EDITOR,--Two recent articles look at use of antidepressants and death from suicide. John A Henry and colleagues have continued previous work and assessed the relation between death and defined daily dose.1 Unfortunately, the defined daily dose for dothiepin used in the paper (75 mg) should be 150 mg (references quoted by Henry and colleagues). The correct figure is then 3.08 deaths per million defined daily doses of dothiepin instead of 1.54--the second highest figure, after that for amoxapine. This agrees with a clinical study that showed that dothiepin has greater toxicity in overdose than other tricyclic antidepressants.2 We are also concerned about the statistical methods used by Henry and colleagues. It is unclear how valid confidence limits are in this context without a clear definition of x and how the standard error (not deviation) was calculated. It is inappropriate to use one tailed tests of significance without a prior hypothesis, and it is incorrect to use Fisher's exact test to compare rates. In the study by Susan S Jick and colleagues we question the inclusion of flupenthixol, which is more likely to be used for schizophrenia than depression and is usually used as a depot preparation.3 Also, the overall suicide rate of 8.3 per 10000 person years of antidepressant use seems remarkably low and is much lower than rates in affective disorder cited in the literature.4 The data from Henry and colleagues' paper give a death rate for poisoning with dothiepin alone of 11.25 per 10000 person years of use.1 These discrepancies raise three possibilities. One is that prescribers were using antidepressants mainly for conditions other than affective disorder. Another is that suicide was considerably underestimated in the cohort. In Henry and colleagues' study only 56% of the deaths due to poisoning were suicides, the remainder being undetermined. To ignore these deaths when comparing outcome may obscure differences in toxicity between antidepressants in overdose. Thirdly, there was systematic under-dosing of the antidepressants in the study compared with the defined daily dose. Obviously, combinations of all three factors may have occurred. Finally, Jick and colleagues comment that their data are consistent with the proposition that those who are determined to commit suicide will find a way that produces the intended outcome. This comment would be supported only if they could show that patients on their database who survived self poisoning went on to commit suicide by other means. This is not supported by the literature, which indicates that most patients who survive after taking an overdose do not subsequently kill themselves by other means.5