BACKGROUND New selective serotonin reuptake inhibitors (SSRIs) are perceived to be much safer in use than older tricyclic antidepressants (TCAs). However, previous assessments of association with fatal toxicity were made too soon after the introduction of the new drugs to permit accurate estimation. OBJECTIVE To determine the level of association of antidepressant drugs with fatal poisoning in the treatment of depression. METHODS National data for England and Wales for three years (1993 to 1995) for fatal poisonings associated with antidepressants were obtained and, together with national primary care data on prescribing, were used to calculate fatality association by antidepressant drug. RESULTS There were substantial variations between drugs in the level of association with fatal poisoning. Assuming an average treatment episode lasted three months, one fatality is associated with 11,800 treatment episodes of antidepressant use (95% CI = 11,120 to 12,580) when only single substance fatalities are considered. For SSRIs as a group the association was one in 411,800 (95% CI = 243,300 to 1.34 million) and for TCAs one in 8130 (95% CI = 7650 to 8670). However, for one of the newer TCAs, lofepramine, the single substance fatality rate associated with its use was one in 233,700 (95% CI = 124,500 to 1.89 million), which is not statistically significantly different from the SSRIs (P = 0.35). CONCLUSIONS Estimated death rates associated with specific antidepressants should be compared with caution because drugs may be used selectively in patients with differing severity of depression. The proportion of these fatalities that could be prevented by switching to safer antidepressants is unclear when so few deaths are recorded as accidental; when there is intent to do self-harm the potential for switching to other means is unknown. However, this approach to relative toxicity may remain the best available since it is unlikely that a randomised trial will ever be conducted with a large enough sample size to obtain experimental data. Fatalities from antidepressant poisoning are very rare but if safety is paramount then lofepramine or an SSRI are justifiable treatment choices.