Personalized Medicine's Bitter Pill

By Stephen S. Hall   

Illustrations by Brian Cairns

Drugs tailored to an individual's genetic makeup promise to be safer and more effective, but they raise tricky economic and ethical questions.

If it were not for the great variability among individuals,” 19th century physician William Osler observed, “medicine might as well be a science and not an art.” There will always be room for art in medicine, but the advent of diagnosis and treatment based on molecular knowledge of diseases is shifting the equation decidedly toward science. Almost from the moment the Human Genome Project completed its draft sequence in 2000, the intimate genetic knowledge it conferred has been accompanied by promises of a powerful, customized form of medicine. Visionaries talk of people carrying their entire genetic sequences on personalized CDs, of medicine artfully tailored to individual anatomies, and of diagnostic tests’ predicting who is likely to respond to a particular medicine, who is likely to react badly, and who is unlikely to benefit at all.

Armed with details of individual variation, biologists could parse patients  into subgroups and predict which is likely to have an aggressive or indolent form of a disease and which would respond to one drug rather than another. Allen D. Roses of GlaxoSmithKline and Duke University School of Medicine has predicted that this approach, called pharmacogenetics, “will change the practice and economics of medicine,” and the popular media have picked up and amplified that message. In 2001 BusinessWeek hailed personalized medicine as an idea that “has captured the imagination of biotech futurists,” and Newsweek suggested that “if pharmacogenetics works, the days of one-size-fits-all therapy could go the way of bleeding by leeches.”

But underneath those extravagantly rosy and somewhat wishful predictions lie important scientific, economic, and  societal questions, beginning with one of feasibility. David Altshuler, director of the Medical and Population Genetics program at MIT’s Whitehead Institute and an endocrinologist at Massachusetts General Hospital, points out that personalized medicine remains “a model, a hypothesis” of the way medical care will evolve. “The genome is going to empower all sorts of things, but it’s not going to happen for 20 years,” he says. Along the way, personalized medicine is likely to raise a number of prickly issues: foremost among them is the paradox that the more personalized the medicine, the less interesting the business. As numerous observers have pointed out, big pharmaceutical companies have become addicted to blockbuster drugs. Targeting a smaller subgroup of a patient population by definition focuses on a smaller market.

Theoretically, one economic advantage of personalized medicine is that clinical trials might be conducted more efficiently and with a greater chance of success when researchers can so specifically select patients for testing. But how small does the pie of potential patients have to shrink before it ceases to be economically viable? Furthermore, personalized medicine is not without social implications. A technology that identifies who will benefit from a new treatment automatically identifies who won’t benefit too.

Researchers, venture capitalists, and economists have been gnawing on these questions and wondering how the field of personalized medicine actually will evolve. Despite the compelling science, some investors find that the economics still leave a lot to be desired—at least in the short term. A venture capitalist who requested anonymity notes, “The vision of personalized medicine is that you’ll go to your doctor’s office, get your finger pricked, give a drop a blood, and it will be put in a machine—right there in the office—which will tell you what drug is going to work for you. But we haven’t seen very many companies that have a viable business model in this area.”

The view from the lab is different. “The treatments we currently use to treat most patients are grossly ineffective. In type 2 diabetes, many people don’t respond,” says Altshuler. “If it were true that you could identify five to 10 percent of the market, identify and treat them in a controlled and perfected way, I think it would be a wonderful thing, and I think you could make money on it too.”

Bumpy Road

Given the uncertainties, those starting down the road of personalized medicine could well learn from Genentech, the South San Francisco, CA, biotech pioneer, and its experience creating a cancer drug called Herceptin. Few people made the connection at the time, but Herceptin’s development—from the discovery of a surface marker on breast cancer cells in 1982 to the U.S. Food and Drug Administration’s approval of a drug targeting that marker in September 1998—is a useful study of the financial risks, clinical problems, social ramifications, and rich rewards of personalized medicine.

The defining characteristic of every form of personalized medicine is its biomarker, a kind of biological fingerprint that distinguishes a subset of the patient population. Herceptin is based on a marker protein that sits on the surface of malignant cells. Called neu when it was first discovered by Robert Weinberg’s group at MIT in 1982 and more popularly known as Her-2 following its independent isolation in 1985 by Genentech scientist Axel Ullrich, the molecule “listens” for signals that tell a cell to grow and multiply. Large numbers of these receptor molecules turn out to be present in certain aggressive breast cancers because the gene for the receptor is “overexpressed.”

As early as 1987 it had become clear that only 25 to 30 percent of women with breast cancer overexpressed the Her-2 gene and might benefit from a drug that blocked the growth signal. But by 1990 scientists at Genentech had developed a drug that would block the Her-2 protein and theoretically would block growth signals to a cancerous cell. “We weren’t really thinking of it in terms of individualized medicine back then,” says Debu Tripathy, an oncologist at the University of Texas Southwestern Medical Center who participated in the early testing of the drug at the University of California, San Francisco.

 A Better Diagnosis
Researchers at universities and biotech and pharmaceutical companies are racing to realize the promise of personalized medicine. Using rapidly improving tools of molecular biology, such as DNA chips that allow biologists to analyze which genes are on or off in malignant or diseased tissue, drug researchers are scrutinizing molecular data for biomarkers that characterize subpopulations of patients. One early conclusion: many deadly diseases actually exist in several molecularly distinguishable forms. By recognizing these differences among groups, researchers hope to better understand why patients respond differently to treatments.

In 2000, for example, Jeffrey Trent at the National Human Genome Research Institute in Bethesda, MD, teamed up with researchers at Agilent Labs in Palo Alto, CA, to examine several types of cancer. The researchers used gene chips to assess the expression patterns of more than 6,000 genes in tumor tissue—including tissue from 31 patients with the skin cancer melanoma—and correlated the patterns with the disease’s progression. “The patients look the same,” says Stephen Laderman, manager of the molecular diagnostics department at Agilent Labs’ Life Science Technologies Laboratory, “but they respond to treatment differently.” The groups discovered in 19 patients a pattern of gene expression that correlated with a less invasive form of the disease. “They were able to distinguish, on the molecular level, these patients from other patients,” says Laderman, “and it was correlated with the behavior of the tumor.” The two labs have mounted a similar effort to distinguish genetic profiles of patient subgroups with breast cancer and sarcoma.

These labs are hardly alone. Millennium Pharmaceuticals, in Cambridge, MA, has identified a set of genes in ovarian cancer that predicts which women will fail to respond to the standard, highly toxic, treatment. While this approach does not immediately provide a new therapy, the technology nonetheless advertises one of the goals of personalized medicine: sparing patients the distress of useless treatments that cause nasty side effects.

Researchers at Genaissance Pharmaceuticals in New Haven, CT, meanwhile, have already launched studies to identify genetic profiles that correlate with good responses to existing drugs for asthma, schizophrenia, and high cholesterol. As was the case with Herceptin, the results could lead first to diagnostic tests that identify differences in patient groups and ultimately to drugs tailored to the newly recognized groups.

With a biomarker in hand, researchers can identify a subgroup more likely to respond to a drug during clinical trials, but the biomarker also segregates the patient population. “Never underestimate the desperation of a patient to obtain a drug,” observes Jan Platner of the National Breast Cancer Coalition, “whether she qualifies for a clinical trial or not.” That lesson of personalized medicine was learned for perhaps the first time, and certainly the hard way, by Genentech.

The company employed a diagnostic test to screen breast cancer patients to see whether they had the Her-2 marker and  thus qualified for trials of the experimental drug. Safety studies began in 1992, and the first, small-scale efficacy studies were completed by 1994. Although the results were preliminary, several women with advanced cancers responded dramatically, and word began to spread through the breast cancer underground. A number of women demanded that their tumors be tested to see whether they would qualify for the experimental drug.

One of those women, Marti Nelson, was a physician from Vacaville, CA. For months, Nelson tried unsuccessfully to get the Her-2 diagnostic test. By the time she finally was tested and learned that her tumor was indeed Her-2 positive, she was terminally ill. She died in November 1994. Her death—and her inability either to get tested or get the experimental drug—crystallized the frustration of women with breast cancer. Whether they had the biomarker or not, whether there was enough of the drug to go around or not, breast cancer patients began to demand access to Herceptin.

In December 1994 several dozen protesters got into their cars and formed a “funeral cortege” to honor Marti Nelson. They encircled Genentech’s headquarters and blocked the parking lot. They leaned on their horns and activated car alarms. They drove up on the lawn in front of the company’s headquarters. Some protesters, many with advanced metastatic cancer, handcuffed themselves to the steering wheels of their cars. The long, unending wail of horns became a kind of plaintive scream on behalf of breast cancer patients who demanded not only access to the drug, but also a role in designing clinical trials and evaluating the data. “It was really that demonstration and the public attention that followed that forced Genentech to sit down at the table,” recalls Barbara Brenner, executive director of Breast Cancer Action, a breast cancer awareness group. “We’d never had anything like that before,” agrees Jennifer Bryson, Genentech’s director of patient advocacy issues. “And that’s certainly what got our attention.”

In some respects, the Herceptin case was simply another example of cancer patients clamoring for a new drug, but it became complicated by the molecular marker that made some patients eligible for clinical trials and others not. Genentech was concerned that including patients who lacked the Her-2 marker would dilute the power of the trials’ statistics to show that the drug worked; indeed, subsequent estimates suggested that if Herceptin had been tested in the general breast-cancer population, only about six percent of the patients would have shown a response, which would have made it seem a marginally effective drug.

Yet patients view potentially life-saving drugs not rationally, but emotionally. Patient activism nearly torpedoed Herceptin. Just as Genentech was designing its final, pivotal trial to establish whether Herceptin was truly efficacious against breast cancer, activists mounted a campaign, demanding that the company give them increased access to the drug and a role in planning the drug’s trials. Genentech officials demurred, expressing concern that such access would skew the results and asserting that, in any event, there wasn’t enough of the drug to go around. “From the political and sociological standpoint, that was just a very difficult time,” says oncologist Tripathy.

In the summer of 1995 Genentech launched its final, large-scale trials of Herceptin and immediately ran into trouble. During the first five months, only 14 women signed up for a trial slated to enroll 450. In August Genentech announced it would provide compassionate access to the drug and sought the help of breast cancer advocacy organizations to get the trial back on track. Patient advocates ended up helping to redesign the trial, and they worked closely with the company on patient accrual, data assessment, and monitoring. They were, Bryson says, “very, very helpful.” And with the selective power of the diagnostic test, Genentech completed its final trial ahead of schedule, despite the sluggish start.

Splinter Groups

The Herceptin case offers multifold lessons for personalized medicine. But perhaps the most critical of those lessons concern what might be called the sociology of diagnostics. For all its power to help doctors target treatment, the precision of molecular genetics can easily generate a residue of medical frustration. “What do you say to someone who doesn’t qualify for the drug?” asks the National Breast Cancer Coalition’s Platner. In that sense, developers of personalized medicine can inadvertently dispossess subgroups of patients. And as its power to fractionate increases, personalized medicine could have the unintended effect of creating many slivers of groups too small to warrant the economics of further drug development. One patient advocate says, “The downside is: What about something that’s very effective, but only for one percent of the patient population? Is that going to be developed?”

Scientists at the National Human Genome Research Institute are exquisitely aware of the possibility of such problems. “The role of government is to work on the fragments of the patient populations that pharma isn’t going to pick up,” says one government researcher who has been participating in the development of the institute’s five-year plan. “Those discussions are still early, but that is already a concern, and it’s a real issue.”

Another collateral issue of personalized medicine is the accuracy of the diagnostic tests. Yet again, the Herceptin example offers a cautionary lesson. The first test that was developed to measure a woman’s Her-2 status typically identifies 10 to 20 percent of patients incorrectly, says Tripathy. In other words, the diagnostic was hardly definitive, and both false positives and false negatives caused a lot of frustration among breast cancer patients. Last August the FDA approved the use of an improved gene-based test for determining which patients qualify for Herceptin. The new test is “probably a little more accurate,” says Tripathy. But in that no man’s land between statistics and human emotion, the lives of many patients may be convulsed by inaccurate diagnoses. “I can foresee that happening with a lot of drugs, where the diagnostic doesn’t give a simple yes-or-no answer,” says Platner.



Axel Ullrich, a Genentech researcher, clones the Her-2 receptor




Using the Her-2 marker, Genentech selects hundreds of patients for its clinical trials of a targeted breast cancer therapy and wins FDA approval.




Todd Golub of the Whitehead Institute, collaborating with Affymetrix, shows that gene chips can be used to analyze the molecular characteristics of cancer tissue.




The Human Genome Project completes a draft sequence of all human genes.



Gleevec, an anticancer drug developed by Novartis, receives FDA approval for treatment of leukemia patients with a specific gene mutation.

Even an imperfect diagnostic test, however, can give a rough answer to another important nonmedical question: it can identify the potential size of the market, a factor big pharma is considering earlier and earlier in the drug development process. In the cold, hard arithmetic of marketing, the fact that only 25 to 30 percent of breast cancer patients possess the Her-2 marker basically quarters the universal market of 180,000 women who are diagnosed with breast cancer each year. Robert Bazell’s book, Her-2, a definitive account of the development of Herceptin, describes the initial skepticism inside Genentech. John Curd, then a medical director at the company, believed Herceptin’s hypothetical market did not justify the more than $150 million it ultimately took to get it there; Bazell’s book quotes Curd as saying that “today if Her-2 came forward, you couldn’t get it into development.” But oncologist Tripathy says, “I think Genentech’s thinking was, ‘Look, it’s still 20 to 30 percent of the breast cancer market, and that’s still a big market.’” Genentech’s Bryson adds, “We have often done work in very small patient populations. Genentech employees are very ideologically motivated, and part of our mission is unmet medical need.”

Each company is likely to have its own answer to the question of whether a targeted patient population is too small, but every answer is an economic work in progress and doesn’t necessarily stem strictly from market size. “To be honest, we’re still trying to get our arms around this issue,” says Geoff Ginsburg, vice president of molecular medicine and research strategy at Millennium Pharmaceuticals in Cambridge, MA. Personalized medicine’s advantages, he suggests, are as much qualitative as quantitative. A targeted treatment allows a more focused clinical trial,which if it is successful, permits a company to charge “premium prices,” particularly for a drug that is highly effective against a grave and often fatal illness. Herceptin treatment, for example, costs about $3,000 a month, and many patients are on it for years. Personalized medicine also allows a company to establish a therapeutic beachhead in a disease population, with opportunities to expand the market. And it offers advantages to patients: patient compliance is likely to increase, says Ginsburg, if there are data suggesting that the patient is likelier to respond to treatment.

Another possibility, more humanitarian than economic, is that patients will be spared highly toxic treatments if, say, genomic analysis indicates they won’t respond to chemotherapy. And Ginsburg believes that focusing on the nonresponder population might eventually provide clues for developing treatments that apply to the entire patient population. But even success would introduce its own set of problems. Altshuler says that “one of the hard issues we’ll have to confront is not treating people because they won’t benefit. If a treatment is working on some people and not others, are we going to be saying to patients, ‘I’m not going to even let you have a chance to respond’? That’s not how we usually do things.”

Promise and Peril

The Herceptin story offers an encouraging epilogue about the development of small-market drugs. Annual sales of the drug began modestly at $188 million in 1999. But sales have climbed steadily to about $346 million in 2001, and Genentech has deftly marketed the drug and expanded its possible uses. Originally approved for patients with the Her-2 marker who had developed metastatic breast cancer and had failed to respond to all other forms of chemotherapy, Herceptin is being tested as supplementary therapy following surgery for breast cancer and in cases of ovarian and lung cancer in which Her-2 is overexpressed.

Rituxan, an anticancer drug developed by IDEC Pharmaceuticals and marketed by Genentech, has followed a similar pattern. The FDA approved the drug in 1997 for non-Hodgkin’s lymphoma, a cancer that affects certain immune-system cells with a surface marker called CD-20. That marker allows for a targeted attack on malignant cells. Genomic analysis is identifying gene expression patterns that correlate with the disease’s progression; such analysis will ultimately affect treatment decisions. From initial sales of $162.6 million in 1998, its first full year on the market, Rituxan racked up sales of $818.7 million in 2001 and was well on its way to becoming a billion-dollar drug with expanded use by the end of 2002. The cells targeted in non-Hodgkin’s lymphoma may play a role in other diseases, such as chronic lymphocytic leukemia and rheumatoid arthritis, so the market may well extend even further.

So the pioneering drugs of personalized medicine convey messages of both caution and promise: Caution about the effect targeted biomarkers can have on a patient population, as well as the way the reaction of those patients can in turn influence the testing of drug candidates and the public’s perception of a company. Promise that a patient population can be identified and treated with a targeted drug that offers greater efficacy, as well as that nonresponders may be spared the ravages of toxic, ineffective treatments. And promise, too, that even a drug targeted to specific biomarkers can have economically bright prospects. The trick will be to capitalize on the promise without leaving too many disenfranchised patients behind.

As bumpy and fitful as was Herceptin’s journey to market, it serves as an inspiration to latter-day practitioners of personalized medicine. “I’m sure Genentech was having to wrestle with the decision of, ‘If only 20 to 25 percent of the population responds, how are we going to compete?’” says Millennium’s Ginsburg. “But somebody had to be first, and I think it’s great that they did it.”

Despite the economic challenges and societal issues associated with personalized medicine, its role in the future of medical care seems all but assured. Not only is personalized medicine going to happen, argues Stephen Laderman, manager of the molecular diagnostics department at Agilent Labs’ Life Science Technologies Laboratory, in Palo Alto, CA, but it is going to happen soon, in part because more and more patients are learning to demand whatever knowledge will empower their decisions about disease treatment. As patients learn more about their options and researchers learn more about the molecular specifics of their patients’ diseases, personalized medicine may still be a bitter pill for many to swallow, but it will be one well worth taking as an antidote to the ravages of disease.