For those who are hoping to commercialize a mass spec pattern-based cancer diagnostic in the near future, navigating the waters of FDA approval is not likely to be a clear-cut process. As Correlogic Systems discovered last month — when the FDA sent the company a letter warning that the ovarian cancer test it had licensed to clinical reference laboratories LabCorp and Quest may be subject to premarket review — it’s not enough to simply call your diagnostic a “home brew” and hope the FDA takes your word for it. (see PM 2-27-04). Preparing for FDA approval, on the other hand, involves entering a realm that even the FDA itself doesn’t appear to understand.
“We just don’t know [what is required], because this has never been done before,” said Jeff Gibbs, a lawyer who specializes in FDA regulation of medical devices at the firm Hyman, Phelps, and McNamara. While Gibbs said companies seeking FDA approval should consider the intended use of the test; what exact claims it is making as to what the test can do; and whether there is a predicate device to which the company is comparing its test, the FDA has not issued any guidance on what special issues might exist for mass spec-based protein pattern tests, and appears to only now be trying to figure this out.
Almost exactly a year ago, the FDA took its first step in addressing multiplexed tests with the release of a draft guidance for the submission of multiplexed assays such as microarrays. This guidance mostly focused on DNA-based applications and did not discuss mass spec-based pattern proteomic tests (click here to view the guidance). In November 2003, the FDA followed with a draft guidance for the submission of pharmacogenomics data that pointedly excluded proteomics data from the guidance entirely — a move that met with objections from some commenting on the guidance, including the NIH and GlaxoSmithKline (see SNPTech Pharmacogenomics Reporter 3-4-04).
At the IBC Biomarkers conference held last month, Steven Gutman, director of the Office of In Vitro Diagnostics at the FDA, said in a presentation that the FDA had received about “a dozen” comments on the multiplex guidance and that the OIVD was in the process of deciding how to edit and reissue it. He also said “future documents are being planned.” Joseph Hackett, associate director of the division of clinical laboratory devices at the FDA, said that this may include a separate guidance document for proteomics tests (see PM 2-13-04).
Although Gibbs suggested that much of the guidance directed toward microarrays could also apply to mass spec pattern tests, Timothy Alcorn of the regulatory consulting firm Biotech Consultant said that microarrays and mass-spec based tests were not completely analogous to one another, and separate standards may apply. However, it is not necessarily the FDA who will first define these, Alcorn said, pointing to the clinical laboratory standards organization NCCLS as a potential pioneer in the area.
In the meantime, Gutman said the FDA will ask questions “grounded in user need” when considering these types of new tests for approval — the same questions, he said, that someone “who might pay for the test would ask.” Although the data might be more complex than in the case of single analyte tests, “no matter how complex the data, the bottom line is ‘what does a positive or negative test mean?’” he said.
This isn’t your same old ELISA
But whether your test is a DNA microarray, a protein array, or a mass spec-based pattern test, complexity does matter to the FDA, according to Tom Collonna, manager of Biotech Consultant. “The whole issue is, these are multiplexed, so you have potentially thousands of variables to control in very small reaction vessels,” Collonna said. “The FDA hasn’t really approved anything like this yet.”
The issue is mainly one of comfort, Collonna said. The more complex, the more automated, the smaller the assay, the less comfortable the FDA feels about approving it. “In the ‘70s and ‘80s, they had the ELISA on these much larger 96-well plates, which they grew comfortable with,” he said. “But that was a whole generation ago.”
Also, the more an assay relies on a machine rather than on human intervention, the more wary the FDA becomes, Collonna said. This is particularly true, he said, when a test relies almost solely on a “secretive, fancy, proprietary algorithm” — as some might say Correlogic’s test does. While simple algorithms that basically speed up what has already been done by hand are generally not of concern to the FDA, for more complex algorithms, “the FDA wants to see that software — they don’t necessarily want to see every line of code, but they want to see how that software was programmed, and the different modules, and particularly how it was tested, and that it was validated and verified,” he said.
Medical devices — a category into which all of these tests could fall when undergoing premarket review — belong to one of three classes, depending on how much scrutiny that FDA thinks they deserve. Class I devices are not subject to review. Class II devices require 510(k) approval, which involves either showing substantial equivalence to an existing device, or arguing that a new device — while it may not be substantially equivalent to anything that came before it — shouldn’t be subject to the highest level of rigor. This latter type of 510(k) approval is called a de novo 510(k). Class III devices require premarket approval, or PMA. It is into this third class, Collonna said, that many multiplex assays would fall — rendering them subject to especially lengthy (180 days) and rigorous review.
Still, some groups have hopes that their tests will qualify for some sort of 510(k): Emanuel Petricoin and Lance Liotta have said that they are working with Gutman to seek this level of approval..
A Wolf in Sheep’s Clothing
Because multiplexed tests would be subject to such rigorous standards if reviewed through the official channels, “you have all these companies spending the better part of the last five years trying to get around the process by calling them home brews and selling them as an ASR,” Collonna said. An ASR, or analyte specific reagent, is a type of home brew technology. But it is only now, Collonna said, that the FDA is putting its foot down and demanding that devices that don’t fit the definition of an ASR seek official FDA approval. The agency did just this last October when it sent a letter to Roche, telling the company that its AmpliChip, a DNA microarray test, was not an ASR.
Indeed, Gutman warned at last months’ Biomarker conference that claiming a product as an ASR was “not entirely a get-out-of-jail-free card,” and that some companies were trying to pass off class III tests “in the lamb’s clothes of an ASR.”
Ciphergen recently said that it will market its own mass spec pattern-based ovarian cancer test first as an ASR, while seeking official FDA approval as a PMA or 510(k) (see PM 2-20-04).
But it might not make sense to call this type of test an ASR, Alcorn said, as mass spec tests themselves do not technically use any analyte specific reagents. “It’s simply separating out biological entities from the specimen and testing it in a mass spectrometer,” he said. “I think that if the test involves, for example, an antibody specific to something they wanted to run through the mass spec, that antibody would be an ASR. But [I] would be hard-pressed to identify outside of that what the ASR is.”
The FDA gives the following definition for an ASR: “Antibodies, specific receptor proteins, nucleic acid sequences, and similar biological reagents which through chemical binding or reaction with substances in a specimen are intended for identification and quantification of an individual chemical substance or ligand in biological specimens.” Gibbs said, however, that this definition should not be taken too literally, and that a mass spec-based pattern test “could well fall within the definition of an ASR.”
For Correlogic, though, an attempt to claim home brew status for its technology was “probably wishful thinking,” according to Collonna. The company’s software was probably complex enough to require FDA review as a device (see PM 2-27-04), and a home brew has to be developed in-house or at least commissioned only for a single laboratory. “When you’re a third party company licensing to multiple vendors, multiple users, that’s not home brew,” Collonna said.
This multiple-vendor licensing is what the FDA may have taken issue with: “The [OIVD] has reviewed reports indicating that Correlogic Systems is contemplating or has begun the commercial distribution to two major commercial laboratories of a test intended for use in detecting ovarian cancer,” the letter to Correlogic read.
But Correlogic’s “home brew” claim should be evaluated on exactly “how much was done by the lab and how much by a third party,” Gibbs said. Although the test itself needed to be conducted at the reference lab, there “could be components supplied by a third party, but assembled and validated by the lab,” he said. If it is really just the algorithm that Correlogic licensed, and not actual instruments or kits, “I think there is some precedent that that could be considered a home brew assay,” Alcorn added.
Correlogic would only say last week that it had licensed the pattern recognition process and “related technologies for use in development of an ovarian cancer test.”
What does it all mean?
It is still unclear how all these considerations might eventually translate for other companies who want to market a pattern test. By sticking its head out with early test release plans before the FDA issued any sort of formal guidance, Correlogic seems to be in the position of setting a precedent for everyone else. Still some issues may be Correlogic-specific, since its test deals with patterns of unknown biomarkers. “Here’s the big issue that has to be resolved by the FDA [in this case]: Will the FDA allow a diagnostic test to go forward where the specific analyte is unknown?” Alcorn asked. “I would be surprised if the FDA allows Correlogic to go forward with approval strictly on the pattern recognition.”
Other companies will be watching closely. “Until we understand the complete landscape, we frankly don’t know [how we will proceed],” said John Rakitan, senior vice president and general manager of Eclipse Diagnostics, a spin-off from Large Scale Biology that also hopes to market a pattern recognition-based ovarian cancer test. “We expect that the FDA will want to foster these sorts of things, but I have no idea what it all means,” he said.