Abstract:
|
The principle that benefit-risk (B-R) decisions about medical treatments should, when possible, be based on statistically-significant evidence is well-established. However, the value of statistical significance is greatly weakened without control for multiple comparisons. Nevertheless, a review conducted by the QSPI B-R Working Group found minimal literature addressing multiplicity in this context. With technical advances creating increasingly fine subgroups, it is vital to have a comprehensive framework for dealing with multiplicity in B-R assessment. In this presentation, we will outline the questions that must be addressed, such as: If the planned analyses for efficacy and/or safety account for multiplicity, are additional controls needed at the stage of integrated assessment? If a B-R method allows users to set their own weights, does the multiplicity of possible weights imply a need for adjustment? Does every party assessing a treatment start with new alpha that may be spent?
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.