Diagnostic test quality may not be something that pharmaceutical and biotechnology companies have had a lot of experience dealing with, but shortfalls in a diagnostic development program can undercut efforts to streamline clinical trials for personalized medicines.

The need for a high-quality diagnostic test development program was a key message delivered by a representative from FDA’s devices center at a Sept. 14 conference on drug/diagnostic co-development.

The conference, sponsored by the Friends of Cancer Research and Alexandria Real Estate Equities Inc., served as a discussion forum for a multi-stakeholder draft proposal aimed at enabling late-stage clinical development of drugs with companion diagnostics by quickly identifying a patient subset most likely to benefit from treatment while minimizing exposure of patients least likely to benefit.

FDA officials were generally receptive to the draft ideas put forward. However, they highlighted the threshold need for a well-characterized, high-quality diagnostic that can accurately distinguish between those patients well-suited for treatment and those unlikely to derive any benefit.

“If you mess up the test and you need one, then it doesn’t really matter how well you’ve designed the trial,” said Elizabeth Mansfield, director of personalized medicine in the Center for Devices and Radiological Health’s Office of In Vitro Diagnostics Device Evaluation and Safety.

FDA and industry representatives also shared real-world examples of the complexities inherent in the development of precision approaches to treating diseases, particularly when a biomarker emerges after a drug is approved.

Draft Paper’s Proposals

The draft conference paper presented at the meeting was developed with the goal of guiding design of Phase III trials to evaluate a drug and companion diagnostic in situations where prior studies do not provide a clear definition of the diagnostically selected population and sufficient evidence to restrict development.

The draft is intended to fill some of the stakeholder-identified gaps in FDA’s July 2011 draft guidance on vitro companion diagnostic devices. In that 12-page guidance, FDA explained how it defines a companion diagnostic and its expectations for simultaneous development and approval of a drug with the accompanying test (“FDA Releases Companion Dx Guidance Broad In Scope, Limited In Detail” — “The Pink Sheet,” Jul. 18, 2011).

However, the guidance left many in the drug and device industries wanting more details, such as the regulatory ramifications for drug/diagnostic combinations that are not developed and reviewed in parallel fashion and how laboratory-developed tests fit into the agency’s paradigm for companion diagnostics (“Views Differ On How Lab-Developed Tests Fit In FDA Companion Dx Guidance” — “The Pink Sheet” DAILY, Oct. 24, 2011).

The conference draft paper addresses some of the more granular process questions that arise during the development of a drug and companion diagnostic, such as the relevant study population. The paper offers guidelines for sponsors to consider in deciding whether to include all comers in a Phase III trial or merely those who test positive for the biomarker of interest.

“Patients have to be protected by appropriately balancing the strength of the diagnostic hypothesis with the need for thorough data generation and evaluation,” the draft paper states. “We believe the appropriateness of including marker-negative patients primarily depends on the strength of the science in support of the diagnostic hypothesis (including but not limited to mechanism of action, pre‐clinical efficacy and, if known, class effect), the potential for risk to patients, and clinical data available to date,” the draft paper states.

The draft contains a grid with recommendations on the extent to which both marker-negative and marker-positive patients should be evaluated in a registration trial.

The paper also contains recommendations on defining the diagnostically selected population for purposes of labeling, including re-adjusting the threshold on a biomarker prospectively specified in the Phase III trial and evaluating the Phase III data with regard to a biomarker not specified in the protocol.

The development and validation of diagnostics that assess multiple biomarkers also is addressed in the draft paper. The composite biomarker, rather than the individual components, should be validated in the Phase III trial, the paper states. “Provided there is appropriate pre‐clinical evidence and scientific rationale, clinical validation of each biomarker separately should not be required; rather, it should be sufficient to validate the combination of markers, essentially treating the pool of biomarkers as one,” the draft states. “This general proposal is applicable to both single (e.g. RT‐PCR panel) and multi‐platform Dx (e.g. IHC and mutation), although different practical considerations may apply.”

Real-World Experiences

Panels comprising representatives from industry, FDA and other government agencies, academia and patient groups discussed the draft proposals in the context of real-world experiences with the development and approval of targeted therapeutics.

Richard Buller, VP of translational oncology at Pfizer Inc., said clinical trial patient selection is a critical decision in drug/diagnostic co-development.

“Whether or not to include biomarker-negative patients in your studies can really precipitate pretty vigorous discussions from, on the one side, it’s not ethical to, on the other side, you really need a four-arm study with biomarker-positive, biomarker-negative and your test agent versus the conventional approach,” Buller said.

He cited several recent approvals that reflect the varying levels of evidence associated with bringing a personalized medicine to market or refining its label afterward.

First he pointed to Pfizer’s own Xalkori (crizotinib), which received accelerated approval in August 2011 for metastatic or locally advanced non-small cell lung cancer that is anaplastic lymphoma kinase (ALK)-positive as detected by a companion diagnostic (“Pfizer’s Crizotinib Eases Past FDA With Targeted Population” — “The Pink Sheet,” Aug. 29, 2011). The approval included a post-marketing commitment to study crizotinib’s efficacy in ALK-negative patients.

Also approved in August 2011 was Roche’s Zelboraf (vemurafenib), for treatment of patients with unresectable or metastatic melanoma with the BRAF V600E mutation as detected by a companion diagnostic (“Zelboraf Approval Hastened By FDA Officials Impressed With Early Efficacy” — “The Pink Sheet,” Aug. 22, 2011). Buller said Phase I testing showed no activity in patients who were negative for the BRAF mutation; consequently, there was no post-marketing commitment for a study in those patients.

Finally, he pointed to the “saga” of Bristol-Myers Squibb Co./Eli Lilly & Co.’s Erbitux (cetuximab), approved in July for first-line metastatic colorectal cancer negative for KRAS mutations detected with a companion diagnostic. Approval of the supplemental indication was aided by a retrospective analysis of tissue samples from a placebo-controlled pivotal study, which showed no efficacy in patients with KRAS mutation-positive tumors (“Can Erbitux Make Inroads In First-line Colon Cancer With Approved Companion Dx In Tow?” — “The Pink Sheet” DAILY, Jul. 10, 2012).

Erbitux and another epidermal growth factor receptor antagonist, Amgen Inc.’s Vectibix (panitumumab), had previously added labeling language stating that patients with KRAS mutations would not benefit from treatment. Those labeling changes followed from the Oncologic Drugs Advisory Committee’s recommendation, made at a December 2008 meeting, that retrospective studies can be used to support labeling for predictive biomarkers (“Committee OKs Retrospective Studies, But With Caveats” — “The Pink Sheet,” Dec. 22, 2008).

Buller said several factors can impact the decision on whether or not to study a drug in the biomarker-negative population. Among these is the nature of the biomarker itself, such as whether it is a continuous variable or binary, targeting a single marker or a spectrum of markers.

Selection also depends upon how well the patient population is understood, as well as the nature of the test itself. “Everybody expects it should be locked down, technically validated, sensitive and specific before you go to test patients. However, no test is perfect and each development pathway can be quite varied,” he said.

In general, Buller said, it is best to include biomarker-negative patients at least through Phase II, particularly in situations where the test threshold may not be clear.

FDA Office of Hematology and Oncology Products Director Richard Pazdur highlighted several basic principles that FDA considers when evaluating the use of biomarkers in applications submitted to the agency.

The first among these is that the companion diagnostic must be essential for use of the drug, Pazdur said. “It’s not like we’d like to have it or it might be nice to have it or we might in the future need it. It should be essential for the use of the drug when it is licensed.”

Furthermore, use of the drug and accompanying diagnostic must be placed in the context of existing therapies, Pazdur said. The final consideration is “efficacy trumps all,” Pazdur said. “If you really have a very efficacious agent, one would question whether they even need an in vitro diagnostic especially if it’s superior to all approved therapies in the general population.”

How these three considerations factor into FDA’s decision-making were exemplified through two examples provided by Pazdur – one involving a drug with modest activity and a poorly characterized in vitro diagnostic, and another drug with impressive efficacy that extended beyond the biomarker-selected population (see sidebar).

Concerns On The Device Review Side

Shortfalls in diagnostic development programs were a key concern for CDRH’s Mansfield, who said sponsors need to pay close attention to how the diagnostic is used in clinical trials.

“I think really a lot of attention needs to be paid to how you’re actually going to implement this test. You need a protocol, you need consistency so that we can actually interpret the drug outcome as well as how well the diagnostic worked. It’s really critical to start with that, then maybe simultaneously or subsequently worry about the clinical trial. Make sure you can run that test in a uniform way that can be interpreted. That is just so important, and we’ve seen a lot of sort of weak test development in that aspect.”

Mansfield said the agency most commonly sees the clinical trial model wherein only those patients who are positive for the biomarker of interest are enrolled. “That works and it seems to get drugs to the market really fast. But as we’ve discussed here, there probably are better designs that are more complex, a little harder to carry out, but don’t actually ignore the marker negatives,” she said.

She also stressed the need for drug developers to plan for the possibility that a biomarker will be discovered after a drug is on the market, resulting in changes in clinical practice that complicate the ability to validate that marker.

She pointed to the difficulty of retrospectively analyzing the lack of efficacy for EGFR inhibitors in patients with KRAS mutations.

“For KRAS, essentially once the ODAC was held, at least, maybe earlier than that, everybody stopped treating people who had KRAS mutations and there were no longer any samples available for people who had actually been treated that you actually had evidence that KRAS mutation mattered,” she said.

When clinical practice changes as a result of emerging biomarker information, “then we have no samples left that can actually validate the test on which we can say, ‘Hey, guess what, that test worked,’” Mansfield said. “With KRAS that was really a huge problem for us, and we stood on our head and we did all kinds of statistical tricks just to make ourselves feel comfortable that we had gotten the answer right. I think really the answer here is that pharma has to plan for this, we can’t just pull it out of the hat.”

She also sought to tamp down the myth that FDA will not consider a diagnostic targeting multiple markers.

“I’m dying for somebody to bring forward a test that has multiple markers on it, whether it’s for lung cancer, whether it’s for any kind of cancer you can think of, every marker that you can test for. Because we can do that,” she said.

“We get to work on what people bring to us, and I think there’s a folklore out there that this isn’t possible through the regulatory mechanism. I think it’s actually not only possible but imminently doable and it’s going to be required. How much tissue do you have? You can’t test one piece of tissue infinitely with every different test on the market. So from my point of view, you know, bring it. We’ll deal with it.”