Life is launched and propelled by two groups of big, complex molecules: nucleic acids that are assembled into genes and the proteins those genes define. All life uses this bootstrap code to lift itself up from the stillness of inanimate matter.
Like software, this wetware can be erased, corrupted, infected, hijacked, and edited, letter by letter, word by word. Parts of our immune system subtly reconfigure their own code on the fly, in random ways; so do cancer cells. The power in this code resides in molecules that, with the technologies we now have in hand, are as easy to read, copy, and manipulate as silicon chips, or soon will be.
By assembling reams of fragmentary data to map out genomes, then dredging medical and genealogical databases to expose links between genes and disease, powerful computers are helping discover which molecules drugs should target. Drug designers then use that information to unleash two astonishingly brilliant and powerful tools for designing magic-bullet molecules to modulate the chosen targets. One enlists digital code to mimic life: Computers play a large role in designing drugs precisely matched to designated targets. Alternatively, molecular magic bullets are being created using the carbon-based biological code in laboratory animals. Either way, the vital core of medicine is now on the same plummeting-cost trajectory as microchips and software.
But the Food and Drug Administration rules that govern the rollout of these 21st-century drugs were designed for the far less powerful 20th-century tools of pharmacology. The regulations were cobbled together at a time when nobody could read the molecular-scale code that controls so much of health and disease. Drugs were designed mainly by hunches and guesswork, and very few worked well. The safe and effective use of medicine depended on gathering purely statistical information about how drugs affect high-level clinical symptoms.
The old system assumed broad areas of biochemical uniformity among patients where we now know there is none. It conflated differences between people and outcomes, steering medicine relentlessly toward generic drugs for generic patients. And by focusing on clinical symptoms and effects, which often take a long time to surface, it was often very slow. The approach we need going forward must be able to deal systematically with complex biochemical diversity that often lurks underneath a single set of clinical symptoms. And it must do so efficiently enough to take full advantage of modern pharmacology's power to develop a vast array of precisely targeted drugs.
As it happens, important elements of such a system were developed in the 1990s after Washington was suddenly smacked in the face by the horrifying biochemical reality of a retrovirus called the Human Immunodeficiency Virus, or HIV. Biochemists immediately began developing treatments using tools that would have, in the National Academy of Sciences' phrase, a "revolutionary effect on modern drug design." Meanwhile, the Food and Drug Administration (FDA) concocted a bunch of clever ways to dodge its own rules, thereby unleashing the full fury of our biochemists against this quiet, slow-to-kill virus. In doing so, the government hinted at a regulatory framework for unleashing the far greater power of today's molecular medicine against the flaws in human chemistry that sicken and kill us.
The AIDS Revolution
With HIV, nature cooked up an all-purpose anti-vaccine, so tiny and gentle that it spread unnoticed for decades, so innocuous that it never quite got around to killing you. It left that messy job to the bacteria, protozoa, and viruses that take advantage of disabled immune systems to feast on brains, lungs, blood, livers, hearts, bone marrow, guts, skin, and eyes.
On June 5, 1981, the U.S. Centers for Disease Control reported five cases, two of them fatal, of a rare form of fungal pneumonia that had struck "previously healthy young men" living in Los Angeles. A second report, issued a month later, noted additional clusters of Kaposi's sarcoma-a rare, aggressive form of skin cancer caused by a herpes virus. "The fact that these patients were all homosexuals suggests an association between some aspect of homosexual life-style or disease acquired through sexual contact and Pneumocystis pneumonia in this population," the first report declared.
These "previously healthy" patients had in fact been mortally ill for years. Isolating the underlying cause took another three years, by which time the stealth epidemic had been creeping its way across America for almost two decades. HIV has been found in the remains of a resident of Kinshasa, Congo, who died in 1959. By 1985, when Rock Hudson's death gave the virus a dose of Hollywood publicity, 12,000 Americans were already either dead or dying. By 1992 AIDS was the leading killer of young American men.
HIV isn't like cholera, smallpox, or any of the other one-size-fits-all infectious scourges of the past. Contagious properties aside, HIV is more like cancer: It briefly causes some flu-like symptoms soon after it enters the body, then typically hides unnoticed for five to 10 years. The disease discriminates fiercely-based on ancestry, sexual lifestyle, needle-sharing habits, or disability (it killed many hemophiliacs), and against babies born of the wrong mother. It typifies the diseases of the future: slow, subtle, complex, and rooted in lifestyles and genes.
"One in five-listen to me, hard to believe-one in five heterosexuals could be dead of AIDS in the next three years," Oprah Winfrey declared in February 1987. Oprah was wrong. HIV was and remains tightly linked to lifestyles shared by several comparatively small, discrete groups of Americans. The virus doesn't spread fast and indiscriminately; that's the old strategy for staying out ahead of the human immune system. It specializes instead in adapting itself to cooperative hosts and outrunning their immune systems.
HIV isn't so much a virus as a system for spawning new viruses. It replicates fast-so fast that, although the immune system fights back, it can't keep up. Each new particle formed has, on average, one mutation in its 10,000 units of nucleic acid code. "As death approaches," the geneticist Steve Jones writes in his 1999 book Darwin's Ghost, "a patient may be the home of creatures-descendants of those that infected him-as different as are humans and apes."
This process allows HIV to adapt quickly to the lifestyles of different communities. On its trek out of Africa, it found shelter in polygamous communities whose naming and coming-of-age traditions often involve the sharing of mother's milk and blood. It then split into two strains, HIV-2, which predominates among African heterosexuals, and HIV-1, which favors homosexuals. New subsidiary clusters-HIV-1A, 1B, and so forth- continue to emerge.
The predominant form of the virus found in Thailand soon after its arrival in the early 1990s was "something of a specialist at travel by the anal route," Jones writes. A decade later, "in its new nation of sex tourists," that strain had given way to a variant that "prefers conventional sex." In varying degrees, all sexually transmitted germs do much the same. "Every continent, with its own sexual habits, has its own exquisitely adjusted set of viruses."
In the 1970s, San Francisco's gay community offered HIV so many opportunities to spread that its best strategy was to attack aggressively, replicate fast, and move on, so it grew steadily more lethal. When its hosts grew more cautious, HIV spawned new strains that killed more slowly, giving the virus more time to spread.
Human genetic diversity also influences how quickly HIV kills. Some people develop AIDS quite soon after they are infected, others much more slowly; "elite controllers" carry genes that apparently allow their immune systems to keep HIV under control indefinitely. Three genes appear to be involved; about a quarter of all Asians, and smaller fractions of Africans and Europeans, carry favorable variations. One is apparently the relict of an ancient viral infection, possibly smallpox. In 2007 doctors transplanted stem cells donated by an elite controller into the bone marrow of an HIV-positive leukemia patient. The patient stopped taking HIV drugs the same day. Four years later, doctors were unable to find any trace of the virus in his body.
The HIV pandemic will now be exerting selection pressure on humanity, favoring the genes that allow its hosts to survive long enough to bear children. Some of those hardy children will carry imprints of HIV itself in their genes. Like other retroviruses, HIV replicates by inserting a template of itself into the host cell's DNA, creating a hybrid human-virus cell that then churns out new copies of HIV in a cancer-like frenzy.
Retroviral diseases and cancers thus have something fundamental in common: Both are propelled by flaws embedded in our own genes. Which helps explain why a failed cancer drug, zidovudine, emerged as the first successful HIV treatment. It had been sitting on the shelf, unlicensed, for more than 20 years.
The FDA Adapts
When AIDS surfaced, Washington begged drug companies and researchers with virus-killing expertise to send in whatever they might have on the shelf for testing in the government's secure HIV labs. A biochemist at Burroughs Wellcome sent zidovudine-better known today as AZT-to scientists at the National Cancer Institute and Duke University. In lab tests, the drug looked promising.
But HIV drugs presented a delicate problem. HIV is invisible and usually harmless for the first five or so years of the infection. What if zidovudine caused grave side effects that took four years to surface? In Washington, a drug had to prove, in meticulously choreographed trials, that it would deliver clinical benefits to live patients. Zidovudine couldn't prove it was good for patients, at least not to Washington's satisfaction, any faster than HIV killed them.
Presented with a dreadful but slow-motion disease, the first antidote that showed real promise in the lab, and plausible biochemical logic for why that drug might work, the FDA scrambled to draft new protocols that would allow clinicians to dodge questions that conventional trials couldn't answer quickly-questions about long-term safety effects, among others.
But many patients who had been infected years earlier were going to die before the new protocols could be finalized and used to evaluate zidovudine. So the FDA approved a first trial of the drug limited to patients suffering from fungal pneumonia, one of the most common killers when the infection turns into full-blown AIDS. The zidovudine trial had to be terminated prematurely when the dead-patient count reached 19 to 1 against the placebo; doctors cannot ethically keep prescribing a placebo just to run up the score once it becomes clear, to them at least, that the drug being tested works.
Thus a drug that had arrived at the HIV lab in February 1985 was licensed in March 1987, far more quickly than any comparable drug had been approved since President John F. Kennedy signed the 1962 amendments to the federal drug law-the amendments that, among other things, had spawned the FDA rules spelling out what kind of proof is required to get a drug licensed.
In dealing with AZT, Washington had greatly accelerated the licensing process by allowing the drug to be tested in patients who had been infected not only with the slow-to-kill virus but also with a secondary fungal infection that kills quickly. AZT was licensed accordingly, officially for use only by late-stage HIV-positive patients who had also been exposed to the fungus. But as the FDA knew full well, other HIV patients weren't going to wait. There immediately followed what has been described as a froth of therapeutic euphoria. Three years later, the FDA broadened AZT's license to cover early-stage treatment as well.
It had taken 30 years for humanity to discover that a dreadful new virus was quietly spreading around the globe, and another five for the United States to start fighting the virus with a 25-year-old drug. HIV then outwitted AZT in 20 months.
That outcome should have come as no surprise. After all, the virus had already outwitted the human immune system, which has the capacity to churn out vast numbers of different magic bullets in a matter of days.
While HIV was laughing its way past AZT, biochemists were analyzing other aspects of the virus's protein chemistry. They isolated the gene for a protease enzyme that HIV uses to assemble its protein shell. Then they started manufacturing the enzyme itself, worked out its three-dimensional structure, identified a key point of vulnerability, and developed the first HIV-protease inhibitor. The drug completed a lightning-fast trip through the FDA in 1995. Other protease inhibitors soon followed.
Within months HIV was developing resistance to one of the protease drugs too. All over the world, the virus picked up the same four mutations at different points in its genome. It had to pick them up in a particular order, and it did so, everywhere. The four mutations couldn't all have occurred in any single patient-even HIV doesn't mutate fast enough for that to have happened by chance. But the virus didn't have to finish the job inside a single patient.
A third class of HIV drugs that target another fragment of HIV's chemistry emerged while all this was happening. Doctors and patients began mixing up three-drug cocktails, zeroing in on one that reduced virus counts in the blood to undetectable levels. Today's HIV drugs target nucleosides, nucleotides, and three other classes of enzymes (transcriptase, protease, and integrase); "fusion" or "entry" inhibitors bind to glycoproteins in ways that stop the virus from prying its way into our cells. For now, at least, we have enough HIV drugs and cocktails to ensure that the HIV virus is no longer a nearly automatic death sentence for people with access to modern medicine.
The Dawn of Precision Molecular Medicine
The HIV drugs that followed AZT were the beneficiaries of two major loopholes in the federal drug law created soon after HIV surfaced. The rapid arrival of these precisely targeted HIV drugs heralded a new paradigm for how modern molecular medicine creates safe and effective treatment regimens.
Although not framed in these terms, the loopholes got the FDA fairly close to what might be called toolkit licensing: License a drug not as an antidote to clinical symptoms but as a molecular scalpel or suture, and let doctors take it from there. Doctors prescribe the drug to patients whose disorder presents the target that the drug is known to hit, perhaps in combination with other drugs directed at other targets. They work out the connections between molecular and clinical effects on their own, one patient at a time. The FDA-approved label plays little if any role.
Enacted in 1983, the Orphan Drug Act took the first significant step toward toolkit licensing. The original idea was to help resurrect "orphan drugs," those dropped by pharmaceutical companies because too few patients needed them. But the law ended up covering other drugs that addressed rare and currently untreatable diseases. It directed the FDA to be very flexible when licensing such drugs, basing approvals on the strength of favorable case reports, animal models, or even in vitro studies when no good animal model existed.
While the FDA was relying on a fungus to help rush AZT through Washington, the agency was also drafting its "accelerated approval" rule, which was finalized in late 1992, then codified and somewhat expanded by Congress in 1997. When the disease is sufficiently serious and available treatments are inadequate, a new drug can get to market by demonstrating that it does indeed produce its intended molecular-scale effect or, more generally, produces favorable changes in what the FDA calls "biomarkers" or "surrogate end points."
Biomarkers and other surrogates allow the FDA to make a first call about the drug's efficacy much earlier, without waiting for clinical effects to surface and persist for some arbitrary period of time. The FDA must be persuaded that the microscopic changes are connected to the clinical symptoms, but "reasonably likely" will suffice. These truncated front-end trials need not resolve concerns about how the drug's performance might be affected by many aspects of genetic or lifestyle diversity; "differences in response among subsets of the population," in FDA parlance, may be addressed later. So too may open-ended questions about distant side effects. The manufacturer must still complete controlled trials, but it does so after the drug is licensed-and thus does so after, or in tandem with, the wider use of the drug by unblinded physicians who can investigate why the drug works in some patients but not others. The FDA rescinds the license if things don't pan out.
Almost all of the early HIV- and AIDS-related drugs were designated as orphans. Most were rushed through the FDA under the accelerated approval rule. And almost all were widely prescribed off label.
HIV's endless mutability-and the arrival of targeted drugs that were licensed quickly, largely on the strength of low-level effects-left doctors in charge of working out the molecular details that determine how to combine drugs into the cocktail that produces the best clinical outcome for the individual patient. We have since learned how that process unfolded and where it ends: It has unfolded remarkably well, and it doesn't end.
A couple of dozen HIV drugs have been approved worldwide; they are typically used in about 10 fairly standard cocktails. The efficacy of each cocktail depends on which strain launched the infection and how it has evolved inside the patient. Different forms of the disease predominate in different countries; the strains also track gender, sexual practices, needle use habits, and other factors, including the variations in human genes that give different individuals more or less inborn resistance to the infection.
When HIV is viewed from the treatment perspective, we see that medicine is now dealing with at least a dozen different diseases, each forever poised to mutate into some new, untreatable form. Treatments work best when the doctor selects just the right trio of scalpels from the molecular tool kit. Choosing them isn't easy, because so many different variables come into play. Until recently, trial and error played a large role. The doctor started with one mix, monitored viral loads and other biomarkers in the patient's bloodstream, and adjusted the treatment accordingly.
Monitoring and adjusting on the fly remain essential, but the process is now often guided by sophisticated analytical engines fueled by huge collections of patient records, including data on HIV genotypes, treatment histories, and responses, along with patient age, gender, race, and route of infection entry. Patient genotypes are certain to be added soon.
As of late 2011, the largest such engine-the one that powers Europe's EuResist network-was using data from 49,000 patients involving 130,000 treatment regimens associated with 1.2 million records of viral genetic sequences, viral loads, and white blood cell counts. When presented with 25 actual case histories that weren't already in its database, EuResist beat nine out of 10 international experts in predicting how well the treatments would perform.
Here, then, we have a medical world that stands Washington's old regulatory science on its head. Whatever we may call it up here, there is in fact no single disease down there, and the disease down there tomorrow will be different from today's. We have treatments that work, but no single HIV drug can honestly be called "safe" or "effective." All have nasty side effects, and when they are used one at a time, each may fail the individual patient-and endanger others by helping to breed a new drug-resistant strain of the virus.
The FDA has licensed or at least tentatively approved at least eight cocktails. Controlled clinical trials were completed before most of the later-developed drugs were licensed, and once it was clear that only cocktails worked, new drugs had to be tested as part of a cocktail from the get go. But nobody even bothered to pretend that when the FDA licensed them, it had in hand what the federal drug law still requires: "substantial evidence" about how the cocktails would perform when they met any substantial fraction of all the variations in HIV and patient chemistry that they might encounter in the future.
The virus continues to evolve, so the cocktails will remain "safe" and "effective" in any meaningful sense of those words only so long as we continue to prescribe them as directed by continuously updated databases. HIV can always gain time by killing its host more slowly; that gives it more time to evolve in today's host, and also more time for that host to infect his or her successor.
Unblocking Medical Acceleration
The revolution in molecular science and technology has had a dramatic impact in many other areas of medicine, most notably oncology. "Once we understand a cancer cell, we can come up with a treatment very quickly," says Mark Kris, a lung cancer specialist at Memorial Sloan-Kettering Cancer Center in New York.
For a time, many cancer drugs also got the benefit of the Orphan Drug Act and the FDA's accelerated approval rule, and as a result many drugs, after a first round of screening for safety issues, were licensed largely on the strength of early indications of efficacy in even a minority of the patients treated.
Oncologists also routinely take advantage of the older and much broader loophole that allows doctors to prescribe licensed drugs off label. They work out, one patient at a time, how best to use these treatments, which they do by fitting the molecular logic of the targeted drug to the biochemistry presented by the patient's cancer. Genetic matching is becoming as routine as matching a blood transfusion to blood type.
At Massachusetts General Hospital, oncologists have begun sequencing the complete genome of different parts of the tumor to launch a systematic search for multiple targets to guide treatment. Sequencing technologies, together with tools for spotting the best targets for drugs to attack, are now fast, powerful, and cheap enough to make this feasible, and they are getting better and cheaper by the day.
What oncologists currently lack are enough different drugs to fire at enough of the right targets. Tumor sequencing has recently revealed, for example, that 50 to 100 different molecular flaws can make a colon cell cancerous-far more than the 15 to 18 previously suspected. Many of these flaws, however, seem to cause trouble only when paired with one of two other mutant genes, which apparently don't cause cancer on their own. These facts point to various promising treatment strategies, but few of them can yet be tested.
Working with the drugs they do have, oncologists routinely prescribe cancer drugs and cocktails far outside the boundaries that were tested in the FDA-scripted licensing trials. A nonprofit alliance of 21 leading cancer centers evaluates and publishes information on off-label uses of drugs. Off-label and cocktail therapies sometimes end up being steered through the rigid, slow, and expensive trials scripted by the FDA. But as a practical matter, the vast majority never will be: Cancer cells, like HIV, have limitless power to reconfigure their chemistry, they do so in ways that vary too much from patient to patient, and there are therefore too many combinations of drugs, dosages, and patient profiles to explore and calibrate.
The Orphan Drug Act and the accelerated approval rule have now been on the books for more than two decades. By early 1998 the FDA had granted accelerated approval to some 27 cancer and HIV drugs, plus 16 drugs for other conditions, most related to either cancer or AIDS.
In a September 2012 report, President Obama's Council of Advisers on Science and Technology (PCAST) concluded that the accelerated approval rule has "allowed for the development of pioneering and lifesaving HIV/AIDS and cancer drugs over the past two decades" and recommended that the FDA make "full use" of accelerated approval "for all drugs meeting... an unmet medical need for a serious or life threatening illness." At the FDA, however, the trend in much of the past decade has been in the opposite direction, back toward conventional clinical trials.
Critics of accelerated approval have focused principally on the follow-up conventional trials that the rule requires after licenses are issued. The trials aren't always completed, and when they are, they sometimes fail to persuade the FDA that the drug should have been licensed in the first place. The initially promising molecular effects don't always lead to the clinical benefits that ultimately matter.
A first, short rejoinder is contained in the 2012 PCAST report. While there is "some risk" in using predictive biomarkers to accelerate drug approvals, the practice is justified by "the opportunities for progress against serious or life-threatening diseases." While it is impossible to quantify how many lives have been saved or extended by accelerating patient access to HIV and cancer drugs, most of those drugs have since proved their worth in follow-up trials and use by practicing specialists.
Between 1992 and 2010, the rule accelerated patient access to 35 cancer drugs used in 47 new treatments. Twenty-six of those treatments had gone on to complete conventional follow-up trials, trials that had required a median time of almost four more years of investigation. There is little doubt that the rule has spurred rapid and important innovation in the treatment of cancer and HIV, and that it has accelerated access to drugs that have collectively done far more good than harm.
More fundamentally, we should now be using our experience with accelerated approval to take a closer look at the FDA's standard licensing protocols for all drugs. Those protocols, as the PCAST report notes, don't allow the participating doctors to systematically explore the many molecular factors that may determine why a drug performs well in some patients and not others. Most of those factors therefore end up being explored and understood after doctors prescribe the licensed drug and researchers assemble and analyze data from these sources. But those factors play an essential role in prescribing drugs safely and effectively, which means that many good drugs won't make it through the licensing process until enough of the relevant molecular factors have been identified.
The accelerated approval framework launches the search for such factors earlier by issuing licenses on the strength of molecular and other low-level indications of efficacy or, alternatively, clinical benefits in a minority of patients. The license then gives many unblinded researchers and doctors the opportunity to start developing the rest of the drug-biomarker science much sooner. The PCAST report recommends that the FDA adopt new trial protocols that would permit more systematic searches for biomarkers from the outset.
The FDA should allow other drugs aimed at other complex diseases to follow the trail that HIV and, to a lesser extent, cancer drugs have already blazed. We can now see, and can probably find ways to control, almost any of the molecules of life. The hard part is reassembling the pieces-working out the connections between what we can see and probably control down there with what we wish to accomplish up here. We need policies that will allow biochemists, doctors, and patients to collaborate in ways that will unravel how human bodies work without the process taking so long and costing so much that it never gets started at all.