cancer clinical trials with Rob Califf, now the FDA commissioner. Hirsch practices medicine in the same Dallas group as Steve Perkins, and he is also senior medical director at Flatiron Health, a cancer data analysis firm in New York City.
Hirsch speculates that the dropout problem could abate as more trials start to compare one immunotherapy to another, instead of an immunotherapy to chemotherapy. But until then, one way to assuage some patients when faced with a randomized trial is more flexibility within the trial itself—perhaps letting patients cross over from the control arm to the experimental immunotherapy arm when the control is clearly having no effect. But crossover trials are much harder to design if the main outcome being measured is how long a drug can extend people’s lives—which is often the case when testing drugs to treat aggressive cancers, like pancreatic or brain tumors.
Crossover trials can work better when measuring a different outcome, the length of time a drug holds a tumor in check (or shrinks it), called progression-free survival. “There are challenges to that, too, but we have to have smarter trial design,” says Paul Sabbatini, a gynecological cancer specialist at Memorial Sloan Kettering Cancer Center in New York who also helps steer the center’s clinical standards and practice.
One of those challenges, for clinical trials and regular medical practice, is a direct product of the new immunotherapy drugs. They can create an effect called “pseudo-progression” or “tumor flare”—meaning the immune boost from the drug makes the tumor, measured with scans, look bigger—as if it were growing unchecked. “There’s discomfort watching something grow and wondering if it will respond to the treatment,” says Sabbatini. He notes that doctors have to be confident not only that the treatment will kick in, but that patients will be strong enough to stay on what might be a longer course of therapy before it takes effect.
The old standard for measuring tumor progression, known by the acronym RECIST, is “archaic,” says Perkins. New standards to measure the immune system’s effectiveness in fighting cancer are necessary.
For these two clinical immunotherapy problems—patients dropping out of trials and the need to rethink the measurement of tumor response—there is, generally speaking, one solution: Better diagnostics.
Doctors nervously watching a tumor expand after immunotherapy would feel a lot more confident, Sabbatini says, if the treatment was reserved for patients identified ahead of time as highly likely to respond, perhaps because of an underlying genetic mutation. Pembrolizumab and others like it seem to work better in patients whose tumors carry a high amount of the protein PD-L1 on their surfaces, for instance, but it’s not that simple. Even among “high-expression” patients those drugs don’t work all the time.
As more is understood about the intricate dance between tumor cells, the immune system, and other factors, researchers hope to create diagnostics to better predict which of the available therapies stand the best chance of fighting each individual patient’s cancer.
Researchers and doctors would also like to use diagnostics to make big, randomized trials less necessary. New experimental drugs would need to be tested for safety, of course. But if those drugs are well designed to match the complicated profiles of a subset of patients, could those Phase 1 tests provide enough evidence of a drug’s efficacy without having to do larger comparisons to older treatments, like chemotherapy? “I think we’re moving to a place where you’re looking for great effects in Phase 1b, or Phase 2 nonrandomized trials,” says Gary Gilliland, president and director of the Fred Hutchinson Cancer Research Center in Seattle, who oversaw development of pembrolizumab while a top oncology executive at Merck (NYSE: [[ticker:MRK]]).
What about drugs that don’t show big early “signals” that they are head and shoulders above the standard of care? Will drug developers want to keep spending R&D dollars on them? If a drug needs to be compared in a randomized trial to a standard treatment, is it likely only to show an incremental benefit? Will insurers want to pay for it? Will doctors want to prescribe it, or patients want to take it?
“We have more drugs and combinations than we can possibly test,” says Gilliland. “It’s OK to leave some stuff on the shelf.” He argues that choosing patients more wisely for trials, combined with more sophisticated statistical analysis, will allow everyone in the field to make decisions based on smaller trial sizes.
If the current “gold standard” is to compare two types of cancer treatment in a randomized trial, perhaps a future gold standard would be to compare a new treatment to the historic performance of the standard of care. Pembrolizumab was first approved—to treat advanced metastatic melanoma that other drugs had failed to halt—without an active control arm. (The 173 patients were randomized to either a low dose or a high dose of pembrolizumab.) The large Phase 1b trial showed enough of an advance over the historical record that the FDA gave the green light, upon condition that Merck follow up with a randomized trial that included the standard of therapy.
A world of new, powerful cancer drugs that, on early blush, seem remarkable and are conditionally approved without randomized testing makes some practitioners nervous. Once a drug is approved, says UCSD’s Scott Lippman, it’s difficult to reel it back, even with safety surveillance systems like the FDA Sentinel program.
“Traditional clinical trials aren’t going anywhere,” says Sabbatini of Memorial Sloan Kettering. “How many times have we done Phase 1 trials that look great, then in Phase 2 you get a 50 percent response, then in Phase 3 it’s a failure?”
New FDA commissioner Rob Califf has encouraged discussion about the use of real-world historical data to supplement clinical data, but keeping the “gold standard” of randomization in large, late-stage trials.
In June, he told the crowd at a conference on big data and healthcare that ultimately he’d like to see a mix of historical and up-to-the-minute clinical data in the evaluation of a drug or medical device. But he delivered that assessment with a generous dose of caution: “If we compromise safety in the interest of innovation it will set us all back.”
