Adaptive subgroup analysis in clinical trials

Alexandra Graf, Franz Koening and Martin Posch
medical University of Vienna

How to deal with subgroup analysis has been controversially discussed between academia,
industry, and regulators (EMA, 2010). On the one hand ignoring a relevant subpopulation one could miss a treatment option due to a dilution of the treatment effect in the full population. On the other hand selecting a spurious sub-population is not without risk either: it might increase the risk to approve a inefficient treatment (inflating the type I error rate), or may wrongly lead to restricting an efficient treatment to a too narrow fraction of a potential benefiting population. Therefore we will investigate under which circumstances it is better to setup a trial with testing the full population, a subgroup only, or both (compare to Stallard et al., 2009, for two treatment groups). Furthermore, we will investigate the impact of performing an adaptive interim analysis to decide which patient population(s) should be investigated and tested for the remaining part of the trial. Applying conventional test statistics, naively ignoring the adaptive character of the trial, will substantially inflate the type 1 error rate. We will focus on how much the type 1 error rate can be inflated if no correction for multiplicity at all is done to account for possible inflation due to the sub-group analysis and/or adaptations. How to deal with adaptations in ongoing trials without compromising the type 1 error has inspired methodological research. In the particular setting of testing in a sub-group and the full population, adaptive designs based on the combination test method (Bauer and
Koehne, 1994) and the closed testing principle (Marcus et al., 1979) have been proposed
to control the family wise type 1 error rate (the probability of rejecting at least one true null hypothesis) in the strong sense (Brannath et al., 2009; Posch et al., 2005; Bretz et al., 2009). The operating characteristics of such (adaptive) designs allowing subgroup analysis will be explored.

References
Bauer,P. and Koehne,K. (1994). Evaluations of experiments with adaptive interim analysis.
Biometrics, 50, 1029-1041. Brannath,W., Zuber,E., Branson,M., Bretz,F., Gallo,P. Posch,M. and Racine-Poon,A. (2009). Confirmatory adaptice designs with Bayesian decision tools for a targeted
therapy in oncology. Statistics in Medicine, 97, 236-244.
Bretz,F., Koenig,F., Brannath,W., Glimm,E. and Posch,M. (2009). Adaptive Designs
for Confirmatory Clinical Trials. Statistics in Medicine, 28, 1181-1217.
EMA (2010). Concept paper on the need for a Guideline on the use of Subgroup Analyses
in Randomised Controlled Trials. Doc. Ref. EMA/CHMP/EWP/117211/2010,
available on http://www.ema.europa.eu/docs/en GB/document library/Scientific
guideline/2010/05/WC500090116.pdf
Marcus,R., Peritz,E. and Gabriel,K. (1976). On closed testing procedures with special
reference to order analysis of variance. Statistics in Medicine, 63, 665-660.
Posch,M., Koenig,F., Branson,M., Brannath,W., Dunger-Baldauf,C. and Bauer,P.
(2005). Testing and estimation in flexible group sequential designs with adaptive
treatment selection. Statistics in Medicine, 24, 3697-3714.
Stallard,N., Posch,M., Friede,T., Koenig,F. and Brannath,W. (2009). Optimal choice of
the number of treatments to be included in a clinical trial. Statistics in Medicine, 28,