NEW YORK (GenomeWeb) – Research enabled by GenomeDx's Decipher GRID data-sharing network continues to yield data on both new and existing gene expression signatures with prognostic or predictive value for prostate cancer patients.
The most recent insight, discussed earlier this month in a study in JAMA Oncology, speaks to the power of the open-access model GenomeDx has pursued since launching its Decipher prostate cancer prognostic four years ago.
In this case, using the company's GRID database, researchers were able to apply a version of the existing PAM50 gene expression-based classifier — commercially developed by NanoString Technologies as the Prosigna test in breast cancer — to nearly 3,800 prostate cancer samples, showing for the first time that the signature also shows promise in classifying prostate cancer cases in ways that could predict response to post-operative androgen deprivation therapy.
However, the results also highlight how the unique model of the GRID program presents challenges to traditional diagnostic development models, and raises questions about the barriers between research and clinical medical practice.
Decipher GRID (which stands for Genomics Resource Information Database), is made possible by the fact that GenomeDx runs its clinical Decipher prognostic test using whole-genome microarrays. Though the company only reports gene-expression results that make up the Decipher assay as part of its clinical report, it nevertheless collects the full array data for every patient it analyzes.
Through GRID, researchers can apply to access that growing database of array results alongside data on patients' clinical characteristics and outcomes, allowing retrospective analyses of potentially any existing or novel expression signature.
All physicians who order a Decipher test can access the Decipher GRID data for their own patients on a research-use basis. Those who want to use more of the resource must sign user agreements, though access is free for all non-commercial investigations.
With new insights continuing to emerge from GRID research, GenomeDx is also continuing to update the GRID report it returns to physicians. Though the information is designated as research-use only, doctors can review how their individual patient scores on a variety of different signatures, even, in the case of PAM50, signatures that have been commercialized by other companies.
In practice, GenomeDx returns a clinical test report for only its Decipher prognostic test, but it also outlines a wealth of other information that has not yet met the threshold for clinical reporting but could be compelling. This includes new signatures that have come out of research using the larger GRID database. For example, a team from the University of Michigan and the University of California, San Francisco published last year on a signature they dubbed PORTOS (Post-Operative Radiation Therapy Outcomes Score) that is predictive of prostate cancer patients' response to post-surgical radiation therapy.
The GRID research-use report also includes scores reflecting a variety of existing classifiers that have been published in the literature — not just PAM50, but also scores based on the collection of genes used in commercial prostate cancer risk tests from other companies like Myriad Genetics and Genomic Health.
Anyone can go and read a peer-reviewed publication and analyze a list of genes," GenomeDx President and Chief Scientific Officer Elai Davicioni said in an interview. "This isn't any different. It's the markers from Prolaris or from [Oncotype DX] but it's not their coefficients or their proprietary methods."
Physicians have always been able to research and consider investigational information in the context of deciding how to treat their patients. But instead of having to search literature, the Decipher GRID report allows doctors to examine how their patient scores on various investigational signatures in a collated and clearly presented report.
This stands in contrast to traditional diagnostic development paradigms in which companies work to develop and incrementally validate new genomic markers or signatures behind the scenes, in many cases only publicizing them when they are clinically validated or close to being so.
"I think what it does is set the stage for clinical analysis. It allows us to begin to test these things prospectively in trials, but in terms of acting on the data, I don't think we [should] in terms of standard of care decision-making for patients," Jefferson University physician Robert Den, an early adopter of Decipher, said this week.
"We use Decipher as its been clinically validated, but some of the other signatures begin to provide us some insight in terms of how patients respond and what their scores spoke to when we go back and analyze them retrospectively," he explained.
Johns Hopkins physician Ashley Ross described the utility of GRID in slightly grayer terms. For example, he said, the GRID scores, though research-use-only, can potentially help solidify a therapeutic or other clinical decision in cases where other factors, like the Decipher prognostic score, conflict with various clinical assessments such as tumor stage, grade, and Gleason score.
"Sometimes a patient has aggressive pathology: Gleason 9, T3A, an older individual, but the Decipher score is really low … One thing GRID has helped me with is, [I can] look at all the other prognostic signatures and see if all of them are tracking in the same way or if some of them are off," Ross said.
Den agreed, to a certain extent. "Maybe if you were on the fence it may assist in thinking about the right way to go, especially in a situation with real clinical equipoise," he said. "If there isn't level one evidence to go one way or another it may tip your hat a little bit."
Physicians are not robots that weigh only the most solidly validated and accepted evidence according to perfect algorithms. They are human beings who rely on what Ross called "clinical acumen" — the knowledge and insight that allows decision-making on complex issues, where information might be incomplete or unknown.
"It's hard to see something that you know is good data with something like PAM50 or the PORTOS score, and not be influenced, at least subconsciously. You'll go into that conversation with your patient with more confidence," Ross said.
"It's always going to be that the technology is ahead of level-one evidence," he added. "In my practice, for example, we use molecular imaging tools. We are still doing the studies that establish the sensitivity and specificity of these tracers but I still use them to guide treatment decisions."
GenomeDx's Davicioni concurred. "Within certain limits, physicians can use any type of information. They can flip a coin or look at a magic 8 ball … so for us it's about physician education and making clear that this isn't validated yet, but also listening to these collaborators and working to validate the things that they see a need for as quickly as possible things so that they actually can use it in clinical decision-making," he said.
Regardless of the gray areas created, Ross said that the model GenomeDx has developed by coupling a dynamic and expansive research-use resource with its clinical diagnostic test is immensely valuable for accelerating clinical research.
"For a long time we didn't know how Gleason score, grade, margins, etc. affected outcomes … and the only way to eventually develop pathologic or clinical nomograms was collecting that data and looking back retrospectively to see what happened," Ross explained. "So having these signatures on the GRID facilitates that greatly."
Previously, he added, researchers might publish a signature in the literature and that would be the end of it. "But with getting this report, strictly on the research side, you now have these numeric scores and you can put them into your database and in a few years, we can do retrospective analysis. That's really what this system is made for," Ross said.
Ross said that in his individual opinion, one more independent validation of the PAM50-based subtyping or of the PORTOS signature should be enough to merit consideration by GenomeDx toward adding these signatures to its clinical test report, or for enterprising physicians to consider them in a more formal way even if they remain research-use-only.
According to Davicioni, the company has seen a lot of progress in taking these next steps just in the last few months. "There will be new stuff on the Decipher clinical report even sooner than I think I would have thought a year ago," he said.
In the mean time, GenomeDx has not put an additional charge in place for the Decipher GRID research-use scores. Patients are charged for the Decipher prognostic signature, but if physicians choose to access the additional GRID information they don't pay any extra.
Davicioni did not comment on how pricing might change in the future with additional analyses added to the clinical test report.