A New Class of Completely Self-Designing Clinical Trials

• Published in: Biometrical journal Vol. 45; no. 1; pp. 3 - 19
• Main Authors: Hartung, Joachim, Knapp, Guido
• Format: Journal Article
• Language: English
• Published: Berlin WILEY-VCH Verlag 01.01.2003; WILEY‐VCH Verlag; Wiley-VCH
• Subjects: Adaptive designing; Adaptive interim analyses; Applications; Biology, psychology, social sciences; Clinical trials; Exact sciences and technology; Group sequential trials; Mathematics; Medical sciences; Parametric inference; Probability and statistics; Sampling theory, sample surveys; Sciences and techniques of general use; Self-designing; Statistics; Biometrics; Prior information; Sample size; Prior distribution; Gaussian distribution; Non parametric estimation; Statistical estimation; Multivariate analysis; Adaptive method; Learning; Statistical method; Hypothesis test; Statistical test; Medical science; Sampling theory; Sequential method; Experimental design; Clinical trial; Sample survey; Learning algorithm
• ISSN: 0323-3847; 1521-4036
• DOI: 10.1002/bimj.200290014

A flexible method is proposed for group sequentially performed clinical trials which allows for an adaptive, data‐driven sample size reassessment at each stage. By also adaptively assigning different weights to the several stages the total number of study parts can be steered to an intended early or late end of the trial in dependence on all information available prior to a stage. Although at each stage the null hypothesis is tested on rejection, the full level‐α‐test is preserved at the end of the study. The proposed method is not restricted to normally distributed responses. The discussed adaptive designing is a useful tool provided that a priori information about parameters involved in the trial are not available or subject to uncertainty. The presented learning algorithm enables the complete self‐designing of a study.