Not all hazards can be identified from clinical studies prior to marketing of medicinal products. Pre-marketing large-scale trials for cisapride did not report any serious cardiac arrhythmias. After a long period of availability in several countries it was withdrawn in 2000 because of reports of serious, and in many cases fatal, cardiac events. Whilst spontaneous reporting systems for adverse drug reactions (ADRs) have limitations such as under-reporting, they are an effective system for signal generation, particularly of rare ADRs. Pharmacoepidemiological studies aim to identify and calculate the incidence of adverse reactions, with increased sensitivity to less common ADRs compared to randomised controlled trials, yet cohort sizes may be insufficient to detect very rare ADRs such as drug-induced Torsade de Pointes, with an estimated incidence of the order of 1 per 12,000 to 1 per 120,000 patients. Several pharmacoepidemiological studies investigated adverse events associated with cisapride, one of which specifically examined the association between serious cardiac arrhythmias and cisapride. These observational studies were conducted using large population databases, but each failed to identify sufficient cases to establish a causal relationship. Explanations include that the cohort sample sizes were too small, and either under-, or mis-reporting of events of interest may have occurred. To estimate the risk of very rare adverse events, pharmacoepidemiological studies require very large numbers. Furthermore, the events in question need to be clinically recognisable by doctors and adequately documented in patients' notes, computer records, or on study questionnaires. The establishment of a national registry for drug-induced QT prolongation to identify cases and correlate clinical information may help to better identify these rare ADRs earlier. Such proactive surveillance could avoid unnecessary delays for other drugs where QT prolongation and serious cardiac arrhythmias may be an issue.