Identification of adverse reactions to new drugs. III: Alerting processes and early warning systems.

The process of discovery for 18 important adverse drug reactions was analysed in terms of study designs and other features which determined whether published reports of suspected adverse reactions were convincing or provided valid evidence of causality. First alerts were almost always anecdotal reports. For earlier alerting during clinical trials event recording is needed, associated with dose-response data and the use of patients as their own controls.