Before the Conclusion | After Approval: Medicines Reconsidered, Case Four
If a cancer medicine can keep a tumor from worsening for longer, that may certainly have value. But if it has not been shown to help patients live longer or clearly improve their quality of life, and carries serious side effects, how should we decide whether that “delay in progression” is worth it?
The US breast cancer indication for bevacizumab, sold as Avastin, once brought this question before patients, researchers, and regulators.
In 2008, the FDA granted accelerated approval for its use with paclitaxel in HER2-negative metastatic breast cancer in patients who had not received chemotherapy for metastatic disease. In 2011, the FDA withdrew this indication while retaining its other cancer uses. This history therefore needs to be described precisely as “one use losing approval,” rather than the drug being withdrawn from the market entirely.[1]
The original hope rested on specific research
Bevacizumab is an antibody against vascular endothelial growth factor A, or VEGF-A. Tumors need a blood supply to grow, and interfering with blood-vessel formation is one approach researchers have pursued to control them. But these signals also contribute to normal tissue function, so disrupting them can cause harm as well.[2]
The pivotal study supporting the breast cancer use, E2100, compared paclitaxel alone with paclitaxel plus bevacizumab. The original paper, published in 2007, reported median progression-free survival of 5.9 months and 11.8 months, respectively, favoring the combination. It did not show a statistically significant improvement in overall survival.[2]
Progression-free survival, or PFS, is generally measured from randomization until disease progression or death. It is neither “time cured” nor the same as overall survival. The study provided a comparison of disease control, rather than establishing how much longer every individual patient lived.
That does not mean PFS has no value. If progression brings pain, organ damage, or a more difficult next round of treatment, delaying it may matter a great deal. What needs to be established is whether the size of the delay in this trial is reliable, whether it comes with benefits patients can feel, and how much it costs them in harm and burden.
In 2008, the FDA accepted this favorable signal, granted accelerated approval, and required further studies to verify clinical benefit. The original approval itself therefore carried an obligation to continue testing.[1]
The later studies did not simply say “no effect at all”
The later AVADO and RIBBON-1 trials also observed improvements in PFS, but their overall magnitude was smaller than the expectations raised by the original trial. They did not provide evidence of longer overall survival, while serious adverse reactions still had to be weighed.[1,3,4]
This matters. If we describe the later results as “failure on every efficacy measure,” we cannot understand why the dispute continued. The company and supporters of continued use could point to improved disease control. The FDA had to decide whether the size and certainty of that improvement justified patients' continued exposure to risk.
The chemotherapy partners also differed between these studies. Different combinations and study conditions may affect the observed results. We therefore cannot simply assemble the medians from several trials as if they were repeated measurements made under identical conditions.
Yet because the later evidence came from different studies, the initially large-looking effect also had to be reassessed within the evidence as a whole. Medical judgment cannot retain only the most positive study and explain away every other comparison as an imperfect exception.
Why the FDA found insufficient grounds for continued approval
Bevacizumab's risks include severe hypertension, bleeding, heart-related harm, and gastrointestinal perforation. In its 2011 decision, the FDA explained that the later studies showed a smaller effect on delaying tumor growth, with no evidence that patients lived longer or had better quality of life. It concluded that these findings were insufficient to outweigh the serious risks.[1]
This was not a decision based solely on checking a P value. A difference can reach statistical significance while remaining too small to justify the costs of treatment. Conversely, a nonsignificant difference in overall survival cannot be translated directly into proof that no patient could possibly benefit.
The question was whether, on all available evidence, this particular regimen still offered sufficiently credible and important net benefit in this group of patients. The magnitude of the PFS effect, serious toxicity, quality of life, and available alternative treatments all formed part of that question.
After a hearing and public debate, the FDA made its final decision on November 18, 2011, withdrawing the breast cancer indication. The decision addressed the evidence and benefit–risk assessment for that specific use.[1]
Does a lack of proven survival benefit always rule out use?
It is easy to reduce this case to an overly simple rule: if a cancer drug does not extend overall survival, it is not worth approving. That formulation loses the genuinely difficult part.
Overall survival is a very important endpoint, but in some diseases mature results take a long time to obtain. Subsequent treatments can also affect comparisons between groups. If a medicine reliably reduces suffering, prevents major complications, or substantially delays clinically meaningful deterioration, those outcomes may also constitute patient benefit.
The bevacizumab dispute therefore asks us to examine what PFS actually represented in this setting: where progression occurred, whether symptoms accompanied it, how large the delay was, whether the results held up under different assessments, and how much additional burden treatment imposed.
A scan showing that a tumor has not yet met progression criteria cannot, by itself, tell us how the patient is living. Quality-of-life measurement thus has value of its own. But a lack of clearly measured improvement cannot be packaged as “patients definitely feel better,” and benefits the study did not establish cannot be treated as established facts.
My view is that the more we rely on an endpoint that represents patients' circumstances indirectly, the more carefully we need to explain its relationship to symptoms, function, and survival in that particular disease. Research need not always wait for the same endpoint, but every alternative basis for judgment needs appropriate justification.
Why Europe and the United States made different choices
There is another branch of this history worth discussing. In its 2010 review, the EMA judged the benefit–risk balance of bevacizumab plus paclitaxel to remain favorable and retained that breast cancer treatment option, while reaching different conclusions about other chemotherapy combinations. When checked for this article, the EMA's medicine record still listed first-line metastatic breast cancer treatment in combination with paclitaxel.[5,6]
This shows that regulators considering similar evidence may weigh the clinical value of disease control and acceptable risk differently. We can compare their reasoning, but the fact that one is more permissive or more restrictive does not by itself establish that the other disregards science.
Nor does disagreement mean every conclusion is equally valid. What we can meaningfully examine is which studies each side relied on, which endpoints it emphasized, how it weighted the magnitude of benefit against risk, and whether it required further verification.
For readers, this is also a reminder to notice where a news event occurred. Saying “the breast cancer indication was withdrawn” without specifying the United States can suggest a uniform decision across the world.
Why one molecule needs several clinical answers
Bevacizumab's approvals in other cancers cannot automatically supply sufficient evidence for its use in breast cancer. Tumor biology, treatment combinations, and alternatives differ, and patients may obtain different benefits.
Likewise, losing the US breast cancer indication cannot automatically erase its value in other uses. The boundaries of a medicine's evidence must be drawn separately for each disease, patient population, combination, and stage of treatment.
This case ultimately returns us to a specific question: compared with other feasible regimens, what additional benefits do patients obtain by receiving this treatment, and what additional burdens do they bear? An impressive PFS number is part of the answer, but other evidence is needed to understand what it means for people.
For people already receiving treatment, an indication change also raises questions about prescribing, payment, and treatment options. Those questions differ from whether compensation for injury is available. A withdrawal decision alone cannot tell us that all of them have been resolved. The concluding article, After a Drug Is Withdrawn, Who Bears Patients' Losses?, will examine these different consequences.
The next article turns to Mylotarg. Why did a drug that left the US market over subsequent trial results and safety concerns return seven years later? This time, the change involved both the evidence available and the way the medicine was dosed and used clinically.
Principal sources
[1] FDA, November 18, 2011: Explanation of the final decision withdrawing Avastin's breast cancer indication. Covers the 2008 accelerated approval, subsequent verification, risks, and retention of other cancer uses.
[2] Miller et al., The New England Journal of Medicine, 2007: The E2100 randomized trial. The 11.8- and 5.9-month figures in the text use this paper's analysis and are not mixed with medians from later independent review or other data cutoffs.
[3] FDA, December 2010: Evidence-review memorandum recommending withdrawal of the breast cancer indication. Includes the interpretation of AVADO and RIBBON-1.
[4] Robert et al., JCO, 2011: The RIBBON-1 randomized trial.
[5] EMA, December 16, 2010: Conclusions of the review of bevacizumab in breast cancer. This is a historical decision; its conclusions about other combinations should not be treated as the complete current indication list.
[6] EMA: Avastin's current medicine record and indications.
Sources checked through October 28, 2026. This article is for the exchange of medical and scientific information and is not individual medical advice. Discuss specific treatment changes with a doctor.
Before the Conclusion: there are still questions worth pursuing.
Medicine · Science · Ideas and Reading
Follow Before the Conclusion to stay curious and think independently together.
Text and image use statement
The text of this article was written originally by Before the Conclusion. You are welcome to share the article's web link. Without separate permission, please do not republish or excerpt its text by copying, reposting, publishing screenshots, or similar means. Unless otherwise indicated, its illustrations may not be separately republished, copied, modified, or used commercially. Rights in cited materials belong to their respective rights holders.