Avastin Debate Could Provide Clarity On Use Of Progression-Free Survival Endpoints

Rather than damaging the use of progression-free survival as an endpoint in oncology clinical trials, as some have feared, the battle over Avastin’s (bevacizumab) metastatic breast cancer claim could give sponsors some clarity as to how FDA expects the magnitude of PFS benefit to be defined.

Discussion at FDA’s June 28-29 hearing on Avastin’s MBC claim suggests that while PFS remains a usable endpoint in some cancer settings, sponsors should not pin their hopes for a drug’s regulatory success on a strong PFS hazard ratio. Rather, they need to look at the entire picture, including a treatment’s impact on median PFS values.

Despite Genentech’s protestations that the hazard ratio, a statistical measure for calculating risk reduction, is a more appropriate measure of a therapy’s efficacy than the median, Center for Drug Evaluation and Research officials were unequivocal in saying that the clinical significance of a hazard ratio is not interpretable unless one also looks at median PFS, which provides a temporal context for judging the clinical relevance of a treatment.

Most of the debate occurred in retrospect, with CDER and Genentech focusing on data from the E2100 trial upon which accelerated approval in MBC was granted and the subsequent AVADO and RIBBON1 studies. Nevertheless, the dispute has implications for Genentech going forward as it attempts to design a new trial aimed at providing the level of evidence CDER says is needed to support the MBC claim (see related story, (Also see "Genentech Losing Its Legs? Avastin Study Proposal Undercut By Feasibility, Scientific Concerns" - Pink Sheet, 11 Jul, 2011.) ).

The debate is sure to have implications beyond Avastin, too, given the continual calls for CDER to provide more clarity and ensure greater consistency in its approval requirements.

PFS Still An Approvable Endpoint

CDER’s proposal to revoke accelerated approval of Avastin’s first-line MBC claim, coupled with the oncology review division’s growing unease with sponsors’ reliance on PFS in certain settings, has spurred concerns within industry and the patient community that PFS is on its way out as an acceptable endpoint for oncology approvals.

CDER Director Janet Woodcock addressed such concerns at a recent Friends of Cancer Research briefing, stating that PFS remain a viable basis for seeking accelerated approval, depending upon the disease setting and drug effect seen (Also see "Accelerated Approval Remains “Viable Pathway,” FDA’s Woodcock Says" - Pink Sheet, 27 Jun, 2011.).

In CDER’s written summary of evidence for the Avastin hearing, the regulator said it has neither changed its views regarding the usefulness of PFS as a clinical endpoint for cancer drug approvals nor determined that an overall survival benefit is always needed in addition to a PFS improvement.

Office of Oncology Drug Products Director Richard Pazdur reaffirmed this view at the hearing. However, whether PFS can serve as the basis for approval of a particular therapy “would have to be considered in the context of a risk-benefit assessment, and one would strongly consider the magnitude of effect in making a decision on that,” he said.

CDER granted accelerated approval for the MBC claim in 2008 based upon the robust magnitude of effect seen in E2100, wherein Avastin in combination with paclitaxel was associated with a significant 5.5-month improvement in median PFS compared with paclitaxel alone, with a hazard ratio of 0.48 (a risk reduction of 52%).

Two confirmatory studies, AVADO and RIBBON1, tested Avastin in combination with non-paclitaxel chemotherapy partners. The addition of Avastin was associated with smaller, though still statistically significant, median PFS improvements, ranging from 0.9-2.9 months, with hazard ratios of 0.62-0.69. However, CDER determined these studies failed to confirm the magnitude of PFS benefit seen in E2100 and sought to withdraw the MBC indication.

The agency’s Oncologic Drugs Advisory Committee served as the figurative “jury” for the two-day hearing and voted 6-0 against retaining the MBC claim while Genentech conducts a new confirmatory trial. FDA Commissioner Margaret Hamburg will make the final decision on the indication’s fate (Also see "Avastin’s Breast Cancer Claim: Will FDA’s Hamburg Take A Middle Road?" - Pink Sheet, 4 Jul, 2011.).

Genentech Criticizes Focuses On Medians

As expected, there was extensive debate at the hearing between CDER and Genentech over what the regulator told the company, and the differences in expectations each had, with regard to the data needed to convert accelerated approval into full approval (Also see "Avastin Breast Cancer Dispute Centers On “Magnitude Of Benefit” Questions" - Pink Sheet, 30 May, 2011.).

Genentech argued that CDER’s views on the evidence needed to confirm the PFS benefit seen in E2100 have evolved since the MBC claim was approved, with the agency now focusing primarily on the median PFS.

Prior to granting accelerated approval, CDER reviewed top-line PFS data from AVADO. This led Genentech to believe that the agency considered the study supportive of the E2100 results based on the magnitude of the PFS hazard ratio, Sandra Horning, Genentech’s global head of clinical development for hematology/oncology, said.

CDER is misplaced in its reliance on median PFS in determining that AVADO and RIBBON1 did not confirm the benefit in E2100, Genentech reps asserted.

“Solely focusing on the median differences is flawed, as they can under- or at times overestimate the true treatment effect,” Genentech Executive VP for Global Product Development and Chief Medical Officer Hal Barron said. “In addition, this focus on median difference as the primary method of assessing magnitude of benefit appears to represent a change in CDER’s thinking since 2008.”

Under cross examination by CDER’s counsel, Barron said: “I think it’s important to point out that the AVADO and RIBBON trials were designed with the intent to be able to observe a hazard ratio between 0.7 and 0.75, and exceeding that threshold with a positive trial can be described as a magnitude of benefit."

Genentech “has been acting on the belief that hazard ratios are the best way to make cross-trial comparisons,” EVP Barron said.

“What we’re trying to explain,” he continued, “is that we have been acting on the belief that hazard ratios were the best way to make cross-trial comparisons in terms of the magnitude of treatment effect. We believe that reflects the entire curve and more data points, and it’s the evolved focus on medians that we’re debating as new.”

Genentech counsel Michael Labson of Covington & Burling said CDER never stated “that only a median of 5.5 months would be considered clinically meaningful” until the July 2010 ODAC meeting, at which the committee voted 12-1 to withdraw the indication.

Office of New Drugs Director John Jenkins challenged the company’s assertion that CDER’s views had changed and questioned how one can put a hazard ratio into perspective without looking at the difference in median PFS.

“You can’t,” Genentech Global Head of Oncology Biostatistics James Reimann replied. “You need to look both at hazard ratios and absolute benefits. But I think it’s important to look at absolute benefits in time, not just at the median point. … It’s the early benefits, median benefits and late benefits.”

Although CDER’s presentation of the Avastin data included both hazard ratios and median PFS rates, “I think it’s quite prominent in CDER’s conclusions and summarization that the focus moves to the medians,” Reimann said.

CDER Division of Biometrics V Director Rajeshwari Sridhara said that to understand the temporal implications of a drug’s clinical effect one needs to look at medians.

“We have had applications where the hazard ratio was 0.5 and, in fact, the difference in PFS was just two weeks. And so where do we take that then? Yes, the hazard ratio was small enough, but the difference in medians was too small to be clinically meaningful,” she said.

Pazdur Won’t Be Pinned Down

In a series of exchanges, Genentech reps repeatedly questioned Pazdur on the hazard ratio and median PFS benefit that would be needed to confirm Avastin’s clinical benefit for MBC. Pazdur, however, would not be pinned down.

An approval decision is based on a risk-benefit analysis, Pazdur said. “The decision to approve a drug is not based solely on a hazard ratio. It is not based solely on the median difference. It is not based on a p-value.”

“We cannot give you a specific number or a specific hazard ratio that would warrant an approval or a non-approval,” he continued. “It has to be placed in the context of a risk-benefit analysis,” which includes the agent’s safety profile, context of the disease setting and existing therapies, among other factors.

Seeking to turn the tables on Genentech, the oncology review division head suggested the company should look to its own data for the answers it seeks from FDA.

“If you’re looking for what is the PFS magnitude that we would be looking for … I would refer you to the ODAC meeting of December 2007, the original ODAC meeting, where your own consultants were here and were exulting the benefits of a 5.5-month advantage in PFS, as well as the corresponding hazard ratio that went with it,” Pazdur said.

If data confirmed either the median PFS benefit or hazard ratio seen in E2100, “we would not be here today,” CDER’s Pazdur said.

Under further questioning by Genentech, Pazdur said: “I believe if we had data that confirmed either the median or the hazard ratio of the original E2100 we would not be here today.”

“An Honest Person’s View” Of The Data

ODAC member Mikkael Sekeres, Cleveland Clinic, said Genentech should have known the AVADO and RIBBON1 results were not enough to confirm the benefit seen in E2100.

“I find it difficult to believe that anybody, if you just taken an honest person's view of this, would think that … as long as I have a significant progression-free survival, that should be enough to satisfy the FDA and to satisfy ODAC for full approval,” he said.

“We can debate exactly who said what at the original ODAC meeting and what was communicated between the FDA and the company, but an honest person would take a step back and say, ‘You know, they’re probably looking for about the same magnitude of progression-free survival.’”

Looking Beyond Avastin

Debate over the merits of hazard ratios versus median PFS values is not a subject unique to Avastin or breast cancer.

In research presented at the recent annual meeting of the American Society of Clinical Oncology, Alberto Sobrero, head of the Medical Oncology Unit at Italy’s Ospedale San Martino, presented an analysis of 15 pivotal Phase III trials for nine biological cancer treatments that found that impressive hazard ratios did not necessarily translate into impressive absolute gains in PFS or overall survival.

Both hazard ratios and absolute gains need to be considered in the context of the specific tumor type when making a value judgment about a clinical benefit, he said (“Live From ASCO: Time To Cool Down?,” The In Vivo Blog, June 6, 2011).

At the start of the two-day Avastin hearing, several cancer organizations testified during the open public session to express concerns about a lack of clarity and consistency in FDA’s approval standards for oncologics.

Bob Erwin, president of the Marti Nelson Cancer Foundation, said FDA “should clarify objective metrics of clinical performance, other than overall survival, sufficient for accelerated approval, if necessary, on a disease and stage-specific basis.”

Tim Turnham, executive director of the Melanoma Research Foundation, questioned whether the approval process for Avastin in MBC has been “clear, consistent and reasonable.”

The hearing may have provided these groups with further insight on FDA’s views as to how the PFS magnitude of benefit should be defined, although it seems unlikely the calls for greater regulatory clarity will stop with Avastin.

By Sue Sutter