NEW YORK (GenomeWeb) – The Clinical Pharmacogenetics Implementation Consortium hopes to issue guidelines for genes that the expert body has determined lack evidence and shouldn't be used to guide treatment decisions but that companies nonetheless have incorporated in testing and are heavily marketing.
At a meeting of the consortium earlier this month in Memphis, Tennessee, CPIC Co-chair Mary Relling enumerated a long list of things the group has set its sights on achieving, including issuing "guidelines for non-actionable but heavily marketed genes."
"Part of our job as clinical pharmacologists is to advocate for the appropriate, evidence-based use of drugs and tests and to counter detailed marketing that's inappropriate," said Relling, who also chairs the pharmaceutical sciences department as St. Jude Children's Research Hospital. "We would love to be able to help with that."
To an extent, CPIC already calls out specific diplotypes or phenotypes within a gene-drug pair that lacks evidence for clinical actionability. But there are many more genes not specifically addressed in CPIC guidelines that have lackluster evidence, but labs are marketing tests for them anyway.
MTHFR is one such example. A quick internet search yields numerous sites, such as this one, that claim that mutations in MTHFR are associated with a host of diseases, including heart disease, stroke, high blood pressure, pregnancy complications, psychiatric disorders, autism, and certain types of cancer. But studies have found no evidence backing such gene-disease associations, and numerous expert bodies, including the American College of Medical Genetics and Genomics, advise against routinely testing for MTHFR variants.
CPIC hasn't issued guidelines on this gene but lists it alongside a number of drugs that it has been associated with in the literature, such as the rheumatoid arthritis drug methotrexate and the chemotherapy oxaliplatin. However, these gene-drug associations currently have a "C" level rating from CPIC, meaning the gene-drug pairs don’t have enough evidence for clinical actionability.
"We've learned from surveys of our members and others that they would very much like to have guidance on which gene-drug pairs are not actionable," Relling told GenomeWeb.
While CPIC has been considering issuing guidelines on non-actionable genes for some time, the amount of work it has on its plate related to the list of PGx genes that are actionable has hindered progress. However, Relling believes there is more urgency for the group to address non-actionable genes more directly now that the US Food and Drug Administration is paying more attention to how pharmacogenetic tests are being marketed.
The FDA considers "valid" gene-drug PGx associations to be those that it has reviewed and referenced in drug labeling. Last year, the FDA issued a safety alert cautioning the medical community that labs were selling tests for gene-drug associations that aren't in drug labeling and reached out to several firms about their marketing practices. According to the FDA, most of the companies it reached out to agreed to its demands to remove mentions of specific drugs from promotional materials and test reports, but Inova Health System in Northern Virginia didn't comply to the agency's satisfaction.
This resulted in the agency sending a warning letter Inova. Since then, the health system entirely halted PGx testing. While CPIC agrees that some entities may be marketing PGx testing for indications that lack evidence, the expert group understandably departs with the agency's view that only FDA-approved drug labels are a source of valid gene-drug associations.
In a blog post, Relling and Teri Klein from Stanford University, who are CPIC's co-principal investigators, responded to FDA's warning letter specifically challenging the agency's assertion in the letter to Inova that "the relationship between CYP2C19 genotype and drug response to escitalopram and sertraline is not established." They pointed out that CPIC experts have concluded after a review of the literature that CYP2C19 phenotypes are actionable for these two drugs.
According to Relling, CPIC has also sent FDA a letter and reached out to leaders at the agency inquiring about its rationale behind the warning letter to Inova. Currently, an FDA official is listed as a CPIC member with observer status. "We have reached out to members of the FDA to ask them to continue to participate in CPIC and offer to meet with them regularly to discuss issues," she said. "We want to keep the lines of communication open with the FDA."
In addition to engaging with the FDA on policy issues that impact PGx testing, CPIC has a lengthy list of action items from making its guidelines more computable, including the development of an application programming interface, to expanding standardization of PGx terms in the community, and working with other groups like ClinGen and ClinVar to increase the visibility and use of its guidelines in the field.
It is largely accepted in healthcare that pharmacogenetic information is most useful when clinicians have this information at the point of care and embedded within electronic health records and clinical decision support. However, the lack of standardization among labs in reporting allele function and phenotype terms is a major hindrance to the uptake of pharmacogenetic testing.
"You can imagine if every hospital has their own set of terms," Relling said. "That's going to [make it] impossible for EHR vendors or for other … software developers to come up with a standardized system of driving the clinical decision support, when we can't even agree on the test names or the basic interpretation of the test results."
To address this problem, CPIC initiated a survey among pharmacogenetics experts a few years ago and took a crack at standardizing PGx terms for certain genes. They then published the standard terms in a 2017 Genetics in Medicine paper, and Relling said that since then there has been good uptake of these terms.
However, in that initial process, CPIC focused on the genes that were the easiest to tackle, such as the drug metabolizing enzyme, transporter, and HLA genes. CPIC plans to engage the community in a similar effort to try to achieve standard terms for genes that experts couldn't agree on in that first process, such as VKORC1 and certain pharmacodynamic genes.
"Okay, we did the easy genes," Relling said. "Now, it's time to tackle some of these more difficult genes and see if we can come up with consensus terms for allele function and phenotype for them."
Reaching consensus is challenging when experts have come up with their own terms for well-known genes, used them for decades, and published papers on them. For example, although the term "extensive metabolizer" had long been commonly used in pharmacogenetics research to mean "normal metabolizer," CPIC's survey revealed that most experts wanted to do away with the former term and just use the latter to avoid confusion in the clinical setting.
Based on the consensus agreement CPIC achieved, the group will now use the term "normal metabolizer" in guidelines and is expecting others in the field to do the same, but Relling said that doing so also risks angering the many scientists who have been using the term "extensive metabolizer" over the last 40 years.
Additionally, CPIC has been working with other groups, including ClinGen and ClinVar to ensure that its guidelines are incorporated into those widely used resources. The two NIH-funded efforts are working together to improve the field's knowledge of genomic variants in medicine. ClinVar is a database that aggregates information on genomic variants and their relationship to human health. ClinGen convenes experts to curate knowledge on genes and variants and is fast becoming a central resource for defining the clinical relevance of genes and variants used in precision medicine and research.
ClinGen publishes summaries for genes and the associated phenotypes, and CPIC wants to work with the group to ensure that when it comes to pharmacogenes, the resource references the consortium's guidelines. CPIC and PharmGKB, the online knowledgebase that aggregates and curates information on genetic variants and drug response, have also been working for several years to figure out how to get information from CPIC into ClinVar.
However, the structure of how variants are displayed in ClinVar isn't very amenable to how pharmacogenes are reported, according to Relling. At a very basic level, the clinical significance of variants in ClinVar are described as either likely benign, benign, variant of unknown significance, likely pathogenic, or pathogenic. "Those terms just don't have meaning for pharmacogenes," she said.
Moreover, CPIC's guidelines are gene-centric, whereas ClinVar is variant-centric. When describing variation in pharmacogenes, it's important to note the diplotypes, but it's been challenging to figure out how to get information on diplotypes into ClinVar. "It's just not a trivial matter in order to figure out how to get the information from CPIC guidelines into ClinVar in a useful way," Relling said. "But we're not giving up."
Given the long list of projects CPIC wants to tackle in addition to its core work of advancing guidelines on pharmacogenes, the consortium is also exploring future funding models. To date, CPIC has been funded via grants from NIH, and the consortium is currently operating with a five-year grant that began in July 2018.
However, the NIH is "under quite a bit of pressure," Relling said, to stop funding database-type resources. As a contingency plan, in case NIH funding should ever be lost, CPIC has been considering how it would continue its work. CPIC has strict rules limiting conflicts of interest of those who author its guidelines, and if the consortium were to start accepting donations, for example, the donors would "have to be somehow kept at arm's length," according to Relling.
"It's critical we not have conflicts of interest," she emphasized to the attendees of the CPIC meeting in Memphis. "It's critical that our guidelines are developed by people who aren't going to make a dime based on whether the testing is done or not. It's been a challenge to figure out how to do it, and we will be asking for help from all of you to try to figure it out."