By Deborah Borfitz
May 3, 2022 | Over the next few years, pharmacogenomics could start to become a standard component of clinical care to improve drug efficacy and avoid adverse responses when certain medicines are prescribed. Nearly everyone is unknowingly walking around with a handful of potentially actionable variants that could help steer the course of treatment toward the best possible outcomes, according to Steve Scherer, Ph.D., a professor of molecular and human genetics and the Human Genome Sequencing Center of Baylor College of Medicine.
Preemptive pharmacogenomics testing is already hypothetically touching every one of the 1.3 million patients being treated by the Mayo Clinic every year, says Richard Weinshilboum, M.D., professor of medicine and pharmacology. Mayo physicians are proactively being given the information through an alert in the electronic health record (EHR) system whenever they write a prescription for a drug whose effects have known genetic influences.
Under the leadership of Liewei Wang, M.D., Ph.D., director of the pharmacogenomics program as well as the Center for Individualized Medicine at the Mayo Clinic, more than 20 drug-gene pairs have been implemented into the EHR system. The intent is to remind physicians of recognized variations in drug metabolizing enzymes and transporters and to order genetic testing, she says, which could guide decision-making about drug choice, dose, and timing of administration.
The pharmacogenomics program at Mayo was made possible by the "Right Drug, Right Dose, Right Time: Using Genomic Data to Individualize Treatment" (RIGHT 10K) study, which recently published in Genetics in Medicine (DOI: 10.1016/j.gim.2022.01.022). Results point to the value of “preemptive testing,” particularly if it extends beyond DNA variants already known to influence drug metabolism, says Wang.
As with much previous research, the RIGHT 10K study also points to the wide applicability of pharmacogenomics to drugs that are being prescribed, adds Scherer. Less than 1% of people do not have any clinically actionable variants within the 13 genes written to the EHR and the conservative estimate is that almost everyone is carrying around at least three. Moreover, study participants on average were found to have at least another three predicted harmful rare variants in these genes.
A deleterious effect may never be an issue if the indicated drug never gets prescribed of course. But if it is, physicians need to know and consider the consequences, he says.
The importance of the pharmacy in bringing pharmacogenomics to the bedside can’t be understated, says Weinshilboum, which was one of the surprising findings of the RIGHT 10K study. “Before this project, we had one full-time pharmacist who specialized in pharmacogenomics, and we now have three … [who] consult with physicians and patients.”
Although the Mayo Clinic has a large cadre of genetic counselors, they generally haven’t had a pharmacology course and felt a “sense of relief” when they realized pharmacists viewed pharmacogenomics as part of their job description, Weinshilboum says.
The study also “informed pharmacists [at large] that this is an area of medicine where they have unique training and expertise,” he continues. While at other institutions nurses may assume the “gatekeeper” role in pharmacogenomics, at the Mayo Clinic pharmacists have had the biggest impact.
An educational program “sensitized” physicians to the preemptive sequencing platform, says Weinshilboum. But, unlike the pharmacists, many doctors don’t have the time or inclination to understand what the various gene-drug associations mean—and how much evidence exists to support acting on the information.
Scherer says he has given lectures on pharmacogenetics and its implementation to students at Baylor College of Medicine who are looking to work as genetic counselors and physician assistants. “There is a lot of big-time interest in this and, among medical students, the younger they are, the keener they are to really start ramping into this.”
RIGHT 10K recruited 10,077 long-term Mayo patients and used their stored blood samples from the Mayo Clinic Biobank to isolate DNA for the study. Targeted sequencing was done on 77 pharmacogenes, but some of them were tied to multiple drugs so only 13 of them—including CYP2D6 (involved in up to one-quarter of all drugs marketed) and human leukocyte antigen (implicated in several of the most severe adverse event episodes observed to date)—became part of the drug-gene pairs deposited preemptively in the EHR, explains Weinshilboum.
All the sequencing work was done at the Human Genome Sequencing Center at Baylor in a partnership that had an unprecedented level of cross-institutional collaboration, says Weinshilboum. The original list of the 77 targets came from evidence-based guidelines of the Clinical Pharmacogenetics Implementation Consortium, which examines data in support of the clinical utility of pharmacogenomic markers, where the organization has achieved consensus with its European counterpart, the Dutch Pharmacogenetics Working Group.
Targeted sequencing rather than genotyping allowed rare genetic variants, as well as the more common ones, to be captured, says Scherer. “We also made these targets backward compatible with known, large genotyping panels and made sure we included some other targets suggested by the Pharmacogenomics Research Network [of the National Institutes of Health, or NIH] as well as barcoding targets so we could follow individual samples through the entire sequencing process and assign them to the correct subject.”
The Human Genome Sequencing Center was part of the NIH network and one of three large programs of its type tasked with demonstrating the value of sequencing over genotyping, Scherer says. One of its first initiatives was to sequence a collection of network-recommended targets that were adopted by the agency’s Electronic Medical Records and Genomics (eMERGE) demonstration project.
That target set was subsequently revamped, first to make it more applicable to clinical-grade guidelines regarding drug-gene pairs and later for the RIGHT 10K study to make the targets a “bit more focused, thus driving down the cost because we could multiplex this assay at a higher rate,” says Scherer. “We were therefore able to generate sequencing information that was much deeper than genotyping for approximately the same price.”
Rare variants are typically uncharacterized but more numerous and may impact future drug prescribing practices, Scherer says. “At some point down the road they will be characterized... with further sequencing projects done throughout the world.”
Wang’s lab, for its part, is applying deep mutational scanning to assess variants of unknown significance in terms of their impact on cell function and protein levels, including that of messenger RNA. It draws on high-throughput DNA sequencing to assess the functional capacity of many variants simultaneously, she explains.
The rationale for the RIGHT 10K study is tied to the creation of the Center for Individualized Medicine at the Mayo Clinic a decade ago, says Wang. “We were trying to translate genetic testing… [to] the bedside.”
Building the required infrastructure was the first order of business, she continues. This included integration of data pipelines, interpretive reports, and best practice alerts into the EHR.
A pharmacogenomics task force—comprised of physicians, pharmacists, and researchers—was also created to vet the pharmacogenetic information that would be shared with prescribing clinicians, says Wang. Most of the early gene-drug pairs had already been recommended by the U.S. Food and Drug Administration, various nonprofits, professional societies, and working groups.
The alert system was initially integrated into clinical practice at all of Mayo’s major campuses (Rochester, Minnesota; Scottsdale and Phoenix, Arizona, and Jacksonville, Florida) about six years ago but its benefits were limited to patients who had already been genetically tested, she says. At the point of care, physicians as well as patients generally don’t want to delay the start of treatment, as reflected by the RIGHT 10K finding that primary care physicians were accepting only 54% of pharmacists’ 2,782 semi-urgent eConsult recommendations based on the sequencing data.
That gave rise to the idea of preemptively doing the sequencing—particularly given increasing adoption of the technology in the clinic, says Wang. “Once physicians prescribe a drug that has pharmacogenomic [testing] indications, now that information is already there, so they don’t have to wait.” An alert from the EHR can “directly recommend a therapeutic decision.”
Moving forward, if any of the participants in the RIGHT 10K study are prescribed an antidepressant, Weinshilboum interjects with a for-instance, those with variants in cytochrome P450 genes will be flagged as being at increased risk for experiencing side effects. In the past, when drugs weren’t working or were causing an adverse response, the problems were typically “stumbled upon” by physicians and a change was immediately made.
The expectation is that the EHR alerts will increase adoption of pharmacogenomic testing in the clinical setting, Wang says. Whether that testing enhances therapeutic efficacy and reduces drug toxicity, as well as changes physician behavior, can also now be tested in a prospective fashion as new prescriptions are written. “RIGHT 10K was just a first step.”
Pharmacogenomics will more often be helping to guide physician prescribing habits, predicts Scherer, noting that most drug-gene pairs have a “bit of subtlety to them.” Patients who need to have their blood thinned will often be prescribed warfarin, for example, which initially involves a lot of blood tests and guesswork on the part of physicians to get the dosing right. Now, information that can help improve that original guess—and eliminate the uncertainty and wasted time and energy—will be right at their fingertips.
While scientific evidence determines the clinical utility of drug-gene pairs, cost is also a factor when it comes to “how wide and how broad” pharmacogenomic testing can be, says Wang. But for the discovery of rare variants that might contribute to phenotypes related to drug response, the more data the better.
Over time, the focus on preemptive sequencing will expand to the development of biomarkers based on a combination of genomics, metabolomics, proteomics, and clinical variables such as diagnosis and demographics, she adds. “Computer algorithms will integrate all these different components for prediction.”
Researchers in Mayo’s Center for Individualized Medicine have in fact already started down that path, says Wang. As they showed in a study published last year in Neuropsychopharmacology (DOI: 10.1038/s41386-020-00943-x), a machine learning algorithm could help clinicians accurately and efficiently predict whether patients with major depressive disorder will respond to an antidepressant based on specific symptoms and thresholds of improvement.
The RIGHT 10K study effectively utilized sequencing technology as a steppingstone toward using the entirety of the genome, says Scherer, which will require long-read sequencing techniques to get a complete set of the body’s genetic instructions. And this is but one data feed into algorithms that will also be considering a host of other variables, including gene-gene and drug-drug interactions as well as rare variants and their impact on gene function and expression.
Institutions looking to implement any sort of drug-gene testing program will have a few challenges to face, says Scherer, including physician education and information technology to get results incorporated into the EHR. They’ll also need to decide who—pharmacists, nurses, physician assistants, or genetic counselors—will serve as support staff.
Preemptive testing requires a big investment, and payers are only now starting to see the value in reimbursing for the service, Scherer notes. Some insurance companies and Medicare subcontractors are starting to offer coverage, and more are likely to follow suit as evidence emerges that the data “really moves the needle with respect to healthcare.”