Making causal assertions from complex cross-sectional data is risky and may lead to erroneous clinical advice. Although a negative association has been revealed between alcohol ingestion in self-poisoning and taking psychiatric medications (particularly a tricyclic or a typical antipsychotic), ^{Reference Chitty, Dobbins, Dawson, Isbister and Buckley1} individuals who are prescribed these medications may be different from those who are not, even after adjusting in covariate analysis for a generic category of ‘psychiatric diagnosis’. This association even led Chitty et al to speculate that D2 antagonists might reduce the use of alcohol. However, there is evidence to the contrary: flupenthixol led to more drinking when tested in randomised controlled trials (RCTS), ^{Reference Wiesbeck, Weijers, Lesch, Glaser, Toennes and Boening2} and olanzapine caused a similar trend. ^{Reference Guardia, Segura, Gonzalvo, Iglesias, Roncero and Cardus3} In the remaining 10 of 13 RCTs found in a systematic review, antipsychotics did not reduce drinking. ^{Reference Kishi, Sevy, Chekuri and Correll4}

Clearly, there are various interpretations of the association that was found. For example, perhaps people who have access to highly sedating and potentially lethal drugs such as tricyclics and antipsychotics can self-poison seriously without recourse to added alcohol.

While Chitty et al raise some interesting questions, we are concerned that those reading the abstract alone might misperceive a role for antipsychotics in drinkers. Suicide rates in people who drink heavily might be best prevented by improving treatment and access to treatment for alcohol use disorders.