SSRIs and deliberate self-harm

Markowitz (2001), commenting on a study by Donovan et al (2000), which indicated that selective serotonin reuptake inhibitors (SSRIs) were more associated with presentations of deliberate self-harm to accident and emergency departments than were other antidepressants, suggested that it was ‘astounding’ that Donovan et al had not taken into account the fact that their results might stem simply from a preferential prescribing of SSRIs to patients with borderline personality disorder — a patient group particularly prone to self-harm. Dr Markowitz's points about the Donovan et al study come down essentially to two — that this study was not randomised and that there was no placebo control. There are, however, data in the public domain bearing on these points to which he may not have had access.

Khan et al (2000) published a meta-analysis of randomised controlled trials submitted to the US Food and Drug Administration (FDA) showing suicides and suicide attempts on a number of recently licensed antidepressants. Requests to the FDA under freedom of information provisions (further details available from the author upon request) indicate that three of five suicide attempts characterised as placebo suicide attempts in the sertraline trial programme reported by Khan et al occurred during washout rather than while on placebo. Similarly, in the paroxetine trial programme reported by Khan et al, both suicides and three of six suicide attempts characterised as placebo suicides and attempts appear to have been washout rather than placebo suicides or attempts. Taking this information into account and analysing the drug v. ‘true placebo’ data for absolute numbers of patients reveals a statistically significant general increase in the risk of suicide acts on novel antidepressants compared with placebo and a specific increase for paroxetine.

In the light of these randomised, placebo-controlled findings it would seem that Dr Donovan et al were cautious and understated in their discussion of their results.