One evening in late May, four senior employees of Merck, the pharmaceutical company, sat in the bar of a Hilton Hotel in Rockville, Maryland, wearing metal lapel pins stamped with the word “TEAM.” They were in a state of exhausted overpreparedness. The next morning, they were to drive a few miles to the headquarters of the Food and Drug Administration and attend a meeting that would decide the future of suvorexant, a new sleeping pill that the company had been developing for a decade. Merck’s team hoped to persuade a committee of seventeen, composed largely of neurologists, that suvorexant was safe and effective. The committee, which would also hear the views of F.D.A. scientists, would deliver a recommendation to the agency. If the government approved suvorexant—whose mechanism, inspired partly by research into narcoleptic dogs, is unlike anything on the market—it would be launched within a year. Some industry analysts had described it as a possible blockbuster, a term usually reserved for drugs with annual earnings of a billion dollars. Merck had not created a blockbuster since 2007, when it launched Januvia, a diabetes drug. The company was impatient. A factory in Las Piedras, Puerto Rico, was ready to start production.
David Michelson, who runs Merck’s clinical research in neuroscience, said of suvorexant, “It’s huge. It’s a major product.” He was sitting perfectly still in his chair; his hair flopped a little over his forehead. He looked as if he were waiting in an airport for a very late flight.
For months, in rooms across Merck’s archipelago of mismatched buildings north of Philadelphia, Michelson had taken part in role-playing rehearsals for the F.D.A. meeting. The focus had been on readying Joe Herring, another Merck neuroscientist; he would be the primary speaker, having run the later clinical trials of suvorexant. Herring, a straight-backed, athletic-looking man in his fifties, had just gone up to his room, for an early night. “Joe had to find a way to be authentic,” Michelson recalled. “He had to find a way to engage with the audience without becoming too informal.” During the meeting, Herring would have access to a library of twenty-one hundred and seventy PowerPoint slides.
The Merck team was frustrated. The F.D.A. had just shown them the draft of a presentation, titled “Suvorexant Safety,” that would be delivered by Ronald Farkas, an F.D.A. neuroscientist who had reviewed thousands of pages of Merck data. In a relentless PowerPoint sequence, Farkas made suvorexant sound disquieting, almost gothic. He noted suicidal thoughts among trial participants, and the risk of next-day sleepiness. He quoted from Merck’s patient notes: “Shortly after sleep onset, the patient had a dream that something dark approached her. The patient woke up several times and felt unable to move her arms and legs and unable to speak. Several hours later, she found herself standing at the window without knowing how she got there.” A woman of sixty-eight lay down to sleep “and had a feeling as if shocked, then felt paralyzed and heard vivid sounds of people coming up the stairs, with a sense of violent intent.” A middle-aged man had a “feeling of shadow falling over his body, hunted by enemies, hearing extremely loud screams.”
An F.D.A. presentation that focusses on individual “adverse events”—and draws attention to patients feeling “hunted by enemies”—is discouraging to a drug’s sponsor. Michelson called the presentation “somewhat unusual,” and emitted a dry laugh.
Darryle Schoepp, the head of Merck’s neuroscience division, was at the other end of the table. During the human trials of suvorexant, he noted, it had been taken two hundred and seventy thousand times, and “every time you take a drug it’s an opportunity for something to happen that the user can report.” He added, “Go back to the early days of Ambien. I wonder how many patient days of data they had with Ambien.”
Ambien, which is now available generically as zolpidem, is one of America’s most popular drugs, and it played a role—silent or spoken—in many conversations that I had heard on visits to the Merck offices. Zolpidem was the cheap drug that suvorexant had to take on, if not unseat, in order to succeed in the sleep-medication market. In addition, rising public worry about risks associated with taking Ambien—ranging from amnesiac devouring of Pop-Tarts to premature death—had reduced the F.D.A.’s tolerance for side effects in sleep medications.
John Renger was also at the bar. A forty-four-year-old neuroscientist, he has a round face, cropped hair, and a neat goatee. He helped lead the company to the suvorexant molecule, and ran the first tests on rats, mice, dogs, and rhesus monkeys. He, too, was politely indignant about the F.D.A. “They’ve taken the emphasis off efficacy,” he said, adding, “They’re saying any residual effects are bad. But they’re not looking at the balance—‘What is the improvement in this mechanism?’ ”
The central nervous system is in an ever-adjusting balance between inhibition and excitation. Ambien, like alcohol or an anesthetic, triggers the brain’s main inhibitory system, which depends on binding between GABA—gamma-aminobutyric acid, a neurotransmitter—and gaba receptors on the surface of billions of neurons. gaba receptors can be found throughout the brain, and when they’re activated the brain slows. Ambien encourages the process by sticking to the receptors, holding open the door to the neurotransmitter. Suvorexant, which Merck describes as “rationally designed”—rather than stumbled upon, like most drugs—influences a more precise set of neurotransmitters and receptors. Orexin neurotransmitters, first identified fifteen years ago, promote wakefulness. When suvorexant is in the brain, orexin is less likely to reach orexin receptors. Instead of promoting general, stupefying brain inactivity, suvorexant aims at standing in the way of a keep-awake signal. This difference may or may not come to mean a lot to insomniacs, but Merck’s marketing is likely to encourage the perception that suvorexant ends the dance by turning off the music, whereas a drug like Ambien knocks the dancer senseless.
If the Merck scientists succeeded at the F.D.A., they would be the first to bring an orexin-related drug to market. “It’s an amazing achievement,” Richard Hargreaves, the fourth colleague at the Hilton, said. “Everyone should be really proud.” But, he added, “my worry is that a new mechanism is being evaluated on the science of an old mechanism.”
“With Ambien, you’ve got a drug that’s got basically only onset,” Renger said, dismissively. That is, it sends you to sleep but might not keep you asleep. “Suvorexant has the onset, but it has the great maintenance, especially in the last third of the night, where other drugs fail.” And even though suvorexant keeps working longer than Ambien, suvorexant patients don’t feel groggier afterward, as you might expect. Impassioned, Renger imagined himself addressing the F.D.A.: “Why aren’t you giving this a chance?”
“Drugs usually have some side effects,” Schoepp said. “It’s all benefit-risk.” He added, “There is some dose where suvorexant will be ultimately safe—because nothing will happen. If you go low enough, it becomes homeopathic.”
They stood to go to their rooms. Schoepp murmured, “I’d love to take it right now.”
Jean-Pierre Kaplan lives in a southern suburb of Paris. When I visited him this summer, he tricked his elderly dog into thinking there was a cat in the front yard that needed chasing, and then we sat down to lunch with Marie-Louise Pelus-Kaplan, his wife. Kaplan is seventy-four, and when he retired, in 2000, he was a patent lawyer. Before that, he was a chemist in the pharmaceutical industry, a career that ended unhappily. In the late seventies, while working in a laboratory a few miles from where we were eating, he co-invented the drug that became known as Ambien. Kaplan’s name is one of two on the French and American patents.
In 2006, Ambien’s manufacturer estimated that it had been taken twelve billion times worldwide. The drug was worth two billion dollars a year in American sales. (Ambien, which was patented in the U.S. in 1981, went generic in 2007.) Last year, there were sixty million prescriptions for sleeping pills in the U.S., and forty-three million of them were for some form of zolpidem, including Ambien C.R., a deftly repatented controlled-release pill. Over lunch, I asked Kaplan, who has not previously given interviews, if he’d ever taken Ambien. “Never,” he said, in accented English. “I sleep very well.”
Pelus-Kaplan, a retired professor of early-modern history, teasingly explained that her husband almost never takes medication. He allows his doctor to write prescriptions, and he even picks up the pills at the pharmacy, “but he never eats one. He says, ‘Too dangerous.’ ”
“If I need a drug, I would take it, but I don’t need it!” Kaplan said. “When I get flu, I stay in bed. Drugs are very important when you don’t want to lose time, but when you have plenty of time you stay in bed.” His annual drug intake, he estimated, was no more than ten over-the-counter painkillers.
Zolpidem is part of a third generation of synthetic compounds that treat insomnia by attaching to GABA receptors. Such drugs were first introduced a century ago, long before the gaba system was identified. The first generation, barbiturates, effectively induce sleep, but can be addictive, and it’s easy to overdose on them (Marilyn Monroe, Judy Garland, Jimi Hendrix, Jean Seberg). In the second, safer generation are benzodiazepines, a class that has some mixture of sedative, muscle-relaxant, anticonvulsant, anti-anxiety, and amnesiac effects. These were invented in the fifties by Leo Sternbach, a chemist at Hoffmann-La Roche in New Jersey. He synthesized Librium and then Valium, which, between 1968 and 1981, was the most frequently prescribed drug in the Western world. Valium was marketed as a treatment for anxiety, but insomniacs also used it. The first benzodiazepine explicitly approved by the F.D.A. as a sleep medicine was Dalmane, launched by Hoffmann-La Roche in 1970. Halcion, a benzodiazepine from Upjohn, became the world’s best-selling sleep aid after its launch, in 1982.
In the early seventies, Sternbach visited the offices of Hoffmann-La Roche in Basel, Switzerland, and ran into Jean-Pierre Kaplan, who then worked there. Sternbach shook Kaplan’s hand, and wished him well. Kaplan, who had grown up in Paris, was then a few years out of college. A mountain climber, he was long-haired and instinctively unaccommodating. “I felt very free,” he said. “I had a very different comportment from the Swiss researchers. I did not fear anybody.” (He was once told that he was the first Jewish scientist to be employed at the site. He is not Jewish.)
In 1973, Kaplan took a new job, at Synthélabo, in Bagneux, near Paris. L’Oréal had just bought a majority stake in the firm, and wanted to turn it into a major pharmaceutical force. After he identified some compounds with anticonvulsive properties, he felt that one of them was being improperly accelerated toward commercial development. He considered the drug ineffective at safe doses—“a big waste of money for the company.” (The drug, progabide, was eventually approved as an epilepsy treatment in France, but not elsewhere.) Kaplan says that this disagreement, along with his activities in a trade union, had already begun souring his relations with the company when, in 1978, a colleague made a passing suggestion: why not try to build something “a little like zopiclone”?
Zopiclone, a compound that had been created several years earlier by a rival company, was an interesting oddity. Although its chemical structure was quite unlike that of a benzodiazepine, it acted just like one. It eventually beat zolpidem to market, as the first in a new category of sleep medication: “z-drugs,” or non-benzodiazepines. (Lunesta, approved by the F.D.A. in 2004, is a close variant of zopiclone.) Kaplan recalled, “I thought, O.K., if zopiclone and benzodiazepines act on the same brain receptor, why don’t I try to make another drug—a hybrid? That was the gist of the invention.”
The molecule, when finished, had another important characteristic. At the lunch table, Kaplan began sketching in my notebook the chemical structure of LSD. He drew hexagons attached to other hexagons. One of them had a tail of nitrogen, oxygen, and carbon. This tail helps to make LSD unusually effective at reaching the brain from the bloodstream. He said of LSD, “I knew that this kind of structure was very active in the brain.” A similar tail was incorporated into zolpidem. “This was important, to have activity in the brain. Maybe it increases the activity one hundred or one thousand times.” (I asked Kaplan if he had taken LSD. “Never!”)
He and Pascal George—a younger colleague whom Kaplan described as “sympathetic and brilliant”—started by building wooden models, including ones for Valium, Halcion, and zopiclone. Colored one-inch spheres, representing atoms, were connected by thin rods, creating models the size of a shoebox. This was a more empirical, architectural approach than is typical in a lot of pharmaceutical chemistry. Kaplan and George tried to identify what these molecules had in common, structurally, that allowed them to affect the brain in the same way. Kaplan told me that their thinking wasn’t wildly creative, but it was agile: “You know, at that time it was maybe clever, because you have no computer. Now it’s routine work.”
George wrote a report describing a few possible types of new chemical compounds. Working separately, they built molecules of the first two types: about ten of one, five of the other. These were unpromising. A third series, made by George, looked better. When it was tested on animals, Kaplan said, “it was clear that it would be a great success. After the very first compound, I knew.” But in 1980, while this work was still under way, Kaplan was taken off the project. In his account, Synthélabo, eager to get rid of him, “didn’t want to give me the merit of the invention.” From then on, George ran the research. Kaplan heard only rumors about how the compounds were testing.
That fall, Synthélabo applied for a French patent on a series of seventy-seven compounds. The company knew that one of the compounds had far more pharmaceutical promise than the others, but did not need to disclose this to industry competitors. So the star molecule was also hidden from Kaplan, even though his name was at the top of the document. He showed me the patent. “I was named the first inventor, but did not have the results of the compound I proposed!” he said. He looked down a list of seventy-seven chemical formulas, and pointed to the seventy-fifth: this was Ambien.
Although Kaplan felt increasingly unwelcome at Synthélabo, he would not resign. The company eventually moved him to an office in central Paris, to a phantom job in an empty room. He left Synthélabo in 1984, and never worked as a scientist again, considering himself blacklisted. “He was furious,” Marie-Louise Pelus-Kaplan said. “He went biking in the woods to think. . . . And then he decided to study law.”
In a friendly tone, Pascal George told me, “Jean-Pierre was very intelligent, but very suspicious. I would say paranoiac. Since he was paranoiac, he was very happy to be frustrated—it was a part of his happiness.” (There is no evidence of ill will between him and Kaplan.) After the patent filing, it took some years before zolpidem reached the market. George recalled that “the internal resistance” at Synthélabo was “rather strong.” Among other things, zolpidem had been conceived by chemists, not biologists, which was unusual. He said that drug development accelerated in 1985, when a company pharmacist, preparing a batch of syrup for the first human trial, accidentally swallowed a teaspoonful of the drug. He immediately fell asleep.
Synthélabo also established that zolpidem was “selective” in its influence on the GABA system. Zolpidem had more impact on sleep than on amnesia, muscle relaxation, and the other effects associated with drugs that bind to gaba receptors. In theory, at least, this selectivity meant that the drug would have fewer undesirable outcomes.
Zolpidem was launched in France in 1988. Five years later, it was brought to America, with the name Ambien, in a joint venture between Synthélabo and Searle. George became the drug’s acknowledged inventor, and Kaplan was sometimes left out of official accounts—“like in the former Soviet Union,” he said. Pelus-Kaplan once attended a conference, incognito, to confirm that her husband was being overlooked. It was George who built the molecule, but Kaplan argues that the initial collaboration created the blueprint for all that followed. George agrees.
Ambien had the good fortune to reach the market just as the reputation of Halcion, which had been promoted as safer than barbiturates, collapsed. The public was concerned about Halcion’s perceived side effects—including amnesia and panic—and about reports that Upjohn had suppressed unfavorable data from its trials. William Styron, in his 1990 memoir, “Darkness Visible,” blamed Halcion for amplifying his suicidal thoughts. Philip Roth, in “Operation Shylock,” drew on his own reaction to Halcion, describing a “mental coming apart” that was “as distinctly physical a reality as a tooth being pulled.” In 1991, Upjohn settled a suit brought by a woman who had shot and killed her eighty-two-year-old mother after taking Halcion, and Time ran a story on “The Dark Side of Halcion.” That year, the drug was banned in Britain. (It remains available in the U.S., but it is no longer a best-seller.)
Searle and Synthélabo presented Ambien as a safe alternative to Halcion. Jed Black, a sleep specialist at Stanford’s medical school who has worked in the pharmaceutical industry, recently recalled being visited by Ambien salespeople: “They would say to me, with a very straight face—and I think they believed it completely—‘This is not a benzodiazepine, and therefore it’s safer.’ ” Ambien did send people to sleep quickly, and the human body broke it down after a few hours, so there was a limited hangover effect. And a fatal overdose would be very hard, if not impossible, to engineer. (Ruth Madoff told “60 Minutes” that she and Bernie Madoff failed to commit suicide with Ambien: “We took pills and woke up the next day.”) But, like benzodiazepines, Ambien sometimes caused amnesia and confusion. According to Black’s reading of published data, the drug was selective—focussed on sleep—in its action on GABA receptors, but only in doses that were too low to induce sleep. At useful doses, it “became indistinguishable” from a benzodiazepine. Nevertheless, Ambien was accepted as a better drug. “Everyone bought into it,” Black said. The situation hasn’t changed. He noted that, when he lectures to physicians at Stanford, “I’ll say, ‘Who here would be equally happy to prescribe Halcion and Ambien?’ And none of them raise their hand. Then I show them the data.”
Ambien quickly became the national best-seller in its category. As Black recalled, “Everybody switched allegiance—most physicians did—and then nothing came along that was any better.” Customers were satisfied, because the drug reliably induced sleep, and, as Black noted, sleep drugs that target GABA receptors “impart a sense of feeling a little less stressed, like you’ve had a drink or two.” And Ambien, in common with many other drugs, can be tricky for some patients to give up. Those who stop abruptly may experience “rebound” insomnia that is worse than when they started. Black said, “And they inaccurately assume, ‘Oh, my insomnia’s really bad still.’ ” He laughed. “It’s actually a nice feature for a drug to have, from a pharmaceutical perspective.”
By the turn of the century, there were more U.S. prescriptions for Ambien than for all benzodiazepines combined, and Ambien’s benign reputation seemed to help normalize the idea of medical assistance for insomnia. (In 1998, Kathy Giusti, at the time a Searle executive, explained to an interviewer, “We had to change consumer perception about the sleep category in general, to eliminate the stigma.”) Between 1993 and 2006, the number of times a year that a U.S. doctor gave a diagnosis of insomnia rose from fewer than a million to more than five million.
In 1995, Kaplan negotiated a payment—about thirty thousand dollars—from his former employers. George, who stayed at the company, happily, until his retirement, in 2010, received a little less. After Kaplan retired from his career in law, he formed an organization that lobbies on behalf of people who invent things while working as a salaried employee.
Kaplan described zolpidem as a “professional disaster.” He added, “It’s not lifesaving, it does not treat cancer, it does not treat malaria, it does not treat Alzheimer’s—the most difficult illnesses to treat. Therefore, I call it a comfort drug.”
A woman recently posted online a description of her Ambien experiences:
Ambien can be disinhibiting and depersonalizing. Or, to quote from the label of a bottle of sleep medication used by Tina Fey’s character, Liz Lemon, on “30 Rock”: “May cause dizziness, sexual nightmares, and sleep crime.” Zolpidem enters the gut, passes into the bloodstream, squeezes through the liver, and then crosses the blood-brain barrier, to make GABA receptors more receptive to gaba. When the neurotransmitter sticks to its target, negatively charged chloride ions flow into cells, making the inside of the cells more negative, and less likely to fire. Traffic is interrupted, signals don’t reach their destinations, and the brain starts to quiet. Many people experience this as a contented swoon that silences inner chatter while giving a half glimpse of childhood; they are overtaken by sleep, like a three-year-old in a car seat.
But others resist sleep and embrace the woozy, out-of-body license. To some, this is an opportunity to take part in what Rachel Uchitel, a former girlfriend of Tiger Woods, has reportedly described as “crazy Ambien sex.” At the London Olympics, some Australian swimmers took Ambien to build team spirit. After taking the drug, they larked around and knocked on the doors of other athletes. As one of them later put it, they allowed themselves “to be normal for one night.” Because the drug had been banned by the Australian Olympic Committee, and because the team failed to win medals that it was expected to win, this became a national scandal.
But for many Ambien users, like the eBay shopper, their activities on the border of wakefulness and sleep are less purposeful. Drew Fairweather, an online cartoonist, has described the phenomenon in a popular series of panels in which a walrus addresses a human companion with such suggestions as “Take some more Ambien and cut off all your hair, man. Let’s do this.” In 2006, Patrick Kennedy, then a congressman, crashed his Mustang into a barrier near Capitol Hill, in the middle of the night; he told police, inaccurately, that he was late for a vote. He had Ambien and an anti-nausea medication in his body. By the following spring, the F.D.A. had heard enough about Ambien-related sleep-driving, sleep-eating, and sleep-walking—accompanied by amnesia—to require new warnings. The drug’s label now refers to the risk of “preparing and eating food, making phone calls, or having sex.”
This kind of behavior can occur during dreamless, slow-wave sleep—the state of an unmedicated sleepwalker—or, more commonly, Jed Black suspects, while someone is awake but disinhibited, by Ambien alone or by Ambien and alcohol. Black noted that this altered state can be mischaracterized as sleep by people who have forgotten their adventures. A recent study, described in European Neuropsychopharmacology, suggests that these phenomena affect five per cent of users. (Other studies have reported lower numbers.) Zolpidem’s reputation for outlandish side effects may be inflated by gossip—by the interaction of medication and the Internet. Thomas Roth, the director of the sleep center at Henry Ford Hospital, in Detroit, who has consulted for Merck and other pharmaceutical companies, told me he has not yet seen persuasive evidence that there is more of this behavior among Ambien users than among the rest of the population (which includes drinkers). The F.D.A.’s 2007 warnings were prompted by doctors’ reports, not by peer-reviewed data. But amnesiac confusion certainly occurs, and zolpidem’s popularity makes misadventures commonplace, to the point that it’s hard to use Ambien in a criminal defense. Defendants must argue that they were involuntarily intoxicated—that they couldn’t have foreseen the possible consequences of taking Ambien, alone or with drinks—despite the warnings delivered both by their doctor and by Charlie Sheen, who called the drug “the devil’s aspirin” after an incident, in 2010, involving a porn star and a damaged chandelier, in the Eloise Suite of the Plaza Hotel.
There may be other risks associated with zolpidem. In a recent paper in the online edition of the British Medical Journal, Daniel Kripke, a professor emeritus at the University of California San Diego School of Medicine, examined five years of electronic medical records collected by a health system in Pennsylvania. He compared more than ten thousand patients who had been prescribed a sleep medicine—most commonly Ambien—and more than twenty thousand patients who had not. After adjusting for age, gender, smoking habits, obesity, ethnicity, alcohol use, and a history of cancer, and after controlling, as much as possible, for other diseases and disorders, Kripke found that people who had taken sleeping pills were more than three times as likely to have died during the study period as those who had not. Those on higher doses of the drugs were more than five times as likely to have died.
“My best estimate is that drugs like zolpidem are killing as many people as cigarettes,” Kripke told me recently. That is, more than four hundred thousand Americans a year. “And suppose they’re only killing a tenth as many people—you still wouldn’t want them on the market.” Echoing Ambien’s co-inventor, Kripke called the risks unnecessary. “Nobody dies because they didn’t take a sleeping pill,” he said.
Kripke acknowledges that his study did not identify the cause of any death; ill people take more sleeping pills than others, and some users might have had illnesses that were undiagnosed, and therefore not controlled for in the study. And insomnia itself could present a significant health risk, although Kripke resists that idea. Jed Black finds the data interesting but too inconclusive. A representative of Sanofi—the company that Synthélabo became part of, after various mergers—told me that Sanofi stood behind its Ambien safety data, which had satisfied the F.D.A.
Other research has linked zolpidem and similar drugs to depression, suicide, and car accidents; there are also data connecting zolpidem to cancer. (Such numbers do not establish causation.) The U.S. Substance Abuse and Mental Health Services Administration recently reported that E.R. cases involving zolpidem had risen from six thousand, in 2005, to nineteen thousand, in 2010.
If the public has largely overlooked such data, even as it pays attention to Patrick Kennedy—or to his cousin Kerry Kennedy, who was arrested last year with zolpidem in her body, having driven for several miles on a shredded tire after colliding with a tractor-trailer—it may be because Ambien deaths are disguised by circumstances. “The people who die after taking sleeping pills tend to be older and obese, and to have multiple illnesses,” Kripke said. “So if they happen to die in the middle of the night nobody supposes that it’s from the sleeping pill. And there’s no way of proving that it was.”
John Renger, the Merck neuroscientist, has a homemade, mocked-up advertisement for suvorexant pinned to the wall outside his ground-floor office, on a Merck campus in West Point, Pennsylvania. A woman in a darkened room looks unhappily at an alarm clock. It’s 4 A.M. The ad reads, “Restoring Balance.”
The shelves of Renger’s office are filled with small glass trophies. At Merck, these are handed out when chemicals in drug development hit various points on the path to market: they’re celebrations in the face of likely failure. Renger showed me one. Engraved “MK-4305 PCC 2006,” it commemorated the day, seven years ago, when a promising compound was honored with an MK code; it had been cleared for testing on humans. Two years later, MK-4305 became suvorexant. If suvorexant reaches pharmacies, it will have been renamed again—perhaps with three soothing syllables (Valium, Halcion, Ambien).
“We fail so often, even the milestones count for us,” Renger said, laughing. “Think of the number of people who work in the industry. How many get to develop a drug that goes all the way? Probably fewer than ten per cent.”
In 1998, when Renger was in Japan, finishing his postdoctoral work, two groups of scientists announced almost simultaneously that they had identified, in rodents, a previously unknown neurotransmitter. One group, in San Diego, called it hypocretin, after the hypothalamus, the area of the brain where it is produced. The other team, in Dallas, called it orexin, as in “orexigenic,” which means “appetite-stimulating.” Its primary function was thought to be the regulation of food intake. Orexin-abundant mice gained more weight than others on the same diet. (The naming question has still not been settled, although “orexin” is more widely used in nonacademic circles, including pharmaceutical companies. Renger referred to hypocretin partisans, affectionately, as “a stubborn group.”)
The orexin papers were widely noticed, in part because of the connection to feeding. Several pharmaceutical companies, including Merck, began investigating possible obesity treatments. A year later, a remarkable paper from Stanford sent everyone in another direction.
Since the seventies, Stanford sleep scientists, led first by William Dement, had bred narcoleptic dogs. This was an achievement in itself. The animals suffered from extreme daytime sleepiness and had a propensity for mid-coital collapse: at moments of high emotion, the dogs, like narcoleptic humans, experienced sudden muscle weakness, or cataplexy. The first Stanford dog was a poodle named Monique. Later, there were other breeds; the Stanford colony, mostly Dobermans, had eighty dogs at its peak. Narcoleptic dogs gave birth to narcoleptic puppies; the disorder in canines has a single genetic cause. In 1999, after a decade-long search, a team led by Emmanuel Mignot, a researcher at Stanford, located the damaged gene, and reported that it encoded a receptor: the same one that had just been identified by the work done in California and Texas. Narcoleptic dogs lacked orexin receptors.
Mignot recently recalled a videoconference that he had with Merck scientists in 1999, a day or two before he published a paper on narcoleptic dogs. (He has never worked for Merck, but at that point he was contemplating a commercial partnership.) When he shared his results, it created an instant commotion, as if he’d “put a foot into an ants’ nest.” Not long afterward, Mignot and his team reported that narcoleptic humans lacked not orexin receptors, like dogs, but orexin itself. In narcoleptic humans, the cells that produce orexin have been destroyed, probably because of an autoimmune response.
Orexin seemed to be essential for fending off sleep, and this changed how one might think of sleep. We know why we eat, drink, and breathe—to keep the internal state of the body adjusted. But sleep is a scientific puzzle. It may enable next-day activity, but that doesn’t explain why rats deprived of sleep don’t just tire; they die, within a couple of weeks. Orexin seemed to turn notions of sleep and arousal upside down. If orexin turns on a light in the brain, then perhaps one could think of dark as the brain’s natural state. “What is sleep?” might be a less profitable question than “What is awake?”
Mignot had done something very unusual: he had discovered the genetic cause of a condition, helped to reframe thinking about a fundamental human behavior, and revealed clear pharmaceutical opportunities. An orexin receptor is the kind of place that many existing drugs are designed to reach. As Mignot put it, “This was druggable.” (That is often not the case: researchers know the genetic cause of Huntington’s disease but have nothing to target.) A drug that activated orexin receptors might help treat narcoleptics, and a drug that blocked orexin receptors, if introduced to a brain producing orexin at unwelcome times, might help insomniacs, perhaps without intoxicating them. Pharmaceutical companies were reluctant to give up their obesity-drug ambitions, but it seemed that the orexin mice described in 1998 were fat because they stayed up late and had more time to eat.
Research at Merck had long focussed on eleven diseases, including Alzheimer’s and diabetes. Insomnia was not one of them. Renger, who joined the company in 2001, recalled, “The perception at that time was, You have a lot of medications available—should we be working on this?” How large was the population of insomniacs poorly served by Ambien? Should Merck invest in a market dominated by a drug that, within a few years, would become a cheap generic? The need for an “orexin-antagonist” sleep aid was neither commercially overwhelming nor clinically pressing. (Indeed, one detects a little professional defensiveness from the suvorexant team. Renger can sound effortful when describing the distress of insomniacs: “We’ve got to think of the patients! That’s why we make medicines.”)
But orexin-related work promised pharmaceutical novelty, which is extraordinarily uncommon. Most new drugs are remixes of old drugs—clever circumventions of patent protections. The last truly original medicines in neuroscience were triptans, for the treatment of migraines, introduced in the early nineteen-nineties. “The science is really what drove us,” Renger said. “To have a new target—to know the genetics of the brain’s control system and to be able to focus on that specifically to control sleep—is a pretty rare event. It’s like the thing people keep promising: you know, the ‘cancer gene.’ This was the first time there was the ‘sleep gene.’ ”
The work was also feasible. It’s easier to observe sleep than, say, a reduction in anxiety or depression. Renger, upon his arrival at Merck, had set up a sleep laboratory that could make very fast, semiautomated measurements of the sleep patterns of rodents and monkeys. The lab was designed to identify sleep-related side effects of Merck compounds, but was well suited for testing insomnia treatments. “With sleep, you can do an EEG study in a few days, and it’ll tell you whether or not we’re having an impact. I could do these studies”—he snapped his fingers—“and get an answer.”
Merck has a library of three million compounds—a collection of plausible chemical starting points, many of them the by-products of past drug developments. I saw a copy of this library, kept in a room with a heavy door. Rectangular plastic plates, five inches long and three inches wide, were indented with hundreds of miniature test tubes, or wells, in a grid. Each well contained a splash of chemical, and each plate had fifteen hundred and thirty-six wells. There were twenty-four hundred plates; stacked on shelves, they occupied no more space than a filing cabinet.
In 2003, Merck conducted a computerized, robotized examination of almost every compound in the library. At this stage, the scientists were working not with Renger’s animals but with a cellular soup derived from human cells and modified to act as a surrogate of the brain. Plate by plate, each of the three million chemicals in the library was introduced into this soup, along with an agent that would cause the mixture to glow a little if orexin receptors were activated. Finally, orexin was added, and a camera recorded the result. Renger and his colleagues, hoping to find a chemical that sabotaged the orexin system, were looking for the absence of a glow.
I visited the room in which this work had been done. Yellow robotic arms, on the same scale as car-assembly robots, were moving the trays from here to there, making bursts of sound like a nut being loosened in a tire shop. “Summertime” played on a radio. A computer monitor showed enhanced images of reactions on the plates: a fuzzy grid of light and dark dots, like a blurry telescope image of distant stars.
The robots ran through Merck’s collection in about three weeks. “If something’s interesting, you grab that by the neck,” Renger said. The molecules that best blocked orexin receptors were re-screened, in various ways. Chemists then modified the most promising candidates, much in the way that the Synthélabo chemists had worked twenty-five years earlier: they induced chemical change by heating and mixing, to build families of drug-like compounds. These were then tested on human liver cells, in vitro, and on animals in Renger’s sleep lab.
Renger took me to see the rats and monkeys. The lab has soundproofed walls built out of the kind of air-infused blocks used in bomb shelters. The rats were transmitting live EEG data, wirelessly, from brain implants. So were the monkeys; they also had touch-responsive screens in their cages, on which they sometimes played games, for rewards of juice. A red square might appear on the screen and then disappear; after a pause, a red square might appear alongside a yellow square, and the monkey would be rewarded for touching the red one. (“It’s like drinking soda and playing a little bit of Assassin’s Creed,” Renger explained.) With these games, Renger could simultaneously measure wakefulness and cognition. During the orexin research, when it was necessary to intrude on the monkeys’ sleep, he played the amplified sound of a tiger’s growl.
The work went back and forth between the chemists and the biologists: compounds were improved and tested. In December, 2006, Renger put on a good suit and drove with his team to Merck’s offices in Branchburg, New Jersey. At the monthly meeting of the pre-clinical-development review committee, they pitched their best bet to the company. “We had what we thought was a fantastic molecule,” Renger recalled. “It had all the properties we thought we would need, and it was going to look like a drug.” It seemed likely that suvorexant would have a far longer half-life than Ambien, which implied a risk of next-day effects. But Renger wanted the drug to extend sleep. “We wanted to have something that covers this system for the entire night,” he said.
Merck approved the compound. “At that point, the might of the corporation swings in behind the science,” Richard Hargreaves, who helped run the meeting, told me. The company was now likely to fund at least a year or two of work. To bring a drug to market now costs an average of about two billion dollars, Hargreaves said. Renger and his team celebrated with drinks at the Cock ’n’ Bull, in Lahaska, Pennsylvania.
Despite years of sleep problems, Samar Chatterjee, a seventy-year-old environmental engineer, had until recently never taken a sleep aid. Chatterjee, who lives in Washington, D.C., told me that he had feared “getting hooked on the drugs, and getting dozy and dopey.” He referred to the extreme example of Michael Jackson, who, at the time of his death, in 2009, was taking a general anesthetic, apparently as a remedy for insomnia. But, in 2010, Chatterjee saw an advertisement for a sleep-medication trial, at the Center for Sleep & Wake Disorders, in Chevy Chase, Maryland, and he applied. He thought that the study might benefit society, and he hoped to learn if he had sleep apnea: people with the condition would not be allowed to participate. After being monitored over two nights of imperfect sleep, at the Chevy Chase center, Chatterjee learned that he did not have sleep apnea, or other complicating conditions, and that he was sufficiently insomniac to join the trial. (The center, one of many contracted by Merck, heard from five hundred applicants, but found only seventeen who met all the criteria.) He took a tablet every night for three months; he was usually at home, but sometimes in a bed in Chevy Chase, where EEG readings, and other measurements, recorded “sleep efficiency” (percentage of time in bed spent asleep); L.P.S. (Latency to Persistent Sleep: the speed with which a person falls asleep); and WASO (Wake After Sleep Onset: the time spent awake in bed after initially falling asleep). When Chatterjee slept at home, he delivered an account of his night to the center, through an automated telephone questionnaire. He suspected, correctly, that he was taking a drug rather than a placebo. He fell asleep faster than usual, and stayed asleep. This seems to have pleased him, but left him ambivalent about insomnia medication. I asked him about side effects. “Nothing major,” he said. “Some constipation. Maybe some dizziness or pain. Headache, that type of thing.” He also experienced some sleepiness in the afternoons.
Drug trials usually have three phases, and Chatterjee had taken part in the final phase of the suvorexant trials. The Phase I trials, begun in 2007, tested for safety. Non-insomniac volunteers—the researchers called them “healthies”—took the drug at high or low doses, or with other drugs, or the night before a supervised, hour-long highway drive in which they were told not to drift out of their lane.
These trials had barely begun when, in February, 2007, Nature Medicine published a paper, “Promotion of Sleep by Targeting the Orexin System in Rats, Dogs and Humans,” written by scientists at Actelion, the Swiss pharmaceutical company. Merck knew that other firms had built orexin antagonists, but Actelion’s paper showed that it was clearly ahead of Merck, perhaps by a year or two.
“We thought, O.K., great,” Renger recalled, with a sigh. But the news also galvanized the Merck team: “We were already highly motivated, but seeing someone jump in front gives you that extra kick in the ass.”
In 2008, results from Phase I studies of suvorexant showed that it was safe enough to go forward. The data also provided enough indications of efficacy—by sending “healthies” to sleep—to allow Merck to accelerate its process, and skip a formal proof-of-concept stage. In late 2008, suvorexant began a Phase II trial, involving two hundred and fifty-four insomniacs in the U.S. and Japan. The results would establish the doses for much larger, and more expensive, Phase III trials, whose results are at the center of any submission to the F.D.A. In Phase II, Merck tested the drug at ten, twenty, forty, and eighty milligrams. Sleep measurements were taken by observing patients in the lab, and by collecting sleep diaries.
Daniel Kripke, of U.C. San Diego, argues that the effectiveness of insomnia treatments should be judged by patients’ ability to function the next day. But the pharmaceutical companies, and the F.D.A., judge a sleep drug by its impact on sleeplessness. That impact is assessed objectively, with electronic monitoring, and subjectively, using patient reports. Objective data show that insomnia medications, on average, provide a gain of only ten or twenty minutes in total sleep time. But a patient’s perception of improved sleep is also a recognized part of the clinical data. In this framework, insomnia is a condition not just of losing sleep but of being disturbed by sleeplessness. Indeed, most people with prescriptions for insomnia never visit a sleep lab, trusting their own assessment of a sleep deficit. This emphasis on the subjective also makes the amnesiac effect of sleep drugs oddly advantageous to those who manufacture them: the drugs inhibit people from creating memories of waking during the night.
The Phase II results were strong: suvorexant worked on insomniacs. Renger recalled that the team was ebullient: “A novel mechanism in neuroscience—whatever happens from there on, you’ve done something in your career.” They took a day trip, with families, to an aquarium in Camden, New Jersey.
By then, the company had begun considering which of the four doses of suvorexant it should take into Phase III. The placebo effect of sleep drugs is powerful. A recent paper in the British Medical Journal suggested that it accounts for half the effect of z-drugs. So insomnia medications need to be quite potent to distinguish themselves from a placebo in clinical trials. The Phase II results showed that, at ten milligrams, suvorexant had an effect that could be measured in a sleep lab, but the dose had no advantage over a placebo in the subjective measures—patients’ estimations of their own speed in falling asleep, subsequent wakefulness in bed, and total sleep time.
Merck then made an important decision. For Phase III, starting in late 2009, it would drop ten and eighty milligrams in favor of twenty and forty milligrams, with forty regarded as the likely standard dose. In Phase III, Merck would also test fifteen- and thirty-milligram doses on patients sixty-five and older, who were more sensitive to the drug. The Chevy Chase sleep center, along with more than a hundred other facilities around the world, was contracted to test the four doses. Eighteen hundred patients participated in the trial.
At the time, Jed Black, the Stanford sleep specialist, was on a two-year leave of absence, working full time on almorexant, the rival drug made by Actelion. Phase III trials of the drug were under way. This work has not been published, and Black cannot discuss it, although he recently described almorexant as having “an absolutely remarkable profile” that was likely to outperform zolpidem in sleep maintenance.
But, in early 2011, Actelion announced that it was halting the drug’s development, because of an undisclosed possible safety issue. Merck’s scientists speculated about the nature of the concerns, and feared for the future of suvorexant. Black said that the problem was “straightforward,” but that Actelion had decided to pause and take its time. “I don’t think almorexant needs re-tinkering at the molecular level,” he said, implying a problem of drug delivery. Black, who is back at Stanford, suspects that almorexant will be launched, and is certain that such drugs will eventually become dominant. (GlaxoSmithKline recently published results, from Phase II studies, of its own orexin antagonist.)
Actelion’s Phase III trials had included a comparison of its drug’s performance with zolpidem’s. Merck used zolpidem in two tiny studies, but not in larger ones. This omission might seem surprising. If suvorexant really was a possible Ambien killer, then couldn’t its superiority have been demonstrated in comparative studies? Merck scientists sometimes seemed evasive in their responses to this question, but an answer eventually came into focus. On the core issues that interest the F.D.A.—efficacy and safety—a de-facto head-to-head would become available; anyone could compare published data about the two drugs. But it was risky to go beyond those requirements, even if such trials might have demonstrated other possible strengths of suvorexant: a lower chance of nighttime confusion, perhaps. The trials would have slowed suvorexant’s sprint to market, and they would have been very hard to engineer: Merck would have had to use safe, low doses of the two drugs, and the differences between them might have been subtle, if they existed at all. Suvorexant might even have lost the contest, and Merck would have been obliged to include that information in its filing with the F.D.A. “We were in competition, and we were behind,” Joe Herring said to me. “We wanted to get across the line with a lean program.”
The real-world test—a double dose, three glasses of wine, and a laptop—would take place after F.D.A. approval. In the meantime, Merck scientists who spoke publicly about suvorexant had to restrict themselves to the data from F.D.A.-sanctioned trials; they could not discuss strengths that the drug seemed only likely to have. They had an impressive narrative about the creation of a rational, novel, and “beautiful” molecule. But they couldn’t display a chart showing that suvorexant was, say, less likely than Ambien to lead to such episodes as “cooking yourself breakfast and forgetting the next day,” as Renger put it.
One possibly significant difference between suvorexant and Ambien may be indicated, informally, by Merck’s Phase III trials, though it wasn’t part of the official results. There were no reports of euphoria—a word that is on Ambien’s label. Thomas Roth, of Henry Ford Hospital, said of suvorexant, “I would not expect any kind of high before sleep.” Most people would regard clear-headedness as a pharmacological virtue, but to some the Ambien buzz is a pleasure enhanced by the comforting promise of imminent sleep.
Merck’s decision to forgo more comparative data was “quite bullish,” Black said, but not unreasonable. “They thought that they had a good safety profile, and that there would be no problems at the F.D.A.” But, as he noted, “the F.D.A., particularly under the direction of Ron Farkas, seems to be raising the bar a bit on safety.” This year, the agency lowered the recommended dose of zolpidem for women from ten milligrams to five.
If there are Merck employees who regret the decision, I didn’t hear them say it. But a recently published paper, written by Renger and others at Merck, offered hints about how suvorexant might have performed in a comparative study. The paper described an experiment involving rodents and monkeys dosed with Ambien, Valium, Lunesta, and a Merck compound called DORA-22—another orexin antagonist that Merck made alongside suvorexant. The dora-22 study first established the amount of each drug necessary to send the animals to sleep, and then—using cognitive tests like the red-square game—measured the extent to which the drugs, soon after ingestion, affected memory and attention span. Renger was “elated” by the results. dora-22 performed far better than its rivals. In one test, monkeys administered thirty times the sleep dose of dora-22 showed no impairment after being woken and given an attention test. The Ambien monkeys were dozily incompetent even at doses too low to have initiated sleep.
The unspoken promise of orexin antagonists, then, is sleep without stupidity. The DORA-22 experiment measured mid-dose confusion. In effect, it was the Patrick Kennedy test. “You can publish this kind of data and get people to think about it,” Renger said, though he emphasized that the animal study had its limits. “You don’t know if it translates,” he said. “Maybe this is a monkey thing.”
Colorcon, the world’s leading supplier of tablet coatings, provides its clients with a pill-color chart. Dots of various hues are arranged in a circle and divided into pizza slices of pinks, blues, and greens, which darken toward the edge. The chart can be overlaid with plastic sheets that are opaque but dotted with clear circles, allowing you to see some of the colors beneath. One sheet reveals the acceptable colors for pills in the E.U. and North America; another—showing intolerance for dark grays, dark greens, and the brightest pinks—also covers Japan.
Rick Derrickson, Merck’s director of project leadership, recently showed me this chart, and recalled meetings, early in 2011, with the company’s experts on drug stability, marketing, and supply chains. As Derrickson remembered it, the issue was: “Do I want to look like something on the market, or do I want to be totally different? Do I want to convey strength or emotion?” He explained, “Reds are culturally not acceptable in some places. It has to do with death. And some colors are viewed as candy. And you don’t see black, either.”
He showed me the finished suvorexant tablets. The forty-milligram version was a pale-green oval. Thirty milligrams was yellow and round. Twenty milligrams was white and oval, fifteen milligrams white and round. “We were looking for non-offensive,” he said. “Hopefully, we won’t run into a country that says, ‘There’s no way we’ll take green.’ ”
At this time, suvorexant had already been “on a tablet path” for two years. “Everyone in the industry tries to gravitate toward a tablet,” Derrickson explained. “It’s tried-and-true technology. When you get into some of the exotics—like putting something under your tongue—people aren’t always comfortable.” In its purest form, suvorexant is a fine crystal, with a texture somewhere between sugar and flour. Merck synthesizes it in Ireland, and ships it across the Atlantic in hundred-and-twenty-litre drums. (Derrickson expected Merck to need “several metric tons” a year.) That active chemical is mixed with a polymer that helps the drug’s absorption by the body. The mixture is heated and then extruded from a machine, like pasta, and flattened between rollers. It cools and flakes, and those flakes are ground very finely, added to filler, and pressed into tablets.
“The U.S. prefers everything in a thirty-count bottle,” Derrickson said. “The rest of the world prefers blisters”—that is, blister packs made of plastic and foil. In 2011, he asked for thousands of suvorexant tablets, in the various doses, to be packaged both ways and placed in several climate-controlled rooms, including one set at 86°F. and seventy-five-per-cent humidity. This was the start of a trial assessing the tablets’ perishability. The F.D.A. asks for at least a yearlong trial, and Merck planned to run the study for three years. In addition to the main batch, some bottled suvorexant tablets were held in other rooms, for an “in use” study, where bottles were opened and closed, manually, on different schedules, as they might be by an occasional user of the drug.
The unopened bottles protected suvorexant well. But some of the “in use” bottles did not: the tablets absorbed moisture, their coating cracked, and they started to crumble. The advantages of an orange plastic bottle over blister packs were so evident to Derrickson—“Cheaper, and more friendly,” as he put it—that he was slow to accept the results. Laura Jacobus, who was in charge of that process, recalled, “He was saying, ‘Are you sure? Check one more time.’ ”
When Merck made its formal submission to the F.D.A., in August of last year—with forty-one gigabytes of material—it proposed selling suvorexant at fifteen, twenty, thirty, and forty milligrams, in blister packs of ten, in a child-resistant plastic case.
People attending the F.D.A. committee meeting on suvorexant passed, in a lobby, a display case of pharmaceutical shame. On shelves, behind glass, were samples of a century’s worth of toxic drugs, including a pack of thalidomide—the sedative and antiemetic, launched in Europe in the nineteen-fifties, that caused thousands of birth defects before it was withdrawn, in 1961. The F.D.A. is proud that thalidomide was never approved for sale in America; in 1962, Frances Oldham Kelsey, the agency reviewer who blocked it, received a President’s Award for Distinguished Federal Civilian Service.
A red rope bisected a large hall. To the left, rows of seats reserved for the public went largely unused. To the right, there was a crush of dark suits: committee members sat at a U-shaped desk, and were flanked, in a kind of parliamentary arrangement, by Merck employees on one side and F.D.A. employees on the other.
Opening remarks were delivered by Russell Katz, the director of neurology products at the F.D.A. He affably described suvorexant as “an exciting compound,” but almost immediately spoke of an emerging F.D.A. preference for drug doses that are as low as one can “get away with.” Without naming zolpidem, he referred to the drug’s recently reduced dosage for women, saying, “We believe this is the right way to go.” He noted that suvorexant was shown to impair next-day driving at a twenty-milligram dose, and perhaps at fifteen milligrams.
Katz also observed that Merck’s Phase II investigation of ten milligrams had shown that it outperformed a placebo in sleep efficiency and Wake After Sleep Onset, although not in the time taken to fall asleep. “These data, in our view, taken together, argue for recommending doses as low as ten milligrams, or even perhaps lower than ten milligrams,” he said.
It was 8:30 A.M. The men and women on Merck’s benches looked grimly composed, like C.E.O.s being scolded at a Senate hearing. For a few days, they had known that the committee was likely to discuss a dose of suvorexant that Merck had examined, and then rejected, four years earlier, at the end of Phase II trials. If this idea was pursued by the F.D.A., then there was the odd possibility that a drug could go to market at a starting dose that Merck had studied only long enough to conclude that it didn’t work. A drug sold at an underperforming starting dose would, of course, be at a disadvantage.
Joe Herring, Merck’s main speaker, was next. Company leaders were watching a live video feed on the F.D.A. Web site, as were pharma investors, and pharma analysts ready to tweet. Herring didn’t need to make the case already made by his Phase III data: suvorexant was effective, particularly in maintaining sleep. His primary task, whose strangeness colored the rest of the day, was to talk down the effectiveness of suvorexant at the ten-milligram dose. He agreed with Katz that the objective Phase II results for ten milligrams were “substantial and encouraging.” But the subjective, patient-reported results were no stronger than those for a placebo, and insomnia “involves patient perception of sleep disturbance and clinically significant distress.” He noted that the F.D.A. had expected Merck to find subjectively significant doses to take into Phase III.
Herring then gave reassuring accounts of the side-effect data connected to the higher doses, and disputed the idea that any reported reactions to suvorexant could be thought of as “narcolepsy-like.” (Herring knew that Ronald Farkas’s unfriendly PowerPoint presentation would make the suggestion.) The direct link between narcolepsy and orexin made such suspicions natural, but Merck, assisted by an external committee, had looked for cataplexy in the data and had not found it. A few episodes of excessive daytime sleepiness, at high doses of suvorexant, and of sleep paralysis, could be explained without reference to narcolepsy. Thomas Scammell, a narcolepsy specialist at Harvard who has published widely on orexin, was sitting on Merck’s benches, as a consultant, and he later spoke to the committee in support of the company’s position. (Emmanuel Mignot, the researcher at Stanford, recently told me that suvorexant seems to produce a rather normal experience of sleep, except that patients are hurried into the REM phase, which is also the experience, in a more extreme way, of narcoleptics. Suvorexant might not be the best drug for people prone to nightmares, he said.)
After Herring finished, David Michelson—the executive who had said that suvorexant was a “huge” product for Merck—spoke. A committee member asked him if suvorexant had been compared to zolpidem in a head-to-head study. No, he acknowledged.
Farkas then gave his PowerPoint presentation. Phrases like “violent intent” appeared in large type. He recommended a ten-milligram dose, and said, “It really does come down . . . to what dose would you want used for your mother?” He seemed to enjoy his role: slightly ill-mannered, and happy to open the door to doubt. He began one sentence, in an innocent tone, “I think we don’t want to raise concerns that suvorexant causes narcolepsy by causing an autoimmune death of cells that produce orexins.” He also took a moment to undermine the importance of subjective results, saying, “Everybody knows that sleep interferes with your ability to know how much time you’ve slept.” John Carroll, an industry analyst, tweeted that the meeting was a “disaster” for Merck.
During a lunch break, Renger ate Doritos and groused: “Ten years of work, all this innovation—novel science—and we’re talking about dose and driving studies!” In the afternoon, Merck continued its effort to undermine the ten-milligram plan: Julie Stone, an expert in statistical modelling at Merck, delivered an elaborate analysis and said, flatly, “We don’t believe that ten milligrams would be an effective dose.” Herring, speaking to the committee again, said, “Ten milligrams is ineffective from a patient perspective.”
In the midafternoon, committee members began to answer a series of questions asked by their F.D.A. hosts. They started with: Was suvorexant effective at the doses suggested by Merck? In two votes, the committee members agreed that it was.
The F.D.A.’s next question began, “The applicant has submitted data supporting the conclusion that ten milligrams is an effective dose.”
This was peculiar. Robert Clancy, a professor of neurology and pediatrics at the Children’s Hospital of Philadelphia, pointed out that it wasn’t true.
Katz agreed: the F.D.A., not Merck, was arguing for the efficacy of ten milligrams. He said, “We shouldn’t couch it in terms of ‘Has the applicant done it?’ Have we done it?”
The committee’s opinion was mixed. Jason Todd, a neurologist from North Carolina, said, “Honestly, it looks like the best treatment, in terms of balancing effectiveness and side effects, is placebo.” Ronald Farkas wondered if ten milligrams would have performed better in a larger study. “A small, underpowered, negative study does not mean the drug does not work,” he said.
The committee was asked to vote on the question: Would a ten-milligram dose require additional studies before it could be approved by the F.D.A.? It voted no. Paul Rosenberg, a psychiatrist at Johns Hopkins, said, “I’m convinced that it maybe works.” Clancy said, “I feel like I’m stuck in an old episode of ‘The Twilight Zone.’ The company’s arguing their drug doesn’t work, and the F.D.A. is arguing, ‘Yes, it does.’ ” He said that he needed a sleeping pill.
By the end of the session, the committee had recommended to the F.D.A. that thirty and forty milligrams should not be approved, for safety reasons. Doses of fifteen and twenty milligrams should be approved, but the F.D.A. should consider instructing Merck to make ten milligrams the drug’s starting dose.
David Michelson sank slowly into a chair in the lobby. “I’m exhausted,” he said. “Just emotionally. You’re up and down, and you don’t know where it’s going to go. You’re forced to sit there and watch it. You’re thinking, This is going south!”—that the committee would vote suvorexant out of existence. “And then it wasn’t going south.” He added, “It’s certainly unusual that they’d be willing to consider approving a dose that had not been extensively studied.”
Jed Black followed the day’s events from afar, and was at this moment wondering if Merck “might just say, ‘Screw this,’ and proceed with another molecule.” I asked Michelson if Merck would pursue ten milligrams, if necessary, despite the company’s public disparagement of the dose. He foresaw discussions. He and his colleagues then walked to their bus, pulling wheeled luggage, in a tight, flight-attendant formation.
A few weeks later, the F.D.A. wrote to Merck. The letter encouraged the company to revise its application, making ten milligrams the drug’s starting dose. Merck could also include doses of fifteen and twenty milligrams, for people who tried the starting dose and found it unhelpful. This summer, Rick Derrickson designed a ten-milligram tablet: small, round, and green. Several hundred of these tablets now sit on shelves, in rooms set at various temperatures and humidity levels; the tablets are regularly inspected for signs of disintegration.
The F.D.A.’s decision left Merck facing an unusual challenge. In the Phase II trial, this dose of suvorexant had helped to turn off the orexin system in the brains of insomniacs, and it had extended sleep, but its impact didn’t register with users. It worked, but who would notice? Still, suvorexant had a good story—the brain was being targeted in a genuinely innovative way—and pharmaceutical companies are very skilled at selling stories.
Merck has told investors that it intends to seek approval for the new doses next year. I recently asked John Renger how everyday insomniacs would respond to ten milligrams of suvorexant. He responded, “This is a great question.” After the approval process is finished, the marketing division of Merck—a company whose worldwide sales last year totalled forty-seven billion dollars—will conduct a different kind of public trial. The study will address this question: How successfully can a pharmaceutical giant—through advertising and sales visits to doctors’ offices—sell a drug at a dose that has been repeatedly described as ineffective by the scientists who developed it? ♦
