NEW YORK — What a certain genetic variant means to one testing lab might not be the same as what it means to another, leaving genetic counselors to work the interpretation out.
At the National Society of Genetic Counselors annual meeting this year, Brian Reys, a genetic counselor at the University of Texas Southwestern Medical Center, presented data from a survey he and his colleague Sayoni Lahiri conducted on discrepant variant classifications. Most — about 88 percent — of the genetic counselor respondents said they assessed lab results they received for discrepant variant calls. Ninety percent said they examined all variants of uncertain significance, 43.5 percent said they examined all positive results, and 47.5 percent said they examined them if the classifications were inconsistent with patients' phenotypes.
These discrepancies in classification arise even after the American College of Medical Genetics and Genomics, Association for Molecular Pathology, and College of American Pathologists issued guidelines in 2015 for the interpretation and evaluation of evidence to classify variants and established standard terminology for those classifications: pathogenic or likely pathogenic variants, variants of uncertain significance, and benign or likely benign variants.
As Lahiri added, "there is a lot of room for subjectivity within those guidelines."
But while discrepant variant classifications are not necessarily an everyday encounter, most genetic counselors do come across them, and they can affect patient care. There are, further, efforts not only to improve variant interpretations but also to encourage sharing of variant classifications.
Not common, not rare
A recent analysis examined agreement among germline variant classifications in ClinVar, a database where labs submit variant interpretations and where many genetic counselors turn to gauge consensus.
That analysis found labs are likely to agree on the classification of most variants. The paper, which appeared in Genetics in Medicine in 2017 from Invitae researchers, found an 81 percent complete agreement on pathogenicity and an 89.3 percent majority consensus on pathogenicity for variants in the database.
Discrepancies, they found, were more likely to affect older classifications. For variants that had multiple submissions from after 2014, 98.4 percent of variants had majority consensus and 97.2 percent of variants had complete agreement.
Meagan Farmer, a genetic counselor at the University of Alabama at Birmingham and Genetic Clinical Operations director at My Gene Counsel, said she found these results reassuring. "It showed more recent variant classification from clinical testing labs really had high overall concordance in many clinical areas, especially in hereditary cancer, which is my area of expertise clinically," she said.
Consensus, though, varied by clinical area, with hereditary cancer having the highest majority consensus rate, while cardiology and metabolic disorders had the lowest.
But most genetic counselors do encounter discrepant variant interpretations during the course of their careers. In a survey reported in Genetics in Medicine in 2019, Karen Wain, a genetic counselor at Geisinger, and her colleagues found that about two-thirds of respondents said they had at some point come to a different conclusion about a variant's classification than a testing lab. But many said that was an uncommon occurrence — 78 percent of respondents said it was rare and 19 percent said it was infrequent.
Nevertheless, encountering such discrepant variant interpretations can affect clinical decisions, UT Southwestern's Reys said. "When you're a clinician, any discordance that impacts how you're making recommendations to patients feels like a big deal," he added. "In a group of six genetic counselors that sees over 6,000 patients a year, even something that only happens one percent of the time really multiplies."
These differences in classifications can spark ethical questions, if, for instance, patients with the same variant have their care managed differently. "When you're going by the test report and you have one patient who has a pathogenic result, then you might have another patient who has an uncertain result," UT Southwestern's Lahiri said. "And so, then you have this conflict of, well, which one is the correct interpretation? How do I manage these patients? Is it ethical to give them different recommendations when they have the same exact finding?"
Aarti Ramdaney, a genetic counselor at UTHealth Houston, recounted a case published in Genetics in Medicine in 2018 in which a discrepant variant could have affected patient care and decisions.
A pregnant woman who underwent carrier screening was told she was a carrier of a likely pathogenic variant for Becker muscular dystrophy, an X-linked condition. Because the fetus was male, she had a 50 percent risk of an affected pregnancy.
But when Ramdaney and her colleagues dug deeper into this variant, they noticed inconsistencies. Upon contacting the reporting lab, they found that the two affected individuals in the literature were actually the same person. Additionally, when Ramdaney spoke to another lab, she was told it would have classified the variant differently.
"To be likely pathogenic, it should ideally be above a 90 percent threshold that the variant is truly disease causing," Ramdaney said. "And I think that's why I was so surprised with how this lab classified that variant. It really did not feel anywhere close to that."
Ramdaney later learned the lab re-classified the variant as a VUS, but by that time, the child was 3 years old.
She added that discrepant variant classifications might be particularly troublesome in the prenatal realm where there are fewer physical clues that clinicians might be able draw upon to bolster their interpretation of a variant. In this case, the couple could have "decided to pursue prenatal diagnosis or if they had changed management plans, that would have been a very different situation," she said.
Efforts to address
While databases like ClinVar and ClinGen have helped gather different interpretations of variants and the relationships between variants and phenotype, and guidelines from ACMG, AMP, and CAP have helped variant interpretation, refinements could still be made.
In a survey conducted by Geisinger's Wain, genetic counselors said that improved lab reports and increased lab submissions to databases like ClinVar could be helpful to support variant interpretations. UTHealth Houston's Ramdaney further noted that while labs do share to the database, it's not always done consistently or in real time.
"I definitely endorse responsible data sharing 100 percent," Wain added.
According to UT Southwestern's Lahiri, much of the issue of discrepant variant classifications could be addressed if labs worked with each other, adding that studies have suggested that the vast majority of discrepant results can be resolved when labs talk to each other. "When those labs come together, they are able to resolve a lot of those discrepancies just through discussion and kind of sharing their internal data," she said.
Rachid Karam, director of R&D for Translational and Clinical Research at Ambry Genetics, said there are such initiatives and that he is co-chair of a variant interpretation committee for CDH1, a gene that causes breast and gastric cancer, for ClinGen. During the meetings, people from different labs exchange data to try to come to an agreement on discrepant variant interpretations within that gene.
"But as you can imagine, it's a challenging project because it's volunteer work," Karam said. "We meet once a month, and the number of variants is huge, and we're talking about one gene."
At the same time, guidelines governing variant interpretation could be improved. "The ACMG/AMP guidelines were a really important step forward," Wain said. But she added that they were a starting point, noting they have been adapted for different genes or for various disease-specific considerations.
"Not every component of the original guideline applies perfectly in every situation," she said.
Karam is also focused on improving variant classifications, by folding in additional data and levels of evidence.
He noted that Ambry's CancerNext RNAinsight test includes both DNA and RNA analyses, which could provide an additional layer of supporting evidence that a variant is pathogenic. For instance, if the DNA findings show the variant affects RNA splicing, the RNA analysis will then provide further data on whether or not it does. "Many times," he said, "this pushes the classification" to one category or another.
Karam and his colleagues are also using a CRISPR-based approach to determine the effects of certain variants to develop a catalog that can be queried when a patient arrives with that variant. As they reported this week in the American Journal of Human Genetics, he and his colleagues at the University of Washington examined whether data like this could help resolve variants of uncertain significance. Using VUS in BRCA1, TP53, and PTEN as test cases, they found that multiplexed functional data could reclassify nearly half of the VUS in BRCA1, 69 percent in TP53, and 15 percent in PTEN.
But Karam noted that while efforts like these might improve variant classifications, they could also contribute to discrepancies in classifications if other labs are not taking similar steps.
Currently, handling discrepant variant interpretations falls to genetic counselors. As part of her survey, Wain assessed whether genetic counselors responded appropriately to different hypothetical patient scenarios for variant interpretation, and she found that genetic counselors generally applied the evidence correctly.
Generally, as the survey conducted by UT Southwestern's Reys and Lahiri found, most genetic counselors, 87 percent, turn to public databases like ClinVar to assess variants, and 62 percent contact the reporting lab, findings similar to what Wain's analysis found earlier.
Wain said her findings indicate that variant interpretation is a key part of genetic counselors' clinical practice and that they should acknowledge their expertise in variant interpretation. By that, she said she means genetic counselors should speak up when they are part of a multidisciplinary team — especially where they might be the only genetics professional present — when an error in interpretation is being made or when a nuance in that interpretation is being overlooked.
"That's one of the reasons that I think it's really important that we recognize this practice and how genetic counselors bring this expertise to a clinical team," she said.
Karam added that he's optimistic that discrepancies in variant classification will eventually be worked out. "I believe that eventually we will have enough data to classify most variants," he said. "We are in the infancy of genetic testing."