Testing for efficacy with multiple endpoints has emerged as an
important
statistical problem due to increasing desire to include secondary
endpoints on drug labels. For example, label for the breast cancer
drug
Herceptin lists a primary endpoint and three secondary endpoints. In
accordance with the Prescription Drug User Fee Act of the
U.S. Congress
(PDUFA IV), The Food and Drug Administration (FDA) will issue a
guidance
on Multiple Endpoints in 2009.
Such statistical problems have defined paths to decision-making.
With primary and secondary endpoints, efficacy in a secondary endpoint
is
only relevant if efficacy in the primary endpoint has been shown.
Similarly, in non-inferiority and superiority testing, superiority
is only relevant if non-inferiority has been shown.
Current approach to such problems is based on closed testing, testing
all
possible intersection hypotheses, and collating the results. For
decision-making to follow pre-defined paths, strategic choices of test
statistics and critical values must be made. As the number of doses
and
endpoints increase, such strategic choices become increasingly
difficult.
Instead of closed testing, we propose to formulate partitioning
hypotheses using the Decision Path Principle. Decision-making from
testing
these hypotheses automatically respect pre-defined paths. The
number of hypotheses tested is also drastically reduced, facilitating
implementation and interpretation.
Meet the speaker in Room 212 Cockins Hall at 4:30
p.m. Refreshments will be served.
