Skip to main content

FDA Panel Mulls Criteria for Accepting Retrospective Data for Rx/Dx Products

A panel of independent experts convened by the US Food and Drug Administration this week attempted to define acceptable evidence criteria for drug developers wishing to limit the indication of marketed therapies to a genetically defined subpopulation without sacrificing the agency’s standards for efficacy and safety.
The FDA’s Oncologic Drugs Advisory Committee met in Gaithersburg, Md., to discuss drug makers’ use of retrospective clinical trials to define the genetic subpopulation for targeted treatments. The committee considered two real-world examples Amgen’s Vectibix and ImClone/Bristol-Myers Squibb’s Erbitux, both anti-epidermal growth factor receptors for colorectal cancer — in discussing the issues surrounding retrospective analyses, including assay validation, appropriate sample requisition, and biomarker discovery.
ImClone and Amgen have each submitted retrospective analyses in an effort to update the drugs’ labels to show they are more efficacious in patients with metastatic colorectal cancer who carry a wild-type version of the KRAS gene.
The Oncologic Drugs Advisory Committee did not convene the meeting to decide whether the labels should be updated. Rather, the committee met to discuss the kinds of evidence the FDA should demand from sponsors that want to narrow the patient population for a drug based on the results of a genetic test developed after the therapeutic came to market.
Ideally, the agency prefers that a companion diagnostic is developed simultaneously with the drug, Richard Pazdur, director of FDA’s Division of Oncology Drug Products, said during the meeting. However, as the industry identifies new gene associations and develops new assay methods, concurrent drug/diagnostic development is not always possible, the FDA acknowledged in briefing documents for the meeting.
‘Fudge Factors’
The FDA approved Erbitux (cetuximab) in 2004 and Vectibix (panitumumab)in 2006. The acceptance within the scientific community of KRAS as a predictive marker of response to EGFR-inhibiting treatments in metastatic colorectal cancer came in the years that followed.
Although the panel did not arrive at any definitive conclusions about when and how to conduct retrospective analyses meant to define genetic subpopulations for a drug, several committee members urged the FDA to maintain “common sense” evidentiary criteria. Committee members generally agreed that when considering retrospective data on genetically targeted treatments, the FDA should validate the assay used to define the subpopulation; ensure that tumor samples used in the study represent the disease state of the patients in the trial; and require sponsors to replicate retrospective findings in a prospective trial.
“We want to be careful that we don’t box [FDA] into a corner where with our information we set a bar that’s so high from our advice that they can’t make sensible decisions,” cautioned Derek Raghavan, director of the Cleveland Clinic Taussig Cancer Center and an ODAC consultant. “One of the things that have been lost here today — and this may be the first time I’ve seen this at the FDA — is that two companies have come here today to enter into a situation where they sell less product. That seems like an important thing.
“We need to set rigor, we need to have good assays, and we need to have well-powered studies, but we might create a fudge factor that would allow [the FDA] to look at the overall numbers,” Raghavan added.
He pointed out that the FDA allows drug companies to submit meta-analyses of “poorly conducted, small studies” under certain circumstances. “That’s not a replacement for a well-conducted study, but we [should] set rules where we have common sense in them and allow the FDA discretion to look at what was the intent of the [retrospective] study,” he continued. “If we raise the bar with a lot of clever terms that will actually stop common sense from being implemented, that would be a shame.”
FDA makes clear in its briefing documents that it discourages retrospective genomic-biomarker assessments, since such trials involve the “re-analysis of a ‘failed’ clinical trial in which efficacy is purported to be established in a subset defined by a genomic biomarker/patient characteristics without consideration of multiplicity (i.e., data dredging), substantial missing data, and poorly characterized assays.”
Seeking to update their labels with KRAS testing data, ImClone and Amgen submitted retrospective analyses to the FDA that looked at efficacy outcomes when KRAS genomic status was considered before administering Erbitux and Vectibix.

“One of the things that have been lost here today — and this may be the first time I’ve seen this at the FDA — is that two companies have come here today to enter into a situation where they sell less product.”

FDA agreed to accept the data under the following conditions: the trials must be adequate and well-controlled; the sample size must be sufficiently large to ensure randomization; tumor tissue must be obtained in more than 95 percent of registered study participants and an evaluable result must be available for greater than 90 percent of study subjects; the assay must be reviewed and validated by the FDA; genetic analyses must be performed according to a qualified assay methods by blinded investigators; and that sponsors and the FDA must create an analytical plan to test hypotheses for updating labels and making promotional claims.
ImClone submitted four retrospective studies and Amgen provided two studies. But only one of the Vectibix studies (no. 408) and two of Erbitux’s submissions (CRYSTAL and NCIC-017) reached their primary efficacy endpoints, according to Ruthann Giusti, a medical officer in FDA’s Division of Biologic Oncology Products.
During the meeting, Richard Simon, an ODAC consultant and chief of the National Cancer Institute’s Biometric Research Branch, cautioned the committee against being wedded to the idea that retrospective studies are only possible when prospective trials have met their overall endpoint.
Rather, the FDA should look at whether a sponsor conducted “a focused analysis and accrued enough patients that test positive and negative for the biomarker, and [have] a test that is analytically validated on archived tissue.”
However, ODAC member Gary Lyman disagreed. “If the original endpoints are not reached …to look at subgroups doesn’t hold,” said Lyman, who is director of Duke University Medical Center’s Health Services and Outcomes Research Program in Oncology. He noted that ideally, the FDA should require two prospective studies for stratified patients and ensure that the trials have enough subgroup participants, particularly if the biomarker being studied is meant to be prognostic and predictive.
During the panel meeting, most participants in general agreed that any retrospective study submitted to the FDA should be randomized, blinded, and have appropriate statistical power and control multiplicity.
One major focus of the discussions involved how tumor samples collected during the original study should be used in retrospective trials. Committee members agreed that it is critical that the samples represent the patient’s disease state at the time of the study, since the patient’s disease state could change between the time of the original study and subsequent retrospective analyses.
“I think you’re going to have to specify that the tumor [sample collection] is going to have to be closer in time to when you’re doing the test,” said ODAC committee member Michael Link, who is chief of the Division of Hematology/Oncology at Stanford University School of Medicine. He noted that in the trials submitted by Amgen and ImClone, “it’s not clear that the tumor specimen represented what the patient looks like now at the time he is being treated.”
Several committee members also suggested ways in which the FDA could mandate sample collection. Lyman proposed that sponsors should collect samples for all novel targeted agents during Phase III trials because it is likely that these kinds of candidates will have a companion diagnostic in the future.
Meanwhile, Raghavan suggested the FDA create a biorepository that sponsors can reference when updating new versions of their tests.
“I think the problem is we’re thinking about this the wrong way. … We’re not around this table today going to define the ideal test because it doesn’t exist,” Raghavan said. “The only way to protect the [population] at large is to have some mechanism to go back and check [the sample]. … I’m not suggesting that there should be a gigantic biorepository housed in the Capitol building, but what I do think is that [we remain] very reasonable as we define new rules.
“When a biomarker is influencing an outcome as broad as who gets the drug and who doesn’t, and particularly as the test of 2008 will be supplanted by the test of 2010 … we need some framing reference for doing that,” Raghavan continued. “The definition of the perfect test will need to be functional. … We need a functional definition of difference, and then we’ll need to decide what is a difference that is big enough.”
Another FDA official at the panel discussion noted that the agency could make sample collection part of a sponsor’s post-marketing commitments to ensure that material will be available for future assay development.
Other ODAC members pointed out that mandating sample acquisition is complicated by cost consideration, varying IRB requirements, and difficulties obtaining informed consent.
One of the FDA’s requirements for Amgen and ImClone to submit their retrospective analyses was for them to obtain tumor tissue from more than 95 percent of the registered study participants, and to ensure that an evaluable result is available for more than 90 percent.
Lyman suggested that perhaps all sponsors requesting to submit retrospective analyses on genetic subpopulations for a treatment should be held up to this standard.
“Acquisition of 90 percent of patients can be reached,” Lyman said. “I think we have to set the bar high if we’re going to do these types of retrospective [assessments of] prospective data. If we’re to look at half the samples or less than half the samples, it leaves us open to a lot of confounding factors that can’t be resolved retrospectively.”
Defining FDA’s Stance
The committee’s discussion could shed additional light on the FDA’s position regarding Rx/Dx co-development. The agency issued a white paper on this subject in 2005. Since then, it has discussed issuing a formal guidance on the topic but has yet to do so.
However, at a recent meeting on personalized medicine in Boston, Lawrence Lesko, director of FDA’s Office of Clinical Pharmacology, announced that the FDA had convened a multi-center and multi-disciplinary working group to advance the stalled document. 
More immediately, ODAC’s discussions could help the FDA decide if, when, and how to update Erbitux and Vectibix’s label to reflect KRAS testing.
The KRAS gene is mutated in between 35 percent and 45 percent of metastatic CRC patients, and studies have shown that KRAS testing can better define who will likely benefit from EGFR-inhibiting monoclonal antibodies such as Vectibix and Erbitux.
The European Medicines Agency, meantime, appears to have made up its mind, recommending that doctors establish a patient’s KRAS status before prescribing Vectibix and Erbitux. The FDA has so far held off on issuing similar recommendations, indicating that it is awaiting prospective trial data.
Meanwhile, the scientific community moved ahead with its acceptance of KRAS testing with EGFR-inhibiting monoclonal antibodies for colorectal cancer.
At the American Society of Clinical Oncology’s annual meeting in June, Eric Van Cutsem, a professor at the University Hospital Gasthuisberg in Leuven, Belgium, and colleagues presented data from a large, multinational, prospective clinical trial showing that metastatic colorectal cancer patients who carry the wild-type version of the KRAS gene are much more likely than patients with the mutated form of the gene to benefit from Erbitux.
The National Comprehensive Cancer Network in November recommended that cancer doctors determine the status of the KRAS gene either of a primary tumor or a metastasizing site as part of a pre-treatment work-up for colon cancer patients.
The FDA recognized that “the widespread publication and presentation of the retrospective KRAS analyses have resulted in practice changes in the community, [and] thus, a prospectively designed trial may no longer be feasible.” As a result, the FDA agreed to accept retrospective analysis if certain conditions were met.
The ODAC meeting was held to further inform and refine the agency’s criteria for accepting retrospective analysis for Rx/Dx combination products.
An FDA-Cleared Dx
In both Amgen and ImClone’s clinical trials, patients’ KRAS status was gauged using DxS’ TheraScreen assay. During the ODAC meeting the FDA made clear that an assay cleared by the agency would be required in order to update the labels of Erbitux and Vectibix.
UK-based DxS told Pharmacogenomics Reporter this week that it is working with FDA to gain regulatory approval for its TheraScreen KRAS mutation kit as a companion diagnostic for use with Vectibix in metastatic colorectal cancer patients with wild-type KRAS status.
Currently, Althea Diagnostics, Caris Diagnostics, and the Laboratory Corporation of America, amongst other labs, offer KRAS testing with the DxS assay. Many believe it would likely help drive physician adoption and reimbursement efforts if the FDA updates the label of Erbitux and Vectibix with information on KRAS testing.

Filed under

The Scan

And For Adolescents

The US Food and Drug Administration has authorized the Pfizer-BioNTech SARS-CoV-2 vaccine for children between the ages of 12 and 15 years old.

Also of Concern to WHO

The Wall Street Journal reports that the World Health Organization has classified the SARS-CoV-2 variant B.1.617 as a "variant of concern."

Test for Them All

The New York Times reports on the development of combined tests for SARS-CoV-2 and other viruses like influenza.

PNAS Papers on Oral Microbiome Evolution, Snake Toxins, Transcription Factor Binding

In PNAS this week: evolution of oral microbiomes among hominids, comparative genomic analysis of snake toxins, and more.