Statistical Software for Simulation of Adaptive Two Stage Designs
by Dipl. Ing. Michael Bauer
SA2D is a general simulation program which was written to study the statistical properties of two stage designs with adaptive interim analyses. The user is able to define the design and the adaptation rule in an interactive way. Then the experiment is simulated several times and the statistical characteristics of the design are derived from the Monte-Carlo sample.
Adaptive Designs represent a new philosophy of planning clinical trial designs. The idea of adaptive designs is to use the information of an interim analysis for further planning and still to control the type-I-error probabilty of the whole trial. Adaptive designs as used in SA2D go far beyond sample size adaptation or adaptive randomization (allocation) which are currently summarized under the topic adaptive designs.
 SAMPLE SIZES calculates power and/or sample sizes for the different distribution scenarios and is supposed to be a help tool for the user to evaluate sample sizes for the clinical trial design. It should also help to compare average sample sizes of the adaptive two stage method with the classical method. Sample size calculation is available for tests for trend, contrasts and difference for both normally and dichotomous response variables.
SA2D INPUT is the user interface for the input of the design (including the fixing of the respective distribution parameter for the simulation), the stopping rule and the adaptation rule. The interactive user module SA2D Input should help to find optimal stopping and adaptation rules for the same test problem under the respective constraints e.g. maximize power, maximize treatment safety with respect to side effects or minimize the bias in estimation. After having finished the input data are stored in a file to be used by the program SA2D RUN .
SA2D RUN performs the Monte-Carlo simulation on the given design with the selected stopping and adaptation rule. The seed and the number of simulations have to be entered. To ensure full flexibility the user has the chance to supply his own adaptation rule using certain key variables in c (c++) code. Examples how this can be done are given. Additional analyses on the output of the Monte-Carlo sample of the adaptive two stage design can be performed. |
