SCOTTSDALE, Arizona (GenomeWeb) – After several years of testing patients for inherited disease variants using gene panels and exomes, open questions remain around what types of results to return and how to report them, both in the clinical context and as part of research studies.
During a panel discussion at the Advances in Genome Biology and Technology Precision Health meeting in Scottsdale, Arizona on Saturday, four experts discussed their current practices, what they have learned over time, and what aspects of reporting – including different types of variants, primary and secondary findings, and updates – might warrant changes.
Overall, panelists agreed that clinical reports need to communicate results in a way that is easier to understand for both patients and doctors, allowing them to take appropriate action. One problem is that laboratories are legally prohibited from advising doctors on what to do with the information, said Heidi Rehm, director of the Laboratory for Molecular Medicine at Partners Healthcare Personalized Medicine. The reason for that is that physicians need to take other types of information into account, such as the patient's co-morbidities and family history, to make decisions about their clinical care.
Yet, there seems to be a need for guidance, Rehm said, and one idea has been to issue a second report, written by a clinical geneticist, which could explain the implications of the results, for example, which family members should be tested. However, there is currently no way to bill insurance for such an additional report, she said.
VOUS and GOUS
Regarding what goes into a clinical test report, one issue that remains contentious is the return of variants of unknown significance, known as either VUS or VOUS. While there is a very high chance that these variants will ultimately be downgraded to benign – according to the panel, this happens for more than 90 percent of them – they can also generate valuable leads that may ultimately result in a molecular diagnosis.
The danger of reporting variants of unknown significance is that physicians might misinterpret them as disease-causing and act on them inappropriately. A surgeon, for example, might recommend a mastectomy because a patient has a breast cancer family history and "there is a change" in a breast cancer gene, even if that change later turns out to be benign, said Gail Jarvik, head of the Division of Medical Genetics at the University of Washington.
Nevertheless, all panelists said they do report variants of unknown impact in genes related to the patient's phenotype, since retesting in the future is generally not an option, for lack of insurance coverage. Reporting a variant of unknown significance is often "the only way to provide relevant information that may become important several years down the line," said Julie Gastier-Foster, senior director of the clinical laboratory at the Institute for Genomic Medicine at Nationwide Children's Hospital.
What's important, though, is to provide context for a variant of unknown significance, Jarvik said, for example, pointing out that most such variants will later be reclassified as benign.
One improvement would be the addition of some quantifiable measure of how likely these variants are going to be pathogenic, said Wendy Chung, director of the clinical genetics program at Columbia University. Also, if there was a continuous flow of information between testing labs and ordering physicians, she might be more inclined to hold back reporting variants of unknown significance, but since this is usually not the case, including them in the clinical report "is the only way to get [the result] out."
According to Rehm, some labs, including her own, already subdivide variants of unknown significance into additional categories, depending on how pathogenic they are estimated to be. However, Gastier-Foster cautioned that it can be difficult to make such estimates, and some labs would rather not have more categories than the five recommended by the American College of Medical Genetics and Genomics and the Association for Molecular Pathology – pathogenic, likely pathogenic, VUS, likely benign, and benign.
Overall, variant categories could benefit from more precise definitions, panel members argued. Rehm said that "likely pathogenic," for example, is currently defined as having a 90 percent probability of being pathogenic, but arriving at that number is somewhat arbitrary. An ACMG panel did discuss a "point system" in the past to make the classification of variants into categories more rational, she said, but discarded the idea because assigning points would involve qualitative decisions and would merely create an illusion of being quantitative.
Another question clinical geneticists and testing labs are grappling with is whether to include genes not clearly associated with a disease phenotype, so-called "GOUS", in a gene panel, and whether to report deleterious variants found in such genes, for example in an exome test. Panel members agreed that in general, GOUS are more problematic than VOUS, but there can be a rationale for including them in reports.
Chung, for example, said that since there is usually only one chance to run a genetic test, she favors being "more generous" with the inclusion of genes that have not yet been clearly connected to the disease in question but might turn out to be disease-causing in the future. One way to help decide whether to return a result for such a gene in a research context, she said, is to check a resource like Matchmaker Exchange to determine whether it has been seen in other patients before.
According to Jarvik, patients may find it difficult to understand why a gene that has no established disease connection is included in a clinical report, though she said she has reported de novo variants in such genes that came up in exome tests.
Gastier-Foster said her lab generally does not report variants in GOUS but that an argument could be made for including a gene in the analysis that has not been associated with a disease, for example, but that belongs to a gene family that has a disease connection.
When it comes to the return of results that are not directly related to a patient's phenotype, arguably, a lot of parallels exist between genomic tests and MRI scans. But unlike MRIs, where doctors are required to return all results irrespective of their ability to say how they may impact the patient's health, there has been much debate about which incidental or secondary results to return for genetic tests.
Since the ACMG came out with a list of 59 genes for which findings should be reported, most labs have given patients the choice of opting in or out of such results. However, the process might need to be fine-tuned. One problem, Rehm said, is that it might create many "patients-in-waiting," who have no disease symptoms, no family history of the disease, and may never develop it.
What makes the interpretation of secondary results difficult is that the evidence for pathogenicity and penetrance is not the same for variants in all genes on the ACMG-59 panel. "Not all ACMG genes are created equal," Jarvik said, noting that she has returned pathogenic variants in genes that were later downgraded to variants of unknown significance.
Generally speaking, the evidence tends to be stronger for cancer predisposition genes than for genes involved in cardiac conditions, for example. A pathogenic variant in a cardiac disease gene, thus, may have a lower penetrance, or even be non-pathogenic, in an individual without a family history than in someone with a family history of the disease, Chung said. On the other hand, disease mutations in cancer predisposition genes tend to be truly pathogenic.
Rehm noted that Genomics England decided to report both pathogenic and likely pathogenic variants for cancer predisposition genes, but to report only pathogenic variants in cardiac disease genes as part of the 100,000 Genomes Project.
Another challenge is that the ACMG-59 genes are not equally actionable, and clinical follow-up is easier for some results than for others. There is not much to do for an asymptomatic patient without a family history of heart problems who has a pathogenic variant in a cardiac disease gene, for example, Jarvik said, whereas doctors can easily measure patients' cholesterol levels if they have a result in a gene associated with hypercholesterolemia.
In addition to returning results from the ACMG-59 gene set, there continues to be discussion about what other information to report to patients, such as disease carrier status and pharmacogenomic variants.
Jarvik said she is not in favor of returning results in a medical setting that are not useful to the patient, for example, disease carrier results for patients who are past their reproductive age. Given that healthcare resources are limited, she said, it can be difficult to justify explaining results that will likely not be used.
Chung agreed that it is not clear how useful the provision of additional information to patients is. For example, she has returned both carrier status and pharmacogenomic results to healthy individuals who underwent exome sequencing but had the impression that the information "did not sink in" and that patients did not know how to use it.
Also, people might not understand that the information they receive could be incomplete, for example, because the sequencing coverage was too low to determine carrier status for certain genes, Gastier-Foster cautioned.
Pharmacogenomic variants, in particular, might only become useful once they are added to a patient's electronic medical record, so their doctors can see them before they prescribe a drug. However, in many cases, that infrastructure is not available yet, which is why Gastier-Foster's lab currently does not report PGx results.
Jarvik said that her institution does return a limited panel of pharmacogenomic variants on a research basis, which go into the EMR and have a clinical decision support system connected. Returning PGx results in a way that's not usable might create liability issues, she said, if a doctor prescribes the wrong drug because he or she did not see the results.
However, it might take a long time before systems to make PGx results are available as part of EMRs. Ideally, Chung said, results would be accumulated in a database in the meantime, and when the infrastructure is ready, they could be added by "flicking a switch."
When it comes to returning raw sequence data, which clinical labs are required to make available to patients upon request, several panel members said they are happy to provide patients with BAM files – which contain sequence alignment data – but they do not give out VCF files, which contain variant data. The reason is that patients might try to look up variants on their own using the Internet, which might provide them with misleading information, whereas the reinterpretation of BAM files would require the involvement of an expert.
Panelists disagreed, though, about the return of genomic data from research projects. While some institutions, including Rehm's, believe that only data generated in a CLIA laboratory may be returned to individuals, others, for example, the University of Washington where Jarvik works, argue that there is an exception in the law that allows the return of data generated in a non-CLIA lab in the context of research.
Another point of contention is the return of results from fetal samples. Chung argued that as much data as possible should be returned prenatally because it allows parents to make more informed decisions, including planning for the care of a sick newborn. On the other hand, prenatal genomic test results may provide ambiguous information and can "put parents and OB/GYNs in a quandary," Gastier-Foster said, which is why her institution takes a more conservative approach to reporting prenatal results.
Updating test reports
Another discussion point of the AGBT panel was the reanalysis of genomic testing data over time. New discoveries keep adding disease-causing genes and variants to the literature, so the interpretation of the same data might change over time. However, requirements for reanalysis and for how to communicate updated test results don't currently exist.
Gastier-Foster's lab, which only recently started offering clinical exome testing, currently does not perform an automated re-evaluation of test results because of a lack of resources, she said. However, if a patient returns with an additional phenotype, the lab will reanalyze the data.
Jarvik said her team asks patients to return a couple of years later, at which point new results may be available. Also, if the testing lab sends an updated report to them, they try to recontact patients.
Chung said that if all else is equal, she prefers to send tests out to laboratories that offer to reanalyze the data at certain intervals and send out updated results. For research studies she is involved in, she said, she reanalyzes exome data every six months.
Reanalyzing genomic data is very labor-intensive, but what is easier to update is the status of variants that were included in the original test report, Rehm explained. Her lab has an automated system that sends out updated reports to physicians whenever a variant is reclassified.
But even that can be problematic – for example, if a physician no longer takes care of a patient, so the updated report never reaches the patient. An alternative might be to notify patients of updated reports directly, she suggested. Rehm also noted that patients have become savvy enough to look up their variants in databases such as ClinVar on their own.
Finally, panelists discussed data security and privacy, and how to maintain it. The reality is that data breaches are possible despite all precautions, and patients need to be aware of this. However, patients sometimes don't seem to be overly concerned about privacy. Jarvik noted, for example, that her institution takes data security very seriously, "but, patients go home and put [their result] on Facebook. They are happy to share this information."