By Deborah Borfitz
September 28, 2020 | Last month’s Next Generation Dx Summit brought together leading experts on the role of companion diagnostics (CDx) in precision medicine, including Wendy Rubinstein, M.D., Ph.D., director of personalized medicine at the U.S. Food and Drug Administration (FDA) in the Center for Devices and Radiological Health (CDRH). She covered the use of routinely collected real-world data (RWD) to support regulatory submissions, focusing on in vitro diagnostic devices (IVDs), and later joined a panel session on the current state of CDx in and beyond oncology.
In a regulatory context, RWD is intended to overcome the limitations of randomized controlled trials—notably, the time and expense in conducting them, the difficulty in collecting rare outcomes and the generalizability of results, says Rubinstein. Benefits and risks of a device can be better inferred in real-world environments, and an analysis of that information qualifies as valid scientific evidence provided it comes from “well-controlled investigations of significant human experiences.”
Formal guidance on the use of real-world evidence (RWE) in regulatory submissions was finalized in 2017, she notes. And just last month, the Medical Device Innovation Consortium (MDIC) issued its framework for incorporating RWD and RWE into pre- and post-marketing regulatory decision-making for IVDs.
For pre-marketing submissions, leftover (aka surrogate) samples whose clinical status is known may be used with the candidate test to determine test performance, she cites as an example. IVD clinical performance studies involving RWD are often non-observational because unbiased estimation of clinical performance requires a study design that includes collection of additional samples. For an observational (“virtual”) clinical performance study, patients might be tested with a candidate test followed by a comparator test (e.g., gold standard) according to routine clinical practice.
Over 140 Emergency Use Authorizations (EUAs) for molecular diagnostic tests have been issued by the FDA since the start of the COVID-19 pandemic, plus a fair number of serology tests, Rubinstein continues. These are largely collaborative activities that are leveraging RWD to demonstrate clinical performance.
The newly launched COVID Shield initiative, targeting lab data disharmony and portability issues specific to SARS-CoV-2, should provide a workable solution for “the rest of medicine,” says Rubinstein.
To conclude her presentation, Rubinstein cites several examples of IVD regulatory assessments that have used RWE. These include Illumina MiSeqDx CF products, designed to detect genetic changes associated with cystic fibrosis (CF), whose clinical validity was supported by evidence of variants detected from dried blood spot specimens collected through routine newborn CF screening.
Similarly, Memorial Sloan-Kettering Cancer Center submitted RWE to support market authorization of MSK-IMPACT (Integrated Mutation Profiling of Actionable Cancer Targets), a next generation sequencing-based tumor profiling test, she says. It uses a publicly available, precision oncology knowledge base, OncoKB, to facilitate the clinical interpretation of detected mutations.
Importance Of Early Planning
The “real question” with RWE, says Rubinstein, is this: Is the research question supported by the data at hand? She advises sponsors to get the FDA’s feedback through its Q-Submission Program, which includes pre-submissions as well as additional opportunities to engage with the agency. Only a “handful” of IVDs have made premarket submissions supported by RWD, she adds, “so we’re still learning.”
Rubinstein was joined in the panel session by Mark Curren, vice president of companion diagnostics at Janssen R&D; Neeraj Adya, Ph.D., head of Genmab US, and Jason Christiansen, chief technology officer for Boundless Bio. Renee Yura, director of diagnostics at Pfizer, served as moderator.
One of the biggest lessons learned in the oncology space is the importance of timeline alignment “early on” if a drug product will be paired with a companion diagnostic, says Christiansen. The idea is to avoid unwelcome “surprises”—all too commonly, translational work that doesn’t translate well into clinical trials.
Knowing the number and types of patients and what kind of samples are needed, and how that translates into actual clinical use, is key to designing validation experiments, Christiansen says. Initially, “cast a wide net” when it comes to selecting biomarkers, he adds. The field can be narrowed when the clinical strategy is being formulated. “Understand what it is you’re validating, what is… going to drive the safety and efficacy of the test you’re building, and how that meshes with all the other data you want versus what you need.”
Familiarity with the complex regulatory process for companion diagnostics is lacking outside of oncology, says Rubinstein. “Tap the knowledge of in-house oncology experts,” she suggests.
Nailing The Markers
The ability to identify highly predictive, accurate markers that can be replicated has made it difficult to find the right areas to study, says Curren. People get identified as either permanent non-responders, or respond one time and not another, limiting the size of cohorts for validation studies. The final cohort ends up with errors in phenotype. “It’s tantalizing biology, but it’s not clinically relevant and useful.”
To get around these phenotype variances, great pains have been taken at Janssen to develop the right cohorts, continues Curren. Since the company can’t do big enough clinical studies required to “nail the markers,” it has moved into the consortia space.
Curren’s prediction is that significant breakthroughs will occur in three areas—including IVDs, where a robust biosignature (gene expression) in tissue can identify patients who will have a poor response to any of the available mechanisms. “The key is to move from biopsy-driven tests into peripheral blood tests or other easy-to-get samples and start using that in trials and in care.”
In Europe, rheumatoid arthritis work with joint tissue has uncovered some markers that are now being validated by a consortium formed through the Innovative Medicines Institute (IMI), he adds. Scientists are also closing in on the different cells types involved in lupus, most recently cytotoxic-type signatures of the disease.
In immunology—his longtime area of expertise—there has also been breakthroughs with new mechanisms and a movement from biologics to orals, says Curren. “The community has made significant progress and we’re also starting to see real interesting combination approaches. I think we’re going to circle around and see biomarkers and diagnostics become even more important, so we don’t keep cycling through monotherapies that don’t work.”
In terms of a diagnostic development process for IVDs, whether in or outside oncology, Adya points to an FDA diagram around design control that begins with identifying user needs and translating those into design inputs for verification and validation testing. As is now broadly recognized, precision medicine can help competitively differentiate an assay from others on the market.
CDx, even in the oncology space, has started to move beyond the single biomarker stage, says Adya. The FDA has approved a spectrum of gene expression profile assays for breast cancer, but up to now these have mostly been prognostic tests approved through the 510(k) premarket submission process.
Composite Biomarkers
To win marketing approval for a complex biomarker designed to serve as a companion or complimentary diagnostic requires further segmenting patient populations (e.g., immune-inflamed vs. immune-desert tumor, or subtypes of the former) to isolate the unmet need population, Adya says. In conjunction with digital pathology, gene expression profiling can identify the right patient population that has the particular biology being addressed. “Single biomarkers are not going to be effective.”
The FDA has published helpful guidance specific to IVD multivariate index assays, Adya notes. Examples in this category include Cologuard (colorectal cancer screening test) and some human papillomavirus assays that generate a composite score from various biomarkers, which have been approved under the agency’s more rigorous premarket approval process for class III medical devices.
On the autoimmune side, Adya continues, interest in interferon genes is high. AstraZeneca, for instance, recently published on its four-protein score of type 1 interferon activity. The complexity in this realm is that patients are often already on steroids, which can impact their immune status and make it difficult to define a baseline for dosing and how they might respond to a particular drug. A “much broader” assay may therefore be required that represents both steroid and interferon signatures.
Partnering Up
To ensure they position on the right platform and technology, Adya advises companies to develop their diagnostics strategy ahead of time, in conjunction with their clinical and biomarker strategy, to avoid the need to do bridging studies or change cut-off values in the move from one platform to another. “It’s imperative to the success of the drug to be launched.”
In oncology, the window of opportunity is tighter for ensuring clinical trials have the right patients and cutoffs, and “tissue is an issue,” he adds. Blood or plasma is more often the sample type in other therapeutic areas, including lupus, and endpoints tend to take much longer to assess.
Those early engagement efforts should include clinicians who may be unaccustomed to doing personalized testing, says Christiansen, to identify potential barriers to entry.
“Tissue is an issue for us,” Curren agrees, in terms of both the amount available and the heterogeneity of the samples collected. Clinical studies need to be well planned out starting at the preclinical stage or “a rescue strategy will not work, and you will never catch up.”
Part of the issue, says Rubinstein, are subpopulations that may have “a phenotype” and, in oncology especially, the fact that testing gets done at the local level. “How does that translate into getting a diagnostic done later on? The only way to solve these issues is to head them off at the pass, to start planning from the beginning.”
How partnerships between pharma companies might reduce complexities, and bring some cohesiveness to the CDx development process, was the focus of Adya. Already, multiple companies (including Merck and Roche) have developed gene signature assays around PD-L1 with different cutoffs and scoring algorithms, and he predicts the same phenomenon will occur with tumor mutational burden. This is going to hamper standardization efforts across signatures and indications, he says.
Janssen’s practice has been to set up consortiums to answer specific questions, says Curren, noting the lessons of EGFR (lung cancer biomarker) and KRAS (for colorectal cancer) have needlessly been learned again and again. A decade ago, it was part of the BATTER-UP (Biomarkers of Anti-TNF Therapy Efficacy in Rheumatoid Arthritis to Define Unresponsive Patients) Consortium that brough together a half dozen biotech and pharma companies to try (ultimately unsuccessfully) to validate the growing number of markers appearing in the literature.
Janssen has subsequently been a part of other consortiums through the public-private Accelerating Medicines Partnership (U.S.) and IMI (EU), Curren says. “You can shape it if you participate in it.” That means contributing scientists, samples and data—not just writing a check.
Moving Toward Standardization
It’s critical to nail down some of the pre-analytics early on, says Adya, including variances that might be attributed to a particular assay. For gene expression profiling, especially in oncology, a common definition for the invasive resection margins would make results much more standardized and comparable.
In all likelihood, diagnostic tests available on a global scale will all end up being run on the same platforms, says Curren. Higher variability in test results will be a factor primarily in settings where non-certified labs are developing their own assays.
Labs regulated by the Clinical Laboratory Improvement Amendments program is “where all this will end up,” says Christiansen, suggesting the consortium idea be extended to standard sample sets. “It may not be a perfect bridge, but… then you don’t have to worry about pharma having to agree with each other.”
Large payer networks may well get together to produce standards, says Curren. Composite biomarkers of the future could also shift some of the focus away from molecular biology and toward digital health. A point-of-care test would be “much more impactful,” interjects Adja.
Rubinstein concluded the panel session with a few words on the merits of standardization. In oncology, the MDIC has supported the cause with somatic reference samples supporting the engineering of actual samples with specific variations through techniques such as CRISPR, she says. Of additional note is that liquid biopsy and pharmacogenetics communities are already collaborating pre-competitively, giving a broad set of stakeholders a voice on certain thematic topics. While not run by the FDA, the agency’s website has information on how to get involved in these collaborative efforts.