Forget whether progression-free survival (PFS) should be a surrogate endpoint in cancer trials, the FDA said as it charged the Oncologic Drug Advisory Committee (ODAC) Tuesday to look at reducing the central review audit burden as a way to streamline cancer trials.

To weed out investigator bias that may have crept into unblinded trials, the FDA currently requires an independent review committee (IRC) to assess all the radiographic images used to determine PFS in trials involving solid tumors. The agency is considering reducing that audit to a central review of a sampling of the images as a way to lower the cost of the trials.

If a trial is truly blinded, no review is needed. But such trials are "few and far between," Richard Pazdur, director of the FDA's Office of Hematology & Oncology Products, said, adding that many factors can unmask a trial.

An IRC costs $4,500 to $7,500 per subject, which means a full review could add up to $1 million to $3 million, depending on the trial size and vendor, Cynthia Dinella, president of Advyzom LLC, told the committee.

Using a sample instead of a full audit could achieve 20 percent to 30 percent in cost savings, which sponsors could use for other trials, she said. But how much would actually be saved would depend on how the FDA implemented the sampling.

Sponsors of multisite trials still could face a logistic burden if they were required to collect all the scans. And if the FDA decided retrospectively that a full review were needed, it could be costly and delay patient access to new cancer drugs, Dinella said.

For the sample-based audit to truly streamline the process, industry would need clarity from the FDA on whether the investigator-assessed PFS would be the primary endpoint in Phase III solid tumor studies and whether the audit would be considered a secondary endpoint, Dinella said. The agency also would need to provide clear guidance up front on how the process would work and what would be expected.

Some of the concerns about the details were shared by ODAC members, who, for the most part, seemed to favor the sample method. However, they stressed that it shouldn't be a truly randomized sample because representation would be needed from all the sites, given the variability from site to site.

Much of the discussion centered around that variability and the reasons for it. Rajeshwari Sridhara, director of the Division of Biometrics V at the FDA's drug center, cited an average discordance of more than 30 percent between investigator assessments of PFS and that of the IRCs. In one case, that discordance was as high as 50 percent.

While the discordance shows variability in how the scans are read and interpreted, there is no evidence indicating a systematic investigator bias in randomized PFS trials, Dinella said. In some instances, it's a difference of how a lesion is measured – or the IRC measuring a different lesion in the scan than what the investigator measured.

Nevertheless, a few experts referenced those differences during the public comment period to push for more training of site readers, increased involvement of trained radiologists and imaging standardization in trials.

Panelist Aman Buzdar, vice president of clinical research at the M.D. Anderson Cancer Center, agreed with those experts. While he thought a sample-based audit was a good idea, he said standardization is needed to address the differences in the quality of scans. "We need to raise the bar to make it much more robust," he added.

But transferring the savings from a full audit to site training and standardization would encumber sponsors more than having a full IRC review, said Wyndham Wilson, chief of the Lymphoma Therapeutics Section at the National Cancer Institute.

Although no vote was taken, the committee seemed to agree that the sample-based audit made sense for some cancer trials that have PFS as an endpoint – so long as the threat of a full audit remained. However, the panelists didn't think it would be appropriate for small trials or those in which the PFS was modest. And it would depend on the tumor type, which could be determined on a case-by-case basis.

Comparing the IRC audit to an IRS audit, David Steensma, associate professor of medicine at the Harvard Medical School, said with each layer of complexity required, the FDA would make it more difficult to complete cancer studies. Acknowledging that a one-size-fits-all approach won't work, he said the agency needs the flexibility to request a full or a sample-based audit as it sees fit. But it also needs to allow investigators to do their job, he added.