Clinical Trials For Companion Diagnostic Approval

By Deborah Borfitz 

September 2, 2020 | The regulatory framework around drug and diagnostic co-development, and how to overcome the inevitable hurdles, was the subject of multiple presentations at the 12th Annual Next Generation Dx Summit held virtually last week. Representatives from pharma and FDA weighed in on the clinical trials required to get companion diagnostics to market. 

It was a logical point of focus, given that personalized medicines have accounted for more than 20% of all new molecular entities approved by the U.S. Food and Drug Administration (FDA) for the past six years and the size of the global biomarker market is projected to double by 2026, according to speaker Eunice Lee, executive director for companion diagnostics (CDx) and in vitro diagnostics (IVD) in the global regulatory affairs group at Merck. 

Advances in technology, clinical screening, and high-throughput sequencing have created a treasure trove of genomic data that could revolutionize precision medicine, Lee says. Biomarkers can play a starring role in guiding decision-making and tailoring disease treatments to individual patients, provided they’re both highly accurate and clinically actionable.   

The FDA’s regulatory framework for companion diagnostics was finalized in 2014, Lee notes. Approval policies stipulate that the therapeutic product and diagnostic test should cross-reference each other in labeling and that the two products be approved simultaneously, although there are exceptions for drugs that treat a life-threatening condition. 

Notable approval “firsts” for the FDA include CDx (1998), real-time polymerase chain reaction test (2011), and next-generation sequencing (2017), and across the board oncology has been the focus, says Lee. Movement on the CDx front began with HerecepTest—a single test tied to a single drug and indication, which quickly gave way to multiples of each. 

The increasing complexity, especially for molecular genetic tests, pushed regulatory innovation, continues Lee. In 2015, FDA regulatory status extended to complementary diagnostics, referring to tests used to improve disease management, early diagnosis, patient risk stratification, and drug monitoring. Up until then, the term referred to tests used to improve disease management and they didn’t require a regulatory link to a specific therapeutic. 

The development pathway for complementary diagnostics and companion diagnostics are distinct but similar, says Lee, and ideally happen concurrently with a drug approval—meaning, lead candidate and biomarker selection happens at the same time, as does investigational new drug (IND) and investigational device exemption studies, regulatory submissions, market launch, and post-surveillance studies. 

In reality, though, they often are not aligned, she says. The drug development process may be accelerated so that the assay used in clinical trials differs from the test ultimately approved as the associated CDx. Additional time is needed to select a companion diagnostic partner and, if needed, do clinical validity and utility testing. 

On a few occasions, the FDA has approved a drug with the pharmaceutical company making a post-marketing commitment to develop a companion diagnostic, Lee says. But that doesn’t guarantee approval of the CDx. “Best practice is to develop a strategy early in drug development, work together and communicate. A strategy beyond the U.S. is also important,” she adds, noting that Australia, the EU and China all have their own regulatory policies for companion diagnostics. 

Clinical Validation 

Analytical and especially clinical validation of CDx is the specialty of conference speaker Reena Philip, director of the FDA’s division of molecular genetics and pathology. Both are areas where companies need more education and better pre-planning, she says. 

Key analytical validation studies include sensitivity, specificity, precision, accuracy (“use clinical trials samples for original CDx”), and robustness/stability, says Philip, and should “as far as possible” utilize the intended use specimen type. Studies should include samples around the clinical decision point. For rare biomarkers, contrived samples with functional comparability to clinical samples (cell lines) can be used as a supplement, she adds. 

Most FDA-approved companion diagnostics have a selection claim measuring the marker because marker-positive trials of a specific therapy were conducted, and the test identifies a subset for whom the therapy has a favorable benefit-risk ratio, she continues. Clinical validation of the device is supported by results of the therapeutic clinical trial. 

When clinical trial assays are used for study enrollment, Philip says, a single, analytically validated test should be used either to screen all patients (recommended) or as a central confirmation test for other enrollment tests. If available, a cleared or approved IVD should be used; otherwise, the enrollment test(s) should be analytically validated. “A minimum standard for analytical performance at local sites will support a successful companion diagnostic.” 

Keep in mind, Philip says, that “use of a variety of clinical trial assays [CTAs] with different performance characteristics can result in unique intent-to-diagnose populations at each test site.” 

For rare biomarkers, several local tests may be used for enrollment, she says, with the following caveats: drug developers need to provide minimum analytical validation requirements to enrollment sites and should only enroll patients from sites that meet the acceptance criteria; all sites should follow the same protocol for specimen collection, banking and processing; each assay should be fully specified and cutoffs established and locked down prior to registrational study; information on the assays (e.g., test methodology/name, specimen type used) should be available to evaluate if test results are comparable and are identifying the same group of patients that are receiving the drug; and every effort should be made to bank all marker positive samples and sufficient representative marker negative patients for a bridging study. 

Bridging Studies 

Ideally, the drug and companion diagnostic should be validated in the same clinical trial, Philip says. Otherwise, the agency would like to see a bridging study to demonstrate the efficacy observed with the CTA is maintained in the CDx and the efficacy for CDx-positive subjects is statistically and clinically significant. All CTA marker-positives and sufficient representative CTA marker-negatives should be retested using the final CDx assay.  

Within a bridging study, the primary sensitivity analysis uses a formula that calculates the efficacy for double positives and the unenrolled subgroup, factoring in positive percent agreement (PPA), negative percent agreement (NPA) and prevalence, says Philip. In the worst-case scenario, the efficacy for the unenrolled subgroup is the same as the control group observed efficacy and, in the best-case scenario, the assumption for the unenrolled subgroup is the same as the double positive group observed efficacy. For samples with missing CDx results, this primary sensitivity analysis should be done to evaluate the impact on the efficacy calculation and study conclusion. 

The agency has noted a trend among pharmaceutical companies and diagnostic partners of not having enough test-positive or test-negative cases for clinical validation of a companion diagnostic, she continues. Sponsors may therefore need to seek additional evidence to support CDx effectiveness, which may delay approval timelines. 

“The higher the missing rate, the more difficult it is to justify the comparability between the CDx analysis dataset and the CTA analysis dataset,” says Philip. “In addition, the available samples may not be representative of the clinical trial samples. A high missing data rate can be a serious problem that undermines the scientific credibility of causal conclusions from clinical trials.”  

Sample ascertainment rates seem to be poor when local laboratory tests (LLTs) are used as CTAs to enroll subjects, she notes. Sponsors should be sure to obtain informed consent to allow retesting and retain all screened samples (positive and negatives) in sufficient volume for bridging studies. When it is not feasible to retain LLT-negative samples, LLT samples may be procured. Preferably, these would be screen-failed samples from clinical trials with the same eligibility criteria. If not, sponsors will need to demonstrate that the procured negative samples are representative of the LLT screen-failed population. 

Sponsors should in all such cases communicate with the FDA’s Center for Devices and Radiological Health (CDRH) and get pre-agreement about their bridging strategies, Philip says. 

High concordance between the CDx and CTA is expected. When the CDx is more sensitive, many CDx-positive but CTA-negative subjects will be eligible for a therapy who were never evaluated in a clinical trial and thus additional clinical evidence will be needed about the CDx, she stresses. Sponsors expecting a sensitive CDx should use a CTA with similar sensitivities to enroll subjects to support CDx effectiveness.  

For CDx targeted for biomarkers with very low prevalence, the specificity of the CDx should be very high, Philip says. High NPA between CDx and CTA is expected; otherwise, false-positives will dominate the CDx-positive population that was not evaluated by the trial.  

Importantly, bridging/accuracy studies need a sufficient number of negative samples. The selected CTA should also have high specificity, or the clinical study will be dominated by false-positives due to low prevalence. 

CDx Case Studies 

Philip next briefly discussed two recently approved CDx devices, one for tumor-agnostic tumor mutation burden (FoundationOne) and the other a liquid biopsy next-generation sequencing (NGS) oncology panel (Guardant 360). Among the key considerations for tissue-agnostic biomarkers and CDx are that they are best supported by strong underlying biological and clinical rationale, trials demonstrate efficacy across multiple tumor types, and the assay performs well across tumor types. 

One of the most important aspects of development and validation is the alignment of biomarker rules for the enrolling assays and to-be-marketed CDx, Philip adds. Otherwise, the efficacy calculation may not be reliable. 

Liquid biopsy NGS oncopanels include category reporting, says Philip. Category 1 indicates prescriptive use of a CDx with a specific therapeutic. Category 2 signifies strong evidence of clinical significance from other FDA-approved liquid biopsy CDx. Category 3A products come with evidence of clinical significance presented by tissue-based, FDA-approved CDx or professional guidelines which—unlike Category 3B products—includes blood to tissue concordance. Category 4 products are backed by evidence from peer-reviewed publications for genes/variants in tissue and variant information from well-curated public databases or in vitro preclinical models. 

Analytical evidence is required across the board, but clinical evidence is presented only for CDx in the first category, she says. Strength of reliability of analytical validity varies by category. 

When bridging studies are required for liquid biopsy NGS-based oncopanels, often the CDx is compared to tissue CTA if enrollment was based on tissue-based testing, she says, and it should demonstrate high PPA and NPA to the enrolling tissue CTA.  

Follow-on CDx tests for already-approved liquid biopsy CDx indications get validated through bridging studies or concordance to the originally approved CDx and measure non-inferiority based on discordance using a procured clinical sample set that is the same as the target population, Philip continues.  

Among the FDA resources she highlighted was final guidance recently published on therapeutic group labeling, describing an approach and specific conditions where device sponsors may request broader labeling for CDx to include multiple drugs in the same therapeutic group. 

Two Guidances  

An update on selected precision oncology initiatives at the FDA was provided by Julie Schneider, associate director for research strategy and partnerships in the agency’s Oncology Center of Excellence created in response to the 21st Century Cures Act. The institute oversees all clinical medical oncology reviews regardless if the product is a drug, device, or biologic. The focus of her presentation was on guidance and policy on low-frequency subsets and IVD risk in oncology trials.  

Genomics tests are used in oncology drug development as enrollment/inclusion criteria for clinical trials, prognostic biomarkers useful for enrolling a high-risk population, pharmacodynamic/response biomarkers, pharmacogenomics/drug metabolism, safety biomarkers, efficacy biomarkers/complementary diagnostics and predictive/selection biomarkers that frequently require contemporaneous approval of a CDx when essential for the safe and effective use of the drug, she says. Medical oncologists are involved in the FDA review process for INDs and also work on approvals for drug/diagnostic pairs. 

Most approved CDx are for oncology indications, says Schneider. In cases where a diagnostic test is being used for trial enrollment purposes and the variants of interest are very rare, the low-frequency subset guidance outlines how patients with different molecular alterations may be grouped if there is strong scientific rationale to expect they will have similar responses to a drug.  

The case for extrapolation gets made with the FDA at the study design stage, Schneider says. The agency will “usually approve the drug for all patients who meet the inclusion criteria for the trial based on the pre-specified criteria, irrespective of the extent to which patients with various molecular alterations were represented.” This is not a substitute for developing a companion diagnostic if one is required for the safe and effective use of the drug. “Co-development will proceed as normal.” 

As an example of the rare subsets guidance in action, she points to revisions to the original label for Gilotrif (afatinib), used to treat metastatic non-small cell lung cancer, to define the indication for use as “non-resistant epidermal growth factor receptor (EGFR) mutations” in lieu of the previous reference to “exon 19 deletions or exon 21 (L855R) substitution mutations.” The clinical pharmacology section of the label was expanded to describe the scientific rationale and define the term non-resistant EGFR mutations, says Schneider. 

She next discussed guidance about the optional streamlined process for study risk determination of IVD devices in oncology trials. Medical oncologists are actively involved in determining if an IVD presents significant risk, nonsignificant risk or is exempt from further pre-market review. The FDA was aware that sponsors frequently had to complete two processes when an IVD was to be used in an oncology trial, she says, including a Study Risk Determination application through a Q-Submission to CDRH as well as an IND to the Center for Drug Evaluation and Research. The two offices frequently work together to determine if an investigational IVD presents significant risk. 

The optional streamlined process allows the drug and device sponsor to submit all of the information through the IND, explains Schneider. The process is intended for oncology co-development programs, only pertains to new IND submissions, and does not apply if an invasive biopsy is taken for investigational IVD testing.   

As clarified in the final guidance, the IND submission would need to include information about how the companion IVD will be used in clinical trials, she adds, but proprietary information about the device is generally not required to make a risk determination. The FDA’s turnaround time with feedback on the diagnostic is within the 30-day review time for the IND. 

Reference Samples  

The public-private Medical Device Innovation Consortium was created to advance medical device regulatory science and the “clinical diagnostics” part of that effort is a Cancer Genomic Somatic Reference Samples project specific to NGS-based oncology tests, says Schneider. It’s focused on developing tissue-based reference samples for solid tumors to support efficient test development and validation and potentially help streamline the regulatory review process. 

The first, completed phase involved defining a group of actionable variants, which is focused on formalin-fixed paraffin-embedded cell lines, and developing a detailed request-for-proposal for selecting a vendor to produce them. During the second phase, now underway, those cell lines will be manufactured and characterized using multiple types of sequencing and ultimately be made available to the research community. 

The role of the Oncology Center of Excellence on the project was to contribute to the list of actionable variants, Schneider says. This was to ensure the variants covered a variety of genes, the technical challenges for detecting them, and would further science across several cancer types.  

Editor’s Note: Did you miss the 2020 Next Generation Dx Virtual Summit? Because the event was virtual, you can still access the event including all of the recorded sessions, presentations, and materials. Register for PREMIUM POST-EVENT ON-DEMAND access.