Last month, an FDA advisory committee voted 12 to 1 to recommend the agency withdraw approval for the multi-billion-dollar Avastin cancer drug to be used to treat breast cancer. The move came after two recently released studies - which were undertaken as a condition of approval - found that patients given Avastin and chemotherapy didn’t survive longer than those given chemo alone. And Avastin patients also suffered more serious side effects.
Two years ago, an FDA panel voted against approval (see here), but the agency endorsed treatment for breast cancer anyway. And later this week, the FDA is expected to decide again what to do about the breast cancer indication for the drug, which is already approved to treat colon, lung and brain cancer. The impending decision is sparking heated debate about the wisdom of the FDA’s accelerated approval program and the extent to which progression-free survival is a worthy endpoint. We spoke with Barbara Brenner, a breast cancer survivor and the outgoing executive director of Breast Cancer Action, a patient advocacy group, about the implications…
Pharmalot: Two years ago, you were adamantly against FDA approval for the breast cancer indication and now two recently released studies appear to offer some vindication. What do you make of that? Brenner: We argued in 2008 that the study on which the FDA was relying was insufficient for its own purposes to figure out efficacy and safety. Under the Bush administration, there were two registration trials needed for approval. But somewhere along the line, one became enough. And the results of one trial can be an accident, which is why we need two confirming trials. These subsequent studies – one of which was called Avado – were under way already. But the FDA set it up so that (the company) had to do these studies, under accelerated approval. And they were wrong to do so. If they had waited a year, when Avado came out…they would have seen that this drug did not do what they thought it could do for breast cancer patients.
Pharmalot: What exactly are you referring to? Progression-free survival? Brenner: Yes, that was all they had - a nine-month improvement with Avastin in a clinical trial, but no improvement in survival and no data about quality of life. PFS is just a surrogate marker and we use those, you know, because it takes a long time to get survival data. But everyone’s impatient, including drug companies…So things are used that we can measure to stand in, but surrogates are only useful it they are actually good indicators of the thing you really want to look for. For Avastin, it doesn’t indicate anything about survival, which should be the gold standard for drug approval…Every drug will be a little different and every cancer will be a little different. I think it’s possible that some drugs will be developed for which PFS is a good surrogate for survival, but just not this drug.
Pharmalot: So you’re saying the FDA needs to re-evaluate its approach, yes? What about the people for whom the drug is working, though? Brenner: There’s a big focus at the FDA on surrogates and markers, but these are only valuable if they stand in for what you’re looking for. So let’s not get too excited about surrogates and markers until we know they can accurately predict what you hope to find. I think they’re right for some patients, if this drug works for them. But almost every drug that was ever approved worked for somebody. But the FDA’s job is to protect the population at large. It’s not my personal FDA. It’s the country’s FDA…That’s not the basis on which we must make these decisions on a policy or federal level – to show one person benefitted. That’s not the test. There could have been 100 people harmed and that’s what we have to look at…It’s their job at the FDA that when a drug comes to market that they determine it will help the population on a wide basis more than they hurt…
I hope people who have been on Avastin and believe they do well on the drug will continue to do well….But we have to keep another thing in mind and that is Avastin is not coming off the market. It’s on the market for colon and lung cancer, so doctors will prescribe on and off label…They were prescribing it for breast cancer before it got that label…So don’t tell me they won’t prescribe off label. They do and they will. They believe what they believe. I hope nobody dies. The drug has serious side effects…the continued use of this drug in early stage disease will have some very serious consequences…
Pharmalot: So what’s your take on the accelerated approval program? Brenner: In general, it’s bad for cancer drugs. Just look at the avastin example. Whatever you think of the drug, assume for a minute they were right to approve it. But then you get more data and a lot of people are on it because it’s been heavily marketed. For two years, a lot of people were getting the drug. Some will do fine, but then you get more data, and the FDA looks at it and maybe says not good enough and then pulls the label. So you’ll have outrage and that would be appropriate. If we don’t know it works, why should it go on the market? Wwe’re talking about breast cancer. We’re not talking about pancreatic cancer, a disease for which there are hundreds of treatments. I just don’t think the FDA should ignore that context…It’s a set up that’s bad for patients who come to rely on things they then later can’t get. And in cancer it’s just too hard to do things that fast.
Pharmalot: On a related note, Genentech and Roche were unexpectedly surprised when the FDA refused to accept the application for their T-DM1 (background here). What do you make of that? Brenner: This was a single armed stage II trial with very few patients. Success was measured using PFS instead of overall survival rate. The FDA has to insist on Phase III trials using controlled trials that measure success by standards that matter to patients. That standard is improved quality of life or overall survival rate…Instead, this was a single-armed Phase II trial. They have PFS data and no survival data, but wanted accelerated approval. You cannot do that. We understand there’s a group of patients who need something they don’t have, but don’t give them something that doesn’t work. I understand you’re dying and there is a drug that may or may not help for which a whole lot of money is charged at the end of your life. I understand why patients act as they do, but drug companies should do something for the patient – either extend life or improve quality of life. I just can’t believe Genentech is continuing to pursue this past.
7 Comments
Genentech should use a cell-based functional profiling assay like the AngioRx to identify a potential target population of breast cancer patients that is thinks will benefit from the drug and then conduct a randomized clinical trial among this group.
Maybe this is not only a drawback to accelerated approval program but a drawback to the clinical trial system, a scientifically bankrupt paradigm which has not fostered either efficient or humane progress in cancer chemotherapy.
The problem is not with using the prospective, randomized trial as a research instrument, the problem comes from applying this time and resource-consuming instrument to address hypotheses of trivial importance: do cancers prefer Coke or Pepsi?
My 2 cents: It is not a question of all or none. SOME patients can probably benefit from being treated with Avastin. How about letting the attending physician/ expert call it?
You think the attending physician is going to know anything? For any drug? And in this case, certainly not more than Brenner and Breast Cancer Action.
It is very clear that some medications benefit some patient populations more than others. This is most aparent in cancer where what looks like the same disease may have different genetic causes. Through genetic profiling, the concept of personalized medicine can be applied. The result would be identification of patient populations most likely to benefit from specific therapeutics - including Avastin.
With advances in genetic sequencing, the concept of true personalized medicine is coming closer to reality.
Genetic profiling is testing (theoretically) candidates for a mutation-targeted therapy. It cannot test sensitivity to any of the mutation-targeted therapies. Genetic (molecular) profiling is testing for gene expression, whether or not a gene is expressed, meaning is RNA being make from the gene. So gene expression assays can be either probing for the specific RNA messengers (mRNA) or it can mean looking for the proteins themselves.
If you believe in the molecular approach, you at least need a gold standard. You have patients not just with breast cancer, but with virtually all solid tumors who are potential candidates for VEGF targeted therapy. How are you ever going to validate a predictive test for colon, lung, pancreatic or melanoma? Do single agent trials correlate assay results?
There are advantages in having cells which you can define as mutation-targeted sensitive or resistant and studying what exactly is different. Then there is testing to circumvent resistance. Then there is testing combinations.
The importance of mechanistic work around targeted therapy as a starting point should be downplayed in favor of a systems biology approach were compounds are first screened in cell-based assays, with mechanistic understanding of the target coming after validation of its impact on the biology of the cancer cells.
Many of these targeted drugs cry out for validated clinical biomarkers to help set dosage and select patients likely to respond. Optimal and reproducible targeted testing continues to evade the diagnostics of the disease. Numerous genes, tumor, and patient factors contribute to the risk of the cancer coming back and the effectiveness of chemotherapy.
It could be vastly more beneficial to measure the net effect of all processes (systems) instead of just individual molecular targets. The cell is a system, an integrated, interacting network of genes, proteins, and other cellular constituents that produce functions. One needs to analyze the systems' response to drug treatments, not just one or a few targets (pathways/mechanisms).
There are many pathways/mechanisms to the altered cellular (forest) function, hence all the different "trees" which correlate in different situations. Improvement can be made by measuring what happens at the end (the effects on the forest), rather than the status of the indivudal trees.
Genetic profiling of populations that both respond and do not respond to specific therapies may reveal information that can guide treatment regimens for future patients. Certainly this approach would require clinical studies that allows evaluation of both responders and non-responders. I am not suggesting mutation-targeted therapies. There are already examples of such treatments relative to use of herceptin for breast cancer.
A March 16, 2010 study in the Journal of the National Cancer Institute looked at the value of various gene tests and concluded none of the studies showed a clear usefulness.
Genomic tests provide lots of information about a patient's genes. However, the Journal article pointed out, there are so many sequences in our DNA which influence disease, that attempting to unravel such complexity just produces more and more information without a particularly useful benefit. While genes may provide a recipe, they do not determine the end results and cannot predict how an individual will respond to a specific treatment.
Like the various influences on a flower seed that cause one blossom to turn out differently from another, there are biological processes in the body that affect the development of cancer in each patient and determine how that patient's cancer cells will uniquely react to treatment.
Despite its allure, the "genetic" path is not all that personalized. Treatment based on genetic testing is still a guessing game. Only a treatment regimen based on a "functional profile" - a real-time test of chemotherapy on the actual cancer tissue - can predict with accuracy an individual's response to chemotherapy.