Suppose you've just been diagnosed with lung cancer, which is fatal in most people within two years. Your doctor tells you there is a new drug that has kept some patients alive for as long as five years, but it can have serious side effects, including liver and eye damage. Worse, it works in only 10 percent of the people who receive it. You would probably respond, Let's see what else you've got--which is more or less what the FDA said about just such a drug, Iressa, when it was approved in 2003 as a so-called third-line therapy. This means it should be given only to patients who have failed to respond to at least two other therapies. By that point, of course, it could be too late to save patients who might have been successfully treated at an earlier stage.

Now, suppose there's a genetic test that can predict whether you're in the 10 percent for whom the drug works.

The day after researchers at Massachusetts General Hospital and Dana-Farber Cancer Institute announced they would begin studying such a test, "the phones were ringing off the hook," says geneticist Raju Kucherlapati. Doctors from all over the world were clamoring to get their patients into the study, which, if successful, will mark another advance in the emerging paradigm for 21st-century medicine: personalizing drug regimens and nutritional advice based on genetic profiling.

We are already seeing the benefits of personalized medicine in the form of enhanced screening for individuals at risk for disease. The high-biotech way to do this is to look for the presence or absence of specific genes, but it is equally valid to test for the biochemical markers that indirectly reflect an individual's genetic makeup and medical history. "We do this today with blood lipids," says Robert McBurney, the chief scientific officer of BG Medicine, a Massachusetts biotech company. "When you go for a physical, the doctor measures HDL, LDL, triglycerides and total cholesterol, and prescribes Lipitor, even though you have no symptoms. You have the biochemistry of the disease; treat it now and maybe you'll never develop the symptoms."

The other thing the doctor might do is take your family history, which is a crude but time- honored form of genetic screening. If your father died of a heart attack in his 50s, that may be a sign of danger for you, too. And the prescription may include diet and exercise as well as drugs. Everyone knows these measures are good for you, but personalized medicine holds out the promise that people might actually comply if they can be shown they are personally at risk for a specific problem. Or, conversely, they might be relieved of the burden of having to follow the latest nutritional fads if they don't stand to benefit. Don't mention this to your spouse, but some studies have even found that people with a particular genetic makeup may derive little or no cardiovascular benefit from exercise. The future of medicine, McBurney believes, lies in expanding those simple concepts.

The idea isn't new; as far back as the 1960s, pediatricians were testing newborns for congenital metabolic disorders that can cause mental retardation unless they receive a special diet. Since then, researchers have identified the genetic or biochemical markers for dozens of rare conditions, such as one in which children lack an enzyme to utilize stored fat for energy. These children can die in a matter of hours, if they stop eating when they're ill. Probably a third of them do die. But Joshua Allen, a 7-year-old boy in Wilson, N.C., is alive today because he was diagnosed at birth with this condition, known as MCAD deficiency. His mother, Sharon, knew enough to rush him to the hospital two years ago when he contracted a low-grade fever that suppressed his appetite. His life was saved because North Carolina screens newborns for MCAD, but only half of states do. An FDA panel is expected to recommend soon that all 50 states require screening at birth for 29 rare genetic diseases.

The most exciting work in personalized medicine is at the intersection of genes and medicine, the field known as pharmacogenomics. One goal of this research is to cut down the number of patients who die from side effects of drugs--a number that by some estimates exceeds 100,000 a year, not counting those killed by outright mistakes. Moreover, the risk of dangerous side effects makes doctors cautious about using otherwise effective treatments. Thousands of lives could be saved if doctors could calculate in advance the potential risks and benefits for each patient.

The related field of nutritional genomics studies how genes interact with food. A gene for the protein apo E, which plays a major role in regulating cholesterol levels, comes in three forms, or "alleles." One of them, E4, can be deadly: it raises total cholesterol, increases the risk of diabetes and heart disease, and reverses the usual protective effect of alcohol. Yet those dangers are not inevitable. People with the E4 gene will get the most benefit from lifestyle changes that doctors routinely recommend for everyone--stop smoking, lose weight, exercise and eat a diet low in saturated fat. "We can counteract the genetic predisposition to disease," says geneticist Jose Ordovas of the Friedman School of Nutrition Science and Policy at Tufts.

Personalized medicine will change the way doctors prescribe drugs. Are you taking Prilosec for gastric reflux, Prozac for depression or codeine for pain? The ideal dose of these drugs and many others is largely a factor of which alleles you possess of two different genes. These genes do not regulate the action of the drugs directly, but they control how quickly your body metabolizes them, rendering them both harmless and ineffective. In general, the rule is that the faster you metabolize drugs, the larger the dose you should receive. People vary widely in this crucial parameter, but until recently there was no easy way to test for it, so the drugs have been either prescribed on a one-size-fits-all basis, or else in dosages adjusted by a time-consuming process of trial and error. A few months ago the FDA approved a single test (the AmpliChip CYP450 from Roche Diagnostics) that predicts how individuals will metabolize many of the most commonly prescribed drugs. "In 10 years," says Dr. David Mrazek, chair of psychiatry and psychology at the Mayo Clinic, "this will be a routine procedure at birth. People will know this information the way they know their blood type."

Gene tests for drug-metabolizing enzymes are just the beginning. In the future, others will help judge the efficacy or safety of a drug for a particular patient. A recent study shows how this might work for someone being treated for congestive heart failure with a medication known as Toprol-XL. A 200mg dose is optimal, according to Julie Johnson, director of the Center for Pharmacogenomics at the University of Florida in Gainesville. But the drug can initially cause a worsening of heart failure in some people, so cardiologists usually start patients at just 12.5mg and work their way up, doubling the dose every two weeks. That, of course, requires five visits over two months before most patients even start to see benefits. But in fact, Johnson says, some patients never improve because "physicians are a little scared of the drug. Many start patients on 12.5mg and just keep them there." In her study of 61 patients, the likelihood of responding well or getting sicker corresponded to the form of a particular receptor gene that a person carried. If the results are confirmed by additional studies, doctors may one day be able to predict from a drop of blood how their patients will react.

These techniques are already in use for some diseases. Herceptin is a powerful drug for a particularly aggressive type of breast cancer, in which tumor cells have an excess of so-called HER-2 receptors for growth factors. The drug doesn't work against other breast cancers, which account for about 70 to 80 percent of cases. But a gene test of the tumor cells can identify patients who will respond. Genetic testing, says Dr. Wendy Rubinstein, director of the Center for Medical Genetics at Evanston Northwestern Healthcare, "is a wonderful tool that keeps us from wasting time waiting for drugs to work when they're not going to." Another example is acute lymphoblastic leukemia, the most common childhood cancer. It can be cured with chemotherapy that includes the drug mercaptopurine. But about 10 percent of patients have only one good copy of the gene for the enzyme that metabolizes the drug, instead of two. These children should receive half the standard dose. About one in 300 has two defective copies of the gene, and can tolerate only a small fraction of the usual dosage. "A standard dose can be fatal to them," says Dr. William Evans, director of St. Jude Children's Research Hospital in Memphis. "In the old days, we didn't understand why. It just happened. If you saw toxicity, you would drop the dose 25 percent, then 50 percent, and you'd still have problems. Now we know we have to cut it by 95 percent, and we do it right away."

Eventually, pharmacogenomics holds out the possibility of making more new drugs available, and available sooner, than under the current system. Half of all drugs in development fail safety testing. Many others don't meet the statistical standards for effectiveness. But a drug that works safely in only a fraction of the population may still save lives, if you can identify that fraction. "This could save three to five years in development times," which currently run between 10 and 15 years for most drugs, says Robert Goldberg, an expert in health-care trends at the Manhattan Institute.

For that matter, pharmacogenomics might have prevented the costly debacle of Vioxx, the pain reliever that Merck pulled from the market after studies showed it increased the risk of heart attacks and strokes in a small group of patients. Arthritis sufferers who found relief with the drug were dismayed; so were researchers who believed this class of drugs could prove useful against cancer and Alzheimer's. One day, doctors might be able to determine in advance which patients are at risk--or pinpoint signs of emerging danger in patients on the drug--and let everyone else keep taking it. That's the theory of Dr. Garret FitzGerald, director of the Institute for Translational Medicine and Therapeutics at the University of Pennsylvania. "This could save a whole class of drugs," he speculates. And that is part of the promise of pharmacogenomics: to save more lives with the drugs we already have, and with the lifestyle and diet changes we already know about. There is, researchers believe, plenty of room for improvement.