton, chair of AMIA's Genomics and Translational Bioinformatics Working Group, called the Intermountain technology infrastructure "a step forward in enabling whole-genome sequencing as a first-line test" for all kinds of medical conditions.
"Testing is cheap. The interpretation is what's expensive," Walton said. "It really makes sense to do things that are going to offer more benefit for the patient, to give them information about other disease risk and pharmacogenomics, and eventually polygenic risk scores."
Intermountain has not actually implemented clinical decision support for pharmacogenomics, however, because the organization's Oracle Cerner electronic medical record does not yet support it. Walton came to Intermountain in early 2020 from Geisinger Health System, which has an Epic Systems EMR that does offer such functionality.
Oracle purchased Cerner for $28.3 billion in a deal that closed in June. The North Kansas City, Missouri-based EMR vendor had significant leadership turnover in the wake of the death of founding CEO Neal Patterson in 2017. David Feinberg, whom Walton worked for at Geisinger, became CEO of Cerner in October 2021 and was elevated to chairman of Oracle Health a year later.
The population genetics programs at Intermountain and Geisinger are among the most comprehensive in the US, surpassed only by All of Us, which is run by the US Department of Health and Human Services rather than a health system.
"What I've done is essentially take the experience from the two largest [population genomics programs at health systems] and come up with a solution to scale, because you have to have that when you're working with that many patients and that much data," Walton said.
Walton said that Intermountain is doing some custom programming with Cerner application programming interfaces, but he has faced a bit of a leadership vacuum in the vendor's genomics division. His earlier contacts on the genomics team have left the company, but one is back by virtue of the Oracle purchase. However, integration and restructuring of the Cerner assets continue.
Another complication with interoperability is that Intermountain now supports three different EMR brands thanks to recent mergers. SCL Health, which Intermountain took over in April, has Epic, while the former HealthCare Partners Nevada, which joined Intermountain in 2019, uses an Allscripts Healthcare Solutions EMR.
Walton said that the central architecture that manages software at all Intermountain facilities interfaces to each EMR but is independent of it rather than tied to a specific vendor.
"Actually it's sort of advantageous to have all these different EMR systems as we build this out because it gives us the ability to scale it even beyond Intermountain eventually," Walton said.
With the technology in place today, as test results come in, Intermountain puts the data into a large repository that is then searched for evidence of actionable genetic variants.
The platform is mostly automated, with human intervention necessary in a small number of cases. Walton said that, for example, any BRCA1 mutations with a ClinVar three-star rating, meaning that the variants are highly reliable predictors of pathogenicity, go through automatically.
Anything with a lower ClinVar rating for BRCA1 goes through review and classification by an external lab. "That will filter out some variants that will remain in our system because they may become valuable down the road as more knowledge is there," Walton explained.
These variants then go back to Walton and his colleagues in the oncology department for review. Some are submitted for an extra check by the organization's variant review committee if the evidence linking variant to condition or risk is not strong enough.
"The majority of variants actually don't go through variant review committee," Walton said. "But we have to be pretty conservative when it comes to oncology because you don't want to be returning a result and having someone get a mastectomy because you gave them a false positive result."
With familial hypercholesterolemia, a false positive will not lead to a radical intervention like a mastectomy, so the team can be a little less conservative, verifying the condition with a lab test.
Walton last month presented about this infrastructure to University of Utah Health, which is somewhat but not fully related to Intermountain, and the feedback was similar to what was discussed at AMIA.
"They don't want to deal with all these secondary findings that come from their genetic testing," Walton said. "They would love to just send an exome on a patient, get these secondary findings from [a BRCA mutation], and just plug you into this framework," which is what Walton is already able to do in his own whole-genome sequencing clinic at Intermountain.
In the Journal of Personalized Medicine article, Intermountain examined 29 genes in four clinical domains, but the program now includes more than 200 genes, with plans to grow even larger.
Walton said that he and his colleagues have not gotten many scathing critiques on the paper yet, but he did try to address in the article a criticism that Intermountain did not include the full list of actionable ClinGen-curated genes.
"We have to take into account that the people who put these genes on [the Intermountain list] are the domain experts," Walton said. "They're the ones who are going to be managing them, so our list is dependent on the clinical domain expertise at this time. Going beyond that right now doesn't make a lot of sense because we have to have a clinical pathway to manage it."
Walton said that the population genomics program will eventually get there.
He is excited for the expected June 2023 launch of Illumina's NovaSeq X line of sequencers, which promises to bring the cost of WGS down to about $200. The priority right now, though, is to extend the program into additional clinical domains and conditions, as well as further develop the IT architecture to support scaling. Increasing integration with the EMR systems is also high on the to-do list.