NEW YORK (GenomeWeb) – Using data from the Pharmacogenomics arm of the National Human Genome Research Institute's Electronic Medical Records and Genomics (eMERGE) project, a new study has illustrated some of the pitfalls of analyzing and reporting rare disease variants in an unselected healthy population — challenges that will be important to overcome if secondary findings are to be returned to sequencing subjects more widely in the future.
The study, published yesterday in JAMA by a team led by Dan Roden and Sara Van Driest of the Vanderbilt University Medical Center, focused on two genes with links to cardiac arrhythmias. The genes are among those that the American College of Medical Genetics and Genomics recommends be returned to patients when incidentally identified in sequenced data.
In the new study, the eMERGE PGx team looked at how variants in SCN5A and KCNH2 were classified and then compared the patterns of reportedly pathogenic variants with about 2,000 participants' electronic medical records, attempting to correlate the phenotypes of the cohort — their clinical symptoms (or lack thereof) — with their genotypes.
The results revealed that not only was there a lack of consistency in whether variants were deemed pathogenic between different labs participating in that analysis, but also that individuals with a purportedly pathogenic variant in one of the two genes most often did not have any symptoms or signs of the associated conditions.
"A common vision is that at some point people will have swaths of genomic information embedded in their electronic medical records to be used when appropriate, but if you are going to start to do this, there are big questions that have to be answered," Roden told GenomeWeb. "When we get to the point of returning sequence data to patients, it's not totally clear what we would say to a patient who has one of the variants we studied here, and by extension, many other genes that fall into the same kind of category of producing severe phenotypes in some subjects."
JAMA Associate Editor William Gregory Feero echoed Roden's conclusion in an editorial accompanying the Vanderbilt study, saying the findings "expose some of the shortcomings" in our current ability to meaningfully predict the consequences of genetic variation thought to be causally related to serious disorders.
The two genes studied by the Vanderbilt group are part of a larger PGx panel that was implemented under the PGx arm of eMERGE in conjunction with the Pharmacogenomics research network.
Roden further told GenomeWeb that the issues this new study addresses are not necessarily new.
The field is currently grappling with issues of consistency in variant calling and interpretation across a range of clinical sequencing efforts. Meanwhile, the notion that the clinical implications of incidentally discovered variants are poorly defined has been a longstanding point of controversy. It's also a central focus of the third phase of eMERGE which was cleared late in 2014 at a meeting where Teri Manolio, director of the division of genomic medicine at NHGRI, presented some early data from the SCN5A and KCNH2 analysis.
Grants for the new third phase were awarded last September.
"Establishing the clinical validity of genetic variations proposed as biomarkers for important health conditions can be technically challenging, time-consuming, and expensive … [and this report] … provides a glimpse of a potential future in which EMR data might be used to define the clinical validity of biomarkers, genetic or otherwise, more rapidly, and at potentially lower cost, than is possible via traditional approaches," Feero wrote in his commentary.
According to Roden and his colleagues, they are the first to take advantage of the growing links between genetic data and EMRs being collected by eMERGE partner institutions to attempt to shed light on the pathogenicity of rare variants.For their study, they looked at EMR data for 2,022 individuals who were genetically tested as part of the eMERGE PGx project.
Three laboratories — GeneDx, Transgenomic, and the Mayo Clinic Windland Smith Rice Sudden Death Genomics Laboratory — assessed variants in SCN5A and KCNH2 for potential pathogenicity. All variants designated as "pathogenic" or "likely pathogenic" by any one of the three or by ClinVar were considered to be putatively pathogenic in the study.
Among the 2,022 study participants, 223 individuals (11 percent of the study cohort) had a total of 122 rare variants in the two arrhythmia susceptibility genes, the authors reported.
The subsequent review and analysis of these variants designated 42 as potentially pathogenic. However, the three laboratories doing the analysis disagreed significantly on the pathogenicity of some of the variants.
More importantly, when the researchers reviewed patients' EMR data for evidence of arrhythmia or abnormal ECG findings, they saw little difference in the prevalence of diagnoses or relevant phenotypes between participants with pathogenic variants compared to those without. Among patients with purported pathogenic variants, only 35 percent had evidence of any arrhythmia or other ECG phenotype, the group reported.
"That was surprising," Roden said. "While we recognize these diseases are not completely penetrant, the extent we saw that this was true was much larger than I would have thought."
As part of the initial eMERGE PGx effort, Roden said the group analyzed about 9000 subjects in total using a panel that included six genes on the list recommended for incidental return by ACMG.
"I'd love to say yeah we are definitely going to do look at these two genes and the other four in the other 7000 patients, but just this study was an immense effort," he added.
Meanwhile, as the new third phase of eMERGE moves forward, the team is working out their strategy for a project Roden said will hopefully yield a huge multiplication of the type of data he and his colleagues just published in JAMA.
"The centerpiece of eMERGE III is to take 100 genes, including the two we looked at, and sequence those in 25,000 subjects. So we are going to do the same study, but times 12.5 in patients and times 49 in genes," he added. "We have the gene list, the sequencing centers, and we have decided on a variant calling algorithm to use across the network, but the other aspects of how we are going to do it, most importantly [what we are going to return to participants,] is something we are still working out."