Interim Analysis. Alpha Spending Function Approach
According to the American Food and Drug Administration’s directives as expressed in the International Conference of Harmonisation (ICH) Guidance (see underneath), interim analysis must be distinguished from monitoring.
© Springer International Publishing Switzerland 2017
TJ. Cleophas, A.H. Zwinderman, Understanding Clinical Data Analysis,
The ICH guidance, section E6, says, that monitoring is for the following purposes:
- - to maintain quality of the trial
- - to ensure that the protocol is followed
- - to ensure that in-/exclusion are appropriate
- - to check the availability and consistency of the data sampled
- - to check accrual rate
- - to check success to keep patients in the trial
- - (to check trial assumptions, and, perhaps, for sample size adjustments)
- - to monitor, simply, because it is essential for good quality.
It does not require an (independent) data and safety monitoring board (DSMB). In contrast, interim analysis does so. The DSMB is for the following purposes:
- - for analyzing efficacy and/or side-effects requiring de-blinding
- - for ethical concerns
- - for stopping, if there are too many side-effects
- - for stopping, if the effect is much larger than anticipated
- - for efficiency reasons
- - for stopping, if the effect is much smaller than anticipated
- - for checking assumptions as made in the design phase of the study.
We should emphasize, that it is useful, only, when decisions can be made! This chapter will review the above purposes, as well as many more relevant issues, including increased risks of type I errors, alpha spending functions to adjust these increased risks, stopping rules, special sample size requirements, decisions otherwise than stopping, special designs addressing interim analysis problems, like continuous sequential procedures, and triangular tests.