In 1991, several news reports mentioned individuals in whom suicidality became problematic after starting an SSRI (e.g., Masand and Dewan, 1991). However, that same year, an ad hoc U.S. Food and Drug Administration panel found no connection between SSRIs and suicide (Harris, 2003). Nevertheless, in 2003 the United Kingdom Committee on Safety of Medicines recommended that physicians avoid most SSRIs for treating major depression in patients under age 18 (Duff, 2003). Similarly, the U.K. Medicines and Healthcare products Regulatory Agency warned doctors against prescribing paroxetine(Drug information on paroxetine) (Paxil) for children (Vedantam, 2003). This year a "healthy" participant given duloxetine(Drug information on duloxetine) (Cymbalta) committed suicide four days after discontinuation and was the fifth suicide in premarket trials of duloxetine (Harris, 2004). Both sides of the suicide/violence issue are represented by a variety of publications.
Healy (2003) claimed an excess of suicides and suicide attempts in participants taking SSRIs compared to a placebo group. A National Institute of Mental Health-sponsored analysis found no significant difference between fluoxetine(Drug information on fluoxetine) and placebo for suicide attempts and completions (Leon et al., 1999). However, violence should theoretically be less prevalent in the treatment group because it is a symptom of the illness that is being treated. The lack of a decline in suicidality with drug treatment implies that the therapeutic effect is incomplete, at least for a group whose symptoms include suicidality.
Significant versus Reliable
On one side of the debate are reports of suicide in patients given SSRIs (The Guardian, 2003). On the other side are reports about patient groups for whom average instances of violence and suicide are not increased with SSRIs (Leon et al., 1999). If SSRI-induced violence is genuine but rare, a study of patient group averages will not reflect it. Rather, it will be hidden by patients who respond to SSRIs.
To illustrate, suppose that 30% of patients show decreased irritability with treatment, 65% hardly change and 5% show increased irritability, while average irritability falls significantly. If the consequences of increased irritability are clinically unacceptable, the statistical significance is irrelevant. A more effective treatment might add a drug that reliably decreases irritability.
Rather than group averages of violence-related behaviors, the incidence of violent acts must be described. Indeed, the reported decrease in suicides since SSRIs were first introduced suggested that these drugs diminish more violence than they might provoke (Walsh and Dinan, 2001). On the other hand, violent methods of completed suicides in patients with detectable fluoxetine levels were about three times the incidence of violent suicides of patients with tricyclic antidepressant levels (65% versus 23%) (Frankenfield et al., 1994).
Incidence of ViolenceThe fact that patients on SSRIs may commit more violent acts does not mean that the SSRIs caused them. A simpler explanation is that violence becomes worse by leaving somatic tension unrelieved. Tension does provoke violent acts such as suicide (Busch et al., 2003), and symptoms left untreated can make themselves worse. However, the sedative properties of some TCAs (e.g., amitriptyline(Drug information on amitriptyline) [Elavil, Endep]) can actually decrease tension. That is, there may be more frequent violent suicides of patients on SSRIs such as fluoxetine than on TCAs, because TCAs may offer greater tension relief. Another explanation is that physicians are more reluctant to give TCAs to patients who appear suicidal because of the toxic overdose potential (Warshaw and Keller, 1996). Both explanations presumably contribute to the finding that more violent suicides may be seen with SSRIs than with TCAs.
Studies that do not reflect positively on SSRIs are unlikely to be published, and this undermines the proper understanding of these drugs. GlaxoSmithKline acknowledged that eight of its nine studies of paroxetine in children and adolescents were not published, and the FDA assessed the overall result to be that paroxetine was not effective for depression in patients under age 18 (Harris, 2003). This nonpublication is a deliberate bias in favor of SSRIs. It obstructs us from making an impartial assessment of positive and negative effects of SSRIs by considering collections of published results, including meta-analyses. The professional ethics of withholding publication are questionable. We are left to make only impressionistic judgments. We psychiatrists seem to be overly comfortable doing so, to such an extreme that DSM expects us to overlook evidence and apply subjective impressions instead (see Psychiatric Times April 2003, p30; and June PT 2003, p23).
On the opposite side, biases against SSRIs are notorious, specifically those propagated by antipsychiatry organizations. When we dismiss these groups for their wrongful predation of the mentally ill or for posing as a religion, it is easy to also dismiss the issues they raise. This is not logically correct.
Not Just Akathisia
It is clear that SSRIs induce somatic tension in the form of akathisia (Hamilton and Opler, 1992). From what I have seen, patients who already have observable tension have the most problematic SSRI-induced akathisia. Figuratively, adding the SSRI was salting an open wound. These patients ended up discontinuing the SSRI on their own within days. Patients who can tolerate an SSRI for months seem unlikely to have problematic akathisia from it. So, it is implausible to attribute substantial violence or suicidality to akathisia, because these patients will not tolerate the drug long enough to commit violence or suicide. Nevertheless, the possibility of SSRI-induced akathisia is an additional reason to monitor for and treat somatic anxiety symptoms. As with antipsychotic-induced akathisia, patients with akathisia from SSRIs should tolerate medication better when the akathisia is relieved.
Of note, patients who do not achieve remission on fluoxetine have substantially more somatic anxiety than patients who remit (Fava, 2003). This meshes with my experience that patients referred by primary care physicians after inadequate SSRI responses show prominent restlessness. That buspirone(Drug information on buspirone) (BuSpar) does not decrease tension as benzodiazepines do probably explains why patients do not accept it (Rickels et al., 2000) or SSRIs as substitutes. Similarly, the sedative action of clomipramine(Drug information on clomipramine) (Anafranil) presumably explains its superiority over SSRIs for tension relief in OCD (e.g., Geller et al., 2003). In addition, tension mitigation surely underlies the mushrooming use of antipsychotics for nonpsychotic patients, and their recent concurrent use with SSRIs (Corya et al., 2003).
