July 1, 2013 (Vol. 33, No. 13)
Demonstrating clinical utility is critical for moving tests out of research-use only mode, according to this article from our July 1 issue.
On the first of this year, clinical laboratories began billing for more than 100 molecular pathology tests using new CPT codes, which replaced the previous stacked-code system.
The transition has already been rocky—the U.S. Centers for Medicare & Medicaid Services (CMS) has asked local administrative contractors to develop their own payment rates. The change has also stirred up concern over the fate of molecular testing in the clinical setting as R&D initiatives press on in the pursuit of personalized medicine.
“It is very exciting to see how much progress we are making so, in the future, we can end up with better patient care. To make these tests available to a wider audience they need to move from a Research Use Only (RUO) status to a clinical one,” says Olive Joy Wolfe, president, Clinical Consultants, of the speakers at CHI’s recent “Biomarkers and Diagnostics World Congress”.
“This may require collaboration across laboratories, working with groups like the NCI [National Cancer Institute], and working closely with the FDA early on to facilitate the process. The first question is defining the use of the test. Is it for diagnostics, therapy choices, patient intervention, or something else?”
Biomarker assays can be multiplexed and performed on complex platforms and instrumentation that the average laboratory technician may be unfamiliar with. In addition, if the assay is multistep, it may require a combination of instruments and techniques.
As MD Anderson Cancer Center’s Samir Hanash, Ph.D., notes, in a discovery mode, the sensitivity that can be achieved for cancer protein biomarkers by mass spectrometry is quite deep. The challenge is to come up with implementable assays that achieve the same level of sensitivity.
“You need samples first to test the process, then if you are taking it to the next level and looking at further development for determining biomarker values in the clinical setting you have to go through validation, just as for any other assay under the CLIA rules,” adds Wolfe.
“It is not as simple as it was before, but there are guidelines. It will not be as foreign to researchers who went through several of these necessary steps when we first came up with other methodologies, such as the RIA [radioimmunoassay], because they are aware of the work involved,” she says. “Yet, it is more complicated because we may be looking at multiplex proteomic-based tests.”
Demonstrating Clinical Utility
Simply put, the question of clinical utility asks: Does the test actually work and provide value to the patient, the physician, and the system?
As the number of biomarker assays available for the study and evaluation of human disease increases so does the need to understand their clinical utility.
“If you are applying a biomarker to determine if a patient has a better outcome, less toxicity, or a better quality of life, that is different than asking about the observed biological phenomena, the pathways, and mechanisms of action,” explains Steven Gutman, M.D., strategic advisor at Myraqa. “They are both valuable pieces of information, they just play in different arenas.”
The FDA looks for a circumspect demonstration that the product is either substantially equivalent to a predicate for products requiring 510(k) review, or safe and effective for products reviewed as a premarket approval (PMA) application.
“A company does not need to establish clinical utility to market a test. Insurance coverage decisions and physician test usage drive clinical utility. Both are more likely to respond to evidence that a test works and that it does not just generate a number,” Gutman adds.
In the past, less attention was paid to clinical utility; the recent reimbursement changes have brought this issue to the forefront.
“There is unfamiliarity with the language of design of clinical utility and a lot of hard questions on the table,” he says. “It is a question of will. We do not want regulation to block but, rather, to foster clinical utility. It is a messy complicated field.”
“Multiple professional groups are now worried about clinical utility. It is conscious and visible. If you raise the standard for payment but assure tests with demonstrated clinical utility are fairly reimbursed, there is potential that things could work out very well,” Gutman says.
From an analytical perspective, a robust biomarker assay can accurately differentiate a positive signal from a negative one time and again.
“However, a successful biomarker relies not only on a robust assay that can be implemented easily, but also, more importantly, on the underlying biological hypothesis,” says Tammie C. Yeh, Ph.D., oncology biomarkers team lead at Merck. “The most notable success stories recently are the identification of BRAF mutations in melanoma for vemurafenib and the identification of ALK translocations for crizotinib.”
The clinical setting incurs additional challenges—specimen variability, equipment limitations, varying levels of technical expertise, and laboratory personnel availability. Some clinical assay decisions are based on logistical feasibility, not strictly science. Keeping sample collection and processing as simple as possible is key to high data quality, especially when working across multiple clinical sites.
For example, developing a patient-selection biomarker is a huge undertaking. The principal challenge is identifying the right biomarker to justify the resources and risk. Other requirements include identifying a CLIA laboratory for prospective patient enrollment, establishing that the biomarker is predictive and not prognostic, and co-developing a companion diagnostic, which brings additional regulatory hurdles.
For patient-selection biomarkers, the final assay platform ought to be compatible with FFPE-processed tumor tissue simply because that is what is most often available.
Contrast this with pharmacodynamics (PD) biomarkers, which are used to establish that the compound is hitting its biological target. A variety of assay platforms can be used; tumor biopsies can be fresh, frozen, or FFPE-processed, or surrogate tissue can be used. The challenge in this case is knowing which combination of tissue/readout/assay platform to use, so the expected biological effect significantly overcomes the expected intradonor variability arising from both technical and biological sources.
“Next-gen sequencing is already starting to have an impact in the medical field and will continue to do so. The immediate challenge is to ensure consistency of data quality and interpretation, as well as alignment with regulatory agencies and payers,” Dr. Yeh says.
“As biomarker-driven therapeutics increase, there will not be enough tumor biopsy tissue to support independent screening, let alone the costs associated with each individual test,” she adds.
Another topic of interest is the use of laboratory-developed tests (LDTs).
“Labs will use LTDs, particularly when developing companion diagnostics,” says John L. Allinson, vp, biomarker laboratory services, ICON Development Solutions.
“If you look at the diagnostics we have today, at some point they were tests developed by a laboratory and not available commercially. Each individual laboratory would validate those assays for the appropriate clinical use,” Allinson says. “If you develop an in-house method because of lack of a platform to use a commercial assay, or you believe your assay is better, you still have to validate to the same degree of rigor as using an accredited diagnostic.”
Companion diagnostics can be used to identify populations that will respond to a particular drug. In the early stages of clinical trials, they begin as a research assay used to assess certain biomarkers.
Typically, if initial results are promising, collaborations are formed with a diagnostic company to develop the final test and platform. Working together to show clinical utility in trials, the drug-development company files the drug submission and, simultaneously, the diagnostic company files the PMA application, each depending on its own data.
In the past, the focus has often been on single surrogate-endpoint biomarker discovery, biomarkers that substitute for the clinical endpoint. Surrogate-endpoint biomarkers are specific to one clinical condition—only about half a dozen exist to date.
More recently, biomarker panels, which can improve specificity and sensitivity of diagnostic determinations, are being investigated. The most popular biomarkers measured today, excluding surrogate markers, provide only a part of a clinical picture, and can show changes in multiple pathologies.
“Technology is going to drive the improvements that we see over the next few years, for instance, novel technologies that are smaller, and have the ability to measure relatively large numbers of different biomarkers, on very small samples,” Allinson says. “New technology adoption needs to be encouraged. It has real potential for positive impacts in the industry.”
Accounting for Variability
Biomarker analysis is a common practice in pharmacokinetic/pharmacodynamics (PK/PD) modeling to learn about a drug’s mechanism of action.
Biomarker assays and kits can be 510(k)-approved, lab-developed, or RUO kits. Several venders produce RUO kits. Multiple lots are generated yearly, with a shelf life less than a year. Since clinical trials can extend for years, a system is needed to link data generated from one kit lot to another.
Each kit may contain several key components, and testing each individual reagent would be unfeasible. To address RUO kit variability, BioAgilytix Labs has implemented the concept of lot bridging.
To quantify the kit as a unit to assure performance consistency, a series of stability QCs and fresh samples are tested over multiple days, both on the previous and new lots. A correction factor—a numerical multiplier—is generated and used to normalize the new lot with the validation, or a previous lot. Lot bridging reduces noise in the resulting data, making it more meaningful for statisticians and clinicians to interpret.
“Even if you have a beautifully validated assay, lot-to-kit-lot variability influences results and has to be accounted for. Kit variability has nothing to do with the drug effectiveness and may point you in the wrong direction if quality data is not generated,” says Afshin Safavi, Ph.D., svp, bioanalytical operations, BioAgilytix Labs. “When you are supporting PK studies conducted under GLP, with specific regulatory requirements and guidelines, typically the drug acts as the calibrator or standard. The assays can be set up in a way that removes, or minimizes, the critical reagents lot-to-lot variability.”
“Much biomarker work is secondary, to learn about drug mechanism of action. Calibrators and controls may, or may not, be available commercially, and may come in a variety of forms. For most biomarkers, there is no universal standard available,” Safavi adds. “Therefore, the concept of ‘fit for purpose’ is used for biomarker assay validation and analysis support.”
“At BioAgilytix Labs, we validate the assays, treating them as if they were almost a regulated study with the proper documentation. This is how we discovered the extent of biomarker kit lot variability in the market today,” he explains. “Lot bridging becomes especially important if you want to compare studies, perform multiyear studies or add more cohorts to ongoing studies, which requires the purchase of additional kits from a different lot.”