The following material is excerpted and revised from Brain-Disabling Treatments in Psychiatry, by Peter R. Breggin, M.D. (Springer Publishing Co. 1997, 2008). Presented here with the permission of Springer Publishing Company, 536 Broadway, New York, NY 10012-3955.
Review of Behavioral Effects of Benzodiazepines With an Appendix on Drawing Scientific Conclusions from the FDA’s Spontaneous Reporting System (MedWatch)
By Peter R. Breggin, M.D.
The benzodiazepines (BZs) have for several decades been recognized in the literature and clinical practice for their capacity to cause mental and behavioral abnormalities. Xanax (alprazolam), and to an even greater extent, Halcion (triazolam), have a significantly different profile from other benzodiazepines due to their greater capacity to bind to receptors and their shorter half-life. Halcion’s very short half-life led to the hope that it would make a particularly good sleeping medication but it has proven especially dangerous.
The brain-disabling or toxic effects of the BZs in general can be divided into several somewhat overlapping categories:
(1) The primary clinical effect of inducing sedation (tranquility) or hypnosis (sleep), which is indistinguishable from a toxic effect except in degree;
(2) Cognitive dysfunction, ranging from short-term memory impairment and confusion to delirium;
(3) Disinhibition and other behavioral aberrations — including extreme agitation, psychosis, paranoia, and depression, sometimes with violence toward self or others;
(4) Withdrawal, in which the individual experiences a continuum of symptoms from anxiety and insomnia after routine use to psychosis and seizures after the abrupt termination of long-term, larger doses;
(5) Rebound, an aspect of withdrawal, in which the individual develops anxiety, insomnia, or other serious emotional reactions that are more intense than before drug treatment began. Withdrawal can take place between doses during the routine administration of BZs, especially the short-acting ones;
(6) Habituation and addiction, along a continuum from feeling dependent on the drug to compulsively organizing one’s behavior in a self-destructive manner around obtaining large amounts of the agent.
The Mechanism of Brain-Disability
Neurophysiological studies show that the BZs potentiate the neuronal inhibition that is mediated by gamma-aminobutyric acid (GABA). In doses used clinically, this results in a generalized suppression of both spontaneous and evoked electrical activity of the large neurons throughout all regions of the brain and spinal cord (Ballenger, 1995). The binding of BZs to the GABA receptors is most intense in the cerebral cortex. Some BZs, such as Xanax and Halcion, bind especially tightly, increasing their tendency to produce more intense sedation and hypnosis, and also more severe cognitive deficits, behavioral abnormalities, rebound, and withdrawal. Some advocates of the BZs have argued for a specific anti-anxiety effect separate from the general sedative effect, but there’s no substantial evidence for this. Rall (1990) concludes “The question whether the so-called antianxiety effects of the benzodiazepines are the same or different from the sedative and hypnotic effects has not been resolved.”
People who use BZs to calm their anxiety will frequently use alcohol and other sedatives interchangeably for the same purpose, either in combination or at different times. As they switch from drug to drug, they tend to find little or no difference in the anti-anxiety effect. This confirms the brain-disabling principle that BZs have no specificity for anxiety in comparison to other sedative/hypnotic agents.
The Mechanism for Producing Behavioral Abnormalities
There are at least two probable causes for the abnormal behavior produced by BZs. One mechanism is direct intoxication, resulting in impaired executive and cognitive function, including reduced judgment and impulse control. Fogel and Stone (1992, p. 341) observe “Benzodiazepines, given to reduce arousal or possibly to treat a hypomanic state, may aggravate impulsive behavior by impairing the inhibition mechanism of the frontal lobes. Barbiturates may have similar effects.”
Especially in regard to the BZs, a second mechanism, rebound, has been demonstrated, and is a probable cause of many more severe reactions. Rebound or discontinuation symptoms occur when the BZs are withdrawn or when they begin to lose their effectiveness (American Psychiatric Association, 1990). As the BZ disappears from the GABA receptor sites, the receptors may have become down-regulated (less sensitive). Or there may be a reduction in GABA itself in response to the drugs, once again leaving the GABA system relatively inactive. Without the inhibiting effects of the GABA system, the “disinhibited” brain over-reacts. The American Psychiatric Association (1990) task force report, Benzodiazepine Dependence, Toxicity, and Abuse, theorized that discontinuation symptoms arise from the abrupt withdrawal of BZs from their receptor sites. Before GABA can retake these positions, there is an acute reduction of GABA, and a loss of GABA inhibitory tone.
BZ disinhibition differs in some ways from alcohol disinhibition. It can occur without a noticeable sedative intoxication, such as slurred speech, lack of coordination, or impaired consciousness. Furthermore, the BZs are prescribed by a physician, often without providing the patient a warning about possible disinhibition. Unlike the experienced alcohol user, the trusting BZ user has little reason to anticipate “losing control.” Expecting to be helped, and not harmed, by the drug, the patient is less able to understand or manage potentially overwhelming feelings of anger or violence, or other untoward emotional responses. Also unlike alcohol, some of the worst BZ behavioral reactions occur during withdrawal, or in between doses, adding to the patient’s confusion concerning what is happening. At the time, the patient may have little idea what is driving the unfamiliar behavior, and in retrospect it may seem like a fragmented, poorly recalled nightmare.
Adverse Reactions to the BZs as a Group
Standard textbooks and reviews spanning more than two decades, as well as a variety of clinical studies, confirm widespread recognition of BZ-induced behavioral abnormalities (DiMascio and Shader, 1970; Kochansky, Salzman, Shader, Harmatz, and Ogeltree, 1975; Shader and DiMascio, 1977; Rosenbaum, Woods, Groves, et al., 1984; Arana and S. Hyman, 1991; Maxmen, 1991; Maxmen and Ward, 1994; Ashton, 1995).
Salzman, Kochansky, Shader, Porrino, Harmatz, and Swett (1974), in a placebo controlled study, showed that volunteers taking chlordiazepoxide became more hostile when confronted with a situation of interpersonal frustration. In 1988, Dietch and Jennings reviewed the literature concerning reports about disinhibition whose “Manifestations range from irritability to increased verbal hostility and frank assault.” They found a variety of studies that demonstrate an increase in feelings of hostility or in verbal hostility. They did not come to a definitive conclusion concerning the existence of the phenomenon, but estimated it was probably rare. Salzman (1992) reviewed the literature. He too pointed out the controversial nature of BZ-induced violence, but went on to assert, “Recent observations, however, have confirmed that hostility can be seen with all benzodiazepines, including alprazolam and clonazepam” (citations omitted). Writing in Goodman and Gilman’s The Pharmacological Basis of Therapeutics, Rall (1990) summarized:
“Adverse psychological effects: Benzodiazepines may cause paradoxical effects. Nitrazepam frequently and flurazepam occasionally increase the incidence of nightmares, especially during the first week of use. Flurazepam occasionally causes garrulousness, anxiety, irritability, tachycardia, and sweating. Euphoria, restlessness, hallucinations, and hypomania behavior have been reported to occur during the use of various benzodiazepines. Antianxiety benzodiazepines have been reported to release bizarre uninhibited behavior in some users with low levels of anxiety; hostility and rage may occur in others. Paranoia, depression, and suicidal ideation occasionally also accompany the use of these agents.”
Rall believed that “the incidence of such paradoxical reactions is extremely small.” Whether or not that is true, they are extremely hazardous. They are more common in regard to the short-acting BZs. The APA task force report on BZs (1990, p. 18) presents a table of discontinuation symptoms. The complete list of frequent discontinuation symptoms includes “anxiety, insomnia, restlessness, agitation, irritability, muscle tension.” Among many symptoms that are common but less frequent, it lists “depression” and “nightmares,” as well as “lethargy.” Clinical experience indicates that the combination of anxiety, insomnia, restlessness, agitation, irritability, nightmares, and depression can produce a spectrum of behavioral abnormalities, including suicide and violence. Adding to the dangers, the task force’s complete list of uncommon symptoms includes “psychosis, seizures, persistent tinnitus, confusion, paranoid delusions, hallucinations.”
The Production of Mania and Rage
As the above observations confirm, reactions to Xanax and other benzodiazepines can reach psychotic proportions. As noted in the 1996 edition of Drug Facts and Comparisons, the BZs in general can cause serious psychiatric problems, including “psychosis.” They can disrupt central nervous system function, producing, among other things, “disorientation… confusion… delirium…. euphoria… agitation.” A special Precautions section notes “Paradoxical reactions,” including “excitement, stimulation and acute rage” and “hyperexcited states, anxiety, hallucinations…”
Mania, a psychosis, is a special danger in regard to Xanax. The Drug Facts and Comparisons (p. 1440) makes a specific reference to Xanax in this regard, stating that “Anger, hostility and episodes of mania and hypomania have been noted with alprazolam.” The fact that mania is a particular risk with Xanax is generally recognized. As another example, Maxmen and Ward’s Psychotropic Drug Fast Facts (1995, p. 287) states that “manic reactions” are “Most often reported with alprazolam.” It also states that “rage reactions” and “violent episodes” have especially been observed with Xanax and Valium. Yet another example is The Handbook of Psychiatric Drug Therapy (Third edition, 1995, p. 177) by Hyman, Arana, and Rosenbaum. It singles out Xanax to observe “Increased impulsiveness, euphoria, and frank mania have been reported with alprazolam.” Drug Facts and Comparisons, Psychotropic Drug Fast Facts and the Handbook of Psychiatric Drug therapy are intended for ready reference for physicians for the purpose of alerting them to adverse drug reactions they need to be aware of. That all three indicate that mania and uncontrollable rage are special problems with Xanax confirms my own clinical observations that this medication is more apt than many others to produce these hyper-excited, aggressive psychotic states.
The FDA’s “Safety Review and Evaluation of Clinical Data” by J. Knudsen (1989) for Xanax for panic disorder once again highlighted its tendency to produce mania. It included a review of Xanax-induced mania from the literature. It described several cases of mania produced in the panic disorder clinical studies and offered two tables described 14 cases of hypomania and one of mania. Several attacks developed during the first week of treatment at relatively small doses.
The Production of Depression and Suicide
As already noted, there are reports in the clinical literature indicating that the BZs can cause depression. Some reviews mention the phenomenon but express skepticism, while nonetheless declaring it should be taken seriously. Arana and Hyman, for example, stated:
“Depression: All benzodiazepines have been associated with the emergence or worsening of depression; whether they were causative or only failed to prevent the depression is unknown. When depression occurs during the course of benzodiazepine treatment, it is prudent to discontinue the benzodiazepine.”
Ashton (1995) observed that benzodiazepines can blunt the emotions in general, producing “emotional anesthesia.” He reported, “Former long-term benzodiazepine users often bitterly regret their lack of emotional response to family events during the period that they were taking the drugs.” Ashton also observed that BZs can precipitate suicide in already depressed patients. The APA task fore report on BZs (1990), in a discussion of toxicity, also observed: “Benzodiazepines have also been reported to cause or to exacerbate symptoms of depression. This, too, is not a frequent side effect, although the depressive symptoms may be potentially serious.” p. 41 (citation omitted) Great Britain’s Committee on Safety of Medicines (1988) recommended that “benzodiazepines should not be used alone to treat depression or anxiety associated with depression. Suicide may be precipitated in such patients.” Due to the action of the BZs on gamma-aminobutyric acid (GABA) receptors, as well as to research in depression, there is growing interest in the role of a GABAergic systems in mood disorders (Kalpana, Musselman, Schatzberg & Nemeroff, 1995; also see Bartholini, Lloyd & Morselli, 1986). GABA systems have also been associated with the production of anxiety and anger.
Cognitive, Emotional and Behavioral Abnormalities Caused by Halcion and Xanax
Several studies have demonstrated rebound phenomena the same night or the day following the ingestion of the short-acting BZ, triazolam (Halcion). In a controlled study, Moon, Ankier and Hayes (1985) found “The results support previous reports that early insomnia and an increase in daytime anxiety are problems associated with short acting benzodiazepines, such as triazolam.”
De Tullio, Kirking, Zacardelli and Kwee (1989) reviewed the charts of 72 adult male patients taking triazolam for sleep through an ambulatory Veterans Administration clinic. Thirty-nine of the patients were available for telephone interviews. Most of the patients were elderly (aged 60 or older) and almost all received 0.25 mg. Of the thirty-nine patients interviewed, only 4 reported no adverse effects and 23 experienced more than one. The most common were dizziness, rebound insomnia, and nightmares. “Rebound insomnia was defined as waking during the night or waking too early in the morning, and having trouble falling back to sleep.” As a result of the study, the VA facility modified its policies on triazolam administration: “for outpatients on chronic triazolam therapy, a switch to a longer-acting benzodiazepine was instituted with tapering if therapy was not to be continued.” We have already note that the APA task force (1990) on BZs described a variety of symptoms, including depression, anxiety, hostility, and paranoia, and attribute them in part at least to discontinuation or withdrawal. In regard to short-acting BZs, the task force made the following observations:
“Abrupt discontinuation of short half-life benzodiazepines leads to rapid drug removal from the blood and brain, rapid uncovering of the receptor site, and relatively rapid onset of post-drug discontinuation syndromes. … Because of the severity of symptoms related to its half-life, short half-life benzodiazepines given for anxiety are frequently implicated in intense discontinuation syndromes. … With very short half-life drugs such as triazolam, rebound symptomatology has actually been described during the period of ingestion, especially when it is given nightly. p. 39-40 (citations omitted)”
Public and professional awareness of the special dangers of Halcion began in 1978. At that time, C. van der Kroef — a psychiatrist in the Hague, Netherlands — noticed abnormal reactions to Halcion in four of eleven patients he treated with the drug. Here is van der Kroef’s description in one of his patients (quoted in Dukes, 1980):
“The insomnia improved at once, but psychically she rapidly went downhill. Progressively she became paranoid. Several times she asked me what the hypnotic contained–LSD perhaps?–for she felt that she was bordering on psychosis. She felt shut off from the world; it was as if she no longer belonged to society. Her friends asked her what was happening to her, so strangely was she behaving. …. After two months I too began to suspect, particularly in light of experience with an earlier patient, that all this might be a consequence of her taking triazolam. The drug was withdrawn and replaced with nitrazepam. Within a day she felt herself again. The people around her noticed the difference and recognized her old self again. The paranoid traits, the hypermotility urge and the hyperaesthesia disappeared in the course of two days…”
M. N. G. Dukes (1980), a physician with considerable regulatory experience, commented on van der Kroef’s findings. He observed that all of the BZs, including those used to induce sleep (hypnotics), have been known to produce reactions that are “frankly psychotic.” While not common, according to Dukes, “virtually every known drug in this class” has produced “hallucinations, delusions, paranoia, amnesia, delirium, hypomania–almost every conceivable symptoms of psychotic madness…”
According to Dukes, all the BZs used for the control of anxiety were also implicated in causing violence:
“If one — to begin at an arbitrary point — looks to the literature for evidence that the benzodiazepines can unleash aggression then one will find it. More than a dozen papers in the literature speak of irritability, defiance, hostility, aggression, rage or a progressive development of hates and dislikes in certain patients treated with benzodiazepine tranquilizers; all those products which are widespread have been incriminated at one time or another. The phenomenon has been demonstrated in animal studies and it has even been proved possible to show in human volunteers that these drugs can release pent-up hostility, particularly in highly anxious or action-oriented individuals.”
Until the advent of Halcion, according to Dukes, the BZs commonly used to induce sleep were not known to cause violence. We shall find his observations confirmed later on by in-house studies at the FDA indicating that Halcion — but not the older hypnotics, Dalmane or Restoril — caused a vastly increased rate of violent activities. It is, of course, extremely difficult to demonstrate drug-induced behavioral abnormalities in highly selective, short, controlled clinical trials (see chapter 11 for detailed analysis of why this is so). Nonetheless, several studies have confirmed some of the hazards associated with Halcion. Gardner and Cowdry (1985) found an increase in “dyscontrol” in borderline patients taking alprazolam in a double-blind, placebo-controlled crossover study. The dyscontrol included the following: “Overdose, severe”; “Overdose, moderate”; “Deep neck cuts”; “Transverse wrist cuts”; “Tried to break own arm”; “Threw chair at child”; and “Arm and heading banging; jumped in front of car.”
Gardner and Cowdry pointed out that there are some reports of borderline patients also improving on alprazolam. They concluded, “Caution should, however, be exercised, particularly in treating individuals with a substantial history of dyscontrol.”
Bayer, Bayer, Pathy and Stoker (1986) conducted a 9 week double-blind controlled study of triazolam and another hypnotic, chlormethiazole, in the elderly with sleep disturbances. They found daytime withdrawal effects from triazolam but not chlormethiazole. At week three, significantly more triazolam patients were rated as more restless during the day “and they also appeared more hostile, less relaxed, more irritable and more anxious.” Patients on triazolam also had more adverse events related to the central nervous system, requiring four of twenty-two patients to withdraw from the study, three of whom recovered after terminating the medication. One patient felt the tablets were “making him nervous.” The others individually developed paranoid delusions, “increasing confusion and irritability,” and irrational, irritable, and uncooperative behavior.
Adam and Oswald (1989), in a double blind, placebo controlled study of triazolam and lormetazepam with forty subjects in each of the three groups, found that “triazolam takers became more anxious on self-ratings, were judged more often to have had a bad response by an observer, more often wrote down complaints of distress, and suffered weight loss. After about 10 days of regular triazolam they tended to develop panics and depression, felt unreal, and sometimes paranoid.” According to the authors, “Subjects’ written comments suggested that from about 10 days after starting triazolam, they became liable to panic attacks, feelings of despair and derealization. There were descriptions of panic episodes in public places in seven subjects during triazolam intake, but none during placebo or lormetazepam. … Several reported their family relationships were changed. … A number of triazolam subjects became paranoid. … Two men developed paranoid psychoses.” During the withdrawal period, the anxiety of the triazolam patients “fell quickly to normal levels.” Soldatos, Sakkas, Bergiannaki, and Stefanis (1986) reported on serious adverse drug reactions in all five psychiatric inpatients during a clinical trial of 0.5 mg. triazolam and placebo. The patients and nurses were blind in the study, but not the physician with medical responsibility for the patients. The study consisted of one week of placebo baseline, two weeks of triazolam administration, and one week of withdrawal on placebo. All five patients developed severe reactions to triazolam. Case 1 developed “anxiety and hallucinations on the last two days of triazolam administration and the first withdrawal day.” Case 2 had a sudden increase in anxiety and became “irritable, uncooperative, and depressed.” She became withdrawn and cried, and showed “considerable impairment of memory and orientation.” On withdrawal of triazolam, “she became more incoherent, expressing paranoid ideas of persecution that persisted about a week.” She required Haldol to control her “delusions.” Case 3 developed severe insomnia during withdrawal and “reported considerable anxiety and irritability along with an uncontrollable fear of death, which persisted to the next day when she additionally manifested a marked degree of memory impairment.” Case 4, by the end of the second week of triazolam administration, “became more depressed and manifested increasingly irritability and hostility…” Case 5, on the second week of triazolam administration, “experienced increasing daytime anxiety and he became, for the first time since admission, irritable, hostile, and somewhat guarded and paranoid towards the unit staff.” The authors suggest that some of the symptoms may be related to disinhibition. They warn that these serious side effects “may not be rare when triazolam is used in patients” with “major psychiatric conditions.”
Rosenbaum, Woods, Groves et al (1994) found that 8 of 80 (10%) of patients treated with alprazolam in an outpatient clinical setting developed extreme anger or hostile behavior.
In the FDA’s “Safety and Review Evaluation of Clinical Data” by J. Knudsen (1989) of Xanax for panic disorder, the issue of disinhibition from Xanax was addressed. It stated “The emergence of the rather paradoxical reaction of disinhibition often characterized by agitating, aggressive behavior, rage, hostility has been reported in patients treated with alprazolam. Knudsen described several reports of adverse drug reactions described as “disinhibition, rage, hostility” occurring early in treatment at relatively low doses. One physician reported three different patients as suffering from these attacks.
Evidence from the FDA’s Spontaneous Reporting System (SRS)
In 1987, Bixler, Kales, Brubaker and Kales, from the Sleep Research and Treatment Center and Department of Psychiatry at the Pennsylvania State University College of Medicine, reviewed adverse reactions to BZs recorded in the FDA’s spontaneous reporting system. They compared triazolam with two other BZs commonly used to induce sleep, temazepam (Restoril) and flurazepam (Dalmane). They controlled the reports for the number and size of prescriptions for each of the three drugs. They found:
“In general, triazolam had much higher overall rates than did the other two drugs. Hyperexcitability and withdrawal effects were greatest for triazolam and least for flurazepam. Amnesia was reported almost exclusively with triazolam. Rates for other cognitive as well as affective and other behavior effects were also much greater for triazolam and about equal for the other two drugs.”
The “affective and other behavioral disturbances” category of adverse drug reactions included “Depression, Psychotic Depression, Emotional lability, Euphoria, Hostility, Personality disorder, and Decreased libido.”
Epidemiological studies at the FDA have consistently shown that alprazolam and especially triazolam produce more frequent and more serious adverse central nervous system effects, including drastic and life-threatening behavioral changes, than any other BZs. I have reviewed the in-house memos with detailed analyses generated by the Division of Epidemiology and Surveillance, which is responsible for the SRS. This division has consistently shown more concern about triazolam than has Paul Leber’s Division of Neuropharmacological Drug Products, which originally approved the drug (see below; see chapter 11 for more about Leber and the FDA). The data from the epidemiology studies will now be described in detail for the first time in the literature. Robert “Bob” Wise (1989), in a working paper for the FDA’s Division of Epidemiology and Surveillance, made an “executive summary” of reports of hostility associated with exposure to triazolam. Wise addresses a syndrome that “consists of anger or rage, aggression, and some actual assaults and murders.” He stated:
“More such reports of this type have been received by the FDA for triazolam and alprazolam than for any other drug product regulated by the Agency. Reporting rates, which adjust for differences in the extent of each drug’s utilization, reveal much higher ratios of hostility reports to drug sales for both triazolam and alprazolam than for other benzodiazepines with similar indications. The public health importance of these reactions lies in their severity, with occasionally lethal behavior unleashed, in the context of large population exposures as the popularity of both drugs continues to rise.”
After a brief history of the FDA’s increased focus on BZ-induced hostility, Wise explains “Our concern with such reactions then broadened to the class of triazolobenzodiazepines, when another Increased Frequency Report included a reaction in which a 57 year old woman fatally shot her mother two hours after taking one-half milligram of triazolam. When we looked at reports received during 1988, we found that triazolam’s 1988 reporting rate for hostility reactions was more than twice as high as alprazolam’s.”
According to Wise (1989):
“In the entire SRS (spontaneous reporting system) during early August, 1989, triazolam was the suspect drug in 113 reports coded as hostility, more than any other medication. It was followed by alprazolam, which accounted for 78 reports. Only nine other drugs were suspected in more than ten cases each. Another 318 drug products had fewer hostility reports, most often one (60.4 percent of 318) or two (14.8 percent).”
Three fatalities were reported to the SRS for triazolam and one for alprazolam. Five reports of alprazolam overdose were associated with assaults, including two murders. Reactions were reported across the dose range. Males (29) and females (26) were almost evenly distributed. Four alprazolam cases showed a reduction in hostility and rage reactions with a reduction in dose, confirming the drug’s role in producing the behavior. Wise summarized, “This apparently excessive number of rage and similar reports with triazolam and alprazolam, after adjusting the differences in frequency of drug use, provides strong suspicion that a causal relationship may obtain.” It should be added, the dose dependency seen in several cases further confirms causation. Wise concluded that these reports cannot “prove the presence of a causal relationship” to the drug, but that they do “imply a substantial public health importance for the potential hostility syndrome.”
On April 21, 1989, Wise wrote an Increased Frequency Report for the FDA on the subject of “alprazolam and rage.” Wise explained that the analysis was undertaken because “Over a 12 month period, Upjohn received six reports of rage, agitation, anger, aggression, and similar behavioral and emotional symptoms after exposures to alprazolam.” All but one involved “manifested or verbalized murderous impulses.” According to Wise: “From spontaneous reports alone, we cannot estimate the actual incidence of alprazolam-induced rage reactions. But in light of the widely acknowledged, substantial under-reporting to spontaneous surveillance systems in general and the to the FDA’s SRS in particular, it is entirely possible that six reports of this kind of reaction within a single year might reflect sixty or more in reality.”
After reviewing all reports made to Upjohn and the FDA, Wise concluded:
“An increase in annual frequency of “rage” reports with alprazolam promoted us to compare hostility reports more generally across several anxiolytic benzodiazepines. Alprazolam appears to have an excessive reporting rate for events coded with “hostility,” even after adjusting for differences in the extent of each drug’s utilization. The numbers and potential gravity of these reactions and their possible relationship to dosage all appear to conflict with current labelling’s brief description of “paradoxical effects” that occur only “in rare instances and in a random fashion.”
On October 17, 1988, Charles Anello, Deputy Director of the Office of Epidemiology & Biostatistics, referred to an earlier FDA comparison of spontaneous reports concerning triazolam to two other BZs used to treat insomnia, temazepam (Restoril) and flurazepam (Dalmane). Anello stated that there was a proportionately increased number of reports concerning abnormal behavior in regard to triazolam. Anello reported on a further analysis comparing triazolam and temazepam, showing that for triazolam the FDA receives proportionally more adverse drug reaction reports (ADRs), more serious ADRs, and more reports of five selected “behavioral” drug reactions.
On September 12, 1989, Anello reported within the FDA on “Triazolam and Temazepam–Comparison Reporting Rates.” He found that adverse drug reactions were reported eleven times more frequently with triazolam than with temazepam. The relative reporting rate was 46 to 1 for amnesia, 9 to 1 for “agitation, anxiety and nervousness,” 16 to 1 for psychosis (“psychosis, hallucinations, paranoid reaction, and acute brain syndrome”), and 19 to 1 for “hostility and intentional injury.” Anello’s analysis indicated that there were no convincing explanations for these differences other than actual drug effects; but he did not make a formal determination of causality. However, in a handwritten analysis attached to the document that was obtained through the Freedom of Information Act, there is a summary entitled “Other Evidence in Favor of Effect of Triazolam” which we produce in full:
“1) Temporal relationship of reactions to initial dose
2) Large proportion of spontaneous resolution with drug withdrawal pos[itive] dechallenge [In dechallenge, a drug is withdrawn to see if an adverse reactions then stops, confirming that it was drug induced.]
3) a few reports of positive rechallenges [In rechallenge, after an adverse drug effect disappears upon withdrawal of the drug, the drug is given again to see if the reaction can be repeated and reconfirmed.]
4) reports of reactions in otherwise normal individuals
5) corroborating reports in literature (including WHO data–similar magnitude of reactions in Canada in data through 3/87)”
Their analysis indicates some of the logical, scientific steps by which data from spontaneous reporting were used by an unidentified FDA official to confirm causality in regard to Halcion and adverse behavioral effects (for a further discussion of the scientific process in regard to epidemiological studies, see appendix).
In an August 11, 1989 in-house FDA memo from Paul Leber and Thomas Laughren to an upcoming Psychopharmacological Drugs Advisory Committee Meeting (PDAC), the FDA officials commented that “Spontaneous reporting of adverse events in the United States subsequent to the marketing of Halcion has consistently revealed a pattern of excess reporting of events for Halcion compared to the two other marked benzodiazepine hypnotics.” He brought up the earlier hope that the recommendations to reduce the dose of Halcion would result in a reduction of the high rates of reported adverse behavioral reactions. However, this proved to be a false hope. Leber observed, Halcion continues to exhibit an excess of adverse events reported compared to one comparitor agent [Restoril]” in regard to “the cluster of behavioral adverse reactions.” In 1991, Diane Wysowski and David Barash, also from the FDA’s Division of Epidemiology and Surveillance, published a report in the Archives of Internal Medicine. A footnote stated, “This article contains the professional views of the authors and does not constitute the official position of the Food and Drug Administration.” Using the FDA’s spontaneous report system, they compared triazolam and temazepam through 1985 for “confusion, amnesia, bizarre behavior, agitation, and hallucinations.” They concluded, “Considering the extent of use, reporting rates for triazolam were 22 to 99 times those for temazepam, depending upon the reaction.” Echoing the handwritten remarks appended to the in-house report by Anello (1990), the authors summarize:
“Factors that indicate a causal association between triazolam and adverse behavioral reactions include corroborating case reports and sleep laboratory studies in the literature, reports of reactions in otherwise normal persons, acute onset and temporal relationship to reactions with initial dose, spontaneous recoveries and return to normalcy with drug discontinuation, and occurrences of positive rechallenge. Also, the high benzodiazepine receptor affinity with triazolam has been postulated as a possible biological mechanism.” [citations omitted]
While unable to “completely exclude the possibility that some selection factors are operating to produce higher reporting rates for triazolam,” nonetheless they find that the “evidence suggests” a greater occurrence with triazolam and than with temazepam. Andreadis and Schirmer (1992) responded critically for Upjohn with a letter and Wysowski and Barash (1992) then tried to answer their objections.
American and British Responses Diverge
Finally, in November 1991, the FDA approved new labelling for Halcion (Food and Drug Administration, April 1992). The new label emphasizes that triazolam is indicated for short-term use, and specifies 7-10 days. Treatment lasting longer than 2 to 3 weeks requires a complete reevaluation of the patient. In addition, the label emphasizes the use of the lowest possible dose.
Here is the new warning in the Halcion label as found, for example, in the 1994 Physicians’ Desk Reference:
“A variety of abnormal thinking and behavior changes have been reported to occur in association with the use of benzodiazepine hypnotics, including HALCION. Some of these changes may be characterized by decreased inhibition, eg, aggressiveness and extroversion that seem excessive, similar to that seen with alcohol and other CNS depressants (eg, sedative/hypnotics). Other kinds of behavioral changes have been reported, for example, bizarre behavior, agitation, hallucinations, depersonalization. In primarily depressed patients, the worsening of depression, including suicidal thinking, has been reported in association with the use of benzodiazepines.”
The warning concludes with the following:
“As with some, but not all benzodiazepines, anterograde amnesia of varying severity and paradoxical reactions have been reported following therapeutic doses of HALCION. Data from several sources suggest that anterograde amnesia may occur at a higher rate with HALCION than with other benzodiazepine hypnotics.”
The final label change was negotiated and approved under the authority of Paul Leber, director of the Division of Neuropharmacological Drug Products, the division responsible for Halcion’s original approval. In several ways, the label seems to fall far short of conclusions generated by both the literature and the division responsible for postmarketing surveillance. The FDA label does mention the disproportionate reporting of amnesia; but by omission it leads the reader to believe that the behavioral effects did not occur with increased frequency. Instead of linking directly to Halcion the enormously increased risk for violence, psychosis and other behavioral extremely hazardous abnormalities, label notes these changes have been “reported in association with the use of benzodiazepine hypnotics, including triazolam.” As we documented earlier in this chapter, Charles Anello, Deputy Director of the Office of Epidemiology & Biostatistics, reported on a comparison of adverse drug reaction reports for Halcion and Restoril. For Halcion versus Restoril, the relative reporting rate for “agitation, anxiety and nervousness” was 9-1; for psychosis, 16 to 1; and for “hostility and intentional injury,” 19-1. Great Britain took a stronger stand and ended up banning Halcion. On October 1, 1991, the Committee on the Safety of Medicines (CSM) gave notice of the withdrawal of Halcion from the market because of concerns about safety, especially in regard to causing memory loss and depression (Brahams, 1991; Asscher, 1991). On December 9, 1991, the Committee on Safety of Medicines (1991) responded to Upjohn’s appeal with a definitive scientific conclusion about the dangers of Halcion (p. 1). It found what it called a clearly established causal relationship between Halcion and adverse psychiatric effects. These adverse effects occurred, in the CSM’s opinion, far more frequently than with other BZs. The CSM declared that the spontaneous reporting system data from the US and England confirmed or strengthened the connection between Halcion and various psychiatric side effects. Concerning the FDA epidemiological data, the CSM observed that despite differences of opinion within the FDA, the US data provided a signal requiring further investigation (p. 10).
Other Risks of Benzodiazepine Use
BZs As Instruments of Suicide
Some of the tricyclic antidepressants and barbiturates are probably more toxic than BZs taken alone. But when BZs are combined with other drugs, such as alcohol, their lethality is increased. Overall, the BZs account for many more suicides than most physicians probably realize.
A survey in Britain covering the decade of the 1980s demonstrated large numbers of successful suicides using BZs, alone and in combination with alcohol (Serfaty and Masterton, 1993; also see Buckley, et al. 1995). Serfaty and Masterton found 891 fatalities with BZs alone and 591 in combination with alcohol. The total of all poisonings attributed to BZs was 1576 during the ten year period, putting them ahead of aspirin/salicylates at 1308 as well as amitriptyline (1083) and dothiepin at 981. That latter two drugs accounted for over half the fatal poisonings attributed to antidepressants.
Among the BZs, two commonly prescribed for sleep, flurazepam (Dalmane) and temazepam (Restoril), had the most deaths per million prescriptions (15.0 and 11.9 respectively). They were more dangerous than about half the antidepressants surveyed by the same methods. Triazolam (Halcion) had far fewer deaths per million prescriptions (5.1) than Dalmane or Restoril; but it was still above the mean for anxiolytic BZs (3.2). In estimated deaths per million patients, the rank order among all BZs in Britain was dominated by the hypnotics. Dalmane (90 per million) was first; Restoril (71) was second; the British hypnotic, flunitrazepam (Rohypnol) (49), was third; and Halcion was fourth (30). Another British hypnotic, nitrazepam (Mogadon and others) (26) was fifth. In death per million patients, among the anti-anxiety drugs, prazepam (Centrax) (25 estimated deaths per million patients) and alprazolam (Xanax) (24) were close behind triazolam and nitrazepam.
BZ-Induced Cognitive Dysfunction
Cognitive impairment, including memory impairment and confusion, is a well-known phenomenon in association with BZs (Ashton, 1995; APA, 1990; Golombok, Moodley and Lader, 1988; Hommer, 1991). Xanax and Halcion are especially prone to produce cognitive deficits. Individuals who take Halcion to sleep on an airplane may end with a blank in their memories for the period surrounding the trip. Students who take BZs before exams in order to relax or to get sleep are in danger of losing some of the material they have been studying. Triazolam produced more sedation and greater impairment of psychomotor performance at the same dose in healthy elderly persons than in healthy young persons (Greenblatt et al., 1991).
Effects on Sleep and the EEG
The effects of the BZs on the EEG resemble those of other sedative-hypnotic agents, including decreased alpha activity and increased low-voltage fast activity, especially beta activity (Rall, 1990). Their effects on sleep are also similar to those other CNS depressants, and provide a window into the dysfunctions they produce (Rall, 1990). Before the brain rebounds after one or more doses, the BZs decrease sleep latency (the time it takes to fall asleep) and reduce the number of awakenings. The overall time in REM sleep is usually shortened, but the number of cycles of REM may be increased later in sleep. Total sleep duration is usually increased. There are complex effects on the dream process.
Within a short time of starting Halcion, rebound begins to the dominate the clinical picture, and insomnia worsens. Nishino, Mignot and Dement (1995) observe that short-acting BZs were initially preferred for elderly patients. They remark, “However, it has since been found that short-acting benzodiazepines induce rebound insomnia (a worsening of sleep beyond baseline levels on discontinuation of a hypnotic, rebound anxiety, anterograde amnesia, and even paradoxical rage.” In general, the usefulness of BZs in insomnia is temporary at best. They do not provide for normal sleep, but rather for a disruption in various aspects of the normal cycle.
The Probability of Permanent Brain Damage
There have been relatively few studies on the persistence of cognitive deficits following termination of BZ treatment. Despite the analogy with alcohol-induced cognitive dysfunction and dementia, most review articles and book chapters do not raise the issue. As I have previously noted (1991), patients on high doses of BZs develop chronic cognitive impairments (Golombok, Moodley, and Lader, 1988; Lucki, Rickels, and Geller, 1986). There is little literature concerning these effects. At least two reports indicate brain atrophy in association with long-term BZ use (Lader and Petursson, 1984; Schmauss and Krieg, 1987). There seem to be no follow up studies concerning these critical questions. Ashton (1995) is among the few reviewers to show concern about the possibility of BZ-induced persistent cognitive deficits or brain damage. He notes the lack of studies and comments, “It remains possible that subtle, perhaps reversible, structural changes may underlie the neuropsychological impairments shown in long-term benzodiazepine users.”
Dependence and Withdrawal
Alcohol-like withdrawal symptoms from the long-term use of excessively high doses of BZs is well-established. Withdrawal can take two to twenty days to develop after abrupt termination of the drug, depending on the half-life of the particular BZ. The first signs may be insomnia, irritability, and nervousness. This can progress to a state that includes abdominal cramps, muscle cramps, nausea or vomiting, trembling, sweats, hyperarousal and hypersensitivity to environmental stimuli, confusion, depersonalization, anxiety and obsessional states, psychosis and organic brain syndrome, and seizures. Occasional case reports suggest that even slow withdrawal may not obviate serious withdrawal symptoms. It is unclear if gradual withdrawal merely extends the process over time rather than avoiding it (Noyes, 1992). Many symptoms can take weeks or months to fully subside, leaving the patient with prolonged anxiety or depression (Ashton, 1995). I have seen patients who felt they had not regained their pre-drug condition many months or years later. Kales, Manfredi, Vgontzas, Bixler, Vela-Bueno, and Fee (1991), in a placebo controlled sleep lab study, showed that even under “brief, intermittent administration and withdrawal” of triazolam, and to a lesser extent temazepam, patients experienced rebound insomnia, “thereby predisposing to drug-taking behavior and increasing the potential for drug dependence.”
There are estimates that 50% or more of patients taking BZs in therapeutic doses over a year will become physically dependent, developing withdrawal symptoms on abrupt cessation (Noyes, 1992; Ashton, 1995). Among the BZs used primarily for the treatment of anxiety or panic, alprazolam seems to have an especially bad record. In the field of drug addiction, Xanax is the most frequently implicated psychiatric drug (Breggin, 1991). Often it occurs in cross-addiction with alcohol and other sedatives. Withdrawal problems and rebound increases in anxiety and panic were so extreme in key studies used for FDA approval of Xanax for panic disorder that many or most patients had more frequent or severe symptoms at the end of the studies than before they took the drug (reviewed in Breggin, 1991).
Some patients can find it difficult to withdraw from as little as 0.5 clonazepam each night for sleep. Even motivated patients have sometimes developed such a fear of trying to sleep without BZs that they cannot undertake a serious effort. The fear is in part psychological. But it is also based on the frightening experiencing rebound insomnia. Physicians erroneously prescribe BZs in ever-increasing doses, mistakenly thinking that their patient’s anxiety was spontaneously increasing rather than rebounding from the drug. Even if the ultimate dose remains within the recommended range, patients can roller coaster with anxiety or other mental aberrations through each day between doses. The patients lives can become devoted to “finding the right drug” and “taking it at the right time.” It requires a physician’s patience and understanding, and often a period of many months, in order to wean some individuals off the BZs. At the end of the weaning, patients may discover that nearly all of their symptoms were drug-induced. I discuss the general principles of drug withdrawal in outpatient practice in greater detail in Talking Back to Prozac. Patients taking large doses of BZs need detoxification in a hospital setting.
Where patients have not been properly monitored by physicians, they may end up taken large doses of BZs for prolonged periods of times. Their daily lives may cycle from periods of excessive sedation when they appear “drunk” to periods of hyperarousal and anxiety as they undergo partial withdrawal. Friends and family may attribute their symptoms to “mental illness” until, for example, the patient begins to stumble about in a drunken manner or collapses in a stupor after one alcoholic drink during a holiday dinner. In retrospect, it will be apparent that the patient was, for months, too intoxicated to properly evaluate his or her own condition or to exercise judgment in regard to the drug effects. Often the patient’s memory for the period of time will be severely impaired. Sometimes they will have committed irresponsible and even illegal acts. In summary, the benzodiazepines can produce a wide variety of abnormal mental responses and hazardous behavioral abnormalities, including rebound anxiety and insomnia, psychosis, paranoia, violence, antisocial acts, depression, and suicide. The BZs impair cognition, especially memory, and can cause confusion. They probably can cause brain atrophy. They are habit-forming and addictive, and when taken in large doses for a period of time, severe withdrawal syndromes can develop on termination of the drug. Mixed with alcohol and other sedatives, their hazards multiply. Halcion and Xanax so commonly produce a variety of potentially disastrous adverse reactions that they should not be prescribed by physicians. If a physician or patient decides to use another of the BZs, the drug should be prescribed in the smallest possible dose for the shortest possible time, usually no more than a week at a time. The physician, the patient, and the family or close friends should be alert for possible adverse drug reactions.
Drawing Scientific Conclusions from the FDA’s Spontaneous Reporting System (MedWatch)
There are a number of approaches that can be used to confirm from spontaneous reports that a drug is actually causing a particular adverse reaction with which it has been associated in the reports. This paper described how FDA officials went about confirming for themselves a possible or probable causal relationship between Halcion (and Xanax) and various behavioral abnormalities, including violence. These principles will now be elaborated upon. To confirm causality, some of the following factors are useful:
(1) a disproportionately high frequency of reporting or disproportionately large number of reports in comparison to other drugs, especially in the same or similar class of drugs;
(2) a meaningful or strong enough association as reflected in epidemiological and clinical data;
(3) an absence of alternative explanations for the increased frequency or number of reports;
(4) reports indicating a temporal relationship between the adverse reactions and initial doses of the drug or increased doses of the drug;
(5) reports of dose-dependent reactions, that is, increased frequency or numbers of adverse reactions with higher dosages;
(6) reports of positive dechallenge — resolution of the reaction following drug withdrawal;
(7) reports of positive rechallenges–the reaction is provoked once again by resuming the drug;
(8) reports of adverse reactions in individuals;
(9) corroborating clinical experience (published and unpublished);
(10) data from clinical trials, including controlled trials;
(11) a rational medical and/or neurochemical explanation for a causal connection between the drug and the adverse reaction, and the corresponding absence of a better explanation.
The Federal Judicial Center (Bailey, Gordis, & Green, 1994) has proposed a series of criteria that compact many of the points I have made. The difference in approach is in part due to their epidemiological emphasis in contrast to my clinical emphasis. Drawing on Koch’s postulates, they state that “seven factors should be considered when an epidemiologist determines whether the association between an agent and a disease is causal.” Put in the form of questions, they list the following factors:
1. How strong is the association between the exposure and the disease?
2. Is there a temporal relationship?
3. Is the association consistent with other research?
4. Is the association biologically plausible?
5. Have alternative explanations been ruled out?
6. Does the association exhibit specificity?
7. Is there a dose-response relationship?
None of the above individual criteria is an absolute requirement for coming to a scientific conclusion. One must weigh the best available evidence and come to as sound a conclusion as possible. Commonly or even typically, decisions with a high degree of probability will be made with an incomplete set of data.
While it would be helpful to have confirmation from controlled clinical trials, it is typically impossible to obtain it, even in regard to known or proven adverse drug reactions (Leber, 1992). Negative findings from controlled clinical trials involving a drug cannot be used to rule out a causal connection between a drug and an adverse reaction. In fact, the vast majority of changes made drug labeling during the post-marketing phase is based on the kind of data that is generated by the spontaneous reporting system and almost none of the changes result from controlled clinical trials. Even when drugs are taken off the market, they FDA often acts on the basis of spontaneous reports. In describing the impact of the spontaneous reporting system, the FDA’s director, David Kessler (1993), said:
“In response to voluntary reports from physicians to the FDA or the manufacturer, the FDA has issued warnings, made label changes, required manufacturers to conduct postmarketing studies, and ordered product withdrawals that have ultimately prevented patient deaths and suffering.”
A recent FDA MEDWatch (FDA, 1995) publication makes clear that the spontaneous reporting system is the most important source of postmarketing information on adverse drug reactions. It frequently leads to scientific determinations for the need to modify drug labels or to withdraw drugs from the market.
According to a 1990 Government Accounting Office (GAO) report, more than 50% of all drugs approved by the FDA between 1976 and 1985 were found during postmarketing to have previously undetected “serious” side effects, sometimes requiring removal from the market. Fifteen psychopharmaceuticals were approved during this period, nine of which turned up to have serious risks during postmarketing, leading in one case to removal from the market (p. 25 & p. 74-78). Since then, yet another drug from that era has been withdrawn from the market. The antidepressant nomifensine was found to cause massive intravascular hemolytic anemia–but only after it had been on the market worldwide for eight or nine years (Leber, 1992, p. 6).
I have reviewed the entire list of serious adverse reactions to psychiatric drugs detected during the postmarketing period in the GAO study. It seems probable that every one of them was discovered and confirmed through a combination of the spontaneous reporting system and general clinical experience. As far as I can ascertain, not one of these adverse reactions was primarily if at all identified by means of a controlled clinical trial. Among the nine drugs, alprazolam (Xanax) had rage added to the label as a paradoxical reaction and amoxapine (Asendin) had neuroleptic malignant syndrome added.