Home
Word Document





William James once wrote, "if we believe that no bell in us tolls to let us know for certain when truth is in our grasp, then it seems a piece of idle fantasticality to preach so solemnly our duty of waiting for the bell" (1896, p. 31). In clinical psychology, we must often make decisions before the entire truth is known. This could not be more evident than in the controversy surrounding Selective Serotonin Reuptake Inhibitors (SSRIs). Whether or not SSRIs cause suicidality in some cases is an open question--no prospective studies of adequate power have set out to investigate the matter (Healy, 2004). Yet, we live in a world where millions of people are prescribed such medication. This paper provides background information on suicide and SSRIs, reviews the available evidence of SSRI-induced suicide, and notes the implications of these findings. This report aims to make sense of the conflicting evidence surrounding SSRI-induced suicidality so that one can act accordingly, while waiting for the bell.


"Suicide" is a value-laden, inherently ambiguous term (Joiner, 2005, p. 27). It is often difficult to determine if an act of killing oneself is, in fact, a suicide. For example, the deaths of the terrorists who hijacked airplanes on 9/11 could be considered suicides or war casualties, depending on the perspective. Similarly, the death of a drunk driver in a car crash could be considered a suicide or merely an accident. Although the term is somewhat controversial, for the purposes of this review, the Mental Health Center for the Prevention of Suicide's definition is used, which characterizes suicide as the "fatal self-inflicted destructive act with explicit or inferred intent to die" (Goldsmith, Pellmar, Kleinman, & Bunney, 2002, p. 27).

Having set forth a working definition of suicide, its various risk factors can be described. Foremost, having a psychological disorder puts one at risk for suicide. In particular, 60% of those that commit suicide meet clinical criteria for an affective disorder, such as major depression or bipolar disorder (DeLeo & Evans, 2004, p. 92). Suicide is also related to specific symptoms, such as feeling hopeless, like a burden, or that one does not belong (Goldsmith, 2002; Joiner, 2005). Psychotic symptoms also put one at risk for suicide. Among schizophrenics, those that are paranoid, as opposed to having predominantly negative symptoms, are much more likely to commit suicide.

Similarly, substance abuse also increases the likelihood of suicide. Alcoholism is the most common type of substance abuse predisposing one to suicide, but the abuse of opiates, cannabis, cocaine, and amphetamines also increases risk (DeLeo & Evans, 2004). This increase likely occurs because such substances may increase impulsivity or cause psychotic symptoms.

Approximately 30% of people who commit suicide meet criteria for a personality disorder (DeLeo & Evans, 2004). Among the personality disorders, those with Borderline Personality Disorder are most likely two commit suicide. Those with Antisocial Personality Disorder also have an elevated risk.

Anxiety disorders also represent a significant risk factor for suicide, which is true for panic disorder, posttraumatic stress disorder, obsessive-compulsive disorder, and social phobia (DeLeo & Evans, 2004).

Other important risk factors include childhood trauma, having a past suicide attempt, and having available means and opportunity (Goldsmith, 2002; Joiner, 2005). In contrast, factors such as coping skills and social support are protective factors, which decrease the risk of suicide (Goldsmith, 2002).

Thus, many factors contribute to suicide, but depressive symptoms are among the most prominent. Treatments that reduce depressive symptoms, then, also ought to decrease risk for suicide. The following section provides an overview of SSRIs, which are commonly used to treat depression, and should theoretically also reduce suicide risk.


SSRIs are most-often used to treat depression, but are also used to treat obsessive-compulsive disorder, aggression, self-injurious behavior, and various other symptoms (Reiss & Aman, 1998). The drugs are selective in that they specifically block the reuptake of serotonin in the synaptic clef. However, they may also indirectly impact other neurotransmitter systems. There is little theoretical basis for their treatment of depression and other disorders. Further, evidence for their efficacy is inconsistent, though they are thought to reduce depressive and other symptoms in some individuals (Reiss & Aman, 1998).

Drugs in this class include Prozac (fluoxetine), Zoloft (sertraline), Paxil (paroxetine), Luvox (fluvoxamine), Celexa (citalopram), Lexapro (escitalopram), and others (Breggin, 2004). These drugs have slightly different efficacies and side effect profiles but are generally regarded as similar enough to be treated as a single category. For example, major side effects found in one are generally also found in the entire class of SSRIs (Breggin, 2004). Evidence that they might actually increase suicidal acts was first reported in a series of case studies.


The initial evidence for SSRI-induced suicide was unexpected and controversial. It came from an article describing several case reports. Teicher, Glod, and Cole (1990) first reported on the possibility of SSRI-induced suicidal ideation among six cases. Teicher was a well-respected biologically-oriented psychiatrist at Harvard. Glod was a nurse, and Cole worked in the McLean Hospital and was instrumental in studying the side effects of chlorpromazine in the 1950s, as he helped bring attention to the side effect of tardive dyskinesia. Although case studies are not overly persuasive in establishing scientific fact, because these researchers were fairly well-respected, their observations raised significant concern over the possibility of SSRI-induced suicidality. Among the six patients, some had never had suicidal thoughts prior to taking the SSRI. Strikingly, they developed suicidal ideation approximately two weeks after starting the drug. Teicher, Glod, and Cole had seen a lot of suicidal patients before, but these cases seemed categorically different. Some had suicidal preoccupations to the extreme, despite never having had thoughts of suicide before. One woman who had never thought about owning a gun was suddenly preoccupied with getting one. Their suicidal and aggressive thoughts were extremely violent. In all cases, suicidal ideation stopped within a few days or weeks after discontinuing the SSRI.

The Teicher, Glod, and Cole (1990) report was instrumental in raising awareness on the issue, and additional case studies were soon reported by other psychiatrists. For example, Wirshing et al. (1992) observed SSRIs to increase suicidality in five of their patients. None of the patients had a history of suicidality. All described an intense desire to pace, which was paralleled by a sense of restlessness and agitation. Insomnia and intense suicidal preoccupations were common. A couple weeks after starting the SSRI, one patient reported feeling "100 times worse than anything I've experienced before" (Wirshing et al., 1992, p. 580). Suicidal thoughts disappeared approximately two weeks after discontinuing the SSRI in each case.

The case reports were not limited merely to adults. For example, King et al. (1991) observed SSRI treatment to induce suicidality in six children. Increasing dosages led to worsening of symptoms. Discontinuation of the SSRI led to an absence of suicidality, and restarting the SSRI led to a recurrence of suicidal symptoms.

Prior to these case reports, SSRIs were believed to reduce all depressive symptoms. There was simply no reason to believe that SSRIs might reduce some symptoms, while at the same time exacerbating the problem of suicidality in some individuals. These reports do not provide conclusive evidence that SSRIs induce suicidality, but they did help to show that the link between SSRIs and suicide was in need of further evaluation.


One way of studying the association between SSRIs and suicide is to examine epidemiological evidence. This involves comparing suicide rates among those on SSRIs to those in some other group, such as those on non-SSRI antidepressant medication or placebo. The problem, however, is that there is no random assignment. The group of people taking SSRIs may be very different from those taking other types of medication. For example, there may be differences in personality, severity of symptoms, age, gender, or past number of suicide attempts. Thus, a central difficulty in epidemiological studies is showing that differences in suicide are due solely to differences in medication, rather than some extraneous variable. This is done through statistical methods, which involve partially out the effect of extraneous variables.

Grunebaum, Ellis, Li, Oquendo, and Mann (2004) found that from 1985-1999, the suicide rate in the United States was inversely related to SSRI prescription rates. This effect was still present after holding constant unemployment and alcoholic beverage consumption. Yet, this model assumes that prescription use, employment status, and alcohol consumption are the only important variables affecting suicide rates during that time period. Other variables, such as demographic shifts, differences in non-medical treatments of depression, or changes in cultural norms might lead to a decrease in suicide over this time period. In essence, it is difficult to determine the exact impact of any one variable on an overall suicide rate.

Gibbons, Bhaumik, and Mann (2005) similarly found evidence for the efficacy of SSRIs, but also failed to control for important variables. They examined data from all U.S. counties, provided by the Centers for Disease Control and Prevention. They found that people taking SSRIs had lower suicide rates than people taking older antidepressants. However, they had no way of controlling for symptom severity, so it is basically impossible to draw any conclusions from their study. People who are more depressed may have been more likely to use older antidepressants and commit suicide.

In contrast, Jick, Dean, and Jick (1995) controlled for several important variables and found an elevated risk of suicide among those using fluoxetine over older antidepressants. They looked at patient records among 172,598 patients taking antidepressants in the United Kingdom. They found that patients taking fluoxetine were twice as likely to commit suicide as those taking older antidepressants. For a subset of the patients, information on symptom severity, demographic variables, and number of past suicide attempts was available. After holding constant these variables, the effect size was the same; however, because only a subset of the population was used in the controlled analysis, the effect was no longer statistically significant (p = .08) at conventional levels. The authors, thus, concluded that all antidepressant medications worked equally well. However, because the effect size did not change in the controlled subsample, they should have concluded that their initial analysis was valid and that fluoxetine carries twice the relative risk of suicide, as compared to other antidepressants. This study shows that SSRIs carry an increased risk of suicide versus older antidepressants. Donovan, Kelleher, Lambourn, and Foster (1999) conducted a similar study in the United Kingdom and Ireland. After controlling for symptoms severity, they found that SSRIs were 3.5 times more likely to lead to suicide than older antidepressants. However, these studies do not provide evidence one way or the other with regard to whether any of these medications are better than none at all. It may be that both types of medications decrease suicidality, with SSRIs being slightly less effective than other antidepressants, or it may be that all of these drugs increase suicidality, with SSRIs being the most deadly.

Thus, epidemiological studies provide mixed evidence on the relationship between SSRIs and suicide. There are articles both for and against an association between SSRI use and increased suicide rates, but the best studies, those that control for symptom severity, tend to show that SSRIs carry twice the suicide risk (if not more), as compared to other antidepressants.


Most clinical trials tend to show no difference in suicide rates among those randomly assigned to an SSRI, placebo, or another antidepressant. For example, data from six early trials show no difference between SSRIs and older antidepressants in suicidality (DeWilde, Mertens, Fredricson Overo, & Hepfner Peterson, 1981; Gonella, Baignoli, & Ecari, 1990; Montgomery, McAuley, Rani, Roy, & Montgomery, 1981; Mullin, Pandita-Gunawardena, & Whitehead, 1988; Nathan, Perel, Pollock, & Kupfer, 1990; Perez and Ashford, 1990). However, none of these studies contained a sample of more than 73 individuals. They lacked power to detect slight-but-important differences in suicide rates. For example, Montgomery et al. (1981) promoted the safety of the first-ever SSRI, zimelidine, which was later banned worldwide for its association with a fatal nerve disease as well as suicidality. Studies utilizing small samples are rampant in SSRI research and have repeatedly lead researcher to reach dangerous, incorrect conclusions.

To summarize past research findings, Beasley et al. (1991) conducted a meta-analysis. They found no differences in suicidality between patients on SSRIs versus placebos. The study is among those most cited, yet has numerous flaws. They retrospectively analyzed suicidality using item 3 of the Hamilton Depression Rating Scale. This was a poor method of measurement because multi-item measures would be more valid indicators of suicidality; further, item 3 is relatively insensitive to some important changes in suicidality (Mann & Kapur, 1991). Additionally, because the study lacked adequate power, Beasely et al. made incorrect conclusions regarding their findings. Specifically, they found a .34% suicide rate on SSRIs versus a .17% suicide rate on placebo. Their own data show that the relative risk of suicide is twice as great for those on SSRIs, compared to those on placebo. Because these differences do not reach conventional levels of statistical significance, they incorrectly conclude that no important differences exist.

The same mistake is made in a recent meta-analysis by Khan, Khan, Kolts, and Brown (2003), which examined nine FDA trials, including 48,277 participants. The study illustrates the difficulty of studying suicide, for it is such a rare event. Only 77 of the participants committed suicide. They conclude that suicides are no more likely among patients on SSRIs, compared to those on placebo. However, .59% of participants taking SSRIs committed suicide, compared to .45% of those on placebo. This indicates a relative risk of 1.31 for those taking SSRIs. This figure is somewhat smaller than what has been reported in some of the aforementioned studies, but still important. This indicates that for every 3 patients who die on placebo, 4 would die on SSRIs. More disturbingly, Healy (2004) reanalyzed the very same data. He found that the data for several patients who had committed suicide on SSRIs during the study had been "lost." Additionally, several patients who became suicidal on SSRIs had their medication discontinued and shortly thereafter committed suicide anyway--their data had been originally coded with the placebo group because the suicides had occurred while they were no longer on medication. After correcting for these errors, Healy determined that there is actual a 3-fold risk of suicide while taking SSRIs, versus placebos. This goes to show that the data on SSRI induced suicide is difficult to classify and interpret. These two figures, 1.3 and 3.0, likely provide a reasonable interval in which the actual relative risk is likely to lie.

These figures are likely somewhat similar for children. For example, in meta-analyzing data on British children, Gunnell and Ashby (2004) found that in comparison to placebo, children randomly assigned to an SSRI are 1.66 times more likely to commit suicide. It is striking that data from a variety of analyses and sources, from SSRI advocates and critics, have arrived at similar conclusions regarding the relative risk of suicide on SSRIs. Across the aforementioned studies, researchers have found that the risk of suicide is somewhere between 1.3 and 3.5 times greater for patients on an SSRI than for those on placebo or other antidepressants. This is startling, given that SSRIs seem to be useful in reducing depressive symptoms in some patients. The relative risk of suicide would theoretically be less than 1.0 if SSRIs decreased depressive symptoms in a somewhat uniform manner. Even if SSRIs do decrease suicidality for some patients, it is evident that they actually increase suicide risk for others. The potential mechanisms by which SSRIs lead to suicide are set forth in the following section.


An assumption of this section is that in some cases SSRIs do, in fact, increase suicidality. Several researchers have proposed mechanisms that may account for the elevated risk of suicide among patients on SSRIs. Perhaps the most comprehensive evaluation of mechanisms underlying SSRI-induced suicide comes from Teicher, Glod, and Cole (1993). They outlined nine potential mechanisms, which are proposed to mediate the relationship between SSRI use and suicidality: uneven symptom relief, paradoxical reactions, akathisia, panic, mania, insomnia, obsessive and compulsive reactions, borderline states, and EEG alterations.

Foremost, it is possible that SSRIs may increase energy in some people before decreasing depressive symptoms. This could result in a greater tendency to act on depressive symptoms, via suicidality. Secondly, all psychotropic drugs are capable of causing paradoxical reactions in some individuals (Reiss & Aman, 1998). It is possible that SSRIs may have a reverse effect for some people, actually increasing depressive symptoms, which would also increase suicide risk, as depression is a major risk-factor for suicide. Akathisia is another possible mechanism, which is characterized by a state of intense internal agitation, often accompanied by physical restlessness. SSRI-induced akathisia is not fundamentally different from neuroleptic-induced akathisia, and neuroleptic-induced akathisia has already been linked to suicide (Mann & Kapur, 1991). Thus, it seems likely that akathisia may explain one mechanism by which some patients on SSRIs become suicidal. Panic attacks have also been shown to occur in some patients taking SSRIs, and panic attacks actually rival depression in terms of predisposing one to suicide (Mann & Kapur, 1991).

SSRIs are designed to elevate mood. To the extreme, symptoms of heightened mood, energy, alertness, and impulsivity mirror mania. Manic and mixed states, and particularly impulsivity, have already been discussed as risk factors for depression. Insomnia is somewhat related to these stimulating, elevating effects of SSRIs. Teicher, Glod, and Cole (1993) propose that when SSRIs cause patients to have insomnia, they may think less clearly or become overly upset with their inability to sleep and overdose on sleeping pills or alcohol.

A startling characteristic of SSRIs is that they sometimes induce obsessive suicidal preoccupations. This effect is extremely bizarre and is likely what led experts to reports such patients in the case reports described earlier. Teicher, Glod, and Cole (1993) posit that everyone has fleeting thoughts of death or suicide from time to time, and perhaps some mechanism in the brain prevents us from becoming too preoccupied with these thoughts. They counter that in some cases SSRIs may disrupt those normal processes, such that people cannot get the normally-transient suicidal thoughts out of their mind. Whereas, a non-medicated person may have a fleeting thought of buying a gun, some patients on SSRIs could, for example, become preoccupied with buying a gun, with needing it, with wanting to use it. The act of suicide may represent a means of trying to alleviate those obsessions.

Borderline personality disorder is a significant risk-factor for suicide, and Teicher, Glod, and Cole (1993) also point out that SSRI-induced borderline symptoms also lead to suicide. They describe a number of case reports in which patients with no history of borderline symptoms become obsessed with self-mutilating, shortly after starting an SSRI trial. In one case, a patient was so self-damaging that she required extensive plastic surgery to cover her self-inflicted wounds. In these cases, borderline symptoms disappeared upon medication discontinuations.

Finally, Teicher, Glod, and Cole (1993) note that many patients taking SSRIs have been found to develop EEG abnormalities. They point out that such abnormalities are also common in epilepsy, and patients with epilepsy are more likely to commit suicide, though they are somewhat vague on why this might occur.

Nevertheless, it seems clear that there are several defensible mechanisms, which explain why some people may become suicidal on SSRIs. However, Teicher, Glod, and Cole (1993) point out that there is no way of predicting in advance which patients will become suicidal or which mechanism will be relevant in any particular case.


Based on the above research, clinicians should assume that SSRIs carry double the relative risk of suicide, until well-designed research studies provide evidence to the contrary. In order to obtain a more accurate estimation of the relative risk of SSRI-induced suicide, large, methodologically-rigorous studies are needed. SSRIs should be compared to both placebo and older antidepressants to determine the suicide risk for each. Multiple measures of suicidality should be used to adequately measure the construct. Additionally, to ensure adequate power in detecting differences in suicidality, large samples should be used. Finally, the A-B-A or challenge-dechallenge-rechallenge design might be especially useful for providing convincing evidence that SSRI-related suicidal symptoms are caused by the medication and that in absence they would diminish (Healy, 2004; Mann & Kapur, 1991). The problem, however, is that such studies are extremely costly. Due to President Nixon's distrust of scientists, most psychiatric research is conducted using private, rather than public, funds, and the drug companies are the only ones capable of investing enough money to adequately investigate SSRI-induced suicide (Healy, 2004). Perhaps, a change in federal policy is necessary to ensure that research with potentially negative implications for big business is capable of being conducted.

Additionally, researchers should conduct a cost-benefit analysis to determine if the advantages of SSRI medication outweigh the side effects, including elevated suicide risk. Healy (2004) estimates that 20,000 additional suicides have been caused by SSRIs that would not have occurred had the medication not been prescribed. The question is, at what point do the deaths by suicide in some individuals outweigh the advantages of relieved depression in other folks? Perhaps the deaths do outweigh the advantages, and if it is impossible to predict who will suffer SSRI-induced suicidality, then perhaps they are not worth prescribing. Future studies need to address this concern.

Clinicians must adapt their practice to cope with this evidence (all-be-it imperfect evidence) that SSRIs cause suicidality in some individuals and that it is currently impossible to predict a priori who will develop such tendencies. Gunnell and Ashby (2004) suggest that psychiatrists should only prescribe SSRIs in cases of severe depression. Among those with only minor depressive symptoms, the risk-to-benefit ratio is likely too high, as the potential for suicide likely outweighs the minor potential benefits. Of concern, SSRIs are being prescribed more and more to those with few symptoms.

Additionally, psychologists can play a significant role in reducing SSRI-induced suicidality. Such symptoms have generally been reported to begin within a few weeks of starting an SSRI (Mann & Kapur, 1991). Psychiatrists do not always see patients regularly, so the educated clinician may be better able to monitor sudden, bizarre changes in the client's mood. Awareness of the potential for SSRI-induced suicidality is of fundamental importance, as psychiatrists often increase dosages in response to worsening symptoms. But when suicidality is triggered by the medication, the exact opposite action is necessary--the medication needs to be reduced or discontinued with care.

Finally, the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2000) provides clinicians with the opportunity to diagnose one mechanism of SSRI-induced suicide--akathisia. The manual states that SSRIs "may produce akathisia that appears to be identical in phenomenology and treatment response to Neuroleptic-Induced Akathisia" (p. 801). Although the symptoms are identical to 333.99 Neuroleptic-Induced Acute Akathisia, the diagnosis of 333.90 Medication-Induced Movement Disorder Not Otherwise Specified is to be used instead. Interestingly, the diagnosis can be given if either movement related symptoms occur or agitation occurs--both are not necessary, despite that it is labeled a movement disorder. Hopefully, future editions will provide diagnostic options for other important SSRI side effects.

In summary, the existing research on SSRIs shows that they carry twice the relative risk of suicide, as compared to placebo. In general, they increase suicidal tendencies by causing an increase in various symptoms that put one at risk for suicide, such as by increasing depression, mania, panic, agitation, and impulsivity. Because studies investigating SSRI-induced suicide have not been perfect and because many experts fail to acknowledge the link, well-designed studies are needed. In the meantime, psychiatrists should limit SSRI prescriptions, and clinicians should look for evidence of SSRI-induced side effects in their clients.