Science Editor

LOS ANGELES - Two studies presented at the American Association for Cancer Research annual meeting are important for clinicians. But just as much, they offer lessons on the surprises that can beset clinical trials.

Both trials report on the use of cetuximab (Erbitux, ImClone Systems Inc.) in metastatic colorectal cancer. One trial met its primary endpoint; the other failed. But those disparate results may have more to do with endpoints, and with what happened after the trials, than with any differing efficacy of the drug in the two trials.

In the first trial, researchers from the National Institute of Canada Clinical Trials Group and the Australasian Gastro-Intestinal Trials Group reported that when used as a third-line agent in metastatic colorectal disease, cetuximab plus best supportive care significantly improved overall as well as progression-free survival compared to best supportive care alone. Derek Jonker of the Ottawa Regional Cancer Centre, who presented the results at a clinical plenary session, said the study marks "the first time that a biologically targeted therapy, given on its own, has improved survival in colorectal cancer. It is also the first time an EGFR-targeting drug has achieved this goal."

On the heels of the first study, Alberto Sobrero from the University of Padua in Italy presented results from a second trial, the so-called EPIC trial, which compared cetuximab plus irinotecan vs. irinotecan alone as second-line treatment for metastatic colon cancer. For this study, cetuximab improved progression-free survival, but showed no benefit in overall survival.

On the one hand, the results were unambiguous. The drug missed its primary endpoint. "Irinotecan remains the standard of care," a stoic Sobrero told the audience at the plenary session.

But both Sobrero and Richard Goldberg from The University of North Carolina at Chapel Hill, who put both studies into a larger perspective, argued that it is plausible that the failure may be due to what happened after the trial. Nearly 50 percent of irinotecan-only patients started receiving cetuximab - a phenomenon known as "crossover" in the world of clinical trials. When such crossovers were excluded in posthoc analysis, patients receiving cetuximab in addition to chemotherapy had a clear survival benefit.

Sobrero noted that the benefit of hindsight precludes such a posthoc analysis from proving anything. "It is a flawed analysis," he said.

But, he added, such an analysis is a legitimate exploratory analysis of clinical trial data to guide further research. And such a survival benefit "is biologically and medically plausible, given the results of the Canadian study."

Sobrero said his group had struggled with the best way to describe the bottom line for clinicians. The formulation they arrived at is that "cetuximab is a key therapeutic agent for the optimal treatment of colorectal cancer."

Goldberg said that the results bolster the argument that overall improved survival, while obviously the goal, may not be feasible to demonstrate in some clinical trials. "I would argue, and I hope that the FDA is listening, that progression-free survival should replace overall survival" as primary endpoint in studies where crossover is a possibility, he said.

The EPIC study, as Goldberg said, might have been "destined to fail" despite being well-designed. Its results also will not have an effect on the drug's approval status. But the same is not necessarily true of other failed trials.

FDA-approved oncology drugs beat survival odds that are significantly worse than those of a patient diagnosed with most early stage cancers. Just 5 percent of cancer drugs make it from the preclinical cradle to a productive life as an FDA-approved therapeutic.

Even worse from an economic point of view is that while the failure rates of cancer drugs are comparable to other classes in preclinical and early stages, the cancer drug success rate in Phase III trials, at a mere 40 percent, is noticeably lower than in other therapeutic indications.

And many of those trials fail, as several sessions at the meetings addressed, because of flaws in drug discovery that range from the preclinical studies all the way to Phase III trials.

At a Saturday session titled "Why Oncology Drug Development Usually Fails: Bad Drugs, Bad Designs or Bad Patient Selections," Mark Ratain from the University of Chicago blamed what he called "oncothink" and defined it as a subset of groupthink practiced by those in oncology drug development.

Ratain described several aspects of what he considers to be oncothink. He said he believes the plight of Phase III oncology drugs can be partly traced to what he called "the prioritization of speed over results:" pushing on with Phase III trials on the basis of weak Phase II data.

Speaking in the same session, Rachel Humphrey from Bristol Myers Squibb Co. described methods for improving Phase II trials by enriching patients based on their clinical response in a so-called randomized discontinuation strategy.

In a nutshell, Humphrey summed up the strategy as, "If you're not sure [a drug] is working, take it off in half the patients and see what happens." Patients who clearly respond stay on the drug, and those who clearly don't respond discontinue its use. Ambiguous cases are randomly assigned to either keep using the drug or switch to a placebo, and the two groups are then compared. "Randomization offers a lot of opportunities to understand your clinical data better," she told the audience.

Humphrey noted there are limitations of that approach. Patients who obviously are progressing are taken off the drug rather quickly, which - much like the posthoc analysis of crossover data - affects the trial. That makes it hard to determine the size of the Phase III group that is needed for a successful trial. But, she added, it offers a way to explore a drug's utility on many different cancers, and limits exposure to placebo - which can be a stumbling block in enrolling patients in a clinical trial.