All Times EDT
Keywords: Alpha spending functions, boundaries, group sequential monitoring, primary and secondary endpoints, type 1 error rate inflation
It is well-known that repeated monitoring of a clinical trial with no adjustment for multiple comparisons inflates the type 1 error substantially. It has also been shown that using group sequential monitoring for a primary endpoint and then spending full level alpha on a secondary endpoint at the time the primary endpoint crosses its boundary inflates the type 1 error rate. This talk reviews and unites these seemingly different topics and shows that the maximum inflation of error rate is actually identical in the two settings.