Cheeta et al (2004) present data on antidepressant-related deaths in England and Wales, 1998–2000. They report information about antidepressant deaths collected by the National Programme of Substance Abuse Deaths (np-SAD database). The aims of their paper were: (a) to investigate relative toxicities of major classes of antidepressant (compare accidental/intentional deaths); and (b) to analyse deaths where other drugs were also present at post-mortem examination. We are concerned that the data-set used for this analysis is unable to address these objectives.
The Office for National Statistics (ONS) database of drug-poisoning deaths holds information about all such deaths since 1993. This includes textual information from the coroners’ reports about the types of substances taken. We have recently conducted an analysis of antidepressant-related poisoning deaths for a 10-year period from 1993 to 2002 (Morgan et al, 2004).
Between 1998 and 2000, Cheeta et al reported 468 deaths involving antidepressants recorded in the np-SAD database. Our analysis of the ONS database found 1452 deaths involving antidepressants for the same period. Clearly a large number of antidepressant-related deaths are missing from the np-SAD database. This is probably because about 80% of antidepressant-related deaths are due to suicide and not substance misuse (the data collected by np-SAD). Furthermore, Cheeta et al found that 93% of deaths involving selective serotonin reuptake inhibitors (SSRIs) also involved other drugs. In our study, this figure was 75%, suggesting that the np-SAD database is less likely to contain deaths involving SSRIs taken alone. This is likely to introduce bias into Cheeta et al's study, leading to an underestimation of the relative toxicity of SSRIs compared with tricyclic antidepressants (TCAs). This is borne out when death rates per million prescriptions are compared between the np-SAD study and our study. The np-SAD study gives the overall death rate per million prescriptions for SSRIs as 1.4. In our study this was 5.1. Our study showed TCA rates to be around 8 times greater than rates for SSRIs; in the np-SAD study TCA rates were 10 times greater.