Like benzodiazepines, neurosteroids can have antianxiety, antiseizure, and sedative effects. These compounds are formed inside the brain, rather than in the gonads or adrenal glands. Some neurosteroids can make GABA receptors more active, and others can push in the opposite direction. One of the most potent and best-studied members of the neurosteroid family is allopregnanolone, whose levels rise throughout pregnancy and fall after giving birth. Allopregnanolone levels in the blood also rise rapidly following acute stress, and deficiencies in allopregnanolone have been observed in several diseases, such as depression and Alzheimer’s.35 Neurosteroids can be found in both men and women, and their fluctuations may account for some of the effects of menstrual cycles on the nervous system—including a rare form of hypersomnia known as “menstruation-related hypersomnia.”36

  However, neurosteroids bind at separate sites on GABA-A receptors from where benzodiazepines bind, according to biophysical studies (figure 9.2).37 Whether flumazenil interferes with a molecule like allopregnanolone depends on the experimental setup.38 We can’t cleanly explain flumazenil’s effects on the nervous system—and especially, its ability to wake some people up—through the displacement of neurosteroids.

  FIGURE 9.2. Diagram of GABA-A receptor assembly and ligand binding sites.

  Source: Illustration by Juan Gaertner, based on data from Shaotong Zhu et al., “Structure of a Human Synaptic GABA-A Receptor,” Nature 559 (2018): 67–72. SciencePhoto.com.

  ELIMINATE THE WORMS

  At Roche, management didn’t jump at the prospect of a benzodiazepine antidote when flumazenil was first identified, according to Haefely and Hunkeler’s account. Interest came from another arm of the company aiming to treat schistosomiasis, a waterborne parasitic infection that affects millions in developing countries. Unexpectedly, a benzodiazepine called meclonazepam could paralyze the worms that cause schistosomiasis, but the amounts needed to eliminate the worms made patients groggy.39 Roche didn’t pursue meclonazepam further, but it kept the ball rolling for flumazenil when an antidote’s commercial potential was uncertain.

  While flumazenil was undergoing early clinical tests, Roche had introduced midazolam as a successor to diazepam but with a more limited set of uses. Midazolam was meant for sedation during medical procedures, not as a mass-market antianxiety pill. Around this time, Imperial Chemical Industries was developing the anesthetic propofol, which wore off more quickly than other anesthetics. With competition in mind, executives at Roche looked more favorably at flumazenil, which could make midazolam’s effects dissipate in a few minutes. Accordingly, when flumazenil was introduced in European countries, it was marketed together with midazolam.40

  Meanwhile, doctors’ unfamiliarity with midazolam and its greater potency had caused problems, which first surfaced in Europe in the mid-1980s. At a congressional hearing, midazolam’s allegedly inadequate labeling was blamed for overdoses and respiratory and cardiac arrest, leading to dozens of deaths.41 Most occurred in connection with sedation during endoscopies, usually with elderly patients. These incidents made the need for a benzodiazepine antidote clearer. In 1988, flumazenil won the Prix Galien, an Academy Award for the pharmaceutical industry, in the category of neuropsychiatry. In 1991, flumazenil was approved as a benzodiazepine antidote by the U.S. Food and Drug Administration.

  ENHANCING VIGILANCE AND SOBERING UP

  If flumazenil’s properties were intriguing, then GABA receptor “inverse agonists” represent an even stronger version of what might have been. In the 1980s, researchers at Roche’s competitors, such as Schering and Ciba-Geigy, were studying flumazenil-related compounds, with some reports claiming that they displayed memory- or learning-enhancing effects in animals.42 These so-called inverse agonists pushed GABA-A receptor sensitivity in the other direction, compared to benzodiazepines.

  Since benzodiazepines seemed to embody the opposite of anxiety, neuroscientists used their opposites, inverse agonists, to create chemically induced models of anxiety in animals. Inverse agonists acquired an aura of danger because they could make the nervous system more sensitive to seizures. One such compound brought on “intense motor unrest” and “an impending fear of death or annihilation” when it was injected into human volunteers.43

  A less potent but longer-lasting inverse agonist called 3-HMC attracted the attention of Wallace Mendelson, who worked with Steven Paul as director of the sleep lab at NIMH. They reported in Science that 3-HMC kept rats in a state of “quiet wakefulness,” and it did not induce agitation, like amphetamines or caffeine did.44 It appears that 3-HMC was never tested in human studies, although Mendelson saw the possibilities. He told the Washington Post: “Here is a drug that has several effects that are opposite of the benzodiazepines. So we wondered, if the benzodiazepines put you to sleep, will this drug wake you up? … Maybe these chemicals would be useful drugs for people with disorders in which they get too much sleep.”45

  Around this time, Roche was investigating a relative of flumazenil called sarmazenil. The company got as far as conducting small studies with sleep-deprived volunteers, finding that the drug could reduce sleepiness and improve reaction time and performance on other cognitive tests.46 In a 1989 report to his colleagues, Roche research manager Peter Schoch proposed that sarmazenil or a related compound could be developed as a potential “vigilance enhancer” for indications such as Alzheimer’s disease. He noted that two other companies were pursuing similar compounds.47 “It could be of clinical use in narcoleptic, depressed and/or geriatric patients with a reduced level of arousal.… Evaluation of the pharmacological properties and therapeutic potentials of compounds like Ro 15-3505 [sarmazenil] has to proceed rapidly if Roche is not to lose this field to its competitors,” Schoch wrote.

  Sarmazenil might have become a competitor for wake-promoting medications such as modafinil, if it had progressed. But company documents reveal that Roche chief Jürgen Drews doubted that vigilance enhancement “fulfills a real medical need—Such a drug would most probably face strong regulatory resistance anyway.”48 Some volunteers reported unpleasant side effects such as dizziness, and there were hints from EEG studies that seizure risk might be elevated. Sarmazenil was registered for reversing benzodiazepine sedation, but for veterinary use only.

  Another tweak of flumazenil’s structure led to an unexpected effect. Roche scientists attached a light-sensitive chemical group to flumazenil, with the aim of creating a probe that could label benzodiazepine receptors in cells. They discovered that the tweaked version, called Ro 15-4513, could reverse the behavioral effects of alcohol in animals—something flumazenil can’t do.49 Ro 15-4513’s properties excited some on Haefely’s team. However, he thought that wide availability of an antidrunkenness pill might encourage risk taking and promote alcohol consumption, and the company decided not to develop it further.50 Other researchers later used Ro 15-4513 as a probe to dissect how alcohol exerted its effects on GABA receptors and the nervous system.

  After the early 1990s, Roche management paid little attention to flumazenil, sarmazenil, or Ro 15-4513, according to research steering committee meeting minutes in the company’s archive. Several strands of research on flumazenil were clipped. For example, epilepsy studies were phased out because “the results accumulated suggest that flumazenil exhibits too weak an antiepileptic activity.” Möhler returned to academic research at the University of Zürich, and Roche’s central nervous system unit underwent reorganization.

  During the 1990s, gene cloning techniques advanced, revealing a forest of genes encoding different GABA-A receptor subunits. Some varieties were present only in particular regions of the brain. The dominant form was sensitive to benzodiazepines, but others were not. Their receptors’ distribution in the brain was more complicated than anticipated, slowing efforts to devise drugs that were specific to one receptor or only a few. “It was a complete shock,” John Kemp, Haefely’s successor as head of central nervous system research, told Forbes.51

  Haefely’s original goal of finding alternatives to Valium was achieved by competitors, through the development of drugs such as Xanax and Ambien. His last paper mentioning flumazenil, published two months after his death in 1993, stated that the drug “exhibits virtually zero intrinsic efficacy.”

  THE PATIENT OPENED HER EYES

  A separate line of research on flumazenil in people with hepatic encephalopathy unfolded in the 1980s and 1990s. Roche sponsored some of the studies but didn’t push for them especially hard. This use for flumazenil emerged unexpectedly, through risk taking in the hospital. In comparison with Anna’s experience, there was less precedent, and the patients were more vulnerable.

  Hepatic encephalopathy can develop as a result of alcoholic cirrhosis, viral hepatitis, acute poisoning, or surgery that bypasses the liver. Alcohol consumption, gastrointestinal bleeding, or having to digest a large amount of protein can trigger an episode. Sometimes the symptoms—sleepiness, agitation, erratic behavior, memory problems, tremors—creep up slowly or fluctuate; people with liver damage can be unaware they have it. At its most extreme, hepatic encephalopathy results in an unwakeable stupor or coma.

  In the early 1980s, it was known that people with liver disease were more sensitive to benzodiazepines, and hepatologists were exploring various explanations.52 Perhaps more GABA, produced by gut bacteria, was leaking into the nervous system, or maybe GABA receptors in the brain were becoming more sensitive? Research on rats with acute liver failure was suggestive enough for some clinicians to try flumazenil in humans with liver diseases.53

  Giuseppe Scollo-Lavizzari, a neurologist in Basel, was the first to test flumazenil in a hospital’s emergency department on people with benzodiazepine overdoses.54 Encouraged by his experiences, Scollo-Lavizzari tried again, this time with a twenty-five-year-old woman who was infected with hepatitis B as a result of heroin addiction.55 Her liver disease had left her in a coma, motionless and unresponsive to painful stimuli. She had not been given benzodiazepines, he noted in a 1985 letter to Lancet.

  After intravenous flumazenil, “the patient opened her eyes, reacted to verbal commands, and moved spontaneously and in response to painful stimuli, but she did not speak.” Flumazenil’s effects were reproducible, but in contrast, the opiate antagonist naloxone did not affect the twenty-five-year-old, who died two weeks later. Before his experiment, Scollo-Lavizzari had asked permission from the head of the hospital’s emergency department, but regulators might not look kindly on the venture today, he told me. Scollo-Lavizzari was one of David Rye’s forebears as a flumazenil enthusiast. He also tested flumazenil as an anticonvulsant and in people who were intoxicated by alcohol. In addition, he obtained a patent for flumazenil in the treatment of stroke.56

  In the same issue of Lancet, doctors in Zürich reported similar results with four cirrhotic patients. One of the Zürich authors was the self-experimenter and Roche employee Walter Ziegler. The Lancet letters led to other small-scale trials of flumazenil in people with hepatic encephalopathy. Flumazenil does not resolve underlying liver damage, so why go through the trouble and risk? It provides symptomatic relief, makes managing the patient’s care easier, and gives time for other interventions.

  One spectacular case report came from Newcastle, England, where a forty-one-year-old woman had been drinking a bottle of vodka daily for several weeks. Her pupils reacted to light, but she would only move her limbs in response to deep pain. In response to intravenous flumazenil, “the patient wiped her eyes and mouth, exclaimed that she was ‘starving,’ and was verbally abusive to the nurse.” A minute later, she could answer questions and focus visually. The woman stayed awake for about two hours. Flumazenil woke her up a second time the next day. Her condition gradually improved, and she was able to leave the hospital twenty-five days later.57

  In Vienna, using flumazenil, physicians were able to stave off coma for almost two years in a woman who had her liver bypassed after gallstone surgery.58 Two decades before Anna Sumner’s experience, this was a rare example of a patient chronically treated with flumazenil. The woman had been in and out of the hospital several times, where she had been treated with lactulose enemas and antibiotics. Her long treatment and its twists and turns make it unlikely that she was somehow deceiving her doctors.

  The Viennese woman was able to avoid coma in two stretches in 1985 and 1986, interrupted by a serious intestinal infection. While taking flumazenil, she did not need to avoid large amounts of dietary protein, which would normally lead to encephalopathy. One drawback was that she consistently became anxious for about half an hour after taking the drug. She was taking 50 mg per day—more than Anna. The Viennese doctors concluded: “In contrast to when she was not taking the drug, while receiving flumazenil she was able to lead a normal life.”

  STORM RUNOFF

  Despite some intriguing parallels, we can only learn so much about idiopathic hypersomnia from looking at people with hepatic encephalopathy, who are seriously ill and have a high risk of mortality. Given how the liver cleanses the body, studying liver dysfunction is like analyzing agricultural runoff or sewer water after a storm. High levels of ammonia are a problem, but many other metabolites build up as well. Intestinal barriers break down, allowing bacteria and their waste products access to the nervous system via a “leaky gut.” How should researchers pick out which toxin is the most important?

  The reports of patients with liver failure waking in response to flumazenil presented the same mystery as with Anna. They were not known to have any benzodiazepines in their systems, so why did flumazenil wake them up? One proposed explanation was that endogenous benzodiazepines were clogging up the brain’s GABA receptors, and flumazenil displaced them. Researchers at NIMH led by Phil Skolnick, a colleague of Paul’s, obtained brain tissue samples from people who had died of acute liver failure after acetaminophen overdose. None of the deceased had received benzodiazepines while in the hospital, the researchers reported in the New England Journal of Medicine. Analyzing liver and brain samples, the authors used mass spectrometry to show that one of the peaks represented diazepam, although not all of the peaks were identified.59 Skolnick told the press: “This is the first evidence that benzodiazepines found naturally in the brain play a role in illness. This study provides a rational basis for a cause and a cure for hepatic comas.”60

  Skolnick’s 1991 paper represents a peak of enthusiasm for flumazenil’s use in hepatic encephalopathy. His colleagues later found hints from animal models of liver failure that intestinal bacteria could be producing the benzodiazepine-like compounds, but they were not able to identify the specific chemicals responsible.61

  As research progressed, a less intriguing explanation loomed larger: hepatic encephalopathy patients’ previous medication intake. When liver function is compromised, benzodiazepines last longer in the body because the drugs are normally broken down by the liver. To know that something generated within the body was causing patients’ stupors, doctors needed to make sure that the patients had not previously been given synthetic benzodiazepines. In earlier studies, adequate screening tests were not performed.

  The Montreal-based hepatologist Roger Butterworth supervised the first randomized, double-blind trial of flumazenil in hepatic encephalopathy. His study, published in 1994, found that a minority (40 percent) of patients in hepatic coma displayed clinical improvement in response to flumazenil.62 A 1998 follow-up concluded: “These findings do not support a role for ‘endogenous’ benzodiazepines in the pathogenesis of HE in chronic liver disease, but suggest that pharmaceutic benzodiazepines administered to cirrhotic patients as sedatives or as part of endoscopic work-up could have contributed to the neurological impairment in some patients.”63

  Early studies, conducted without placebo controls, claimed that a majority of patients responded favorably to flumazenil. The numbers fell as more rigorous studies were performed, in which thorny issues emerged. Was it better to enroll comatose patients or those who were still wakeable? Patients with acute liver failure or chronic liver disease? If someone became encephalopathic because of gastrointestinal bleeding, maybe they would have recovered anyway, once the bleeding was controlled.

  The largest study of this type recruited more than five hundred participants with cirrhosis in Italy and found neurological improvements in less than 20 percent of those treated with flumazenil. Others obtained similar results, and the hepatology field began to drift away from flumazenil.64 Butterworth later reexamined the issue and proposed that neurosteroids were likely to be playing a major role in amplifying GABA signals in hepatic encephalopathy, rather than endogenous benzodiazepines.65

  A CONCLUDING NOTE ABOUT ENDOZEPINES

  Endozepines’ relevance for sleep disorders remains tenuous. Diazepam binding inhibitor, the protein originally identified by Costa and Guidotti, does loosely fit the information we have about the hypersomnia somnogen, the GABA-enhancing substance in patients’ spinal fluid. Peptide fragments of DBI are around the right size, and it binds to GABA-A receptors in a flumazenil-sensitive mode.

  When Rye raised money in 2014 to look for the unknown somnogen (see next chapter), testing DBI fragments was the first task on his list.66 Rye and Jenkins were encouraged when epilepsy researchers at Stanford found that DBI was capable of acting analogously to diazepam within the reticular thalamic nucleus: a region of the brain with a central role in sleep oscillations.67 The Stanford findings confirmed the role that Costa had envisioned for DBI years before. But when Jenkins obtained some DBI peptide, it was inactive in his patch clamp experiments, he said. Based on that negative result, DBI does not account for the GABA-enhancing activity of CSF samples from hypersomnia patients.68