As next-generation sequencing continues to advance into the clinic, how the US Food and Drug Administration will regulate diagnostic tests based on the technology remains a major question among corporate and academic test developers.
While a number of laboratory-developed tests use next-gen sequencing, none yet has FDA approval. However, in the last year, the agency has begun to take note of the technology's potential, including holding a workshop with stakeholders to discuss sequencing in clinical diagnostic applications (CSN 6/29/2011).
At Cambridge Health Institute's Clinical Genome Conference, held last week in San Francisco, Zivana Tezak, associate director for science and technology in the FDA's office of in vitro diagnostics, discussed the issues around regulating NGS-based diagnostic tests.
The technology poses hurdles for the FDA to regulate because next-gen tests are inherently multi-analytical, the technology is changing so rapidly that there are no performance standards, and the tests can turn up results with uncertain diagnoses.
"We're dealing with new, rapidly developing, disruptive technologies," said Tezak in a presentation. "They raise new regulatory questions."
Next-gen sequencing tests would be regulated under the office for in vitro diagnostics as a medical device, Tezak said. The office currently regulates reagents, instruments, and systems intended for use in the diagnosis of disease.
The issue with next-gen sequencing tests is not that they are genetic or genomic tests — the FDA has already cleared a number of genetic and genomic tests, including both single- and multi-gene tests and expression tests — but that there is not yet a clear cut way to analytically assess the technology's performance.
And while the definitions for evaluating safety and effectiveness of devices are wide and open for interpretation, "not everything that's been applied to [other approved genetic] tests will be applied to next-gen sequencing."
One major issue with getting approval for next-gen sequencing tests is that all tests must be approved for a specific use. The FDA evaluates the diagnostic for that use only. "But for people in the next-gen sequencing arena, when they're talking about the claims of a test, they're talking about very wide claims," said Tezak.
"We look at very specific intended uses, because then a manufacture can validate it for that specific use," said Tezak. "We don't know what the performance is for wider uses."
For instance, she said, in areas like cancer where large gene panels are sequenced and analyzed, the same panel can be used on a variety of different cancers and can uncover variants that indicate response or resistance to a variety of different drugs. The panels can also turn up variants that may influence how a cancer is stratified or disease prognosis, or whether a patient is a good candidate for surgery.
Another issue specific to next-gen sequencing has to do with how test performance is evaluated. For genomic tests the FDA has evaluated so far, it is not enough to give an overall accuracy measurement of an assay, said Tezak. "We need accuracy for each allele."
For instance, say the total accuracy of a test is 98 percent. That could equate to an accuracy of 100 percent for alleles one through 20, an accuracy of 87.5 percent for allele 22, and an accuracy of 66.67 percent for allele 23. These varying performance metrics make it difficult to state accuracy of a test as a whole.
It would be impossible to require accuracy for each allele in even a targeted next-gen sequencing test, not to mention a whole genome sequencing test, which interrogates three billion base pairs, she said. So, a new method of evaluating these tests is needed.
Additionally, even if a precise measurement can be given for calling a certain variant, interpreting what that variant means for disease state is not always straightforward. For example, she said, what happens when a novel mutation is found in a gene that predicts the risk of developing cancer? Even if that variant has been accurately called, its diagnostic relevance may be unknown.
"There's so much that we don't know," she said. "There are so many variants of unknown significance."
Finally, not only is there much that is not understood about the results of next-gen sequencing, but there are also no standards for evaluating the performance of the technology itself. Different machines have different error profiles and perform differently depending on the area of the genome that is being sequenced or even the software that is being used to analyze the results.
For instance, homopolymeric regions, indels, repeats, copy number variants, and redundant sequence are all notoriously difficult areas to analyze with next-gen sequencing, said Tezak, and different machines perform differently within each of these regions.
Therefore, in order to understand the performance of a platform as a whole, "what percentage of the genome needs to be covered, and should different regions of the genome have different performance requirements?" she questioned.
Tezak said that while there is currently "no set FDA policy" on next-gen sequencing, it recognized the potential for the technology to impact medicine, and that it is actively working with manufacturers and individual groups that are looking to develop next-gen sequencing-based tests.