By Deborah Borfitz 

September 13, 2023 | The U.S. Food and Drug Administration (FDA) has been making moves signaling its intention to better align the regulatory framework for companion diagnostics (CDx) to what is happening in real-world clinical practice. The key showpiece at the moment is a pilot program, announced in June, which seeks to reduce the risks associated with using laboratory developed tests (LDTs) for selecting certain oncology drug treatments.  

A brief summary of the pilot was offered yesterday by Dun Liang, Ph.D., executive director of regulatory affairs for companion diagnostics for Loxo@Lilly, the oncology branch of Eli Lilly, during his presentation at the Next Generation Dx Summit 2023 in Washington, DC. The talk focused on current CDx guidance and the challenges and implications of some of the agency’s proposed changes from the pharma perspective. 

Before joining Eli Lilly, Liang worked as an FDA reviewer of companion diagnostics. Earlier, he was a lab director responsible for bringing LDTs into clinical use.   

The FDA has to date issued four CDx guidances, the first foundational document in 2014 followed two years later by how-to type guidance around CDx development and regulation, he says. In 2020, the agency published group labeling guidance and, most recently in June, particulars on the voluntary pilot program. 

In the first two guidances, the FDA addresses three CDx regulatory models—co-development, bridging, and follow-on—only the first of which involves using the diagnostic test in a clinical trial, Liang points out. As originally envisioned, the concept of companion diagnostics revolved around multiple sites sending patient samples to a central lab where a single CDx candidate would be used to determine if a patient was eligible for a study. The clinical validity (i.e., efficacy and safety) of the CDx would be evaluated at the same time as the investigational drug, with analytical validation completed well before then. 

This bumped up against the reality in most oncology trials where the decision about whether to enroll patients is an urgent matter without the luxury of a three- to four-week wait to get them through central testing, says Liang. As a matter of practicality, sites will therefore enroll patients using LDTs.    

When doing this, samples get banked and sent to the CDx sponsor with whom a drug manufacturer has contracted to do a bridging or follow-on study, Liang says. That is, the clinical validity of the CDx is “extrapolated” by one of these means rather than being validated in a drug trial. 

Local tests come in many varieties, he continues, among them tests purely developed in the lab, LDTs designed to be “research use only” (RUO) products and those built on FDA-approved diagnostic kits, in vitro diagnostic tests without a CDx indication or used without a drug indication, and clinical trial assays (CTAs) that are developed as a “prototype” companion diagnostic but are changed after their use in a clinical trial. The FDA’s major concern with the CTAs is that they are being used in clinical trials establishing the efficacy and safety of a drug or to define the intend-to-treat population but were never themselves formally reviewed. 

The Disconnect 

For sponsors, the gap between the reality and the foundation of the CDx regulation is causing confusion, says Liang. Based on the idea that there is a one-to-one relationship between a drug and companion diagnostic, FDA guidance indicates that the drug label needs to reflect the CDx and the agency expects in most cases that the drug and diagnostic will be co-approved. But in the real world, multiple device and drugs have become the mainstays of cancer treatment. 

“Practical use of local tests hasn’t been properly engineered in the current regulatory framework,” he says, “and in the end the patient’s access to the approved agent is actually limited.” Moreover, CDx is regulated as a class III device, the agency’s “highest and most complex” category requiring a premarket approval (PMA)—meaning, a more rigorous validation process requiring at least twice as much replication as the 510(k) premarket submission pathway—a bar that few device sponsors can reach.   

Significant differences also exist between the biomarker rules at play for a drug versus a device submission, adds Liang, which tends to complicate the review process. The situation is “even worse” in cases where both the drug and CDx in a treatment trial are considered investigational because they are regulated separately under, respectively, investigational new drug (IND) and investigational device exemption (IDE) application processes. 

“Compliance with IND does not necessarily fulfill compliance with IDE,” he notes. “It is one clinical study, two regulations,” one each for the drug and device sponsor.   

Current regulations are leaving sponsors with a few big questions, starting with whether a CDx is required for a targeted drug approval. In the case of an all-comer trial where the drug demonstrates significant efficacy in the biomarker-positive population, he offers as an example, existing FDA regulations suggests that the agency might want to limit the indication to the biomarker-positive population as identified by the CDx. But by the time the drug sponsor learns this and wants to engage in CDx development it is “almost too late.” 

Then there’s the question of how to properly use a local test for an oncology drug trial, Liang continues, or more precisely how to bridge a set of CTAs through one candidate CDx. The answer would typically be to use the representative method, but that variety would be hard to characterize when trials might be using 100 different types of assays. 

How to meet significant risk device (SRD) requirements under the IDE approval process is yet another conundrum, he adds. The question is whether using an FDA-approved test could effectively help de-risk a drug study and, in practice anyway, the answer is probably no since it is not authorized for the CDx indication. But since it has gone through “reasonably rigorous validation” for accuracy and precision, requiring two sets of reviews—IND and IDE—seems unnecessarily wasteful. 

Finally, sponsors are wondering how to go about selecting the right CDx partner or device, says Liang, which in the real world is not necessarily a choice based on the quality of the technology or commercial capacity. The requirements of the preferred co-development model often eliminate many of the new platforms with bigger business potential simply because device companies can’t meet the speed to approval of drug developers.   

Pilot Particulars 

Under the FDA’s new pilot program, some precision oncology drugs can be approved without a CDx, says Liang. “The drug label will no longer be tied to the specific CDx at the time of drug approval and drug sponsors need to submit performance of the clinical trial assay used in the trial.” Performance metrics on CTAs will be posted online by the FDA. 

The rationale for the pilot, he notes, is that enrollment based on centralized testing potentially restricts patient access to clinical trials and delays treatment. The FDA is worried about the lack of transparency regarding its oversight of LDTs, particularly when they are being used to identify participants in pre- and post-marketing studies.   

To date, the FDA has reported approving 40 targeted therapies without mandating CDx approval, says Liang. These include cases where sponsors have made a post-marketing commitment to later bring a CDx through the approval process. 

Liang next points to a few of the key differences when comparing the current CDx paradigm with the new pilot. Today, the drug label would read that a particular biomarker is detectable by an FDA-approved test, and the device label would indicate that the test has been approved through the PMA pathway for use in identifying patients for cancer treatment with a certain drug. Device sponsors would provide the required clinical data directly from a drug trial or extrapolate the data via bridging or follow-on studies, and tests would be reviewed as class III devices. 

Under the new pilot, Liang says, one drug will be matched to multiple tests exhibiting FDA-recommended performance characteristics. Drug labeling will no longer refer to the need for an FDA-approved test, but instead CTA biomarker tests that meet the agency’s performance bars. No device labeling would be needed since these are LDTs. 

The drug/device review process will likely be combined and involve consultation with the FDA’s Center for Devices and Radiological Health, he speculates, which would eliminate the need for separate clinical data. How the SRD process will go under the pilot is not addressed in the guidance. 

Among the key questions on the minds of drug developers is whether they need to reach out to labs outside the U.S. to obtain needed information, and the FDA’s tolerance for missing performance data on some CTAs given the difficulty of the collection task, says Liang. 

Up for discussion by labs are their willingness to work with pharma sponsors to provide these data and the incentive to do so, as they might view the performance characteristics of their tests as proprietary information, he adds. Labs that didn’t develop a test from scratch might also wonder if they can simply submit performance data on the RUO kit they modified. 

The FDA, for its part, may be pondering how to align its pilot program to LDT rulemaking, says Liang. A mechanism for integrating the review of select diagnostic data during the drug review process is also still missing.