By Virginia Postrel
 |
Apostolia Tsimberidou, of the
University of Texas
MD Anderson Cancer Center
in Houston, is on the front lines
of drug research to treat
narrowly targeted groups of
cancer patients.
|
When Apostolia Tsimberidou was a young hematologist, a diagnosis of chronic myelogenous leukemia meant a patient had only a few years to live.
The median survival time when she started medical school in 1985, she recalls, was just 3.5 years. Then came Novartis’s Gleevec, or imatinib, which the Food and Drug Administration approved in 2001. Unlike traditional chemotherapy drugs, which work by poisoning the body’s fast-growing cells, Gleevec is a so-called biologic that works by altering the behavior of abnormal protein molecules — in this case, inhibiting an enzyme that makes the cancer cells proliferate.
With Gleevec, the death rate for patients with the disease plummeted to only 1 percent or 2 percent a year. The estimated eight-year survival rate has increased from 6 percent before 1975 to 87 percent since 2001. The drug, says Tsimberidou, “changed dramatically the survival of patients with this disease.”
That striking success made her want to find more molecularly tailored treatments when she moved in 2007 from hematology to designing “Phase I” human trials in a new department at the University of Texas’s MD Anderson Cancer Center in Houston. But chronic myelogenous leukemia is an unusual case, because almost all patients have the same molecular abnormality and therefore are candidates for Gleevec.
When analyzed at the molecular level, a cancer that has traditionally been viewed as a single disease commonly fragments into many different subtypes, each possibly requiring a different treatment. There are now tests for about 200 different such abnormalities, which may occur by themselves or in combination.
“We should realize first that every patient is different,” says Tsimberidou. “We cannot treat all patients with, say, colorectal cancer the same or think, for instance, that all metastatic liver disease is the same. In addition to the standard diagnostic procedures, we should perform a more refined tumor molecular analysis to better characterize every patient’s disease, and we have to tailor the treatment to the specific tumor and patient characteristics.”
The molecular understanding of cancer means both good news and bad news for improving treatment.
The good news is that more cures should be possible, with less waste from giving the wrong patients drugs that won’t work in their particular cases. That has the potential to save money and significantly to reduce suffering.
By matching patients and drugs, Tsimberidou and her team have already gotten impressive, if preliminary, results.
The patients in Phase I trials, which test brand-new drugs for safety and dosage rather than efficacy, are usually in a bad place. Their cancers are advanced, and they’ve already had lots of treatments. In short, they’re dying. They’re willing to be guinea pigs on the off-chance that something new might buy them some time.
Traditionally, “there was no particular expectation that you would even see any responses in a Phase I setting,” says George Sledge, a professor at the Indiana University School of Medicine and the former president of the American Society of Clinical Oncology.
But Tsimberidou believed she could do better, using molecular information, even at the earliest stage of drug trials. In research reported in September in Clinical Cancer Research, she and her co-authors first performed molecular analyses of patients’ tumors to identify genetic abnormalities in patients with advanced cancer who had volunteered for Phase I trials. They then assigned patients to trials based on the tests.
The results were striking. Only about 5 percent of patients who weren’t assigned to drugs based on molecular profiles responded to treatment, which is typical for Phase I trials. By contrast, 27 percent of patients in the matched therapy — those who got therapy targeting the specific molecular abnormalities in their tumors — responded.
Other measures tell the same story. The median time to treatment failure was 5.2 months for the matched-therapy group, compared with 2.2 months for unmatched patients, and the median survival time was 13.4 months, compared with nine months.
The two groups were not randomly assigned, so it’s of course possible that some hidden factor accounts for the disparate results. But given that these were very sick patients who had already had lots of treatments, and that they had many different kinds of tumors, even with a nonrandomized group the differences are great enough that it looks like Tsimberidou’s team is on to something real.
“When you’re talking about a five-fold improvement in a Phase I response rate over what you see historically, that implies that very early in the development process now we should be able to get some sense of whether or not a drug is active,” Sledge says.
Tsimberidou is now developing a randomized trial to test the concept.
Since June 2011, she and her team have also doubled the number of patients they’ve tested, finding even more molecular aberrations they might potentially match with new drugs. About 52 percent of patients have shown at least one abnormality, 11 percent have two, and 2.5 percent have three or more. A recent patient turned up with 10 molecular aberrations.
Therein lies the bad news.
The first problem is that not every abnormality has anything to do with the cancer. Some are just, in Tsimberidou’s phrase, “cosmetic” — correlated with a cancer but not causal. That poses a scientific challenge. To develop effective treatments, researchers have to figure out which biomarkers are relevant and should therefore be attacked with drugs.
Then there’s the economic problem.
It costs something like $1 billion to develop a new drug and bring it through testing to market. That cost, plus profit, needs to be spread over a lot of patients.
For blockbuster biologics like Gleevec or Genentech Inc.’s Herceptin, which treats HER2-positive breast cancer, the right patients are numerous and easily identifiable. But most mutations — or combinations of mutations — are much rarer, making the markets for drugs to address them much smaller. Sledge points, for example, to Pfizer’s Xalkori, or crizotinib, which treats lung cancer in just 5 percent of patients with a particular mutation.
As cancers and treatments are defined more and more precisely at the molecular level, nearly every form of cancer could become an “orphan disease” with a narrow, potentially unprofitable market for drugs.
“Your markets become a lot smaller,” said Meredith Buxton, the program director of I-SPY, a University of California at San Francisco program doing human trials of potential breast cancer drugs. “That’s the dilemma. The more people learn that breast cancer is not a homogeneous disease — that it’s many different little diseases — the less the value for a company to put in the $1 billion for a drug that’s going to be for a fairly specific and small-market.”
I-SPY is trying to turn the problem around by matching drugs to biomarkers and speeding up human trials. Testing drugs on patients who have the wrong kind of tumors is, after all, expensive and inefficient.
“The high cost of oncology drug development is not only an issue of finance but also occurs because many cancers are heterogeneous,” wrote Laura Esserman, the UC San Francisco professor who founded I-SPY, and Janet Woodcock, the director of the U.S. Food and Drug Administration’s Center for Drug Evaluation and Research, in a December 2011 journal article. They called for “new clinical trial designs that account for the heterogeneity and complexity of the specific disease at the outset.”
I-SPY does just that, using two novel approaches to speed up Phase II trials testing the efficacy of about a half-dozen new drugs for breast cancer. Although the I-SPY program focuses on breast cancer, Esserman’s specialty, researchers hope oncologists treating other types of cancer will adopt the tools and processes it has developed.
Patients in the program have been newly diagnosed, but they already have large tumors. Everybody gets the standard treatment, but some patients are randomly assigned to also get one of the new drugs.
The first twist is that instead of treating patients with drugs after they’ve had surgery, and then waiting five years to see whether the cancer recurs, I-SPY uses the drugs first and tracks the size of the tumor over the six or seven months until surgery. If the tumor disappears by the time of the surgery, that qualifies as a “pathologic complete response.” Enough such results allow a new drug to move on to Phase III trials — years earlier than the traditional approach.
The second twist is an adaptive, or Bayesian, design. Patients’ tumors are analyzed for biomarkers when they start the program. If patients with certain abnormalities do particularly well on a certain drug, new patients coming into the trial with that same biomarker will have a higher probability of being given that drug. Over time, drugs that do badly are dropped and those that do well progress. The experimental design cuts the time and number of patients needed for testing. In theory, drugs that graduate to Phase III trials should have an 85 percent chance of succeeding in that final, tougher and much more expensive stage — much better odds than the typical trial.
“If we find a faster, less expensive way to provide these drugs,” said Buxton, “then companies will feel like it’s worth it to continue to test them.”
Understanding the molecular differences among cancers may be interesting science. But without economically feasible treatments, it won’t do much for patients.
