False alarm rates of statistical methods used in determining increased frequency of reports on adverse drug reaction

Several statistical methods have been proposed or used in the determination of an increase in frequency of reports of adverse drug reactions. Some of the methods are based on large sample approximation or equal sample sizes assumption. The actual type I error may differ significantly from the predetermined nominal level of 5% when the methods are applied to rare adverse events. Simulated false alarm rates of the methods were compared under the assumption of 10 or less expected reports in the reference period.