Celera Diagnostics’ head of discovery research talks about FDA leading the charge, the need for biomarker validation, and why the conversation should be about targeted — but not personalized — medicine.
As vice president of discovery research at Celera Diagnostics, John Sninsky has a bird’s eye view of how biomarker-based molecular diagnostics are faring in these early days. GT’s Jeanene Swanson caught up with Sninsky for a chat about biomarker validation, getting diagnostics through FDA and to physicians, and more.
Genome Technology: Looking at the molecular diagnostics market, how do you expect these new tests to get to the consumer or physician — through traditional diagnostic labs, clinical labs at hospitals, direct from vendors, or some other route?
John Sninsky: We don’t expect the natural progression or the availability of diagnostic tests to change substantially from the way it’s being done now. The natural progression of diagnostics is from large clinical reference laboratories as laboratory-developed assays, or sometimes referred to as home-brews, and then to evolve into regulatory approval of in vitro diagnostic products to the laboratories with lower throughput. Sometimes these laboratory-developed assays are provided by smaller service laboratories when they get launched. And then the offering as a service really provides accelerated availability due to the timing that’s required for product development, manufacturing, and registration. So a service gets the information out there sooner, but the advantage of the in vitro diagnostic product is that it’s more widespread in its availability, it’s improved standardization as you compare information coming from one laboratory to another, and lastly, it usually proves to be more cost effective because there are a larger number of people carrying out the test.
With the availability of the Internet and with the strength of advocacy [groups], rather than the patients or the clinicians being passive participants … my expectation is that patients will be better informed in terms of where things are and what can be done, and they will be true participants in the decision.
GT: A major challenge with any of these diagnostics is validating the biomarkers. In your experience, is there a class of biomarkers that looks most promising?
Sninsky: Validation is one of the key elements of biomarker acceptance. It used to be there were scientific papers with relatively small numbers of patients. So it’s difficult to extrapolate how something’s going to perform in a large dataset and to make sure that what is observed in the smaller set isn’t in some way biased. Replication is really key.
[Another] important element that we need to ensure is whether or not the biomarker actually addresses an unmet diagnostic need. Sometimes there’s science that gets done; it’s really interesting, but it doesn’t impact a routine clinical question. It’s also important that it be actionable. It’s fine to take a piece of information and say, well, that’s really interesting, but what does a person do differently — whether it be the patient or the clinician — is really important.
And then lastly, an important analysis is sometimes referred to as the multivariate analysis, and what a statistician means by that is whether or not the information is redundant with the information that’s already being used [in] conventional markers or diagnostics, or whether or not it’s providing new information.
We have work being done in proteins, in mRNA, and in DNA in gene variance, and we hope they all look promising. We don’t think that any one of those analytes have a corner on the market. We are a player in all three of those and they all look promising to us.
GT: From what we’ve heard, scientists in clinical labs seem reluctant to embrace this new breed of molecular diagnostics. They say they want a lot more data before they’ll move forward with it. Is that a trend you’re seeing? If so, what’s the strategy to get those labs to buy into these tests?
Sninsky: What’s most important these days is translational medicine. Translational medicine really is where unproven observations advance to increasingly large and more informative studies to provide definitive and comparative information for a medical decision in routine clinical practice. Everyone who contributes along that process has to weigh and balance the information so that the new information will not be used prematurely but at the same time, equally importantly, what [is] the delay if that information is actually useful for medical decisions. And all medical decisions need to weigh the level and the nature [of] risk of not using that information with the level and nature of risk [of] using that information. So in the end, no one piece of information, whether it be the conventional diagnostic test or the new test, will suffice for medical decisions but will require gathering and reflecting on as much of the information as possible.
What our experience has been is it’s not the clinical labs who will decide whether new biomarkers are used, but instead will be the practicing clinicians, the reimbursement agencies, and the patients who see value in them that will make those decisions. But I will credit the clinical labs that run these tests [with] highlighting the importance of replication and interaction analysis to make sure that the information isn’t redundant with what we’re already collecting. One of the things that I think is really good news is the stand that the US regulatory agency has taken in terms of saying we’re going to help this process along. If someone would have said to me 15 years ago that the FDA was going to be leading the charge of the application of biomarkers, I would have been pessimistic about that.
GT: How have FDA’s guidelines affected the molecular diagnostics field? Is it more challenging to get a diagnostic approved now than it was a few years ago?
Sninsky: We think there are two changes in diagnostics. One is that the FDA has created a separate and distinct office for what’s been referred to as companion pharmacogenomic tests, where it’s a combination of a drug and a diagnostic to identify segments of the population who are particularly good at or not as good at responding to a drug. And that office has begun to approve tests. An area that will merit watching carefully is the criteria established to identify the appropriate drug label. There are three categories that the FDA has identified: the test is going to be required, the test is recommended, or it’s only being used for information on the drug label. And right now there’s not a clear set of thresholds for criteria that puts any diagnostic test in one of those three bins. But I think people are going to have to develop that in order to better understand how those categorizations get put in place.
And lastly, the regulatory draft guidance [has] already appeared with multivariate index assays, and the FDA has decided to have oversight for complex tests that involve multiple analytes in sophisticated laboratories. So they’re going to require 510(k) clearance rather than simply being carried out as a laboratory-developed or home-brew test in the oversight of CLIA.
GT: Some people say that truly personalized medicine is a realistic goal, and others say the best we can hope for is medicine that stratifies a population into a few broad segments. What’s your view?
Sninsky: If you identify personalized medicine as information that’s truly unique to an individual, we actually think that information applicable to segments of the population will prove more timely and more important, and we think that maybe calling it targeted medicine rather than personalized medicine makes the case in point. The validation and demonstration of utility really only comes with replication with large, well-characterized studies. How you gather that data on an individual is less clear. How you gather that on a segment of the population, whether it be five percent, 10 percent, or 50 percent, is much more straightforward. So we think in order to be most effective, the information applicable would usually address about 10 percent of the population. So disease management, or health management, in the context of targeted medicine is the way we think about it.