The lessons of failure
Clinical flops are helping researchers define the keys to successful cancer drug trials
Article first published online: 14 DEC 2012
Copyright © 2012 American Cancer Society
Volume 120, Issue 6, pages 359–360, 25 December 2012
How to Cite
Nelson, B. (2012), The lessons of failure. Cancer Cytopathology, 120: 359–360. doi: 10.1002/cncy.21261
- Issue published online: 14 DEC 2012
- Article first published online: 14 DEC 2012
Last December, Mark Ratain, MD, director of the Center for Personalized Therapeutics at the University of Chicago in Illinois, correctly predicted that a drug called perifosine would fail its phase 3 clinical trial. Part of his reasoning focused on what Dr. Ratain deemed a flawed phase 2 study design by the drug's codevelopers, Keryx and Aeterna Zentaris. The companies, he says, went on a “complete fishing expedition” and failed to enroll enough patients to adequately test their hypothesis that they could achieve a significant survival benefit among patients with advanced colorectal cancer.
Perhaps more surprisingly, Dr. Ratain says that the clinical trial flameout, which led to steep plunges in the shares of both companies, could have been anticipated through a simpler proxy: the relative lack of interest from major investors. In an editorial in the Journal of the National Cancer Institute that he coauthored with The Street biotech analyst Adam Feuerstein, Dr. Ratain compiled 10 years of data that revealed a 100% failure rate for all 21 phase 3 drug trials performed by micro-cap developers, or those with less than $300 million in market value.1 Conversely, drug developers with more than $1 billion in value achieved success in 21 of 27 phase 3 trials.
Dr. Ratain says big pharmaceutical companies are so hungry for drugs with a high probability of success that the majority of promising candidates are snapped up well before they reach phase 3 trials. Unlike the average analyst, large firms can base their investment decisions on detailed clinical trial data accessed through confidentiality agreements. “So if a small company is doing a phase 3 trial on its own, there's a reason for that,” he says.
Poor design, questionable targets, and weak drug-target interactions can all raise red flags. In addition, as the pharmaceutical industry tries to improve its ability to predict, prevent, and learn from the overwhelming majority of trials that still end in disappointment, experts have suggested a long list of potential fixes. To avoid the same fate as perifosine and countless other drugs, however, oncologists say clinical trials must first become better at determining molecular pathways of disease, and at defining those patient populations most likely to benefit from an intervention.
Despite lingering skepticism over the predictive capabilities of preclinical models, Lillian Siu, MD, a medical oncologist at Princess Margaret Hospital and Ontario Cancer Institute in Toronto, Ontario, Canada, says it may be time to revisit tools such as annotated cancer cell lines and xenografts of human tumors in mice. The molecular profiling could provide critical information about the potential for drug resistance or sensitivity. “I think a lot of the failures could be prevented by just doing a little more due diligence before entering into patient testing,” she says.
Although the rapid growth of genomic sequencing has helped to unearth more rare mutations, in many cases oncologists do not know which are the driver mutations and which are merely passengers. “We're still learning a lot about the biology of cancers that will keep us busy for a while,” Dr. Siu says.
A stronger focus on drug and disease pathways is already paying off. Razelle Kurzrock, MD, vice center director for clinical science at the University of California at San Diego Moores Cancer Center, links recent improvements in clinical trial success rates with a better understanding of how the drugs actually work. “The best successes, of course, are where the patients have actually been selected based on the pathway impacted,” she says.
As a classic example, Dr. Kurzrock cites the drug crizotinib, which gained approval from the US Food and Drug Administration in 2011 for the treatment of anaplastic lymphoma kinase-rearranged nonsmall cell lung cancer. The genetic rearrangement has been found in approximately 4% of patients with lung cancer, a subpopulation that responds well to crizotinib. “In the old days, which is not very long ago, patients would be treated in an unselected manner, so you would have had at best a 4% response,” Dr. Kurzrock says. If the drug development somehow made it to phase 3, she says, “probably the drug would have been trashed.”
For other drugs, such as vemurafenib for the treatment of BRAF-mutated metastatic melanoma, Dr. Siu says the science has been so compelling that clinical decisions are relatively easy. But a no-brainer in one context can be a head-scratcher in another. Researchers recently discovered that vemurafenib utterly failed in the treatment of patients with colon cancer. Subsequent sleuthing revealed that the unexpected failure was due in part to the drug's activation of the cancer-related epidermal growth factor receptor (EGFR) gene in colon tumors.
Failure as “Gold Dust”
If clinicians collect relevant tissue and other biomaterials from all enrolled patients in a standardized method, even failures can still offer “gold dust” by revealing the mechanism of resistance, says Jorge Reis-Filho, MD, a cancer pathologist at the Memorial Sloan-Kettering Cancer Center in New York City.
An informative failure, though, often requires a biological marker that can shed light on differing patient outcomes. Some experts, for example, say the recent woes of the drug Avastin (bevacizumab) have been compounded by the inability to find biomarkers that might explain disappointing responses by many patients.
In July, the United Kingdom's National Institute for Health and Clinical Excellence declined to approve Avastin, which is manufactured by Genentech (South San Francisco, Calif) for use in women with advanced breast cancer. Among its reasoning, the regulatory agency cited studies that demonstrated no significant improvement in overall survival, and a lack of evidence concerning whether the drug could improve quality of life. The regulatory body had already declined to approve bevacizumab for the treatment of cancers of the colon, kidney, and lung.
“My impression is that Avastin is really a very powerful drug and helps some patients a lot. But the problem is that we can't actually select the patients,” Dr. Kurzrock says. The incremental advance in survival noted in some phase 3 trials may represent a small subset of patients who are benefiting greatly, and a majority who derive no benefit at all. However, with no associated biomarkers to distinguish between the 2 groups, researchers have been left with frustratingly ambiguous results.
Experts are still debating whether drug developers should be pressured to define the subset of patients most likely to be helped before the drug is allowed to move forward. Dr. Kurzrock admits to being conflicted herself. If a company struggles with that definition requirement, patients may be deprived of an important therapy, perhaps for years. But if the US Food and Drug Administration approves a drug without requiring any definition, she says, the company would have no incentive to define it after the fact, because a narrower set of likely-to-benefit patients could constrict the potential market.
Dr. Reis-Filho is clear about how similar trials should be handled in the future. “I don't think that we should be allowed to carry out clinical trials where patients are not a priori stratified based on a biomarker that will tell us whether or not the patient is likely to respond,” he says. “Currently, biomarkers are almost afterthoughts: ‘We've got a really exciting drug, now let's find a biomarker.’” With a standardized system in place to test additional markers, he says, clinicians may find that a patient is more likely to benefit from a separate drug under development, and can immediately reallocate that patient from one trial to another.
Researchers also are realizing that virtually all cancers are composed of multiple clones whose aberrations occur in only a subset of each patient's tumor cells. Understanding this genetic heterogeneity within tumors will be critical in cutting off all avenues of escape and eventually developing a cure.
New technology, such as whole-genome sequencing, might help to open other target pathways and identify patient subpopulations with a strong response to a specific drug. However, Dr. Ratain says the available methods of analyzing biomarkers must still improve if clinicians are to understand tumor variation enough to ward off future failures. “Current technologies are fairly static, but what may be more important than a current snapshot is the whole moving picture,” he says. “In other words, we know that tumors are highly heterogeneous both in space and time. Until we have a better way of understanding that variability, it's the difference between a photograph and a movie. And we need the movie.”
BRYN NELSON IS A FREELANCE MEDICAL JOURNALIST.
Erratum: Please see p. 424 for a correction to the October 2012 issue of “CytoSource.”
Cytosource Reader Poll #9:
Clinical Trial Failures
Q: What do you see as the single biggest contributor to failure among anticancer drug trials?
A. Improperly defined patient population.
B. Inadequate understanding of drug/disease pathway.
C. Flawed study design.
D. Ineffective or irrelevant drug-target interactions.
E. Lack of follow-through due to financial considerations.
Take the poll online at www.cancercytojournal.com.
The results will be published in the April 25, 2013 issue.
AUGUST POLL RESULTS
The biggest contribution of comparative effectiveness research is likely to be:
0% A means for generating better hypotheses, thereby leading to better clinical trials.
0% A solid foundation for more personalized medical care.
100% A decisionmaking aid to help physicians and patients make more informed choices.
0% A way for medical organizations to weigh benefits and harms when issuing guidelines.