To explain why anti-depressant treatment with selective serotonin reuptake inhibitors (SSRIs) does not improve cardiac prognosis, Korszun et al suggest that SSRIs may not alter the mechanisms through which depression leads to increased cardiovascular morbidity and mortality. However, two other explanations may be more plausible. First, the effects of antidepressant treatment on depression itself are not strong enough. In both the ENRICHD and SADHART trials, the response rates of patients in the active treatment arm hardly exceeded those of patients receiving usual care or placebo. Second, the cardiotoxic effects of depression are limited to patients for whom antidepressant treatment is not effective (Reference Grace, Abbey and KapralGrace et al, 2005; Reference De Jonge, van den Brink and Spijkermande Jonge et al, 2006a ). We have shown that the cardiotoxic effects of depression are concentrated in incident post-myocardial infarction depression, whereas results from the SADHART study have indicated that sertraline is only effective in non-incident depression (of interest, Korszun et al mention mechanisms related to recurrent depression, which appears not be to cardiotoxic). If antidepressant treatment is only effective in non-cardiotoxic depression, no effects on cardiovascular prognosis can be expected.

Shetty raises ethical concerns about our study, because we used Zelen's method of randomisation. Controls were not told about their diagnosis of depression and, as argued by Shetty, therefore may have ‘missed’ an offer of adequate treatment. However, we feel that in 1999, when the study started, Zelen's method was both scientifically useful and ethical because no controlled comparative studies had yet investigated the clinical efficacy and safety of antidepressant drugs in depression post-myocardial infarction. At that time, the proportion of myocardial infarction patients with depression who were treated for their post-myocardial infarction depression was negligible. In addition, serious concerns existed about the safety of antidepressant drugs in cardiac patients. Moreover, in our study patients with a significant risk of suicide or severe depression were excluded from randomisation and referred for psychiatric treatment outside the study. Finally, all patients received usual care, i.e. had cost-free access to all usual treatment facilities such as normal cardiac rehabilitation programmes and healthcare from family physicians. We therefore feel it was ethical to use Zelen's method in our study and scientifically useful as our control patients were truly representative of patients receiving usual care.

We agree with Dr Kho that we need to develop new treatments for depression post-myocardial infarction, but believe it is premature to consider electroconvulsive therapy (ECT) as an effective alternative. In our experience those types of depression that are least similar to those seen in general psychiatry (i.e. incident depression occurring for the first time after myocardial infarction; Reference De Jonge, Ormel and van den Brinkde Jonge et al, 2006b ) and those that are dominated by feelings of exhaustion rather than negative self-esteem or suicidality (Reference De Jonge, van den Brink and Spijkermande Jonge et al, 2006a ) are the most cardiotoxic. To our knowledge the mechanism(s) explaining this remain unclear. Similarly, it is not known whether ECT is effective in these subtypes (although it appears that antidepressive medication is not). In fact, the studies mentioned by Dr Kho suggest increased rather than decreased cardiovascular events.

New, effective treatments for depression post-myocardial infarction will improve quality of life but perhaps also survival, as rightfully argued by Dr Kho. Carney et al (Reference Carney, Blumenthal and Freedland2004) showed that responders to anti-depressive medication had a better cardiovascular prognosis than non-responders, using data from the ENRICHD trial. In the MIND–IT study we recently confirmed that non-response to mirtazapine and citalopram was associated with more cardiovascular events compared with responders and even untreated controls, a finding that remained after controlling for several confounders, including early cardiovascular events (Reference De Jonge, Honig and van Mellede Jonge et al, 2007). However, as it is unclear what factors are related to response to antidepressive medication (these may well include the presence of somatic symptoms of depression; Reference Tylee and GandhiTylee & Gandhi, 2005), it also remains uncertain whether it might be an improved state of the heart disease that influences depression or reversely that treatment of depression results in an improved cardiovascular prognosis. However, although causality remains unproven it suggests that more effective treatments may have cardiovascular effects as well. We are not yet convinced that this will be ECT but we encourage researchers to explore this possibility.