By Clifton Leaf
Adjust font size:
The models of cancer stink
Outside Eric Lander's office is a narrow, six-foot-high poster. It is an org chart of sorts, a taxonomy, with black lines connecting animal species. The poster's lessons feel almost biblical--it shows, for example, that the zebrafish has much in common with the chicken; that hedgehog and shrew are practically kissing cousins; and that while a human might look more like a macaque than a platypus or a mouse, it ain't that big of a leap, really.
The connection, of course, is DNA. Our genomes share much of the same wondrous code of life. And therein lie both the temptation and the frustration inherent in cancer research today. Certain mutated genes cause cells to proliferate uncontrollably, to spread to new tissues where they don't belong, and to refuse to end their lives when they should. That's cancer. So research, as we've said, now revolves around finding first, the molecular mechanisms to which these mutated genes give rise, and second, drugs that can stop them.
The strategy sounds obvious--and nobody makes it sound more so than Lander, the charismatic founding director of the Whitehead Institute's Center for Genome Research in Cambridge, Mass., and a leader of the Human Genome Project. The "Prince of Nucleotides," as FORTUNE once called him, sketches the biological route to cancer cures as if he were directing you to the nearest Starbucks: "There are only, pick a number, say, 30,000 genes. They do only a finite number of things. There are only a finite number of mechanisms that cancers have. It's a large number; when I say finite, I don't mean to trivialize it. There may be 100 mechanisms that cancers are using, but 100 is only 100."
So, he continues, we need to orchestrate an attack that isolates these mechanisms by knocking out cancer-promoting genes one by one in mice, then test drugs that kill the mutant cells. "These are doable experiments," he says. "Cancers by virtue of having mutations also acquire Achilles' heels. Rational cancer therapies are about finding the Achilles' heel associated with each new mutation in a cancer."
The principle is, in all likelihood, dead-on. The process itself, on the other hand, has one heck of an Achilles' heel. And that takes us back to the six-foot poster showing the taxonomy of genomes. A mouse gene may be very similar to a human gene, but the rest of the mouse is very different.
The fact that so many cancer researchers seem to forget or ignore this observation when working with "mouse models" in the lab clearly irks Robert Weinberg. A professor of biology at MIT and winner of the National Medal of Science for his discovery of both the first human oncogene and the first tumor-suppressor gene, Weinberg is as no-nonsense as Lander is avuncular. Small and mustachioed, with Hobbit-like fingers, he plops into a brown leather La-Z-Boy that is somehow wedged into the middle of his cramped office, and launches into a lecture:
"One of the most frequently used experimental models of human cancer is to take human cancer cells that are grown in a petri dish, put them in a mouse--in an immunocompromised mouse--allow them to form a tumor, and then expose the resulting xenograft to different kinds of drugs that might be useful in treating people. These are called preclinical models," Weinberg explains. "And it's been well known for more than a decade, maybe two decades, that many of these preclinical human cancer models have very little predictive power in terms of how actual human beings--actual human tumors inside patients--will respond." Despite the genetic and organ-system similarities between a nude mouse and a man in a hospital gown, he says, the two species have key differences in physiology, tissue architecture, metabolic rate, immune system function, molecular signaling, you name it. So the tumors that arise in each, with the same flip of a genetic switch, are vastly different.
Says Weinberg: "A fundamental problem which remains to be solved in the whole cancer research effort, in terms of therapies, is that the preclinical models of human cancer, in large part, stink."
A few miles away, Bruce Chabner also finds the models lacking. A professor of medicine at Harvard and clinical director at the Massachusetts General Hospital Cancer Center, he explains that for a variety of biological reasons the "instant tumors" that researchers cause in mice simply can't mimic human cancer's most critical and maddening trait, its quick-changing DNA. That characteristic, as we've said, leads to staggering complexity in the most deadly tumors.
"If you find a compound that cures hypertension in a mouse, it's going to work in people. We don't know how toxic it will be, but it will probably work," says Chabner, who for many years ran the cancer-treatment division at the NCI. So researchers routinely try the same approach with cancer, "knocking out" (neutralizing) this gene or knocking in that one in a mouse and causing a tumor to appear. "Then they say, 'I've got a model for lung cancer!' Well, it ain't a model for lung cancer, because lung cancer in humans has a hundred mutations," he says. "It looks like the most complicated thing you've ever seen, genetically."
Homer Pearce, who once ran cancer research and clinical investigation at Eli Lilly and is now research fellow at the drug company, agrees that mouse models are "woefully inadequate" for determining whether a drug will work in humans. "If you look at the millions and millions and millions of mice that have been cured, and you compare that to the relative success, or lack thereof, that we've achieved in the treatment of metastatic disease clinically," he says, "you realize that there just has to be something wrong with those models."
Vishva Dixit, a vice president for research in molecular oncology at Genentech in South San Francisco, is even more horrified that "99% of investigators in industry and in academia use xenografts." Why is the mouse model so heavily used? Simple. "It is very convenient, easily manipulated," Dixit explains. "You can assess tumor size just by looking at it."
Although drug companies clearly recognize the problem, they haven't fixed it. And they'd better, says Weinberg, "if for no other reason than [that] hundreds of millions of dollars are being wasted every year by drug companies using these models."
Even more depressing is the very real possibility that reliance on this flawed model has caused researchers to pass over drugs that would work in humans. After all, if so many promising drugs that clobbered mouse cancers failed in man, the reverse is also likely: More than a few of the hundreds of thousands of compounds discarded over the past 20 years might have been truly effective agents. Roy Herbst, who divides his time between bench and bedside at M.D. Anderson and who has run big trials on Iressa and other targeted therapies for lung cancer, is sure that happens often. "It's something that bothers me a lot," he says. "We probably lose a lot of things that either don't have activity on their own, or we haven't tried in the right setting, or you don't identify the right target."
If everyone understands there's a problem, why isn't anything being done? Two reasons, says Weinberg. First, there's no other model with which to replace that poor mouse. Second, he says, "is that the FDA has created inertia because it continues to recognize these [models] as the gold standard for predicting the utility of drugs."
'We have a shortage of good ideas'
It is one of the many chicken-and-egg questions bedeviling the cancer culture. Which came first: the FDA's imperfect standards for judging drugs or the pharmaceutical companies' imperfect models for testing them?
The riddle is applicable not just to early drug development, in which flawed animal models fool bench scientists into thinking their new compounds will wallop tumors in humans. It comes up, with far more important ramifications, in the last stage of human testing, when the FDA is looking for signs that a new drug is actually helping the patients who are taking it. In this case, the faulty model is called tumor regression.
It is exciting to see a tumor shrink in mouse or man and know that a drug is doing that. A shrinking tumor is intuitively a good thing. So it is no surprise that it's one of the key endpoints, or goals, in most clinical trials. That's in no small part because it is a measurable goal: We can see it happening. (When you read the word "response" in a newspaper story about some exciting new cancer drug, tumor shrinkage is what it's talking about.)
But like the mouse, tumor regression by itself is actually a lousy predictor for the progression of disease. Oncologists can often shrink a tumor with chemo and radiotherapy. That sometimes makes the cancer easier to remove surgically. If not, it still may buy a little time. However, if the doctors don't get every rotten cell, the sad truth is that the regression is not likely to improve the person's chances of survival.
That's because when most malignant solid tumors are diagnosed, they are typically quite large already--the size of a grape, perhaps, with more than a billion cells in the tumor mass. By the time it's discovered, there is a strong chance that some of those cells have already broken off from the initial tumor and are on their way to another part of the body. This is called metastasis.
Most of those cells will not take root in another tissue or organ: A metastasizing cell has a very uphill battle to survive once it enters the violent churn of the bloodstream. But the process has begun--and with a billion cells dividing like there's no tomorrow, an ever-growing number of metastases will try to make the journey. Inevitably, some will succeed.
In the end, it is not localized tumors that kill people with cancer; it is the process of metastasis--an incredible 90% of the time. Aggressive cells spread to the bones, liver, lungs, brain, or other vital areas, wreaking havoc.
So you'd think that cancer researchers would have been bearing down on this insidious phenomenon for years, intently studying the intricate mechanisms of invasion. Hardly. According to a FORTUNE examination of NCI grants going back to 1972, less than 0.5% of study proposals focused primarily on metastasis--trying to understand, for instance, its role in a specific cancer (e.g., breast, prostate) or just the process itself. Of nearly 8,900 NCI grant proposals awarded last year, 92% didn't even mention the word metastasis.
One accomplished researcher sent an elegant proposal into the NCI two years ago to study the epigenetics (changes in normal gene function) of metastases vs. primary tumors. It's now in its third resubmission, he says. "I mean, there is nothing known about that. But somehow I can't interest people in funding this!"
M.D. Anderson's Josh Fidler suggests that metastasis is getting short shrift simply because "it's tough. Okay? And individuals are not rewarded for doing tough things." Grant reviewers, he adds, "are more comfortable with the focused. Here's an antibody I will use, and here's blah-blah-blah-blah, and then I get the money."
Metastasis, on the other hand, is a big idea--an organism-wide phenomenon that may involve dozens of processes. It's hard to do replicable experiments when there are that many variables. But that's the kind of research we need. Instead, says Weinberg, researchers opt for more straightforward experiments that generate plenty of reproducible results. Unfortunately, he says,"the accumulation of data gives people the illusion they've done something meaningful."
That drive to accumulate data also goes to the heart of the regulatory process for drug development. The FDA's mandate is to make sure that a drug is safe and that it works before allowing its sale to the public. Thus, the regulators need to see hard data showing that a drug has had some effect in testing. However, it's hard to see "activity" in preventing something from happening in the first place. There are probably good biomarkers--proteins, perhaps, circulating in the body--that can tell us that cancer cells have begun the process of spreading to other tissues. As of yet, though, we don't know what they are.
So pharma companies, quite naturally, don't concentrate on solving the problem of metastasis (the thing that kills people); they focus on devising drugs that shrink tumors (the things that don't).
Dozens of these drugs get approved anyway. At the same time, many don't--and the FDA is invariably blamed for holding up the War on Cancer. The fault, however, is less the umpire's than the players'. That's because many tumor-shrinking drugs simply don't perform much better than the standard treatments. Or as Rick Pazdur, director of oncology drugs for the FDA, puts it, "It's efficacy, stupid! One of the major problems that we have is dealing with this meager degree of efficacy." When it's clear that something is working, the agency is generally quick to give it priority review and/or accelerated approval, two mechanisms that speed up the regulatory process for compounds aimed at life-threatening diseases. "We have a shortage of good ideas that are likely to work," agrees Bruce Johnson, a Dana-Farber oncologist who runs lung-cancer research for institutions affiliated with the Harvard Medical School, a huge partnership that includes Massachusetts General Hospital, Brigham and Women's Cancer Center, and others.
That is also the devastating conclusion of a major study published last August in the British Medical Journal. Two Italian pharmacologists pored over the results of trials of 12 new anticancer drugs that had been approved for the European market from 1995 to 2000, and compared them with standard treatments for their respective diseases. The researchers could find no substantial advantages--no improved survival, no better quality of life, no added safety--with any of the new agents. All of them, though, were several times more expensive than the old drugs. In one case, the price was 350 times higher.
Why the new drugs disappoint
Flawed models for drug development. Obsession with tumor shrinkage. Focus on individual cellular mechanisms to the near exclusion of what's happening in the organism as a whole. All these failures come to a head in the clinical trial--a rigidly controlled, three-phase system for testing new drugs and other medical procedures in humans. The process remains the only way to get from research to drug approval--and yet it is hard to find anyone in the cancer community who isn't maddeningly frustrated by it.
In February 2003 a blue-ribbon panel of cancer-center directors concluded that clinical trials are "long, arduous," and burdened with regulation; without major change and better resources, the panel concluded, the "system is likely to remain inefficient, unresponsive, and unduly expensive."
All that patients know is that the process has little to offer them. Witness the fact that a stunning 97% of adults with cancer don't bother to participate.
There are two major problems with clinical trials. The first is that their duration and cost mean that drug companies--which sponsor the vast majority of such trials--have an overwhelming incentive to test compounds that are likely to win FDA approval. After all, they are public companies by and large, with shareholders demanding a return on investment. So the companies focus not on breakthrough treatments but on incremental improvements to existing classes of drugs. The process does not encourage risk taking or entrepreneurial approaches to drug discovery. It does not encourage brave new thinking. Not when a drug typically takes 12 to 14 years to develop. And not with $ 802 million--that's the oft-cited cost of developing a drug--on the line.
What's more, the system essentially forces companies to test the most promising new compounds on the sickest patients--where it is easier to see some activity (like shrinking tumors) but almost impossible to cure people. At that point the disease has typically spread too far and the tumors have become too ridden with genetic mutations. Thus drugs that might have worked well in earlier-stage patients often never get the chance to prove it. (As you'll see, that may be a huge factor in the disappointing response so far of one class of promising new drugs.)
The second problem is even bigger: Clinical trials are focused on the wrong goal--on doing "proper" science rather than saving lives. It is not that they provide bad care--patients in trials are treated especially well. But the trials' very reason for being is to test a hypothesis: Is treatment X better than treatment Y? And sometimes--too often, sadly--the information generated by this tortuously long process doesn't much matter. If you've spent ten-plus years to discover that a new drug shrinks a tumor by an average of 10% more than the existing standard of care, how many people have you really helped?
Take two drugs approved in February for cancer of the colon and rectum: Erbitux and Avastin. In each case it took many months just to enroll the necessary number of patients in clinical trials. Participating doctors then had to administer the drugs according to often arduous preset protocols, collecting reams of data along the way. (ImClone's well-known troubles with the FDA occurred because it had not set up its trials properly.)
And Erbitux? Although it did indeed shrink tumors, it has not been shown to prolong patients' lives at all. Some certainly have fared well on the drug, but survival on average for the groups studied didn't change. Still, Erbitux was approved for use primarily in "third line" therapy, after every other accepted treatment has failed. A weekly dose costs $ 2,400.
Remember, it took several years and the participation of thousands of patients in three stages of testing, tons of data, and huge expense to find out what the clinicians and researchers already knew in the earliest stage of human testing: Neither drug will save more than a handful of the 57,000 people who will die of colorectal cancer this year.
You could say the same for AstraZeneca's Iressa, another in the new class of biological wonder drugs--compounds specifically "targeted" to disrupt the molecular signals in a cancer cell. Not a single controlled trial has shown Iressa to have a major patient benefit such as the easing of symptoms or improved survival--a fact that the company's upbeat press releases admit as if it were legal boilerplate. Even so, the FDA okayed the pill last year for last-ditch use against a type of lung cancer, citing the fact that it had shrunk tumors in 10% of patients studied.
"Very smart people, with a lot of money, have done trials of over 10,000 patients around the world--testing these new molecular targeted drugs," says Dana-Farber's Bruce Johnson. "AstraZeneca tested Iressa. Isis Pharmaceuticals and Eli Lilly tested a compound called Isis 3521. Several different companies ended up investing tens of millions of dollars, and all came up with a big goose egg."
The one targeted drug that clearly isn't a goose egg is Novartis's Gleevec, which has been shown to save lives as well as stifle tumors. The drug has a dramatic effect on an uncommon kind of leukemia called CML and an even more rare stomach cancer named GIST. Early reports say it also seems to work, in varying degrees, in up to three other cancers. Gleevec's success has been held out as the "proof of principle" that the strategy we've followed in the War on Cancer all these years has been right.
But not even Gleevec is what it seems. CML is not a complicated cancer: In it, a single gene mutation causes a critical signaling mechanism to go awry; Gleevec ingeniously interrupts that deadly signal. Most common cancers have perhaps as many as five to ten different things going wrong. Second, even "simple" cancers get smarter: The malignant cells long exposed to the drug (which must be taken forever) mutate their way around the molecular signal that Gleevec blocks, building drug resistance.
No wonder cancer is so much more vexing than heart disease. "You don't get multiple swings," says Bob Cohen, senior director for commercial diagnostics at Genentech. Use a drug that does not destroy the tumor completely and "the heterogeneity will evolve from the [surviving] cells and say, 'I don't give a rat's ass! You can't screw me up with this stuff.' Suddenly you're squaring and cubing the complexity. That's where we are." And that's why the only chance is to attack the disease earlier--and on multiple fronts.
Three drugs, four drugs, five drugs in combination. Cocktails of experimental compounds, of course, were what doctors used to control HIV, whose rapidly mutating virus was once thought to be a death sentence. Virtually every clinician and scientist interviewed for this story believes a similar approach is needed with the new generation of anticancer drugs. But once again, institutional forces within the cancer world make it nearly impossible.
Combining unapproved drugs in clinical trials brings up a slew of legal and regulatory issues that cause pharma companies to squirm. While many drug-company oncologists are as committed to the public's well-being as government or cancer-center researchers, they have less flexibility to take chances on an idea. Ultimately, they need FDA approval for their investigational compounds. If two or three unapproved drugs are tested in concert, it's even harder to figure out what's working and what isn't, and whether one drug is responsible for side-effects or the combination. "It becomes much more challenging in the context of managing the databases, interpreting the results, and owning the data," adds Lilly's Pearce.
Over dinner at Ouisie's Table in Houston, M.D. Anderson's Len Zwelling, who oversees regulatory compliance for the center's 800-plus clinical trials, and his wife, Genie Kleinerman, who is chief of pediatrics there, have no trouble venting about the legal barriers that seem to be growing out of control. It takes no more than ten minutes for Kleinerman to rattle off three stories about trying to bring together different drug companies in clinical trials for kids with cancer. In the first attempt, the trial took so long that the biotech startup with the promising agent went out of business. In the second the lawyers haggled over liability concerns until both companies pulled out. The third, however, was the worst. There were two drugs that together seemed to jolt the immune system into doing a better job of targeting malignant cells of osteosarcoma, a bone cancer that occurs in children. "Working with the lawyers, it was just impossible," she says, "because each side wanted to own the rights to the combination!"
Additional reporting by Doris Burke