Advances and Applications in Statistics
Volume 12, Issue 1, Pages 25 - 36
(June 2009)
|
|
EVALUATION OF STOPPING RULES FOR STOCHASTIC CURTAILMENT
Demosthenes B. Panagiotakos (Greece) and Urania Dafni (Greece)
|
Abstract: Interim analysis with a view at
early stopping is often used in clinical trials. A group-sequential analysis
approach and one based on stochastic curtailment are the two main methodologies
applied to reach a decision on the early termination of a trial. Stochastic
curtailment uses the data already collected along with simulated data under
several hypotheses for the examined treatment differences. Decision is reached
based on the calculation of conditional power (CP) at the end of the study.
Critical values for the rejection or not of the hypothesis of interest are
chosen as an ad hoc. Aim of this work is to evaluate critical values for the
stochastic curtailment procedure in order to identify an objective stopping
rule. For this purpose we used data from an AIDS Clinical Trials Group study (ACTG
243), which was terminated early based on the stochastic curtailment procedure
(at 57.5% information ratio). Keeping the information ratio constant,
simulations were performed under several possible scenarios regarding the number
of events in the two treatment arms. According to our results, critical values
and
are chosen as most appropriate. Based on these cut-off points the decisions from
the two approaches, i.e., sequential analysis and stochastic curtailment, are
almost identical. |
Keywords and phrases: interim analysis, clinical trials, biostatistics. |
|
Number of Downloads: 348 | Number of Views: 1219 |
|