Dr Healy responds to concerns raised by Markowitz (2001) about the potential for misinterpretation of the Donovan et al (2000) study. Markowitz points out that the patient populations receiving SSRIs and tricyclic antidepressants were not similar, and accordingly that comparisons of the effects of the two classes of antidepressants on suicide risk are not meaningful. Dr Healy suggests that there are data in the public domain bearing on this issue, citing the meta-analysis performed by Khan et al (2000) and data obtained from the US Food and Drug Administration. Khan et al found no difference in suicide or suicide attempts with the use of antidepressants compared with placebo. Dr Healy claims that suicides and suicide attempts during ‘washout rather than while on placebo’ invalidate the results of Dr Khan et al's analysis.

With respect to paroxetine, Dr Healy misstates the scope of the Khan et al meta-analysis, and the conclusions he draws lack scientific substantiation. Dr Healy fails to recognise that the exposure time of patients on paroxetine in the clinical studies was substantially different and far greater than that on placebo — under these circumstances an analysis of absolute numbers of patients with no consideration of time of exposure is not meaningful. Furthermore, contrary to Dr Healy's implication, the Khan et al report was not limited to randomised, placebo-controlled studies. In the case of paroxetine, the studies covered included open label extensions, studies without placebo arms, and studies that were not randomised. When one considers only the randomised, controlled portions of the placebo-controlled trials (excluding events occurring during placebo run-in) included in the Khanet al analysis, there are no statistically significant differences in suicides or suicide attempts between paroxetine and placebo, either in absolute numbers of patients or when adjusted for time of exposure.

Donovan et al caution about the conclusions that should be drawn from the study. They point out that physicians are following guidance to prescribe antidepressants that are purportedly ‘safer in overdose’ to patients who are perceived to be at greater risk of deliberate self-harm. Consistent with Dr Markowitz's comments, this prejudices against SSRIs when associations are made between their use and deliberate self-harm. Donovanet al also note that it is problematic attributing the cause of deliberate self-harm to antidepressant treatment when such behaviour occurs as a symptom of depressive illness itself and that establishment of cause and effect is ‘almost impossible’.

The ‘drug v. “true placebo” ’ analysis Dr Healy describes is not only scientifically invalid, but also misleading. Major depressive disorder is a potentially very serious illness associated with substantial morbidity, mortality, suicidal ideation, suicide attempts and completed suicide. Unwarranted conclusions about the use and risk of antidepressants, including paroxetine, do a disservice to patients and physicians.