Lessons For Drug Development From The Aveo Buzzsaw
Drug development is a tough business. Small companies that dare to try often end up in public train wrecks when they head to Washington to meet the FDA. It’s a shame, a colossal waste, every time a company gets that far and stumbles. And it doesn’t have to be that way.
The folks at Cambridge, MA-based Aveo Oncology (NASDAQ: AVEO), and their partners at Japan-based Astellas Pharma, know the feeling. They suffered last week as an FDA advisory panel voted 13-1 against Aveo’s application to start selling tivozanib to treat kidney cancer in the U.S. The drug had passed a large, well-controlled clinical trial and appeared to offer a small, but meaningful benefit for patients. Now, partly because of some mistakes the company made, and longstanding flaws in the way drugs get developed, it looks like the FDA is going to kick this drug to the curb. This painful verdict came down after 11 years of R&D and more than $320 million of investment.
Although Aveo appears to have fallen into some common errors that little companies make when they look to save time and money on trials, the experience truly exposes some deeper dysfunction in how clinical trials get conducted. The only way our society is going to get out of this bind, and start getting more of the cost-effective cancer drugs we demand, is by making some changes to the way we evaluate safety and effectiveness, and the ultimate bargain that’s being made with patients who participate in clinical trials.
Consider the Aveo experience. It had good reason to think it had a chance of winning FDA approval. It ran a 517-patient clinical trial that randomly assigned patients to get its drug or a different FDA-approved product. This was no straw man trial, designed to beat a mere lame placebo. Aveo’s drug kept tumors of kidney cancer patients from spreading for a median of 11.9 months, compared with 9.1 months for those on an active comparison agent—Bayer and Onyx Pharmaceuticals’s sorafenib (Nexavar). Side effects for the new compound were minimal. Aveo hit the main goal of the study, so it was time to go the FDA and ask for approval to hit the U.S. market.
While that all looked fine and good, Aveo’s application got ripped apart by the FDA advisors and staff for a couple of reasons. One is that too many patients—88 percent in the pivotal study—came from Central and Eastern Europe. Only about 8 percent of the patients in the key study came from North America and Western Europe, according to an FDA staff briefing document. That’s a problem, because the standard of care is different in the U.S. than it is in Eastern Europe, where patients tend to be poorer, medical standards are different, and there’s less access to advanced targeted drugs. That means it’s hard to extrapolate that a drug which performs a certain way there will perform the same way here.
The other problem in the Aveo application is that cancer drugs need to show they can extend lives, not just hold back tumors for a couple extra months. When the final survival numbers were tallied, they contradicted what was seen on the tumor progression scores. Patients lived a median of 28.8 months on the Aveo drug, and 29.3 months in the control group.
How could that be, that a drug which slows down tumors doesn’t help extend life? This happens occasionally in clinical trials, and sometimes the reasons elude scientists. But there was another explanation here, a quirk in the study design in which patients were allowed to “cross over.” What that means is that once patients were in the study, and saw their disease worsen, they were able to “cross over” to get another drug that might help.
Turns out that almost two-thirds of the patients who originally got the Bayer/Onyx drug (63 percent) ended up getting the Aveo drug later. What about the poor folks who were assigned the Aveo drug in the first place, and then had their disease worsen? They basically got nothing. Only 16 percent ended up getting any extra therapy after their disease worsened, according to the FDA staff document.
Now, no one will ever know for sure why those patients in the control group ended up living longer. Aveo advanced the notion that they lived longer because the Aveo drug works, and so many of the control group patients ended up getting it later. That might be right, but there’s no way of knowing for sure from this study.
The “crossover” clinical design has always had its critics. The purpose of running most cancer clinical trials is to get a clear answer of whether Drug A is better than Drug B. But, there are human beings involved in a study, and they must be treated ethically. It’s not ethical to keep a dying patient on a placebo when you know they could get potentially life-extending drug, all in the name of keeping your statistical database clean.
The thing that doesn’t make sense here is that Aveo already had an ethical study design from the start, when it offered patients one good drug (Bayer/Onyx) or the chance to get another drug that might be better (Aveo’s). There was no need to offer patients a “cross over” option to get the Aveo drug later. In fact, the original pivotal study protocol Aveo discussed with the FDA in May 2009 didn’t contain a “cross over” provision, and this item was only inserted later as an amendment to the protocol, according to the FDA. It’s hard to understand why a company might make such a change.
Aveo wasn’t saying much after the FDA panel, and didn’t take questions from investors or the media. It offered some statements about how it was dismayed by the FDA panel vote, and isn’t ready to throw in the towel.
By now, there are a few obvious questions. Why did Aveo feel the need to tilt its trial so heavily to sites in Eastern Europe? Why did it change the design to allow the patients to “cross over” in the middle of the study, running the risk of muddying up its survival statistics? Was it under so much deadline pressure from shareholders that it offered the “cross over” option to patients to help speed up patient recruitment? Why didn’t it follow the FDA’s advice in May 2012 to run another clinical trial with a set of patients more directly comparable to those found in the U.S? Why didn’t more patients who got the Aveo drug first end up crossing over to get the Bayer/Onyx drug later?
We might never the true answers to those questions. But many other biotech companies have been down this road before, and confronted the complex set of factors around time, money, and ethics that govern how clinical trials get done—both in the U.S. and overseas.
Cancer drug developers are often told by contract research organizations (CROs) that they can get a big clinical trial done for one-tenth the price in Eastern Europe compared with sites in the U.S., says David Miller, president of Biotech Stock Research in Seattle. These CROs also promise that they’ll be able to recruit patients much faster than in the U.S., and as any businessperson knows, time is money. Many cancer patients in Eastern Europe and Russia are poor, and have limited treatment options. When someone comes to them and says they can get a good treatment, or a potentially better experimental treatment, for free, that’s a compelling offer. It’s part of the reason companies are often looking to enroll patients in China, Brazil, and elsewhere around the world.
Contrast that to the U.S., where our clinical trial system has its own problems. Costs are sky-high. Patients, those with insurance anyway, still often get the full kitchen sink of treatments before getting to the end of the line and considering enrolling in a clinical trial. Patients are often reluctant to participate, especially if there’s a chance they might get a placebo.
The U.S. also has some of the world’s best medical centers, which are great places to run clinical trials. But Big Pharma companies run so many studies there that for certain diseases like breast or prostate cancer, they have essentially got the first shot to recruit from a small pool of patients. The Big Pharma pig pile at these medical centers—in which top medical centers and leading investigators have all kinds of financial conflicts of interest—makes it next to impossible for upstarts like Aveo to get in the door and do business in a timely manner.
While most everybody complains about high drug prices, this is one of the factors driving up the cost of R&D that gets little notice. The clinical trial system is broken in the U.S. The FDA often sends confusing, shifting signals about what constitutes valid medical evidence. Doctors outside top medical centers often aren’t equipped, or motivated in many cases, to participate in trials because they’re so busy haggling with insurance companies every day.
The only way we’re going to break out of this jam is to re-think the way trials get done. The prospective, randomized, well-controlled clinical trial may have served us well in the past, giving us a safe and predictable set of information about new drugs. But it’s not perfect. Those standards created a system that’s so costly, and so time-consuming, and corrupt in many cases, that it has created an unsustainable system for new drug development.
A better way to do things would be to take advantage of our growing capacity for using “big data.” Miller, for one, says he can imagine moving to a world in which the gold standard—prospective, randomized, controlled trials—become obsolete. They could be replaced by matched case-control studies, which have long been used in the field of public health.
This way, a patient, with, certain characteristics (55-year-old, non-smoking white male with advanced kidney cancer) could just be offered an experimental drug for free. The sponsor of the study would then go find a patient with matching characteristics somewhere else who’s getting the standard treatment around the same time. You follow them up, compare their outcomes, and keep repeating from patient to patient until an answer emerges.
In a world with truly networked national medical records, and sharing of emerging data from genomics, all the confounding variables that might skew a study could be scrubbed out by supercomputers, leaving a well-controlled body of evidence for a new drug that could be compared with standard of care. Patients would be motivated to enroll quickly in trials, knowing they have a good chance of getting a better drug, not a placebo. Companies would get good data on their drugs early in the game, before blowing huge amounts of money. And if safety problems or wrinkles in a drug’s efficacy profile appear after it hits the market, regulators and physicians would get updates in real-time, enabling them to make adjustments accordingly.
“With big data, that should be enormously easy to do,” Miller says.
I know people have been talking about ideas like this for a long time. Last week, the leaders of UC San Francisco, the FDA, and the National Institutes of Health advanced some similar ideas for improving healthcare under a broad banner of “precision medicine.” Surely, a lot of readers are still thinking this is a trip into fantasy land. We’re a long way from accepting this kind of system today, but it’s not an impossible dream. It’s technically doable. It requires leadership and collaboration among payers, healthcare providers, regulators, and drug companies. It’s time to come up with something different. Otherwise we can keep building up, and tearing down, companies like Aveo while patients wait for something better.