Life is unfair, and while others have suspected as much before, biochemists can now prove it. You have colon cancer—possibly because a flawed APC gene failed to produce the protein that helps prevent the disease. When the cancer spreads to your liver, you need Pfizer’s Camptosar. But if you’re the one-in-ten patient with a flawed UGT1A1 gene—find out with a Food and Drug Administration–approved test kit—you lack an enzyme to purge the drug from your body before it accumulates to toxic levels. Your oncologist may be able to adjust the dose so you can take Camptosar anyway. Or maybe not.

Washington can’t help. The Fourteenth Amendment doesn’t guarantee equal protection at the pharmacy. No privacy-protecting, discrimination-banning law, no promise that someone else will pay, will ensure that a drug that suits others will suit your genetic profile too. If Pfizer can’t make a gentler Camptosar, it will only do business with tougher patients. Meet “pharmacogenomics”—eugenics for drugs.

This is where diversity blather gives way to the rigorous diversity science that’s taking over the medical show. Drugs supply almost all the real health care these days, because human hands are too big to grapple with the microscopic things that cause most of our problems. Eugenic drugs reflect how biochemically separate and unequal people are. Some, indeed, target genes that track sex, race, or ethnicity; their FDA licenses affirm truths unmentionable in polite society and approve conduct illegal in every other sphere of commerce and public life. All are terrible news for anyone determined to pull people together, pool medicine’s costs, equalize its benefits, and lose diversity in the crowd. The doctors of equity promise universal access to the Mayo Clinic, where the real doctors now brew discriminatory cures and card your genes at the door.

So the stage is set for a long battle between radically new medical science and a senescent, unscientific vision of how diseases are cured and what the “health-care system” ought somehow to deliver. Much of the battle will be fought at the FDA, which is able to see things both ways, because it now has two separate brains humming away under its hat. What health care most needs is less of the old brain and more of the new. That policy alone will improve the quality of medicine and lower its cost more than any development since germs were exposed and immunology became a science almost a century and a half ago.

Soon after Watson and Crick published the blueprint for the double helix in 1953, George Hitchings and Gertrude Elion at Burroughs Wellcome began designing drugs systematically around the biochemistry presented by their intended targets. This logical approach foreshadowed the future of medicine and (in a weirdly circuitous way) the reinvention of the FDA almost three decades later. Scientists caught on much sooner, and “structure-based” drug design advanced rapidly as biochemists acquired the tools needed to read, configure, and build the molecules that choreograph life.

Drug designers take diseases apart. Iressa, for instance, targets a single receptor that proliferates in the most common form of lung cancer. The FDA licensed the drug in 2003, after clinical trials yielded good results. Follow-up trials involving almost 2,000 patients suggested, however, that the drug wasn’t working after all. Further analysis then revealed that Iressa had indeed worked—but only on receptors found in 300 patients of Asian ancestry. Similar variations apparently explain significant differences in the efficacy of drugs used to treat many other cancers. White Americans have less tolerance for some antidepressants, antipsychotics, and heart-disease drugs, while blacks respond poorly to certain drugs for high blood pressure and hepatitis. Eleven variations in just one gene affect responses to common antidepressants.

Drug designers take the rest of the patient apart, too. Tolerance for many drugs often hinges on how well patients metabolize and expel them, which seems to depend on a couple of thousand variants in a couple of hundred different genes. What were once inexplicable “side effects” are now predictable interactions between the drug’s chemistry and healthy parts of the patient’s. That lets medicine keep the drug and vote the too-delicate patient off the island.

Designers are also finding out when to leave well enough alone. A gene variant discovered in early 2008 apparently protects about 40 percent of African-Americans from heart disease as well as certain drugs do, by tinkering in much the same way with adrenaline’s effect on heart cells. Adriamycin has been routinely used against late-stage breast cancers, even though it can cause serious heart problems and spawn other cancers. A test that profiles 21 genes now identifies patients who will do as well on milder drugs, and others who can skip chemo altogether.

And that’s it. Two thousand years after Democritus postulated a material universe made up of indivisible atoms, science can finally track life, too, down to its basic elements. Health once depended on four fickle humors, which apothecaries rebalanced with eye of newt, adder’s fork, and fillet of the fenny snake. Cholera was just one ill-humored disease among many that left patients lying in puddles of their own waste—then Koch and Pasteur found V. cholerae in the puddle, and others found tetracycline to kill it. Until quite recently, heart attacks were just bad luck—now they’re clogged arteries saved by cholesterol-busting Lipitor. Until even more recently, Camptosar’s toxic side effects were a mystery—now they’re a gene on a dipstick.

At each step, medicine has advanced by disassembling—the old swellings, fluxes, and fevers into hundreds of discrete germs, the new into thousands of genes and other biochemicals. Pharmacology has found better ways to tame smaller shards of hostile life while dodging friendly molecular bystanders. And in scrambling to do all that, it has revealed that we aren’t all the same deep down, neither in sickness nor in health.

Charles de Gaulle once wondered how anyone could govern a nation that had 246 different kinds of cheese. Designer medicine could probably stock that many varieties on just the cholesterol shelf of its fromagerie. The simplest fix: fewer cheeses.

In most developed countries, the fix happens at the national health pharmacy, which stocks the government’s favorite Brie and skips the Bleu de Termignon. The price of the drug falls sharply when the owner of its patent is permitted to sell to just one buyer. And sharply again when the government migrates patients to a generic alternative, which it does as quickly as possible. Formularies, wholesale purchases, and uneven copayment schedules accomplish much the same in the United States. All such schemes favor the cheapest pills that can help the most patients—which means older drugs that serve the biochemical mainstream.

The government has compelling reasons to push things that way, because drugs are mostly know-how—the first pill costs a billion dollars, but copies cost only pennies to manufacture. So most drugs end up being sold close to cost when patents expire and generic copies flood the market. Every new sliver of human diversity offers drug companies, physicians, and patients a new opportunity and reason to go their own, separate, patented, and therefore expensive ways. Uniform crowds, one size fits all, and enough already are much cheaper.

Boundless diversity is ungovernable as well as costly. Setting national health-care priorities in Washington and herding people into line was straightforward when infectious germs threatened everyone and everyone could beat them with the same handful of vaccines and antibiotics. Most of today’s drugs aim at differences within the herd itself. The fragmentation of diseases and cures leads inevitably to fragmentation of economic and political interest. That leaves drug companies in control of which patients—or make that biochemical profiles—the health-care system will help next, and companies are free to favor profiles that pay their bills. The global war against germs reached that point some time ago: the World Health Organization is very interested in malaria, but U.S. investors aren’t. Genes are next. Progress toward universal health care now depends on a pipeline of drugs controlled by Wall Street, not Washington.

Happily for patients, Wall Street prizes diversity—the real thing—a lot more than Washington does. To be sure, drug companies make excellent money selling one-size cures for very common problems. But the stupendous diversity of human chemistry is the only thing that keeps the business profitable in the long run, because the clock is always ticking on patents already bottled, and at midnight the profits turn into pumpkins.

There is, moreover, lots of money to be made in tracking unhealthy differences down to their fragmented root causes. Drugs that target fatty blood, a malignant gene, or an insidious virus long before it morphs into a plaque, a tumor, or full-blown AIDS must often be prescribed for decades, and thus end up very profitable even when they address problems that aren’t very common at all. Wall Street adores them.

Diversity also lets drug companies take an almost free ride on innovation pioneered by their rivals. By tweaking a pioneer drug’s chemistry, a me-too competitor can dodge the patent while saving itself the cost of reinventing all the chemistry. About 4 percent of HIV patients on an AIDS drug called saquinavir develop very high cholesterol levels; these patients tolerate two statin drugs quite well but react badly to two others. The second-generation painkillers (Vioxx among them) cause less stomach irritation and bleeding than aspirin and, unlike Advil, can be taken in tandem with blood thinners. And so many people use cholesterol drugs and painkillers that making small changes in the chemistry of established drugs can do more good and earn more money than developing completely new drugs that target less common diseases.

The money is also happy to build the patient-profiling test kits—automated, microscopic laboratories constructed on silicon chips—that expose diversity and take markets apart. One developed by Roche and approved by the FDA in late 2004 detects 31 genetic variations that determine how well patients metabolize certain drugs. Eli Lilly has another that categorizes patient chemistry during clinical trials. “Theranostics,” the combination of drugs with patient-screening kits, makes sense because greatly improving the product at modest cost almost always makes sense in a competitive market. Excluding the unsafe or ineffective drug-patient conjunctions during clinical trials can also cut hundreds of millions of dollars off the front-end cost of licensing. And if Pfizer itself doesn’t develop the kit that tells patients that they can’t metabolize Camptosar, Lilly or Roche will take care of it.

The FDA revolves around a requirement, signed into law by President Kennedy in 1962, that every new drug must first prove itself in clinical trials. Congress neglected, however, to specify just how many different flavors of human diversity have to be invited to a trial before the drug is good to go.

This oversight wasn’t surprising. Between 1906 and 1962, when most of the FDA’s governing statute was cobbled together, the diseases that mattered were caused mainly by infectious germs, and the FDA knew better than to license bigoted drugs to fight them. Cholera causes epidemics because one size fits all—and tetracycline kills the bacterium wherever it finds it. An antibiotic can, in principle at least, work equally well and safely in every patient because it need not touch human chemistry at all. And the medical science of that era knew too little about the deep roots of heart disease, cancer, and most chronic and degenerative diseases to think about them very differently. Even less was known about diseases caused by drugs themselves—side effects.

In this state of ignorance, it was reasonable to hope that simple trials of modest size would suffice. Pollsters, after all, call a two-party national election by interviewing just a few thousand voters; each disease and its antidote seemed to present an equally simple, binary contest. The winner would be decided by tracking fevers, fluxes, lumps, morbidity, and other clinical symptoms, because germs aside, too little was known about the microscopic causes of diseases and side effects to proceed otherwise.

The years passed, and then the diversity police piled on. Their motives were probably more political than medical, but their demands made rough scientific sense, too. If patients are visibly different, they must also be somewhat biochemically different. Thus, six times between 1988 and 2002, Congress or the FDA itself demanded broader representation of race, ethnicity, sex, age, and “population subgroups” in clinical trials. As recently as 2005, the FDA licensed BiDil, a mixture of two older drugs for heart disease in “self-identified black patients”; a few months later, it directed drug companies to analyze clinical data using six sex/race/ethnic categories defined by the Office of Management and Budget to enforce civil rights laws in education. As the FDA sheepishly acknowledged at the time, however, those categories are “sociocultural construct[s],” not science. The patient, not the doctor, decides which box to check, and (s)he may check more than one.

Meanwhile, scientists had been excavating the biochemical constructs of diversity. What they found was (and remains) alarming: with tens of thousands of bystander molecules inside each patient, and no two patients quite the same, any drug that targets human chemistry and gets widely prescribed will almost inevitably sideswipe some innocents. Such drugs are risky even when they hit only the right target. In their 1985 Nobel lecture on the lipid, Michael Brown and Joseph Goldstein noted that cholesterol is a “Janus-faced molecule”—an indispensable component of our cell membranes, but lethal in excess. So, too, then, is Lipitor, and every other drug that tinkers with what makes us tick. Some bodies will inevitably stare down the dark face of the Janus-faced drug better than others.

To find out which ones, old-guard clinicians stage bigger trials and run them long enough for bad human chemistry to turn into bad clinical symptoms. Two decades ago, the typical trial involved about 1,000 patients; today, it’s well over 4,000; and tens of thousands of patients can be involved in testing drugs intended for widespread use. The length of these massive trials keeps growing as well.

Costs have risen in tandem. In a 1994 directive explaining how much diversity must be tested in trials that they fund, the National Institutes of Health found it necessary to add that “cost is not an acceptable reason for exclusion.” For drug companies, however, the cost of inclusion became a compelling reason to pursue only those drugs that could be sold to lots of patients once licensed—which gave the FDA still more reason to require even bigger, longer trials. Arm in arm, the FDA and its wards tracked diversity into a quagmire of human trials that never stop growing, take forever, and cost the earth.

The drugs that survive clinical trials involving huge, indiscriminately assembled crowds will inevitably be those that subdue the most common forms of the disease and that the most common brands of stomach, liver, kidney, heart, and immune system will tolerate. The minority of patients may be allowed to veto a drug that causes sufficiently nasty side effects, but the last drug left on the shelf will still be the one that best suits the majority. And even those that suit the majority quite well will often fail to win by margins big enough to persuade the FDA that the benefits to some outweigh the risks to others.

While Congress certainly didn’t anticipate or intend this result in 1962, “safety first” has emerged as the most potent and politically seductive policy for narrowing choice, saving money in the short term, and promoting more uniform (though less effective) care. It favors older drugs over me-too novelties still under patent, because the older ones have had years of additional vetting in the market. It slows licensing to a crawl, because some bad side effect may always still lurk just over the horizon. And the safety-first mind-set can launch runaway litigation that can knock out entire classes of drugs. Vioxx, we learn, apparently boosts the risk of heart attack or stroke. Merck and the FDA were negligent or worse. Lawsuits will be filed against Vioxx and all similar painkillers. We should have stuck with generic ibuprofen.

A safety-first policy presents its biggest hurdle at the threshold: even volunteers can’t be poisoned willy-nilly just to establish that a drug should not be licensed, so human trials can’t get started at all until lab and animal tests confirm that a drug shows promise and isn’t too toxic. In 1964, a cancer drug called zidovudine missed the cut.

Synthesized by Jerome Horwitz, zidovudine was an early product of the new, logical approach to drug design pioneered by Hitchings and Elion. Zidovudine is a “nucleoside analogue,” a subtly flawed version of a molecule used to assemble nucleic acids like DNA, which cancer cells must clone every time they divide. The idea has been likened to bankrupting a bakery by supplying defective yeast. But 1964 wasn’t a good time to be pursuing drugs that caused birth defects, even if only in cancer cells. The 1962 drug law had been propelled through Washington by a drug called thalidomide, which had a dreadful power to halt fetal limb development during the early stages of pregnancy. Burroughs Wellcome picked up zidovudine after Horwitz abandoned it, but then set it aside.

Then one day it became necessary to rethink everything. On June 5, 1981, government epidemiologists reported five cases, two of them fatal, of a rare form of fungal pneumonia in “previously healthy young men” living in Los Angeles. It took three years to isolate the underlying cause—a virus that destroys immune systems—by which time the stealth epidemic had been spreading across the United States for well over a decade. President Reagan’s FDA blasted an HIV-only tunnel through President Kennedy’s law, put the thalidomide past behind it, and embraced the molecular future.

Retroviruses like HIV reproduce by hijacking other cells and inserting their own blueprint into the host’s DNA. This hybrid human-viral genome may then lie dormant for years before beginning to churn out billions of new virions in a cancer-like frenzy. Antibiotics can’t fix infected DNA; what was needed was something more like a cancer drug. A biochemist at Burroughs Wellcome figured that zidovudine might work. He sent it to scientists at the National Cancer Institute and Duke University and suggested that they give it a whirl in their HIV lab glassware. It looked promising. The FDA immediately authorized clinical trials.

In the 1962 way of doing things, the clinicians wouldn’t have been able to prove that zidovudine would help the typical HIV-positive patient any faster than the untreated virus was likely to kill him, and the killing typically took ten years. Presented with a dreadful but slow-motion disease, the first and only drug that showed real promise against it in the lab, and solid biochemical logic for why the drug should thwart the virus, the FDA scrambled to draft new “fast-track” protocols that would allow clinicians to dodge questions that conventional trials couldn’t answer quickly. HIV, zidovudine, Ronald Reagan, and the gay community thus converged to invent the first major lobe of the FDA’s new brain.

As later formalized in FDA-speak, “fast-track” trials may focus on “biomarkers” and “surrogate endpoints” rather than clinically observable effects. A drug, in other words, may prove its stuff against the low-level chemistry of disease, the viral loads, cholesterol levels, or rogue proteins; the FDA won’t always require evidence that it beats the whole AIDS, heart attack, cancer, arthritis, or Alzheimer’s. Microscopic changes, which happen much faster, will suffice when they “reasonably suggest” whole-patient benefits in the future, however distant.

The rock-bottom biomarkers are genes. In guidelines published in 2005, the FDA lists dozens of examples of genetic differences that may cause “interindividual variations” in drug performance—by affecting how a patient absorbs, metabolizes, or excretes a drug or how the drug interacts with any diseased or healthy chemistry in the patient’s body. Drug companies may use these differences in selecting patients for clinical trials, setting dosages, analyzing results, or writing warning labels. Companies are encouraged to develop diagnostic kits to identify biomarkers that matter.

The fast-track rules also changed the FDA’s approach to safety. Zidovudine subverts HIV’s genetic chemistry, but does it slip safely past all 25,000 human genes and their countless biochemical progeny scattered, in all their variety, through the rest of every last HIV-positive body? For all anyone knew in 1987, dreadful side effects would show up five years later. To dodge that concern, the first-round clinical trials focused on patients with advanced AIDS who were likely to die soon of fungal pneumonia. But everyone knew that once licensed, the drug was going to be used far more widely. The FDA licensed zidovudine anyway, though officially still clinging to the fiction that it would be used only by patients with terminal AIDS. The fast-track rule finalized very shortly after announced that whenever a “life-threatening” disease was involved, long-term safety questions could be resolved after the drug was licensed. Later versions of the rule said “serious” or even just “severely debilitating.” Progressive blindness will do, but baldness won’t. If the new drug can beat a seriously bad biomarker, patients get it now; what it does to bystanders will be worked out down the road. Safety later.

The zidovudine trial had to be shut down prematurely in 1986, when the dead-patient count reached 19–1 against the sugar pill and in favor of the drug. Doctors treating real live patients can’t ethically keep prescribing placebos just to run up the score once they’re personally convinced that the drug works. Thus, a drug for a new disease that had entered the lab in February 1985 was licensed in March 1987, far more quickly than any comparable drug had been cleared in the preceding two decades. A year later, Hitchings, Elion, and a third structural design pioneer shared the Nobel Prize for Medicine. Zidovudine is better known today as AZT.

For countless thousands of HIV patients, the fast-track rule turned out to be just barely fast enough. An AZT-resistant strain of HIV quickly emerged. Drug designers isolated the enzyme that HIV uses to assemble its protein shell, analyzed its three-dimensional structure, identified a key point of vulnerability, developed the first protease inhibitor—saquinavir—and completed a fast-track rush through the FDA in 1995. A third class of HIV drugs that target another bit of HIV’s chemistry soon followed. Three-drug cocktails have proved effective ever since.

In the past decade, according to one recent study, 25 trials of drugs that treat cancer of the breast, bowel, lung, kidney, ovary, and gastrointestinal tract were stopped earlier than planned, all but one because the drug was working too well. Five hadn’t yet enrolled even half the number of patients planned. Their designers were surely more gratified than surprised. Tracking pneumonia to AIDS to HIV to a protein to an enzyme provides the blueprint for saquinavir. The drug then gets licensed because it suppresses a biomarker just a small step or two back up that same chain. Biochemical logic substitutes for extra years at the Mayo Clinic. The dealer is allowed to peek at the cards in this kind of poker, and that improves his game considerably.

He can also stack the deck, inviting to the trial only patients who present the precise biomarker that the drug was designed to beat. Because most problem biomarkers don’t correlate with race or sex, a trial stacked this way won’t usually look discriminatory to the naked eye. But sometimes they do, so sometimes it will. Too bad—clinicians will stack the biomarkers regardless, or disgrace their profession.

Once a drug is licensed, doctors may then prescribe it to any patient to treat any disease—“off-label” prescriptions are perfectly legal. The 1987 AZT license approved prescriptions only to patients with the secondary infections that accompany full-blown AIDS. But there immediately followed, as one critic put it, a “froth of therapeutic euphoria,” a rapid accumulation of evidence that the drug worked, and further (superfluous) clinical trials to keep FDA lawyers happy. Three years later, the FDA broadened the AZT license to cover early-stage treatment of any HIV-positive patient. Off-label bartenders then developed the three-drug cocktails. Letting doctors prescribe drugs indiscriminately causes trouble, too, but good outcomes spur rigorous trials that expand the license. That, in turn, allows broader marketing of the drug, expands insurance coverage, and often extends patents.

Biochemistry guides much of the off-label experimentation. Herceptin, first licensed to treat advanced breast cancers, was then found to cut recurrence rates in half for some patients with earlier-stage tumors. In the 1960s, the treatment of leprosy was revolutionized by one doctor’s serendipitous prescription of a notoriously dangerous sedative. Two decades later, another scientist identified a protein that the sedative suppresses, the protein was then linked to other diseases, and by 2006, the drug had ten FDA licenses on its wall. Several of them cover diseases that occur mainly in AIDS patients. Having jolted the FDA out of its thalidomide-induced sleep, the virus thus helped rehabilitate thalidomide itself.

Thalidomide and AZT both got special tax credits and exclusive marketing rights under the 1983 Orphan Drug Act, yet another pathbreaking change signed by President Reagan in the shadow of AIDS. The original idea was to help resurrect drugs dropped by pharmaceutical companies because too few patients needed them, but the law ended up also covering any new drug that treats a disease afflicting fewer than 200,000 U.S. patients. In 1987, AIDS still seemed rare enough meet that threshold.

Under the orphanage’s oversight, clinical trials grow smaller, not bigger. The tiniest orphan licensed so far was tested by a single doctor in eight of the 14 U.S. patients suffering from an exceedingly rare immune-deficiency disease. Orphans often get licensed quickly, too, because many orphan diseases are caused by flawed genes, many genetic diseases look serious enough to the FDA for the fast-track rule to kick in, and every unique gene defines a unique protein immediately downstream, which can serve as a biomarker to track. And orphan trials can be quite cheap. Getting a drug designated an orphan isn’t difficult, and the FDA then helps design the tests needed to get it approved and offers some direct grants to help even the solitary doctor convert an off-label practice into a license.

The law lets rich parents adopt orphans, too, and that has allowed big drug companies to show how much they can accomplish under an FDA that welcomes small trials and makes quick decisions that are often predicated as much on biochemical logic as on clinical results. This has proved fortunate for the orphans, because exploring the genetic bottom where orphans abound is very expensive. Little orphan Gleevec was painstakingly designed to suppress a rogue human protein associated with a rare form of leukemia. The FDA approved quick trials, reviewed the results in three months, conceded it didn’t yet know whether the drug would keep patients alive longer, and licensed it anyway. Science, the agency declared, now has “tools to probe the molecular anatomy of tumor cells in search of cancer-causing proteins.” Gleevec is “proof that molecular targeting works.”

Some of the little orphans then become billionaires. One year after securing its leukemia license, Gleevec landed a second to treat a rare gastrointestinal cancer. Five more orphan licenses followed, and more are expected. Gleevec currently rakes in several billion dollars a year, and its revenues continue to rise fast. Traveling the other way down the same road, some billionaires declare themselves orphans. Viagra is a certified orphan because it can treat a rare form of hypertension. So is a quite widely used acne drug, when directed against a rare cancer. Over half of the orphans end up as wards of Big Pharma, and the most successful end up treating big crowds.

The orphan-billionaire reflects the gulf between the old medicine and the new. While new and improved on other matters, the orphanage still defines disease from the top down—pimples aren’t cancer, bone cancer isn’t gut cancer, and tense blood isn’t flaccid sex. The biochemists who fit drugs to diseases look for a common molecular problem lurking underneath, and if the same problem lurks under seven diseases, Gleevec earns seven-orphan profits. The biochemists also know that if a serious two-bad-gene disease hits 200,000 Americans, a milder one-gene version of the disease will hit 15 million, and that helps attract investors. Designer science doesn’t just pull diseases apart; it can also pull them back together, albeit in weird new ways.

So the 1962 brain is still there, too, but the FDA has grown a new brain alongside it. Billionaire drugs for common, not-too-serious ailments still creep like unwilling snails through Rogaine brain; orphan-billionaire drugs for rare and serious diseases glide through the lobes of Gleevec. In Gleevec brain, solitary doctors and corporate behemoths pursue biochemical hunches, old and new, and the FDA eggs them on. A drug gets prescribed wildly off-label, to treat diseases that at first glance seem to have nothing in common, and the FDA waves this by. Gleevec brain takes the FDA back to its childhood. Performing as promised—a requirement dating back to the 1906 truth-in-labeling law that created the FDA—gets the drug licensed. Nailing down safety—the requirement first codified in 1938 and relentlessly expanded after 1962—comes later.

The sequencing of a first human genome (with 6 billion and counting to go) wasn’t finished until 2003, and the FDA’s Gleevec brain is still under construction. It will take some years for drug companies to assimilate the new science, new rules, and new economics. Then the dam will burst. One recent estimate suggests that 6,000 diseases that collectively threaten 25 million Americans qualify as orphans, and new links between genes and disease are discovered almost every day. About 1,400 orphan drugs have been certified, over 280 have been licensed, and they have been used to treat over 14 million patients since 1983. Far more drugs will be required to treat uncommon genetic diseases that aren’t quite rare enough to qualify as orphans—roughly speaking, flaws that show up (as single copies) in more than 5 percent of Americans. And countless other genetic differences will be implicated in side effects, driving still more fragmentation in the pharmacy. Gleevec apparently works less well in patients with too much of a protein called IGF-1R.

If the law lets them, biochemists will design and bottle as many potent, dangerous drugs as it takes to span the vast breadth of human diversity and all the forever mutating microbes that afflict us—which, of course, means that the biochemists will be at it forever. Potent drugs because they are perfectly designed to jigger one key molecule in the molecular infrastructure of life. Dangerous because they go after such delicate, important stuff. And safe only when prescribed to exactly the right patients. Designed drugs are intensely discriminatory.

And also not discriminatory at all. Saquinavir doesn’t care a fig about sexual orientation; it hates HIV protease and nothing else. And if—as now appears likely—seven other protease inhibitors, each slightly different, hate hepatitis C, herpes, the common cold, or a key link in the chemistry of osteoporosis, inflammation, strokes, or Alzheimer’s, the hatred won’t hinge on standard forms of bigotry. BiDil, the drug for “self-identified black patients,” didn’t sell and was withdrawn three years after it was licensed. Its backers blamed racial bias in the health-care system. The next BiDil will be for people who match a color-blind biochemical profile diagnosed by a dipstick.

The future is also economically indiscriminate, because copying knowledge is so cheap, even when it’s printed in chemistry. People with biochemical profiles not yet covered by good drugs are victims of our collective ignorance, not their individual poverty. New knowledge, the main economic ingredient of every drug, ends up shared for free with everyone when the patent expires.

No clinical trial can prove that researchers, drug companies, doctors, and patients should be allowed to communicate and collaborate freely, that academics should be permitted to patent drugs developed under federal grants (as yet another Reagan-era law allows), that drug patents in general should be strengthened, that more drugs should be declared orphans, that more should be fast-tracked, that me-too drugs should be welcomed, that patients and doctors should be given even more discretion, and that we should celebrate dipsticks and digital networks that tell small groups of patients what they need and mobilize them to fight for it.

Nor can it be denied that this agenda has an ideological slant. It favors dispersion of information, authority, and economic interest. It relies less on electing drugs in national referenda run from Washington, and more on town meetings convened by biochemists and doctors. It benefits those who add their own intelligence to the drug’s and endangers those who don’t. It requires our parents to pay more for patented pills today, to get cheap generics to their grown children and new and better drugs to their grandchildren. It accepts that even while still under patent, medicine brewed by the vat with Wall Street’s money provides far more health care, far more cheaply, than any alternative.

In support of this agenda we can, however, invoke the biochemical logic of drugs and patients. The patient’s chemistry matters as much as the drug’s. Americans are biochemically diverse. Only so much can be learned at the Mayo Clinic; the rest has to be learned from patients whose chemistries weren’t invited to the trial. Trying to invite them all leads to quagmire and stifles learning before it begins. Getting from where we are now to universal care at the pharmacy will involve far more information than Washington can ever hope to assimilate.