The long and growing list of non-antiarrhythmic drugs associated with prolongation of the QT interval of the electrocardiogram has generated concern not only for regulatory interventions leading to drug withdrawal, but also for the unjustified view that QT prolongation is usually an intrinsic effect of a whole therapeutic class [e.g. histamine H(1) receptor antagonists (antihistamines)], whereas, in many cases, it is displayed only by some compounds within a given class of non-antiarrhythmic drugs because of an effect on cardiac repolarisation. We provide an overview of the different classes of non-antiarrhythmic drugs reported to prolong the QT interval (e.g. antihistamines, antipsychotics, antidepressants and macrolides) and discusses the clinical relevance of the QT prolonging effect. Drug-induced torsade de pointes are sometimes considered idiosyncratic, totally unpredictable adverse drug reactions, whereas a number of risk factors for their occurrence is now recognised. Widespread knowledge of these risk factors and implementation of a comprehensive list of QT prolonging drugs becomes an important issue. Risk factors include congenital long QT syndrome, clinically significant bradycardia or heart disease, electrolyte imbalance (especially hypokalaemia, hypomagnesaemia, hypocalcaemia), impaired hepatic/renal function, concomitant treatment with other drugs with known potential for pharmacokinetic/pharmacodynamic interactions (e.g. azole antifungals, macrolide antibacterials and class I or III antiarrhythmic agents). This review provides insight into the strategies that should be followed during a drug development program when a drug is suspected to affect the QT interval. The factors limiting the predictive value of preclinical and clinical studies are also outlined. The sensitivity of preclinical tests (i.e. their ability to label as positive those drugs with a real risk of inducing QT pronglation in humans) is sufficiently good, but their specificity (i.e. their ability to label as negative those drugs carrying no risk) is not well established. Verapamil is a notable example of a false positive: it blocks human ether-a-go-go-related (HERG) K(+) channels, but is reported to have little potential to trigger torsade de pointes. Although inhibition of HERG K(+) channels has been proposed as a primary test for screening purposes, it is important to remember that several ion currents are involved in the generation of the cardiac potential and that metabolites must be specifically tested in this in vitro test. At the present state of knowledge, no preclinical model has an absolute predictive value or can be considered as a gold standard. Therefore, the use of several models facilitates decision making and is recommended by most experts in the field.