Donovan et al (Reference Donovan, Clayton and Beeharry2000) make interesting points about deliberate self-harm (DSH) and antidepressant drugs, but their report as written is open to grave misinterpretation. Indeed, a reporter brought the article to my attention wanting to know why selective serotonin reuptake inhibitors (SSRIs) increased suicide risk relative to tricyclic antidepressants (TCAs).

A key problem with this cross-sectional, naturalistic study of DSH and antidepressant medications at emergency department presentation is that patients were not diagnosed. The authors write as if antidepressant medications are almost invariably prescribed to treat depression, yet clearly this is not always true. Even within mood disorders, patients may differ greatly in suicide risk. The authors found fragmentary evidence that patients on SSRIs may have been relatively treatment-resistant.

Moreover, SSRIs are prescribed for a growing spectrum of psychiatric illnesses beyond depression. The authors hint at the multiplicity of indications, mentioning enuresis as an indication (presumably for TCAs). Astoundingly, however, they never mention borderline personality disorder (BPD). Patients with BPD, known for their frequent parasuicidal gestures (Reference Davis, Gunderson, Myers and JacobsDavis et al, 1999), are more likely to receive SSRIs than TCAs: partly because of their safety in overdose, partly for their benefit for impulsivity independent of mood disorder. Hence BPD and other patients at higher risk for DSH may have received SSRIs rather than TCAs. The authors mention this briefly (“… the question of whether patients prescribed TCAs were similar in terms of DSH risk to those prescribed SSRIs”, p. 553) but fail to emphasise how crucial this issue is. (Neither do they mention substance misuse, a further risk factor for self-destructive behaviour.) Given this likely diagnostic and prescriptive imbalance, it is unsurprising that more suicidal patients presenting at emergency departments were taking SSRIs.

In summary, without knowing that equivalent patient populations were receiving the two classes of medications, we cannot compare their effect on suicide risk.