On June 28, 2011, ten police cars descended upon the headquarters of the Food and Drug Administration. Scores of protesters carrying loudspeakers and donning pink shirts had attempted to enter the building, but were thwarted by the officers. One man held a sign that read: “My wife is not a statistic: Save Avastin.”
“I don’t really recognize the agency for which I once worked, because they’re making all these crazy decisions.”
Avastin was one of the world’s best-selling cancer drugs, first approved in 2004 to treat advanced colon cancer, with high expectations. The first drug of its kind, it was designed to block the blood supply of growing tumors, and it appeared to work well for colon cancer and lung cancer. Riding on a swell of optimism, the FDA decided in 2008 that Avastin could also be used to treat a certain type of advanced breast cancer — but just three years later, the agency seemed poised to reverse that decision.
Inside the building that June morning, a public hearing on that decision was being held. Some members of the audience shouted “Don’t take my drugs away!” Others came with photo albums in the hopes of showing the expert panel weighing the drug’s fate how the cancer drug had saved them from the ravages of breast cancer, and allowed them to live longer, fuller lives.
The pleas didn’t work. The panel concluded that Avastin didn’t improve chances for surviving breast cancer, and in fact, determined that the drug was so toxic — it sometimes caused high blood pressure, heart attacks, and ruptured intestines — that it could be more lethal than the cancer itself. Four and a half months later, the FDA officially rescinded the approval of Avastin for treating breast cancer. The FDA commissioner at the time, Margaret Hamburg, admitted that new evidence showed that the once-promising drug was not actually effective in treating breast cancer. “It is clear that women who take Avastin for metastatic breast cancer risk potentially life-threatening side effects without proof that the use of Avastin will provide a benefit… that would justify those risks,” Hamburg said at the news briefing.
The anger and disappointment of the gathered patients was palpable, but the stunning reversal, and Hamburg’s own words, raised a troubling question: If there was no evidence that Avastin was effective against breast cancer — and even some evidence that it was explicitly harmful — why was it approved as a breast cancer treatment in the first place, and why was the company that made the drug, Genentech, permitted to market Avastin to doctors and breast cancer patients for the better part of three years?
It’s a question that cuts to the heart of a program that allows the FDA to approve drugs using a lower standard of evidence. Under what’s known as the Accelerated Approval Program, the FDA can reduce the bar for approval in cases where there is an unmet medical need for a serious condition. In such cases, a drug manufacturer need not show that the drug works. It only needs to demonstrate some reasonable expectation that the drug ought to work.
By definition, that’s a much more subjective threshold, but according to Janet Woodcock, the director of the FDA’s Center for Drug Evaluation and Research, the benefits of accelerated approval more than justify the problems that might come with lowering standards of scientific evidence — particularly when desperate patients are willing to gamble on the additional risk. “It’s not unusual to have differences of opinions about accelerated approval because it’s more uncertain,” Woodcock said. “The patients are saying ‘we want to accept the tradeoffs, we’ll accept more uncertainty.’”
Today, the FDA is increasingly proactive in bringing drugs to market short of full approval and uses accelerated approval to get new drugs to people suffering from devastating diseases. Since 2003, more than 16 percent (66 of 404) of all new drugs were approved through the Accelerated Approval Program, and it seems to be a more popular option. Between 2003 and 2013, about three drugs were approved each year through this expedited route. But during each of the last three years (through 2016), that number has increased to more than seven drugs per year.
The FDA is candid about its commitment to expedited approval programs — in part to speed up what is often characterized as a notoriously drawn-out and bureaucratic approval process. The agency’s former head, Hamburg, wrote about the FDA’s intention of getting new drugs to people as “quickly” as possible, and the FDA’s new leader, doctor and cancer survivor Scott Gottlieb, bemoans the FDA’s slow-moving approval process. While a fellow at the conservative American Enterprise Institute in 2012, Gottlieb lamented the “increasingly unreasonable hunger for statistical certainty on the part of the FDA.” And at Gottlieb’s confirmation hearing last May, he rejected the idea that speeding up drug approvals would compromise their safety, calling it a “false dichotomy that it all boils down to a choice between speed and safety.”
But the increasing reliance on accelerated approval and other means of expediting drug approval have many critics worried — particularly given that the interests most readily served by fast-track approvals are those of the pharmaceutical industry. David Gortler, an associate professor of pharmacology at Georgetown University and a former FDA medical officer, is one such critic. He fears that the drive to get drugs out faster with weaker scientific evidence is already taking a toll — not just on consumers who are taking drugs that should never have been approved, but also on the agency’s credibility.
“I don’t really recognize the agency for which I once worked,” said Gortler, “because they’re making all these crazy decisions.”
The essential problem is that when it comes to drug approvals, speed and certainty are fundamentally at odds. It typically takes years of testing in large numbers of patients to determine if a drug provides a meaningful benefit — including improving an individual’s odds of survival. And it’s impossible to detect potential side effects until a sufficient number of patients have been monitored carefully, and for enough time, for such problems to truly surface. It’s extremely time-consuming to show, scientifically, that a drug really works — and to understand its risks.
But not all patients have the luxury of time. In the early 1980s, for example, an HIV infection meant certain death, as the virus devastated a person’s immune system, leaving them with wasting, sore-ridden bodies. Without drugs available to combat the virus, the afflicted were pushed toward desperate, ineffective treatments. Some tried cooking medicine themselves, while others heard stories of potent drugs and sought them out on black markets. None of it, of course, worked.
Responding to the crisis, the FDA began experimenting with what would formally be called accelerated approval. Instead of requiring ironclad evidence that a potential anti-HIV drug prolonged patients’ lives, the agency asked for indirect evidence that the drug was working as it was supposed to. In 1992, the anti-HIV drug ddC was approved because initial trials demonstrated that patients using it showed an increase in the number of a certain type of immune-system cells in their bloodstream. It wasn’t proof that the drug actually helped patients — the agency asked for a number of follow-up studies to establish that fact — but the consequences of the disease were so grave that the agency decided that it was worth the risk to approve ddC.
“We approved the AIDS drugs … and over time it came about that those were correct decisions and the epidemic was controlled.”
In this case, the gamble paid off. The drug worked. “We approved the AIDS drugs … and over time it came about that those were correct decisions and the epidemic was controlled,” Woodcock said. It also set a pattern for the future. Under certain circumstances — only having to do with serious diseases and unmet needs — the FDA can allow a drug to come to market with a lesser standard of evidence. Instead of proving that a drug prevents heart attacks, a pharmaceutical company might only need to show a reduction of fatty cholesterol molecules in patients’ bloodstreams. Instead of proving that a cancer drug extends lives, the company might have to show only that the chemotherapy delays tumor growth for a while. And instead of showing a direct benefit to a patient, an applicant might need only demonstrate that its drug meets a “surrogate endpoint” that suggests the drug is helping people who take it.
Using these surrogate endpoints saves time; it might take only weeks or months to show that a drug affects patients’ blood chemistry, whereas it takes years to gather enough data — and enough deaths — to determine whether a drug can actually extend a patient’s lifespan. The promise of surrogate endpoints is getting drugs to patients who are in desperate need of them quicker, but the downside is that there’s less careful testing of whether the drug actually works, or whether it kills patients instead of helping them.
“It is likely” said Vinay Prasad, a hematologist-oncologist at Oregon Health and Sciences University, “that many people are being treated with drugs that actually do not make them live longer or live better.”
In September 2016, the FDA approved the drug eteplirsen, designed to treat Duchenne muscular dystrophy (DMD), which primarily affects young boys. DMD is an invariably fatal disease that slowly destroys the muscles of its victims. None of the disease’s variants has a cure, although drug therapies like steroids can help slow muscle deterioration. Eteplirsen, however, is the first drug approved to target the root of the disease. Despite the huge expense — starting at around $300,000 per year — eteplirsen is the only real source of hope for children dying of DMD.
The root of the disease is a protein called dystrophin that is necessary for building muscle fibers; but those with DMD have a genetic mutation that interferes with the normal production of this protein. Eteplirsen can be used to specifically counter the effects of one of these mutations.
To speed up the approval process for struggling boys and hapless parents, the FDA put eteplirsen on the accelerated approval pathway. During the drug’s testing, the trials were too short to determine if the drug actually improved how long the boys lived, so “dystrophin production” was used as a surrogate endpoint. The study’s scientists reasoned that if they found increased levels of the protein, it was a good indication that the medication would ultimately improve the children’s condition.
As it was, evidence did suggest that eteplirsen increased the levels of dystrophin in patients’ muscles, but the increase was tiny — far below what would be expected to have any clinical effect. Scientists were baffled, in fact, by the miniscule quantity of dystrophin the drug produced.
“I find it difficult to conceive how a treatment effect of three parts per thousand could confer clinical benefit,” wrote Ellis Unger, the director of the FDA’s Office of New Drugs, and who oversaw eteplirsen’s scientific review. “If there were 10 inches of snow on a sidewalk that needed to be cleared, three parts per thousand would amount to 1/32nd of an inch.”
Unger wrote that a drug would need to increase dystrophin levels to around 10 percent of normal, healthy levels to even be considered “reasonably likely” to offer “measurable clinical benefit.” This means that even if dystrophin levels were 32 times higher than the trial results, there would still be big question marks about eteplirsen’s effectiveness.
Indeed, given that the drug didn’t even seem to meet the weakened standard of a surrogate endpoint, Unger and the other FDA scientists on the review panel wanted the FDA not to approve eteplirsen. Still, their recommendation was overruled by Janet Woodcock, who concluded the results met FDA effectiveness and safety standards for drugs on an accelerated pathway. (Two of the eteplirsen panel members resigned after this decision.) DMD patients across the country are now fighting with their insurance companies to get coverage for the expensive new drug.
“They may very well be paying $300,000 for some snake oil treatment,” says Gortler at Georgetown University. “It’s mean too, because these people are very desperate. They’ve been given a death sentence and they want to have hope.”
Prasad also warns that the increasing reliance on surrogate endpoints risks eroding the overall standards of new drugs that are allowed to appear on the market. If the bar for approval becomes so low, he reasons, pharmaceutical companies aren’t incentivized to make any of them truly safe and effective. “We want A-plus drugs, not D-minus drugs. So why are we accepting it?”
Not everyone takes such a hard line on the risk-benefit calculations at the heart of accelerated approval processes. Mikkael Sekeres is an oncologist at the Cleveland Clinic who served on the FDA’s advisory committee for Avastin on whether or not it should be withdrawn. While approval for use of the drug in treating breast cancer was ultimately rescinded, Sekeres says the FDA is right to try certain drugs — based on surrogate endpoints — if a disease is bad enough. “If the risk of disease is awful, then [the FDA] might be willing to consider a drug with more risk than with a disease that isn’t as awful,” Sekeres said.
He pointed by way of another example to lymphoma where patients might survive only 10 years with the disease — a rather long disease course. Running a complete clinical experiment to see if a drug can prolong cancer patients’ survival might take 15 years, Sekeres noted, meaning many patients would be dead before they can ever try the drug.
“So you make an approval based upon a clinical maker,” he said. And if a drug does prove to be harmful, the FDA’s accelerated approval program is designed to pull the drug from use — just like the FDA did for Avastin. “It’s a demonstration that the system works when the FDA reserves its rights to recommend removing the drug from approval,” said Sekeres. “I think the evidence is working and the FDA is exercising its duty to protect the public.”
To be sure, the FDA emphasizes that accelerated approval is designed to help patients with no other options. “This pathway allows for flexibility in new drug approvals for serious diseases with no satisfactory therapies while meeting the appropriate standards for safety and effectiveness,” FDA spokesperson Sandy Walsh wrote in an email. “Accelerated approval recognizes that physicians and patients are generally willing to accept greater risks or side effects from products that treat life-threatening and severely-debilitating illnesses, than they would accept from products that treat less serious illnesses.”
As it stands, the FDA approved 86 cancer drugs in the 15 years prior to October 2017. Twenty-nine cancer drugs were granted accelerated approval in the same time frame, representing about a third of all cancer drug approvals. And cancer therapies are a substantial component of all drugs brought to market, accounting for nearly 22 percent in the last 15 years.
The FDA approved 86 cancer drugs in the 15 years prior to October 2017. A third of these were granted accelerated approval.
Of course, even under the best circumstances, Walsh noted, the accelerated approval process can lead to approvals that are later revoked. “These limitations are a reason accelerated approval is available only for a limited group of drugs,” Walsh said, including “those intended to treat serious or life-threatening illnesses when the drug is expected to provide a meaningful benefit over existing therapy.”
Companies (Sarepta included) that approve drugs based on a surrogate endpoint are still required to conduct clinical trials, called Phase 4 confirmatory trials, demonstrating that the medication provides the intended benefit, Walsh added.
Still, there is some evidence that this may well depend on the drug. The drug Mylotarg was originally granted accelerated approval in 2000 with a surrogate endpoint showing a decrease in patients’ leukemia. Following its approval, however, further trials revealed the drug was not only ineffective, but could also be lethal. So in 2001, the FDA issued a so-called black box warning — a notice appearing on a drug’s label that warns users of “life-threatening risks.” Nine years after that, in 2010, Pfizer withdrew the drug completely.
Despite this, Mylotarg is back. The FDA, designating a lower dose and new population for the drug, approved it to treat certain patients with acute myeloid leukemia. But the new approval still relied on surrogate endpoints, which critics say prioritizes speed over compelling proof. One of the Mylotarg clinical trials, for example, found that the drug increased event-free survival by nearly eight months, but did not impact overall survival at all.
“The problem with surrogate outcomes is they’re more convenient because you can observe them sooner, but it’s very, very easy to get a big difference in the surrogate outcomes that does not translate into a big difference in survival,” says Peter Thall, a biostatistician and expert in clinical trial design at the MD Anderson Cancer Center. “The whole game of saying ‘This is statistically significant’ is grossly misleading. This is done again and again and again in oncology.”
This is also what happened in the case of Avastin. The drug was brought to market for the treatment of colon cancer and then lung cancer specifically, and in both cases, there actually was clinical data showing the drug marginally extended patients’ lives. But the 2008 approval of Avastin for breast cancer didn’t have that sort of data, so the FDA put Avastin on the accelerated approval pathway. Early evidence showed that when Avastin was used in combination with the chemotherapy drug paclitaxel, patients experienced an improvement in progression-free cancer survival over those that took the chemotherapy drug alone. However, clinical tests on cancer patients didn’t show that breast cancer patients actually lived any longer overall. Rather, the surrogate endpoint “progression-free survival” only meant that patients lived longer with the tumor after treatment.
Patients taking Avastin did seem to do better on this particular measure, so the FDA took the gamble to let patients start taking the drug. But in this case, follow-up studies showed that the drug didn’t actually increase overall survival. What’s more, Avastin proved unacceptably toxic, including producing gastrointestinal perforations in some patients. The FDA rescinded the approval, much to the disappointment of scores of breast cancer patients who were convinced that Avastin had saved their lives.
The uncertainty that comes with a surrogate-endpoint-based approval is a huge problem when it comes to treating cancer, said Oregon Health and Sciences University’s Prasad. “If you look at big randomized studies in oncology over [the] last few decades, they used to look at survival more. But survival as the endpoint of randomized studies has fallen and it’s largely been replaced by surrogates like progression-free survival,” Prasad says.
“They used to look at survival more. But survival as the endpoint of randomized studies has fallen and it’s largely been replaced by surrogates.”
What’s worse: These surrogate endpoints often don’t mean the patient benefits. Prasad and his colleague Chul Kim investigated drugs for 55 cancer indications approved by the FDA based upon surrogate endpoints and discovered that only about half of those drugs had any sort of proof that the surrogate endpoint really helped patients in any meaningful way. The analysis was published in the journal Mayo Clinic Proceedings in 2016. “For almost half of these established surrogates, there is no published study ever showing what their correlation is with survival or quality of life,” Prasad said. “We just can’t even find documentation.”
Kim, lead author on the report and an attending physician at MedStar Georgetown University Hospital admits to being surprised. “It was sobering to see,” he said. “I’d expect to see more high-level evidence to support the use of surrogate endpoints. Because progression-free survival and response rate have been used for a long time.”
This lack of proof creates a layer of uncertainty that is extremely hard to explain to patients, Prasad says. People with cancer come to him desperate to find a drug that will prolong their lives. When he explains that a drug was shown to slow tumor growth, patients immediately ask, “Does that mean I live longer?” Prasad simply can’t answer that. And thanks to surrogate endpoints, FDA approval of a cancer drug doesn’t quite mean that it is proven “effective” in the way that patients naturally think it is.
Nonetheless, accelerated approval is likely to become increasingly common, experts say, because patients, politicians, and pharmaceutical companies are all shouting for swifter drug approval. “There’s been an ongoing beating of the drum,” said Joseph Ross, an associate professor and physician at Yale University, “about the FDA needing to be faster,”
But Ross urges caution before the approval process is kicked into an even higher gear. “All evidence suggests that the FDA is reviewing and approving drugs faster than any other peer regulator in the world,” he said. “But we need to do more studies now to show what this means for safety and efficacy.”
About half of new drugs, Ross says, are now approved on standards that might look impressive in the laboratory but might not actually benefit patients. “These drugs are being approved on the basis of lab measures, but will they improve symptoms and mortality?” Ross asked. “We don’t know.”
That’s why Kim and other experts would like to see more study of the impacts of accelerated approval and surrogate endpoints overall. “If a drug is approved on a surrogate endpoint, we must do a follow-up study to make sure that it was not a false sign of efficacy,” Kim said.
“Cancer drugs can be toxic,” he added. “Without those hard outcomes, we may just cause harm to patients.”
This article was produced by students in the Science, Health & Environmental Reporting Program at the NYU school of journalism.