This paper describes a new statistical test of non-inferiority in clinical trials of anti-infective medications. The test allows the inferiority limit, delta, to vary with the underlying success rates. The test is a simple modification of the standard fixed-delta test, and its parameters can be adjusted to yield the fixed-delta test, a test similar to the FDA's Points-to-Consider procedure, or a test with nearly constant power values over a range of success rates. The test satisfies the size and power criteria for a valid statistical test. The effects on the test of lowering success rates through stringent criteria or through error are discussed.
Copyright 2002 John Wiley & Sons, Ltd.