NEW YORK – An ongoing program from the Mayo Clinic seeks to provide evidence that pharmacogenomics data can be implemented within a healthcare system to change the course of patient care.
A paper published late last month in Genetics in Medicine provided an overview of the Right 10K study, conducted by Mayo Clinic's Center for Individualized Medicine and the Baylor College of Medicine Human Genome Sequencing Center, and laid out how Mayo has implemented preemptive pharmacogenomics data into routine clinical care practices.
DNA from 10,077 volunteers with samples in the Mayo Clinic Biobank was sequenced for 77 genes known to play a role in drug response with a capture panel designed by researchers including Steven Scherer, a professor in Baylor's department of molecular and human genetics. Scherer said that Baylor's role in the study was to design, clinically validate, and run the test and develop software to incorporate predicted drug metabolism phenotypes for 13 of those genes into Mayo's electronic medical record system (EMR).
The sequencing panel was designed to target genes that the Clinical Pharmacogenetics Implementation Consortium (CPIC) had published clinical guidelines for and address 21 recognized drug-gene pairs, Scherer said. Once Baylor put the panel together and tested the samples, with sequencing completed at the beginning of 2019, Mayo "took the ball and ran with it," he said.
Scherer noted that the study revealed that almost everyone carries at least one clinically actionable PGx variant, with 79 percent of patients sequenced having at least three. While genotyping has typically been the method used for PGx testing, Scherer's group used sequencing because it reveals novel variants and is cost competitive — each panel was completed for about $300 per sample, because the researchers were able to multiplex the panel and run 48 samples per test. Genotyping a sample would cost about the same, but offers less information, he said.
Once Mayo got ahold of the data and transferred it into the EMR — not an easy task, Scherer said — the implementation began. That step required significant investments in infrastructure, said Liewei Wang, chair of Mayo's department of molecular pharmacology and experimental therapeutics and one of the leaders of the program. The implementation step "takes an army of investigators and physicians, as well as IT and our healthcare providers," she added, since IT systems need to be added to support and host the data in the EMR and clinicians need to be educated on how best to use the data. That implementation process is now complete, although it will need to be adapted as new drug and gene information comes out.
Although physicians at Mayo can order genotyping tests to get PGx data, the information currently has a limited application, since many clinicians may not want to wait for the results before prescribing, especially if the therapy is time-sensitive, Wang said. The new "preemptive" system provides a message directly in the EMR when a doctor is prescribing a medication with a PGx indication that encourages them to either order a PGx test or use the genetic information already available in the EMR to determine the best drug.
Mayo is not the only healthcare system to incorporate PGx alerts into EMRs. In 2016, researchers from the Cleveland Clinic published a paper in Pharmacotherapy explaining their implementation of PGx data into their EMR, and in 2017 a team from the University of Chicago published a study in Clinical Pharmacology and Therapeutics that found point-of-care decision support altered drug prescribing.
According to Richard Weinshilboum, a consultant in Mayo's division of clinical pharmacology and another author on the paper, another key finding was the "critical role" of genetic information provided by a consulting pharmacist. When patients were at significant risk, 60 percent of pharmacist recommendations were immediately accepted by the physician, Weinshilboum said. Mayo now has three full-time experts who are able to consult with clinicians. There is also a training program in place for pharmacists to learn more about PGx, and 392 of Mayo's 452 pharmacists are trained in PGx.
Future plans
Wang echoed Scherer's opinion on the benefits of sequencing versus genotyping, saying that with genotyping you "know only what you know," meaning you can only measure known gene variants. Sequencing provides "much more additional information, newer information, that currently we don't know." All of the sequencing information for the known drug-gene pairs found through the study is in Mayo's EMR, with the rest of the data captured by Baylor's panel in a "holding pattern" for research use, Wang said. When there's more info on the clinical utility of those genes, the info can be moved into the medical record.
That additional info is being used for research at Mayo to assess variants of unknown significance and determine whether they may have clinical implications via high-throughput mutational scanning. If they do show potential clinical implication, the researchers tell the clinical lab so the lab can conduct further testing, take it into consideration, and potentially add a note informing the clinician that there is relevant preclinical data. The physician can then take that info and use it in their clinical decision-making, Wang said, although she noted that the addition of that data doesn't necessarily mean the drug will be changed.
Furthering the potential use of PGx information, Weinshilboum said that Mayo has a plan in place to ensure that by 2030, every patient will have the majority of their DNA sequenced.
Mayo is also using the data to make strides in its artificial intelligence programs. There is still "a lot of other factors that we haven't really taken into account" when it comes to PGx, including other lab values and demographic information, Wang said. Mayo plans to use sophisticated AI tools to derive predicted algorithms that integrate molecular biomarkers, symptoms, and demographics to predict a patient's response to a medication.
The healthcare system already has one such algorithm in place for antidepressants in adult patients, with plans to create more. Mayo is also considering implementing the algorithm across the entire system. In a year, Weinshilboum said he hopes the whole clinic will be using the antidepressant algorithm.
Each drug type needs a separate algorithm and the development takes "years of research" to ensure there's no bias and enough information, but the basic idea remains the same, she said. Pharmacists will also play a key role in helping interpret the findings of the algorithms.
The team is working to figure out how to "use the minimal amount of information … related to an individual's healthcare to give you the maximal predictive value," she said.
Questions remain
There are still questions to be addressed after the study has been completed, however. While Mayo has made strides in determining how this info can be implemented, the research still doesn't answer how exactly adding genetic data can change physician behavior, Wang said. Because the samples used were already present in the biobank, most of the patients with available data are long-term patients who have been on their drugs for a while and are on the right dosage. In this case, additional PGx info wouldn't be particularly useful.
A relatively small portion of clinicians have changed or adjusted prescriptions based on this data, but the real value going forward is for new patients, Wang said. The question of clinical utility is "from the research side [and] is also very important," and the team plans to ask whether the data "shorten[s] the time … to achieve the appropriate or best dose range for these patients," she added. Once there is additional data, there will also be more information on how the tests might best be adopted.
Weinshilboum noted that this study wasn't designed to answer those questions — the emphasis was on the implementation of this preemptive system and the sequencing of the samples. "We're looking to have this ready to go when the drug's first prescribed, not waiting for some time and then figuring out there's a problem." Wang added that the team was specifically trying to show preemptive sequencing is feasible and has a clinical benefit.
Baylor's Scherer also mentioned the importance of demonstrating the efficacy of providing this information, which will require following study participants and tracking people as they're prescribed drugs. Future research and larger studies will need to show that providing this information early will result in less toxicity and more efficiency, he said.
Other studies have shown, however, that there are benefits to providing PGx data. The University of Chicago study found that in 2,279 patient encounters, medications with high pharmacogenomic risk were changed significantly more than drugs without PGx information. Medications with cautionary PGx information were also changed more frequently, and no pharmacogenomically high-risk medications were prescribed when clinicians used the PGx decision support tool.
There is also the question of money: Will insurers pay for these tests? In the past, Scherer said payors have not often reimbursed for these kinds of tests, although some are beginning to pay. However, the preemptive aspect of this testing may be a hard sell, as that's "not exactly the way insurance generally works."
"Everybody walking in the door really ought to have this done to begin with to where it's in the [EMR] to begin with, so that you're not interrupting the clinical flow when somebody goes to write a prescription."
Providing this sequencing would "require an investment upfront in order to show the maximum benefit," which payors may not want to offer until there's more proof that it improves efficiency, he said, although he "firmly believes" in the efficacy of preemptive testing.
Wang said that her team's future work will also look at whether there is an economic or financial benefit to preemptive PGx testing, as it will be difficult for places to adopt the practice if there's no payment.
Medicare Administrative Contractor (MAC) Palmetto GBA released a local coverage decision in 2020 that offers limited coverage for single-gene and multi-gene PGx tests as a medical decision-making tool. Noridian, another MAC, later followed suit.
One more hurdle is getting buy-in among clinicians. Scherer noted that this study has demonstrated "the need for some sort of middleman," such as a pharmacist or genetic counselor to interpret the info. In his experience, younger physicians are on board with PGx testing, while older doctors tend to be more resistant. Baylor currently doesn't have a PGx program, although Scherer said he's working on it. There are a "fair number of hills to be climbed in terms of gaining acceptance," he said.
In the US, there are attempts at improving awareness of PGx testing underway. Legislators in February introduced the Right Drug Dose Now Act, which aims to advance PGx-related awareness campaigns for the general public, as well as education programs for healthcare providers. The bill also provides funding for PGx implementation research; improving integration of PGx information into EMRs; and ensuring drug-gene interactions and drug-drug-gene interactions are more readily factored into the federal government's drug adverse events tracking system.
Mayo's program indicates further progress toward getting sequencing and genetic info implemented into regular clinical workflows. PGx has been an area of genomics that has promise in helping healthcare providers use this testing at the point of care to individualize therapies, Wang said.
It's "leading the way" toward a world where "DNA sequence data can be used for virtually the entire population," Scherer said.