A sequential procedure to test the progressive decrease of rare adverse events

A program to construct the Rapid Sequential Test (R-ST), a sequential procedure to test the decrease of the rate of rare adverse events, is described. This sequential procedure is constructed with the Sequential Probability Ratio method. It relies on a model for the progressive decrease in the rate of adverse events from a given initial rate to a target rate, during a transition period. The program allows one to calculate the bounds of rejection of the R-ST, and to study its power under various types of decrease. It is shown that the R-ST gives a conclusion faster than standard sequential procedures for the same type I and II error, saving at least one quarter of the trials completed during the transition period. The use of the software to construct R-STs is discussed with respect to the specification of the model.