The recent case—control study by Reilly et al (Reference Reilly, Ayis and Ferrier2002) showing an association between probable sudden unexplained death and current treatment with thioridazine in psychiatric in-patients underscores two fundamental issues essential for dealing with potential risks of this kind in the context of new, atypical antipsychotics. The first is that the abandonment of older drugs that have been used successfully such as thioridazine, even with the availability of newer compounds, can have profound consequences for some patients, as has been shown in the case of discontinuation of thioridazine in patients with learning disabilities (Reference Davies, Cooke and MooreDavies et al, 2002). For those drugs that are being used successfully to control serious conditions, caution may be warranted in making changes that lead to the removal of drugs purely because of potential torsadogenic risks. Even those drugs that block the human ether-a-go-go-related gene (HERG)-encoded K⁺ channel, which is thought to mediate many of the cases of drug-induced long-QT syndrome, must be considered cautiously in this respect. Some pharmacovigilance estimates based on spontaneous reports of adverse reactions suggest that the risk of torsades de pointes for non-antiarrhythmic drugs, even those considered to be associated with risk, may be as low as 0.10 per million defined daily dosages, and that the consequent risk of sudden death may be as low as 0.025 per million (e.g. Reference Lindquist and EdwardsLindquist & Edwards, 1997); even assuming a 1% reporting rate, this risk remains small.

The second issue is that given the fact that torsades de pointes is so rare, it has been proposed that this kind of arrhythmogenesis may be a ‘multi-hit’ phenomenon in which several risk factors must simultaneously be present (Reference Keating and SanguinettiKeating & Sanguinetti, 2001). Furthermore, it has been suggested that in the normal ventricle, there is little risk of developing torsades de pointes because the normal functioning of the robust repolarising currents ensures a large repolarisation reserve, and it is by the co-occurrence of risk factors (e.g. female gender, hypokalaemia, bradycardia), which reduce the repolarisation reserve, that the likelihood of torsades de pointes is greatly increased (Reference RodenRoden, 1998). For example, we and others have previously pointed out that low serum potassium attenuates HERG activity and that increasing serum potassium has been used to correct quinidine-induced acquired long-QT syndrome (Reference Choy, Lang and ChomskyChoy et al, 1997; Reference Hancox and WitchelHancox & Witchel, 2000). Given that Reilly et al's two control groups differed radically in their use of diuretics, and given that their model using control per case showed a significant odds ratio > 100 associated with the use of diuretics, it may be important to determine potential synergies for risk mediated by hypokalaemia directly, as well as including the use of diuretics in regression analyses.

As several of the new, atypical antipsychotics recently have been shown to block the HERG K⁺ channel, the clinical implications are that without a more complete understanding of the mechanism of risk, further studies examining this association for new atypical antipsychotic agents will require, where possible, prospective studies that can be used to determine the synergistic action of other known risk factors to be measured directly. Although mortality, even a low risk of mortality, is an unacceptable effect for a drug used to treat a non-fatal condition, the successful use of drugs such as thioridazine militates against the wholesale elimination of these drugs without due consideration for individual cases.