Lessons and Questions from the E2100 Avastin Breast Cancer Trial

by Robert Erwin

The Food and Drug Administration (FDA) granted accelerated approval this past February to biotech company Genentech to market Avastin (bevacizumab) for the treatment of metastatic breast cancer. The FDA defines “accelerated approval” as “a provisional approval with a written commitment to complete clinical studies that formally demonstrate patient benefit.” The word “accelerated” notwithstanding, the events leading up to this regulatory decision were complex and played out over a long period of time. Here is a simplified chronology of some of the key events prior to this decision:

  • 2000—Clinical trial AVF2119g began enrollment (Xeloda with or without Avastin in pretreated metastatic breast cancer).
  • 2001—Clinical trial E2100 began enrollment (Taxol with or without Avastin as first-line therapy for locally recurrent or metastatic breast cancer).
  • 2002—Failure of Avastin to provide significant benefit to patients in AVF2119g trial reported at San Antonio Breast Cancer Conference.
  • 2005—Significant impact of Avastin on progression-free survival of patients in E2100 trial reported at the annual meeting of the American Society of Clinical Oncology. No significant increase in overall survival was reported.
  • 2006—Genentech filed a supplemental biologics license application with the FDA seeking approval to market Avastin for first-line treatment of metastatic breast cancer.
  • 2006—FDA issued a “complete response” letter requiring Genentech to provide additional data and resubmit its application.
  • 2007—Genentech submitted a new application, and the FDA sought input from its Oncologic Drugs Advisory Committee (ODAC). Although an independent review of the E2100 medical records confirmed the significant improvement in progression-free survival, Avastin still had no significant impact on overall survival. The committee voted 5 to 4 against approval.


The arguments for and against approval of Avastin for treatment of metastatic breast cancer encompass a number of considerations. These include the risks of the drug—which are substantial for some patients—the lack of overall survival as the primary endpoint in a clinical trial, and the wisdom of authorizing the marketing of an expensive drug (approximately $100,000 a year) that has not been shown to extend survival nor provide an objectively measurable increase in quality of life. However, an important factor in the debate is the fact that in the E2100 trial, Avastin treatment increased progression-free survival more than any other drug that has been tested so far. Many patients and physicians consider the absence of tumor progression alone to be an important contribution to overall quality of life and thus an adequate basis for regulatory approval.

Ultimately, the trade-offs in using or not using any drug are very personal, individual decisions. Statistics do not apply to individuals, and data from clinical trials can be used only to guide judgment, not to predict the benefit or harm that any specific patient will experience. Despite these limitations in the use of clinical data, are there steps that could be taken to improve the value of clinical trial results to individual patients and increase the confidence with which we make decisions? Let’s consider a few possibilities:

  1. Test the new experimental agent with standard treatments vs. standard treatments alone (with placebo). Most patients with cancer who enroll in a clinical trial are seeking therapeutic benefit and do not want to be randomized to receive a placebo instead of the experimental agent under study. This is true even if the placebo is combined with other currently approved treatments, as is almost always the case in cancer clinical trials. While patients’ desires to obtain access to new treatments are understandable, the aim of a clinical trial is to determine if the experimental drug is actually effective and/or safe, and many treatments tested in clinical trials are later found to be unsafe and/or ineffective. Credible data on important clinical questions, such as whether an experimental treatment improves quality of life or even whether it stops tumor growth, can only be collected through blinded and randomized trials.
  2. Provide a cross-over provision in the clinical trial design. The best way to be fair to trial participants and also speed the accrual of trials is to offer patients the opportunity to cross over to the experimental therapy either after a prespecified time or upon disease progression. Many experts argue that crossover may obscure the benefit of an active agent and make data harder to interpret, but if the experimental agent is active enough to be of real clinical value, this should not be a problem. The value of a significantly effective agent should be relatively easy to document.
  3. Track and obtain medical information beyond initial treatment. When patients are allowed to cross over to the experimental agent, or when there are multiple treatment choices available to patients if disease progression occurs (as is now the case with first-line metastatic breast cancer), it would be logical to track each patient after the treatment has been completed or terminated. Maintaining accurate records of what subsequent therapy was used, including crossover to the experimental agent, would help to clarify whether failure to objectively document a survival benefit was due to use of the experimental agent by patients initially randomized to the standard treatment arm. This was not done in the E2100 trial.
  4. Establish independent third-party verification of endpoint measurements. If the primary endpoint of a clinical trial is potentially subjective, such as radiographic measurement of progression, then it is important to use a mechanism of independent third-party verification of investigator observations. Independent third-party analysis is particularly important for progression-free survival assessment because of the inherent limitations of the technology for detecting progression.1 This should be standard practice in clinical trials designed with progression-free survival, or any other subjective endpoint, as the primary endpoint. This was not done in the E2100 trial until after the FDA demanded such an analysis as a condition for reviewing the marketing application.
  5. Adequately fund the clinical trial. Third-party verification of data adds to the cost of running a clinical trial, but the increased reliability of the conclusions justifies the expense. In addition, ensuring that a large enough number of patients is accrued to the trial to obtain solid statistical results is very important. To run a trial with too few patients to obtain statistically meaningful results is to gather data that cannot be interpreted  and to waste the precious time of trial participants. As taxpayers and potential beneficiaries of cancer clinical trial data, we should demand adequate funding for cooperative group trials such as E2100, whether industry is involved or not.
  6. Adequately pay the investigators’ institutions. Two of the main reasons some clinical trials are slow to accrue patients is that the trial is uninteresting to an informed patient or the physician—unfortunately at times for financial reasons. The amount paid to cover costs per patient can play a significant role in the speed of trial accrual as indicated by the generally faster accrual to industry-sponsored trials than to National Cancer Institute Cooperative Group trials. The difference in accrual rate is probably more the result of the higher amount per patient paid to the medical institution by industry sponsors than with the inherent value of the medical question being explored in the trial. This adds to the cost of a clinical trial, but considering the value of the time that may be lost, the price is justified.
  7. Adequately fund a strong, independent FDA. The physicians and scientists in the FDA responsible for the review and approval or disapproval of applications to market drugs and biologics are probably the best objective defense patients have against sloppiness and dishonesty in the drug development and marketing process. With adequate funding to hire and retain smart, motivated, and well-trained medical reviewers, delays in decision making can be reduced, and the role of politics in determining what products can be sold to patients can be substantially minimized.


An important, complex, and even emotional issue highlighted by the recent FDA decision to approve Avastin for metastatic breast cancer is what the standard should be for allowing the sale of another drug for the treatment of cancer. The benefit of Avastin as measured by the increase in progression-free survival demonstrated in E2100 is considered major progress by some and is considered pitifully inadequate by others.

Some patients clearly will receive benefit from this drug, and for them the FDA made the correct decision, especially if they do not currently have insurance that will pay for off-label use of expensive drugs like Avastin. For other women seeking treatment for newly diagnosed metastatic breast cancer, Avastin is going to cause them harm without providing any measurable benefit. Unfortunately, at the current state of scientific knowledge, it is impossible to predict in advance which patients will benefit and which will be harmed by Avastin or by most other biotechnology products.

Considering that the cost of cancer treatment is rising much faster than the efficacy of cancer treatment, perhaps a final lesson from the E2100 trial is that more money must be invested in developing the technology to identify which drugs will work for which people and to ensuring that the standards necessary to change medical practice are no lower than those required to bring a new treatment to market. Perhaps one way to rapidly stimulate efficient research to address these goals would be to establish a new rule for the sale of cancer treatments—market them like any other consumer product and require manufacturers to offer a warranty or a money-back guarantee. The lure of high prices and profits is drawing a lot of money to the high-stakes world of new cancer drug development. Would a requirement to ensure product performance for each individual, in addition to obtaining impressive p-values in clinical trials, provide an incentive to do the research necessary to prove what works for whom? I know, this seems like a crazy idea—but it’s OK to dream, isn’t it?

Robert Erwin is the president and founder of the Northern California–based Marti Nelson Cancer Foundation, whose primary goal is to help cancer patients by speeding the development of, and access to, significantly better treatment options.


1 See Panegeas, KS, et al., “When You Look Matters: The Effect of Assessment Schedule on Progression-Free Survival,” J. Natl. Cancer Inst. 2007 99: 428-432.


This entry was posted in Articles.