Back to 2.9

2.10 Prozac, suicide and aggression

The complexities that underlie claims made for the superior safety of SSRIs are also well illustrated in relation to risk of suicide, notably because fluoxetine and other SSRIs have proved safer in overdose than tricyclic and other antidepressants. This has been highlighted as an important reason for prescribing the newer drugs: a Lilly-sponsored symposium concluded, for example, that for both legal and practical reasons, "it is difficult to justify the first line use of toxic antidepressants when safer alternatives are available" (Montgomery, 1994). However, it is not that simple: SSRIs are less toxic than tricyclics in overdose, but may not reduce the risk of suicide overall:

"While it is accepted that fatal overdosage (with SSRIs) is less of a problem, the overall incidence of death by suicide does not appear to have been reduced as patients have resorted to other means of suicide". (Reynolds, 1996)

Great emphasis has also been placed on the need to prevent suicide through better recognition and treatment of depression. Failure to treat, and undertreatment, are regarded as major risk factors, and the perceived level of risk is high - "with a 15% risk of death from suicide with more severe forms of depression". (National Depressive and Manic Depressive Association, 1997) However, this widely cited figure would be less relevant in general practice as it refers to the fate of patients hospitalised for depression. They would include many resistant cases, people who hadn’t responded to drugs.

The essential proposition is that "depression probably precedes the large majority of all completed suicides" (Paykel & Priest, 1992) and that the SSRIs treat "depression" most effectively of all. There is therefore some implied linkage between risk of suicide and low levels of brain serotonin. Indeed, this has promoted much experimentation - including measurement of the levels of the principal metabolite of serotonin (5-HIAA) in the cerebrospinal fluid (CSF) of depressed, impulsive and aggressive patients, also of suicide victims. The clinical significance of the overall findings is uncertain. Apart from the obvious difficulties of distinguishing between cause and effect, no linkage was found in most patients, but has been found with disorders other than depression:

"In brief, most authors conclude that a subgroup of depressed patients (35%) fall into a low CSF 5-HIAA group and that patients with low CSF 5-HIAA are more prone to impulsive, violent suicide. This finding has not been restricted to patients with depression but is also present in patients with other psychiatric illnesses (arsonists, some alcoholics and some schizophrenics) who are suicidal or impulsive (Åsberg et al., 1987; Roy et al., 1990)." (Delgado et al., 1992).

By contrast, the following accounts emphasise the role of personal circumstances and social factors in increasing the risk of suicide. The first is a reflection dating from before the introduction of SSRIs, on "how to identify and deal with the suicide-prone"; the other is a more recent account of the problem as seen in the casualty department of St Mary’s Hospital, London. The writer of the first account is co-author of the second. He was also Chairman of the Defeat Depression Campaign and principal author of the Royal Colleges’ guidelines for the treatment of depression. See 2.12, 2.13 below.

"If we want to pick out the person who will kill himself, many studies have shown whom to look out for. The vulnerable patient is male rather than female, old rather than young, with a history of drug dependence, alcoholism or mental illness. He is childless and he is single, divorced or widowed. He will be found living alone in a cheap hotel in a densely populated part of a big town. He gives a history of a broken home in childhood, and recent break of routine (especially loss of job or retirement), and recent bereavement is common. He is likely to have had some conflict with the law, to be geographically mobile, and to be suffering from physical illness ..." (Priest, 1979)

"A prospective study was conducted of all referrals to the emergency psychiatric service of an inner-London hospital over one year. There were 53 individuals who presented with the specific and spontaneous complaint of suicidal ideation without any accompanying act of self-harm. The main diagnoses in this group were personality disorders (40%) and alcohol dependence (15%); only 13% were suffering from depressive illness. Members of the group differed from the other 369 presenters to the service in that they were less likely to be accorded a diagnosis of a defined mental illness, twice as likely to have a criminal record, and more likely to have a previous history of deliberate self-harm. A quarter of the suicidal complainants were admitted to hospital following assessment." (Hawley et al., 1992)

The question has also been raised, whether fluoxetine more than other SSRIs might induce "suicidal ideation" and occasionally precipitate suicide attempts. The manufacturers have denied it and regulatory authorities have agreed. The reason the debate persists seems essentially to do with the difference between risk and harm, there being good theoretical evidence of one but no compelling empirical evidence for the other. This could mean there was no problem or that the problem was rare, but would also reflect the many possible complications in research. Jick and colleagues, for example, found that the suicide rate with fluoxetine "seems to be substantially higher than that of the other antidepressants", but they concluded otherwise:

"... when the analysis was restricted to those without a history of having felt suicidal or who had only taken one antidepressant, the increased risk for those who took fluoxetine was reduced. We conclude that the increased risk associated with fluoxetine in the current studies may be explained by selection bias. Even after removing from the analysis subjects with a history of being suicidal or taking multiple antidepressants, there may have been residual factors which reflected a higher risk of suicide for subjects taking fluoxetine." (Jick, et al., 1995)

This population-based study concluded that "the risk of suicide was not determined by the antidepressant prescribed", and estimated the overall incidence to be one suicide per 1200 patient years. This would represent thousands in a population of millions of users, but the role of the drug is uncertain and many factors might affect the numbers involved.

The research team at the centre of this controversy has acknowledged that "the overwhelming preponderance of data indicate that these drugs are relatively safe and of unquestionable value" and "have provided countless patients with undeniable relief". Nevertheless, they suggest problems might be masked. Following an extensive review, this team identified a range of clinical mechanisms that might promote suicidal tendencies and concluded thus:

"Although antidepressants diminish suicidal ideation in many recipients, about as many patients experience worsening suicidal ideation on active medication as they do on placebo. Furthermore, at least as many patients attempted suicide on fluoxetine and tricyclic antidepressants as on placebo ... These observations suggest that antidepressants may redistribute risk, attenuating risk in some patients who respond well, while possibly enhancing risk in others who respond more poorly. Sophisticated studies will need to be conducted to meaningfully explore this possibility". (Teicher, et al., 1993)

Has there been any discernible effect on suicide rates, since the start of the new war on depression ? Suicide rates in the USA, (Mrela, 1997) where SSRIs have been most used, and in England (Department of Health, 1997) give no evidence of any national dose-response.

The possibility that fluoxetine (among other SSRIs) may trigger aggression and hostility has often been discussed, but the issue remains open to question. (DTB, 1992) The many anecdotal reports of such effects are impossible to evaluate individually, as are the occasional reports of atrocities in which Prozac is alleged to have played some part, even when documented with great care. (Cornwell, 1996) The complexities of analysis obscure almost everything, bar the feeling that it would be mad to assume they were all groundless. Collectively, they add to the impression that all is not well, not that the courts would be the place to establish what might have gone wrong. (Ibid.)


Contents page