FDA's Avastin Breast Cancer Withdrawal Sets New Standard, Deters Drug Development

FDA’s proposal to withdraw the approval of Avastin for MBC is based on a new standard for the efficacy showing that a therapy must make in first-line MBC. The Agency previously -- and appropriately -- followed a more flexible approach based on data showing a medicine’s effect on PFS [progression-free survival] or other disease progression endpoints, together with data on OS [overall survival].

Is FDA requiring Avastin to beat the drug's own preliminary results in progression-free survival in metastatic breast cancer and exceed the performance of other drugs already approved for the indication? That is the regulatory challenge that the company believes it is facing. Roche/Genentech says that FDA's new standard for Avastin in breast cancer is too demanding and ill-defined and will discourage future sponsors.

Sandra Horning, Michael Labson, Paul Schmidt

On January 16, Genentech formally requested a hearing on FDA's proposed withdrawal of the accelerated approval for Avastin in metastatic breast cancer. The following excerpt from the request for a hearing was written Genentech Senior VP Sandra Horning, MD, and outside counsel from Covington & Burling, Michael Labson and Paul Schmidt.

The filing has been edited, with new section headlines, citations and footnotes removed.

FDA's proposal to withdraw the approval of Avastin for metastatic breast cancer (MBC) is based on a new standard for the efficacy showing that a therapy must make in first-line MBC. The Agency previously -- and appropriately -- followed a more flexible approach based on data showing a medicine's effect on progression-free survival (PFS) or other disease progression endpoints, together with data on overall survival (OS).

Now FDA has stated that improvement in PFS is not adequate unless it is of a large (but unspecified) magnitude (such as the substantial 5.5-month improvement observed in E2100 [a multicenter, randomized, controlled open-label Phase III study on more than 700 patients, sponsored by National Cancer Institute and conducted by Eastern Cooperative Oncology Group]) and that otherwise there must be an improvement in OS to support approval.

FDA's newly articulated standard carries potential risks for the future availability of medications for the first-line treatment of patients with MBC.

No other drug has shown a magnitude of effect on median PFS as large as seen for Avastin with weekly paclitaxel in E2100. At the same time, there are substantial scientific obstacles to the detection of statistically significant improvements in OS in the first-line MBC setting. Existing drugs for first-line MBC currently on the market were not approved on the basis of such a standard, and few drugs for first-line MBC in the future will be able to make such a showing.

The risks to innovation from the standard FDA is now imposing are further exacerbated by the lack of clarity in FDA's current approach. Genentech recognizes the need for a certain degree of case-by-case evaluation based on the particular circumstances presented by a given compound and dataset. However, FDA is requiring a large magnitude of improvement in PFS without any statement of what particular magnitude must be shown.

FDA is also imposing this requirement without having presented a proposed framework for feedback to the broader stakeholder community. The Agency is applying this new standard to Avastin without having communicated it in a transparent manner to Genentech during the discussions that outlined the conditions for converting from accelerated to full approval in MBC.

Roche Says Prior FDA Approvals Based on Less Demanding Standard

FDA has accepted disease progression endpoints—such as PFS and time to progression (TTP) —as the basis of regulatory approvals in MBC without requiring a specific or large magnitude of effect as large as that observed in E2100 or a demonstration of OS benefit.

The following table identifies FDA-approved agents in MBC, the endpoint supporting approval, and the magnitude of effect shown. ( See Exhibit 1.)

Specifically in the first-line MBC setting, drugs have received full approval on the basis of an effect on disease progression well below that seen in E2100 with no established statistically significant improvement in OS.

FDA has repeatedly drawn a distinction between the efficacy data for Avastin and the data that underpinned approval of Gemzar and Hereceptin (trastuzumab), citing Gemzar and Herceptin as examples of recent approvals in first-line MBC that were purportedly supported by both OS and PFS improvements. FDA made this assertion at the July 2010 ODAC meeting that considered E2100, and FDA repeats it in the Decision Memorandum (from Dr. Richard Pazdur, Director, Office of Oncology Drug Products to Dr. Janet Woodcock, Director, Center for Drug Evaluation and Research, December 15). The data do not support this distinction.

Gemzar, in particular, does not meet the criteria that FDA is now setting forth – i.e., a large magnitude of PFS benefit and/or a statistically significant benefit in OS. Herceptin was pproved on the basis of a TTP benefit (4.2 months with paclitaxel) larger than that of Gemzar, but approval was not supported by a statistically significant improvement in OS at the time of its approval, and in any event is in the distinct HER2-positive patient population.

The following table presents the efficacy results from the pivotal studies supporting the initial approvals of Avastin and Gemzar. ( See Exhibit 2.)

When Gemzar was approved in 2004, the OS data that were available at the time of approval were based on an interim analysis and the p-value associated with the hazard ratio (0.049) did not reach statistical significance for the interim review. More importantly, when the OS data matured, the effect on OS continued to be not statistically significant. Thus, the same conclusion must be drawn from the OS data supporting the current Gemzar approval and the OS data from E2100 -- i.e., the hazard ratios favor the test arm, but the effects on OS are not statistically significant.

The divergence between the standard that FDA has articulated for Avastin and the standard that FDA applied to Gemzar is especially striking because FDA granted full approval to this other therapy.

FDA now proposes to withdraw accelerated approval for Avastin, even though Avastin's use in first-line MBC is supported by data that are comparable to the data upon which FDA maintains a full approval for Gemzar. This disparate treatment of Avastin underscores the risk to development presented by FDA's approach to Avastin, and this approach is contrary to FDA's regulations.

If the standard that FDA seeks to apply to Avastin were applied to Gemzar, it arguably would not be approved today. Going forward, the standard may unduly limit future approvals in this area of significant unmet medical need. The disparate treatment of Avastin and Gemzar also fails a basic tenet of administrative law requiring an agency to treat like cases in a like manner.

Disease Progression Endpoints Were Appropriate

PFS should be an accepted objective measure of treatment effect in first-line MBC reflecting a direct clinical benefit. PFS measures tumor control and facilitates clinical trials with reasonable sample sizes and follow-up times.

PFS can be reliably measured using radiographs, and potential investigator bias in assessing PFS can be minimized by implementing a placebo control and/or employing an independent review of tumor assessments (as was done in E2100).

FDA itself has recognized in guidance that PFS can be an endpoint for approval because PFS reflects tumor control, can be assessed before the determination of a survival benefit, is not confounded by subsequent therapy, and is generally based on objective and quantitative assessment. The European Union's Guideline on the Evaluation of Anticancer Medicinal Products in Man (CPMP/EWP/205/95/Rev.3/Corr.2) recognizes PFS as a primary endpoint for drug approval in oncology clinical studies.

Most fundamentally, PFS is a valid measure of direct clinical benefit because prolonging the time to disease progression is clinically meaningful for patients.

As noted by Dr. Marisa Weiss on behalf of Breastcancer.org, "[t]hrough ten years of dialogue with women diagnosed with metastatic breast cancer, Breastcancer.org has learned a great deal from these women about their feelings, perspectives, needs, and experiences with the disease and treatments … Longer progression-free survival is a meaningful benefit for many women." Because MBC is an incurable disease, the therapeutic goals for patients with MBC focus on maintaining tumor control, extending the time to subsequent progression, and delaying the introduction of subsequent lines of therapy. PFS is inherently relevant to these goals.

Lengthening PFS delays the onset of disease-related symptoms and the side effects from new therapy. It also avoids the psychological consequences associated with disease progression and changing therapies, and eliminates uncertainty as to whether additional treatments will be effective.

Finally, the clinical significance of PFS effects can be supported by other related measures, such as 1-year survival. One-year survival reduces confounding of treatment effect by use of subsequent therapies, making it less subject to dilution of treatment effect than OS in an indication in which post-progression survival is long; and 1-year survival may result in faster access to Phase III results and approval of new therapies. A 1-year survival result supporting a positive PFS effect also reflects safety and is clinically meaningful in the first-line setting.

Demonstrating Direct Survival Effect is Increasingly Difficult

The requirement of an OS improvement as a measure of drug effect instead of PFS or other disease progression endpoints risks discarding a clinically meaningful treatment result in the first-line MBC setting. Additionally, although Genentech recognizes and agrees that the ultimate goal of treatment is to prolong survival, there are fundamental challenges with demonstrating that a first-line treatment for MBC improves OS.

The administration of second-line and beyond anti-cancer treatments, with their associated safety and efficacy effects, following a patient's initial disease progression creates one challenge. Once a patient's cancer progresses while on treatment, it is generally necessary on ethical grounds to permit the patient to be treated with whatever therapy the treating physician considers best available care.

For purposes of the clinical trial, the administration of subsequent treatment lines outside the original, controlled study treatment arms may obscure the effect of the initial treatment on the patient's ultimate survival. This is particularly true where the duration of overall survival is long relative to the period of progression-free survival.

A systematic review of 76 randomized clinical trials for MBC, including 45 studies that enrolled only first-line patients, showed that approximately two-thirds of the overall survival time for MBC patients is accounted for by post-progression survival (i.e., survival after initial disease progression). Stated differently, only about one-third of the duration of an MBC patient's overall survival occurs prior to progression on first-line treatment.

Notably, the clinical trials that have shown an OS benefit in MBC were more likely to enroll patients who had progressed on prior therapy for metastatic disease. Of the 13 trials that had published improvements in both PFS and OS, 9 enrolled patients who had previously received treatment for metastatic disease, whereas only 4 of the trials included only first-line patients.

Another challenge is provided by the fact that it is not typical to observe a larger absolute improvement in OS than in PFS with any new therapy, and this effectively translates into a smaller percent improvement in OS that is more difficult to detect.

In other words, one would expect the magnitude of prolongation in OS in months to be similar to the magnitude of improvement of PFS in months. Actual clinical trials in first-line MBC confirm this to be a valid assumption, as the duration of post-progression survival (time from median PFS to median OS) has been similar or slightly longer in the control arm than in the corresponding experimental arm in 33 randomized studies in MBC. Furthermore, median OS in recently conducted first-line studies of patients with HER2-negative MBC is 24 months or longer, resulting in median post-progression survival of 18 months or more.

OS Benefit Studies Might Require 2300 patients; 7.5 Years

These challenges have made it increasingly difficult to conduct a controlled clinical study powered to show an effect on OS. Specifically, Broglio and Berry estimated the impact on sample size to detect an OS benefit assuming that a novel agent has a 3-month improvement in PFS (from 6 to 9 months), with no further incremental survival improvement beyond progression, under varying assumptions of the

duration of survival after progression. They found that longer duration of survival after progression led to greater dilution of the OS hazard ratio, and a study designed to detect this OS benefit with conventional statistical power would need to have 1050 to 2440 patients if the median survival after progression was between 12 and 24 months.

Assuming a typical patient accrual rate of 30 patients a month, this could result in a study that requires more than 6 years to accrue, 5 years longer than a typical 600-patient study designed to detect a PFS benefit. Similarly, simulations performed by Genentech, under the assumption of a larger 4-month treatment benefit in PFS and median duration of post-progression survival of 18 months (typical of studies of chemotherapy for first-line treatment of HER2-negative MBC), found that detection of an OS effect could require up to 2300 patients and take up to 7.5 years to deliver OS results. These illustrative examples demonstrate the increasing difficulty of conducting a controlled clinical study in the first-line setting that is powered to show an effect on OS.

The data and analyses presented above show that although OS remains the ultimate goal of MBC treatments, there are substantial scientific limitations on the ability to detect a significant OS benefit in studies of first-line treatments for MBC.

In addition, strict adherence to an OS improvement standard risks discarding clinically meaningful treatments. For example, drugs that have been approved for MBC over the last 20 years have contributed to cumulative survival benefits, although the magnitude of improvements in disease progression endpoints for these drugs has been modest. Over the past decade, incremental improvements in survival have been achieved in the MBC setting.

This improvement in survival has occurred despite the fact that few randomized trials have demonstrated more than a modest effect on disease progression or an effect on OS. That is, few individual therapies have been shown in clinical trials to confer a disease progression benefit with a magnitude comparable to that observed in E2100, or a clear OS benefit, but median survival of MBC patients has nonetheless improved.

This analysis indicates that it may not be necessary to require a direct showing of increased OS for individual agents in order to extend survival ultimately for MBC patients. In addition, to the extent that FDA's new proposed efficacy standard would prevent the future approval of drugs based on comparable showings to those made in past years, FDA might jeopardize the extension of the survival gains that have been observed going forward. Whereas therapeutic agents came to market previously with more modest showings of clinical benefit but nonetheless contributed to the overall improvement in outcomes for MBC patients, new agents may never make it to the market or be allowed to stay on the market under FDA's new standard.