A recent workshop hosted by the Institute of Medicine highlighted the significant hurdles that sequencing technology faces as it moves into clinical practice, but also revealed a range of approaches labs could take to address these issues in the short term.
The workshop, held last week at the National Academy of Sciences in Washington, DC, was hosted by the IOM's Roundtable on Translating Genomic-based Research for Health — a group of academic, industry, and government representatives that the IOM convened in 2007 with the aim of identifying approaches to improve the translation of genomic research findings into healthcare.
Catherine Wicklund, director of the graduate program in genetic counseling at Northwestern University and co-chair of the workshop, said it grew out of a previous meeting on the value of genetic testing, where many participants raised questions about "how to integrate large amounts of data into clinical practice."
Bruce Blumberg, director of graduate medical education for Northern California Kaiser Permanente and another workshop co-chair, added that the IOM roundtable team began planning the meeting a year ago thinking that clinical sequencing was several years away. During that year, "the future became the present," he said. "The technology has gotten there more quickly than practice, guidelines, and ethics have found answers for."
In a keynote address, Les Biesecker, chief of the National Human Genome Research Institute's Genetic Disease Research Branch and principal investigator for NHGRI's ClinSeq study, outlined a range of promising uses for sequencing in clinical care — from newborn screening to carrier testing to pharmacogenomics and routine healthcare — but stressed that these applications are still "pie in the sky ideas about what the clinician could do if they had at their disposal a genome and a set of analytical tools."
The truth, he said, is that "we don't know how to do any of this."
Challenges include an inadequate infrastructure for storing and distributing clinical genomic data, a lack of "clinician-friendly" analytic software, and deep-seated mistrust among physicians about the advantages of genomic medicine. Even data generation, which on the surface may appear to be "pretty straightforward," is not exactly ready for prime time, he said.
"The sensitivity needs to be increased a bit," he said, noting that sensitivity is currently between 88 percent and 92 percent for genome-wide studies. "That is not acceptable sensitivity for a clinician who is interrogating a patient's genome for particular disease genes," Biesecker said. "Work needs to be done to get that into the high 90s."
These improvements "will take years, if not a decade or so, to implement widely and effectively," he said, acknowledging that this hasn't stopped labs from deploying clinical sequencing in the meantime. "So what should we be doing in the interim to solve these problems?"
Biesecker noted that one area where sequencing has shown clear advantages over the current standard of practice is in rare genetic diseases, and suggested that "we ride the current tiger of rare disease disorders forward."
Rare disease diagnosis is currently a "huge consumer of healthcare resources," he explained, with "patients bouncing from provider to provider for answers, getting test after test after test" — a situation that makes it easy to justify the utility of sequencing-based diagnosis.
"If we could develop a system where patients with rare inherited diseases could be more efficiently diagnosed, we could streamline that. Plus it provides us with an opportunity to build the infrastructure" to move sequencing toward routine care, he said.
"If we were to build out from rare diseases, we could start by building our sequencing and data infrastructure and start with specialists using informatics tools, who would then make them more user-friendly for the non-specialists."
In the process, he said, the community would gain valuable knowledge about one of the biggest issues related to clinical sequencing — so-called incidental findings.
"If you're sequencing a kid for a rare metabolic disorder you're going to find a lot of other things as well," he said. "We have to learn how to deal with those because those incidental findings that we call when we're doing rare disease work — that's the important stuff that we want the clinician to use in routine care in the clinical segment."
The ultimate goal, he said, should be for genomics to "disappear into clinical practice."
The clinician's role will "evolve from a person who uses clinical insight to select a test to someone who takes the available data and integrates it to answer a question," he said. Ultimately, he added, the clinician will "shift from being a phenotyper to being a bioinformatician."
Robert Nussbaum, chief of the division of medical genetics at the University of California, San Francisco, School of Medicine, also proposed that sequencing will increase the demand for data analysis, but saw things evolving slightly differently.
The key change, he said, will be a move toward "ongoing interpretation" of a patient's genome. As a result, sequencing will be "a subscription, not a test. It will be an ongoing relationship between the bioinformatician, the physician, and the patient."
Nussbaum argued for "demonstration projects" that would involve "a large health plan or provider and study the use of complete genome sequencing as part of an overall health assessment though all stages of life."
Without such a demonstration, he noted, it may be difficult to justify the clinical validity and utility of clinical sequencing.
'Maximize the Benefits, Minimize the Harms'
Henry Greely, director of Stanford University's Center for Law and the Biosciences, predicted that sequencing will be a routine part of healthcare within ten to twenty years, but outlined a laundry list of regulatory, ethical, and privacy challenges that the field must address before it gets to that point.
He noted that the healthcare community "can manage the transition in ways that make it less painful or more painful," adding that "it's important to figure out how to use this technology in ways that maximize the benefits and minimize the harms."
On the regulatory front, he argued that "all sequencers should be Class 3 regulatory devices" that will need to pass certain sensitivity and specificity requirements before being used in clinical practice.
In addition, he noted that labs will need to be CLIA certified, but more importantly, "CLIA needs to get more serious about what that means." In particular, he said that CLIA will need to better define "what kind of proficiency standards you actually have to meet before we will allow you to sequence people for clinical purposes."
He added that these standards should "not just be an issue of, 'Can the technicians run the machines properly and can they transfer data files from one place to another?' but also what kinds of systems they've got, what kinds of procedures they've got for, say, rerunning suspicious results, or rerunning particularly sensitive and important positive results to make sure there aren't false positives."
The implementation of these standards, he said, will likely "lead to a significant increase in effective regulatory attention to clinical laboratories."
He also noted that any software developed to analyze genomic data in the clinical setting should be considered a medical device and will need to be validated for sensitivity and efficacy.
Ethical issues include questions about what information physicians should share with their patients, particularly when it comes to incidental findings. In Greely's view, "the point of research sequencing is research, and the point of clinical sequencing is clinical care. The physician or other healthcare provider should have to provide clinically important information, but should not have to provide information that is not clinically significant."
[ pagebreak ]
This same strategy should carry over into consent, he said, where patients should be told what to expect up front. "You should tell them what you plan to look for. You should tell them that if you see anything else that's clinically significant, you'll tell them about it. You should tell them that if you see other things that aren't clinically significant, you probably won't tell them about it," Greely said. "You should ask them if they want to know more than that … but your obligation should be to talk about the clinically important material, and that's it."
There are also privacy challenges. "Confidentiality cannot be guaranteed" in the electronic age when anyone can download information on a laptop and then lose it, he noted.
Another question involves mechanisms for updating patient genomes as new associations are discovered, or as variants are reclassified. While a patient's germline isn't going to change, "the interpretation will change," Greely said, "probably every week as things move from one category to another category."
The community will need to develop standards that not only outline the criteria for reclassifying a benign variant as pathogenic, for example, but also regarding how often a genome should be requeried in light of updated information.
That information, he said, should be passed along to a physician or directly to the patient if the physician can't be found, though he added that it raises the question of what to do when a patient can't be found: "How hard to you have to look?"
Clinical labs that have already implemented sequencing as part of patient care have taken different approaches to handling some of these issues.
For example, there is currently no centralized public database of clinically validated variants for whole-genome analysis, which leaves many labs searching through multiple resources in order to interpret the genomes they sequence. Not only is it time-consuming to sift through a dozen or so databases to cull relevant information for whole-genome analysis, but many of these resources were curated from the literature for research purposes and contain information that has since been proven to be incorrect. In response, several labs are developing clinical-grade internal mutation databases for this purpose.
For example, Madhuri Hegde, scientific director for the Emory Genetics Laboratory at Emory University School of Medicine, said that her team has developed such a database that it uses to help interpret genomes. The Emory lab has been offering NGS-based sequencing tests for congenital disorders for about a year and a half.
The Emory team updates its database when it reclassifies variants and also notifies external databases when it finds errors in them, though Hegde noted that this process poses a number of challenges — not the least of which is a financial one.
"It's not like we have funding to do this," Hegde said. "One of the models we are considering at Emory is actually to hire curators who will curate our databases on a monthly basis, but that's a position the lab has to fund."
Heidi Rehm, director of the Laboratory for Molecular Medicine at the Partners Healthcare Center for Personalized Genetic Medicine, said that her lab has also developed an internal mutation database for interpreting whole genomes. The LMM currently offers an NGS-based cardiomyopathy panel and plans to launch a whole-genome interpretation service in 2012.
In addition to LMM's internal database, Rehm is involved in several ongoing efforts to develop a universal repository for clinically curated variation data. She is a principal investigator for MutaDataBase, an open access clinical-grade variant database managed by a non-profit foundation of the same name. Emory's Hegde is also participating in the MutaDataBase effort and serves on the foundation's board of advisors.
Rehm is also working with the National Center for Biotechnology Information, which is developing a similar database called ClinVar.
These projects are still in the early stage, however, and in the meantime most labs are on their own. "We need to get this data out of the clinical labs and into databases," she said. "Most of it is not going anywhere."
Labs are also developing systems to help physicians order and interpret sequencing-based tests.
Hegde noted that when Emory first launched its sequencing-based tests, most physicians were not providing adequate clinical or phenotypic information to help the lab interpret the results. In response, her team developed an ordering tool that includes check boxes for various phenotypes, such as developmental disorders, dysmorphologies, and neurological features. The software than uses this data to select genes that may be implicated in the disorder based on information in the literature, and then this data is used to support the interpretation of the sequencing data.
At the other end of the spectrum, Rehm's group has developed software that improves the reporting process. The system, called GeneInsight, automatically updates patient reports when variants have been reclassified. Rehm said that physicians in the Partners system are still evaluating the software, but have so far "found it useful."
Labs are also developing strategies for validating their sequencing-based assays, which most acknowledged to be a tricky process in such a quickly evolving field. Most groups are focusing on analytical validation of the platforms rather than clinical validation of specific assays.
"You need to focus on what the analytical platforms are capable of doing versus the very specific test validation that we do right now because we are not going to be able to entertain validating every single base pair that we're analyzing," said Federico Monzon, director of the Research Pathology and Molecular Diagnostics Cores at the Methodist Hospital Research Institute. The goal, he said, is to develop existing next-gen sequencing platforms to the point where labs are as confident in their results are they are in the results of Sanger sequencing.
"We can never validate every position, and can certainly never validate every variant we may identify that we've never seen before," said Rehm. As a result, "we have to use more generalizable approaches to do that where we would look for percent concordance across those results," for example.
Bruce Korf, director of the Heflin Center for Genomic Sciences at the University of Alabama and Birmingham, posited that "there may be limits" to the amount of validation that is possible for sequencing-based diagnostics. "It could be argued that we will simply never, in the history of humankind, explore every possible variant in every possible genome, and there must be a point where we recognize that there are limits to the degree that we can validate things."
That, he suggested, "is really where the practice of medicine takes over from things that are considered to be black and white."
These labs are working out their validation issues well ahead of the regulatory curve. The US Food and Drug Administration recently held an informational meeting on clinical sequencing but has not yet signaled how, or whether, it intends to regulate the field (CSN 6/29/2011). In the meantime, labs are operating under CLIA guidelines and offering sequencing-based assays as laboratory-developed tests.
While some attendees of the IOM meeting expressed concern that FDA involvement could hamper development of the field, most clinical lab representatives were supportive of the FDA's strategy so far and welcomed more regulatory clarity.
The FDA is "trying to understand what the challenges are and what their role is," Rehm said. "We still worry that there's a possibility of stifling innovation, but I haven't seen that happen, or evidence that it's going to."
She added that there should be some regulation in the field "because there are a lot of differences in quality in the environments that are out there."
Likewise, Monzon said that FDA is being "extremely cautious in how they proceed," and said he believes that any policies the agency adopts "will allow us to proceed with the development of these assays."
Hegde noted that until the FDA sorts out its regulatory stance, "the important thing is to use the current rules that are in place when you are developing a test and at least try and meet those conditions." In addition to CLIA guidelines for laboratory-developed tests, she noted that her lab complies with guidelines published by the American College of Medical Genetics in 2007 regarding the reporting of sequence variations identified in the course of clinical lab testing.
Have topics you'd like to see covered in Clinical Sequencing News? Contact the editor at btoner [at] genomeweb [.] com.