Hypnotic Medications and Suicide: Risk, Mechanisms, Mitigation, and the FDA

Our review found one hypnotic medication toxicology study from the United States. Toxicology findings of 187 consecutive suicide victims (137 men and 50 women; mean age=44 years) from the San Diego coroner’s office in 1981 and 1982 indicated that hypnotics were found in 10.7% of cases, and a third of those taking hypnotics also tested positive for alcohol (11). There were no cases in which a benzodiazepine was the only drug detected and also judged to be the sole cause of death. The authors of the study concluded that in cases of suicide related to benzodiazepines, multidrug overdoses are of primary concern.

Our review found two hypnotic medication toxicology studies from outside the United States. A study of 1,467 suicide deaths in Switzerland in 1990 reported 22 cases in which a benzodiazepine hypnotic was the single agent leading to a suicide poisoning (1.5% of all suicide deaths); there were 16 cases involving flunitrazepam, four with midazolam, and two with triazolam (12). However, the authors conceded that the role of alcohol in these single-agent cases may have been underestimated.

A study of 2,245 elderly suicides in 2003 using the Swedish Cause of Death Register found that 548 of the suicides were by poisoning, with benzodiazepines playing a role in 216 cases. A single benzodiazepine was the cause of death in 155 cases, and these included the hypnotics flunitrazepam (N=104), nitrazepam (N=36), and triazolam (N=1) (13). Upon reviewing all suicides in Sweden between 1992 and 1996, the authors found that after correcting for the differences in rates of prescribing for flunitrazepam, nitrazepam, zopiclone, and zolpidem, the associations between these four hypnotics and suicide were about the same, with approximately one suicide per million daily doses of each drug. Together, these toxicology studies show that modern hypnotics are often involved in suicide deaths but are usually taken in combination with other sedatives or alcohol.

We found one retrospective cohort study from the United States. A secondary analysis of the National Comorbidity Survey Replication surveyed 5,692 randomly selected community-dwelling adults between 2001 and 2003, with 5.8% of respondents answering “yes” to a question inquiring about the use of sleeping pills such as zolpidem or zaleplon in the past 12 months (14). Of these, 2.6% endorsed suicidal thinking in the past 12 months, 0.7% endorsed suicide plans, and 0.4% admitted to a suicide attempt. Hierarchical multivariate logistic regression models, with suicidal thinking, plans, and behaviors as the dependent variables, and with hypnotic use as the independent variable adjusting for gender, age, ethnicity, marital status, education, poverty status, general health conditions, mood disorders, anxiety disorders, substance use disorders, and insomnia, indicated odds ratios of 2.2 (95% CI=1.5–3.3) for suicidal ideation, 1.9 (95% CI=1.1–3.3) for suicide plan, and 3.4 (95% CI=1.6–7.4) for a suicide attempt, as compared with persons who did not use hypnotics. Strengths of the study include its representative sampling technique, large size, adjustment for insomnia, and inclusion of the Z-drugs. Weaknesses include lack of consideration of high-risk personality disorders, lack of measurement of the severity of depressive illness, the possibility of recollection bias, lack of measurement of completed suicide deaths, and a retrospective design that makes it impossible to understand the temporal association between hypnotic use and suicidal phenomena. Overall, the study adds credence to the idea that in some instances hypnotics may increase risk for suicide, but the study does not shed light on mechanism. The authors correctly point out that clinical trials are required to show causality between hypnotics and suicide. However, the authors calculated the sample size required to identify a difference between hypnotics and placebo in rates of suicide attempts in a sample of insomniacs, and this calculation yielded a sample size of 1,068 participants (534 per group). It seems unlikely that such a large trial would ever be attempted; therefore, there is justification for trials to detect a proxy risk factor, such as suicidal ideation, rather than suicide attempts.

We discovered two retrospective case-control studies from outside the United States on the topic of suicide and hypnotics. A retrospective case-control study examined medications prescribed to 25 Finnish psychiatric inpatients who died by suicide during their hospitalization, compared with 25 inpatient comparison subjects matched on age, gender, and psychiatric diagnosis (schizophrenia [N=14]; major depression [N=7]; personality disorder [N=3]; “neurosis” [N=1]) (15). The suicide victims were more likely to be receiving a benzodiazepine (72%) than the comparison subjects (44%) (p<0.05). Limitations included lack of detail on whether the benzodiazepines were prescribed for anxiety or insomnia, lack of matching the groups on severity of symptoms, and small sample size.

The second study reported on the association between zolpidem and suicide (defined here as suicide death plus hospitalizations for suicide attempt) from a Taiwanese nationwide register covering 2002–2011. Each suicide case was matched on age, gender, occupation, and urban or rural setting with 10 comparison subjects without a history of suicide attempts (16). The adjusted odds ratio for exposure to zolpidem was 2.08 (95% CI=1.83–2.36) and remained significant regardless of whether depression, insomnia, anxiety, bipolar disorder, or other comorbidities were present. However, the study was limited by imprecision in establishing the diagnoses of depression, insomnia, and other mental disorders and by the lack of measurement of severity of symptoms. The investigators speculated that zolpidem use could lead to a “dreamlike confusion state [that may] cause bizarre behavior [and] lead to potentially dangerous self-harm and even fatality” (16).

Our review produced one prospective cohort study regarding hypnotics and suicide. In 1982, the American Cancer Society evaluated frequency of use of “prescription sleeping pills” in a convenience sample of 1,099,830 participants (17). Respondents self-determined whether they were taking a “prescription sleeping pill,” and hence the data likely mixed benzodiazepines with barbiturates and antidepressants. The sample was followed prospectively until 1986. During this interval, adjusted hazard ratios for suicide were elevated in men (1.32) and women (1.22) who had nightly use of “prescription sleeping pills,” compared with those who did not, after controlling for body mass index (BMI), cigarette smoking, diet, and other mortality predictors. Apart from the nonrepresentative nature of the sample and poor characterization of “prescription sleeping pills,” another weakness of the study included failure to take into account psychiatric diagnoses. The authors called for the measurement of suicidal ideation and behavior in clinical trials.

This review produced three prospective cohort studies from outside the United States that examined hypnotics and suicide. The first of these studies examined 26,952 adult Swedes between 1975 and 1981, 500 of whom (2%) were frequent hypnotic users (345 women and 155 men), and tracked suicides until 1985 in the national Cause of Death Registry (18). Subsequent suicides were found in 1.1% of hypnotic-treated women, compared with 0.1% of the women not taking hypnotics (odds ratio=2.6, p<0.05). No men died by suicide. Study strengths include prospective measurement, which reduces ascertainment bias risk. Weaknesses include the inability to adequately control for depression diagnoses, other psychiatric disorders, and insomnia, which are known risk factors for suicide, and the failure to define “hypnotics.”

In the second prospective cohort study, the rate of suicide attempts among benzodiazepine users was evaluated in persons enrolled in the Saskatchewan Health Databases (19). The study compared 225,796 adults taking benzodiazepines (all new prescriptions from 1979 to 1986) and 97,862 age- and gender-matched comparison subjects. The benzodiazepines included triazolam, flurazepam, lorazepam, diazepam, and oxazepam. Relative to comparison subjects, patients receiving an antidepressant concurrent with a benzodiazepine did not have increased rates of suicide attempt or suicide death over the 60 days following the filling of a new benzodiazepine prescription. In contrast, patients not taking an antidepressant when filling a benzodiazepine prescription had six times the odds (odds ratio=6.2, p<0.05) for suicide attempt or suicide death during the 60 days after the filling of a prescription, relative to comparison subjects. The authors speculated that “benzodiazepines may cause disinhibition, consisting of poor impulse control, and paradoxical reactions that may induce suicidal impulse” (19). Weaknesses of the study include the confounding of treatment (benzodiazepines and antidepressants) with diagnoses, as mood disorder diagnoses were not considered in the analyses.

In the third prospective cohort study, cause-specific mortality was tracked for 20 years in a randomly selected sample of 1,750 men and 1,773 women who responded to a mail survey in Sweden in 1983 asking, “How often do you use sleep medication?” (20). Data were collected regarding sleeping problems, medical diagnoses, depression, and BMI. There were 19 suicide deaths (10 men and nine women). The adjusted hazard ratio for suicide among regular male hypnotic users was 21.2 (95% CI=2.6–173.7), and the hazard ratio in women was 24.3 (95% CI=4.4–132.2). Limitations include the small number of observed events, the lack of measurement of alcohol and substance abuse, and uncertainty as to what medications were being included among “sleep medications.” The authors concluded that regular hypnotic use is an indicator of increased risk of mortality, including suicide, but that the study design did not show causality. They called for clinical trials to clarify the directionality of these relationships.

Collectively, these prospective cohort studies show an association between hypnotic use and suicide, but all of these studies suffer from confounding of hypnotic use with psychiatric and substance abuse diagnoses.

There are two cases in the medical literature of suicide from temazepam as a single agent (24, 25), four cases reported by poison control centers (26–29), and two more adverse event reports from the FDA, altogether accounting for eight cases of suicide death from overdose of temazepam as a single agent. These cases included six men (ages 49, 66, 73, 74, 77, and 92 years) and two women (ages 83 and 85 years). The suicide of the 73-year-old man occurred after he took temazepam for only 2–3 nights, and he gave no prior indication that he was suicidal.

There are no published cases of suicide with quazepam. A search of the FDA’s adverse event files failed to reveal any possible single-agent suicide deaths related to quazepam.

There are no published cases of suicide with estazolam. A search of the FDA’s adverse event files failed to reveal any possible single-agent suicide deaths related to estazolam.

Apart from suicide deaths, there are numerous publications of suicide attempts that occurred during intoxication with zolpidem. A compilation of spontaneous adverse event reports from Australia between 2001 and 2008 indicated that zolpidem had higher odds for “parasomnia, amnesia, hallucinations, and perhaps suicidality compared to either all other drugs or hypnotics” (36). This study calculated an odds ratio of 8.3 (95% CI=5.4–12.8) for suicidality related to zolpidem compared with other psychotropic drugs in the survey. The authors acknowledged that a role for alcohol in the findings could not be ruled out, and they implicitly raised the concern that some of the unexpectedly high risk for suicide attempts with zolpidem could be related to a higher rate of parasomnia behavior during the period of peak drug effect. This possibility is supported by several detailed case reports.

A report of 91 coroner cases from New South Wales, Australia, between 2001 and 2010 noted two cases of “bizarre, dissociative” behavior leading to suicide death in decedents with toxicology-proven zolpidem use (37). In one case, the person was dancing on a high ledge before falling, and in another case, the person impulsively jumped out of a window during an argument. A 27-year-old man with no known history of psychiatric illness, suicidal ideation, or sleep disorder took 20 mg of zolpidem, in the absence of alcohol, and then took an additional 40 mg sometime in the middle of the night. The patient’s only explanation was that he remembered feeling suicidal later in the night (38).

The FDA’s postmarketing safety review of the Z-drugs in 2006 included additional adverse event reports for zolpidem that were not found in the medical literature (10). The FDA report describes a total of 49 suicidal-related adverse events in which zolpidem was judged to be a factor, including suicidal ideation (N=12), suicide attempt or self-harm (N=16), and suicide deaths (N=21). In nine of the 49 cases, there was no prior history of psychiatric illness, and in eight cases, there was no other use of psychotropic medication. In one instance, a 33-year-old man with no history of mental illness took zolpidem for 4 or 5 nights, then on the last night took 10 mg of zolpidem at 2:30 a.m., awoke at 6 a.m. (3.5 hours later), went to work, and the next thing he remembered was “standing on a pile of boxes with a noose around his neck.” In another case, a 38-year-old man with no history of mental disorder began using 10 mg of zolpidem and died by suicide on the fourth day (method not described). Finally, a 39-year-old man with no history of mental illness took 10 mg of zolpidem for 3 nights. On the third night, his wife “found him sleep walking in the closet with no memory of how he got there.” The next morning he “could not focus” and walked into the woods and shot himself. The FDA report summarized the zolpidem cases by noting that the suicidal events often occurred within a week of initiation and that many were accompanied by bizarre behavior, hallucinations, or paranoia.

While there are no published cases of suicide with zaleplon, the FDA’s postmarketing safety review described one case of suicidal ideation, three cases of self-harm or suicide attempt, and three cases of suicide death related to zaleplon; however, only one of these cases had no prior psychiatric history. All of the zaleplon adverse events occurred within 14 days of drug initiation (10).

Sixty instances of poisoning with eszopiclone as a single agent were reported in Texas between 2005 and 2006 (39). There were no instances of death when eszopiclone was taken as a single agent in an overdose. However, the FDA’s postmarketing safety review described 16 suicide-related adverse events, including seven cases of suicidal ideation, five cases of self-injury or suicidal attempt, and four suicide deaths. A psychiatric history was reported in all but one case, and in 13 cases the adverse event occurred within 1 month of initiation. Several cases were accompanied by “irritability, agitation, confusion, or worsening depression” before the event (10).

There are no published cases of suicide with ramelteon. A search of the FDA’s adverse event files failed to reveal any possible single-agent suicide deaths related to ramelteon.

There are no published cases of suicide with doxepin prescribed in low dosages for insomnia.

In 2012, the FDA recommended prospective, systematic assessment of suicide in all clinical trials used in the registration of any new psychoactive medication. The Columbia-Suicide Severity Rating Scale was specifically named as a suitable instrument for this purpose (40). Consequently, the research and development of suvorexant followed these FDA recommendations, and the findings were reported at the FDA suvorexant advisory committee meeting in 2013 (41). Per the 2012 FDA recommendations, data were pooled across all phase III trials. In the case of the new drug application for suvorexant, this meant the collation of one combined-age, blinded, randomized, high-dose trial and one blinded, randomized, combined-age, low-dose trial. High dose and low dose were defined as follows: young adults: low dose of 20 mg and high dose of 40 mg; elderly (individuals older than age 65): low dose of 15 mg and high dose of 30 mg. A total of 11 persons expressed suicidal ideation across both studies, as defined by the Columbia-Suicide Severity Rating Scale (Table 1).

While there were no suicidal behaviors reported, and while the majority of suicidal ideation events lasted “minutes to hours and were mild to moderate in intensity,” the FDA interpreted the results as supporting a dose-response relationship between suvorexant exposure and suicidal ideation (41). As a result, the package label for suvorexant carries the unique warning that “in clinical studies, a dose-dependent increase in suicidal ideation was observed in patients taking Belsomra as assessed by questionnaires.” Examination of the specific cases of suicidal ideation as described in the advisory committee meeting shows that the data collection method failed to record whether suicidal ideation occurred in temporal proximity to drug ingestion (i.e., at night, within a few hours of ingestion) or later in the day after the drug effect would have dissipated. Finally, a search of the FDA’s adverse event files failed to reveal any possible single-agent suicide deaths related to suvorexant.